I N V E S T I G AT I N G T H E P L A S M O D I U M I N VA D O M E U S I N G

H I G H - T H R O U G H P U T R E V E R S E G E N E T I C S

theodore richmond sanderson

Clare College, University of Cambridge

This dissertation is submitted for the degree of Doctor of Philosophy

September 2015

Theo Sanderson: Investigating the Plasmodium invadome using high-throughput

reverse genetics, September 2015

This dissertation is the result of my own work and includes nothing

which is the outcome of work done in collaboration except as

declared in the Preface and speciﬁed in the text.

It is not substantially the same as any that I have submitted, or, is

being concurrently submitted for a degree or diploma or other

qualiﬁcation at the University of Cambridge or any other University

or similar institution except as declared in the Preface and speciﬁed

in the text. I further state that no substantial part of my dissertation

has already been submitted, or, is being concurrently submitted for

any such degree, diploma or other qualiﬁcation at the University of

Cambridge or any other University of similar institution except as

declared in the Preface and speciﬁed in the text.

It does not exceed the prescribed word limit for the relevant Degree

Committee.

A BS T R A C T

The invasion of host erythrocytes by Plasmodium merozoites is an es-

sential event in both the parasite life cycle and in malaria pathogen-

esis. Many Plasmodium genes have been identiﬁed that play a role

in this process, but bioinformatic analysis suggests that much larger

numbers of as yet uncharacterised proteins may also be involved.

I set out to use the PlasmoGEM reverse-genetic approaches that

are being developed at the Wellcome Trust Sanger Institute to sys-

tematically investigate genes that have been associated with erythro-

cyte invasion. I ﬁrst conducted a screen of 145 putative invasion-

related genes in the rodent malaria model Plasmodium berghei, using

a recently developed barcode-sequencing approach. Parasites were

transfected with a pool of high efﬁciency knock-out constructs, each

tagged with a unique barcode. The quantities of different barcodes

in the population were counted over the course of the infection over

the course of the infection, and these data were used to identify the

viability and relative ﬁtness of knock-out parasites for each gene in

this panel. This work tripled the number of putative invasion-related

genes that have been investigated genetically, and I demonstrated the

viability of knock-outs for three genes previously described as essen-

tial.

I then explored approaches to adapting PlasmoGEM methodology,

which is currently only in use in rodent Plasmodium systems, to a

human-adapted line of the zoonotic parasite Plasmodium knowlesi that

can grow in human erythrocytes in vitro. I developed the vectors

needed to use PlasmoGEM technology in this system, and tested a

number of constructs generated to target putative invasion-related

genes. I was able to use these linear vectors both to knock-out and

to C-terminally tag genes in this species, which opens up potential

for future scale-up of PlasmoGEM in a human Plasmodium species for

the ﬁrst time.

Finally I developed a number of in silico approaches to analyse

the wealth of data developed by these approaches at scale. Knock-

out phenotypes were investigated in the context of a previous pu-

tative protein-protein interaction network and a correlation between

the phenotypes of connected genes was observed. In addition, a new

online database was created to facilitate the collation of mutant phe-

notype data in a systematic and machine readable fashion.

C O N T E N T S

i background to the project 1

1 introduction 3

1.1 Current control measures 5

1.2 Parasite life-cycle 7

1.3 Species of malaria parasite infecting humans 9

1.4 Parasite species in the lab 14

1.5 New sequencing technologies enable new approaches 19

1.6 The biology of invasion 20

1.7 Approaches to studying invasion 24

1.8 Experimental genetics in Plasmodium 29

1.9 The PlasmoGEM project 34

1.10 Approaches to enhance transfection further 39

1.11 Aims and objectives 44

2 materials and methods 45

2.1 Production of DNA vectors 45

2.2 Plasmodium knowlesi in vitro culture 50

2.3 Plasmodium berghei 58

2.4 Data analysis 61

ii results 63

3 initial investigations: gene selection and p. berghei

screen 65

3.1 Assessing the Hu et al. invadome 65

3.2 Selecting the core invadome 68

3.3 Barseq in P. berghei 71

3.4 Barseq growth phenotypes 77

3.5 Assessment of barseq phenotypes en-masse 93

3.6 Conclusions 97

4 the transfer of plasmogem technology to p. knowlesi 99

4.1 Introduction 99

4.2 System development 100

4.3 Preparation of PlasmoGEM vectors 114

4.4 Plasmodium knowlesi gene targeting 116

4.5 Transfection of PlasmoGEM vectors 122

4.6 Towards CRISPR/Cas9 in P. knowlesi 139

4.7 Generation of Cas9 mother vectors 143

4.8 Enhancing pJazz transfections with Crispr/Cas9 143

4.9 Discussion 146

5 new approaches and tools for large scale phe-

notypic analysis in silico 151

5.1 Barseq in P. berghei 151

x contents

5.2 Network-based analysis of the invadome 151

5.3 Large scale analysis of mutant phenotypes 154

5.4 Building the infrastructure to collate reverse genetic

data across species 165

5.5 Applying PhenoPlasm data to invasion 170

5.6 Discussion 172

iii discussion 177

6 general discussion 179

6.1 Achievements 179

6.2 Screening limitations 181

6.3 Extensions of work in P. berghei 185

6.4 Extensions of work in P. knowlesi 187

6.5 Final thoughts 190

bibliography 190

iv appendix 223

a tables 225

a.1 PlasmoGEM vectors used in this work 225

b supplementary figures 229

c raw barseq data 233

L I S T O F F I G U R E S

Figure 1 The decline in malaria deaths over the last cen-

tury 4

Figure 2 Distribution of gene descriptions in the P. falci-

parum genome 5

Figure 3 The life-cycle of the malaria parasite 8

Figure 4 Phylogenetic tree illustrating the evolutionary

relationship between Plasmodium species 10

Figure 5 Long-tailed macaques, Macaca fascicularis 12

Figure 6 P. knowlesi growth curves after episomal and

integrative transfections 18

Figure 7 The key steps of merozoite invasion 21

Figure 8 Schematic of a merozoite 22

Figure 9 Model for the motor complex involved in inva-

sion 26

Figure 10 Construction of PlasmoINT network 29

Figure 11 Schematic of ends-in integration 30

Figure 12 Synthesis-dependent strand annealing 32

Figure 13 PlasmoGEM procedure for generating targeting

vectors 36

Figure 14 PlasmoGEM procedure for generating targeting

vectors 36

Figure 15 Schematic to explain barcode-sequencing 38

Figure 16 Schematic of CRISPR/Cas9 43

Figure 17 Giemsa-stained smears from the stages of the

tight synchronisation protocol 52

Figure 18 Histogram showing the peak expression of all

Plasmodium falciparum genes and the subset that

are part of the invadome 66

Figure 19 Venn diagram of Plasmodium genes showing

1:1:1 orthology. 69

Figure 20 Phenotypes for core invadome genes as already

known from RMGMDB 70

Figure 21 Gene abundances in a single mouse for the ﬁrst

invadome barseq experiment 72

Figure 22 Fold-changes of all genes in barseq experiment

1 74

Figure 23 Fold-change on successive days in barseq ex-

periments 75

Figure 24 Fold-change values for the six control constructs

are as predicted 76

xii List of Figures

Figure 25 Strand-speciﬁc AP2-O transcription for AP2-O

locus 86

Figure 26 Individual knock-outs of three genes conﬁrms

barseq results which contradict previous stud-

ies 88

Figure 27 Comparison of barseq phenotypes with previ-

ous RMGMdb phenotypes 89

Figure 28 Barseq phenotypes of rhoptry-localised genes 90

Figure 29 Barseq phenotypes of IMC-localised genes 91

Figure 30 Barseq phenotypes of miscellaneous other genes 92

Figure 31 Proportion of various phenotypes in the ocore

invadome barseq 93

Figure 32 Known phenotypes 97

Figure 33 The effect of different sera on P. knowlesi growth 101

Figure 34 The effect of atmosphere on P. knowlesi growth

over 48 hours. 103

Figure 35 The effect of shaking vs. static culture on P.

knowlesi growth 104

Figure 36 Luciferase assays demonstrate high efﬁciency

of P. knowlesi transfection 106

Figure 37 Effect of DNA quantity on transfection efﬁciency 107

Figure 38 GFP expressing parasites are seeen one day af-

ter P. knowlesi transfection 108

Figure 39 The effect of electroporation on parasite mor-

phology 109

Figure 40 P. knowlesi is sensitive to DSM1, pyrimethamine

and G418, but with IC

values spanning four

orders of magnitude. 113

Figure 41 Crystal structiure of PfDHODH 114

Figure 42 Premature stop codon in an ApiAP2 gene in

A1.H1 isolates 115

Figure 43 Segmental duplication in my population of the

A1.H1 clone 116

Figure 44 Possibilities for library expansion 117

Figure 45 Phylogenetic analysis of sequences homologous

to PKNH_0941100 119

Figure 46 Phylogenetic analysis of PKNH_1207600 orthologs

in Plasmodium 121

Figure 47 Phylogenetic analysis of DHHC3 orthologs in

Plasmodium and T. gondii 122

Figure 48 Positive integration PCR results 123

Figure 49 Maps of double crossover integration strate-

gies using PlasmoGEM vectors both for knock-

outs and tagging in P. knowlesi 124

Figure 50 An example of the genotyping data produced

by qPCR. 127

List of Figures xiii

Figure 51 qPCR results reveal four modiﬁcations which

are dominant in their populations after trans-

fection. 128

Figure 52 Immunoﬂuorescence of 3xHA-tagged apical sushi

protein in P. knowlesi reveals the expected api-

cal localisation. 130

Figure 53 Alignment showing the C-terminus of PkDHHC3 131

Figure 54 Elevated dhfr copy number in non-targeting

transfections 133

Figure 55 Quantitative PCR conﬁrms attempt to dilution

clone PKNH_0941100 knock out. 134

Figure 56 Whole genome sequencing reveals successful

gene deletion at the PKNH_0941100 locus 135

Figure 57 Whole-genome sequencing reveals a failure to

delete PKNH_1207600 (highlighted in red) 136

Figure 58 Whole genome sequencing reveals failure to

target DHHC3. 137

Figure 59 Coverage of selection cassette as compared to

the general genome in whole genome-sequenced

samples. 138

Figure 60 pJAZZ arms 139

Figure 61 Distribution of on-target cutting scores for all

possible guide RNAs in the P. knowlesi genome 141

Figure 62 Distribution of predicted CRISPR on-target cut-

ting scores compared to mammalian systems

and analysed on a per-gene basis indicates a

large potential for CRISPR in P. knowlesi 142

Figure 63 Position of Crispr/Cas9 cutting site within PKNH_0941100,

a gene already known to tolerate deletion in P.

knowlesi from work earlier in this chapter. 145

Figure 64 Decreased wild-type presence in Crispr-assisted

pJAZZ transfection 145

Figure 65 Possible trajectories for a parasite after trans-

fection with a linear vector 146

Figure 66 Selecting the 90% conﬁdence subnetwork re-

sults in a network with relatively low connec-

tivity for most genes. 152

Figure 67 Invasion sub-network in context of entire 90%

conﬁdence network 153

Figure 68 Separation by localisation in the invadome sub-

network 154

Figure 69 Testing centrality-lethality in Plasmodium inva-

sion genes 155

Figure 70 Analysing the invadome subnetwork accord-

ing to mutant phenotype 157

Figure 71 The core invadome coloured by P. berghei barseq

mutant phenotypes, and labelled 158

Figure 72 Association plot demonstrates that connected

nodes tend to have the same mutant pheno-

type 160

Figure 73 Essentiality ratios for invadome genes clustered

by barseq mutant phenotype 161

Figure 74 Degree of knock-down in AP2-O knockout (in

published ookinete data) compared to the same

gene’s individual mutant phenotype 164

Figure 75 PhenoPlasm database schema 166

Figure 76 Phenoplasm screenshot 167

Figure 77 PhenoPlasm single gene view 168

Figure 78 Interface for data addition to PhenoPlasm 168

Figure 79 Supplementing invadome barseq data with Pheno-

Plasm database phenotypes 171

Figure S1 AP2-O expression levels by stage 229

Figure S2 Map of Crispr/Cas9 vector 230

Figure S3 Barseq and phenoplasm connectedness analy-

sis 231

L I S T O F TAB L E S

Table 1 Comparison of transfectability of the three par-

asite species analysed in this work 41

Table 2 Annealing reaction for the creation of dsDNA

inserts encoding guide RNAs. 49

Table 3 Recipe for P. knowlesi culture medium 50

Table 4 Barseq growth phenotypes for invadome genes 78

Table 5 Genes for which barseq phenotype agrees with

previous data 84

Table 6 Calculated IC

s in P. knowlesi for three drugs 111

Table 7 Targeting constructs available for initial P. knowlesi

experiments 118

Table 8 Result of pJAZZ transfection attempts 128

Table 9 Summary of results for three genes for which

knock-out attempts have been made in three

Plasmodium species 132

Table 10 Contingency table showing network edges grouped

by mutant phenotypes at either node 159

Table 11 Mutant phenotypes for invasion genes predicted

based on their essentiality ratios. 162

xiv

List of Tables xv

Table 12 Collated mutant phenotypes for Hu et al. in-

vasion genes which were not included in the

1:1:1 core invadome 173

Table S2 DNA oligos used in these studies. Note that

other oligos for PlasmoGEM vectors are not

listed since these are displayed on the Plasmo-

GEM website. 228

Part I

B A C K G R O U N D TO T H E P R O J E C T

I N T R O D U C T I O N

The battle between humanity and malaria parasites has raged for mil-

lennia. Early skirmishes are recorded in the human genome, with the

acquisition 5 – 10,000 years ago of alleles [114, 90] causing glucose-

6-phosphate deﬁciency, alpha thalessemia and sickle cell anaemia.

The diseases these variants cause, in homozygous form, are collateral

damage, representing the overwhelming selective pressure of malar-

ial deaths [114] that has shaped human evolution since the emergence

of agricultural societies 10,000 years ago. [101]

The past century has seen a reversal of fortune, with series of de-

feats for the parasite enabled by two key factors. The discovery of the

parasite lifecycle (section 1.2) led to the draining of swamps and use

of insecticides to reduce populations of the vector. This, combined

with the development of a succession of cheap and effective drugs,

has allowed the elimination of malaria in ninety-three countries since

1900 and a corresponding reduction in deaths (Fig. 1).

But parasites have shown an alarming propensity to develop re-

sistance to any drug thrown at them. Resistance to chloroquine was

ﬁrst observed in Thailand in 1957. It had reached Africa by the 1970s;

where it was the likely cause of increased mortality. [179] Resistance

to sulfadoxine-pyrimethamine developed in the late 1970s, and meﬂoquine-

resistant parasites appeared in the mid 1990s. [200] The emergence of

resistance to these drugs frequently followed the same pattern, with

parasites appearing ﬁrst in Southeast Asia, followed by a spread of

the resistant alleles across the planet.

With the emergence of artemisinin-resistant parasites in Cambodia,

[51] there is an urgent need for new drugs, and ideally for the de-

velopment of an effective vaccine. To date, the most potent vaccines

show only modest reductions in morbidity. [168] As in many ﬁelds,

malarial drug development is shifting to a target-based approach, in

which drugs are speciﬁcally developed towards promising parasite

proteins.

The most obvious prerequisite for a drug target is that its gene be

essential—as we will see in the biology of invasion, parasites often

4 introduction

100

200

300

1900 1930 1950 1970 1990 1997 2013

Year

Annual malaria mortality per 100,000

Geographic region

World excl. SSA

World

Sub−Saharan Africa

Chloroquine developed

First chloroquine resistance

ACTs

Figure 1: Malaria mortality has declined sharply in the last century but remains

high in Sub-Saharan Africa.

The decline of malaria seen between 1930 and 1950 was the result

of many efforts, but the development of chloroquine was a major

factor. The rise in drug resistant parasites from 1954 onwards is

correlated with a resurgence of malarial deaths, warning that the

recent inroads made against the parasite [149] could be rapidly

eroded by resistance to artemisinin. (ACTs: Artemisinin Combina-

tion Therapies, SSA: Sub-Saharan Africa)

have many redundant pathways for achieving a particular biological

process. Beyond this, it is important to understand the role the pro-

tein plays in parasite biology and how likely resistance is to develop.

Identifying and prioritising targets will require detailed insights into

parasite biology at the protein level, but our understanding of the

Plasmodium biology is far from complete. The most common anno-

tation for P. falciparum genes is “conserved Plasmodium protein, un-

known function”, which is the sole description for around 30% of

genes (Fig. 2). Because the apicomplexans diverged from other eu-

karyotes very early in evolutionary history many genes have no ho-

mologs in other systems, and it will be a long time before we under-

stand the role that some of these play in parasite biology.

One particularly important event in the parasite life-cycle, the full

details of which are still to be elucidated, is erythrocyte invasion—the

process by which the parasite enters a new host red blood cell. This

process is considered an important target for drug and vaccine de-

1 1900-1997 data from WHO World Health Report, 1999. I calculated 2013 ﬁgures from

WHO statistics using World Bank population data.

1.1 current control measures 5

conserved Plasmodium protein, unknown function

riﬁn,PIR protein (RIF)

conserved Plasmodium membrane protein, unknown function

conserved protein, unknown function

unspeciﬁed product

erythrocyte membrane protein 1, PfEMP1 (VAR)

Plasmodium exported protein, unknown function

RNA−binding protein, putative

zinc ﬁnger protein, putative

stevor,PIR protein

DnaJ protein, putative

riﬁn, pseudogene,PIR protein, pseudogene (RIF)

probable protein, unknown function

Others

Figure 2: A large proportion of the genes in the Plasmodium genome have entirely

unknown functions.

This chart shows the distribution of P. falciparum genes with the

description provided by PlasmoDB. The ’others’ slice represents

descriptions applying to fewer than 15 genes.[157]

velopment. [41, 104] While the roles of some invasion-related genes

have been outlined, [42] there are hints that there may be far more

involved [92] and this project aims to analyse some of these genes at

scale to better understand their functions and importance.

Somewhere in the milieu of ’conserved Plasmodium proteins, un-

known function’, may lie an invasion-related drug target that will one

day quell the rise of multi-drug resistant parasites, and so it is impor-

tant to gain more data on these proteins. Such useful data includes

mutant phenotypes, protein localisations and genetic interactions.

1.1 current control measures

1.1.1 Prophylaxis and vector control

One of the most cost-effective means of malaria control has been the

prevention of bites by long-lasting insecticide-treated mosquito nets

and indoor residual spraying. [81] However this approach shows no

signs of entirely eliminating transmission, and its success is threat-

ened by increasing levels of pyrethroid resistance. [8]

A number of chemoprophylactic agents exist for travellers visiting

malaria-endemic countries, but attempts to prescribe these on long

timescales result in low rates of compliance, [31, 32] whilst some

6 introduction

cause side-effects and resistance inevitably develops. [31, 131] Nev-

ertheless they may have important potential roles in the protection of

children and pregnant women. [81]

1.1.2 Treatments

Once malaria is suspected in a patient the current frontline treat-

ment is one of the artemesinin combination therapies. These partner

a rapidly-acting artemisinin compound (which has a short clearance

time in the body) with a longer-lasting drug which can clear any re-

maining parasites.

These drugs are highly effective, and their roll-out has helped to

reduce the number of deaths due to malaria over the last decade. But

the emergence of resistance has the potential to threaten these devel-

opments. ’Slow-clearing’ parasites have arisen in Western Cambodia

with mutations that allow them to evade clearance under artemesinin

treatment and to produce recrudescent infection. [182] Such parasites

now give rise to treatment failure under a number of drug regimens.

If the parasite alleles causing these effects spread in Africa, a signiﬁ-

cant increase in mortality seems likely.

1.1.3 Vaccines

There is no vaccine currently available against malaria which shows

both cost-effectiveness and good efﬁcacy. The most developed tar-

geted vaccine to date is the RTS,S subunit vaccine based on the cir-

cumsporozoite protein (CSP), which demonstrated 36% efﬁcacy in

reducing episodes of malaria in children but no signiﬁcant effect on

mortality. [168] A number of other vaccines are under development.

Many of these are targeted to various components of the parasite’s

invasion machinery.

An alternative approach is the intravenous injection of irradiated

whole sporozoites—the stage of the parasite injected by mosquitoes.

In early trials this method provided highly effective protection against

subsequent P. falciparum challenge, [172] but numerous obstacles exist

to its roll-out in endemic settings including the need for manual dis-

1.2 parasite life-cycle 7

sections of thousands of mosquitoes and a liquid-nitrogen cold-chain

distribution system. [69]

1.2 parasite life-cycle

Like many parasites, Plasmodium has an exquisitely complex life-cycle

reliant on more than one host. During this process the single-celled

eukaryote transforms itself into many and varied forms, and journeys

large distances through both an insect and a vertebrate.

1.2.1 In the human host

As a malaria-infected Anopheles mosquito begins to probe for blood

vessels (Fig. 3), it injects saliva containing vasodilators and anticoagu-

lants. Plasmodium hijacks this process to transmit sporozoites, present

in the salivary glands, into a new host. In each blood meal, 15-123

motile sporozoites are injected into the tissue of the skin. [159] These

parasites ﬁnd their way into blood vessels and are carried in the cir-

ulation to the liver sinusoid. Here, they traverse the sinusoidal cell

layer to infect hepatocytes.

Once inside a hepatocyte, hidden from the immune system, a sin-

gle sporozoite divides again and again over 2-16 days (depending on

parasite species) to produce thousands of invasive merozoites. [159]

During this process the parasite induces the death of the host cell

and causes it to detach. The detached cell extrudes merosomes, pack-

ets of merozoites which bud off and enter the bloodstream. In the

blood the merosomes rupture, releasing merozoites which can invade

erythrocytes. [185]

Upon entering an erythrocyte, a parasite initially assumes a ‘ring’-

like morphology—the parasite nucleus appears as the jewel on a ring

of cytoplasm, when Giemsa-stained. The parasite then begins a 24–72

hour maturation process (depending on parasite species) culminating

in the release of new merozoites.

After a period in the ring stage, the parasite becomes a trophozoite

as it digests the cell’s haemoglobin, forming crystals of haemozoin

in the food vacuole. It then develops into a schizont as the genome is

replicated and up to 32 daughter merozoites produced.

8 introduction

1. Sporozoites injected

and enter bloodstream

2. Many rounds of

replication in the liver

Ring

Trophozoite

Schizont

Egress

Invasion

4. Gametocyte

formation

5. Activation

6. Fertilisation

7. Ookinete traversal

Ookinete

Gut wall

Salivary gland

8. Oocyte forms

and matures

9. Sporozoites invade salivary gland

Gut wall

Figure 3: The Plasmodium life cycle

This schematic outlines the major stages of the parasite lifecycle.

Plasmodium parasites transition between a multitude of lifestages

as they traverse the variety of environments in the different organs

of their hosts.

1.3 species of malaria parasite infecting humans 9

Once this development is complete the mature schizont can begin

an organised process of egress in which ﬁrst the parasitophorous vac-

uole membrane, then the erythrocyte membrane, is broken down and

merozoites are released into the bloodstream to begin a new round

of invasion.

While this intra-erythrocytic cycle can continue indeﬁnitely, a small

proportion of infected cells undergo a developmental switch, and be-

gin to express the transcription factor AP2-G. [178, 102] These cells

develop to schizonts in typical fashion, but each merozoite released

is committed not to continue the asexual cycle but to eventually be-

come a gametocyte. It will invade a new red blood cell, and form a

trophozoite, but then not into a schizont but an elongated gametocyte.

These sexual stages have two distinct morphologies, microgametocytes

and macrogametocytes, or male and female respectively.

1.2.2 In the mosquito vector

When these gametocytes are subsequently taken up by a mosquito

as part of its blood meal, the environment of the mosquito midgut

triggers ‘activation’. Male gametocytes exﬂagellate, releasing motile

gametes which can fuse with the female to form a diploid zygote.

This zygote then develops into a motile ookinete which burrows into

the midgut wall and forms an oocyst. Each oocyst grows and divides

to produce thousands of sporozoites which are ultimately released

and make their way to the salivary glands, ready to be injected into

the next human host.

1.3 species of malaria parasite infecting humans

This particular parasitic way of life, with an insect and vertebrate

host, is shared by countless different Plasmodium species. It is now

known that six species of malaria parasite cause signiﬁcant numbers

of human infections.

Because Plasmodium malariae, and Plasmodium ovale (now known to

comprise two distinct species [186]), are of low prevalence and rarely

cause severe malaria, they will not be discussed in detail here. But

10 introduction

P. gallinaceum

P. berghei

P. vivax

P. cynomolgi

P. knowlesi

P. malariae

P. ovale

P. gaboni

P. reichenowi

P. falciparum

Species Host

Laverania

Figure 4: The evolutionary history of malaria parasites in different species

Parasites used in subsequent analysis are denoted by boldface.

Host pictograms denote (in order of appearance): birds, rodents,

humans, monkeys, chimpanzees. The tree illustrates the interest-

ing relationships, or lack thereof, between host and parasite phylo-

genies – P. falciparum is equally closely related to P. vivax (a human

parasite) and P. berghei (a rodent parasite). The tree represents a

subset of a phylogeny of cyt b and cox1 genes by Duval et al.. [56]

to set this work in context, some detail is needed on three parasite

species.

1.3.1 Plasmodium falciparum

90% of malaria deaths occur in sub-Saharan Africa. This is enabled by

the fact that infections in this region are caused almost exclusively by

the rapacious P. falciparum, compounded by often inadequate health-

care provision. This species is a member of the Laverania, a subgenus

of ape parasites that form an outgroup to the other human malaria

parasites and to rodent parasites (Fig. 4). It itself appears to have

originated from this group in a host-switching event from a gorilla-

infecting parasite in Africa, relatively recently in evolutionary history.

[126]

1.3 species of malaria parasite infecting humans 11

Two aspects of P. falciparum’s biology make it especially deadly.

Where many other species show a marked preference for invading

reticulocytes, P. falciparum can invade red blood cells of any age al-

lowing it to produce higher parasitaemias and more severe anaemias.

Another dangerous feature of P. falciparum is its habit of sequestra-

tion. Malaria parasites can be cleared if they pass through the spleen

and so some species cause their host cells to present adhesive pro-

teins and ‘sequester’ within capillaries. In P. falciparum it appears that

sequestration can occur in the vasculature of the brain, where it some-

times combines with activation of the immune system to result in

cerebral malaria. This is the most severe pathology of any malaria par-

asite, which can cause coma and ultimately death. [20] Most sufferers

are children, and even those who survive can have lasting cognitive

deﬁcits. [93]

1.3.2 Plasmodium vivax

The origin of P. vivax, both in terms of geography and ancestral host,

has been hotly contested. [29, 46, 40] One recent view holds that this

parasite too hails from Africa. [127] The recent discovery of a ‘syl-

vatic’ clade of P. vivax-like parasites in African apes has led to the

suggestion that an ancient P. vivax was originally able to infect hu-

mans, gorillas and chimpanzees in Africa.

In ancestral African populations, the selective pressure of P. vivax

morbidity triggered a loss of expression of the host receptor it re-

quires, the Duffy antigen receptor for cytokines (DARC), from ery-

throcytes. [141]

In the new model, this led to the extinction of P. vivax in humans

in Africa, but the parasite lived on in apes as the sylvatic clade and

human P. vivax today represents a population that escaped Africa,

undergoing a genetic bottleneck.

An alternative view is that the parasite emerged in Asia, [46] imply-

ing a previous monkey host for P. vivax. Many of the well-characterised

Plasmodium species related to P. vivax (such as P. knowlesi and P. cynomolgi)

infect monkeys.

Recent observations on Madagascar have challenged the view that

P. vivax is absolutely dependent upon DARC for invasion; with robust

12 introduction

Figure 5: Long-tailed macaques, Macaca fascicularis, which along with pig-

tailed macaques are the primary hosts of P. knowlesi (original pho-

tograph)

data indicating the presence of P. vivax in some Duffy-negative indi-

viduals. [139] This is just one example of how little we know about

P. vivax, despite the substantial morbidity it causes worldwide, [67]

primarily because of the intense difﬁculty of culturing it in vitro.

1.3.3 Plasmodium knowlesi

In contrast to the two previously mentioned human species, P. knowlesi’s

primary hosts are not humans but the macaques of Southeast Asia

(Fig. 5). All extant parasites are derived from a parasite that lived

98,000 - 478,000 years ago. This emergence is before the human settle-

ment of Southeast Asia, meaning that the parasite likely emerged in

macaques. [116]

There are disputes over who was the ﬁrst person to observe P.

knowlesi microscopically, [11] but it owes its name to Robert Knowles.

He and Biraj Mohan Das Gupta described the parasite, which had

been discovered in a macaque imported from Singapore. [108] They

observed that it caused a ’quotidian’ fever in humans (temperatures

peaking once each day), rather than the tertian or quartan fevers ob-

served with other human Plasmodium. [160]

Most early data on human P. knowlesi infection comes from a pe-

riod in which the parasite was used as a pyretic to treat sufferers of

neurosyphilis. [164] In a study in Edinburgh it was passaged from

one patient to the next with decreasing potency on each occasion; but

in Bucharest, where the treatment was most widespread, after 170 se-

1.3 species of malaria parasite infecting humans 13

quential passages the parasite developed such virulence that its use

was ended. [160]

1.3.3.1 P. knowlesi causes large numbers of infections with potentially se-

vere symptoms

The ﬁrst reported natural human infection with P. knowlesi was in

1965 when a ‘U.S. Army Map Surveyor’, working in the jungles of

Malaysia, developed a fever [37] – though given reports that he did all

his work by night, he is more likely to have been a spy. [160] Samples

of his blood were subsequently inoculated into ‘volunteers’, result-

ing in quotidian infections and moderate to severe clinical symptoms.

Sporozoite inoculation revealed that the infection could be passed on

by mosquitoes which had fed on an infected human, proving the theo-

retical possibility of parasite transmission within human populations.

While P. knowlesi enjoyed attention as a model parasite in the 1970s

and 1980s (see later sections), there was little further discussion of

human infections until a seminal discovery in 2004.

Balbir Singh and colleagues were investigating a focus of malaria

infections in Kapit, a divison of Sarawak in Malaysian Borneo. [176]

In Kapit, a ﬁfth of infections were then identiﬁed by Giemsa-smear

as P. malariae, but these infections did not fully resemble this typically

benign parasite species: 18.5% of patients had more than 5,000 par-

asites per µl of blood. The researchers aimed to investigate whether

these infections represented a new P. malariae variant, or an emergent

Plasmodium species. The answer turned out to be neither of these.

After PCR assays for each of the ﬁve major Plasmodium species gave

negative results, the researchers ampliﬁed the SSU rRNA gene from

these parasites and sequenced it. The sequence they found was that

of P. knowlesi, then considered almost exclusively a simian parasite.

P. knowlesi can assume a band-like morphology which is typical of P.

malariae, and this was the cause of the diagnostic confusion. Newly

designed PCR primers then revealed that 58% of malaria infections

in Kapit contained P. knowlesi parasites, and subsequent research has

shown that human P. knowlesi infections are widely distributed in

Borneo and beyond. [44, 100]

Studies have since conﬁrmed the severity of P. knowlesi malaria,

which has a fatality rate of 1-2%. [47] Many of the symptoms of se-

14 introduction

vere malaria can be present including anaemia, jaundice, acute renal

injury, hyperparasitaemia, hyperlactaemia, hypotension and respira-

tory distress. [43]

1.3.3.2 Insights into population genetics

One key early question was whether the population of parasites in

humans represents a single clade of the P. knowlesi radiation. The

observation that most alleles found in human infections are shared

by infections of wild macaques demonstrated this was unlikely to be

the case. [116]

In a separate project we sequenced entire parasite genomes from

patients presenting with P. knowlesi and found that they clustered

into two distinct groups. [156] Dimorphic genes included the invasion

ligand NBPXa, and the alleles of this gene were implicated in changes

in parasite virulence. [2]

At the time the cause of this sympatric dimorphism was unclear,

but recent research has revealed a genomic dimorphism between par-

asites in M. fascicularis and M. nemestrina hosts, [54] and I have since

conducted in-silico PCR analyses to conﬁrm that these reﬂect the two

groups we saw in patient isolates (unpublished data).

1.3.3.3 The mosquito host of P. knowlesi

The P. knowlesi parasite is transmitted by members of the Leucos-

phyrus group of Anopheles mosquito species, predominantly A. latens.

[180] The range of these species largely explains the parasite’s distri-

bution in Southeast Asia, centred around the island of Borneo. These

hosts are exophilic, rarely venturing into houses, and this explains

why most P. knowlesi patients are people who work in forests. [180, 43]

This raises the prospect that changes in land-use might drive a switch

in mosquito behaviour and an increase in P. knowlesi transmission.

1.4 parasite species in the lab

There are a number of model systems of malaria used to investi-

gate different aspects of parasite biology. It is worth considering the

strengths and weaknesses of each.

1.4 parasite species in the lab 15

1.4.1 P. falciparum

As the parasite species causing the most morbidity and deaths world-

wide, P. falciparum has long been the primary focus of malaria re-

search. Early experimental work used human volunteers, but there

were also ape models in chimpanzees and Aotus monkeys.

A signiﬁcant breakthrough was achieved with the development in

1976 of a system that allowed continuous in vitro culture of the par-

asite. [194] This involves a low-oxygen atmosphere, produced using

either malaria gas or a candlejar, and media based on RPMI-1640 –

developed for culturing leukemia cells – supplemented with serum

or its equivalents.

P. falciparum was the ﬁrst parasite to have its genome sequenced

in 2002 [70], but reverse genetic studies of this parasite have lagged

behind the murine malarias. Difﬁculties in the experimental genetic

investigation of P. falciparum result from its very low transfection ef-

ﬁciency (section 2.2.6) and the need to transfect with circular DNA

which forms episomes which can only be removed by a prolonged

period of drug cycling, though recent targeted endonuclease develop-

ments may change this.

1.4.2 P. berghei

In any tropical region, many species of animal will be infected with

malaria parasites, and the thicket rats of central Africa are no excep-

tion. It is from these animals that P. berghei was isolated, but in the lab

it can infect a range of rodents including laboratory strains of mice

and rats. [192]

P. berghei has proved a powerful in vivo model in which parasites

can be transfected with linear DNA and achieve high transfection

efﬁciencies. Parasites grow rapidly (parasitaemia increasing tenfold

per 24-hour cycle during the logarithmic period of infection). The

model allows the study of a number of host-speciﬁc features of par-

asite disease, including cerebral malaria, [130] host nutritional status

and immunology. [112] With knock-out mice now available for large

numbers of genes the model also allows the genetic investigation

of host-parasite interaction from both sides. The RMGMdb database

16 introduction

maintained by Leiden University [117] currently lists 625 attempted

gene disruptions in P. berghei.

1.4.3 P. knowlesi

Although the discovery that P. knowlesi causes a signiﬁcant clinical

burden is very recent, the parasite has long been of interest to para-

sitologists. Before the development of robust culture systems, animal

models were the only means to study malaria, and primate models

considered the most accurate representations of human disease. Ad-

ditionally, in parallel with the development of culture systems for P.

falciparum, it also became possible to culture P. knowlesi in vitro, using

macaque red blood cells.

Kocken et al. used both this system, and growth in animals, to

show that the parasite could be transfected with both circular and

linear DNA, and demonstrated that a number of selectable markers

could be used. [109] However there have been very few publications

in which P. knowlesi parasites are genetically modiﬁed in the macaque

context. This is in part due to the difﬁculty of acquiring macaque

blood with sufﬁcient quantity and frequency for routine culture.

1.4.3.1 Adaptation to human erythrocytes

While by 2004 it was established that P. knowlesi could theoretically

replicate in human red blood cells, indeed that it could cause vast

parasitaemias in patients, attempts to culture the parasite in human

cells always failed. It was clear that parasites had some ability to

invade human cells: if donor parasites from a cynomologous culture

were incubated with human erythrocytes, some infected human cells

were seen. [142] But parasites would quickly die out if continuously

cultured in the human cells.

Moon et al. set out to establish a line of P. knowlesi that could

grow robustly in human erythrocytes. [142] The starting point for

this experiment was the A strain, derived from the H strain ﬁrst

isolated from that ’surveyor’ in Malaysia (section 1.3.3.1) and main-

tained since then by passage in macaques.

2 This ﬁgure includes multiple attempts to disrupt the same gene

1.4 parasite species in the lab 17

Moon ﬁrst conﬁrmed that if he mixed human cells with macaque

cells in a 3:1 ratio he could maintain parasites in this hybrid blood.

The 25% of the erythrocytes which were from macaques were enough

to keep the parasites growing, albeit at a lower rate than usual, and

his hypothesis was that the presence of 75% human erythrocytes

would provide a selective pressure for the parasites to adapt to grow

inside these cells in vitro. Any parasite that developed a mutation

allowing it to do so better, even to a small extent, would have an ad-

vantage over its neighbours which could not exploit most of the cells

surrounding them.

After eight months of growing the parasites in this mixture, he

found that the resultant parasites were capable of replicating well (3–

4ˆ increase per intraerythocytic development cycle (IDC)) in human

blood, although they still grew somewhat better (4-5ˆ) in macaque

cells.

Across the Atlantic, Lim et al. had also been adapting a line of P.

knowlesi to human erythocytes, in a similar fashion. [120] In this case

they found that the changes that occured in this time expanded the

range of ages of erythrocyte that could be invaded.

However in neither case has there yet been a published mutation,

or mutations, shown to be causative in the adaptation of the parasite

to the human cells.

Moon performed experiments to establish whether the human-adapted

line was amenable to transfection. Using schizont transfection he

achieved a transfection efﬁciency of more than 10% on day 1 post-

transfection – a better result than in any previously reported Plasmod-

ium transfection. He also showed that he could transfect with linear

DNA to achieve ends-in integration (see section 1.8.2), a feat not pos-

sible in P. falciparum.

However it is worth noting that in this integration experiment (Fig. 6)

the parasitaemia of transfectants did not reach 2% until day 14, 11

days later than in the transient transfection condition.

The parasite typically multiplies at a rate of 2.5-3.5 fold per 28

hours. 2.5

11ˆ

“ 5649, indicating that of those parasites that take up

DNA, just 1 in 5000 may be integrating it. This is a very different re-

sult to P. berghei where the rate of formation of episomes is reported

to be similar to that for integration. [97]

18 introduction

10 15

Days post-transfection

Circular DNA

Linear DNA

Transfectant parasitaemia / %

Figure 6: Parasites come up rapidly under selection for P. knowlesi episomal trans-

fections but integrations take signiﬁcantly longer.

Figure re-drawn to combine data from two ﬁgures of Moon et al.

[142]

Such differences presumably reﬂect differences in the DNA repair

machinery between the two organisms. But using recently developed

techniques for enhancing the efﬁciency of homologous recombination

(section 1.10.3.1) it might be possible to bring the rate of integrant-

formation closer to the rate of episome formation in P. knowlesi.

1.4.3.2 The rationale for genetic modiﬁcation of P. knowlesi

This project involves an attempt to bring novel experimental genetic

technologies to bear on P. knowlesi. Some might question why we

should we study this parasite, when the majority of those those suf-

fering from the disease may be monkeys. There are at least 3 possible

answers to this question.

The ﬁrst is, ‘for its own sake’: it is clear that P. knowlesi causes signif-

icant numbers of fatalities in Southeast Asia each year, and it is pos-

sible that ecological pressures on its simian hosts could select for its

further adaptation towards humans. In addition, it appears that the

sucess of malaria control in decreasing the numbers of P. vivax and P.

faciparum populations may be resulting in increased P. knowlesi infec-

tions. [205] These factors, and simple scientiﬁc curiosity, may lead us

to want to better understand the biology of this fascinating zoonotic

parasite.

But a second reason to study P. knowlesi is as a model for all malaria

species. Studies of rodent malaria have been crucially important in

1.5 new sequencing technologies enable new approaches 19

understanding malaria as a whole, both because these parasites are

more amenable to genetic manipulation and because they allow host-

parasite interactions to be studied in vivo. Both of these advantages

apply to P. knowlesi with the additional advantage that it can be cul-

tured in vitro in human cells, allowing the scale of experiments to

increase without a corresponding increase in animal use.

A ﬁnal interesting possibility of P. knowlesi research is that it might

allow us to better understand the elusive and important P. vivax. This

species, which causes considerable morbidity, [158] cannot be cul-

tured in vitro effectively which severely limits our understanding of

its biology. As shown in Fig. 4, P. knowlesi is far more closely related to

P. vivax than is any other model (indeed, in terms of evolutionary dis-

tance, even the rodent malaria parasites would be better models for

understanding P. vivax than would be P. falciparum). And P. knowlesi

recapitulates P. vivax biology in some areas. Naturally it shows a retic-

ulocyte preference [120], and its genome has a similar (A+T)-content

to P. vivax, signiﬁcantly lower than that of P. falciparum. It does not

seem that P. knowlesi produces hypozoites, the sleeper cells of P. vivax

which can lie dormant in the liver to produce a recrudescence years

later, but it is quite possible that there are many other features of P. vi-

vax biology shared with the other non-Laverania ape parasites which

we do not yet understand because we lack the ability to culture P.

vivax. P. knowlesi might shed insights on some of these areas.

1.5 new sequencing technologies enable new approaches

Many of the techniques used in this project would not have been pos-

sible a decade ago; the cost of sequencing genomes has been radically

altered by the development of what is often called "next-generation

sequencing". Where sequencing by the Sanger method identiﬁed the

nucleotides in a single stretch of DNA of about 1 kb, these technolo-

gies permit up to 4 billion short reads DNA to be sequenced on a

single ﬂow cell

. [94]

While read-lengths with this technology were initially less than 50

bp, improvements to the sequencing chemistry now allow two 150bp

reads per DNA fragment sequenced. In recent years smaller scale se-

3 Illumina HiSeq 2500

20 introduction

quencing platforms such as the Illumina Miseq have been launched.

While these are more expensive per basepair sequenced, they cost

very much less per run. This has made it feasible to routinely se-

quence small numbers of parasite genomes, and also to use massively

parallel sequencing for other purposes such as barcode-sequencing,

as we will discuss.

1.5.1 Reference genomes

In 2002 the P. falciparum genome was sequenced, this resource has

proved invaluable to malaria research. The P. knowlesi genome, which

represents the H strain, passaged exclusively through macaques was

published in 2008 and there is also limited whole-genome data from

the Nuri strain. [150]

1.6 the biology of invasion

This project is focused on the very smallest developmental form of

the parasite, indeed one of the smallest eukaryotic lifeforms known,

the merozoite. P. falciparum spends less than two minutes per 48 hour

cycle – 0.06% of its time – in this form, [75] and yet its biology is a

focus of intense interest.

The merozoite is a specialised machine designed to overcome one

of the greatest challenges the parasite faces. Once supplies of nutri-

ents in an erythrocyte are exhausted, parasites have to leave the rela-

tive safety of the parasitophorous vacuole and enter the bloodstream

to ﬁnd a new home. There they face the full force of the immune

system: a sea of antibodies, white blood cells and complement. The

merozoite must avoid this adversity and rapidly bind to and enter a

new cell.

Invasion is considered an attractive target for drug and vaccine de-

velopment. The blood stage is the only target for a malaria therapy

that can cure a patient already showing symptoms, and the merozoite

is the only stage of this cycle in which the parasite is outside a host

cell. Thus it could represent the weak link at which the intraerytro-

cytic cycle is most targetable.

1.6 the biology of invasion 21

deformation

tight junction

AttachmentEgress Pre-invasion Invasion Post-invasion

Dynamic `Resting’

Exonemes

Micronemes Dense granules

Rhoptries

Figure 7: The process of merozoite invasion

The key steps of invasion are shown, along with the organellar

secretion that occurs at each stage. Drawn with inspiration and

information from [42].

1.6.1 The key steps of invasion are conserved across Plasmodium

The ﬁrst detailed view of Plasmodium invasion in action came, in fact,

from P. knowlesi. The relatively large size of its merozoites enabled

the ﬁrst video microscopy of invasion in 1975. [57] The researchers

observed that the haemozoin in a schizont would converge to a sin-

gle point; and that soon after the merozoites would separate, and

the erythrocyte swell, then explosively rupture. The merozoites re-

leased made contact with red blood cells, which were then engulfed

by waves of deformation. A short time after these ended the parasite

moved into the erythrocyte, through a pinch-point (the ’tight junc-

tion’), to give rise to a ring parasite. It was also observed that the

apical end of the merozoite had to be facing into the erythrocyte for

invasion to occur.

Subsequently Gilson and Crabb conducted similar experiments in

P. falciparum. [75] After making their observations of invasion and

comparing their results to those seen 34 years prior in P. knowlesi,

they wrote that:

Our observations of P. falciparum indicate that the mor-

phological steps and kinetics of erythrocyte invasion are

remarkably similar to those detected in P. knowlesi. This

conservation across a large evolutionary distance suggests

that the maintenance of invasion rates, including the tim-

ing of the individual steps, has reached an optimum.

22 introduction

Nucleus

Exonemes

Dense granules

Rhoptries

Micronemes

Inner membrane

complex

Surface coat

Figure 8: Schematic of a merozoite

Major oreganelles are labelled, many of these (rhoptries, mi-

cronemes, IMC, surface coat) are intimately involved in invasion.

These approaches, and many others, have led to the model of inva-

sion we have today. (Fig. 7).

1.6.2 The anatomy of egress and invasion

The ultrastructure of the merozoite is divided into organelles which

function at different steps during the invasion process. The proteins

contained in these organelles vary in number, and in sequence, be-

tween the different Plasmodium species, and some mutate rapidly un-

der the high pressure of the immune system. Nevertheless the con-

served appearance of the process as a whole suggests that increased

understanding of the molecular biology of invasion in one species

will have ramiﬁcations for others.

As with most areas of parasite biology, the molecular details of

invasion are currently best understood in P. falciparum and so in this

section I will describe the process as it occurs in this species, and with

its gene nomenclature. Nevertheless many of the genes involved have

homologues across Plasmodium.

Before any invasion can occur, the merozoite must escape the con-

ﬁnes of its host erythrocyte. Triggering this egress process is the role

of the exonemes. [209] As a schizont matures, the calcium concentra-

tion increases causing the activation of cGMP-dependent protein ki-

nase (PKG). Ultimately PKG activation results in the secretion of the

contents of the exonemes into the parasitophorous vacuole. [38] One

crucial component of the exonemes is SUB1.

SUB1 is a serine protease that among other targets processes the

merozoite surface proteins MSP1, MSP6 and MSP7. At least some of

1.6 the biology of invasion 23

these events appear to be essential to prime the merozoite for inva-

sion. [36] SUB1 processing also activates SERA6, a protease essential

for egress. [170] The parasitophorous vacuole is now broken down,

likely by a process involving conscripted host calpain. [33] Finally

the erythrocyte membrane bursts and merozoites are released. [206]

Merozoites are explosively ejected from bursting schizonts. In the

blood they encounter low potassium levels, which cause calcium re-

lease and the discharge of the adhesins stored in the micronemes onto

the merozoite surface. [177]

When a merozoite comes into contact with an erythrocyte, its initial

contact is low-afﬁnity and reversable. This attachment is most likely

mediated by members of the merozoite surface protein family. [42]

During this period the red blood cell ’rufﬂes’ with visible deforma-

tions, and the parasite reorientates until its apical end reaches the

erythrocyte surface.

An irreversible attachment is now formed, and this triggers the

rhoptries to begin to discharge, releasing the Rh proteins. These bind

to various receptors on the erythrocyte surface (Rh5 to basigin, Rh4

to complement receptor 1, and Rh1-3 to as yet unknown receptors).

Together with members of the EBL family the attachment of these

proteins appears to commit the parasite to invasion.

The next player to leave the rhoptries is the RON complex, which

is inserted into the erythrocyte. RON2 forms a complex with AMA1

which in the past been considered has been considered the interac-

tion that provides traction between the parasite and the erythrocyte,

although this is not deﬁnitively established. [16]

The last part of the rhoptry to discharge its contents is the ’bulb’,

containing the lipids and proteins required to make up the para-

sitophorous vacuole, so that as the parasite starts to invade the ery-

throcyte it is able to build its own compartment inside.

The physical force for invasion comes from another parasite or-

ganelle, the inner membrane complex (IMC). The IMC lies beneath the

parasite plasma membrane and is made up of a series of ﬂattened

vesicles, or alveoli, which are intricately connected to the cytoskele-

ton below. [110] It is this feature that gives the Alveolata, the super-

phylum to which Plasmodium belongs, its name. Between the IMC

and the plasma membrane is an actin myosin motor which propels

the parasite into the erythrocyte and the nascent parasitophorous vac-

24 introduction

uole. The IMC forces the parasite through a visible pinch point on the

erythrocyte surface, known as the tight junction.

It is clear that the formation of strong attachments between para-

site and erythrocyte is essential for invasion, but in order for invasion

to occur it is perhaps just as important that these attachments are

cleaved—without this one might imagine the parasite would remain

stuck to the outside of the erythrocyte. This ‘shedding’ of invasion lig-

and peptides into the supernatant is mediated by proteases including

ROM1, ROM4, and SUB2.

As the IMC pulls the parasite into the red blood cell, the tight junc-

tion is the last component to enter and ﬁnally the parasitophorous

vacuole is sealed. The parasite is now becoming a ring and the dense

granules are released to make the erythrocyte resistant to further inva-

sion events and to begin remodelling the erythrocyte to the parasites’s

own ends. The most dangerous period of the parasite’s intraerythro-

cytic cycle is over and the next 48 hours of ordered metabolism and

development can begin.

1.7 approaches to studying invasion

Broadly speaking, proteins involved in invasion can be thought of as

being divided into two categories: adhesins and invasins. [42] The

former bind directly to factors on the erythrocyte whilst the latter are

involved in invasion, but in less direct ways.

1.7.1 Detecting the molecular components underlying known host-parasite

interactions

Historically, the identiﬁcation of machinery involved in invasion has

often relied upon the fact that these proteins target parasites to spe-

ciﬁc subsets of cells. Whenever a difference has been observed be-

tween the ability of parasites to invade one cell type compared to

another researchers have sought to identify the protein-protein inter-

actions underlying this difference, and thus build up a picture of the

invasion process.

For example, the observation that people of African descent are

largely protected from P. vivax malaria led to the discovery of DARC

1.7 approaches to studying invasion 25

as an essential host receptor for these parasites. [141] Analysis of par-

asite proteins which bound differentially to Duffy-positive vs. Duffy-

negative blood, then identiﬁed the parasite protein involved in this

process, Duffy binding protein (DBP). [204] Meanwhile, it had been

observed that P. falciparum, which does not require DARC for inva-

sion, was unable to invade cells which had their surface sialic acid

removed by enzyme treatment. [26] Similar binding experiments to

those undertaken in P. vivax allowed the identiﬁcation of the pro-

tein involved, EBA-175. Comparison of the sequences of the DBPs

of P. knowlesi and P. vivax and EBA-175 from P. falciparum identiﬁed

shared domain architecture that grouped these proteins as the found-

ing members of a family, the erythrocyte-binding like (EBL) family.

[1]

The other major family of invasion proteins, the RBLs, were ﬁrst

identiﬁed in P. vivax in a very similar way. [68] Parasite proteins which

bound only to reticulocytes were identiﬁed on protein gels. Again

here, proteins with similar sequences were identiﬁed in other parasite

species, forming the reticulocyte-binding like (RBL) family [162, 140].

The expansion of the families has sometimes involved bioinformat-

ics preceding biology. For example Rh5 was identiﬁed purely on the

basis of sequence similarity to other RBLs, and then its role in inva-

sion was established, [18] and ﬁnally its host receptor identiﬁed [45].

(i.e. the mirror image of the order of events in the discovery of DBP).

It is likely that there are other parts of the invasion machinery which

have yet to be found because they do not have sequence similarity to

known invasion genes.

1.7.2 The identiﬁcation of components of the actin-myosin motor has pri-

marily relied on inferences of orthology

Many of the adhesins described above differ markedly between Plas-

modium species, and they lack homologues outside Plasmodium. This

is unsurprising given that these components—which interact with the

host, and are exposed to the immune system—are likely to be under

intense selective pressure and hence to evolve rapidly in evolutionary

arms-races between parasite and host.

26 introduction

ErythrocyteParasite

Receptor

MyoA

MTIP

Actin

Adhesin

GAC

GAP45

GAP50

GAP40

Inner IMC Outer IMC Plasma membrane Plasma membrane

GAPM

Figure 9: A possible model for the mechanism of force production by the motor of

the malaria parasite

A molecular motor comprising MTIP and MyoA pushes the IMC

into the merozoite, pulling against ﬁlamentous actin which is an-

chored to receptors in the erythrocyte. This anchoring may take

place through some glideosome adhesin connector (GAC) but this

is not yet deﬁnitively established. Partially adapted from [63]

It took much longer for the invasin genes of Plasmodium to be char-

acterised - the motor complex was ﬁrst reported in 2006, based on

components already characterised in Toxoplasma. [99, 17] The motile

stages of Plasmodium share an actin-myosin motor complex with other

Apicomplexan parasites, and the machinery used for gliding motility

in other stages plays a key role for merozoite invasion. Almost all

invasin biology was discovered in Toxoplasma before its orthologous

constituents were identiﬁed in P. falciparum.

Our understanding of this machinery, and how it links to the ery-

throcyte is still far from complete. The protein linking the motor to

a surface adhesin was originally proposed to be aldolase, but this

has recently been questioned and a candidate to replace it identiﬁed

in Toxoplasma. [173, 95] A very speculative model of how the motor

might function is presented in Fig. 9, based on this new GAC protein,

1.7 approaches to studying invasion 27

but this model pieced together from data from various species cer-

tainly excludes some key components, and may well be fundamen-

tally ﬂawed. Whether the glideosome adhesin connector (GAC) in

Plasmodium is homologous to that in Toxoplasma is not known. Futher-

more the adhesin that links to the GAC is not established. TRAP plays

this role in sporozoites and a merozoite homologue MTRAP is a can-

didate to do so during erythrocyte invasion but supporting data is

currrently sparse.

A recent analysis identiﬁed a second myosin motor MyoB and an

interacting light chain (PF3D7_1118700) as very likely also implicated

in invasion. [211]

1.7.3 Novel approaches may be needed to expand our understanding of

invasion-related components

Thus far, many techniques that have identiﬁed proteins involved in in-

vasion, from the binding experiments of the 1980s to AVEXIS (Avidity-

based extracellular interaction screening) experiments today, have been

biased—for very good reasons—towards detecting adhesins. Despite

an array of challenges, these are typically easier to identify: there is a

target and relatively open-ended experiments are carried out to iden-

tify what binds to it.

Experiments to discover invasins have been generally limited to in-

vestigating sequences homologous to those better understood in other

systems, but this prevents the discovery of any machinery unique to

Plasmodium.

Often experiments to directly identify invasins involve testing very

speciﬁc hypotheses—that X processes Y so that it can interact with

Z. Increasing the scale of invasin discovery will require much more

hypothesis-free approaches.

1.7.3.1 Transcriptomic data can predict protein function

One such approach was developed by Hu and colleagues in 2010

using transcriptomic data. [92] Previously researchers had tried to ex-

ploit the "just in time" nature of Plasmodium transcription to assign

functions to proteins. [22, 163] These approaches essentially treated

co-expression over the course of the intraerythrocytic development

28 introduction

cycle (IDC) as evidence of a possible common function. However

because a number of different processes of parasite biology can be

active at any point in the lifecycle these approaches did not yield

high-conﬁdence predictions. [92]

Hu et al. aimed to add more dimensions of resolution to the tran-

scriptome by exposing parasites to 20 different compounds known to

inhibit parasite growth. They found that each compound left a unique

signal on the transcriptome of the parasites perturbed by it. The idea

here was not to investigate the compounds themselves, but to exploit

the possibility that two genes whose expressions were inﬂuenced in

a similar fashion by exposure to a particular compound might be in-

volved in the same part of parasite biology.

The researchers took 5-10 time points for each compound, giving a

total of 144 microarrays. They combined this data with IDC microar-

rays for ﬁeld and lab strains and then looked for co-expression as

evidence of a functional relationship.

This co-expression evidence was then combined with data from

yeast two-hybrid experiments (which use a heterologous expression

system to detect protein-protein binding), and data on the phyloge-

netic distance between proteins and domain-domain interactions. A

Bayesian approach was used to integrate these four data sources into

a single score for any two proteins’ likelihood of being involved in an

interaction (Fig. 10).

The ‘PlasmoINT’ database of putative interactions that this analy-

sis produced has implications for many areas of parasite biology, but

the researchers chose to validate it by looking at invasion. They chose

25 proteins known to be involved in invasion and then identiﬁed the

subnetwork formed by selecting all proteins linked with > 90% conﬁ-

dence to at least one of these.

This analysis identiﬁed a 418 gene subnetwork, a putative ‘invadome’.

The researchers successfully characterised the localisation of 42 of the

additional proteins identiﬁed. 31 of these were consistent with a role

in invasion (the remainder had mostly cytoplasmic localisations, not

indicative of invasion but not necessary excluding some role in it) ,

giving some conﬁdence to the data. But, as we would expect from

our limited understanding of the genome (Fig. 2), 37% of these genes

have as their only annotation "conserved Plasmodium protein, un-

1.8 experimental genetics in plasmodium 29

Perturbation microarrays

Phylogenetic relationship

Yeast two-hybrid

Domain-domain interaction

gene A

gene B

gene C

gene D

gene E

gene D

gene C

gene B

gene A

95%

50%

98%

95%

98%

50%

Bayesian model

Figure 10: Schematic of approach taken by Hu et al. to construct the PlasmoINT

interaction network

A Bayesian approach was used to optimally combine four data

sources to infer potential protein-protein interactions.

known function". This implies that there are still many components

of the invasion machinery we have yet to understand.

1.8 experimental genetics in plasmodium

1.8.1 The DNA repair machinery available in the parasite informs the ex-

perimental genetic approaches that are possible.

That reverse genetics is possible in any eukaryotic system is the unin-

tended consequence of cellular machinery designed to repair genomes

from natural damage. Each day cellular processes and external fac-

tors such as cosmic rays create an estimated 10 double-strand breaks

(DSBs) in each human cell, and these must be efﬁciently repaired.

[119]

Where this repair machinery differs between organisms this af-

fects the possible strategies for reverse genetics. Crucially, Plasmodium

lacks the machinery for canonical non-homologous end joining (C-

NHEJ), and although it does possess an alternative, microhomology-

mediated, end joining mechanism [107] these events are rare. [167]

Thus Plasmodium is left mostly reliant on repairing its genome by

homologous recombination (HR), a process for which it does possess

the requisite proteins. [115] The presence of this system, and the ab-

sence of C-NHEJ, is in one sense of great assistance to malaria geneti-

cists because it means that almost the only way in which a parasite

can integrate a resistance marker is by homologous recombination,

30 introduction

Homologous region

Homologous region Homologous region

Chromosome

Insert

Chromosome ChromosomeInsert

Figure 11: Schematic illustrating the process of ends-in integration

With this approach the entire plasmid is copied into the genome

with the single homology region replicated.

which necessarily targets the marker to a particular place in the para-

site genome.

DNA repair is a complex process in which a number of different

mechanisms can lead to the same end result in the genome of a trans-

fected parasite. There are, however, just two possible arrangements in

which exogenous DNA can be incorporated into a parasite. These are

sometimes called single and double crossover, but these names are

not accurate because an apparent ‘double crossover’ event can actu-

ally occur without any crossing over. [115] The names I will use here

are ’ends-in’ and ’ends-out’, respectively.

1.8.2 Integration events fall into two main classes

In an ends-in integration event there is a single region of homology be-

tween the target genome and the targeting plasmid. Crossover occurs

somewhere in this region, and the entire plasmid is then copied into

the genome, sometimes multiple times. There are two disadvantages

to this approach as compared to ends-out integration.

Firstly a region can never be actually deleted from the genome by

this method; DNA can only be added. This doesn’t make knock-outs

impossible – a critical part of a gene can be disrupted by an insertion

– but it does introduce complexity.

1.8 experimental genetics in plasmodium 31

Secondly, because the entire plasmid is introduced it is never pos-

sible to use negative selection to distinguish between integrants and

parasites carrying episomes.

In ends-out recombination there are two regions of homology, either

side of the region to be replaced. The result is incorporation of the

DNA between them, in place of that region in the genome.

The limited data available suggests that the primary way in which

ends-out recombination occurs in Plasmodium is by synthesis-dependent

strand annealing (SDSA). [184]

In this mechanism (Fig. 12), a double strand break forms (spon-

taneously, or otherwise) in the genome near the site to be targeted.

There is then resection of both fragments by exonucleases which chew

back on the 5’ (and to a lesser extent the 3’) ends of the break. One

or both of the free 3’ strands can then invade a homologous region

of the transfected plasmid. Here it acts as a primer, to be extended

in the 3’ direction by polymerases. At ﬁrst a region of homology will

be copied, then the heterologous resistance marker and any tags, and

then a ﬁnal region of homology. This ﬁnal region provides material

which can anneal to the other side of the lesion to repair the double

strand break. Any nucleotides absent from the other strand can now

be synthesised on the basis of this completed template.

While this system exists to allow repair from fully homologous se-

quences, for example in other chromosome copies during schizogony,

it is enlisted by the geneticist to incorporate alterations and markers

into a parasite genome.

1.8.3 Genetic screens are crucial for understanding genomes at scale

Large scale genetic screens for particular phenotypes, from develop-

mental defects in Drosophila to metabolism in S. cerevisiae, have made

a more profound contribution to our understanding of model organ-

isms than any other technique. But such techniques are just begin-

ning in Plasmodium, where publications which investigate more than

50 genes are very rare.

There are two broad approaches that can be used in a genetic

screen: forward and reverse genetics.

32 introduction

Gene to be knocked out

Spontaneous or induced DSB

Resection by 5’ exonucleases...

.. and concurrently by 3’ exonucleases

Linear donor DNA, with selection marker. Strand invasion by free DNA ends.

Free DNA ends are extended copying the donor sequence.

Once homology on the other side of DSB has been copied,

strand annealing and ligation can occur.

The incomplete strand is synthesised by copying the newly-synthesised strand

Strand annealing.

Strand is extended by synthesis.

Strand invasion by just of one of the free DNA ends.

Figure 12: Synthesis-dependent strand annealing

This ﬁgure illustrates one of the processes by which ends-out in-

tegration of DNA can occur. A double-strand break forms within

the gene to be deleted, and resection by exonucleases exposes re-

gions homologous to the donor vector. Strand invasion enables

the synthesis of new genomic DNA from the donor template and

ligation repairs the nicks in the genome.

1.8 experimental genetics in plasmodium 33

A forward genetics approach is any in which the discovery of a phe-

notype occurs before the identiﬁcation of the genetics underlying it.

Perhaps the best example is the discoveries of Nusslein-Volhard and

Wieschaus in Drosophila embryology. [146] They chemically induced

random mutations in ﬂies and looked for developmental phenotypes.

The location of the genetic change driving the phenotype was then

mapped by crossing mutant ﬂies with wild-type ﬂies and looking for

markers which co-segregated with the phenotype.

One problem with employing such an approach in Plasmodium is

the intense difﬁculty of crossing parasites due to the complexities of

the sexual cycle. This would likely mean a need for whole-genome

sequencing to identify the mutations present in cloned parasites with

a phenotype. Such an approach might be possible in this age of ever-

cheaper sequencing, but the mutation rate of the parasite (Hamilton,

Claessens, et al., unpublished) means that multiple mutations might

be identiﬁed leaving the causative one unclear.

Transposon mutagenesis has become a favoured way to randomly

create loss-of-function mutations which can be easily mapped by meth-

ods such as inverse PCR. The technique has been employed effectively

in large numbers of organisms including Plasmodium, where it has

yielded many clones with mapped insertions. [14] However until the

development of recent unpublished techniques (similar to barcode-

sequencing) the phenotyping of these parasites has been relatively

laborious.

By contrast, a reverse genetics approach disrupts a known region of

the genome and then looks for any resultant phenotype. This has the

advantage of allowing a targeted investigation of a subset of genes,

for example every kinase in the Plasmodium genome. [189, 80]

The discovery of RNAi has simpliﬁed reverse genetics approaches

in many organisms by allowing easy knock-downs of gene expres-

sion, but Plasmodium lacks the machinery for this process and so it is

not a technique available to malariologists. [19]

Instead, at least until alternative techniques such as dCas9 are per-

fected [133], targeted mutations must be made in the parasite genome.

The targeted nature of these genetic alterations has some advantages

over forward genetics approaches – genes can be deleted, rather than

the creation of a frameshift or the insertion of a premature stop codon.

34 introduction

1.8.4 History of experimental genetics in malaria

When the ﬁrst P. falciparum transfection was carried out in 1995, the

authors wrote that ’one problem has been the cloning of (A+T)-rich

sequences ﬂanking the coding regions of P. falciparum genes, as many

recombinant DNA constructs containing these sequences are unsta-

ble in Escherichia coli’. [208] They had found that the P. falciparum

dhfr gene, which they hoped to use (in mutant form) as a selectable

marker, spontaneously re-arranged during cloning.

In the two decades since these ﬁrst transfections it is estimated that

only ~500 Plasmodium genes have been successfully targeted for gene

disruption, [48] the majority in P. berghei. One challenge for genetic

modiﬁcation has been low transfection efﬁciencies. This is perhaps

unsurprising considering the number of membranes that DNA has

to cross to reach the nucleus. To successfully transfect ring or tropho-

zoite stage parasites, DNA must cross four membranes: the red blood

cell plasma membrane, the parasitophorous vacuole membrane, the

parasite plasma membrane and ﬁnally the nuclear envelope.

A number of techniques have been developed attempting to in-

crease transfection efﬁciency. These include, ‘pre-loading’ erythrocytes

with DNA and then introducing parasites to them, electroporation of

ring stage and schizont stage parasites and ‘double-tap’ combinations

of these. [88, 27] Anecdotally, the schizont appears the parasite stage

most amenable to transfection. [142] Additionally, it seems that the

square wave produced by Lonza electroporators enhances DNA up-

take. [122]

1.9 the plasmogem project

The Plasmodium genetic modiﬁcation project (PlasmoGEM) at the Well-

come Trust Sanger Institute aims to increase the efﬁciency and scale

of Plasmodium genetics using new genetic technologies and plate-

based vector generation techniques.

1.9 the plasmogem project 35

1.9.1 Linear cloning vectors allow the creation of more efﬁcient targeting

vectors

As previously mentioned, one of the problems of working with Plas-

modium is that the parasites have extremely (A+T)-rich DNA which

is difﬁcult to propagate in circular vectors using the conventional lab-

oratory workhorse Escherichia coli.

The ﬁrst important component of PlasmoGEM was the creation of

a large-insert library of P. berghei genomic DNA in the pJAZZ vector

developed by Lucigen. This vector is derived from the N15 coliphage,

which has a linear genome of double-stranded DNA ﬂanked by co-

valently closed hairpins created by the phage-encoded protelomerase

TelN. [78] The pJAZZ vector is a version of this phage which lacks the

genes for the phage’s lytic cycle, replacing them with a drug resis-

tance marker and a multiple cloning site. It is maintained in the TSA

strain of E. coli which has the TelN gene integrated into the bacterial

genome.

The linear topology of the vector means that unlike a closed circular

plasmid it is not possible for supercoils to be formed. Superhelicity

is largely responsible for the instability of (A+T)-rich DNA and so

pJAZZ vectors can allow inserts of up to 28 kb of Plasmodium DNA

to be maintained. [154]

The linear topology of the vector has a secondary advantage. Be-

cause vectors are never circular, even inefﬁcient restriction enzyme

digestion to release inserts will not leave circular vectors with the

ability to form episomes that can confer drug resistance to parasites.

A library of genomic fragments inside pJAZZ backbones has now

been produced covering 91% of P. berghei genes, and smaller pilot li-

braries have been produced for P. knowlesi and P. falciparum. Capillary

sequencing has been used to map the position of each library clone

in the relevant parasite genome.

1.9.2 A robust pipeline for vector production increases the scale of genetic

studies

The second important component of the PlasmoGEM project is a high-

throughput procedure for taking a clone from this library, containing

36 introduction

GOI 5’ UTR

Gene of interest

GOI 3’ UTR

GOI 5’ UTR

Gene of interest

GOI 3’ UTR

Zeo/PheS

recombineering

Gateway reaction

PCR

Zeo/PheS

PCR with primer overhangs

Gene of interest

HA tag 3’ UTR 5’ UTR dhfr/yfcu 3’ UTR

(positive selection with Zeocin)

(negative selection with d,l-p-chlorophenylalanine)

Library clone

Zeo intermediate

Transfection vector

Figure 13: Procedure for the generation of tagging PlasmoGEM constructs

GOI 5’ UTR

Gene of interest

GOI 3’ UTR

GOI 5’ UTR

GOI 3’ UTR

Zeo/PheS

recombineering

Gateway reaction

PCR

Zeo/PheS

PCR with primer overhangs

HA tag 3’ UTR 5’ UTR dhfr/yfcu 3’ UTR

(positive selection with Zeocin)

(negative selection with d,l-p-chlorophenylalanine)

Library clone

Zeo intermediate

Transfection vector

Figure 14: Procedure for the generation of knock-out PlasmoGEM constructs

1.9 the plasmogem project 37

wild-type parasite genomic sequence, and turning it into a target-

ing vector containing a drug resistance marker. For these purposes, a

pipeline involving recombineering and Gateway technology has been

developed. [154]

In this methodology the library clone containing the target locus

is transformed with a plasmid encoding the bacteriophage λ red re-

combinase operon and E. coli recA recombinase. After 16 hours of

incubation to allow this protein to be expressed, the bacteria are tran-

siently transformed with a cassette expressing the zeo-pheS positive-

negative selection marker. The cassette is ﬂanked by short regions of

homology to the genomic clone, added by PCR with long primers.

These permit the recombinase to exchange the cassette into the ge-

nomic clone, replacing the gene of interest in the case of a knockout

vector, or immediately following it in the case of a tagging vector.

The zeo-pheS cassette allows positive-negative selection and is ﬂanked

by attLR sites. This means that the vector containing it is a powerful

intermediate; a Gateway recombinase [213] reaction can be used to

insert any of a range of different cassettes into the vector, each encod-

ing a drug resistance marker and, in the case of tagging vectors, an

epitope tag.

1.9.3 PlasmoGEM vectors carry barcodes which enable novel approaches

to genetic screens

An important component of the PlasmoGEM approach is the barcode

carried in each vector. During the PCR of the Zeo-PheS cassette a

unique 11 bp sequence is added to each vector, ﬂanked by constant

annealing sites. This barcode identiﬁes the gene targeted by the vec-

tor. The barcodes are designed with redundancy such that even the

mutation of two nucleotides will not cause one barcode to become

another.

A PCR reaction with the same primers can be used to amplify bar-

codes from any vector, or from the genome of a parasite transfected

with any vector. The advantage of this approach is that a pool of dif-

ferent vectors can be transfected together into a parasite population.

When selection is applied to kill parasites which have not integrated

a vector, a pool of transgenic parasites remains, with different genes

38 introduction

Gene A: 80%

Gene B: 20%

Gene C: 0%

Gene A: 60%

Gene B: 40%

Gene C: 0%

Gene A: 33%

Gene B: 33%

Gene C: 33%

normal

essentiality

slow

day 5day 4day 0

Barcode counting

by Illumina seq.

Pooled transfection of

schizonts

Infection timecourse

->day 8

PCR PCR PCR

Figure 15: Schematic view of barcode seqencing

A pool is made of barcode-knock out constructs targeting a num-

ber of different genes. In this case there are three constructs, one

targeting a redundant gene (green), one a gene whose knock-out

parasites are attenuated (yellow), and one an essential gene (red).

These constructs are transfected into schizonts and integrate into

the parasite genome, producing a mixed population of knock-

outs with the gene knocked out in each case indicated by the

presence of a barcode. Selection is applied to kill parasites not

carrying a vector, and from days 4 to 8 of the infection a small

sample of blood is taken each day. These samples and the in-

put pool are subjected to PCR to amplify their barcodes. These

are then quantiﬁed by Illumina sequencing. Essential genes can

be identiﬁed by the presence of their barcodes in the input pool

but absence in the alive transfected parasite population (because

all parasites integrating them have died as a result). Attenuating

and redundant knock-outs can be distinguished by comparing

the fold-change on successive days of the infection.

1.10 approaches to enhance transfection further 39

deleted in different parasites. The gene deleted in each case is indi-

cated by the barcode incorporated into that parasite. This means that

the ratio of different barcodes in the population can be used to follow

the relative growth of parasites with different genes knocked out.

Because all barcodes are ﬂanked by the same primer annealing

sites, a single PCR reaction can amplify all of the different barcodes in

the pool, maintaining their relative abundances. Massively parallel se-

quencing (section 1.5) can then be used to sequence many thousands

of these barcode molecules in order to quantify the ratios within

them.

The incorporation of index tags during the preparation of barcodes

can allow at least 32 different barcode pools to be sequenced on a sin-

gle MiSeq lane. These points can represent independent replicates of

an experiment, and also different points in time during an individual

experiment.

This means that a pool of vectors can be transfected into a para-

site, and then DNA extracted from the parasite population daily as it

grows up under selection. When barcodes are counted for each day,

the relative growth rate of parasites with different genotypes can be

inferred. Additionally those genes whose barcodes never appear in

the population can be identiﬁed as refractory to deletion.

This technique has been developed and validated in P. berghei, where

it has been shown that a single pooled experiment recapitulates the

phenotypes found in previous single-transfection experiments. [79]

1.10 approaches to enhance transfection further

1.10.1 Increasing the basal growth rate

What makes Plasmodium berghei such an efﬁcient model system for

malaria experimental genetics? One basic factor which should not be

overlooked is that the parasite grows very fast. During the exponen-

tial phase of an infection parasitaemia increases tenfold each twenty-

four hour cycle. P. falciparum by contrast rarely exceeds an 8-fold in-

crease over a forty-eight hour cycle. Because growth is exponential, in 6

days a single P. berghei parasite will have produced a million progeny

while a P. falciparum parasite has in the same time produced at most

40 introduction

512. Our inability to achieve such growth rates in P. falciparum and P.

knowlesi is likely an absolute limitation of culture systems - although

we do know some ways to increase the re-invasion rate, such as using

an orbital shaker. [7]

Differences in basal growth rate clearly affect how quickly one can

obtain mutants, but they also deﬁne whether one can obtain a par-

ticular mutant. A particular gene, when mutated, might decrease a

parasite’s growth rate to 25% of its wild-type value. For P. berghei this

is a decrease from 10-fold to 2.5-fold growth. In the case of P. knowlesi

this would be a decrease from 3-fold growth to 0.75-fold growth. But

of course a growth rate less than 1 means that the transgenic para-

site will never be seen, however long culture is continued. The higher

the basal growth rate the more genes are experimentally accessible,

making optimising culture conditions especially important.

1.10.2 The rate of DNA integration differs between Plasmodium species

This rapid growth is not the only factor that makes P. berghei so sat-

isfactory to work with. On day 4 after a transfection of P. berghei, a

low parasitaemia is visible under selection. The vast majority of these

parasites (we know from the reliability of barseq experiments) are

the result of correctly targeted double crossover events. Day 4 for P.

berghei approximately corresponds to day 8 for P. knowlesi, due to the

lower growth rate. In optimal P. knowlesi experiments even on day 8

60% of the population does not contain the expected genotype [142],

and this is despite the fact that the raw transfection efﬁciency of P.

knowlesi, the ease with which we can get DNA into the parasites, is

orders of magnitude higher than in P. berghei.

There appears to be a distinction between how easy it is to get DNA

into the parasite and how easy it is to get that DNA integrated into

the parasite genome (Table 1). Now that the former seems very well

optimised for P. knowlesi perhaps there are approaches from other

systems which can help to improve the latter.

1.10 approaches to enhance transfection further 41

parasite species transfection eff. linear dna

P. falciparum Low Very low

P. berghei Medium High

P. knowlesi Very high Medium

Table 1: Comparison of transfectability of the three parasite species analysed

in this work.

1.10.3 Enhancing integration efﬁciencies

A number of methods for generating double strand breaks (DSBs) on

demand have now been developed, initially to assist in the study of

the machinery of recombination, [169] but later to enhance integration

efﬁciencies.

Initially DSBs were created by inserting sites for the site-speciﬁc

eukaryotic nuclease SceI into the region to be studied. Since this point

a succession of nucleases have been developed, each one allowing

more convenient targeting than the last.

Zinc ﬁnger nucleases (ZFNs) were developed by fusing a non-speciﬁc

nuclease domain from the FokI enzyme to zinc-ﬁngers, protein mo-

tifs from transcription factors which bind to speciﬁc triplets of DNA.

Once constructed, these allowed DSBs to be created at any site of in-

terest, allowing signiﬁcant increases in targeting efﬁciency in many

systems [136] including Plasmodium [184, 144].

Once a validated ZFN is produced it is a very effective method of

inducing DSBs. The downsides of ZFNs are practical: they are expen-

sive, requiring synthesis by an external ﬁrm and a number of can-

didates typically need to be tested before one is found which works

efﬁciently. This would make scaling up a ZFN approach to large num-

bers of genes prohibitively expensive.

These problems were initally ameliorated with TALENs which re-

placed the zinc-ﬁngers with DNA-binding domains from the TAL-

effector proteins of plant pathogens. These have a speciﬁc peptide se-

quence deﬁned to bind to each nucleotide triplet, enabling in-house

synthesis from a distributed kit. But TALENs had a limited time at the

technological forefront before the next targeted-nuclease revolution.

42 introduction

1.10.3.1 Crispr-Cas9

In the short time since its development, Crispr-Cas9 technology has

radically changed many ﬁelds of genetic research. The Cas9 protein

is a nuclease which is guided to its target by binding a short guide

RNA sequence (gRNA). These gRNAs can be transcribed from DNA,

meaning that targeting Cas9 to a new section of the genome is as

simple as ordering a pair of oligonucleotides.

A number of studies have now reported success in using CRISPR-

Cas9 to modify malaria parasite genomes. [212, 72, 201] The approach

has allowed linear DNA to be used to target P. falciparum for the ﬁrst

time [72], reducing the problem of episomes, and has also allowed the

selection of marker-free parasites with genomic modiﬁcations. [201]

Because of the absence of NHEJ machinery in Plasmodium (see sec-

tion 1.8.1), not all of the CRISPR approaches used in other eukaryotic

systems are possible in Plasmodium. It is always necessary to include a

template for repair, precluding mass-gene disruption by transfection

with libraries of guide RNAs.

Nevertheless this technology is already transforming Plasmodium

research and is likely to profoundly change approaches to malaria

reverse genetics as it is more widely adopted. In combination with

the high transfection efﬁciency of P. knowlesi it might be a powerful

approach.

1.10 approaches to enhance transfection further 43

Genomic DNA

Guide RNA

Cas9 protein

PAM site (NGG)

Figure 16: Schematic of CRISPR/Cas9 function

A guide RNA containing a binding region and a structural region

binds to the Cas9 protein and directs it to a region of the genome.

Once bound the Cas9 protein’s endonuclease activity cleaves the

DNA to create a double-strand break.

44 introduction

1.11 aims and objectives

My goal in this project was to genetically investigate proteins in-

volved in Plasmodium invasion at the largest scale currently possible,

using PlasmoGEM technology, and in the process to develop new ap-

proaches and tools. To these ends I had three main aims.

1.11.1 AIM 1: To conduct a screen for growth-phenotypes for putative in-

vasion related genes using PlasmoGEM vectors

Since large scale genetic disruption of invasion-related genes has not

previously been conducted, I aimed to perform such a screen using

the rodent model P. berghei. In Chapter 3 I will describe this screen,

which used barcode-sequencing to investigate the effect of deleting

each of 145 genes for which there is evidence of possible involvement

in invasion, using the in vivo barcode-sequencing technique that has

been developed in P. berghei. I also followed up a number of unex-

pected results with individual transfections.

1.11.2 AIM 2: To bring PlasmoGEM technology to P. knowlesi

There are potential limitations to analysing invasion in vivo and so I

also aimed to adapt the PlasmoGEM system for the generation and

transfection of constructs targeting the P. knowlesi genome. In Chap-

ter 4 I will discuss my optimisation of these techniques and the ul-

timately successful deletion and tagging of invasion-related genes in

this parasite with PlasmoGEM vectors.

1.11.3 AIM 3: Bioinformatic analysis of phenotypes based on protein-protein

interactions

Finally, the production of large scale phenotype resources necessitates

the development of new approaches to extract maximal insights from

these datasets. In Chapter 5 I will describe my efforts to uncover pat-

terns within my phenotyping data using protein-protein interaction

and transcriptomic datasets and to predict and explain mutant phe-

notypes on the basis of these relationships.

M AT E R I A L S A N D M E T H O D S

2.1 production of dna vectors

2.1.1 Cloning of P. knowlesi Gateway cassette

The PbGWR6K3HA vector, used for the production of P. berghei vec-

tors was digested with KpnI and PstI to release the P. berghei selection

cassette. Three fragments were ampliﬁed by PCR:

• P. knowlesi Hsp70 5’ UTR (primers: TPR321, TPR322; template: P.

knowlesi genomic DNA)

• Dhfr/yfcu fusion (primers: TPR323, TPR324; template: PbGWR6K3HA)

• P. knowlesi Hsp70 3’ UTR (primers: TPR325, TPR326; template: P.

knowlesi genomic DNA).

These primers added overlaps so that the four fragments could be

assembled by Gibson Assembly [74] (NEB Gibson Assembly kit). The

product was transformed into One Shot Pir2 chemically competent E.

coli (Invitrogen).

The ﬁnal construct is in the R6K backbone, meaning it can only

propagate in Pir+ E. coli ensuring no carryover in vector production.

The plasmid was prepared for subsequent work using a Qiagen Plas-

mid Maxi Kit.

2.1.2 pJAZZ vectors

The PlasmoGEM approach employs recombineering and Gateway tech-

nology to construct high-efﬁciency targeting vectors for Plasmodium

genes (see section 1.9). In the course of this project PlasmoGEM vec-

tors were used to target P. berghei and, for the ﬁrst time, P. knowlesi.

One advantage of this method is that it is highly scalable—constructs

can be produced in a production setting by dedicated staff; but equally

the protocol can be carried out in tubes to process small numbers of

46 materials and methods

samples. Early proof of concept vectors I produced myself as follows.

Later, as scale increased, constructs were produced by essentially the

same procedure but in 96-well plates, by Gareth Girling and Burcu

Anar from the PlasmoGEM team.

2.1.2.1 Initial vector

TSA strain E. coli containing the desired region of Plasmodium genome

in a pJAZZ backbone were streaked from archival plates and acted

as the starting point for construct generation. An overnight culture

was prepared from this clone in Terriﬁc Broth (TB) with 30 µg{ml

kanamycin.

2.1.2.2 Transformation with recombinase expression vector

The next day this culture was made electrocompetent as follows: it

was diluted to an OD

600

of 0.05 and incubated at 37

C with shak-

ing until the OD

600

reached 0.6–0.8. At this point the culture was

placed on ice for 15 minutes before 1.4 ml was centrifuged (5000g, 2

mins, 4

C), and washed with ice cold H

O. The washing procedure

was repeated 2 further times, then cells were resuspended in 50 µl of

0.5 ng{µl pSC101gbdA plasmid and transferred to a 1 mm gap electro-

poration cuvette. The cuvette was pulsed (1800V, 25 uF, 200 Ω) and

950 µl of TB was added; cells were allowed to recover at 30

C with

shaking for 70 mins before the addition of antibiotic containing TB

(ﬁnal concentrations: 30 µg{ml kanamycin, 5 µg{ml tetracycline )

2.1.3 Ampliﬁcation of Zeo/PheS cassette

The position of the ﬁnal Plasmodium cassette is determined by the

location at which a temporary positive-negative bacterial selection

cassette is inserted by recombineering. This is speciﬁed by amplify-

ing the cassette using primers which add overlaps of >50 nucleotides

homologous to the insertion site. These primers also add the barcode

used in barseq experiments.

The reaction is carried out with the following mixture: 15.5 µl H

1 µl 12 ng{µl zeo/pheS plasmid template, 1.5µl Advantage buffer, 2.5 µl

2 µM recUp primer, 2.5 µl 2 µM recDown primer, 0.5µl 10 mM dNTPs,

2.1 production of dna vectors 47

0.5 µl Advantage Taq2 and cycling parameters: 95

C 5’ // 95

C 30" /

C 30" / 72

C 1’30" (x30) // 72

C 10’// 4

C hold.

After ampliﬁcation, residual template was digested by the addition

of 1 µl DpnI and incubation at 37

C for 10 minutes, then salts were

removed by dialysis against water on a 0.1 µM Millipore ﬁlter.

The bacteria transformed with recombinase were diluted to an OD

600

of 0.05 after overnight incubation. They were allowed to grow to an

600

of 0.3–0.4 and then L-arabinose was added to a ﬁnal concen-

tration of 0.2% to induce the pBAD promoter that controls the λ red

recombinase, and the temperature was shifted to 37

C for 40 minutes

(the ﬁrst 5 in a water bath).

The cells were then transfected with the PCR amplicon as described

in 2.1.2.2. This time the recovery period was at 37

C and the ﬁnal

antibiotics added were zeocin and kanamycin.

The next day DNA from the (uncloned) culture of zeocin and kanamycin

resistant bacteria was extracted using a QIAGEN Spin Miniprep kit

(according to the manufacturers’ instructions but with 2x volumes of

P1, P2 and N3 due to the low copy number of pJAZZ vectors).

This DNA is the Gateway intermediate. It can generate a number

of different possible ﬁnal vectors through a Gateway reaction with

different donor vectors.

2.1.4 Gateway reaction and transformation

Gateway reactions were set up with 10 µl intermediate pJAZZ vector

(~30 ng{µl), 1 µl donor (100 ng{µl) in TE, 4 µl LR clonase buffer, 2 µl

LR clonase enzyme mix, 3µl TE. The Gateway reaction was incubated

overnight at 25

C, then 0.5 µl proteinase K was added and the tube

incubated at 37

C for 10 minutes to inactivate the reaction. The prod-

uct was dialysed against ddH2O on Millipore ﬁlters and 5 µl of this

dialysed product was transformed into 50 µl BigEasy-TSA cells with

the parameters described in 2.1.2.2.

The cells were allowed to recover in 1 ml of TB at 37

C for 70

minutes and the bacteria then plated out on YEG-Cl kanamycin agar

plates and incubated overnight at 37

C, to select for bacteria that had

lost the Zeo/pheS cassette in the Gateway reaction (by exchanging it

for the ﬁnal cassette).

48 materials and methods

2.1.5 Clone veriﬁcation

Colonies on plates the next day represented potential ﬁnal targeting

vectors. These were screened by lysate PCR – they were checked for

the presence of a correct recombineering event (primers GW1 and

QCR2), and for the absence of carry-over wildtype plasmid (primers

QCR1 and QCR2) in two seperate PCR reactions.

Reaction components: 5 µl H2O, 2.5 µl colony lysate, 12.5 µl GoTaq

mastermix, 2.5 µl 2µM primer 1, 2.5 µl 2µM primer 2

Cycling parameters : 95

C 5’ // 95

C 30”/ 50

C 30” / 68

C 1’ (x30)

// 68

C 10’ / 4

C hold

2.1.6 Plate-based protocol

Where vectors were generated by plates the method employed was

similar to that above in tubes, but with kits optimised for 96 well

plates. This protocol is described in more detail in [154]. This work

was carried out by Gareth Girling and Burcu Anar.

2.1.7 Quality control

Constructs produced through these processes were veriﬁed by Illu-

mina sequencing (analysis conducted by Frank Schwach) to exclude

constructs in which any of the following events had occurred:

• Any mutations in the barcode or the primer annealing sites

ﬂanking it

• A large deletion

• Any mutation that affects either a protein-coding sequence or

splice site of any gene within the vector

This means that some QC-passed vectors are not base-perfect. The

majority of these cases are insertions or deletions of A or T residues

into homopolymeric repeats. Such changes are probably inevitable

when Plasmodium sequence is maintained in E. coli.

2.1 production of dna vectors 49

reagent volume / µ l

Forward oligo (100µM) 9

Reverse oligo (100µM) 9

10X NEBuffer 2.1 2

Total 20

Table 2: Annealing reaction for the creation of dsDNA inserts encoding

guide RNAs.

2.1.8 Crispr vectors

2.1.8.1 Cloning of mother vector

The starting point for the CRISPR constructs was pDC2 Cas9-U6, de-

veloped by Marcus Lee, with the selection marker changed to yDHODH

by Zenon Zenonos. A PCR reaction was used to amplify the P. knowlesi

U6 promoter with overlaps corresponding to the insertion site in the

destination vector (primers: TPR458, TPR459; template: P. knowlesi ge-

nomic DNA). The two P. falciparum constructs, one with dhfr and one

with yDHODH were cut with SalI and BbsI to remove the P. falci-

parum U6 promoter. I assembled these two fragments together with

Gibson Assembly.

Though heterologous promoters are functional in P. knowlesi, I wanted

to rule out any inefﬁciency due to this and so I ampliﬁed the bidirec-

tional elongation factor 1 alpha promoter primers: TPR467, TPR468;

C 2’ //98

C 20” / 55

C 30” / 60

C 3’ (x35) // 60

C 10’ // and

inserted it with Gibson Assembly between yDODH and Cas9 (after

digesting with AvrII and NcoI).

2.1.8.2 Guide RNA design

2.1.8.3 Cloning of guide RNAs into vectors

Guide RNAs were ordered as two complementary oligos with 4 nu-

cleotide 5’ overhangs complementary to the insertion site. The mother

vector was digested with Type IIS restriction enzyme BbsI to create

the site for insertion of the guide RNA.

The oligos were annealed by setting up the reaction shown in Table

2 and running the following programme in a thermocycler.

50 materials and methods

reagent concentration

RPMI-1640 10.43 g/l

L-glutamine 2 mM

HEPES 7.15 g{l

Hypoxanthine 50 mg{l

Horse serum 10% (v/v)

Albumax 0.5% (w/v)

Sodium bicarbonate 0.24% (w/v)

Table 3: Recipe for P. knowlesi culture medium

1. Incubate at 95

C for 10 minutes.

2. Ramp temperature down by 10

C at a rate of 0.6

C per second.

3. Incubate at this reduced temperature for 1 minute.

4. Repeat steps 2 and 3 until tube reaches 25

From then on the annealed oligos were kept on ice.

The annealed product was diluted 50x in ddH2O, and 1 µl of the

diluted DNA was ligated into 50–100 ng of digested plasmid in a

10 µl reaction with 1 µl T4 ligase (NEB) and then transformed into

OneShot TOP10 cells according to the manufacturer’s protocol.

2.2 plasmodium knowlesi in vitro culture

2.2.1 Routine culture

Parasites were maintained in P. knowlesi culture media (Table 3). The

horse serum was added to the medium last and was not ﬁlter ster-

ilised. All other components were. Except where otherwise speciﬁed,

parasites were maintained at 2% haematocrit.

Leukocyte-depleted O+ blood was obtained from the National Blood

Service approximately weekly. Since P. knowlesi is dependent on the

DARC receptor for successful invasion blood was screened by a ba-

sic invasion assay prior to use in important cultures – schizonts were

puriﬁed and allowed to invade the new blood, then a Giemsa smear

was made to check this was proceeding at normal levels.

plasmodium knowlesi in vitro culture 51

2.2.2 Routine pelleting

When necessary to wash, or change media, parasites were pelleted by

spinning at 1100g for 5 minutes.

2.2.3 Schizont puriﬁcation

A Nycodenz stock solution was made by dissolving Nycodenz (or

alternatively Histodenz, which is chemically identical) at a concen-

tration of 27.6 g{l. HEPES was added to a ﬁnal concentration of 1

mM and the pH adjusted to 7.2 with hydrochloric acid. A cushion

solution was then prepared by making a 55% dilution of this stock in

RPMI-1640.

In order to purify schizonts, cultures were pelleted as in 2.2.2 and

resuspended in up to 9 ml RPMI-1640. This was overlaid onto 5ml

of cushion solution and the gradient was centrifuged at 1500g for 10

minutes with a low brake setting.

Schizonts settled at the interface and were extracted with a pasteur

pipette. They were washed by adding at least one volume of RPMI-

1640 and pelleting as in 2.2.2.

2.2.4 Synchronisation

It is reported (Rob Moon, personal communication), that sorbitol-

lysis of late stages – the primary synchronisation method used for

P. falciparum – is not effective for P. knowlesi. Because of this, the pri-

mary method of synchronisation used for this work was Nycodenz

puriﬁcation of schizonts as described above, and then selecting either

the schizont-containing interface or the pellet containing the rings

and trophozoites for continued culture.

2.2.4.1 Narrow window synchronisation

For some procedures, in particular transfection, it was necessary to

obtain very synchronous cultures in which one could say with cer-

tainty that (almost) all parasites were within a deﬁned range of ages.

Here a double synchronisation procedure was employed.

52 materials and methods

First, schizonts were puriﬁed as described in section 2.2.3. These

schizonts were added to fresh blood and allowed to reinvade for 60

to 90 minutes with shaking. Over this time a signiﬁcant portion of

schizonts rupture to release merozoites which reinvade and result in

rings. The culture was examined to ensure that a signiﬁcant number

of rings had formed and then the schizont puriﬁcation was repeated

but this time the schizonts were carefully removed, and the pellet

resuspended in complete media and returned to culture. This culture

contains only very young rings, as shown in Fig. 17.

(a) Mixed culture with mostly late

stages

(b) Schizonts puriﬁed by centrifuga-

tion onto a Nycodenz cushion.

utes of reinvasion, showing young

rings and unruptured schizonts

(d) Young rings after purifying away

unruptured schizonts. These para-

sites are all synchronised to within

90 minutes.

Figure 17: Representative Giemsa-stained smears from the various stages of the

tight synchronisation protocol

2.2.5 Preparation of DNA for transfection

For transfection of pJAZZ vectors, 100 ml of Terriﬁc Broth (TB) was

supplemented with Kanamycin (30 µg{µl). A 10% w/v stock of L-

arabinose was prepared in water and ﬁlter-sterilised. 100µl was added

plasmodium knowlesi in vitro culture 53

per 100 ml of TB. L-arabinose transiently increases the copy number

of pJAZZ vectors for a higher yield.

Cultures were incubated for 16 hours at 37

C with shaking. DNA

was then prepared using a QIAFilter Plasmid Midiprep kit, accord-

ing to the manufacturer’s instructions. The pellet was resuspended

in 50 µl TE, quantiﬁed, and cut with NotI-HF (NEB). DNA was pre-

cipitated with 10% sodium acetate and 0.7 volumes isopropanol at

´20

C, and washed with 70% ethanol before it was dissolved in 15 µl

sterile TE.

2.2.6 Transfection

Parasites were synchronised within a narrow window (as described

in 2.2.4.1) on the day before each transfection. Around 26 hours af-

ter parasites had been returned to culture they were smeared and

Giemsa-stained. The culture was stained regularly (every hour, then

every 30 minutes) until it reached a stage where the majority of par-

asites were very late schizonts, with haemozoin massing at a single

point per parasite, and the ﬁrst rings were appearing from invasion

of the most rapidly maturing schizonts. At this point, schizonts were

puriﬁed as described in 2.2.3.

Transfections were typically carried out in batches: schizonts were

diluted in complete media such that each 1 ml of media contained

5–10 µl of schizonts. A 1 ml aliquot of this schizont-containing media

was prepared in an eppendorf tube for each transfection.

For each transfection, one of these 1 ml aliquots was centrifuged

(2000 rpm, 90 seconds) to pellet the schizonts. The supernatant was

removed and schizonts were resuspended in 100 µl transfection mix

(100 µl Lonza nucleofector P3 mixed with up to 10 µl DNA in TE).

100 µl of the schizont-nucleofector-DNA mix was transferred to a nu-

cleofector cuvette and then electroporated with programme FP-158

within 30 seconds. The contents of the cuvette were immediately re-

moved and added to a pre-warmed tube containing 650 µl complete

media and 150 µl packed freshly washed erythrocytes. This tube was

placed on a thermomixer at 37

C and incubated with shaking for 30

minutes to one hour. After this time, the contents of the recovery tube

were added to 5 ml of pre-warmed media in a well of a 6-well plate.

54 materials and methods

Selection was initiated on day 1 or 2 post-transfection, with media

changed daily for the ﬁrst 5 days post-transfection.

2.2.7 Limiting dilution cloning

Parasite cultures were prepared at a haematocrit of 2% and a para-

sitaemia of 1 ´ 3%. They were counted by live SYBR-green staining

as in section 2.2.8, or by counting parasitaemia in 1000 erythrocytes

by Giemsa smear, and diluted successively in RPMI-1640 to a par-

asitaemia of 0.5 parasites per 10 µl. 10 µl was then transferred into

200 µl of 2% haematocrit media in a 96 well plate. Media was changed

every 3 days. Approximately 2 weeks after the plates were set up, the

media in some wells would turn yellow due to the presence of a popu-

lation of parasites. These wells were screened by Giemsa smear, found

to be positive and transferred to larger cultures. Negative wells were

left for an additional week and then screened by Giemsa smear to

conﬁrm the absence of parasites—indicating positive wells are likely

to indeed be clonal.

2.2.8 Live SYBR green ﬂow cytometry

The starting point for staining was in each case a 96-well plate with

wells containing 50 µl of parasite culture at 2% haematocrit. Staining

was initiated by adding 15 µl SYBR green diluted 1:1000 in RPMI-

1640. Plates were then incubated at 37

C for 1 hour. They were then

washed twice in 200 µl PBS, and diluted 1:10 into 100 µl PBS before

acquisition using a BD FACScalibur ﬂow cytometer.

Analysis was performed using FlowJo. Cells were ﬁrst gated for for-

ward and side-scatter to separate erythrocytes from any debris. This

erythrocyte population was then divided into infected and uninfected

cells based on SYBR-Green ﬂuorescence detected in the FL-1 channel

to calculate parasitaemia.

2.2.9 Luciferase assay

Culture volume to be analysed was pelleted as in 2.2.2. The super-

natant was removed and 1 ml of 0.15% saponin was added to lyse

plasmodium knowlesi in vitro culture 55

the red blood cells. The tube was incubated at room temperature for

ﬁve minutes, then spun at 2000g for 3 minutes and the supernatant

discarded. The pellet was washed in PBS until the supernatant was

no longer red (2000g, 3 minutes). The parasite pellet was then resus-

pended in 65 µl 1x Passive Lysis Buffer (Promega 5x Passive Lysis

Buffer diluted in water), and incubated with shaking for 10-15 min-

utes at room temperature to lyse parasites. The mixture was pipetted

up and down to mix, then centrifuged (1000g, 3 minutes) to clear the

supernatant. Triplicate 20 µl samples were pipetted into the wells of

a luminometer plate. Immediately before reading 50 µl of Promega

Bright-Glo reagent were added to each well. Luminescence counts

were measured for 10 seconds per well.

2.2.10 Genotyping

2.2.10.1 DNA extraction by phenol-chloroform extraction

Where a maximal yield of DNA was needed from low parasitaemia

samples, DNA was puriﬁed by phenol-chloroform extraction. Red

blood cells were lysed by adding 10 pellet volumes of ammonium

chloride lysis buffer 0.15 M NH

Cl, 0.01 M KHCO

, 1 mM Na

EDTA;

pH 7.4) and incubating on ice for 5 minutes. The lysate was cen-

trifuged (5000g, 5 minutes) and the parasite pellet resuspended in

500 µl TNE buffer (50 mM Tris-HCl, 100 mM NaCl, 5 mM EDTA; pH

8.0)

The lysate was transferred to a PhaseLock gel tube and 500 µl of

25:24:1 phenol:chloroform:isoamyl alcohol was added. The tube was

mixed by thorough shaking and then centrifuged (10000g, 5 minutes).

The aqueuous supernatant was tranferred to a new PhaseLock tube

and 500 µl of chloroform-isoamyl alcohol was added. The tube was

mixed thoroughly and then centrifuged at 10000g for 5 minutes. The

aqueous phase was transferred to a new tube, 3 µl pellet paint was

added and the DNA was precipitated overnight at ´20

C after the

addition of 500 µl isopropanol.

The isopropanol was removed and the pellet dried and then resus-

pended in 50 µl or 100 µl TE.

56 materials and methods

2.2.10.2 DNA extraction by QIAGEN kit

Where lower yields of DNA were viable the Qiagen DNeasy Blood

and Tissue kit was used according to the manufacturer’s instructions.

For whole-genome sequencing large cultures were lysed with 0.15%

saponin to maximise yield without overloading the columns.

2.2.11 Genotyping by short range PCR

To conﬁrm the presence or absence of amplicons shorter than 3 kb,

25 µl PCR reactions with GoTaq Green Master Mix were carried out.

The cycling conditions were: 95

C 2’ //95

C 30” / 55

C 30” / 68

C 1’

per kb (x35) // 68

C 10’ //

2.2.12 Genotyping by single barcode PCR

To ensure parasites which came up were not the result of cross-contamination

with other PlasmoGEM parasites, nor cross-contamination of vector

DNA prior to transfection, the barcode from the PlasmoGEM vector

transfected was on occasion ampliﬁed and Sanger Sequenced (GATC

Biotech.). Ampliﬁcation was carried out with GoTaq green (2.2.11)

and DNA was puriﬁed with a QIAGen MinElute PCR puriﬁcation

kit. The primers used were arg97 and PbGW2, with the sequencing

read made with PbGW2.

2.2.13 Genotyping by integration PCR

To check for the presence of successful integration into the parasite

genome long range PCR was carried out with GoTaq Long PCR Mas-

ter Mix, according to the manufacturer’s instructions.

2.2.14 Genotyping by quantitative PCR

Reactions were set up in white LightCycler plates with 10 µl Bio-

Rad SsoAdvanced universal SYBR Green supermix, 1 µl 10 µM for-

ward primer, 1 µl 10 µM reverse primer, 2 µl DNA template and 6 µl

plasmodium knowlesi in vitro culture 57

nuclease-free water. Where error bars are shown they are the 95%

conﬁdence interval calculated from 3 replicates.

The light-cycler programme was 95

C 5’ // 95

C 10”/ 57

C 10”

/ 72

C 20” (x45)). A melt curve was then carried out to ensure the

presence of a single amplicon.

2.2.15 Genotyping by whole-genome sequencing

Genomic DNA was extracted from parasites by saponin lysis of a

large culture (>50ml at >5% parasitaemia), resuspending the parasite

pellet in 200ul of PBS and processing with a QIAGEN DNeasy Blood

and Tissue Kit as per the manufacturer’s instructions for a blood sam-

ple.

Sequencing was carried out by the Illumina team at the Sanger

Institute. DNA was quantiﬁed by Qubit Fluorometric Quantitation,

then sheared into 400–600 bp fragments. Libraries were prepared

with the NoPCR protocol. [111] Multiplexed samples were sequenced

on the Illumina MiSeq platform for 150 paired ends cycles accord-

ing to the manufacturer’s protocol. Data were analysed from the Illu-

mina sequencing machines using RTA1.6, RTA1.8 or GA v0.3 analysis

pipelines.

All downstream analysis I performed myself, as outlined below.

2.2.15.1 Mapping target

All P. knowlesi H strain chromosomes, and Mitochondrial, Apicoplast

and bin sequences from the March 2015 genome release were down-

loaded in EMBL format. A custom EMBL ﬁle was created by concate-

nating these and also the gateway cassette used in transfection, and

the arms of the pJAZZ-OK vector. A mapping index was constructed

for this custom genome with bowtie2-build.

2.2.15.2 Mapping

CRAM ﬁles produced by the Sanger sequencing pipeline were con-

verted to FastQ ﬁles for customised analysis.

Bowtie2 was used to map these to the customised EMBL ﬁle de-

scribed above as follows:

58 materials and methods

/software/bowtie2/bin/bowtie2 -x union -1 ../$1

1.fastq -2 ../

2.fastq -S $1.sam

samtools view -bS $1.sam | samtools sort - $1.bam.sorted

samtools index $1.bam.sorted.bam



Mapped reads were analysed with Artemis to observe differences in

coverage and points where split-reads indicating junctions between

transfected parasite and reference.

2.2.15.3 Assembly

Reads were also de novo assembled using velvet as follows, using a

k-mer size of 51:

velveth $assemblyfile 51 -shortPaired -fastq -separate $1

1.fastq

2.fastq

velvetg $assemblyfile



2.2.16 Immunoﬂuorescence

2.2.16.1 Fixation and labelling

All steps were carried out with agitation.

Parasites were ﬁxed in a ﬁxing solution comprising 4% paraformalde-

hyde and 0.0075% glutaraldehyde in PBS [193] for 30 minutes. They

were then permeabilised for 10 minutes at room temperature with

0.1% Triton-X in PBS, and blocked for 1 hour with 3% BSA in PBS

at room temperature. They were labelled for 18 hours with FITC-

conjugated goat anti-HA at 4

C and then washed three times with

PBS for ten minutes at room temperature. Parasites were placed on

a slide and allowed to air dry and then mounted with Prolong Gold

mounting reagent containing DAPI. f

2.3 plasmodium berghei

2.3.1 Pooled midiprep

To minimise expense and labour in DNA production, a pooled midiprep

procedure was adopted to produce the DNA for transfection in P.

berghei experiments. Growth blocks were set up with 1 ml of TB-kan

2.3 plasmodium berghei 59

per well, and for each transfection two wells were inoculated with

each vector to be used in an experiment. After 16 hours of growth at

C with shaking, all wells from a single experiment were pooled

together and then processed with a QIAFilter Midi kit (according to

manufacturer’s instructions but using 2x volumes of P1, P2 and P3

due to the use of a low copy vector).

The resulting DNA was spiked with DNA for 7 reference vectors

(prepared by Ana Rita Gomes) – 4 redundant in the blood stages (p25,

p28, soap, p230p) and three known to result in attenuated growth

(plasmepsin IV, PBANKA_140160 and PBANKA_110420). The use

of these universal references allows calibrated comparison across all

PlasmoGEM barseq experiments.

Prior to transfection the pooled DNA was digested with NotI, pre-

cipitated with isopropanol and resuspended in 6 µl TE.

2.3.2 Schizont preparation and transfection

A Wistar rat was infected by intraperitoneal injection and monitored

until blood parasitaemia reached 1-3%. Infected blood was then har-

vested by cardiac puncture and prepared as a schizont culture: 25x

blood volume schizont medium (RPMI 1640 supplemented with 25mM

L-glutamine, 25 mM HEPES, 10mM NaHCO3, 100 U/ml penicillin/strep-

tomycin and 25% foetal bovine serum) was added and the ﬂask gassed

for 90 seconds with malaria gas (1% O2, 3% CO2, 96% N2) and incu-

bated overnight with shaking.

The next day the culture was smeared to check for the presence of

mature schizonts and these were puriﬁed on a 55% Nycodenz/PBS

(v/v) cushion with low brake. The schizont-containing interface was

collected with a Pasteur pipette and washed with schizont medium.

Schizonts were pelleted by centrifugation (500g, 2 mins), and the pel-

let resuspended in 16 µl P3 solution (Lonza). The 6 µl vector pool was

then added to this mixture and transfection was carried out in 16-

well Lonza cuvette strips with programme FI-115. The transfection

mixture was immediately injected intravenously into the tail veins of

6-8 week old Balb/c mice.

The next day drinking water was supplemented with pyrimethamine

to begin selection (7 mg/ml).

60 materials and methods

For each pool three transfections were carried out in parallel and

injected into three mice to act as biological replicates.

The remaining transfection mixture after injection was kept as an

input control for barseq.

Parts 2.3.2 and 2.3.3 of the experiment was conducted by Ellen

Bushell, Ana Rita Gomes and Tom Metcalf; remaining steps I carried

out myself.

2.3.3 Time course

On days 4, 5, 6, 7 and 8, a small sample of blood (<30 µl) was collected

from the tail vein. Samples were collected at the same time on each

day, and Giemsa smears were also made to monitor the infection.

The erythrocytes in these time-point samples were lysed by adding

1 ml of lysis buffer (0.15 M NH

Cl, 0.01 M KHCO

, 1 mM Na

EDTA;

pH 7.4) and incubating on ice for 2 minutes. Parasites were then pel-

leted by centrifugation (3 mins, 1000g). Genomic DNA was isolated

by phenol-chloroform extraction (2.2.10.1)

2.3.4 Barcode ampliﬁcation

PCR was used to amplify barcodes from this extracted genomic DNA.

A 50 µl Advantage 2 Polymerase reaction was prepared according to

the manufacturers instructions with the primers arg91 and arg97.

1 µl of phenol-chloroform extracted gDNA was used as the tem-

plate for the reaction. The protocol for the PCR was as follows: 95

5’ //95

C 30” / 55

C 20” / 68

C 8” (x35) // 68

C 10’ //. A ﬁnal

sample from which barcodes were ampliﬁed in parallel was the ‘in-

put pool’ - the DNA present in the DNA used for transfection. This is

important to ensure that absence of a barcode in the population is not

due simply to a failure of the bacteria carrying that vector to grow.

5 µl of the product from this ﬁrst PCR reaction was then used as

the template for a second PCR which added the Illumina adapters

required for the product to adhere to the Illumina ﬂow cell, and also

index tags to allow the sample corresponding to each read to be de-

1 The collection of the blood from the animals was performed by Ellen Bushell, Ana

Rita Gomes or Tom Metcalf.

2.4 data anal ysis 61

termined after sequencing. The cycling conditions were: 95

C 5’ //

C 30” / 68

C 15”(x10) // 68

C 5’ //

A water control was taken through the entire experiment to ensure

reagents were not contaminated with trace amounts of barcodes.

Following the second PCR, DNA was puriﬁed using a QIAGEN

MinElute PCR puriﬁcation kit. Eluted DNA was quantiﬁed using the

Qubit system and 100 ng of each sample was taken and pooled to-

gether.

Up to 32 timepoints could be pooled on a single MiSeq lane.

2.3.5 Barcode sequencing

Once pooled libraries were further quantiﬁed by qPCR, diluted to

1 nM and spiked with 40-50% of PhiX (due to the low complexity

of the sequence). They were loaded at low cluster density 4 ˆ 10

clusters/mm

). 150 bp paired-end reads were collected. These steps

were conducted by the Sanger Institute Illumina C team.

2.4 data analysis

2.4.1 Barseq

The output from the Sanger sequencing pipeline was converted into

a FastQ ﬁle containing the raw reads using Picard. I used the script

count

tags.pl written by Frank Schwach to read through this FastQ

ﬁle and count the number of barcodes corresponding to each gene for

each time point.

Further analysis was substantially based on ideas developed in [80].

I created an R script

to analyse the barcode data.

The procedure for analysing STM data is as follows:

1. For each day (d, 4-8) and each mouse (m,1-3) calculate what

proportion (p) each gene (g) represents of the total number of

barcodes.

2 8% of lines in the script are unchanged from one written by Ana Rita Gomes, the

rest is original to me

62 materials and methods

d,m,g

“

d,m,g

d,m,G

, where b

d,m,g

represents the number of bar-

codes for gene g counted on day d in mouse m.

We treat any ratio < 0.1% as background and set to 0.

2. For each mouse on days 4-7, calculate the fold-change in bar-

code number for each gene on the next day, i.e. c

d,m,g

“

d`1,m,g

d,m,g

3. For each mouse on days 4-7, calculate a relative ﬁtness for each

gene by normalising this gene’s fold-change to the mean fold-

change of the reference genes. f

d,m,g

“

d,m,g

d,m,R

4. For each gene perform a t-test comparing the dataset represent-

ing the fold-change on days 4-7, for all mice, with the reference

growth rate on days 4-7 for all mice.

5. Assign a phenotype to each gene as follows:

a) If two or more mice have no (or below background) read-

ings on days 4-7, call the phenotype as "Putative essential"

b) If two or more mice lack at least one reading (i.e. these

are below background), call the phenotype as "Inconsistent

(likely essential)".

c) In all other cases at least two mice have a full set of read-

ings on days 4-7. We can now examine the t-test p-values,

adjusted for multiple comparisons with the FDR method.

Where p<0.05 we call the phenotype as "Slow" or (hypo-

thetically) "Fast" depending on whether the relative ﬁtness

is below or above 1. Where p>0.05, we call the phenotype

as "Putative redundant" (in the blood stages).

Part II

R E S U LT S

I N I T I A L I N V E S T I G AT I O N S : G E N E S E L E C T I O N

A N D P. B E R G H E I S C R E E N

3.1 assessing the hu et al. invadome

My overarching aim in this project was to apply high-throughput

reverse genetics to genes hypothesised to play a role in erythrocyte

invasion. As discussed in the introduction, one set of such genes was

described by Hu et al. in 2010 derived from the PlasmoINT interaction

network.

Before beginning experiments I analysed this gene set, to conﬁrm

its suitability for investigation.

3.1.1 ‘Invadome’ genes have expression patterns consistent with a role in

invasion

One naïve heuristic for identifying genes potentially involved in in-

vasion would be to simply select genes expressed late in the intraery-

throcytic development cyle. This is the point at which the merozoites—

invasion machines—are being built. Does the Hu et al. invadome dif-

fer very much from the results of such an approach?

To investigate this I downloaded the 2003 Bozdech et al. microarray

timecourse, which provides values for gene expression in the HB3

strain of P. falciparum at hourly increments over the 48 hour IDC [22].

In this dataset many genes are covered by more than one probe, so

I computed an expression value for each gene at each timepoint by

taking the mean of all probes covering it. I then found the timepoint

of maximal expression for each gene, and subdivided this data de-

pending on whether or not the gene was a member of the Hu et al.

invadome.

Reassuringly this analysis (Fig. 18) showed that the vast majority of

the invadome genes are indeed expressed late in the erythrocytic cy-

cle, as we would expect (these genes are part of a network imputed,

66 initial investigations: gene selection and P. BERGHEI screen

100

150

200

0 10 20 30 40

Timepoint / hours

Genes with peak expression

Figure 18: Genes in the Hu et. al invadome peak late in the IDC, but do not encom-

pass all late-peaking genes.

Histogram showing the peak expression [22] of all P. falciparum

genes and the subset that are part of the invadome (blue).

in part, by correlated expression with other invasion genes which

will themselves peak late in the IDC). The median peak for invadome

genes was at 38 hours. Additionally I saw that only 50% of genes

showing maximal expression at this 38 hour point were part of the

invadome. This may indicate that the perturbation-based approach

used to generate the PlasmoINT network has provided a more spe-

ciﬁc subset of genes than one based entirely on expression over the

course of the IDC (intraerythrocytic development cycle), though it

does not prove this.

3.1.2 Gene Ontology analysis suggests the relevance of the invadome gene

list

I also looked for gene ontology terms which were enriched in this set.

As one might anticipate, the highest enrichment was for "entry into

host" with 8-fold enrichment.

However 11 genes annotated with the “entry into host” term were

not included in the Hu et al. set. I investigated these with literature

searches. Four were sporozoite proteins (CSP, CTRP, TRAP, MAEBL),

3.1 assessing the hu et al. invadome 67

two were involved in midgut invasion (PLP3, PLP5, [58]) and two

have no proven role in invasion other expression at the right stages

(P12, [187] and falcipain 1 [84, 174]).

I had been concerned that given the same organelles are involved

in invasion across the Plasmodium life-cycle, network-based perturba-

tion approach might have identiﬁed proteins which are involved in

cellular invasion, but operate primarily or exclusively in other stages

of the parasite life-cycle. The fact that several such genes are actually

excluded from the Hu et al. invadome provides some reassurance that

this is a relatively speciﬁc erythrocyte invadome.

However, two of the non-included genes with the "entry intro host"

annotation do have clear roles in erythrocyte invasion (falstatin ICP,

[153] and DOC2 [62]). These represent false-negatives, but they ap-

pear relatively few in number.

The other concern is false positives – genes which have no real

role in invasion. A small number of clear false positives stood out

from the rest of the invadome. Pantothenate transporter (PAT), folate

transporter 1 and FCP (which is involved in the food vacuole [138])

all seemed unlikely to actually be involved in invasion.

One undeﬁned question is how intimately a gene has to be involved

in the entry into the host cell to be considered part of the invadome.

There were four ApiAP2 genes in the gene list, these bind to DNA

and so are unlikely to be directly involved in invasion but might well

play key roles in regulating invasion-related genes which would, in

my view, make them rightful inclusions. There were also a number of

genes included with known roles in egress (PKG, SUB1), and it is pos-

sible the PlasmoINT approach does not have the ability to separate

egress and invasion. But these processes are interconnected and may

share components and so analysing them together may be helpful.

3.1.3 A number of PlasmoINT predictions have been borne out by subse-

quent research

There are a number of examples I uncovered in my research of the

invasion literature that anecdotally suggest the predictive power of

the analysis method that produced the PlasmoINT network. Since

the Hu et al. invadome was described in 2010, a number of genes

68 initial investigations: gene selection and P. BERGHEI screen

have been newly identiﬁed to have roles in invasion. Several such

discoveries in recent years were in fact foreshadowed in the Hu et al.

list.

Though the role of MyoA in invasion has been clear for a long

time, Myosin B, and its light chain MLC-B, have only recently been

discovered to have expression and localisation highly suggestive of

an involvement in invasion; [211] both were predicted by Hu et al. to

be involved in invasion purely on the basis of the interaction network.

In addition, the traditional model of the Plasmodium motor complex,

inspired by Toxoplamsa, involved aldolase acting as the link between

the actin cytoskeleton and adhesin proteins. [152] Interestingly the

Hu et al. invasion subnetwork does not include aldolase. Until last

year I would have described such a result as a false-negative. However

in 2014 renewed research in Toxoplasma identiﬁed aldolase as essential

for metabolism but not invasion [173]. In light of this, a new candidate

glideosome-adhesin connector (TgGAC) has been suggested to play

this key role across the Apicomplexa. [95] The P. falciparum orthologue

of this gene had already been included in the Hu et al. invadome,

though it was at the time a wholly uncharacterised protein. Thus

the Bayesian approach excluded a false-positive in the literature and

instead included the real interactor.

In sum, though the PlasmoINT invadome is clearly not a perfect

list of the invasion machinery, it appears a good starting point for

a screen-based approach to invasion. While a curatorial approach

would yield an invadome more in line with the literature, it would

exclude the genes that it is most important to investigate – those we

currently know nothing about.

3.2 selecting the core invadome

Hu et al. deﬁned their putative invadome with 418 P. falciparum genes,

but I wanted to analyse the core invasion machinery shared among

many Plasmodium species. To identify this core invadome, I searched

for those members of the Hu et. al invadome showing 1:1:1 orthol-

ogy between P. falciparum, P. berghei and P. knowlesi. I used a phyletic

pattern search on OrthoMCLDB [118] to retrieve this dataset, sum-

marised in Fig. 19. The search string used was “pber+pfal+pkno=3T

3.2 selecting the core invadome 69

AND pber+pfal+pkno=3”. The ﬁrst half of the expression requires that

an ortholog group contains at least one member from each of the

three target species. Given this, the second half of the expression en-

sures that they do not contain more than one. This approach would,

I hypothesised, select genes which for the most part played identical

roles in the three organisms in question.

P. knowlesi

196

135

100127

4,082

134

P. falciparum P. berghei

Figure 19: The majority of the Plasmodium genome is 1:1:1 orthologous across P.

berghei, P. falciparum and P. knowlesi.

Note that there are two reasons an orthologue group might be

excluded: it could contain no members in the target organism or

it could contain more than one member.

I then identiﬁed the subset of these 4,082 genes contained within

the 418 gene Bozdech invadome. This yielded 320 genes as members

of the ’core invadome’. I hypothesised that such an approach would

select the evolutionarily conserved ancestral genes involved in many

of the key processes of invasion. However I was aware that certain

important invasion genes would be excluded as a result. Invasion

takes place at the parasite-host interface and so involves a number of

rapidly evolving genes families which do not show 1:1:1 orthology.

Notable examples include invasion ligands, where there is a vast se-

lective pressure for the evolution of novelty and evolutionary arms

races take place between parasite and host. However the aim of this

study was to carry out a pan-Plasmodium assessment of conserved

70 initial investigations: gene selection and P. BERGHEI screen

Not essential Deletion failed Not attempted

Figure 20: Prior to this study few attempts to disrupt genes in the core invadome

had been reported.

Each icon represents a single gene. Its color represents known tar-

getability in P. berghei prior to this work. Data are from RMGMdb.

[105]

genes and the >75% of the invadome that is included in this set is a

good list of candidates for such analysis.

3.2.1 Few invadome genes have been previously analysed experimentally

in P. berghei

To put into context my experimental work I ﬁrst searched for what

was known already about the targetability of these genes.

As of July 2015, the RMGM database contained records for at-

tempts to knock out 52 of these 320 genes in P. berghei, with 23 of these

attempts successful and 29 unsuccessful (Fig. 20). Less than 20% iof

the core invadome has therefore previously been studied using exper-

imental genetics, consistent with the limited extent of genetic studies

across the entire genome.

This is likely a slight underestimate as the database may not be

comprehensive. If data from P. falciparum are included also then tar-

getability is described for an additional 18 genes, bringing the to-

tal coverage to 22%. However because P. falciparum has been less

amenable to transfection we can be less confudent in the calls of ’es-

sential’ genes in this species. Regardless, it is clear that high-throughput

1 8 orphan genes were added back for P. berghei phenotyping: PBANKA_031660, PBANKA_090080

PBANKA_100010 PBANKA_133270 PBANKA_134910 PBANKA_110140 PBANKA_102250

3.3 barseq in p. berghei 71

reverse genetics had the potential to reveal a great deal more about

this set of genes.

3.3 barseq in p. berghei

The PlasmoGEM project has proceeded furthest in P. berghei, because

of the ease with which the parasite’s genome can be manipulated.

[154] Knock-out vectors exist for the majority of genes, with coverage

growing each week, and so this system seemed the best starting point

for a high-throughput investigation of the invadome. In particular,

the barseq protocol had been established in P. berghei and used to

great effect. [79]

3.3.1 Barseq experiment 1

I will discuss the ﬁrst barseq experiment in some detail to establish

the procedure followed and its strengths and weaknesses.

All P. berghei vectors were produced by the PlasmoGEM recombi-

neering team using the plate-based recombineering method. [155] I

prepared a pool of 48 vectors (listed in appendix A.1) for this exper-

iment; these were mixed with the seven growth rate control vectors

(targeting p25, p28, soap, p230p, plasmepsin IV, PBANKA_140160 and

PBANKA_110420) and transfected into parasites which were injected

into three mice, as described in section 2.3.

All three mice devel-

oped a parasitaemia under pyrimethamine pressure which became

visible on day 4 and steadily increased. Small blood samples were

taken daily from day 4 to day 8 for DNA extraction. I ampliﬁed bar-

codes from these samples and they were sequenced on an Illumina

ﬂow cell, along with those from the input DNA.

3.3.1.1 Barseq generates a rich dataset

In the ﬁrst stage of analysis I counted barcodes belonging to each

vector at each timepoint in each mouse and normalised these to that

sample’s total number of barcodes, to determine the percentage com-

2 All steps involving animals were conducted by Ellen Bushell, Ana Rita Gomes and

Tom Metcalf

72 initial investigations: gene selection and P. BERGHEI screen

10%

12%

14%

16%

18%

input

day 4

day 5

day 6

day 7

day 8

Gene targeted

Ratio

Figure 21: Mutant abundances at each timepoint in one mouse

This plot illustrates the rich data provided by a single animal in

a barseq experiment. 324 data points represent the proportion of

each barcode at each time point. Three vectors are highlighted,

one essential (red), one redundant (green) and one with attenu-

ated growth (yellow). Control genes are indicated in grey. This

ﬁgure is intended as an illustration of the data provided by a sin-

gle mouse, but raw barcode ratios are shown for every gene in

every mouse in the appendix.

position of each knock-out in the population. These raw ratios from

a single mouse can be seen in Fig. 21.

The proportion of each vector in the input is relatively constant,

indicating that the pooled midiprep procedure (2.3.1) has not overly

biased our experiment. There are 7 genes with lower input values –

these are in fact the 7 controls that were added in. There must have

been an error of quantitation, either an overestimation of these con-

trols or an underestimation of the other vectors which has resulted in

these spike-ins being present at 11% of the abundance of other vec-

tors. Nonetheless this did not prevent the experiment from working,

as these genes still gave correct growth phenotypes (slow/redundant

depending on the vector).

Despite the presence of barcodes for all genes of interest in the

input, by day 4 many vectors show no barcodes whatsoever (e.g.

PBANKA_133890). This is because naked DNA is cleared rapidly

from the plasma in vivo [171], thus the only way in which a barcode

barseq in P. BE RGHEI 73

can persist is if it is carried inside a parasite. P. berghei has a high

propensity for integrating linear homologous DNA into its genome,

meaning that vectors persist by knocking out their target genes (there

is no evidence for episomal carriage of PlasmoGEM vectors in P.

berghei). Where this target gene is essential parasites will die, and ei-

ther lyse or be cleared by the spleen, removing the barcodes from the

circulation. Thus we can speculate that those genes which show no

barcodes on day 4 may be essential.

The analysis is not completely straightforward, however. We see

that even where mutants grow at similar rates, and have similar input

ratios (e.g. compare PBANKA_031480 and PBANKA_133580), they

may have quite different representations on day 4. This demonstrates

that different vectors integrate into the parasite genome with differ-

ent efﬁciencies. Efﬁciency is in part a factor of the length of vector

arms [154] but also may be affected by the state of chromatin around

the gene. We therefore cannot exclude the possibility that we have

false negative results from vectors which are unable to integrate. Dis-

proving this possibility, and conﬁrming essentiality, always requires a

conditional or complementation approach. That said, previous proof-

of-concept experiments for the barseq technique showed concordance

previous screens. [79] To be accurate however, where barcode levels

are below the detection threshold by day 7 genes are described as

“likely essential”, rather than essential, which cannot be deﬁned by

this technique.

By looking at the shape of the graph beyond day 4 we can begin

to get a sense of the dynamics of growth occurring within the mouse.

Some genes show barcode counts which increase continually over the

course of the infection (e.g. PBANKA_031480); these are likely to be

almost entirely redundant.

Others appear to decline in proportion over the course of the in-

fection (e.g. PBANKA_136440). These represent slow-growing knock-

out parasites. Although they are growing, and so persist to day 4 and

beyond, they are continually being outcompeted by the redundant

knock-outs, such that they would ultimately drop below detectable

74 initial investigations: gene selection and P. BERGHEI screen

0.0

0.5

1.0

1.5

PBANKA_010290

PBANKA_020460

PBANKA_030600−ko

PBANKA_030670

PBANKA_031140

PBANKA_041600

PBANKA_052170

PBANKA_072050

PBANKA_080220

PBANKA_080960

PBANKA_082870

PBANKA_083560

PBANKA_090380

PBANKA_091030

PBANKA_091170

PBANKA_102010

PBANKA_103540

PBANKA_111530

PBANKA_113780

PBANKA_122740

PBANKA_133890

PBANKA_122440

PBANKA_072150

PBANKA_140160

PBANKA_082490

PBANKA_124060

PBANKA_110420

PBANKA_120200

PBANKA_103440

PBANKA_081700

PBANKA_110650

PBANKA_090590

PBANKA_100240

PBANKA_051490

PBANKA_136440

p230p−tag

PBANKA_142310

PBANKA_123730

PBANKA_093210

PBANKA_062240

PBANKA_112810

PBANKA_145910

PBANKA_120800

PBANKA_031480

PBANKA_122540

PBANKA_133660

PBANKA_103780

PBANKA_130520

PBANKA_051520

PBANKA_146060

PBANKA_051500

PBANKA_133580

PBANKA_110690

Gene targeted

Fold-change

Mouse

Day

5 - 6

6 - 7

7 - 8

Figure 22: Fold-changes of all genes in barseq experiment 1

Error bars depict 95% conﬁdence interval calculated across all

days and all mice. This diagnostic plot allows a simple assess-

ment of whether growth effects apply across all mice and of

whether such effects change over the course of the infection.

There is more uncertainty over the growth rate of slow-growing

mutants than redundant ones because these growth rates are in-

ferred from smaller numbers of parasites, leaving noise propor-

tionally higher.

levels by barseq (despite the fact that the actual number of parasites

is steadily increasing).

Finally, a group of genes have no barcodes by day 7, and are called

as ”Likely essential”.

These dynamics deﬁne the limit of the most extreme phenotypes

that can be detected. An extremely slowly growing parasite might

already be competed to below the detection threshold early in the

experiment, leading it to be measured as putatively essential.

3.3.1.2 Barseq assigns numerical values to parasite ﬁtness

The style of analysis above is qualitative, but helpful for understand-

ing the strengths and limitations of the barseq approach. However a

key strength of barseq is the ability to construct an analysis pipeline

which with minimal human intervention gives robust and highly

quantitative phenotyping results.

3 The steady nature of the trend in parasite abundance, and the consistency between

mice, demonstrates that these effects are the result of selection and not random

genetic drift.

barseq in P. BE RGHEI 75

0.0

0.5

1.0

1.5

2.0

4 5 6 7

Day

Fold-change

0.6

0.8

1.0

1.2

Mean fold-change

Figure 23: Line graph depicting how fold-changes themselves change over time

Each line depicts a gene, its colour determined by its mean fold

change. It is clear that for likely redundant genes, with fold-

changes around 1, the most consistent data comes towards the

end of the infection where there are large numbers of mutant

parasites, reducing the effect of noise.

The approach I adopted, highly inspired by [80] but re-implemented

in a more self-contained bioinformatic approach, is described in 2.4.1.

I developed a pipeline which in a single click can go from the fastQ

output of an Illumina sequencing machine to assign phenotypes ”Likely

essential”, ”Redundant” and ”Slow growth” and also along the way

produce a number of diagnostic graphs.

The ﬁrst stage is to compute fold-changes in barcode ratios from

one day to the next, i.e. day 4 to day 5, day 5 to day 6, day 6 to

day 7 and day 7 to day 8. These values are shown in Fig. 22 (one

of the diagnostic plots produced by the script). We can see that on

this basis alone we can get a good sense of whether a gene appears

to be redundant, likely essential or will result in slow growth when

disrupted.

However, these fold-change data are not a consistent value of a par-

ticular knock-out but rather the result of an interaction between the

ﬁtness of the mutant and the background against which it is compet-

ing; a mutant can only increase in relative abundance at the expense

of some slower growing mutant. Therefore the distribution of fold-

changes shifts over the course of the infection.

This effect is shown in Fig. 23. On day 4 redundant genes have fold-

changes substantially above 1, as they increase at the expense of slow

76 initial investigations: gene selection and P. BERGHEI screen

Day

Fold- change

4 5 6 7

0.0

0.5

1.0

1.5

2.0

PBANKA_140160

PBANKA_110420

PBANKA_103440

PBANKA_103780

PBANKA_051500

p230p-tag

PBANKA_051490

Figure 24: Fold-change values for the six control constructs are as predicted.

Values shown are the mean from three mice. Predicted (and

demonstrated in [80]) to be normally-growing mutants(shown

in green) do indeed have higher fold-changes than attenuated-

growth controls (shown in yellow).

growing mutants. By the end their fold-change drops towards 1, be-

cause slow-growing mutants are such a small part of the population

that there is very little change in the proportion of normally-growing

mutants over time (e.g. a change from 98% to 99%).

Because of this effect the fold-change can tell us only about the

relative growth of different mutants on each day of the experiment.

3.3.1.3 Control genes allow the normalisation of ﬁtness values

To gain an absolute view of mutant’s ﬁtness, we can normalise its

growth to mutants generated in the same experiment but with vectors

which produce known ﬁtness effects. The same 7 vectors are added

to each barseq transfection to act as a universal comparator between

experiments (see section 2.3.1).

We can analyse these genes as an initial check on the experiment

(Fig. 24). I found that the slow-growth controls did indeed grow more

slowly than their redundant counterparts, giving some conﬁdence in

the results of the experiment. These results reﬂect the ratio between

the abundance on two days. The abundances on the ﬁnal days are

the result of compounding these results over all preceding days and

so the graphs shown in the appendix show a markedly higher ﬁnal

abundance for the normal-growth controls as compared to the slow

controls.

3.4 barseq growth phenotypes 77

One can also see that there is a ’sweet-spot’ for the acquisition of the

best growth phenotype data. On day 4 there is a great deal of noise,

likely due to PCR ampliﬁcation from a low number of molecules and

on day 7 the proportion of slow growing parasites is becoming very

low indeed leading to less accurate results.

With these control genes behaving as anticipated, I used them to

normalise my data. I calculated the mean fold-change of the three

redundant knockouts in each mouse at each timepoint. This can be

set to a ﬁtness of 1, and all fold-changes normalised on this basis.

The ﬁtness values on day 4 are unreliable due to the low para-

sitaemia at this point, so these were discarded. The remaining values

were used to compute a mean ﬁtness for each gene, and to conduct

a t-test against the redundant controls to test for statistically signif-

icant attenuated growth. The resulting p-values were corrected for

multiple comparisons and based on them phenotypes were assigned

as described in 2.4.1.

3.3.2 Barseq experiment 2

Invadome barseq experiments were carried out in two batches. The

ﬁrst, of 48 vectors, was conducted as described above. The second

was carried out to phenotype an additional 97 vectors, with the same

procedure previously described, and early checks again suggested

that the experiment was working correctly (data not shown).

Exactly the same analysis procedure was carried out to identify

statistically signiﬁcant mutant growth phenotypes, and henceforth I

will discuss the entire dataset together.

3.4 barseq growth phenotypes

Both datasets were analysed separately as above and then their growth

phenotypes collated to produce the ﬁnal dataset. In total I was able to

assign potential growth phenotypes to 145 putative invasion-related

genes. Twenty-ﬁve of these already had a growth phenotype described

4 In fact these effects are exacerbated in these genes in this particular experiment,

because the control genes were spiked in at a lower level than other vectors. The

fact that results for these genes are still accurate illustrates the robustness of the

methodology.

78 initial investigations: gene selection and P. BERGHEI screen

on RMGMdb. The remainder are to my knowledge being targeted for

the ﬁrst time in P. berghei.

The results are shown in Table 4. Around half of all genes were

found to be likely essential. Of those that were targetable, 81% showed

no detectable growth phenotype with the remainder showing attenu-

ated growth. We did not detect any genes whose deletion signiﬁcantly

increased growth rates.

Table 4: Barseq growth phenotypes for invadome genes

putative growth

id description phenotype rate

PBANKA_141980 3’,5’-cyclic nucleotide phosphodiesterase, puta-

tive

Essential –

PBANKA_122670 AAA family ATPase, putative Essential –

PBANKA_122760 adenylyl cyclase beta, putative (ACbeta) Essential –

PBANKA_050410 autophagy-related protein 8 (ATG8) Essential –

PBANKA_083560 cAMP-dependent protein kinase catalytic sub-

unit (PKAc)

Essential –

PBANKA_113780 conserved Plasmodium protein, unknown func-

tion

Essential –

PBANKA_072150 conserved Plasmodium protein, unknown func-

tion

Essential –

PBANKA_080960 conserved Plasmodium protein, u. f. Essential –

PBANKA_092940 conserved Plasmodium protein, u. f. Essential –

PBANKA_081510 conserved Plasmodium protein, u. f. Essential –

PBANKA_050180 conserved Plasmodium protein, u. f. Essential –

PBANKA_135610 conserved Plasmodium protein, u. f. Essential –

PBANKA_040640 conserved Plasmodium protein, u. f. Essential –

PBANKA_030110 conserved Plasmodium protein, u. f. Essential –

PBANKA_100920 conserved Plasmodium protein, u. f. Essential –

PBANKA_092250 conserved Plasmodium protein, u. f. Essential –

PBANKA_010260 conserved Plasmodium protein, u. f. Essential –

PBANKA_112410 conserved Plasmodium protein, u. f. Essential –

PBANKA_040690 conserved Plasmodium protein, u. f. Essential –

3.4 barseq growth phenotypes 79

putative growth

id description phenotype rate

PBANKA_091940 conserved Plasmodium protein, u. f. Essential –

PBANKA_134700 conserved Plasmodium protein, u. f. Essential –

PBANKA_051390 conserved Plasmodium protein, u. f. Essential –

PBANKA_100170 conserved Plasmodium protein, u. f. Essential –

PBANKA_141380 conserved Plasmodium protein, u. f. Essential –

PBANKA_010630 conserved Plasmodium protein, u. f. Essential –

PBANKA_141200 conserved Plasmodium protein, u. f. Essential –

PBANKA_101040 conserved Plasmodium protein, u. f. Essential –

PBANKA_122440 ferlin like protein, putative Essential –

PBANKA_133890 glideosome associated protein with multiple

membrane spans 1, putative (GAPM1)

Essential –

PBANKA_103540 glideosome associated protein with multiple

membrane spans 3, putative (GAPM3)

Essential –

PBANKA_111530 glideosome-associated protein 40, putative

(GAP40)

Essential –

PBANKA_091030 guanylyl cyclase, putative (GCalpha) Essential –

PBANKA_040180 HAD superfamily protein, putative Essential –

PBANKA_124060 membrane skeletal protein, putative Essential –

PBANKA_083100 merozoite surface protein 1 (MSP1) Essential –

PBANKA_135570 myosin A (MyoA) Essential –

PBANKA_145950 light chain 1, putative,myosin A tail domain in-

teracting protein MTIP, putative (MTIP)

Essential –

PBANKA_082870 phosphatidylserine decarboxylase, putative Essential –

PBANKA_020460 photosensitized INA-labeled protein 1, puta-

tive

Essential –

PBANKA_102010 PP1-like protein serine/threonine phosphatase,

putative

Essential –

PBANKA_131540 protein phosphatase 2b regulatory subunit, pu-

tative

Essential –

PBANKA_122740 protein phosphatase, putative Essential –

PBANKA_041600 RhopH3, putative Essential –

80 initial investigations: gene selection and P. BERGHEI screen

putative growth

id description phenotype rate

PBANKA_080450 rhoptry associated membrane antigen, puta-

tive (RAMA)

Essential –

PBANKA_093200 rhoptry neck protein 4, putative (RON4) Essential –

PBANKA_071310 rhoptry neck protein 5, putative (RON5) Essential –

PBANKA_030950 secreted protein with altered thrombospondin

repeat domain, putative (SPATR)

Essential –

PBANKA_080220 serine/threonine protein kinase, putative Essential –

PBANKA_090380 serine/threonine protein kinase, putative Essential –

PBANKA_031140 serine/threonine protein kinase, putative Essential –

PBANKA_010290 SPE2-interacting protein, putative (SIP2) Essential –

PBANKA_091170 subtilisin-like protease 2 (SUB2) Essential –

PBANKA_143360 thrombospondin-related apical membrane pro-

tein (TRAMP)

Essential –

PBANKA_052170 transcription factor with AP2 domain, putative

(ApiAP2)

Essential –

PBANKA_133270 Duffy-binding protein* Essential –

PBANKA_135740 conserved Plasmodium protein, u. f. Essential –

PBANKA_092090 conserved Plasmodium protein, u. f. Essential –

PBANKA_092560 conserved Plasmodium protein, u. f. Essential –

PBANKA_130430 conserved Plasmodium protein, u. f. Essential –

PBANKA_103970 conserved Plasmodium protein, u. f. Essential –

PBANKA_134490 conserved Plasmodium protein, u. f. Essential –

PBANKA_144290 conserved Plasmodium protein, u. f. Essential –

PBANKA_030450 conserved Plasmodium protein, u. f. Essential –

PBANKA_133460 diacylglycerol kinase, putative Essential –

PBANKA_092510 leucine-rich repeat protein (LRR11) Essential –

PBANKA_082490 patatin-like phospholipase, putative Essential –

PBANKA_083300 proﬁlin, putative (PFN) Essential –

PBANKA_061970 rhoptry-associated leucine zipper-like protein

1, putative (RALP1)

Essential –

3.4 barseq growth phenotypes 81

putative growth

id description phenotype rate

PBANKA_103210 rhoptry-associated protein 1, putative (RAP1) Essential –

PBANKA_030670 transporter, putative Essential –

PBANKA_093690 actin-like protein, putative (ALP1) Slow 0.68

PBANKA_091150 conserved Plasmodium membrane protein, u. f. Slow 0.82

PBANKA_136440 conserved Plasmodium protein, u. f. Slow 0.89

PBANKA_092170 conserved Plasmodium protein, u. f. Slow 0.78

PBANKA_113880 conserved Plasmodium protein, u. f. Slow 0.66

PBANKA_080130 conserved Plasmodium protein, u. f. Slow 0.80

PBANKA_120200 membrane skeletal protein, putative Slow 0.72

PBANKA_051140 peroxiredoxin, putative (nPrx) Slow 0.73

PBANKA_100240 protease, putative Slow 0.83

PBANKA_083290 Protein MAM3, putative Slow 0.67

PBANKA_110650 rhomboid protease, putative (ROM4) Slow 0.72

PBANKA_090590 transcription factor with AP2 domain(s) (AP2-

Slow 0.68

PBANKA_110140 RAP 2/3 Slow 0.76

PBANKA_110760 6-cysteine protein (P38) Redundant 1.07

PBANKA_100550 AAA family ATPase, putative Redundant 0.98

PBANKA_100360 apical sushi protein, putative (ASP) Redundant 0.88

PBANKA_130340 CCAAT-binding transcription factor, putative Redundant 0.99

PBANKA_090260 CCAAT-box DNA binding protein subunit B,

putative (NFYB)

Redundant 0.97

PBANKA_144720 CG2-related protein, putative Redundant 0.98

PBANKA_031480 conserved Plasmodium protein, u. f. Redundant 1.05

PBANKA_146060 conserved Plasmodium protein, u. f. Redundant 1.07

PBANKA_110690 conserved Plasmodium protein, u. f. Redundant 1.09

PBANKA_122540 conserved Plasmodium protein, u. f. Redundant 1.05

PBANKA_145910 conserved Plasmodium protein, u. f. Redundant 1.03

PBANKA_133660 conserved Plasmodium protein, u. f. Redundant 1.05

PBANKA_112010 conserved Plasmodium protein, u. f. Redundant 0.96

82 initial investigations: gene selection and P. BERGHEI screen

putative growth

id description phenotype rate

PBANKA_100250 conserved Plasmodium protein, u. f. Redundant 1.05

PBANKA_131420 conserved Plasmodium protein, u. f. Redundant 0.96

PBANKA_090610 conserved Plasmodium protein, u. f. Redundant 0.88

PBANKA_050720 conserved Plasmodium protein, u. f. Redundant 1.04

PBANKA_050690 conserved Plasmodium protein, u. f. Redundant 1.04

PBANKA_121060 conserved Plasmodium protein, u. f. Redundant 1.00

PBANKA_071660 conserved Plasmodium protein, u. f. Redundant 0.77

PBANKA_082160 conserved Plasmodium protein, u. f. Redundant 0.94

PBANKA_021450 conserved Plasmodium protein, u. f. Redundant 1.02

PBANKA_121440 conserved Plasmodium protein, u. f. Redundant 1.01

PBANKA_070200 conserved Plasmodium protein, u. f. Redundant 1.07

PBANKA_121180 conserved Plasmodium protein, u. f. Redundant 1.04

PBANKA_100530 conserved Plasmodium protein, u. f. Redundant 0.91

PBANKA_041780 conserved Plasmodium protein, u. f. Redundant 1.07

PBANKA_090840 conserved Plasmodium protein, u. f. Redundant 0.94

PBANKA_091240 conserved Plasmodium protein, u. f. Redundant 1.07

PBANKA_134640 conserved Plasmodium protein, u. f. Redundant 0.88

PBANKA_102760 conserved Plasmodium protein, u. f. Redundant 0.98

PBANKA_143910 conserved Plasmodium protein, u. f. Redundant 0.92

PBANKA_080250 conserved Plasmodium protein, u. f. Redundant 1.04

PBANKA_090860 conserved Plasmodium protein, u. f. Redundant 1.02

PBANKA_140920 conserved Plasmodium protein, u. f. Redundant 1.12

PBANKA_120180 conserved Plasmodium protein, u. f. Redundant 0.87

PBANKA_131280 conserved Plasmodium protein, u. f. Redundant 0.88

PBANKA_103170 conserved Plasmodium protein, u. f. Redundant 0.88

PBANKA_145870 conserved Plasmodium protein, u. f. Redundant 0.91

PBANKA_093210 DHHC-type zinc ﬁnger protein, putative Redundant 0.97

PBANKA_133580 disulﬁde-isomerase, putative Redundant 1.07

PBANKA_101910 DNA-directed DNA polymerase, putative Redundant 1.08

PBANKA_092970 haloacid dehalogenase-like hydrolase, putative Redundant 1.05

3.4 barseq growth phenotypes 83

putative growth

id description phenotype rate

PBANKA_123730 inner membrane complex protein, putative Redundant 0.97

PBANKA_062240 kinesin, putative Redundant 0.99

PBANKA_091510 LEM3/CDC50 family protein, putative Redundant 1.05

PBANKA_051520 MORN repeat-containing protein 1, putative

(MORN1)

Redundant 1.07

PBANKA_110330 myosin pfm-b, putative Redundant 1.00

PBANKA_051535 OTU-like cysteine protease, putative Redundant 1.06

PBANKA_112810 phospholipase (PL) Redundant 1.02

PBANKA_142310 phospholipase, putative Redundant 0.96

PBANKA_072070 regulator of chromosome condensation, puta-

tive

Redundant 1.01

PBANKA_011160 rhoptry protein, putative (ROP14) Redundant 1.05

PBANKA_093190 serine esterase, putative Redundant 1.01

PBANKA_130520 serine/threonine protein kinase, putative Redundant 1.06

PBANKA_112390 sphingomyelin synthase 1, putative (SMS1) Redundant 1.07

PBANKA_081700 sugar transporter, putative Redundant 0.90

PBANKA_120800 tubulin-tyrosine ligase, putative Redundant 1.04

PBANKA_145700 ubiquitin-conjugating enzyme, putative Redundant 1.10

PBANKA_071270 zinc ﬁnger protein, putative Redundant 1.06

PBANKA_031660 RBP, putative Redundant 1.05

PBANKA_100010 RBP, putative Redundant 1.04

3.4.1 Comparison with previous reverse genetic data

Data for a limited number of P. berghei knockouts was already avail-

able on the RMGMdb database. This allowed me to quickly establish

whether phenotypes of putative invasion-related genes were as one

would predict from this literature. In many cases the results of my

experiment agreed with previously published data.

For the 17 genes (of 25 for which there were previous reported P.

berghei transfection attempts) shown in Table 5 we detected pheno-

84 initial investigations: gene selection and P. BERGHEI screen

types identical to those previously described in literature, either by

ﬁnding the gene refractory to deletion, or to have a growth rate indis-

tinguishable from that of a redundant gene. For two further genes we

agreed on targetability but disagreed on whether growth was attenu-

ated, giving an agreement rate on targetability of 76%.

Table 5: Genes for which barseq phenotype agrees with previous data

In each redundant case, the growth rate is statistically indistinguish-

able from redundant controls. In each refractory case barcodes are

not detected in at least 2 timepoints for two out of three mice.

*: Attempt by Ghosh S; de Koning-Ward TF failed (unpublished

data, via RMGMdb)

gene id consensus ref.

DHHC9 PBANKA_093210 Redundant [65]

P38 PBANKA_110760 Redundant [198]

Pb235

PBANKA_031660 Redundant [64]

Phospholipase PBANKA_112810 Redundant [197]

GAMER PBANKA_122540 Redundant [3]

Pp7 PBANKA_1020100 Refractory [87]

Sub2 PBANKA_0911700 Refractory [197]

Pka PBANKA_083560 Refractory [189]

Guanylyl cyclase PBANKA_0910300 Refractory [91]

Ser/thr protein ki-

nase

PBANKA_0311400 Refractory [189]

RON4 PBANKA_093200 Refractory [76]

Proﬁlin PBANKA_083300 Refractory [113]

TRAMP PBANKA_143360 Refractory [191, 64]

RAP1 PBANKA_103210 Refractory [76], *

MSP-1 PBANKA_083100 Refractory [39]

GAPM1, PSOP3 PBANKA_1338900 Refractory [59]

Ser/thr protein ki-

nase

PBANKA_0903800 Refractory [189]

CNA PBANKA_1227400 Refractory [87]

There were also several conﬂicts with previous literature. Some of

these are potentially easily explained, others less so.

The serine/threonine protein kinase PBANKA_130520 was previ-

ously described as essential in P. berghei. [189] However, there was

some evidence for integration (although a clone could not be gener-

ated) and the gene has been successfully knocked out in P. falciparum

3.4 barseq growth phenotypes 85

[181]. It was also detected as potentially redundant by a previous

barseq experiment [80]. I detected it as likely redundant by barcode

sequencing and hypothesised based on the sum of the evidence that

a vector-speciﬁc problem may have resulted in the failure to disrupt

reported in [189]. To conﬁrm this result an individual transfection

was carried out with the PlasmoGEM vector targeting this gene and

I conﬁrmed by qPCR that the knock-out was successful (Fig. 26).

The dipeptidyl aminopeptidase DPAP3 plays a crucial role in asex-

ual growth in P. falciparum, where it is implicated as important to al-

low egress. [53] However the Leiden Malaria Group report (on RMG-

Mdb) that PBANKA_100240 could be knocked out with normal growth

detected from the increase in levels of haemozoin in the blood. We de-

tect slow growth, at a rate of 82%. This is a modest growth reduction

but is highly signiﬁcant (p “ 0.006). I hypothesise that either this re-

duction in growth was too slight to be detected in the previous study,

or potentially that it only appears in a competitive environment.

The ApiAP2 protein AP2-O (PBANKA_090590) is so named be-

cause its primary role is to regulate the ookinete stage of the parasite.

[210] It was therefore a surprise to detect slow growth (rate=0.68) of

this mutant. The canonical model of AP2-O is that it is transcribed in

gametocytes but repressed by DOZI until the ookinete stage, where it

activates a raft of genes important for the ookinete. So why a growth

phenotype in the blood stages? My ﬁrst explanation was that per-

haps a neighbouring gene important in the blood stages had a UTR

which extended into AP2-O and this was disrupted in the knock-out.

I therefore looked at strand-speciﬁc RNA-Seq from the P. falciparum

orthologue (Fig. 25). I found that this was not the case but I did ob-

serve that the gene is signiﬁcantly transcribed in the schizont stage.

We do know that the AP2-O protein is not visible in the blood stages

when tagged with GFP [210], but this does not rule out a low level of

protein being present, which could still play a signiﬁcant regulatory

role. The fact that expression peaks in schizonts correlates (perhaps

coincidentally) with the fact that both the merozoite and the ookinete

are invasive.

PBANKA_113780, a protein containing Armadillo domains rela-

tively uncharacterised in Plasmodium, was present in the input pool

at 2%, but below the detection threshold in all mice by day 6. This

would lead us to believe it to be likely essential. However it is re-

86 initial investigations: gene selection and P. BERGHEI screen

Schizont

Stage V

gametocyte

1 kb

AP2-O

Figure 25: The AP2-O locus in P. falciparum, shows schizont expression and is not

overlapped by any other transcript.

Strand-speciﬁc RNA-Seq data from [128], adapted from Plas-

moDB. The orientation has been reversed from genome orienta-

tion so that AP2-O is facing forwards. ’F’ indicates forward with

respect to AP2-O, ’R’ reverse. No genes have UTRs which overlap

AP2-O.

ported to have been knocked out previously in P. berghei on the basis

of pulsed-ﬁeld gel electrophoresis analysis. [105] This protein is the

orthologue of the TgGAC recently reported as essential in Toxoplasma,

which suggests it might also be essential in Plasmodium (the previous

positive PFGE result could be the result of a segmental duplication).

However more experiments will be needed to conﬁrm which experi-

ment’s results reﬂect reality.

Previously rhoptry-associated protein 2/3 (RAP2/3, PBANKA_110140)

could not be knocked out over three attempts in P. berghei [196]. This

was in a sense a surprising result since RAP1 can be deleted in P.

falciparum [13] resulting in a failure to localise RAP2, without lethal

effects upon the parasite (and RAP3 is redundant in P. falciparum).

We observed apparent slow-growing RAP2/3 knock out parasites in

the barcode-sequncing data. A possible explanation is the higher efﬁ-

ciency that PlasmoGEM vectors allow compared to conventional tech-

nologies. [154] I was able to generate parasites with RAP2/3 knock-

out by individual transfection. (Fig. 26)

Intriguingly RAP1 appeared essential in our P. berghei screen, and

was recently described as such with a single knock-out approach

[105]. This phenotype, combined with the fact that the total loss of

RAP2/3 is viable, suggests that RAP1 plays a role in P. berghei that is

more important than simply trafﬁcking RAP2/3 to the right cellular

location.

IMC1c (PBANKA_120200) could not be knocked out in a previous

P. berghei attempt, [195] although the researchers were able to tag it

3.4 barseq growth phenotypes 87

with GFP, leading them to hypothesise its essentiality. Our data sug-

gest it to be attenuated, with a growth rate 71% of the wild-type. The

authors of the past study reported that parasites came up with in-

sertions of the resistance marker into non-speciﬁc genomic locations.

Perhaps the inconsistency can be explained by these rare parasites

outgrowing the slow-growing knock out parasites, however a single

attempt to knock out the gene with a PlasmoGEM vector did not give

rise to a signiﬁcant knock out population leaving more investigation

needed.

A ﬁnal inconsistency was that we detected ROM4 (PBANKA_110650)

knock-out parasites as attenuated, but viable. This gene has been pre-

viously reported as essential in P. berghei. [124] In P. falciparum, ROM4

was shown to process EBA-175, shedding it into the supernatant as

invasion occurs. This step is thought to be crucial to invasion because

modiﬁcations to EBA-175 which remove the ROM4-processing site

prevent parasite growth, and the ROM4 gene itself has been previ-

ously reported as refractory to deletion in both P. falciparum and P.

berghei. [148] The barseq result was therefore surprising and so was

followed up with a single transfection of the pJAZZ vector into a sin-

gle mouse. When DNA from the parasites that came up from this

transfection was assessed by qPCR there was indeed a major reduc-

tion in the amount of locus present (Fig. 26). The parasites also came

up two days after those of the other vectors transfected in parallel, a

result consistent with attenuated growth.

If the processing that ROM4 performs is an essential process, one

possible explanation for the gene’s non-essentiality is possible shared

substrate speciﬁcity between ROM4 and the other rhomboid proteases

of the parasite.

3.4.1.1 Barseq data for invadome genes largely agrees with previous exper-

imental results

A summary of all genes with data both from RMGMdb and barseq is

shown in Fig. 27. For 72% of genes the barseq result entirely agrees

with that seen previously, and only in 8% does barseq reverse a previ-

ous result. In the remaining cases barseq suggests attenuated growth

where previous techniques suggested redundancy or essentiality; three

such results have been conﬁrmed in individual transfections.

88 initial investigations: gene selection and P. BERGHEI screen

P rim e r

Rela tive a b u n d a n ce b y q PC R

Se r/

hr k

OM4

tro

p 2/3

Se r/

hr k

OM4

tro

p 2/3

Se r/

hr k

OM4

tro

p 2/3

100

Ser/thr kinase KO

ROM4 KO

Rap 2/3 KO

Figure 26: Individual knock-outs of three genes conﬁrms barseq results which con-

tradict previous studies

These data demonstrate three conﬁrmed knock-outs by individ-

ual transfection of a pJAZZ vector. qPCR was conducted on para-

site DNA from uncloned populations. Arrows in each case iden-

tify the bar of interest. (The small apparent reduction in Ser/Thr

kinase level in the ROM4 KO is an artefact of the system, and a

low parasitaemia in this sample, rather than a real mistargeting

effect.)

’Ser-thr kinase’ indicates PBANKA_130520

These results gave me conﬁdence that barseq was proving a sensi-

tive method to detect growth phenotypes in invasion-related genes. I

therefore turned my attention to the majority of the dataset for which

there was not previous reverse genetic data in P. berghei.

3.4.2 Fitness results by organellar localisation

One way in which to break down the growth phenotypes we obtained

by barseq is to group them by their localisation within the parasite.

I combined ApiLoc and RMGMdb data to synthesise the localisation

data that existed for my phenotyped proteins.

3.4.2.1 Analysis of the rhoptry suggests newly essential components

The phenotypes I identiﬁed for mutants knocking out proteins pre-

viously shown to localise to the rhoptries is shown in Fig. 28. This

shows the conﬁrmed previous P. berghei refractory phenotypes for

RON4, RAP1 and TRAMP. RAP2/3 mutants appeared to produce at-

tenuated growth, in contrast to previous data (as discussed above).

3.4 barseq growth phenotypes 89

Prior (RMgmDB) Barseq

Possibly redundant

Attenuated growth

Possibly essential

Figure 27: Comparison of barseq phenotypes with previous RMGMdb phenotypes

This ﬁgure shows all genes analysed by barseq for which data

existed in RMGMdb prior to this experiment. Left hand side of

each gene represents status in RMGMdb, right hand side repre-

sents result by barseq.

Because of their known importance in invasion, the rhoptries have

been intensively studied in P. falciparum. I conﬁrmed that previous

failures to knock out RhopH3, RAMA and RALP1 in this species ex-

tended to P. berghei, giving further conﬁdence to the idea that these

are very likely essential.

I also found growth phenotypes for four genes which I believe have

not yet been characterised by reverse genetics in Plasmodium. RON5

and SPATR appeared refractory to deletion; conﬁrming that the latter

likely plays a role beyond that already established in the sporozoite.

[34] Unpublished data (referenced in [132]) suggests that in addition

to HepG2 cells, PfSPATR also binds erythrocytes and anopheline cells.

If these interactions are functional, they suggest a role for this protein

across the parasite life cycle. ROP14 and ASP appeared to be dispens-

able without any measurable effect on ﬁtness.

There are fewer non-redundant genes among those localised to the

rhoptry than in the invadome as a whole, however this is not a statis-

tically signiﬁcant result.

3.4.2.2 The inner membrane complex: ﬁtness for knock-outs of motor-related

proteins

My analysis included at least 12 inner membrane complex proteins,

very few of which had been previously examined in P. berghei (Fig. 29).

While previous analysis had occurred in Toxoplasma, it has been re-

cently discovered that major IMC components are entirely dispens-

able for invasion in this species, complicating the interpretation of its

phenotypes in a Plasmodium context. [61]

90 initial investigations: gene selection and P. BERGHEI screen

RhopH3

RAMA

RALP1

RON4

TRAMP

RAP1

RON5

SPATR

ROP14

ASP

RAP2/3

Prior (RMgmDB) Barseq

Possibly redundant

Attenuated growth

Possibly essential

Previously unreported

Figure 28: Barseq phenotypes of rhoptry-localised genes

This ﬁgure presents the information about rhoptry genes pro-

vided by this study. The left-hand side of each icon represents any

known phenotype before this study, the right-hand side the phe-

notype called by barseq. Three genes (RhopH3, RAMA, RALP1)

which previously were known to be refractory to deletion in P.

falciparum are now known to be likely essential in P. berghei also.

Two genes previously found to be refractory to deletion (RON4,

RAP1) are conﬁrmed as such. I newly suggest two genes (RON5,

SPATR) as likely essential and two (ROP14, ASP) as redundant.

Where previously attempts to delete RAP2/3 failed in P. berghei I

detect deletion as merely attenuating growth.

I have already discussed the contradictory phenotype I observed

for IMC1c as compared to previous studies, and listed the conﬁrmed

phenotypes, possibly essential and redundant respectively, for GAPM1

and DHHC9.

It was not surprising to identify GAPM3, GAP40, MyoA and MTIP

as likely essential. These are core components of the motor machinery

which drives the parasite into the erythrocyte. However in the light

of the recent discovery in Toxoplasma that some of these genes are

not absolutely required for invasion, a conditional approach might

be needed to check whether mutant parasites these are severely at-

tenuated or entirely non-invasive. [61]

Three of the phenotyped proteins were alveolins (IMC1c, IMC1f,

IMC1g). Two of these were targetable (IMC1c, IMC1f), but resulted

in attenuated growth. Given the structural similarity of the alveolins

one might speculate that there is a degree of functional redundancy

within these proteins. However IMC1g appears to be essential. Some

sources claim that PBANKA_123730 is also an alveolin [125] but very

little is known about this protein. I identiﬁed it as redundant in the

blood stages.

3.4 barseq growth phenotypes 91

GAPM3

GAP40

MyoA

GAPM1

IMC1g

MTIP

IMC1c

IMC1f

ALP1

DHHC9

MORN1

PBANKA_123730

Prior (RMgmDB) Barseq

Possibly redundant

Attenuated growth

Possibly essential

Previously unreported

Figure 29: Barseq phenotypes of IMC-localised genes

Five core glideosome components (GAPM1, GAPM3, GAP40,

MTIP, MyoA) were detected as likely essential, as was the alve-

olin IMC1g. IMC1c and IMC1f produced attenuation when singly

deleted, as did ALP1. MORN1, DHHC9 and PBANKA_123730

appeared redundant.

I identiﬁed MORN1 as redundant in the blood-stages. This is a po-

tentially surprising result given this protein is essential for correct

budding in Toxoplasma. [129] MORN1 is predicted by the Malaria Ad-

hesins and Adhesin-Like Protein Predictor (MAAP) [10] to be an adhesin

(or adhesin-like). (Score: 0.63). The predictor scores the Toxoplasma or-

thologue signﬁcantly lower (0.04), perhaps hinting that this protein

might have been recruited into a new role in Plasmodium. If so its

redundant phenotype would be consistent with that of most other

adhesins.

The measurement of the ALP1 knock-out as attenuated also contra-

dicted previous T. gondii data describing it as essential. [82]

While there are fewer non-redundant genes among the IMC-associated

genes than in the invadome as a whole, this is not a statistically sig-

niﬁcant result.

3.4.2.3 Other proteins of prior interest

There are a number of additional protein of interest for which we

obtained growth phenotypes, shown in Fig. 30.

MAM3 at ﬁrst appeared a strange inclusion in the invadome – its

name comes from its Saccharomyces cerevisiae homologue, which is a

mitochondrial organiser protein. However this protein actually has

similarity to haemolysins.

I therefore hypothesise that it may have

6 Information from S. cerevisiae genome description

92 initial investigations: gene selection and P. BERGHEI screen

Phil1

AP2

SIP2

Sub2

P38

MSP1

MyoB

Mlc-B

ROM4

AP2-O

Proﬁlin

Phospholipase (PL)

Phosopholipase

Peroxiredoxin (nPrx)

MAM3

PBANKA_

052170

PBANKA_

142310

MyoB

complex

Merozoite

surface

Proteases IMC

Phospho-

lipases

DNA

associated

Misc.

Prior (RMgmDB) Barseq

Possibly redundant

Attenuated growth

Possibly essential

Previously unreported

Figure 30: Barseq phenotypesof miscellaneous additional known genes

Miscellaneous proteins approximately grouped by shared fea-

tures areas of biology.

a role in egress, as has recently been suggested for haemolysin III.

[143] This would predict that the slow growth of the mutant is due to

a defect in egress.

It has very recently been found that a second myosin, MyoB is lo-

calised separately to MyoA and based on a failure to knock-out its

interacting light chain this gene was reported to be essential in the

blood stage. [211] Barseq conﬁrmed the essential nature of the rele-

vant Myosin light chain Mlc-B, but surprisingly identiﬁed MyoB as

redundant. However given this combination of phenotypes is incon-

sistent with the model reported in [211], the results should be treated

with skepticism until a cloned knockout of MyoB can be generated.

3.4.2.4 Barseq probes many previously uncharacterised proteins

The advantage of a large scale screening approach is that it assigns

growth phenotypes not only the proteins listed above, about which a

certain amount is known already, but also the bulk of proteins about

which we know very little.

These experiments have assigned growth phenotypes to 104 mem-

bers of the core invadome which were previously almost entirely un-

characterised experimentally. This dataset will be useful in two re-

7 One could speculate that the protein is not essential because its role can be partially

complemented by haemolysin III. It would be interesting to see whether the double

knock-out of haemolysin III and MAM3 is viable.

3.5 assessment of barseq phenotypes en-masse 93

Possibly

essential

Redundant

Attenuated

Figure 31: Proportion of various phenotypes in invadome barseq

spects; ﬁrstly as a source of data on individual genes – with this

dataset available those investigating a candidate gene may simply be

able to refer to this dataset rather than generating their own knock-

out. As further parts of the invasion machinery are identiﬁed and

conﬁrmed by new methods, such as BioID [166], I believe that many

of their mutant phenotypes will already lie in my data, allowing the

prioritisation of more important targets. Secondly I believe that anal-

ysis of this phenotyping together with the protein-protein interaction

network, as I will discuss in a later chapter, has the ability to reveal

new insights into parasite biology.

3.5 assessment of barseq phenotypes en-masse

As well as comparing phenotypes for individual mutants as above,

I also considered summary statistics for how many genes were tar-

getable, in comparison to previous screens. This might tell us both

something about the overall essentiality of genes that comprise the

invadome, and also about the types of data that barseq can supply.

3.5.0.1 Comparison to previous reverse genetic studies in P. berghei

I downloaded RMGMdb data for all 453 genes for which disruption

attempts had been reported in P. berghei. 59% of these attempts were

successful, as compared to 51% in my gene set.

Statistically this is a signiﬁcant difference. While it might repre-

sent increased essentiality in the invadome (which one would predict

given its enrichment for blood-stage genes), it could also be the result

94 initial investigations: gene selection and P. BERGHEI screen

of publication-bias in favour of successful knock-outs, i.e. redundant

genes.

The rate of targetable genes in this putative invadome happens to

be identical to the previous barseq study of the kinome (51%), [79]

and very similar to a conventional screen of the phosphatome. [87] It

is possible that the targetability rate for the genome as a whole lies

in the vicinity of this ﬁgure. If so this would be a ﬁgure much higher

than the 18.2% reported for S. cerevisiae [73] or the 16% reported for E.

coli [71]. It seems likely that when genes preferentially expressed in

non-blood stages are analysed the ﬁgure will decrease, but it seems

unlikely to reach the level seen in these other systems, perhaps re-

ﬂecting the complexity and number of vital events in even just the

intraerythrocytic cycle of parasite development.

3.5.0.2 Comparison to previous forward genetic study in P. falciparum

Perhaps the largest screen published to date in Plasmodium is the for-

ward genetic screen using piggyBac insertional mutagenesis in P. falci-

parum (Balu et al., [15]). In a forward genetic screen there are no nega-

tive results (until saturation is reached). This precludes a comparison

of targetability, but we can compare the different growth phenotypes

seen when parasites are successfully modﬁied.

Of those mutants we classiﬁed viable, just 19% showed an attenu-

ated growth phenotype which we could detect. This is signiﬁcantly

lower than that seen in the Balu et al. study in which around 50% of

clones analysed showed attenuation. I considered a number of possi-

ble reasons for this discrepancy.

• Explanation 1: Barseq does not have the statistical power to de-

tect moderate attenuation

– We can consider this possibility by ignoring the noise in

our data and assuming that are ﬁtness measurements are

accurate (because though noise may change the results for

individual genes, the overall distribution should remain

the same). If we use a simple cut-off as in the Balu et al.

analysis we obtain a similar result, demonstrating that the

t-test is not the cause of the inconsistency. The cut-off for

attenuation vs. redundancy in the Balu et al. analysis was

~80%. If we use such a cut-off rather than a statistical test

3.5 assessment of barseq phenotypes en-masse 95

we ﬁnd just 14% of our mutants to be attenuated—we are

not calling mutants Balu et al. found to be attenuated as

redundant.

• Explanation 2: The forward-genetic screen includes insertions

outside genic regions

– One could hypothesise these are more likely to produce

an intermediate attenuated phenotype, perhaps by simply

changing expression levels. Examining the Balu et. al data

reveals that the 33 insertions within CDSes (coding DNA

sequences) had a slightly lower attenuation rate (45%), but

this is still a signiﬁcant difference from the data in our

approach

• Explanation 3: Invasion genes have a more extreme ﬁtness dis-

tribution than the genome as a whole:

– This is entirely conceivable, but very difﬁcult to test with-

out a larger reference dataset. We do know that the conven-

tional kinase and phosphatase screens presented a similar

phenotype distribution to the invadome.

• Explanation 4: The distribution of phenotypes in P. berghei is

more extreme than in P. falciparum.

– This is very possible, especially considering the differences

between a culture system and an in vivo model in which

the presence of an immune system may be less forgiving to

attenuated parasites, but the hypothesis is very difﬁcult to

test without large comparable datasets for both organisms.

• Explanation 5: Barseq is identifying severely attenuated mu-

tants as ’essential’

– A ﬁfth of genes identiﬁed in the Balu et al. screen were iden-

tiﬁed as ’severely attenuated’, meaning they led to growth

rates below 60% to those of wild-type parasites. We did

not identify any with such a low growth rate. Indeed our

experimental design probably did not allow us to. We re-

quired that to be considered viable, parasites had to be

present at ą 0.1% in the population on day 7 in at least

8 Hypergeometic distribution p ă 0.0004

96 initial investigations: gene selection and P. BERGHEI screen

2 mice. If vectors were to be integrated at roughly equal

rates, any given mutant might start off at 1.6% of the vi-

able population. From this starting point the parasite must

grow at at least 63% of the rate of a wild-type parasite if it

is to remain >0.1% on day 7.

I believe that a combination of explanations 4 and 5 is the likely cause

of the differences in distribution we see. A previous study attempted

to knock out 8 P. berghei genes which had reduced growth described

by Balu et al. in P. falciparum. [123] Half were entirely refractory to

deletion in P. berghei and the remainder gave normal growth (were

redundant); so differences in the experimental systems no doubt play

a role. However it is also important to acknowledge the insensitivity

of our system to very slow growing mutants, and we should accept

that we may be calling some of these as ’likely essential’.

The line in the sand between ’attenuated’ and ’essential’ is always

an artefact of the experimental system used. Any change that reduces

the overall parasite growth rate, for example by varying media or

varying a mouse’s diet, will shift the position of this line and so a

lower resolution here is not deeply concerning. We could detect more

attenuated mutants by removing the requirement that we detect bar-

codes until day 7, but this would be at the expense of increased num-

bers of false-positives. The optimum point to place this division will

depend on the aims of any particular experiment.

3.5.1 The picture today

As a result of the work described in this chapter we have trebled the

number of core invadome genes for which we can suggest a pheno-

type from 52 to 170, out of 320. A comprehensive picture of what

we now know about invadome targetability in P. berghei is shown in

Fig. 32. This can be compared to Fig. 20 in order to visualise the

progress described in this chapter. The genes for which phenotypes

have been newly described include 68 genes with the description

"conserved Plasmodium protein, unknown function". This ﬁrst step to

assign 30 of these genes as likely essential prioritises them for further

study.

3.6 conclusions 97

Likely redundant Attenuated Likely essential

Not attempted

Prior (RMgmDB) Barseq

Figure 32: All putative growth phenotypes for core invadome genes in P. berghei,

including both barseq and previous work

Each icon represents a single gene. The left side represent previ-

ous reported RMGMdb targetability P. berghei [105], the right side

represents barseq phenotyping data.

3.6 conclusions

The PlasmoGEM-barseq methodology has proven powerful when ap-

plied to this set of putative invasion related genes. For the small pro-

portion of these genes which already had described growth pheno-

types, it has fundamentally agreed with previous data. But in a small

number of cases previously described growth phenotypes have been

challenged, and three of these results conﬁrmed by subsequent inde-

pendent transfections.

These data are a useful starting point for further studies of invasion,

and a resource for the community – one important piece of informa-

tion needed when looking at a new gene of interest is its knock-out

phenotype.

98 initial investigations: gene selection and P. BERGHEI screen

3.6.1 Possible follow-up work

We did not obtain any ‘severely-attenuated’ mutants with very slow

growth rates, in contrast to Balu et al. This might be in part due to

such mutants being outcompeted early in the infection, but might

also be a real result of an in vivo system. This results could be further

investigated by barseq conducting a transfection of only vectors tar-

geting ’likely essential’ genes and observing whether parasites came

up, and which barcodes they carried – any newly identiﬁed non-

essential genes would have to be conﬁrmed by independent trans-

fections to exclude the possibility of freak events which incorporate a

barcode without knocking out the target gene [79].

The PlasmoGEM project is forever moving forwards towards com-

plete coverage of the P. berghei genome. There are now a number of

additional invadome vectors available which will in time be analysed

by barseq to begin to ﬁll in the parts of Fig. 32 which for now remain

grey.

Finally there are some limitations to using an in vivo system for

studying invasion by barcode sequencing. The development of an in

vitro approach might unlock new avenues for the application of Plas-

moGEM technology to invasion, as I will discuss in the next chapter.

T H E T R A N S F E R O F P L A S M O G E M T E C H N O L O G Y

TO P. K N O W L E S I

4.1 introduction

The previous chapter described the use of pooled-transfection bar-

code sequencing to assign growth phenotypes to mutants in a large

number of invasion related genes. The scale of these experiments

was made possible by the high efﬁciency of transfection possible in

P. berghei. However we also observed some of the limitations which

come with working in a rodent model.

The assay allowed me to measure the blood-stage ﬁtness of para-

sites lacking each of these genes, and based on the PlasmoINT net-

work and their IDC placement we believe these genes to be involved

in erythrocyte invasion. But the barseq experiments did not provide

any direct evidence that the genes were involved in invasion rather

than, say, metabolism. Nor was I able to identify what stage, if any, of

that process they were involved in.

Many of P. berghei’s strengths as a model come from it being an in

vivo system, but there are also concomitant disadvantages. Because

of sequestration, P. berghei schizonts are not available without in vitro

schizont culture, and these schizonts do not rupture in vitro, render-

ing invasion relatively inaccessible. This clearly limits the number of

invasion-related experiments possible in P. berghei today.

In a model in which invasion was directly accessible, barseq might

be able to be applied directly to the invasion process. A pool of bar-

coded mutants could be used to produce schizonts, and the relative

ratio of different mutants assessed by barseq. The schizonts could be

allowed to rupture and the resulting merozoites puriﬁed and anal-

ysed by barseq. A decrease in barcode abundance as compared to the

schizonts would indicate an egress defect. One could progress further

– merozoites could be allowed to invade erythrocytes, with barcodes

ampliﬁed from the resulting rings. Differences in abundances would

reﬂect differences in the ability to invade.

100 the transfer of plasmogem technology to P. KN OWLESI

Additionally, the barcoded invasion assay could take place into ery-

throcytes treated with enzymes or antibodies to deplete certain recep-

tors, and thus genes within the parasite linked to speciﬁc erythrocyte

components.

These approaches would probably be most powerful when applied

to attenuated mutants – it would not be possible to establish the func-

tion of an essential gene – and this brings us to another limitation

of the approach described in the previous chapter. As we saw, atten-

uated mutants rapidly decrease in the population and so the pooled

transfection approach probably excludes analysis of a sizeable propor-

tion of severely attenuated mutants (which could be ~20% of genes).

An in vitro system might mitigate this by allowing individual trans-

fections to be carried out for each gene without the use of large num-

bers of animals, allowing pooling just before an assay and so avoiding

the problem of attenuated mutants being outcompeted prior to anal-

ysis.

In short, there is much potential from establishing PlasmoGEM

technology in an in vitro system. PlasmoGEM uses linear transfec-

tion vectors, which have the advantage of minimising episomes, but

until the recent development of Crispr/Cas9 approaches linear DNA

had not been used successfully in the primary in vitro system, P. fal-

ciparum. In P. knowlesi, however, linear transfection has been reported

for a long time [109] and the parasite has recently been adapted for

growth in vitro in human red blood cells with high transfection efﬁ-

ciencies reported. This seemed an ideal system, therefore, to attempt

to bring an in vitro approach to PlasmoGEM technology.

4.2 system development

4.2.1 Optimising culture conditions

The P. knowlesi A1-H.1 clone (a descendant of the H strain) was a gift

from Robert Moon at the National Institute for Medical Research and

he provided invaluable advice in the initial stages of my project. My

early work was characterised by difﬁculties in parasite culture. Over

time I optimised a number of aspects, leading to robust growth in my

hands.

4.2 system development 101

25%

50%

75%

100%

No additional

serum

0.5% additional

Albumax

10% Addenbrookes

serum

10% Horse serum

10% Interstate Blood

Bank serum

Additional serum

Relative growth

Figure 33: The effect of different sera on P. knowlesi growth over 48 hours.

Parasitaemias were assessed with SYBR-green ﬂow cytometry. Er-

ror bars reﬂect 95 % conﬁdence-intervals. In each case the basal

media contains 0.5% Albumax.

I set up a number of assays to establish optimised conditions for

P. knowlesi culture. In each case a 0.5% parasitaemia culture of par-

asites was set up with triplicates for each condition. Parasites were

incubated for 48 hours, corresponding to approximately two intraery-

throcytic cycles.

After this time parasites were stained with SYBR green and the

ﬁnal parasitaemia assayed with ﬂow cytometry as described in Sec-

tion 2.2.8.

4.2.1.1 Horse serum permits stable growth of P. knowlesi

P. knowlesi is more demanding of serum than P. falciparum (where

10% serum or 0.5% Albumax are routinely used). The species has in

the past been grown in 20% Rhesus macaque serum, and was then

adapted to growth in 20% human serum. [109] During its adaptation

to human erythrocytes the line was grown in 0.5% Albumax II (w/v)

(which approximately corresponds to an initial 10% serum (v/v) ),

combined with an additional 10% human serum.

102 the transfer of plasmogem technology to P. KN OWLESI

My attempts to cultivate parasites in serum from haemochromato-

sis patients, which is used for routine P. falciparum culture in our

lab, were not successful—parasites grew slowly and often crashed. I

ordered commercial serum from Interstate Blood Bank

, which was

even less successful with parasitaemias crashing immediately when

parasites were transferred into it.

Ultimately I learnt that horse serum had been used quite success-

fully for P. knowlesi culture (R. Moon, personal communication). Al-

though I tried a number of strategies, including screening a num-

ber of different patient sera for growth, in my hands heat-inactivated

horse serum (Life Technologies) has been the most consistently succ-

cessful in culture. The results of a 48-hour growth assay conﬁrming

these results is shown in Fig. 33.

It is worth noting that this assay may understate the effect of serum,

since its effects may take days to become apparent. Equally however,

a caveat is that the parasites used in this experiment had been grown

in horse serum for a number of months before this assay and may

have adapted to it.

Others have achieved culture in human serum that is as good or

better than that in horse serum, but this typically involves screening

many batches whereas I have yet to ﬁnd a batch of horse serum which

is not effective. That equine serum can replace human serum was also

observed for P. falciparum back in 1979. [24]

Given these results, subsequent work described here was carried

out with parasites grown in 0.5% Albumax and an additional 10%

horse serum.

4.2.1.2 Malaria gas is necessary for optimal P. knowlesi culture

‘Physioxia’, the oxygen level in human tissue, is typically 1%-11%.

Given this, it is remarkable that most tissue culture of human cells

is successful at oxygen levels of 19.95%. [28] However Plasmodium

appears to be more sensitive to O

concentration, and the use of a

candle-jar was important for its initial adaptation to in vitro culture

[194]. Today malaria gas, a mixture of 93% N

, 4% CO

and 3% O

is more typically used. I sought to assess the importance of gassing

1 http://www.interstatebloodbank.com/

4.2 system development 103

25%

50%

75%

100%

Air CO

Malaria Gas

Atmosphere

Relative growth

Figure 34: The effect of atmosphere on P. knowlesi growth over 48 hours.

Error bars reﬂect 95% conﬁdence-intervals. In each case, media

contained 0.24% (w/v) sodium bicarbonate. (*: p ă 0.05, **: p ă

0.01)

cultures, and whether a CO

incubator could replace the need for

this.

In a 48-hour growth experiment, the results shown in Fig. 34, I

found that (at least in the absence of time to adapt) parasites grew

signiﬁcantly more slowly in a CO

incubator as compared to sealed

ﬂasks gassed with malaria gas. Nevertheless there was robust growth

in the CO

incubator.

Both these conditions were markedly better than air, though this

will have been in part an effect of the sodium bicarbonate in the

medium, which is intended to buffer 5% CO

In the light of this data I used a CO

incubator as a fall-back when

gas was not available and for very short periods of culture where

gassing was impractical. I have since maintained parasite culture for

more than a week in a CO

incubator when it was necessary to work

in a lab without malaria gas, showing that a low oxygen environment

is not essential.

4.2.1.3 Static culture appears superior to suspension culture for P. knowlesi

For P. falciparum, suspension culture achieved by gentle shaking is as-

sociated with faster parasite growth, lower multiplicity of infection

(MOI) and greater synchrony. [6] Especially signiﬁcantly, this is re-

ported to markedly reduce the amount of time taken for modiﬁed

parasites to grow up after transfection. [6]

I hypothesised that shaking would also improve P. knowlesi culture.

But a 48-hour growth experiment (Fig. 35) actually demonstrated

104 the transfer of plasmogem technology to P. KN OWLESI

***

25%

50%

75%

100%

Shaken Static

Type of culture

Relative growth

Figure 35: The effect of shaking vs. static culture on P. knowlesi growth over 48

hours.

Error bars reﬂect 95% conﬁdence-intervals. ( ***: p ă 0.001)

static culture to be superior in terms of parasite growth — with the

difference sufﬁcient to overcome any improvement in MOI or syn-

chrony, even if this exists for P. knowlesi. This assay was repeated on

three separate occasions with results on each occasion showing static

culture to be superior.

We can speculate on the reason for such a result. The parasite was

adapted in static culture and so is likely to have optimised itself for

these conditions. However even P. falciparum lines which have long

been cultured in static conditions seem to beneﬁt from shaking.

In static culture, the cells settle form a layer of compacted erythro-

cytes, both infected and uninfected. Shaking maintains blood and

medium in a churning suspension. Which of these is closer to the

in vivo case?

Qualitatively, the shaken medium appears much closer—blood is

itself a suspension of cells in serum. But the ratio of cells to ﬂuid

in culture is very different in vitro and in vivo and this may have

important consequences.

Blood has a haematocrit of ~50%. In culture, without kidneys to

remove waste and a constant resupply of nutrients, we must lower

the haematocrit to ~2%. In suspension culture this means an infected

erythrocyte is likely on average to be 10.2 µm

from its nearest neigh-

bouring erythrocyte, as compared to 4.7 µm in real blood. Perhaps

this sparsity prevents as many merozoites contacting erythrocytes in

shaken cultures as compared to real blood.

2 r “ p

4πn

4.2 system development 105

But when erythrocytes settle in static culture they achieve a lo-

cal haematocrit of 100%, with each bursting schizont intimately sur-

rounded by erythrocytes which can be invaded. In this sense the static

culture is closer to the in vivo situation, and perhaps this explains P.

knowlesi’s preference for it.

Why the difference between P. falciparum and P. knowlesi then? Here

I enter the realm of speculation, but one possible explanation is their

different number of merozoites per schizont. If we model packed red

blood cells as packed spheres, each makes contact with 12 neighbours

around it in an optimal packing conﬁguration.

That number, 12, is roughly equal to the number of merozoites in a

mature P. knowlesi schizont. One could imagine this being an optimal

ﬁgure if blood cells are packed – ’one merozoite per erythrocte’. By

contrast P. falciparum schizonts can release up to 32 merozoites. That

could make multiple infections much more likely in a P. falciparum

static culture. The argument can also be made in reverse: perhaps the

larger number of merozoites makes up for the decreased likelihood

of each merozoite making contact with an erythrocyte in P. falciparum

shaken culture.

However many other possible explanations are possible, perhaps

involving a greater sensitivity to mechanical stresses in P. knowlesi.

Regardless, in light of this data I did not shake parasites for routine

culture, nor when bringing up parasites from a transfection.

Despite this result shaken cultures were helpful over short time

periods. My tests showed greater invasion to occur from puriﬁed sch-

izonts over 90 minutes with shaking than without (data not shown).

One explanation is that once placed on an incubator shelf large static

cultures do not actually form a full settled erythrocyte layer for hours,

and schizonts are likely to be the last to settle given their lower den-

sity. In these situations, where a suspension culture is inevitable in the

ﬁrst hour regardless, shaking is likely to accelerate the rate at which

merozoites encounter erythocytes. Shaking also causes the culture to

reach thermal equilibrium with the incubator more quickly which is

important for late stage cultures.

Given these results I used shaking during the tight synchronisation

procedure, and immediately post-transfection. In these short-term sit-

uations haematocrits above 2% were used.

106 the transfer of plasmogem technology to P. KN OWLESI

4.2.2 P. knowlesi permits a high efﬁciency of transfection

For a rapid, quantiﬁable detection of the efﬁciency with which DNA

was taken up by P. knowlesi I employed a luciferase assay.

The plasmid pHLH-1 [49], which expresses ﬁreﬂy luciferase from

a P. falciparum hrp3 promoter, was obtained from MR4. DNA (40 µg)

was transfected into tightly synchronised mature P. knowlesi schizonts

using the Amaxa Nucleofector system as described in [142]. Schizonts

were resuspended in Lonza Nucleofector P3 solution pulsed with

programme FP-158 and added to fresh warm blood and media. Lu-

ciferase activity was assessed 18 hours later, as described in 2.2.9.

The results (Fig. 36) demonstrated a substantially higher transfec-

tion efﬁency than P. falciparum and as a side-effect also show the P.

falciparum hrp3 promoter to be functional in P. knowlesi.

owle

iparum

100

1000

10000

100000

C u ltu re

C ounts per 10 second s

Figure 36: Luciferase assays demonstrate the high transfection efﬁciency of P.

knowlesi.

The P. falciparum result shown here is a typical result obtained

by others in the lab (courtesy of Zenon Zenonos) rather than the

result of my own experiment. Control represents an extract from

untransfected parasites, i.e. background instrument detection. Er-

ror bars represent 95% conﬁdence intervals in luciferase detection

and conﬁdence intervals for transfection itself might be larger.

4.2 system development 107

4.2.2.1 Luminescence does not increase linearly with DNA concentration

Plasmodium transfections typically involve quantities of DNA orders

of magnitude above the manufacturer’s directions for Lonza Nucleo-

fector kits. I sought to assess whether this was detrimental to transfec-

tion. To test this, I transfected the same luciferase construct into the

same batch of parasites, but at two different concentrations tenfold

apart. I found the larger amount of DNA to yield the most lumines-

cence but the magnitude of the increase was not proportional to the

increase in concentration. This suggests that saturation may be being

approached at 0.8 µg but that it has not been reached.

8μg

500

1000

1500

2000

A m o u n t o f D N A

R e la tiv e lu m in e sc e n c e

Figure 37: Effect of DNA quantity on luminescence intensity in a luciferase assay.

Increasing the quantity of DNA transfected in a P. knowlesi trans-

fection from 0.8µ g to 8µg

4.2.2.2 Transfectant parasites are visible one day after P. knowlesi transfec-

tion

To complement the data from luciferase assays I also conducted a sch-

izont transfection with the episomal GFP plasmid described in [142].

16 hours post transfection parasites expressing cytosolic GFP could

be seen in the culture (Fig. 38). I was not able to achieve transfected

parasitaemias as high as the ~0.5% seen in [142] but was able to ob-

tain ﬂuorescent parasitaemias greater than 0.01%, much higher than

for P. falciparum.

108 the transfer of plasmogem technology to P. KN OWLESI

BF GFP

Hoechst Merge

Figure 38: One of many parasites with cytosolic GFP seen one day post-

transfection.

4.2.2.3 Nucleofector solution affects parasite viability

I tested two different nucleofector solutions: nucleofector P1 and nu-

cleofector P3. I found vast decreases in viability using nucleofector P1

with very few parasites surviving transfection even prior to selection.

4.2.3 Mechanics of schizont transfection

There is a lack of information in the literature about what physically

occurs during a schizont transfection, but it seems important that

schizonts, and the merozoites they contain are at a very late, mature

stage for them to survive the process.

One potential explanation for this low viability is that electropora-

tion might lyse one or both of the parasite vacuole membrane (PVM)

and the erythrocyte plasma membrane, and that any schizonts that

do not already contain merozoites sufﬁciently mature for invasion

simply die.

However, this presents something of a puzzle as during normal de-

velopment secretion of the exonemes both begins the egress process,

and is also important for processing merozoite surface proteins to

prime them for invasion.

In order to better understand what actually happens in the cuvette

of schizonts during a schizont transfection, I prepared smears from

4.2 system development 109

cultures immediately before and immediately after transfection of

schizonts in solution nucleofector P3 using programme FP-158 in the

Lonza nucleofector.

(a) Schizont puriﬁcation immediately

prior to transfection

(b) Culture immediately after transfec-

tion, with apparent lysis of the par-

asitophorous vacuole, and in some

cases the red blood cell.

Figure 39: Giemsa smears before and after electroporation with Lonza nu-

cleofector FP-158. Representative cells are enlarged.

Before transfection in typical P. knowlesi fashion the merozoites are

tightly clustered, radiating from the food vacuole. Afterwards most

red blood cell membranes are intact, although understanding the pro-

portions is hampered by the fact that centrifugation may have prefer-

entially pelleted intact cells. However in contrast to the tightly organ-

ised appearance before electroporation, merozoites are distributed

throughout the cell. The appearance suggests to me that electropo-

ration causes the PVM to break down.

There are two ways in which one might imagine this happening:

• The electric shock could directly cause disruption of the PVM

• The pores created by the electric pulse might allow Ca

enter the parasite cytosol from intracellular stores (such as the

endoplasmic reticulum), inducing egress in much the same way

as calcium ionophore. [77]

In my hands, post-transfection parasitaemias achieved are at best

10% of those expected if the same number of schizonts was added

to fresh blood without electroporation. Understanding the source of

this attrition is likely to be an important component of increasing

P. knowlesi transfection efﬁciency in future. If the PVM breaks down

before the egress cascade is properly initiated this might result in the

release of merozoites which are non-invasive.

110 the transfer of plasmogem technology to P. KN OWLESI

I was not able to proceed further with this analysis but I think that a

full understanding of the mechanics of transfection may be helpful for

further increases in transfection efﬁciency; this might be achieved by

using a GFP marker to identify the PVM with analysis by microscopy

immediately post transfection.

If it were found that no ordered egress occurs after a schizont trans-

fection there might be value in using a transfection buffer which in-

duces lysis of the erythrocyte membrane and PVM in order to leave

fewer membranes between the exogenous DNA and the parasite nu-

cleus.

4.2.4 Characterisation of the response of P. knowlesi to selective agents

One pre-requisite for a genetic system system is the availability of

drugs allowing those parasites which have taken up exogenous DNA

to be selected from the background population of wild-type para-

sites. Previously WR99210, G418 and blasticidin have been used for

selection in P. knowlesi, in the context of in vitro culture in macaque

cells. [199] WR99210 was also used in selection in the A1.H1 clone

in human cells. [142] Because drug-sensitivity is often strain speciﬁc,

I decided to determine kill curves for 3 drugs in the A1.H1 clone:

pyrimethamine, G418 and DSM1. Pyrimethamine was selected be-

cause of unpublished reports of a large difference in IC

between

wild-type and transformed parasites, G418 because it has the highest

increase in IC

published in P. knowlesi [199] and DSM1 because it

was already in use in a number of P. falciparum constructs in the lab

and its use would simplify the adaptation of these constructs to P.

knowlesi.

Dilution series for the drugs were set up in 96-well plates in com-

plete media, seeded with parasites and fresh erythrocytes and incu-

bated for 48 hours. Final parasitaemias were assessed by ﬂow cytom-

etry. The results are shown in Fig. 40. I used a 4-parameter logistic

regression to ﬁt a sigmoidal curve to the data, and interpolated IC

values which are shown in Table 6.

4.2 system development 111

drug target ic

95% CI

Pyrimethamine Dihydrofolate reduc-

tase

3 nM 2.8 – 4.2

G418 Protein synthesis 377 µM 248 – 573

DSM1 Dihydroorate dehydro-

genase

1.5 µM 0.57 – 4.24

Table 6: Calculated IC

s in P. knowlesi for pyrimethamine, G418 and DSM1

drugs

s and their conﬁdence intervals, as calculated from the growth

inhibition curves shown in Fig. 40.

4.2.4.1 Pyrimethamine

I calculated an IC

of 3 nM for pyrimethamine, considerably lower

than that in P. falciparum (~40 nM). [161] The IC

for the activity

of the PfDHFR enzyme is reported at 42 nM, [145] whereas that for

PkDHFR was reported as 1 nM ([25]), a ratio consistent with my re-

sults. I subsequently identiﬁed early P. knowlesi literature from 1971

suggesting an IC

close to 1 nM. [86] This increased sensitivity has

important implications for the selective pressure to be used.

4.2.4.2 G418

The IC

I calculated for G418 (geneticin) was 377 µM, within the

range seen for sensitive P. falciparum isolates (3D7 is one of the most

sensitive with an IC

of 300 uM). [135] An IC

of 600 mM (three

orders of magnitude higher) was reported in P. knowlesi, [203] but

from the other ﬁgures described in this publication I believe this was

a copy-editing error and that the authors intended 600 µM. In sum,

it appears that use of the neoR marker might be effective in the P.

knowlesi A1.H1 clone.

4.2.4.3 DSM1

The most surprising result was for DSM1. This drug has an IC

56 nM in P. falciparum. I detected P. knowlesi as 27ˆ less sensitive to

it. Given the large difference seen, and the relative large conﬁdence

interval, I cultured parasites at 500 nM and 10 µM and after 72 hours

observed complete inhibition in the case of 10 µM and normal growth

for 500 nM. I then checked the drug was still functional by observing

112 the transfer of plasmogem technology to P. KN OWLESI

that it killed P. falciparum at 500 nM. This degree of insensitivity (in

P. falciparum the selective marker yDHODH provides an IC

of only

4.6 µM) may rule out the use of this drug as a selectable marker in P.

knowlesi.

I performed a Clustal alignment of the P. falciparum 3D7 and P.

knowlesi H strain sequences, and found that, while the P. knowlesi

genes is well conserved with P. falciparum, there are a number of dif-

ferences, including a C175L substitution. This Cys

175

is within 4Åof

the site of DSM1 binding, [50] and so might well play a role in the

>20-fold reduced sensitivity to DSM1.

These experiments are, to my knowledge, the ﬁrst use of DSM1 in

P. knowlesi and though they reveal it to be effective at killing wild-type

parasites at high concentrations, they suggest a decreased sensitivity

as compared to P. falciparum which may limit the drug’s usefulness

for genetic studies.

4.2.5 Genomic analysis

In the course of this project I performed whole-genome Illumina se-

quencing the genomes of various parasite populations derived from

the H1.A1 clone. I will discuss the rationale, and results, of this se-

quencing in a locus speciﬁc manner later. But I will here provide

some overall observations on the genome of my isolate of the A1.H1

clone (which has undergone more than 6 months of culture in my

hands, and so may not reﬂect that published in [142]).

I am comparing to the reference H-strain, [151] which since it was

isolated had been maintained exclusively in a cynomologous context.

The A1.H1 line does not produce gametocytes. In lab-adapted P.

berghei and P. falciparum and strains which lose the ability to produce

gametocytes, this is often the result of a mutation in AP2-G [179]. I

therefore looked for such a mutation but none was apparent. How-

ever a scan of other ApiAP2 genes did reveal a SNP which changed

a Cys to a premature stop codon in PKNH_1441500. This would be

expected to prevent the translation of the ﬁnal 3.5 kb of this 8.6 kb

gene. Since ApiAP2 genes appear to be some of the master regulators

of Plasmodium transcription this may well be a functional change –

but whether it affects gametocytogenesis remains to be proven.

4.2 system development 113

C o ncentration (μ M)

Fin a l p a ra sita e m ia (%)

Pyrimethamine

G418

(a) Growth inhibition curve for wild-type P. knowlesi parasites

placed under increasing concentrations of pyrimethamine and

G418

D sm 1 co ncentration

(μM)

Fin a l p a ra sita e m ia (%)

-1

Control

(b) Growth inhibition curve for wild-type P. knowlesi parasites

placed under increasing concentrations of DSM1

Figure 40: P. knowlesi is sensitive to DSM1, pyrimethamine and G418, but with

values spanning four orders of magnitude.

Parasitaemias were calculated by SYBR-green ﬂow-cytometry

from parasites grown in serial dilutions of each drug.

114 the transfer of plasmogem technology to P. KN OWLESI

Cys175

DSM1

Figure 41: Crystal structure of P. falciparum DHODH bound by DSM1, highlight-

ing a residue that differs in P. knowlesi

I downloaded the crystal structure from PDB ID 3I65 [50], and vi-

sualised it using Cn3D [202]. I have highlighted in yellow Cys

175

which is in close contact with DSM1 (blue). In P. knowlesi this

residue is substituted by a hydrophobic leucine which may play

a role in increasing the IC

for DSM1 in this species.

In addition, I noted the presence of a 450 kb segmental duplication

in chromosome 14 (Fig. 43). This contains 99 genes, including the

invasion-related gene TRAMP. Recently reports in another human-

adapted P. knowlesi line have identiﬁed duplications of invasion lig-

ands as a potential mechanism of adaptation, [121] meaning that the

99 genes in this region might well be of some interest.

It is important to bear this duplication in mind when designing

gene targeting experiments since replacement of any gene within it

with a selectable marker will still leave one copy of the wild-type

gene.

4.3 preparation of plasmogem vectors

4.3.1 Cloning of Gateway cassette

The protocol for generating PlasmoGEM targeting constructs involves

replacing the region to be targeted with a Zeo/PheS negative selec-

tion cassette using recombineering and then a ﬁnal Gateway reaction

to change this bacterial selection cassette for one which is functional

in Plasmodium. (Fig. 13) My ﬁrst action in adapting the system was to

construct a Gateway cassette that would function in P. knowlesi.

4.3 preparation of plasmogem vectors 115

Figure 42: Premature stop codon in an ApiAP2 gene in A1.H1 isolates

I ampliﬁed the region upstream of the P. knowlesi hsp70 gene adding

overhangs to insert it after the Pb 3’UTR of the Pb3HA-dhfr-yfcu vec-

tor and before the hDHFR/yfcu fusion gene. The hDHFR/yfcu gene

was itself ampliﬁed, as was the 3’ UTR of PkHsp70. These three frag-

ments were assembled in a Gibson Assembly reaction and cloned into

the vector cut at the PstI and KpnI sites. The assembled product was

transformed into PIR2 cells which are permissive for the R6Kγ origin

of replication.

4.3.2 Assessing the pilot pJAZZ library

As I began this project a small library had been produced containing

sections of P. knowlesi genomic DNA in the pJAZZ backbone [78].

This was carried out by Burcu Anar, and clones were mapped to the

P. knowlesi genome by Frank Schwach.

I examined the location of clones on the genome and found that

they covered 876 genes. When I added a requirement that the clone

contain at least 1000 bp upstream of the gene, to act as a homology

arm, this was reduced to 687 genes, just over 12% of the genome. I

performed a simulation to examine how coverage would be increased

by library expansion, shown in Fig. 44. This shows that should exper-

imental efforts be successful with the limited P. knowlesi library, it

should be possible to expand it to be able to target the majority of the

genome with ~10,000 mapped clones – but that 100% coverage will

never be cost-effective.

116 the transfer of plasmogem technology to P. KN OWLESI

Coverage

Genomic coordinates

Figure 43: Segmental duplication in chromosome 14

This segmental duplication is present in all of my isolates of the

H1.A1 clone, with some implications for gene targeting in this

region.

Limited to this pilot library, I identiﬁed those genes covered within

it which were part of the core invadome, as deﬁned in Chapter 3. The

PlasmoGEM team attempted to manufacture 24 knock-out constructs

and 24 tagging constructs for these invasion genes. There is a degree

of attrition in the production process and this ﬁrst run had particular

problems with carryover of wild-type clones. Therefore only 14 clones

passed the ﬁnal QC process. Additionally these clones were produced

using an earlier version of the GW cassette designed to use G418 as a

resistance marker

I took intermediate vectors made by the recombineering team which

had given rise to successfully screened vectors, and used a Gateway

reaction to insert the newly constructed P. knowlesi dhfr/yfcu Gate-

way cassette. I screened the reactions and identiﬁed 9 correct target-

ing vectors, shown in Table 7.

4.4 plasmodium knowlesi gene targeting

All of these constructs were used in pilot P. knowlesi gene targeting

experiments. Whether a construct was for tagging or knock-out was

primarily the result of attrition during the production process rather

than a planned selection.

3 I cloned this cassette and attempted transfections which were not successful, but the

cause of these failures could not be concluded and so those experiments will not be

described in detail

4.4 plasmodium knowlesi gene targeting 117

500

1000

1500

2000

2500 5000 7500

Insert size / bp

Simulated distribution

(a) Hypothetical size distribution for frag-

ments sheared to ’5kb’, assuming a nor-

mal distribution and the ranges seen in

the manufacturer’s shearing protocol.

The shearing may not in fact be nor-

mal, with positive skew likely, but the

decreased transformation efﬁciency of

large-inserts is likely to counteract this,

and so I approximate to a normal dis-

tribution.

(b) As the number of simulated clones in-

creases, coverage increases but the rate

of increase steadily drops (due to mul-

tiple coverage). The red line shows the

amount of the genome expected to be

covered, whilst the blue line predicts

the proportion for which a target cas-

sette will be able to be produced.

25%

50%

75%

100%

5000 10000 15000 20000

Mapped clones

Predicted coverage

Basepair covered

Covered with 1kb flanking region

Figure 44: Possibilities for library expansion

4.4.1 Genes targeted for HA-tagging

Three proteins were targeted for HA-tagging:

• apical sushi protein was initially described as localising to

the micronemes but is now known to reside in the rhoptry-neck.

[183] Its function is not known.

• rop14 is a rhoptry-localised protein. Its exact function is not

known but it appears to be one of the last proteins to leave from

the rhoptries, its secretion probably taking place after invasion.

[214]

• ribonuclease h2 subunit a expression peaks at 46 hours

post-invasion in P. falciparum. Although it forms part of the Hu

et al. invadome it seems likely to be a false positive, since in

other systems it is suggested that it may be involved in DNA

replication.

4.4.2 Genes targeted for knock-out

Six constructs were successfully produced for knocking out P. knowlesi

genes. I will brieﬂy review the proposed functions of the genes these

118 the transfer of plasmogem technology to P. KN OWLESI

gene id description construct

ha tagging constructs

PKNH_0304000 Apical sushi protein TGL58

PKNH_1126300 Ribonuclease H2 subunit A TGL60

PKNH_1136700 ROP14 TGL61

knock-out constructs

PKNH_0406200 Ser/Thr protein kinase TGL59

PKNH_0918700 DHHC3 TGL62

PKNH_0941100 DnaJ protein TGL63

PKNH_1207600 Cons. Plasm. protein, u.f. TGL64

PKNH_1306400 Rhodanese like protein, put. TGL65

PKNH_0423200 Protein kinase G TGL66

Table 7: Targeting constructs available for initial P. knowlesi PlasmoGEM experi-

ments

constructs are intended to knock out, and any expected mutant phe-

notypes.

4.4.2.1 Target 1: A type IV J protein

PKNH_0941100 encodes a DnaJ protein which is conserved in Plas-

modium but lacks orthologues in other Apicomplexa despite the pres-

ence of orthologs in metazoans, as shown in Fig. 45. Transcription of

its P. falciparum orthologue, PF3D7_1143200

, peaks at 33 hours post

invasion and despite initial suggestions of an apicoplast localisation

[21], proteomic data suggested it to be present in the merozoite. [89]

The Hsp40 family, of which this protein is a member, is expanded

in P. falciparum with 43 members and this expansion is hypothesised

to be linked to the parasitic lifestyle and the need to invade and re-

model erythrocytes. [21] It is divided into subfamilies I, II, III and

IV. PF3D7_1143200 falls into the type IV family. 11 of 12 members of

this subfamily have PEXEL export sequences. PF3D7_1143200 is the

solitary exception which does not.

PF3D7_1143200 was conﬁrmed to be an integral membrane protein

which was expressed from 37 hours post invasion until egress and

was absent from rings. [89] Its localisation is initially reminiscent of

4 A PhD thesis [89] provides by far the best review of what is known about

PF3D7_1143200

4.4 plasmodium knowlesi gene targeting 119

Arabidopsis lyrata subsp. lyrata

Daphnia pulex

Tetrahymena thermophila

Ichthyophthirius multiﬁliis

Plasmodium fragile

Plasmodium inui

Plasmodium vivax

Plasmodium cynomolgi

Plasmodium knowlesi strain H

Plasmodium yoelii

Plasmodium berghei strain An a

Plasmodium berghei strain An b

Plasmodium vinckei petteri

Plasmodium chabaudi chabaudi

Plasmodium reichenowi

Plasmodium falciparum isolate Camp / Malaysia

Plasmodium falciparum Santa Lucia

Plasmodium falciparum isolate NF54

Plasmodium falciparum Vietnam Oak-Knoll FVO

Plasmodium falciparum Tanzania 2000708

0.9

0.67

0.69

0.85

0.99

Figure 45: Phylogenetic analysis of sequences homologous to PKNH_0941100

A BLAST search was carried out against the UniProt database.

The closest matching sequences were aligned with Muscle, cu-

rated with Gblocks and phylogeny computed with PhyML. Boot-

strap values were calulated with a aLRT test. [35, 85, 9, 30, 60, 52]

Each Plasmodium species has a single copy of the gene. The near-

est outgroup includes 2 members of the Alveolata, but also a

metazoan and a plant. Other apicomplexans lack a similar gene

entirely, suggesting that it plays a speciﬁc and important role

across Plasmodium.

the endoplasmic reticulum but becomes apical to the nuclei of mero-

zoites in segmented schizonts, localising either to the micronemes or

the rhoptries. [89] The protein is not detected in culture supernatant,

nor on the surface of free merozoites. [89]

In P. falciparum, a single-crossover C-terminal FLAG tag of the gene

was created. Subsequent immunoprecipitation identiﬁed, among other

proteins, rhoptry neck proteins RON3, RON4 and RON5 as possible

interactants. [89]

In contrast, attempts to knock the gene out with a double crossover

approach in P. falciparum were not successful despite three attempts.

Some parasites were recovered in which the 3’ UTR was ablated, but

these still showed normal protein expression levels. This led to the

suggestion that the protein was likely essential in P. falciparum. [89]

In parallel the orthologous gene (PBANKA_0905800) was included

in a P. berghei screen for proteins involved in cytoadherence (which

would reﬂect a non-invasive function). It was included in that screen

because some corresponding peptides were found in proteomic exper-

120 the transfer of plasmogem technology to P. KN OWLESI

iments designed to detect proteins presented on the erythrocyte mem-

brane of schizonts. [64] Two methods were used to identify erythro-

cyte membrane proteins: hypotonic lysis and surface shaving. The

authors were concerned about the possibility of contamination of the

surface-shaving peptides with merozoite membrane proteins and so

they subtracted any proteins previously found in a merozoite pro-

teome. PBANKA_0905800 peptide hits were identiﬁed in the hypo-

tonic lysis set (in which merozoite contamination was not expected),

and not in the surface shaving set, nor the merozoite proteome. This

might suggest a different role for this gene in P. berghei, but could

equally be the result of noise (MTRAP, a known invasion gene, had

the same number of peptide matches as PBANKA_0905800 in the

hypotonic lysis proteome). In this experiment mutants could be gen-

erated but were not cloned by limiting dilution. [64]. No further anal-

ysis was conducted.

4.4.2.2 Target 2: DHHC3

The palmitoyl-transferase DHHC3 appears to be localised to the IMC

in P. berghei, which would be consistent with a role in invasion. [65]

However in P. falciparum the orthologous protein localises to the Golgi.

[188]

In P. berghei DHHC3 could be deleted, with integration of the marker

detected on a Southern blot, but the parasite was not cloned nor phe-

notyped in detail. [65] However an attempt to knock-out the same

gene in P. falciparum with a double-crossover approach was not suc-

cessful – parasites were recovered which carried episomes and inte-

gration could not be forced despite drug cycling and the application

of negative selection with 5-ﬂuorocytosine. [188]

4.4.2.3 Target 3: PKNH_1207600

PKNH_1207600 is a typical "conserved Plasmodium protein, unknown

function" in that we know almost nothing about it. The sequences

from species in the simian clade are predicted to have a transmem-

brane domain and a non-cytoplasmic C-terminus by Phobius (but not

TMHMM). The only orthologs outside Plasmodium on OrthoMCLDB

is a highly divergent Dictyostelium discoideum, there are no orthologs

in the apicomplexa outside Plasmodium (Fig. 46).

4.4 plasmodium knowlesi gene targeting 121

P. knowlesi

P. chabaudi

P. berghei

P. reichenowi

P. falciparum

P. yoelii

P. vivax

P. cynomolgi

Cytoplasmic

Non-cytoplasmic

Figure 46: Phylogenetic analysis of PKNH_1207600 orthologs in Plasmodium

Protein sequences were retrieved from PlasmoDB. Sequences

were aligned with Muscle, curated with Gblocks and phylogeny

computed with PhyML. [35, 85, 9, 30, 60, 52] Phobius was used

to predict trans-membrane domains. One was identiﬁed in the

simian clade only.

4.4.2.4 Target 4: PKNH_1306400

PKNH_1306400 is a rhodanese-like protein. Rhodanese is found in

the mitochondria where it detoxiﬁes cyanide by acting as a sulphur-

transferase, but the domain is also found in other classes of protein in-

cluding phosphatases and ubiquitin C-terminal hydrolases. The gene

is 1:1 orthologous across all Plasmodium genomes. PKNH_1306400

does have a T. gondii homolog, but is not found in Babesia and Theileria.

There is very little additional detail available about this protein.

4.4.2.5 Target 5: A serine/threonine protein kinase

PKNH_0406200 is a putative serine/threonine protein kinase. As ki-

nases have been a focal point of previous reverse genetic research

we have some data from orthologs. The orthologous P. berghei gene,

PBANKA_0311400 has been described as essential in three experi-

ments. In the P. berghei conventional kinome screen the gene could

not be knocked out despite ﬁve attempts. Then in the P. berghei ki-

nome barseq screen too it was described as likely essential, [80] and

122 the transfer of plasmogem technology to P. KN OWLESI

P. vivax

P. cynomolgi

P. knowlesi

P. chaubaudi

P. berghei

P. yoelii 17XNL

P. yoelii 17X

P. yoelii YM

P. falciparum 3D7

P. falciparum IT

P. reichenowi

T. gondii

0.81

0.99

0.93

Figure 47: Phylogenetic analysis of DHHC3 orthologs in Plasmodium and T.

gondii

Protein sequences were retrieved from PlasmoDB, and the T.

gondii DHHC13 sequence added. Sequences were aligned with

Muscle, curated with Gblocks and phylogeny computed with

PhyML. Bootstrap values were calulated with a aLRT test. [35, 85,

9, 30, 60, 52] The tree is that one would predict from the evolution

of the parasite species shown in Fig. 4.

was described thus again in the invadome barseq experiment in the

previous chapter.

In P. falciparum the picture is not so clear. A kinase screen described

the gene as likely non-essential. However this was on the basis of an

intense signal obtained in the integration PCR assay, using primers

either upstream or downstream of the target gene. [181] Successfully

targeted clones could not be recovered by limiting dilution cloning.

This experiment clearly proves that integration is occurring but not

that the integrated parasites are indeﬁnitely viable.

4.4.2.6 Target 6: cGMP-dependent protein kinase (PKG)

PKG is a well-characterised essential protein kinase, with roles in

egress and invasion established in P. falciparum and P. berghei. [4] I

chose to include it as a negative control, to investigate the result of

targeting essential genes.

4.5 transfection of plasmogem vectors

Because P. knowlesi is not a widely used genetic system, there was

no direct data on whether any of the modiﬁcations for which the

constructs in Table 7 were designed would be tolerated. Seeking a

positive control, I decided that the vectors most likely to be tolerated

were the three HA-tagging constructs. These were transfected into

4.5 transfection of plasmogem vectors 123

LadderrTGL58 trans.

4 kb

3 kb

2.5 kb

2 kb

(a) First positive integra-

tion PCR band from

dying parasites ini-

tially transfected with

TGL60.

(b) Subsequent positive

integration PCR from

second round of

transfections.

1kb-

2kb-

3kb-

Trans WT Trans WT

PKNH_

1207600

PKNH_

0304000

Figure 48: Positive integration PCR results

parasites as described in Section 2.2.6 and selected from 1 day after

transfection with 200 nM pyrimethamine.

Parasitaemia dropped rapidly upon selection, but two weeks after

transfection I could see a barely visible parasitaemia in the PKNH_1126300

transfection. I decided to isolate DNA from these parasites and per-

form an integration PCR. After 35 cycles of PCR a band was visible

at the expected size (Fig. 48a).

Despite this, parasites did not come up after 35 days in any of the

transfections.

My interpretation of these results was that correctly targeted par-

asites were being formed but were for some reason not surviving. I

made two changes for all subsequent experiments:

• Drug concentration - I dropped the pyrimethamine concetration

to 100 nM, which my previous characterisation (page 113) sug-

gested would still be selective.

• Erythrocyte replacement - I replaced at least 50% of the erythro-

cytes each week to ensure there were always young erythrocytes

present for optimal invasion.

I used this technique to transfect all nine constructs shown in Ta-

ble 7. Selection was applied one day after transfection with 100 nM

pyrimethamine. The ﬁrst parasites were seen on day 9, and parasites

were up in all nine transfections by day 14.

While the presence of drug-resistant parasites was encouraging, it

was alarming that the knock-out vector for PKG gave rise to viable

parasites when transfected. This vector had been included as a nega-

tive control, and I did not expect to see viable parasites if the locus

124 the transfer of plasmogem technology to P. KN OWLESI

was successfully modiﬁed. Since I had transfected linear DNA I had

not expected any form of episomal carriage. Thus it was essential to

genotype the parasites.

4.5.1 Genotyping

GOI

Dhfr selection cassette

GW1

WT1

5’ int

GW2

WT2

dhfr1

dhfr2

3’ int

Wild-type genome

pJAZZ vector

Modified locus

barcode

(a) Generic map for knock-out vectors.

QCR1

5’ int

QCR2

dhfr1

dhfr2

3’ int

GOI

Dhfr cassette3xHA

QCR2

dhfr1

dhfr2

Dhfr cassette3xHA

Wild-type genome

pJAZZ vector

Modiﬁed locus

barcode

(b) Generic map for HA-tagging vectors.

Figure 49: Maps of double crossover integration strategies using Plasmo-

GEM vectors both for knock-outs and tagging in P. knowlesi with

the primers used for genotyping by PCR and qPCR shown.

A range of genotyping techniques were used to consider the possi-

ble events that had given rise to drug-resistant parasites.

4.5.1.1 Genotyping by single barcode sequencing veriﬁes transfection of

correct constructs

One possible explanation for seeing parasites in all wells is cross-

contamination. I therefore performed testing to exclude any possibil-

ity I had accidentally transferred parasites from a viable transfection

into the PKG well, or contaminated the DNA prior to transfection. As

discussed in 1.9.3, PlasmoGEM vectors carry barcodes between con-

served primer binding sites – primarily for the purposes of barseq on

the Illumina platform. However I was also able to use this feature to

rule out cross-contamination – I extracted genomic DNA from par-

4.5 transfection of plasmogem vectors 125

asites and ampliﬁed a region containing the barcode (as described

in Section 2.2.12), puriﬁed the PCR product and sent it for capillary

sequencing by GATC Biotech.

I streamlined this technique by writing a script which searches

a capillary sequence for barcode primers, extracts the barcode and

queries a table to report the relevant gene to which that barcode cor-

responds.

In all nine cases the correct barcode was detected—albeit in one

case with a single nucleotide error (probably a sequencing artefact)—

giving conﬁdence that parasites were resistant due to transfection

with the correct vector.

4.5.2 End-point PCR demonstrates that linear PlasmoGEM vectors are

able to integrate into the P. knowlesi genome

I screened the isolated genomic DNA by end-point PCR using primers

in the genome outside the homology arms of the vectors, Fig. 49 and

without further opimisation was able to obtain positive integration

PCRs for PKNH_0304000 and PKNH_1207600. Integration PCRs can

fail on pJAZZ vectors due to the long-range of ampliﬁcation necessi-

tated by the long homology arms, which can be exacerbated by Plas-

modium’s (A+T)-rich stretches.

However for all nine transfections PCR to detect the wild-type un-

modiﬁed locus also produced a bright amplicon band on an ethidium-

bromide stained gel, suggesting that in no case was there a clonal pop-

ulation of modiﬁed parasites. It was therefore necessary to use other

approaches to identify which PCR result was most representative of

the population.

4.5.3 Quantitative PCR indicates the proportion of parasites carrying ge-

nomic alterations

I had two concerns about integration PCR as a way of assessing my

ability to alter the parasite genome. The ﬁrst was the possibility of

false negative results, since the use of pJAZZ vectors with long ho-

mology arms necessitates amplifying large regions to reach part of

the genome which was not included in the initial transfection vector.

126 the transfer of plasmogem technology to P. KN OWLESI

It is possible that these amplicons will fail even if the integrated locus

is present.

The second concern was false-positives due to the non-quantitative

nature of an integration PCR. Theoretically a single parasite contain-

ing an altered genome could be enough to allow the many cycles of

a PCR to produce a band demonstrating integration, and while this

is a true result, it is not indicative of the population and will yield a

disappointing result upon dilution cloning. Furthermore the parasite

containing the integration might even be dead, or dying, due to the

modiﬁcation. A ﬁnal concern is that an integration PCR can occasion-

ally show a true integration event, but one which has only been made

possible by the parasite duplicating the region containing the target

gene, [79] so that there has been a real modiﬁcation of the genome

but the result is not the knock-out of the target gene. This latter con-

cern also applies to pulsed-ﬁeld gel electrophoresis, which conﬁrms

merely the chromosome which has been targeted, not the nature of

the event on this chromosome.

I hypothesised that all of these issues would be solved by using

quantitative PCR (qPCR) to amplify a region expected to be removed

in successfully modiﬁed parasites.

In the case of knock-out attempts primers were designed within the

knocked out gene using PRIMER3. In the case of tags the PlasmoGEM

QCR1 and QCR2 primers were used. These bind just upstream of the

tag, and just downstream of the stop codon, thus producing a small

amplicon from wild-type locus. They could in theory also produce an

amplicon from the tagged locus but it would be >5kb. Provided short

extension times are used, this amplicon is not generated.

The abundance of a sequence as detected by qPCR is a factor of

three variables:

1. The parasitaemia of the sample

2. The efﬁciency of DNA extraction

3. The proportion of parasites which still carry the target locus

I used the well-characterised primers Plasmo1 and Plasmo2 [165] to

establish the baseline amount of genomic DNA present in the sample,

controlling for factors 1 and 2.

4.5 transfection of plasmogem vectors 127

P a ra site lin e

R ela tiv e a b un da n ce

0.0

0.5

1.0

1.5

2.0

2.5

Plasmo1, Plasmo2

63 - PKNH_0941100

62 - DHHC3

Figure 50: qPCR shows successful disruption of the PKNH_0941100 gene

Relative abundance was calculated by taking 2

´C

, and nor-

malised for each primer pairto the TGL58 transfection. There

are three primer sets shown here. Plasmo1/Plasmo2 are well-

characterised primers which detect DNA from any parasite

genome. The two amplicons 62 and 63 are located within DHHC3

and PKNH_0941100 respectively. We see different samples have

different abundances for the Plasmo1/Plasmo2 amplicon, repre-

senting parasitaemia and the efﬁciency of DNA extraction. The

two regions of interest are highlighted in boxes. TGL62 has levels

similar to other amplicons. This indicates that although resistant

parasites have appeared the transfection attempt has not success-

fully produced an integrated population. In contrast for TGL63,

wild-type amplicon levels are very much lower than would be

expected from the Plasmo1/Plasmo2 primers and so we can in-

fer that the knock-out has succeeded in the vast majority of the

parasite population.

In initial experiments to establish the effectiveness of this qPCR

method I used each pair of primers on every sample. The results for

three primer pairs are shown in Fig. 50.

The results show that amplicons co-vary between samples, repre-

senting different parasitaemias and efﬁciencies of extraction. All three

primer pairs covary for DHHC3 indicating that most parasites retain

the locus, but for the attempt to knock-out PKNH_0941100 the am-

plicon located within the target is present at a far lower level than

the other primer pairs, showing that most parasites have this region

ablated.

128 the transfer of plasmogem technology to P. KN OWLESI

The Plasmo1/Plasmo2 primers can be used to calculate a normalised

copy number for each gene. I used this to calculate the copy-numbers

of all 9 genes.

Tra n sfe c ta n t

C o py num b er

58B -03

04000

59B - 0406200

60B - 1126300

61B - 1136700

62B - 0918700

63B - 0941100

64B - 1207600

65B - 1306400

66B - 0423200

0.0

0.5

1.0

1.5

Figure 51: qPCR results reveal four modiﬁcations which are dominant in their pop-

ulations after transfection. Amplicons for TGL58, TGL60, TGL61,

TGL63 are very much lower in abundance in their corresponding

transfection than other controls, indicating four successful modi-

ﬁcations: three tags and one knock-out.

Four showed successful modiﬁcation (Fig. 51). Five showed wild-

type amplicons still present at the wild-type level. These results are

summarised in Table 8.

gene id description construct result

ha tagging constructs

PKNH_0304000 Apical sushi protein TGL58 Success

PKNH_1126300 Ribonuclease H2 subunit A TGL60 Success

PKNH_1136700 ROP14 TGL61 Success

knock-out constructs

PKNH_0406200 Ser/Thr protein kinase TGL59 Failure

PKNH_0918700 DHHC3 TGL62 Failure

PKNH_0941100 DnaJ protein TGL63 Sucess

PKNH_1207600 Cons. Plasm. protein, u.f. TGL64 Failure

PKNH_1306400 Rhodanese like protein, put. TGL65 Failure

PKNH_0423200 Protein kinase G TGL66 Failure

Table 8: Result of pJAZZ transfection attempts

It is striking that every HA-tagging attempt was successful, whereas

only one of the knock-outs attempted was. This, combined with the

4.5 transfection of plasmogem vectors 129

failure to knock-out negative control PKG, gives a ﬁrst indication that

these experiments are producing meaningful results.

4.5.3.1 HA-tagging in P. knowlesi is efﬁcient and permits immunoﬂuores-

cence analysis

There are two factors meaning that we would expect tagging to suc-

ceed. C-terminal tagging with a small epitope, is unlikely to affect the

functionality of the majority of most proteins (though in rare cases it

might result in mislocalisation by interfering with a localisation sig-

nal). And secondly, tagging does not remove any part of the genome,

and so the length of the region which must be excised by host nucle-

ases is less than for a knock-out.

As a proof of concept that pJAZZ HA-tagging vectors could be

useful in P. knowlesi I decide to attempt immunoﬂuorescent analysis

of apical sushi protein using the parasites that came up from the

TGL58 transfection which attempted to HA-tag apical sushi protein.

After a number of unsuccessful attempts involving methanol/acetone

ﬁxation I used a paraformaldehyde ﬁxation method [193] and was

able to observe an apical localisation (Fig. 52).

The organisation of P. knowlesi schizont, with merozoites radiating

from the centre of the parasites, appeared to prevent the resolution

of individual nuclei. In part due to the lack of P. knowlesi speciﬁc anti-

bodies and in part because the position of apical sushi protein in the

rhoptries is already well established, I did not attempt co-localisation.

4.5.3.2 Attempt to disrupt DnaJ protein PKNH_0941100 was successful

Quantitative PCR identiﬁed a very signiﬁcant decrease in abundance

of the amplicon located within this gene, indicating that it has been

lost in the majority of the population. This makes it the ﬁrst gene to

be knocked-out in P. knowlesi using a PlasmoGEM approach.

This success in P. knowlesi is in contrast to several attempts in P.

falciparum but in agreement with results from P. berghei. While it is

possible that—like DHHC3 below—this protein plays different roles

in the different species, the fact that it can now be knocked out in

two out of three species would probably justify renewed attempts at

deletion in P. falciparum using newly developed CRISPR approaches.

130 the transfer of plasmogem technology to P. KN OWLESI

BF HA DAPI Merge

(a) The ﬂattened z-stack depicting a mature schizont, with punctate

staining labelling what are presumably the rhoptries.

DAPI HA

Combination

(b) Three-dimensional deconvolution of z-stack

Figure 52: Immunoﬂuorescence of 3xHA-tagged apical sushi protein in P. knowlesi

reveals the expected apical localisation.

4.5 transfection of plasmogem vectors 131

260

270

280

290

P. berghei

P. knowlesi

P. falciparum

Conservation

Figure 53: The C-terminus of PkDHHC3

While the DHHC3 genes of P. knowlesi, P. falciparum and P. berghei

are in general well conserved, and have the same evolutionary

relationship as their respective species (as shown in Fig. 47), the

C-terminus is not well conserved.

4.5.3.3 Attempt to distrupt DHHC3 was unsuccessful

I was not able to knock-out DHHC3. This is in accordance with pre-

vious attempts in P. falciparum, but not with those in P. berghei where

a knock-out was successfully made. [65] It is possible that the P. fal-

ciparum transfection failed because of the greater difﬁculty in geneti-

cally modifying this species, but the fact that DHHC3 is differentially

localised in P. falciparum and P. berghei [188] suggests they play dif-

ferent roles in these species. We do not know, of course, how the

P. knowlesi protein is localised. However there are suggestions from

research on the mammalian palmitoyltransferases suggests that C-

termini may be involved in DHHC localisation. [83] The C-terminus

of PkDHHC3 shows no greater homology to PbDHHC3 than it does

to PfDHHC3 (Fig. 53) allowing the possibility that like PfDHHC3,

PkDHHC3 may be a putatively essential Golgi-targeted enzyme.

It would be interesting to localise DHHC3 in P. knowlesi to shed

further light on these possibilities. This could be easily achieved with

PlasmoGEM tagging vectors.

4.5.3.4 Attempt to disrupt protein kinase G was unsuccessful

We could not detect any signiﬁcant number of parasites with this

locus removed, as we would expect given it is the negative control.

Nevertheless drug-resistant parasites did come up in this transfection

and their origin must be investigated.

4.5.3.5 Attempt to disrupt PKNH_0406200 was unsuccessful

We could not detect any signiﬁcant number of parasites with this lo-

cus removed, which is in accordance with a number of transfection at-

tempts in P. berghei but not with the postive integration PCR reported

in P. falciparum. Given the potential weaknesses of integration PCR

132 the transfer of plasmogem technology to P. KN OWLESI

Table 9: Summary of results for three genes for which knock-out attempts have been

made in three Plasmodium species

p. falciparum p. berghei p. knowlesi

PKNH_0941100 Refractory to deletion Deleted Deleted

PKNH_0918700 Refractory to deletion Deleted Refractory to deletion

PKNH_0406200 Integration PCR but uncloned Refractory to deletion Refractory to deletion

outlined in the chapter earlier, it may be worth revisiting this gene in

P. falciparum and attempting to produce a clone (with the help of new

CRISPR technology). If this does not succeed we can judge the gene

likely to be essential across the three species.

4.5.3.6 Attempts to disrupt PKNH_1207600 and PKNH_1306400 were

unsuccessful

We could not detect a signﬁcant number of knock-out parasites for

PKNH_1207600 or PKNH_1306400. This means there is a possibility

these genes are essential or at least that their deletion has some ef-

fect on parasite ﬁtness. But with a lack of data on the proportion of

targetable genes in P. knowlesi these suggestions are not conclusive.

4.5.4 Investigating the cause of resistant parasites without genomic modi-

ﬁcation

At this point it is clear that attempts to knock-out essential genes in P.

knowlesi can give rise to viable parasites which still possess the target

locus. I considered a number of possible explanations for this result.

4.5.4.1 Parasites have more than one copy of human dhfr per genome

One concern is that these parasites might have spontaneously de-

veloped pyrimethamine resistance, which requires only a single nu-

cleotide substitution [5]. To investigate this possibility I assayed what

proportion of parasites carried the human dhfr gene introduced by

my constructs – or at least what the average copy number of this gene

was. I used PRIMER3 to design primers to the human dhfr gene, and

to the PkHsp70 UTR that is used to express it and performed qPCR

(Fig. 54).

4.5 transfection of plasmogem vectors 133

Parasite line

Relative abundance

0.0

0.5

1.0

1.5

2.0

2.5

WT primer pair

(Plasmo1, Plasmo 2)

dhfr

hsp70

Figure 54: Elevated dhfr copy number in non-integrating transfections suggests

the presence of episomes

Dhfr and hsp70 abundance in genomic DNA from each transfec-

tant is indicated. Primer pairs were normalised to each other by

setting to a relative abundance of 1 for the successful TGL58 trans-

fection.

The primer pairs in this analysis were normalised to a success-

ful transfection - TGL58, which we expect to contain one copy the

Plasmo1/Plasmo2 amplicon, one copy of human dhfr and two copies

of PkHsp70 UTR.

We can make two observations. Firstly it is clear that the resistance

marker is present in signiﬁcant quantities in all samples – sponta-

neous resistance is not at issue. Secondly, we can also see that four

out of ﬁve of the failed deletion attempts have more than 1 copy of

human dhfr per genome. (The Hsp70 amplicon appears at an inter-

mediate level because it is present at two copies in the control and so

each extra plasmid copy increases copy number by 50% of the con-

trol rather than 100%.) This presence of multiple copies could be the

result of multiple insertions into some site in the genome (perhaps

segmental duplications of the target locus), or the formation of epi-

somes (somehow from linear DNA).

I decided to attempt to investigate these possibilities further by

whole genome sequencing, but ﬁrst I dilution cloned parasites in or-

134 the transfer of plasmogem technology to P. KN OWLESI

der to remove potential confounding factors from the analysis. I also

attempted to dilution clone the PKNH_0941100 knock-out to estab-

lish whether a clean knock-out could be generated.

Analysis of the result of dilution cloning parasites transfected with

this construct (Fig. 55) revealed that non-targeted parasites had been

effectively eliminated: a very low signal was seen for the unmodifed

locus, and melt-curve genotyping revealed this to be a different prod-

uct (data not shown), indicating the success of dilution cloning.

Parasite line

Relative abundance

e d

n c

-7

-6

-5

-4

-3

-2

-1

Deleted region (405, 406)

Plasmo1,Plasmo2

(present in all parasites)

Figure 55: Quantitative PCR conﬁrms attempt to dilution clone PKNH_0941100

knock out.

The apparent level of intact locus fell by more than 5 orders of

magnitude (and melt curve analysis revealed the remnant locus

apparently to be the result of non-speciﬁc ampliﬁcation.)

4.5.5 Whole genome sequencing of transfectants

DNA was extracted from ﬁve successfully dilution cloned cultures

(TGL58, TGL60, TGL62, TGL63, TGL64). This DNA was labelled with

index tags and the 5 samples pooled on one lane of a MiSeq.

The sequencing data was analysed in three ways.

4.5 transfection of plasmogem vectors 135

4.5.5.1 Barcode veriﬁcation

Once again, a search was made for barcodes to verify that no cross

contamination had occurred. The grep command was used to search

raw reads for the barcode primer sequence, and the barcode following

this was checked to ensure it corresponded to the intended gene. In

each case it did.

4.5.5.2 Mapping

These genomic DNA reads were mapped with bowtie2 to the P. knowlesi

genome, and some of the common heterologous sequences used in

the transfection, as described in section 2.2.15.2.

This method of analysis is capable of detecting SNPs and copy-

number changes, as well as giving some hints as to the location of

indels. As anticipated, the DNA extracted from parasites in which

we had attempted to knock out PKNH_0941100 had essentially no

coverage in the knocked-out region of the gene (Fig. 56).

Figure 56: Whole genome sequencing reveals successful gene deletion at the

PKNH_0941100 locus (highlighted in red)

The other two sequenced knock-outs, in which qPCR data had sug-

gested that the knock-out had not succeeded, had very different pat-

terns of coverage to this clean knock-out.

In both cases (Figs 57 and 58), the level of coverage at the gene

itself was in line with coverage at any other arbitrary location in the

genome. However regions on either side of the gene had elevated

coverage.

136 the transfer of plasmogem technology to P. KN OWLESI

Figure 57: Whole-genome sequencing reveals a failure to delete PKNH_1207600

(highlighted in red).

The green line indicating the coverage from the parasites of in-

terest, transfected with a pJAZZ vector targeting PKNH_1207600.

Their coverage peaks in regions either side of the gene, rising

to more than 2x normal coverage. Other colours indicate control

populations (parasites transfected with other constructs).

There are two theoretically possible origins for this increased cov-

erage. The ﬁrst possibility is that a segmental duplication occurred

before the targeting construct was incorporated. This would have al-

lowed one of these copies to be knocked out without affecting par-

asite ﬁtness, returning coverage to 1x in that region. The alternative

possibility, given that the entirety of the region at higher coverage is

contained within the homology region that exists on the vectors, is

that the increased coverage may come merely from additional copies

of the vector. These would be present in some form that does not

involve knocking out the target locus: i.e. either an episome or a non-

targeted integration.

The synthetic sequence I constructed for mapping to was made

up of the P. knowlesi genome concatenated to the selection cassette

and this enables the copy number of the cassette to be estimated.

Fig. 59 shows a difference here between the two failed attempts, one

to knock out PKNH_1207600 and one to knock out DHHC3. The copy

number of the selection cassette is much more than 1x in the case of

PKNH_1207600 but at about 1x for DHHC3.

It was necessary to understand the context in which the additional

reads corresponding to the homology regions were found. This could

4.5 transfection of plasmogem vectors 137

Figure 58: Whole genome sequencing reveals failure to target DHHC3.

The targeted gene is shown in red, the blue coverage line is from

the attempt to knock it out and has coverage which increases in

the region either side of the target gene.

indicate whether they corresponded to episomes or to some duplica-

tion event. I started to examine the subset of reads that did not come

from ‘proper pairs’. Proper pairs are reads which map as expected, to

opposite strands a small distance apart. When I investigated reads at

the breakpoint of the increased coverage I found a preponderance of

reads whose mates did not map anywhere in the P. knowlesi genome,

nor to the gateway cassette. I took the sequence of one of these mates

and queried it against the NCBI’s non-redundant nucleotide collec-

tion. The unexpected result was that this sequence belonged to the

backbone of the pJAZZ vector. This result was unexpected because

the PlasmoGEM protocols I followed called for the pJAZZ arms to be

removed by NotI digestion. However in retrospect this should have

been less of a surprise since we know from running the digest by gel

electrophoresis that NotI digestion is substantially incomplete (even

when the amount of DNA is substantially reduced).

I constructed a new synthetic genome to map to, this time contain-

ing the pJAZZ arms, and remapped reads to this new supercontig.

The mapping results for the ﬁve whole-genome sequenced clones

are shown in Fig. 60. Much, but not the entirety of the long pJAZZ

arm is present for PKNH_1207600, but the entire short arm is absent.

Additionally there is a step-change in coverage at one point in the

arm. This must represent a site of rearrangement with multiple copies

of the outer region of the long arm, but the exact structure is hard to

deduce because of the repetitive nature of the sequence.

138 the transfer of plasmogem technology to P. KN OWLESI

Genomic region Selection cassette

Figure 59: Coverage of selection cassette as compared to the general genome in

whole genome-sequenced samples.

All samples show similar coverage with the exception of the at-

tempt to knockout PKNH_1207600 (pink), which shows greatly

elevated levels of the selection cassette.

Fig. 60 demonstrates that successfully targeted parasites lack any

sequences from the pJAZZ arms (as one would expect from double

crossover), and that the other failed knock-out analysed, DHHC3 also

posesses only the pJAZZ long arm. Interestingly, a greater length of

the long arm is included and there is a coverage increase that begins

at a different point, illustrating that the presumed recircularisation

event that gives rise to episomes does not occur in only one way.

The presence of the long pJAZZ arm seems to be a common factor

in the transfectants which come up without targeting the correct lo-

cus, sometimes with noticeable copy number changes. This suggests

to me that P. knowlesi is able to recircularise linear DNA with which

it is transfected and maintain it, although it is not impossible that

a concatomeric form of the cassette is incorporated into the parasite

genome.

4.5.6 pJAZZ transfection results in P. knowlesi are reproducible

Finally, to identify whether the transfection results I obtained were

repeatable I transfected 7 of the constructs (TGL58-TGL65) an addi-

tional time. On this occasion I lowered the amount of DNA trans-

fected to 3 µg to see whether I could reduce the number of non-

targeting events which occurred. The same four successful modiﬁca-

tions were achieved (TGL58, TGL60, TGL61, TGL63), as assessed by

4.6 towards crispr/cas9 in p. knowlesi 139

pJazz long arm pJazz short arm

TGL62

TGL64

No reads:

TGL58

TGL60

TGL63

Figure 60: Genome sequencing reveals transfectants still have regions from

the pJAZZ backbone, despite NotI digestion prior to transfection.

qPCR. Apparent episomes were formed in all other cultures except

TGL62, where parasites did not come up.

This suggests that reduced DNA concentration, or perhaps reduced

transfection efﬁciency achieved in any way, may have a role in increas-

ing the speciﬁcity of targeting. The reproducibility of the successful

results is encouraging.

4.6 towards crispr/cas9 in p. knowlesi

The data from [142] shown in Fig. 6 suggest that P. knowlesi appears

to be the malaria parasite species best able to take up exogenous

DNA. This effect is shown by its maximal transfection efﬁciency of

30% [142], as well as its consistent ability to give rise to drug-resistant

parasites after transfection shown in this chapter. However the speed

with which vectors can be integrated into the parasite does not par-

allel this. Parasites with integrated DNA come up 11 days later than

those with episomes [142] and I estimated in the introduction, based

on this data that just 1 in 5,000 parasites that take up DNA result in

an integration event.

Clearly any strategy to increase this proportion would be very use-

ful both to increase the speed with which modiﬁed parasites could

be generated in P. knowlesi and perhaps to lower the rate of episome

formation seen in this chapter.

140 the transfer of plasmogem technology to P. KN OWLESI

One of the limiting factors on the integration of DNA into the par-

asite genome may be the formation of a double-strand break in the

genomic locus to be targeted. In P. falciparum the use ﬁrst of zinc-

ﬁnger nucleases [184, 144] and then very recently of CRISPR/Cas9

[212, 72, 201] has permitted greater integration efﬁciencies and the

isolation of correctly targeted parasites without drug cycling.

I decided to attempt to bring this technology to P. knowlesi, hoping

that the combination of a high transfection efﬁciency with techniques

to boost the rate of integration might lead to an ideal genetic system.

In this section I will describe the progress made towards this goal.

4.6.1 Bio-informatic prediction of cutting efﬁciency of guide RNAs target-

ing the P. knowlesi genome

Whereas a number of online guide-RNA design tools support P. fal-

ciparum, none were available which supported P. knowlesi as I began

this project. In order to ultimately use CRISPR at scale, and also to

analyse its potential in P. knowlesi, I decided to create a streamlined

process for designing sgRNAs for P. knowlesi. There are three consid-

erations for the design of guide RNAs: potential off-target sites, the

on-target cutting efﬁency and the location of the break.

Based on BLAST searches of candidate guide RNAs, I hypothesised

that, given the much reduced size of the Plasmodium genome as com-

pared to that of mammals, off-target cutting was unlikely to be a con-

cern – the higher (G+C)-content of P. knowlesi also should minimise

off-target activity as compared to P. falciparum. In addition, given

that Plasmodium does not possess machinery for non-homologous end

joining, if off-target cutting occurs in occasional parasites it will sim-

ply result in their deaths rather than in non-targeted insertion events.

I therefore focused on on-target cutting efﬁciency and break-location.

Algorithms to estimate on-target cutting efﬁciency based on se-

quence have been created for mammalian systems [55]. While it is

not clear whether these factors will apply to Plasmodium also, if they

simply describe the interaction between the Cas9 protein and DNA

then it is seems likely that they might. In the absence of speciﬁc Plas-

modium data on on-targeting efﬁciency, I decided to use this scoring

system.

4.6 towards crispr/cas9 in p. knowlesi 141

10000

20000

30000

40000

0.0 0.2 0.4 0.6 0.8 1.0

On−target score

Frequency (sgRNAs)

Figure 61: Distribution of on-target cutting scores for all possible guide RNAs in

the P. knowlesi genome

Plotting the scores from all 1.25 million possible sgRNAs in

P. knowlesi indicates that the majority have scores associated

with potentially low on-target cutting efﬁciency. Nevertheless the

small proportion of high-scoring guides still represents a sizeable

number.

I wrote a script to identify, and score, every possible guide RNA in

the P. knowlesi genome. The strategy employed was to search through

a Fasta ﬁle containing genomic sequences for every P. knowlesi gene

(downloaded from PlasmoDB) and to identify every PAM (Proto-

spacer adjacent motif) site, and the corresponding guide RNA se-

quence. Then it also extracted the 30-mer containing this sequence

in a genomic context and calculated the on-target score according to

[55].

4.6.1.1 Theoretical feasibile scale of CRISPR in P. knowlesi

My analyses identiﬁed 1.25 million candidate guide RNAs within the

coding sequences of P. knowlesi. The vast majority of these have low

on-target scores as shown in Fig. 61. This is also reported in mam-

malian systems and in fact Fig. 62a shows that there are a higher pro-

portion of high-scoring RNAs in P. knowlesi than in M. musculus – this

may be because of a reported ’strong bias against guanine immedi-

ately 3’ of the PAM’[55] which would be reduced in an (A+T)-rich or-

ganism such as P. knowlesi. In practice one is typically using only one

or a handful of guide RNAs at a time, so the most important consid-

eration is what proportion of genes are targetable by high-efﬁciency

guides.

142 the transfer of plasmogem technology to P. KN OWLESI

0.8–1.0

0.6–0.8

0.4–0.6

0.2–0.4

0–0.2

M. musculus P. knowlesi

(a) The left hand side of this plot is adapted from Doench et al. (2014) The

right hand side was calculated by me based on the full 1.25 million sgR-

NAs that exist within the genes of P. knowlesi. [55]

200

400

600

800

0.0 0.2 0.4 0.6 0.8 1.0

Maximum on−target score for gene

Frequency (genes)

First base

(Any)

(b) Histogram showing the distribution of genes according to their maxi-

mum on target score calculated with the Doench et al. algorithm. Both

the distribution allowing any starting base, and with free choice of start-

ing base are shown.

Figure 62: Distribution of predicted CRISPR on-target cutting scores compared to

mammalian systems and analysed on a per-gene basis indicates a large

potential for CRISPR in P. knowlesi.

In order to consider this I selected the highest scoring guide RNA

within each gene and then considered the distribution of genes (Fig. 62).

This revealed that the vast majority of genes had at least one high-

scoring guide. The use of the U6 promoter requires that transcription

begin, as in the native U6, with a guanine residue (although it is pos-

sible to add a non-binding G to the 5’ end of the sequence if required),

imposing this demand reduced the number of genes with high scores

but still left 88% of genes with a maximum score greater than 0.6 and

47% with a score greater than 0.8.

In subsequent work I prioritised guide RNAs which received scores

greater than 0.8: 80% of these should be in the top quintile

for activ-

5 A quintile is a 20% proportion of ranked data, i.e. 20 percentiles

4.7 generation of cas9 mother vectors 143

ity and the remainder in the second and third quintile according to

the analysis in M. musculus.

4.7 generation of cas9 mother vectors

The starting point for the CRISPR constructs was a pDC2-based vec-

tor expressing Cas9, and the sgRNA from a U6 promoter, developed

by Marcus Lee. Since there are no data as to whether the P. falciparum

U6 promoter is functional when used heterologously in P. knowlesi, I

exchanged this promoter for the corresponding 500 bp sequence in

P. knowlesi. A PCR reaction was used to amplify the P. knowlesi U6

promoter with overlaps corresponding to the insertion site in the des-

tination vector (primers: TPR458, TPR459). The P. falciparum construct

was digested with SalI and BbsI to remove the P. falciparum U6 pro-

moter and the PCR amplicon assembled into the backbone by Gibson

Assembly.

Two versions of the construct were produced, one with yDHODH

as the resistance marker (based on a construct from Zenon Zenonos)

and one with hDHFR. A map is shown in Fig. S2.

These constructs contain sites for the type IIS restriction enzyme

BbsI which allow the precise insertion of guide-RNA-encoding DNA

sequences, which are produced by annealing two oligos with 5’ over-

hangs. Cas9-guide RNA constructs were made for use in tandem with

each of the knock-out constructs described in this chapter, by anneal-

ing desalted oligos and ligating with T4 ligase into the BbsI digested

vector.

4.8 enhancing pjazz transfections with crispr/cas9

An initial proof of concept experiment was carried out with the pJAZZ

vector well-established to target PKNH_0941100. This vector, which

was used several times in section 4.5 to successfully delete this gene

was transfected alone, and also with a cassette containing Cas9 and

a guide RNA targetted to PKNH_0941100 under the P. knowlesi U6

promoter. The version of the construct used yDHODH rather than

hDHFR, and no selection was applied with Dsm1 for the presence

144 the transfer of plasmogem technology to P. KN OWLESI

of the Crispr cassette, in part because of the potential problems with

this marker in P. knowlesi outlined earlier in the chapter.

transfection pjazz / ug crispr / ug

pJazz alone 15 0

pJazz & Crispr 15 30

Selection was applied with 100nM pyrimethamine on day 2 post-

transfection. Parasitaemia rapidly decreased upon selection but para-

sites were apparent in both transfections by day 10 post-transfection.

On day 12 after transfection parasitaemias were as follows:

transfection day 12 parasitaemia

pJazz alone 0.28%

pJazz & Crispr 1.2%

DNA was extracted and used for qPCR analysis to detect the level

of ablation of wildtype locus. The results are shown in Fig. 64.

The fact that parasites came up more rapidly, and with lower levels

of intact locus, compared to a typical pJAZZ transfection provides en-

couragement that CRISPR/Cas9 technology may substantially boost

both the throughput and the targetability of P. knowlesi genes by

pJAZZ vectors. However further experiments are needed to prove the

reproducible effect of the technology in this system.

Experiments conducted to attempt to select for the Cas9 vector

rather than an integrated resistance marker, and thus prove the se-

lective force of Cas9-mediated DNA cleavage have not yet been suc-

cessful.

4.8 enhancing pjazz transfections with crispr/cas9 145

Crispr binding site

PKNH_041100 (DnaJ protein)

500

1000

Figure 63: Position of Crispr/Cas9 cutting site within PKNH_0941100, a

gene already known to tolerate deletion in P. knowlesi from work

earlier in this chapter.

PKNH_0941100 amplicon

Plasmo1/Plasmo2

25%

50%

75%

100%

Wild−type

pJazz alone

pJazz + Crispr

Wild−type

pJazz alone

pJazz + Crispr

Sample

Locus presence

Figure 64: qPCR data from a single experiment demonstrates lower levels of wild-

type locus in a pJAZZ transfection when supplemented by Crispr

Results are normalised for primer efﬁciency and absolute DNA

content. Error bars represent 95% conﬁdence intervals calculated

from technical triplicates from qPCR rater than biological repli-

cates)

146 the transfer of plasmogem technology to P. KN OWLESI

Parasite without

DNA

Parasite with

unintegrated

linear DNA

Parasite with

recircularised

episomal DNA

Dead parasite

Parasite with

integration

Transfection

Integration

Selection

Circularisation

Unstable

segregation

Replication

Replication Replication

Replication

Figure 65: The possible paths parasites can take after transfection are many and

various

While use of linear DNA is sometimes thought of as simplifying

the process of transfection there are still many possible avenues

for transfectant parasites.

4.9 discussion

After efforts to optimise the culture, transfection and selection of P.

knowlesi, it is clear from these results that linear pJAZZ vectors can

be integrated into the P. knowlesi genome, creating desired targeting

events. On the other hand, it is evident that an alternative event is also

possible whereby vector DNA is carried without correct integration

into the genome.

I have not established what the ratio between these events is, nor

the order in which they happen. One possible explanation for P. knowlesi’s

ability to be transfected with linear DNA is that it may recircularise

this linear DNA rapidly, allowing it to carry and replicate this until

a spontaneous DSB in the target gene creates the right conditions for

integration.

A summary of the possible dynamics that can occur during the

process of transfection and selection is shown in Fig. 65, but the pro-

portion of parasites making each transitition is unclear. Unstable seg-

regation of episomes pose a signiﬁcant ﬁtness costs on the parasite.

[147] So one model that ﬁts the data is that in each transfection both

episomes and integrants are formed and these then compete to take

4.9 discussion 147

over the population. Where integrants have wild-type ﬁtness (as in

the HA-tags and with blood stage redundant genes), they take over

the population. But if they have a signiﬁcant ﬁtness cost, this may out-

weigh the weaknesses in episomal segregation resulting in episomes

taking over the population. Presumably here integrants will still be

formed continually at a low level but they will never take over the

population.

The ability to efﬁciently epitope tag proteins of interest, combined

with P. knowlesi’s large merozoite size, may make it a powerful system

for the localisation of invasion-related proteins.

In short the ability to transfect with linear DNA is not, in P. knowlesi,

a silver bullet which ensures that only parasites bearing integrations

will come up, but genetic modiﬁcations with pJAZZ vectors are cer-

tainly possible in this parasite.

4.9.1 Prospects for barseq

Where does this leave the idea of barcode sequencing in P. knowlesi?

One important discovery is that barcodes can be carried by P. knowlesi

parasites without the gene they target being knocked out. Clearly it

means that at present barseq will be somewhat more cumbersome in

this species – transfections would have to be carried out for each vec-

tor individually, assessed for successful modiﬁcation by qPCR, and

possibly dilution cloned before being pooled for a barseq experiment.

However none of these challenges are insurmountable in an in vitro

system and such an approach would allow complete conﬁdence in a

gene called as "redundant", avoiding any false positives due to locus

duplication.

On the other hand, it is possible that because of the formation of

episomes we will be unable to obtain populations of parasites with

signiﬁcant ﬁtness defects, these having been outcompeted by their

episomal counterparts. Without mutants with a range of ﬁtness there

would be no differences for barseq to detect.

One key question will be whether technological changes allow the

rate of false-positive formation to be reduced.

148 the transfer of plasmogem technology to P. KN OWLESI

4.9.2 Possible optimisations to increase proportion of parasites with a tar-

geted genome

4.9.2.1 Complete digestion of pJAZZ arms

In both the episomes I sequenced the pJAZZ long arm was present.

This is only circumstantial evidence that the presence of the arms

prevents integration or promotes recircularisation. It is possible that

it just so happens that a majority of the DNA transfected was undi-

gested and that recircularisation would have occurred regardless. How-

ever it is conceivable that repetitive sequences within the pJAZZ telom-

eres might promote circularisation, or that chew back by exonucle-

ases is important to allow integration and that the pJAZZ arms pre-

vent this. Using smaller amounts of DNA might allow digestion of a

greater proportion of vectors and promote integration.

4.9.2.2 Reducing initial population of parasites

I showed in my second round of transfections that dropping the DNA

concentration roughly tenfold resulted in false-positive living para-

sites. Ironically this may mean that all my efforts to optimise transfec-

tion served to increase the rate of false-negatives. Provided that the

rate of integration of linear DNA is higher than the rate at which is

recircuarises, it might be possible to obtain clean populations of mod-

iﬁed parasites by tweaking the transfection efﬁency such that zero

recircularisation events occur. However achieving this balance for ev-

ery vector in every transfection is likely to be difﬁcult.

4.9.2.3 Negative selection against episomes

In the episomes I sequenced, the pJAZZ long arm was present. Had

this backbone contained a negative selectable marker I would have

been able to kill parasites carrying the episome and thus select only

integrants. Negative selectable markers can undergo a variety of mu-

tations to become non-functional, however, which might undermine

such a strategy.

4.9 discussion 149

4.9.2.4 CRISPR/Cas9

Finally CRISPR/Cas9, which has only been brought to Plasmodium

very recently, might be able to increase the rate at which exogenous

DNA, whether linear or circular, is incorporated into the P. knowlesi.

This might allow increased transfection efﬁciency without a concomi-

tant increase in mistargeting events.

My preliminary data suggests that inclusion of Cas9 increases the

efﬁciency of correct targeting, at least at the one locus tested.

4.9.3 Conclusion

As has been previously observed, we found that human-adapted P.

knowlesi offers a high transfection efﬁciency and can be transfected

with linear DNA. This has enabled the use of this system to bring

PlasmoGEM vectors to an in vitro system for the ﬁrst time.

The high-efﬁciency of endogenous tagging by P. knowlesi vectors

suggests an immediate application for protein localisation. The large

size of merozoites in P. knowlesi might also allow increased through-

put of localisation experiments with better resolution of merozoite

substructure than those in other species.

The use of PlasmoGEM vectors in P. knowlesi enabled tagging of ev-

ery gene attempted, and a successful knock-out of a putative invasion-

related gene previously refractory to deletion in P. falciparum. How-

ever given the observation of parasites carrying barcodes even after

the targeting of essential genes, direct use of the P. berghei Plasmo-

GEM strategy of pooled transfection is not appropriate for P. knowlesi

and more development will be needed to establish a combined qPCR-

barseq assay which checks for targetability prior to pooling barcoded

vectors and tracking their growth through time.

More development is needed to fully realise the potential of Plas-

moGEM genetics in P. knowlesi, but the parasite is already a useful

model for many applications.

N E W A P P R O A C H E S A N D TO O L S F O R L A R G E

S C A L E P H E N O T Y P I C A N A LY S I S I N S I L I C O

5.1 barseq in p. berghei

Chapter 3 of this thesis described the largest screen to date of pu-

tative invasion-related genes, in Plasmodium berghei. The intervening

pages have described progress made towards extending this high-

throughput approach to an in vitro system in Plasmodium knowlesi.

But to fully leverage the large datasets that result from such analyses,

new approaches and tools are needed for analysis of the data pro-

duced at scale. This chapter will outline my approaches to some of

these challenges, using the P. berghei invadome dataset and curation

of the invasion literature.

5.2 network-based analysis of the invadome

I have already discussed the individual gene phenotypes for the barseq

dataset I derived in Chapter 3 but, since the ’core invadome’ was de-

rived from an interaction network based on gene expression patterns,

I also wanted to investigate whether combining this interaction data

with the mutant phenotype data would result in a more detailed pic-

ture of the genetic architecture underlying invasion. This is especially

important given how many of these genes have no annotated func-

tion.

5.2.1 Reconstruction of PlasmoINT network

The ﬁrst pre-requisite for such an analysis was to re-build the Plas-

moINT network in a form in which I could analyse it. I queried

the PlasmoINT database [92] for the connections and interaction co-

efﬁcients for every P. falciparum gene and stored this dataset in a tabu-

151

152 new approaches and tools for large scale phenotypic analysis IN SILIC O

lar form, listing the two gene nodes, and the likelihood score of their

being some interaction between them according to the database.

40000

80000

120000

1 2 3 4 5

Log10(Likelihood Score)

Number of edges

(a) Subsetting of 90% conﬁdence-

network (blue) from overall net-

work (grey).

100

200

300

400

0 250 500 750

Connections / gene

Frequency

(b) Connectivity of 90% conﬁdence net-

work.

Figure 66: Selecting the 90% conﬁdence subnetwork results in a network with rel-

atively low connectivity for most genes.

These data comprise the ‘edge-list’ for a network with 410,898 con-

nections (edges) linking a total of 4,655 genes (nodes) each with a like-

lihood score reﬂecting the degree of conﬁdence that there is some link

between these proteins. In order to increase stringency I also isolated

the subset of these edges with likelihood-scores sufﬁcient to qualify

for the 90% conﬁdence network described in Hu et al. (Fig. 66b). This

reduced the number of edges to 145,622 and the number of genes in-

cluded to 3,466. Because the ’invadome’ I analysed was based on this

subset of the network, however, all 418 genes in it are still included.

In order to verify that I had recapitulated the network in a way

that was true to Hu et al., I examined the entire 90% conﬁdence

network with the Gephi graph-visualisation package. The network

was initiated with nodes placed in random positions and then the

Fruchterman-Reingold algorithm used to iteratively reposition nodes

until the energy of the graph was minimised.

This approach is a form of force-directed graph drawing. Nodes are

initially randomly positioned in a compact space on a two-dimensional

plane. It is assumed that each node repels all other nodes with a cer-

tain force, but interactions between nodes are represented as ’springs’

which pull these nodes together. The position of nodes is then itera-

tively simulated and they expand out from the compact space, even-

tually settling in a pattern which best satisﬁed (at least as a local

maximum) the conﬂicting demands of the attracting and repelling

forces.

5.2 network-based analysis of the invadome 153

Invadome

Non-invadome

Figure 67: Invasion sub-network in context of entire 90% conﬁdence network

Fruchterman-Reingold transformed 90% conﬁdence network

from Hu et al. with the invadome highlighted in blue. The cluster-

ing of the invadome illustrates the network has been successfully

reconstructed.

When this graph was visualised the invadome did indeed cluster

well (Fig. 67), which suggested the network had been successfully re-

built. However these visualisation techniques project hundreds, or

thousands, of dimensions onto a two-dimensional plane and thus

there are inevitable compromises made in placement. This explains

the scattering of non-invadome nodes within the invadome, and should

remain in mind throughout this chapter.

5.2.2 Overlaying localisation data on the PlasmoINT network

Now conﬁdent that the network had been reproduced, I next wanted

to look within the network of the invadome module itself. Hu et al.

conducted a series of analyses to show that the genes in this set were

enriched for invasion-related localisations, but I wanted to investigate

whether there was sub-structure within the data. I therefore selected

only the subnetwork containing genes which were part of the 418

gene invadome and repeated the Fruchterman-Reingold layout algo-

rithm. In order to visualise potential substructure I downloaded local-

isation data from my PhenoPlasm database (which will be described

in detail later) and encoded the localisation of the network genes by

colour.

154 new approaches and tools for large scale phenotypic analysis IN SILIC O

Figure 68: Separation by localisation in the invadome subnetwork

In this ﬁgure the 418 genes of the Hu et al. invadome are laid out

using the 90% conﬁdence PlasmoINT network. They have been

coloured according to their localisation.

This analysis, an example of which is shown in Fig. 68, seems to

show clustering by localisation in that the secretory organelles, the

rhoptries and the micronemes, are relatively well separated. However

there is intermingling with other invasive localisations such as the

inner-membrane complex.

One interestesting node here is PF3D7_1143200, the DnaJ protein

whose orthologue I knocked out in P. knowlesi in Chapter 4, which

in P. falciparum had contradictory localisation signals implicating it

as being either either in the rhoptries or the micronemes [89]. This

gene is positioned in the network in a fashion highly suggestive of

rhoptry localisation. Furthermore, when we examine the individual

connections that this protein makes, 8 are to rhoptry proteins and

0 to micronemal proteins. These observations are an example of the

power that pooling these data sources can provide.

5.3 large scale analysis of mutant phenotypes

Now with access to both a large set of phenotyping data, and some

of the network dynamics that may underlie it, I investigated three

broad hypotheses about the structure of the phenotype dataset: the

centrality-lethality rule, the correlation of phenotypes for co-expressed

genes and the connectedness between the relationship between the

phenotypes of regulator knock-outs and those of their targets.

5.3 large scale analysis of mutant phenotypes 155

5.3.0.1 Centrality-lethality

The centrality-lethality rule [98] states that the more interactions a

protein makes with other proteins, the more likely it is to be essential.

It was proposed on the basis of analysis of S. cerevisiae interaction

networks and has since been found to apply also to E. coli and H.

sapiens. [106].

Possessing a large dataset of both phenotypes and putative inter-

actions I decided to investigate whether I could detect evidence for

centrality-lethality in Plasmodium. The phenotype data is from P. berghei

and the interaction data from P. falciparum but as I am investigating

core conserved invasion genes I expected each datum to be represen-

tative of both species. I compared the connectivity (i.e. the number

of connected nodes) in the 90% conﬁdence PlasmoINT data of es-

sential and non-essential genes as deﬁned by the invadome barseq

data (Fig. 69). There is no signiﬁcant difference, on average, between

the connectivity of essential and non-essential genes. However when

one looks at the 6 most-connected genes (circled in the ﬁgure) it is

striking that every single one is essential, and this is unlikely to be a

coincidence.



100

200

300

400

Essential Non−essential

Mutant phenotype

Connectivity

Essential

Non−essential

Figure 69: Testing centrality-lethality in Plasmodium invasion genes

While there is no signiﬁcant difference in the mean connectivity

of essential and non-essential genes it is noteable that the 6 most

connected genes are all essential.

This analysis does not provide strong statistical evidence that the

central-lethality rule applies across the board in Plasmodium, but it

has been noted previously that high-quality protein-protein interac-

tion data is needed to detect the relatively weak effect of centrality-

156 new approaches and tools for large scale phenotypic analysis IN SILIC O

lethality – no such data currently exist for Plasmodium. The hints of

centrality-lethality in our data, as well as the widespread nature of

the phenomenon is in other organisms, suggest that the lack of ev-

idence in this dataset may be a matter of insufﬁcient power rather

than a feature of Plasmodium biology.

5.3.1 Correlated mutant phenotypes for ’interacting’ genes

In Fig. 68 we saw that the invadome subnetwork is structured, in

part, according to proteins’ subcellular localisation. Now that I had

brought the knock-out phenotypes of genes into the model I sought

to investigate whether there might also be structure on the basis of

the PlasmoINT network.

There are a priori reasons to believe that such structure might ex-

ist. We know that in P. falciparum there are redundant and essential

pathways within the invasion process. If modules like these appear,

in some form, in the network then we might see structure.

For this analysis, I selected a further sub-network from the Plas-

moINT network, selecting only those genes in the 320 gene core in-

vadome. I then laid out this network with the ForceAtlas algorithm

– which adjusts results for the connectivity of a node – (though sim-

ilar results are obtained with Fruchterman-Reingold) and this time

encoded mutant phenotype by colour.

The result (Fig. 70a and Fig. 71) did appear to have structure, with

essential genes seemingly enriched at one ’end’ and redundant genes

at the other. Since humans are apt to spot apparent patterns in ran-

dom data (’apophenia’) [66]. I wanted some measure of signﬁcance,

and so placed a line of division where phenotypes appeared to be best

separated. This yielded the split in phenotypes shown in Fig. 70b. The

enrichment for essential genes in the upper portion is signiﬁcant by

the hypergeometric distribution with an initial p-value of 6.2 ˆ 10

´5

However since the data were partitioned at the most signiﬁcant point,

correcting for multiple comparisons is necessary. Methodology for

this situation is unclear but given the low p-value, a conservative Bon-

feronni correction would permit me to draw 800 such lines and test

each by the hypergeometric distribution to retain experiment-wide

5.3 large scale analysis of mutant phenotypes 157

Redundant

Slow

Essential

Unknown

Mutant phenotype

(a) Core invadome subnetwork with barseq mutant phenotypes indicated by colour.

Above ‘the line’

Below ‘the line’

(b) All phenotyped genes above and below the line marked in panel A, laid out side

by side to enable visualisation of the magnitude of the difference in ratios.

Figure 70: Analysing the invadome subnetwork according to mutant phenotype

158 new approaches and tools for large scale phenotypic analysis IN SILIC O

Figure 71: The core invadome coloured by P. berghei barseq mutant phenotypes,

and labelled

5.3 large scale analysis of mutant phenotypes 159

Table 10: Contingency table showing network edges grouped by mutant phenotypes

at either node

target

Essential Slow Redundant

source

Essential 1310 269 859

Slow 269 48 165

Redundant 859 165 722

signiﬁcance at p ă 0.05. Therefore we can conclude that the non-

random distribution is indeed signiﬁcant.

However as I have noted previously these visualisations are impre-

cise representations of complex network data in two-dimensions. I

next sought to pull apart the network into its individual nodes and

connections, and observe whether phenotypes here were correlated

based on interaction. My initial approach to this was to select all

edges within the 90% conﬁdence core invadome which connected

any two genes with barseq phenotypes. I then examined these edges

and binned them into categories based on the source node: redun-

dant, essential and slow growth. For each edge in each of these groups

I then examined the other, ‘target’, node. I counted the number of

redundant, essential and slow-growth targets for each of the source

categories, producing the data shown in Table 10.

A χ

test performed on this contingency table gives p “ 0.00063.

This demonstrates that the mutant phenotype of the node at one side

of an edge is not independent of the phenotype at the other. The di-

rection of the differences, and their magnitude and signiﬁcance can

be seen in Fig. 72. There is an assortative trend, as was seen on the net-

work diagram, with redundant nodes tending to be more connected

to redundant nodes and essential nodes tending to be more connected

to essential nodes. (Attenuated-mutant nodes are more likely to be

connected to essential nodes.)

While this lack of independence is highly signiﬁcant, the effect size

is small. Under the null hypothesis 653 redundant–redundant edges

would be expected but we observe 722. This is a discrepancy of only

10%.

However this analysis considered each edge singly. I next consid-

ered each node on the basis of all nodes connected to it. For each

160 new approaches and tools for large scale phenotypic analysis IN SILIC O

−1.8

−1.5

−1.0

0.0

1.0

1.5

2.0

2.7

Pearson

residuals:

Target node

Source node

Redundant Slow Essential

Essential Slow Redundant

Figure 72: Association plot demonstrates that connected nodes tend to have the

same mutant phenotype

This is a Cohen-Friendly association plot demonstrating a lack

of independence between the phenotypes on either end of a Plas-

moINT connection. The width of each cell represents the expected

size of the cell and the height the signed difference between ob-

served and expected. Edges with redundant genes at one end are

more likely to have a redundant gene at the other end also than

would be expected by chance.

node I counted the number of connected essential nodes (e) and the

number of connected redundant nodes (r). I then computed a simple

metric, which I will call the essentiality ratio (R).

R “

e ` r

I calculated R for all genes which were themselves essential or re-

dundant, and excluded genes where e ` r ă 3 since for these the

metric is likely to be very noisy. The mean R for essential and re-

dundant genes was 0.63 and 0.53 respectively, and signiﬁcantly dif-

ferent (p “ 1.871 ˆ 10

´5

). This increased signiﬁcance suggests that

the essentiality-ratio approach adds more power.

I compared the distribution of essentiality ratios for redundant and

essential genes (Fig. 73). The populations of redundant and essential

genes do appear to have offset essentiality ratios, but the differences

appear most profound at the extremes. In the barseq data any gene in

which more than 82% of phenotyped connected genes were essential,

was itself essential (there were 7 such genes) and any gene where less

than 37% of connected genes were essential was redundant (there

were 3 such genes).

I was interested in the possibility that phenotypes might be able

to be predicted for invadome genes for which we do not yet have

an established phenotype, so I examined those genes satisfying these

5.3 large scale analysis of mutant phenotypes 161

limiting criteria (i.e. would not give false-positive results in the ex-

isting data) in the set of genes not yet phenotyped by barseq. This

allowed the prediction of mutant phenotypes for 23 further genes,

which exceeded these limits and had no known barseq phenotype

(Table 11).



 



Unknown

Essential

Redundant

0.00 0.25 0.50 0.75 1.00

Essentiality ratio

Barseq mutant phenotype

Figure 73: Essentiality ratios for invadome genes clustered by barseq mutant phe-

notype.

Two lines are marked, one at the minimum essentiality ratio for

essential genes, the other at the maximum essentiality ratio for re-

dundant genes. These were used for predicting the possible phe-

notypes of unknown genes.

This table also includes any suggested mutant phenotypes from

other previous experimental work, retrieved from the PhenoPlasm

database. Such data available for six genes, and all of these were pre-

dicted as essential. In three cases the genes had proven refractory to

deletion in previous experimental techniques, but in the remaining

cases there were reports of successful deletion.

This limited experimental data shows that this approach has a sig-

niﬁcant number of false predictions. But I would hope, based on the

profound differences seen in the distribution, that a larger screen of

these predicted genes would demonstrate this approach to have at

least some predictive power. As more data arrives, from future barseq

experiments and more conventional approaches, these issues will be-

come clearer.

Regardless of whether it is useful for prediction, we have clearly

shown than in this subset of the invadome there is correlation of

phenotypes between putatively interacting genes.

162 new approaches and tools for large scale phenotypic analysis IN SILIC O

Table 11: Mutant phenotypes for invasion genes predicted based on their essential-

ity ratios.

Gene Product Prediction Experimental

PBANKA_010990 cons. Plasmodium protein, u. f. Essential -

PBANKA_052030 cons. Plasmodium protein, u. f. Redundant -

PBANKA_060060 NIMA related kinase 3, putative

(NEK3)

Essential Essential

PBANKA_060500 rhodanese like protein, putative Redundant -

PBANKA_070180 cons. Plasmodium protein, u. f. Essential -

PBANKA_071440 cons. Plasmodium protein, u. f. Essential -

PBANKA_080940 P1 nuclease, putative Essential -

PBANKA_081900 secreted acid phosphatase,

putative,glideosome-associated

protein 50, putative (GAP50)

Essential Essential

PBANKA_082810 fumarate hydratase, putative Essential Essential

PBANKA_083270 adenylate kinase-like protein 2,

putative (AKLP2)

Redundant -

PBANKA_090580 DnaJ protein, putative Essential Redundant

PBANKA_092700 tyrosine kinase-like protein, puta-

tive (TKL2)

Essential Redundant

PBANKA_100660 cons. Plasmodium protein, u. f. Essential -

PBANKA_123300 SWIB/MDM2 domain-

containing protein, putative

Essential -

PBANKA_124010 cons. Plasmodium protein, u. f. Redundant -

PBANKA_131330 cons. Plasmodium protein, u. f. Redundant -

PBANKA_132370 FYVE and coiled-coil domain-

containing protein, putative

(FCP)

Redundant -

PBANKA_133170 zinc ﬁnger protein, putative Redundant -

PBANKA_134070 mitochondrial ﬁssion 1 protein,

putative (FIS1)

Essential -

PBANKA_140640 tripartite motif protein, putative Essential -

PBANKA_143900 cons. Plasmodium protein, u. f. Essential -

PBANKA_144330 merozoite surface protein 9, puta-

tive

Essential Redundant

PBANKA_144950 cons. Plasmodium protein, u. f. Essential -

PBANKA_146310 cons. Plasmodium protein, u. f. Redundant

5.3 large scale analysis of mutant phenotypes 163

5.3.2 Investigating the connection between a regulator’s phenotype and

those of its targets

Among the invasion network are four ApiAP2 genes, putative tran-

scription factors in Plasmodium. For one of these, AP2-O, in Chapter 3

I detected a viable knock-out in the blood stage. This viability has

already been well described but I found that the mutants appeared

to grow with reduced ﬁtness in the blood stage. As I discussed there,

this is a controversial result given the primary role of this gene is in

the ookinete. I wanted to investigate these dynamics further at scale

and see whether a causative downstream gene could be identiﬁed.

In a simplistic model in which each AP2 gene acts alone to activate

downstream genes in an all-or-nothing fashion, it would logically fol-

low that it must be possible to knock out every gene targeted by AP2-

O to generate viable parasites, since the effect of an AP2-O deletion

would be to effectively knock-out each of these genes.

My ﬁrst attempt to test this idea was to download the targets of

AP2-O, suggested recently on the basis of ChIPSeq [103]. When I

analysed those genes in this set for which I had barseq-phenotyping

data, there was a trend towards increased numbers of viable knock-

outs in the targets of AP2-O compared to the background but the

effect was not statistically signiﬁcant.

However, the model given above is overly simplistic. In fact some

genes regulated by AP2-O might have their expression only subtly al-

tered by its binding, and what effect this binding has on transcription

may depend on other transcription factors and on the regulatory ele-

ments of the gene in question – some of these genes might be activated

by AP2-O binding.

I therefore chose to consider data which recorded the actual ef-

fect of the knock out on transcription. I analysed ookinete microarray

data from [103]. This comprised microarrays from 7 AP2-O knockout

ookinete cultures and from 5 wild-type controls. I used the Gene Ex-

pression Omnibus to calculate the log fold change for each probe, and

then performed further analysis in R. Where multiple probes covered

the same gene the median of their log fold-changes were taken. These

values were compared to the mutant ﬁtness (in the blood stage) of the

knock-out of the individual gene by barseq. This allowed me to ex-

164 new approaches and tools for large scale phenotypic analysis IN SILIC O

amine mutant ﬁtnesses in the context of the degree of knockdown of

their gene in the AP2-O knock-out (see top panel of Fig. 74).

In this analysis it stood out that none of the genes strongly affected

by AP2-O knock down were essential in the blood stage, which is the

result one would predict from the AP2-O mutant’s viable phenotype.

I decided to extend this analysis to include targetability data available

from RMGMdb also (see middle panel) which makes the same case.

Unknown

PKB

Figure 74: Degree of knock-down in AP2-O knockout (as published in [103]) com-

pared to the same gene’s individual mutant phenotype

All severely knocked-down genes are non-essential, consistent

with the fact that the AP2-O mutant is viable in the blood stage.

Protein kinase B is an essential gene which shows two-fold down-

regulation in an AP2-O knockout and so may have a role in ex-

plaining the apparent slow growth of these parasites.

This argument may seem self-fulﬁlling – one could argue that we

already knew these genes were non-essential when we saw their tran-

script level reduced in the knock-out parasites. But this analysis re-

mains useful. The microarray data on the knock-out is from ookinetes,

not from the blood stages. This analysis could have proved that AP2-

O must act differently in the blood stage, compared to the ookinete,

if it was found that there are genes essential in the blood stage which

are highly activated by AP2-O in the ookinete. Since this was not

found, the idea that the targets of AP2-O in the blood stage are the

same as those in the ookinete remains a possibility (albeit unproven).

This analysis was also useful for an attempt to investigate the rea-

son for apparent attenuation in the AP2-O knockout. Had I identiﬁed

a single gene which had signiﬁcantly reduced expression in the AP2-

O knockout and gave slow-growth in the blood stage when mutated,

5.4 building the infrastructure to collate reverse genetic data across species 165

this gene alone would then have been an explanation for the mecha-

nism of attenuation of my apparent slow growth in AP2-O parasites.

We did not observe such a gene. It could yet exist, in the signiﬁcant

portion of genes we have yet to probe by barseq. But alternatively it

is possible that though individual knock-outs of genes whose expres-

sion is strongly governed by AP2-O have no effect in the blood stage,

the loss of all these genes at once has a measurable effect – either

through epistasis effects or just the additive inﬂuence of many very

small growth defects. A ﬁnal possibility is that the small effect of AP2-

O knock-out on expression of essential genes may be the root of the

effect. Protein kinase B, an essential protein known to be involved in

invasion, is two-fold down regulated in the AP2-O knock out and so

could represent the causative gene.

I think that this approach of bringing together large scale regula-

tory and ﬁtness datasets will add to our understanding of both com-

ponents. For example we might be able to infer the likely blood-stage

non-essentiality for the other genes heavily downregulated in the

AP2-O knock-out for which we have no phenotyping data at present

(see bottom panel of Fig. 74).

5.4 building the infrastructure to collate reverse ge-

netic data across species

In the introduction I cited an estimate from [48] that approximately

500 genes have been successfully targeted in malaria parasites. This

is an estimate because there is no database that compiles attempts to

knock out Plasmodium genes, other than the excellent Rodent Malaria

genetically modiﬁed Parasites database (RMgmDB) [105] which applies

only to the rodent parasites. The result is that when researchers be-

gin to examine a large set of genes, as I wanted to do for the in-

vadome, they must trawl through the literature to uncover attempts

to disrupt these genes. If a minute is spent researching each gene

of the invadome this adds up to 7 hours of intensive research, and

even then only one species has been investigated. There was a clear

need for a database that allowed an instant view of the transfection

attempts carried out for a large set of genes, in any of the Plasmodium

166 new approaches and tools for large scale phenotypic analysis IN SILIC O

Figure 75: PhenoPlasm database schema

The database is designed to allow painless migration to new gene

models in subsequent genome releases

genetic systems: P. falciparum, P. berghei, P. chaubaudi, P. yoelii and P.

knowlesi.

Additionally, since the ApiLoc database ceased to be curated there

is no up to date database of localisations for Plasmodium genes.

Therefore, before embarking on a comprehensive search of the in-

vasion literature for previously described mutant phenotypes, I built

a database in which this information could be deposited in a machine-

readable form, so that my data can be accessed by others and they in

turn can if they wish contribute such data for other genes.

5.4.1 Technology

I used MySQL to construct an online database to hold both phenotyp-

ing and localisation data, which I call PhenoPlasm (http://phenoplasm.org).

An empty database of genes and their orthological relationships was

set up on the basis of data from PlasmoDB. The structure of the tables

is as shown in Fig. 75.

I designed the system to be backwards compatible with the for-

malised description of phenotypes established by RMGMdb. This

meant that I could initially populate it with the data already avail-

able in that database.

5.4 building the infrastructure to collate reverse genetic data across species 167

Figure 76: The PhenoPlasm database allows queries of all experimental genetic data

on a gene to be assessed quickly

This picture depicts a search for some members of the invadome.

This is a search for P. falciparum genes and so opaque icons rep-

resent P. falciparum phenotypes. Semi-transparent icons represent

the phenotypes of 1:1 orthologues in another species, in these

cases P. berghei. In this case we see that a DnaJ protein was refrac-

tory to deletion in P. falciparum but targetable in P. berghei with no

detectable phenotype in the blood stage.

A PHP/XHTML front-end was then built in the Bootstrap frame-

work to present this data in an easily comprehensible way. Screen-

shots of the user interface, which has been designed to show only

speciﬁc types of data in a minimalist format, are shown in Figs. 76 –

78.

5.4.2 Key features

There are a number of key features that make this database different

to previous endeavours.

orthology The database is constructed such that searches for genes return

both phenotypes in the species requested, and phenotypes in

1:1 orthologous genes from any other species. It is the only tool

at present which allows a potentially comprehensive view of

reverse genetic data for a set of genes across Plasmodium. This

allows even researchers of the genetically intractable P. vivax

to enter gene IDs from this species and receive the results of

experiments in P. falciparum, P. berghei and P. knowlesi.

168 new approaches and tools for large scale phenotypic analysis IN SILIC O

Figure 77: Single-gene view on PhenoPlasm

This allows the source of a claim to be checked for citations and

veriﬁcation.

Figure 78: Interface for data addition to PhenoPlasm

5.4 building the infrastructure to collate reverse genetic data across species 169

speed In the world of -omics there are many occasions where pheno-

types for a large set of genes will need to be queried. RMGMdb

becomes unwieldy when more than ~500 genes are queried, tak-

ing minutes to load and sometimes timing out. The database

structure of PhenoPlasm is organised such that even phenotyp-

ing data for the ~5,000 genes in an entire genome takes less than

1.5 seconds to return.

crowd-sourcing To facilitate the capture of large amounts of data the site allows

any user to instantaneously add phenotyping data for a mutant.

The wisdom or otherwise of such an approach will only become

apparent with time, but I hope the ease with which mutants can

be added will help the database to keep up to date.

referencing Like an encyclopedia, the database does not aim to be a pri-

mary datasource but a collation of references to others. This

means that data such as blot images is not currently included,

as it is on RMGMdb. It is expected that this data will be pre-

sented in a publication and referenced there. This is a necessary

consequence of a non-curatorial approach. Every phenotype is

referenced with a PubMed ID or other hyperlink where this is

unavailable (e.g. conference abstracts).

5.4.3 A systematic review of experimental genetics in Plasmodium inva-

sion

The absence of semantic digital identiﬁers for genes referred to in

papers makes ﬁnding every reference to attempts to mutate a gene

potentially challenging. I constructed a feature within PhenoPlasm to

search Google Scholar (which contains full-text from many journals)

for both the gene ID of a particular gene, and also every previous

iteration of that name (PFDXXXXC, MAL13P1.XXX, etc.)

Using this feature I searched for references to each of the 418 genes

listed in the Hu et al. invadome. Where a gene also had a common

name I additionally searched for this name to include publications

not providing a full gene ID.

Of the 418 genes in the invasion-network list, for 74 were reported

knock-out attempts were found. Exactly 50% of these were successful,

with the remainder refractory to deletion. There was no data in P.

170 new approaches and tools for large scale phenotypic analysis IN SILIC O

falciparum for the remaining 82% of genes, but some of these had

data available from orthologues imported from RMGMdb.

I also examined the immunoﬂuorescence and immuno-EM images

available at Malaria Metabolic Pathways to ﬁnd localisations for this

list of proteins and where I could see a clear organellar localisation I

added this information to the database.

5.4.4 Database statistics

I have since supplemented the database with additional knock-out

attempts reported in the literature and at scientiﬁc conferences. As of

September 2015, PhenoPlasm contains records of 164 genes for which

attempts to disrupt have been reported in P. falciparum, all manually

curated from the literature into a systematic machine-readable form.

(78 of these were successfully disrupted, 86 are described as refractory

to deletion). I believe that this includes the vast majority of genes tar-

geted to date. There are an additional 457 genes for which phenotypes

have been reported when data from P. berghei (provided by RMGMdb)

and P. knowlesi is included.

5.5 applying phenoplasm data to invasion

The PhenoPlasm database has applications to every area of parasite

genetics, but we can illustrate some of these with the example of inva-

sion. The core invadome presented already with barcode sequencing

data, can be supplemented by PhenoPlasm data generated from pre-

vious single transfection experiments as seen in Fig. 79. This is useful

to ﬁll in parts of the network which are not covered by pJazz vectors

at present. A caveat is that where such data comes from P. falciparum

we can be less conﬁdent that a gene refractory to deletion is in fact

essential, given the greater difﬁculty in achieving genetic manipula-

tions in this organism.

One approach to this challenge is to only include data points from

P. falciparum that indicate a successful knock-out has occurred. I used

this approach to merge barseq data with the phenotypes gathered

from the literature in PhenoPlasm and repeated some of the analyses

performed earlier in the chapter to see if this additional data could

5.5 applying phenoplasm data to invasion 171

Source of phenotype

Barseq (Pb)

Curated literature on PhenoPlasm

(includes Pf and Pb)

Mutant phenotype

Redundant

Attenuated

Essential

Figure 79: Supplementing invadome barseq data with PhenoPlasm database

phenotypes

increase the power of the analysis. I found the same direction of corre-

lation in linked nodes (Fig. S3) but with this dataset the overall effect

was not statistically signiﬁcant by χ

test. However an analysis for

the difference in essentiality-ratio between essential and redundant

genes gave a result of p “ 0.04, again illustrating the increased power

of this analysis method. Nevertheless the effect was noticeably less

strong compared to the dataset with barseq data alone.

While this could simply represent regression to the mean, i.e. indi-

cating that the effects we saw in barseq data were pure ﬂuctuations,

the p-value of p “ 1.871 ˆ 10

´5

seen in the barseq data suggests to

me that this is not the case. Rather I think the reduced power in the

subsequent analysis illustrates the difﬁculties in pooling data from

disparate laboratories – many factors such as the selection of these

172 new approaches and tools for large scale phenotypic analysis IN SILIC O

genes for study, and publication bias in favour of successful mutants

may be playing a role in confounding results. These factors make

hypothesis-free screening approaches particularly important. As such

efforts continue on invasion genes it will be important to revisit these

analyses and conﬁrm how well the effects I observed are reproduced.

5.5.1 Non-conserved invasion ligands (and others)

In this thesis I have primarily focused on the core invadome because

it can be investigated using P. berghei and P. knowlesi. Of course the

genes within this core set must interact with rapidly evolving parasite

proteins at the interface with the host in order for invasion to occur. A

complete understanding of invasion requires understanding the role

of these proteins as well. The data so far available for mutants in these

additional genes is shown in Table 12. (This shows the already well-

established phenomenon [207] that the majority of these genes are

dispensable, with the notable exception of the complex comprising

Rh5, CyRPA and RIPR.)

The three CDPKs may seem an odd inclusion in this list: their se-

quences are so similar that they are placed in the same group by Or-

thoMCL and therefore not considered 1:1 orthologs between species.

One possible improvement to the methodology would be to use syn-

teny to assist in the calling of 1:1 orthologs, which I will consider for

future work.

5.6 discussion

5.6.1 Towards a holistic view of invasion

Earlier chapters have introduced two systems by which we may be

able to obtain phenotyping data for large numbers of genes using

high-throughput approaches. In this chapter I have developed a tool

that enables any researcher to retrieve this data, regardless of what

species of malaria they work on, in a matter of seconds. I have also

outlined how new approaches to analysing this data, leveraging in-

formation about protein-protein interactions or transcriptomics, may

allow additional insights to be drawn into parasite biology.

5.6 discussion 173

Table 12: Collated mutant phenotypes for Hu et al. invasion genes which

were not included in the 1:1:1 core invadome so far discussed. For

references please see the relevant page on PhenoPlasm.

Gene ID Description Mutant phenotype

PF3D7_0217500 calcium-dependent protein kinase 1 (CDPK1) Refractory

PF3D7_0717500 calcium-dependent protein kinase 4 (CDPK4) Viable

PF3D7_1337800 calcium-dependent protein kinase 5 (CDPK5) Refractory

PF3D7_0423800 cysteine-rich protective antigen (CyRPA) Refractory

PF3D7_0302500 cytoadherence linked asexual protein 3.1 (CLAG3.1) Viable

PF3D7_0302200 cytoadherence linked asexual protein 3.2 (CLAG3.2) Viable

PF3D7_0935800 cytoadherence linked asexual protein 9 (CLAG9) Viable

PF3D7_1035700 duffy binding-like merozoite surface protein (DBLMSP) Viable

PF3D7_1301600 erythrocyte binding antigen-140 (EBA140) Viable

PF3D7_0424300 erythrocyte binding antigen-165, pseudogene (EBA165) Viable

PF3D7_0731500 erythrocyte binding antigen-175 (EBA175) Viable

PF3D7_0102500 erythrocyte binding antigen-181 (EBA181) Viable

PF3D7_1035300 glutamate-rich protein (GLURP) Viable

PF3D7_1035600 merozoite surface protein (H101) Viable

PF3D7_1036000 merozoite surface protein (MSP11) Viable

PF3D7_1035400 merozoite surface protein 3 (MSP3) Viable

PF3D7_0206900 merozoite surface protein 5 (MSP5) Viable

PF3D7_1035500 merozoite surface protein 6 (MSP6) Viable

PF3D7_1335100 merozoite surface protein 7 (MSP7) Viable

PF3D7_0316000 microneme associated antigen (MA) Refractory

PF3D7_1476300 Plasmodium exported protein (PHISTb), unknown function Viable

PF3D7_0202100 Plasmodium exported protein (PHISTc), unknown func-

tion,liver stage associated protein 2 (LSAP2)

Viable

PF3D7_1335400 reticulocyte binding protein 2 homologue a (RH2a) Viable

PF3D7_1335300 reticulocyte binding protein 2 homologue b (RH2b) Viable

PF3D7_0402300 reticulocyte binding protein homologue 1,normocyte bind-

ing protein 1 (RH1)

Viable

PF3D7_1252400 reticulocyte binding protein homologue 3, pseudogene

(RH3)

Viable

PF3D7_0424200 reticulocyte binding protein homologue 4 (RH4) Viable

PF3D7_0424100 reticulocyte binding protein homologue 5 (RH5) Refractory

PF3D7_0323400 Rh5 interacting protein (RIPR) Refractory

PF3D7_0501500 rhoptry-associated protein 3 (RAP3) Viable

PF3D7_0208000 serine repeat antigen 1 (SERA1) Viable

PF3D7_0207800 serine repeat antigen 3 (SERA3) Viable

PF3D7_0207500 serine repeat antigen 6 (SERA6) Refractory

PF3D7_0802100 transcription factor with AP2 domain(s) (ApiAP2) Refractory

PF3D7_1235200 V-type K+-independent H+-translocating inorganic py-

rophosphatase (VP2)

Viable

174 new approaches and tools for large scale phenotypic analysis IN SILIC O

When the barseq results described in Chapter 3 are added to Pheno-

Plasm, a search for the 418 genes of the P. falciparum invadome will

return 214 results. More than 50% of these result from my barcode

sequencing experiments, the remainder are derived from my survey

of the P. falciparum literature and those imported from the RMGMdb

database. This collection of phenotypes for what may be the majority

of proteins involved in invasion represents an important resource for

researchers studying this crucial part of the parasite lifecycle.

Already in this chapter we have seen the applications that such a

dataset can provide. Based on the localisations recorded in Pheno-

Plasm, and the PlasmoINT network, I was able to suggest that a pro-

tein for which immunoﬂuorescence analysis gave an unclear apical lo-

calisation is very likely to be localised to the rhoptries. As additional

protein-protein interaction resources are developed by tools such as

BioID, and integrated into interaction networks, these approaches are

likely to become more and more powerful.

Having observed an AP2 gene giving an unexpected apparent blood-

stage phenotype, by integrating phenotype data with transcriptomic

data generated (by others) from parasites lacking this gene, I was able

to suggest a possible candidate downstream gene responsible for this

phenotype.

By analysing the substructure within phenotyped genes I was able

to observe some limited evidence for the phenomenon of centrality-

lethality in Plasmodium invasion. This type of analysis also enabled

the observation that a gene’s mutant phenotype is correlated with

that of those which interact with it. This enabled attempts to pre-

dict the phenotypes of genes as yet uncharacterised experimentally,

though the accuracy of these predictions remains unclear.

The range of approaches described above is just a hint of the possi-

bilities that large-scale phenotypic data can provide when made eas-

ily available to the wider research community. I hope that Pheno-

Plasm, and some of the approaches to analysing its data discussed in

this chapter will contribute to the realisation of some of these possi-

bilities.

5.6 discussion 175

5.6.2 Areas for follow-up

5.6.2.1 Protein-protein interaction network

Much of the analysis in this chapter, and this project, relies on an in-

teraction network which is now ﬁve years old. Clearly the quality of

interactions available has major implications for what can be inferred

in these analyses. In the coming years I anticipate a number of pub-

lications using new approaches (such as protein tagging with biotin

ligases or peroxidases), as well as more conventional immunoprecipi-

tation, to enable large numbers of new, high quality, interactions to be

established. These may not be sufﬁcient to generate a genome-wide

interactome but properly integrated into the other data sources used

by Hu et. al they will provide heavily weighted known interactions

that improve the network as a whole.

There are also a number of new larger scale datasets that should be

added to such an analysis. Co-expression in AP2 knockouts (such as

the one discussed earlier in this chapter) may indicate an increased

likelihood of interaction, as should shared presence in the proteome

of a certain parasite stage.

These must all be integrated in the next generation of Plasmodium

interaction networks, and one important result will be an increase in

the proportion of the genome covered in high conﬁdence predictions

above the two-thirds coverage that exists today.

Ultimately such a dataset may produce a network diagram in which

clusters of genes are very well separated by parasite stage, and when

phenotyping information is overlaid on this the sexual and mosquito

stages will light up with the green of blood-stage redundant genes.

5.6.2.2 Improvements to PhenoPlasm

There are a number of additional features that could be added to

PhenoPlasm. The database already collects information on the type of

experimental approach taken to establish a phenotype, but does not

yet allow searching based on this. This will become important as more

and more high-throughput approaches are developed – a user may

want to retrieve only genes where cloned lines have been produced

carrying a knock out, or for other purposes they might be prepared to

consider a gene redundant if dCas9 targeted to its genomic location

176 new approaches and tools for large scale phenotypic analysis IN SILIC O

was tolerated by the parasite. Collecting all this data and then making

it appropriately searchable will be essential.

Part III

D I S C U S S I O N

G E N E R A L D I S C U S S I O N

6.1 achievements

6.1.1 New phenotyping data for putative invasion genes may inform target

selection and models of the invasion process

Invasion genes are key targets for vaccine development, and may also

represent good drug targets. In Chapter 3, 71 likely essential putative

invasion-related genes were identiﬁed. 54 of these were being inves-

tigated for the ﬁrst time in any Plasmodium species, and each one is

now part of a very long list of potential drug/vaccine targets. These

data represent a very small ﬁrst piece in a long checklist of require-

ments for a drug target, but is nonetheless nonetheless an essential

step.

In addition to establishing new potential targets, my data also rules

out some genes which might previously have been seen as candidates.

As a result of the open-ended screening in this study we overturned

phenotypes for 3 genes previously considered essential, including

ROM4 which is one of a class of rhomboid proteases considered

possible therapeutic targets in Apicomplexa. [23] The use of efﬁcient

PlasmoGEM vectors, ﬁrst in barseq experiments and then in single

transfection follow-up demonstrated knock-outs for these genes to

be viable (but in some cases attenuated).

Another key role for the identiﬁcation of phenotypes for uncharac-

terised genes is to limit the search space in investigations which try

to ﬁnd the gene that ﬁlls a particular role in invasion. As a theoretical

example, the identify of the Plasmodium glideosome adhesin connec-

tor is not yet deﬁnitively established. One approach to identify this

would be to pull down proteins binding to parasite actin. However

such a biochemical approach would probably identify a large num-

ber of proteins, many of which would be false positives. Most models

would assume the GAC to be essential for invasion and so the incor-

poration of my phenotyping data would allow any proteins identiﬁed

179

180 general discussion

by biochemical means to be ﬁltered to a more manageable number –

genes with viable knock-outs could be excluded as candidates, and

uncharacterised genes shown by barseq to be essential could be pri-

oritised for further investigation.

6.1.2 A new model system for PlasmoGEM

After a number of experiments to optimise the system for parasite

culture, I have brought the PlasmoGEM approach which has already

made a profound impact on P. berghei genetics to a new species, P.

knowlesi. This new system has signiﬁcant additional beneﬁts. While

the P. berghei system is clearly more efﬁcient, and so has a certain

level of genetic power that cannot yet be matched in P. knowlesi, my

adaptation of PlasmoGEM to P. knowlesi has two key advantages over

P. berghei – a robust in vitro culture system and the presence of hu-

man erythrocytes. The in vitro nature of this system should simplify

tagging experiments, reduce animal usage and enable more detailed

phenotyping experiments (such as discriminating between genes in-

volved in egress and invasion), and the presence of human erythro-

cytes clearly will make some results more clinically relevant to other

human malaria parasites. Finally, the emergence of the threat to hu-

man health posed by P. knowlesi in Southeast Asia, and an increased

appreciation of the potentially severe clinical complications of P. knowlesi

malaria, means that results obtained in this species may have direct

clinical relevance in their own right.

6.1.3 New in silico approaches and a new tool for the experimental genetics

community

Finally I have investigated the patterns within my data at scale. Such

analyses are made difﬁcult by the uncharacterised nature of these pro-

teins, for which there are not existing mapped pathways. However

my use of previous bioinformatically inferred protein-protein interac-

tion has revealed patterns within phenotyping data, and allowed new

insights, such as the suggestion of possible models for the attenuation

observed in an AP2 mutant by barseq.

6.2 screening limitations 181

Finally the development of a database in which the data generated

in P. berghei and P. knowlesi can be deposited, and its supplementa-

tion with a full literature review of the experimental genetics of in-

vasion genes in P. falciparum may prove a valuable resource to other

researchers.

6.2 screening limitations

6.2.1 False negatives may occur in barseq approaches

In any conventional knock-out approach it is very difﬁcult to distin-

ugish between a failure to obtain mutants due to gene essentiality,

and a failure due to an inability of the targeting construct to inte-

grate. These issues are evidenced by the examples of genes refractory

to deletion in previous published experiments where PlasmoGEM

vectors, with their long homology arms, have now enabled disrup-

tion – including three veriﬁed examples described in Chapter 3. Even

with the efﬁciency of PlasmoGEM technology a non-zero number of

our vectors will have insufﬁcient homology to integrate, or contain

spontaneous mutations that disrupt their ability to create resistant

parasites. Thus we cannot be certain that any gene described as es-

sential by barcode sequencing is certainly essential, but we can say

that there is a high probability that the gene will be essential or that

its knockout will result in parasites that are severely attenuated. Val-

idation with conditional technologies, which could also be scaled us-

ing PlasmoGEM and are discussed further below, will be needed to

conﬁrm essentiality.

6.2.2 False positives are a possibility

There is also a potential for false-positive events in barcode-sequencing.

In very rare events, a locus can be spontaneously duplicated (we saw

an example of this in the parental A1.H1 P. knowlesi line in Chapter 4).

In P. berghei such an event has previously allowed an essential gene

to be duplicated so that one copy could be deleted and replaced by

a barcode. [79] The only way to exclude such a result is to follow up

with an experiment to clone out the parasite, and perform a Southern

182 general discussion

blot or qPCR. However the typical success of these follow-ups gives

considerable conﬁdence to an initial ’redundant’ call by barseq.

One advantage of the qPCR approach which I developed as part of

my application of PlasmoGEM to P. knowlesi is that it avoids any pos-

sibility of false-positives, by directly assaying gene deletion, but this

approach cannot be used in pooled transfections. Because P. knowlesi

is an in vitro system the single knock-out approach might be scalable

even to a genome scale, especially with the application of automated

cell culture facilities.

6.2.3 On the meaning of ‘essentiality’

The difference between severe attenuation and ‘essentiality’ (which

technically perhaps means a generational growth rate ă 1) is impossi-

ble to establish by barcode-sequencing. There are two reasons for this.

Firstly essentiality is a relative term, and dependent on context. For

example, the host immune system exerts pressure on the parasite, and

a gene that appears essential in our experiments might be merely at-

tenuated in an immunodeﬁcient mouse. A separate issue is that very

slow-growing parasites are rapidly outcompeted from the population

and may reach levels undetectable by our sequencing methodology,

preventing their measurement and leading them to be called as likely

essential. These effects are not of major concern since merely know-

ing that loss of a gene causes a severe decrease in ﬁtness provides a

great deal of information. But this does make it more difﬁcult to test

models predicting that a particular gene plays an indispensable role

in invasion. Conditional approaches, which I will discuss, are likely

to be the solution.

6.2.4 ..and ‘redundancy’

40% of P. berghei core invadome genes proﬁled were identiﬁed by

barseq as redundant with no statistical evidence of attenuation. It is

tempting to think that these genes are of little importance. However it

is worth considering that estimates put the divergence time between

P. falciparum and its closest relative, P. reichenowi at 5-7 million years

ago. The divergence time between the murine malarias is estimated

6.2 screening limitations 183

at more than 24 million years ago. [175] The split therefore between

P. falciparum and the phylogenetic branch containing P. berghei and P.

knowlesi likely occurred much earlier in evolutionary history. Every

gene in the core invadome has been conserved, with 1:1 orthology,

throughout these many millions of years and this provides strong ev-

idence that any parasites which lost them were at a signiﬁcant ﬁtness

disadvantage in at least some circumstances.

6.2.4.1 Genetic interactions

In some cases the appearance of redundancy will be because the func-

tion of the gene in question gives only a moderate growth advantage -

slightly increasing adhesion to the erythrocyte, for example. In these

cases many such genes would have to be knocked out to create an

additive effect which abolished erythrocyte binding and became es-

sential.

In other cases it may be that the gene is involved in an alternative

pathway, necessitated only in certain conditions (for example the par-

asite encountering a host which has built up an immune response

to one invasion pathway). This would mean that a single knock-out

would have wild-type ﬁtness but that the additional knock-out of a

second ‘gene’ might prevent invasion entirely.

Ultimately these factors mean that to fully dissect the genetics of

parasite machinery it will be necessary to invesigate synthetic lethal-

ity, and synergistic genetic interactions more broadly.

6.2.4.2 Understanding the interplay between invasion and the immune sys-

tem

Invasion is a unique point in the intraerythrocytic cycle where the

parasite is outside a red blood cell and directly accessible to the im-

mune system. This is a powerful selective force which has doubtless

had profound effects on parasite evolution.

It is likely that many merozoite proteins may have functionality

that becomes apparent only in the presence of a functional immune

system. In the P. knowlesi in vitro system, there is of course no immune

system, and complement present in the serum is destroyed by heat-

inactivation, so any such effects would not be observed.

184 general discussion

While the Theiler’s Original mice used in my P. berghei ANKA ex-

periments all possessed an immune system, they are unable to control

P. berghei infections and would die if infections were allowed to con-

tinue. This is unlikely to reﬂect the natural African setting for the

parasite where chronic infections are presumably favoured for trans-

mission. The clear failure of the mouse to clear the parasite may be

indicative of a limitation of even this model to identify genetic effects

which result from interactions with the immune system.

In addition some aspects of immune evasion would not be ex-

pected to give cell-autonomous genetic effects (e.g. those involving

immunomodulation). If a mutant’s defect can be rescued by the pres-

ence of parasites of a different genotype in a co-infection then this

defect will not be visible in a barcode-sequencing approach.

6.2.5 Barseq in P. berghei fails to directly assay invasion

Balu et al. [15] wrote that:

’Malaria parasite growth in the intraerythrocytic stages is

determined by multiple factors: the length of the asexual

cell cycle, the number of merozoites produced per sch-

izont, the efﬁciency of merozoite release from host cells,

and the invasion efﬁciency of new host cells by the newly

formed merozoites.

Their argument was that measuring growth rates allows access to

these many different processes. But this also means that if one mea-

sures an attenuated growth rate, even for a mutant in a gene expected

to be involved in invasion, one cannot be sure it is the result in a de-

fect in invasion.

Other than a single immunoﬂuorescence experiment, and sanity-

checking bioinformatics, I have done very little in this thesis to verify

that the genes with which I have been working actually have a role

in invasion. The data such as they are are still very helpful, but the

future development of invasion-speciﬁc assays will be important.

Expanding these studies into in vitro systems, such as was the focus

of the P. knowlesi work, will be necessary to make such stage-speciﬁc

functional inferences.

6.3 extensions of work in p. berghei 185

6.2.6 Crossing species barriers

In this thesis I have worked across three species of malaria parasite,

and been able to exploit the strengths of each – the wealth of bioin-

formatic data available for P. falciparum, the robust genetic system

available in P. berghei and the potential for high efﬁciency in vitro

transfection in P. knowlesi. There are clear advantages to such an ap-

proach, but caution must also be exercised.

Ultimately all research into malaria is undertaken because of the

possibility that the results generated may someday have a role to

play in managing or decreasing the hundreds of thousands of deaths

caused by the parasite each year. The vast majority of these deaths are

caused by P. falciparum (though the morbidity due to P. vivax should

not be neglected).

Therefore where model systems are used it is important that dis-

coveries made will be of relevance to human parasites. I believe that

our approach of ﬁnding 1:1:1 orthologues will mean that for the most

part results are applicable to P. falciparum. But we do have examples,

such as the palmitoyl-transferase DHHC3, of proteins with 1:1:1 or-

thology but whose localisation, and likely hence also function, differs

between the species. These possibility of these genetic ’false-friends’

must be considered when inferences are made across species.

6.3 extensions of work in p. berghei

There are a number of ways in which I believe the P. berghei invadome

experiments could be extended.

A key limitation of time-point based barseq is that it gives an indi-

cation only of mutant ﬁtness, but not of the part of parasite biology

in which the gene is involved. However with novel approaches there

may be the potential to gain a more detailed view of the biological

mechanisms underlying invasion. Even in P. berghei, in which inva-

sion is often thought inaccessible there may be some potential for

this.

186 general discussion

6.3.1 Assaying P. berghei invasion in vitro

While P. berghei is primarily an in vivo system it may be possible to

conduct assays of invasion with an in vitro approach. In this system

a pooled barcode transfection of the invadome would be carried out

and grown up in a mouse. At a late stage of infection the mouse

would be bled and ring parasites cultured to schizont stage, some

of which would be analysed by barseq. Mechanical disruption could

then be used to release merozoites, and their invasive capacity as-

sayed by counting barcodes from the fresh rings formed after inva-

sion. It appears that the use of an orbital shaker to induce disruption

of the schizonts [96] might achieve relatively efﬁcient invasion into

reticulocytes with minimal mortality.

6.3.2 Comparative assay of P. berghei invasion in vivo

An alternative approach would be to investigate the connection be-

tween parasite genes and erythrocyte receptors by performing com-

parative invasion assays, which could potentially be performed in

vivo. In such an approach ﬁrst a pooled transfection of the entire

invadome would be carried out as already described. A sample of

the blood would then be taken on, say day 6 post-invasion. This

would contain ring parasites, and erythrocytes could be labelled with

ﬂuorescent-conjugated antibodies, for example with anti-CD71 ([137])

which would identify populations of CD71 reticulocytes from other

red blood cells. The two erythrocyte populations could then be sepa-

rated by ﬂuorescence-activated cell sorting (FACS). After this process

DNA could be extracted from the two subpopulations and barseq

performed.

Mutants differing signiﬁcantly in abundance between the two pop-

ulations would likely be indicative of genes permitting invasion into

the two cell-types. P. berghei shows a signiﬁcant reticulocyte prefer-

ence and so one might expect some genes to be identiﬁed, potentially

with important implications for reticulocyte-restricted P. vivax [134].

Extending this in vivo approach, we need not necessarily limit our-

selves to testing the repertoire of cells existing within a mouse – a

further possibility would be to treat P. berghei erythrocytes in vitro

6.4 extensions of work in p. knowlesi 187

with enzymes such as trypsin, chymotrypsin and neuraminidase in

order to deplete certain surface receptors, then to label these various

populations with ﬂuorescent dyes ([190]). These different target blood

cells could be pooled and injected into a mouse midway through an

invadome barseq transfection. A few hours later, to allow reinvasion,

blood could be taken from this mouse, sorted by FACS to separate

the labelled cells and DNA extracted and barcodes counted. In theory

this might allow an in vivo invasion assay identifying genes involved

in binding to certain host cell receptors.

6.4 extensions of work in p. knowlesi

6.4.1 Barcode-sequencing with technology developed to date

As I discussed at the end of Chapter 4, we are already in a position

to begin barseq experiments in P. knowlesi. However with existing

technology these will be more complex than those in P. berghei due

to the need to perform individual transfections and culture for each

vector, before pooling.

The power of this approach will depend upon the level of attenua-

tion of mutants that can be recovered with PlasmoGEM transfections

in P. knowlesi. Testing this will require further experiments. In the

event that it is discovered that only the ﬁttest of knock-outs can be

recovered, it may be important to ﬁrst investigate ways of enhancing

the efﬁciency of genetic modiﬁcation, as I will discuss later.

Effective barseq established in vitro would enable a wealth of ex-

periments because of the easy accessibility of invasion. Barseq exper-

iments could be used to identify mutants speciﬁcally depleted after

invasion into enzyme treated erythrocytes, erythrocytes of different

ages and even different species (Homo sapiens, Macaca fascicularis and

Macaca nemestrina).

6.4.2 Potential for large scale endogenous tagging

With the genetic modiﬁcation technology developed in Chapter 4, it is

already clear that tagging transfections in P. knowlesi are highly effec-

tive. In every tagging experiment the vast majority of the transfected

188 general discussion

population carried the correct modiﬁcation. There are two major ways

in which I can imagine tagging contributing to our understanding of

invasion. The ﬁrst is localisation – invasion involves a coordinated se-

quence of events involving a series of organelles, from the merozoite

surface to the micronemes to the rhoptries to the inner membrane

complex. Understanding where a protein is localised provides a great

deal of information about its potential role in invasion and so high-

througput tagging is likely to be an important component of better

understanding invasion at scale. This could take place either exactly

as we carried it out in this study – with an epitope tag – or using

ﬂuorescent protein fusions which might permit analysis in live cells.

6.4.3 Conditional analysis to allow the investigation of essential genes

A very different application of tagging would be for establishing the

functional importance of genes with a conditional mutagenesis ap-

proach. Given the observation that episomes can be established in P.

knowlesi, there is a potential concern of parasite harbouring episomes

outcompeting knock-outs and so causing barcode-sequencing to give

false-positive results in a pooled transfection. This concern could be

mitigated with a conditional approach. In such a system the integra-

tion of a barcode would not knock out a gene but tag it with a se-

quence allowing conditional knock-down such as the destabilisation-

domain (DD). [12] Such parasites would be expected to have neg-

ligible ﬁtness defects in permissive conditions (in the presence of

the compound Shld-1) and so to outcompete episomes (due to seg-

regational instability). Thus a mixed population could be established

and barcodes counted in these permissive conditions. Shld-1 could

then be removed and barcodes counted after a round of invasion.

Any difference in barcode abundance between permissive and non-

permissive conditions would necessarily indicate a ﬁtness defect due

to the knockdown. It is possible that this strategy, or a similar ap-

proach with one of the other Plasmodium conditional systems [48]

would also help to separate a role in invasion from a more general

role in metabolism by allowing induction only at the late schizont

stage.

extensions of work in P. KN OWLESI 189

6.4.4 Approaches to improve the rate of DNA integration

The observation that P. knowlesi is capable of carrying a resistance

marker from a transfection without necessarily undergoing the in-

tended ends-out integration event suggests that the robustness of

transfection could be increased by selecting against such parasites.

6.4.4.1 Negative selection marker

One possible means of achieving this is to transfect with vectors, ei-

ther circular or linear, which carry a marker outside the homology

region that allows parasites carrying it to be selected against. This

would select against circular episomes and therefore (in combination

with positive selection) in favour of ends-out integration events. The

thymidine kinase negative selection marker has already been shown

to function in P. knowlesi [199] and so would be a sensible candidate

for such an approach.

6.4.4.2 Crispr/Cas9

A Crispr/Cas9 approach may kill two birds with one stone by pro-

viding negative selection while also enhancing recombination into

the target locus.

Provided there is a good efﬁciency of cutting (which my in silico

analysis sugggested should be achievable for the vast majority of the

P. knowlesi genome), Crispr/Cas9 will provide a constant supply of

double strand breaks in unmodiﬁed loci. In theory given the lack

of non-homologous end-joining in this system these will be resolved

either by the correct integration of a targeting construct or by the

death of the parasite.

Thus selection for a cassette encoding the Cas9 protein and a guide

RNA may substantially boost transfection efﬁciencies as the presence

of double-strand breaks recruits DNA repair machinery to the target

site. My experiments to date have suggested a more modest improve-

ment, but trouble-shooting is needed to investigate whether this is

simply due to technical hurdles.

If Crispr approaches were highly successful there might be a value

in encoding a guide RNA from directly within each PlasmoGEM vec-

tor. This could be achieved by using a long oligo during the recombi-

190 references

neering process containing the guide RNA sequence driven by a T7

promoter. Such an approach might make a pooled barseq experiment

possible in P. knowlesi by recreating the condition present in P. berghei

that the presence of a barcode deﬁnitively indicates the knocking out

of a gene in the parasite.

6.5 final thoughts

2015 is an exciting time to be studying invasion in Plasmodium. It is

clear just how much there is still to be understood about this crucial

event. Simple probability means that some of the vast number of un-

characterised Plasmodium genes must be involved in this process, in

ways we do not yet understand.

The high-throughput tools now becoming available should allow

the extension of my results to identify not only the importance of

these genes but to prove that their role is in invasion. High-throughput

localisation should even be able to indicate at what stage of the intri-

cate process they play their role.

As our understanding of invasion is ﬁlled in with increased resolu-

tion, the power of large scale bioinformatic analysis based on protein-

protein interaction data can be expected to multiply the effect of each

discovery by pulling previously uncharacterised genes into increas-

ingly sophisticated models. These approaches will have to bring to-

gether data on a protein’s interactions on a molecular scale, on its

organellar localisation, and on its the phenotypes of parasite created

when its gene is disrupted – in terms of the classes of cell that can be

invaded and their relative efﬁciencies.

The technology for generating each of these classes of data at scale

is just starting to be employed in laboratories around the world. In

the coming years, we can hope that at their intersection will emerge

a much clearer understanding of the complex molecular machinery

that underlies erythrocyte invasion.

B I B L I O G R A P H Y

[1] J. H. Adams, B. K. I. M. L. E. E. Simt, S. A. Dolan, X. Fang,

D. C. Kaslow, and L. H. Miller. A family of erythrocyte binding

proteins of malaria parasites. 89(August):7085–7089, 1992.

[2] A. M. Ahmed, M. M. Pinheiro, P. C. Divis, A. Siner, R. Zain-

udin, I. T. Wong, C. W. Lu, S. K. Singh-Khaira, S. B.

Millar, S. Lynch, M. Willmann, B. Singh, S. Krishna, and

J. Cox-Singh. Disease progression in Plasmodium knowlesi

malaria is linked to variation in invasion gene family mem-

bers. PLoS neglected tropical diseases, 8(8):e3086, Aug. 2014.

ISSN 1935-2735. doi: 10.1371/journal.pntd.0003086. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=4133233&tool=pmcentrez&rendertype=abstract.

[3] K. A. Akinosoglou, E. S. C. Bushell, C. V. Ukegbu,

T. Schlegelmilch, J.-S. Cho, S. Redmond, K. Sala, G. K.

Christophides, and D. Vlachou. Characterization of

Plasmodium developmental transcriptomes in Anophe-

les gambiae midgut reveals novel regulators of malaria

transmission. Cellular microbiology, 17(2):254–68, Mar.

2015. ISSN 1462-5822. doi: 10.1111/cmi.12363. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=4371638&tool=pmcentrez&rendertype=abstract.

[4] M. M. Alam, L. Solyakov, A. R. Bottrill, C. Flueck, F. A. Siddiqui,

S. Singh, S. Mistry, M. Viskaduraki, K. Lee, C. S. Hopp, C. E.

Chitnis, C. Doerig, R. W. Moon, J. L. Green, A. A. Holder, D. A.

Baker, and A. B. Tobin. Phosphoproteomics reveals malaria par-

asite Protein Kinase G as a signalling hub regulating egress and

invasion. Nature communications, 6:7285, Jan. 2015. ISSN 2041-

1723. doi: 10.1038/ncomms8285. URL http://www.nature.com/

ncomms/2015/150706/ncomms8285/full/ncomms8285.html.

[5] A. P. Alker, V. Mwapasa, and S. R. Meshnick. Rapid real-time

PCR genotyping of mutations associated with sulfadoxine-

pyrimethamine resistance in Plasmodium falciparum. An-

timicrobial agents and chemotherapy, 48(8):2924–9, Aug. 2004.

ISSN 0066-4804. doi: 10.1128/AAC.48.8.2924-2929.2004. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=478490&tool=pmcentrez&rendertype=abstract.

[6] R. J. W. Allen and K. Kirk. Plasmodium falciparum culture: the

beneﬁts of shaking. Molecular and biochemical parasitology, 169(1):

63–5, Jan. 2010. ISSN 1872-9428. doi: 10.1016/j.molbiopara.2009.

09.005. URL http://www.ncbi.nlm.nih.gov/pubmed/19766147.

191

192 references

[7] R. J. W. Allen and K. Kirk. Plasmodium falciparum culture: The

beneﬁts of shaking. Molecular and Biochemical Parasitology, 169

(1):63–65, 2010. ISSN 01666851. doi: 10.1016/j.molbiopara.2009.

09.005.

[8] P. L. Alonso and M. Tanner. Public health challenges and

prospects for malaria control and elimination. Nature medicine,

19(2):150–5, Feb. 2013. ISSN 1546-170X. doi: 10.1038/nm.3077.

URL http://dx.doi.org/10.1038/nm.3077.

[9] M. Anisimova and O. Gascuel. Approximate likelihood-ratio

test for branches: A fast, accurate, and powerful alternative. Sys-

tematic biology, 55(4):539–52, Aug. 2006. ISSN 1063-5157. doi:

10.1080/10635150600755453. URL http://www.ncbi.nlm.nih.

gov/pubmed/16785212.

[10] F. A. Ansari, N. Kumar, M. Bala Subramanyam, M. Gnanamani,

and S. Ramachandran. MAAP: malarial adhesins and adhesin-

like proteins predictor. Proteins, 70(3):659–66, Feb. 2008. ISSN

1097-0134. doi: 10.1002/prot.21568. URL http://www.ncbi.nlm.

nih.gov/pubmed/17879344.

[11] S. Antinori, L. Milazzo, and M. Corbellino. Plasmodium

knowlesi: An Overlooked Italian Discovery? Clinical infec-

tious diseases : an ofﬁcial publication of the Infectious Diseases

Society of America, 53(8):849; author reply 849–50, Oct. 2011.

ISSN 1537-6591. doi: 10.1093/cid/cir527. URL http://cid.

oxfordjournals.org/content/53/8/849.1.full.

[12] C. M. Armstrong and D. E. Goldberg. An FKBP destabiliza-

tion domain modulates protein levels in Plasmodium falci-

parum. Nature methods, 4(12):1007–9, Dec. 2007. ISSN 1548-7105.

doi: 10.1038/nmeth1132. URL http://www.ncbi.nlm.nih.gov/

pubmed/17994030.

[13] D. L. Baldi, K. T. Andrews, R. F. Waller, D. S. Roos, R. F.

Howard, B. S. Crabb, and A. F. Cowman. RAP1 controls

rhoptry targeting of RAP2 in the malaria parasite Plasmod-

ium falciparum. The EMBO journal, 19(11):2435–43, June

2000. ISSN 0261-4189. doi: 10.1093/emboj/19.11.2435. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=212767&tool=pmcentrez&rendertype=abstract.

[14] B. Balu, C. Chauhan, S. P. Maher, D. A. Shoue, J. C.

Kissinger, M. J. Fraser, and J. H. Adams. piggyBac is

an effective tool for functional analysis of the Plasmod-

ium falciparum genome. BMC microbiology, 9:83, Jan.

2009. ISSN 1471-2180. doi: 10.1186/1471-2180-9-83. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=2686711&tool=pmcentrez&rendertype=abstract.

references 193

[15] B. Balu, N. Singh, S. P. Maher, and J. H. Adams. A ge-

netic screen for attenuated growth identiﬁes genes cru-

cial for intraerythrocytic development of Plasmodium

falciparum. PloS one, 5(10):e13282, Jan. 2010. ISSN

1932-6203. doi: 10.1371/journal.pone.0013282. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=2952599&tool=pmcentrez&rendertype=abstract.

[16] D. Bargieri, V. Lagal, I. Tardieux, and R. Ménard. Host cell

invasion by apicomplexans: what do we know? Trends in para-

sitology, 28(4):131–5, Apr. 2012. ISSN 1471-5007. doi: 10.1016/

j.pt.2012.01.005. URL http://www.ncbi.nlm.nih.gov/pubmed/

22326913.

[17] J. Baum, D. Richard, J. Healer, M. Rug, Z. Krnajski, T.-W.

Gilberger, J. L. Green, A. A. Holder, and A. F. Cowman. A con-

served molecular motor drives cell invasion and gliding motil-

ity across malaria life cycle stages and other apicomplexan par-

asites. The Journal of biological chemistry, 281(8):5197–208, Feb.

2006. ISSN 0021-9258. doi: 10.1074/jbc.M509807200. URL

http://www.ncbi.nlm.nih.gov/pubmed/16321976.

[18] J. Baum, L. Chen, J. Healer, S. Lopaticki, M. Boyle, T. Triglia,

F. Ehlgen, S. a. Ralph, J. G. Beeson, and A. F. Cowman.

Reticulocyte-binding protein homologue 5 - an essential ad-

hesin involved in invasion of human erythrocytes by Plasmod-

ium falciparum. International journal for parasitology, 39(3):371–

80, Feb. 2009. ISSN 1879-0135. doi: 10.1016/j.ijpara.2008.10.006.

URL http://www.ncbi.nlm.nih.gov/pubmed/19000690.

[19] J. Baum, A. T. Papenfuss, G. R. Mair, C. J. Janse, D. Vlachou,

A. P. Waters, A. F. Cowman, B. S. Crabb, and T. F. de Koning-

Ward. Molecular genetics and comparative genomics reveal

RNAi is not functional in malaria parasites. Nucleic Acids

Research, 37(11):3788–3798, Apr. 2009. ISSN 0305-1048. doi:

10.1093/nar/gkp239. URL http://nar.oxfordjournals.org/

content/37/11/3788.full.

[20] A. Berendt, G. Tumer, and C. Newbold. Cerebral malaria:

The sequestration hypothesis. Parasitology Today, 10(10):

412–414, Jan. 1994. ISSN 01694758. doi: 10.1016/

0169-4758(94)90238-0. URL http://www.sciencedirect.com/

science/article/pii/0169475894902380.

[21] M. Botha, E.-R. Pesce, and G. L. Blatch. The Hsp40 proteins

of Plasmodium falciparum and other apicomplexa: regulating

chaperone power in the parasite and the host. The international

journal of biochemistry & cell biology, 39(10):1781–803, Jan. 2007.

ISSN 1357-2725. doi: 10.1016/j.biocel.2007.02.011. URL http:

//www.ncbi.nlm.nih.gov/pubmed/17428722.

194 references

[22] Z. Bozdech, M. Llinás, B. L. Pulliam, E. D. Wong, J. Zhu, and

J. L. DeRisi. The transcriptome of the intraerythrocytic develop-

mental cycle of Plasmodium falciparum. PLoS biology, 1(1):E5,

Oct. 2003. ISSN 1545-7885. doi: 10.1371/journal.pbio.0000005.

URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=176545&tool=pmcentrez&rendertype=abstract.

[23] F. Brossier, T. J. Jewett, L. D. Sibley, and S. Urban. A spatially

localized rhomboid protease cleaves cell surface adhesins es-

sential for invasion by Toxoplasma. Proceedings of the National

Academy of Sciences of the United States of America, 102(11):4146–

51, Mar. 2005. ISSN 0027-8424. doi: 10.1073/pnas.0407918102.

URL http://www.pnas.org/content/102/11/4146.abstract.

[24] G. a. Butcher. Factors affecting the in vitro culture

of Plasmodium falciparum and Plasmodium knowlesi.

Bulletin of the World Health Organization, 57 Suppl 1

(Suppl 1):17–26, Jan. 1979. ISSN 0042-9686. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=2395724&tool=pmcentrez&rendertype=abstract.

[25] W. Campbell. Chemotherapy of Parasitic Diseases, vol-

ume 11. Springer Science & Business Media, 1986.

ISBN 1468412337. URL https://books.google.com/books?id=

MRvaBwAAQBAJ&pgis=1.

[26] D. Camus and T. J. Hadley. A Plasmodium falciparum Antigen

That Binds to Host Erythrocytes and Merozoites. Sceince, 327

(1981):1982–1985, 1985.

[27] F. Caro, M. G. Miller, and J. L. DeRisi. Plate-based trans-

fection and culturing technique for genetic manipulation of

Plasmodium falciparum. Malaria journal, 11(1):22, Jan. 2012.

ISSN 1475-2875. doi: 10.1186/1475-2875-11-22. URL http:

//www.malariajournal.com/content/11/1/22.

[28] A. Carreau, B. El Hafny-Rahbi, A. Matejuk, C. Grillon, and

C. Kieda. Why is the partial oxygen pressure of human

tissues a crucial parameter? Small molecules and hypoxia.

Journal of cellular and molecular medicine, 15(6):1239–53, June

2011. ISSN 1582-4934. doi: 10.1111/j.1582-4934.2011.01258.

x. URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=4373326&tool=pmcentrez&rendertype=abstract.

[29] R. Carter. Speculations on the origins of Plasmodium vivax

malaria. Trends in parasitology, 19(5):214–9, May 2003. ISSN 1471-

4922. URL http://www.ncbi.nlm.nih.gov/pubmed/12763427.

[30] J. Castresana. Selection of conserved blocks from multiple align-

ments for their use in phylogenetic analysis. Molecular biology

and evolution, 17(4):540–52, Apr. 2000. ISSN 0737-4038. URL

http://www.ncbi.nlm.nih.gov/pubmed/10742046.

references 195

[31] CDC. Perspectives: Malaria in Long-Term Travelers & Expatri-

ates - Chapter 8 - 2016 Yellow Book | Travelers’ Health | CDC,

2016. URL http://wwwnc.cdc.gov/travel/yellowbook/

2016/advising-travelers-with-specific-needs/

perspectives-malaria-in-long-term-travelers-expatriates.

[32] CDC. Long-Term Travelers & Expatriates - Chap-

ter 8 - 2016 Yellow Book | Travelers’ Health | CDC,

2016. URL http://wwwnc.cdc.gov/travel/yellowbook/

2016/advising-travelers-with-specific-needs/

long-term-travelers-expatriates.

[33] R. Chandramohanadas, P. H. Davis, D. P. Beiting, M. B.

Harbut, C. Darling, G. Velmourougane, M. Y. Lee, P. A.

Greer, D. S. Roos, and D. C. Greenbaum. Apicomplexan

parasites co-opt host calpains to facilitate their escape from

infected cells. Science (New York, N.Y.), 324(5928):794–7, May

2009. ISSN 1095-9203. doi: 10.1126/science.1171085. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3391539&tool=pmcentrez&rendertype=abstract.

[34] R. Chattopadhyay, D. Rathore, H. Fujioka, S. Kumar, P. de la

Vega, D. Haynes, K. Moch, D. Fryauff, R. Wang, D. J. Carucci,

and S. L. Hoffman. PfSPATR, a Plasmodium falciparum pro-

tein containing an altered thrombospondin type I repeat do-

main is expressed at several stages of the parasite life cycle

and is the target of inhibitory antibodies. The Journal of biologi-

cal chemistry, 278(28):25977–81, July 2003. ISSN 0021-9258. doi:

10.1074/jbc.M300865200. URL http://www.ncbi.nlm.nih.gov/

pubmed/12716913.

[35] F. Chevenet, C. Brun, A.-L. Bañuls, B. Jacq, and R. Chris-

ten. TreeDyn: towards dynamic graphics and annota-

tions for analyses of trees. BMC bioinformatics, 7:439, Jan.

2006. ISSN 1471-2105. doi: 10.1186/1471-2105-7-439. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=1615880&tool=pmcentrez&rendertype=abstract.

[36] M. A. Child, C. Epp, H. Bujard, and M. J. Blackman. Regulated

maturation of malaria merozoite surface protein-1 is essential

for parasite growth. Molecular microbiology, 78(1):187–202, Oct.

2010. ISSN 1365-2958. doi: 10.1111/j.1365-2958.2010.07324.

x. URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=2995310&tool=pmcentrez&rendertype=abstract.

[37] W. Chin, P. G. Contacos, G. R. Coatney, and H. R. Kimball. A

Naturally Acquired Quotidian-Type Malaria in Man Transfer-

able to Monkeys. Science, 149(3686):865–865, Aug. 1965. ISSN

0036-8075. doi: 10.1126/science.149.3686.865. URL http://www.

sciencemag.org/content/149/3686/865.1.abstract.

196 references

[38] C. R. Collins, F. Hackett, M. Strath, M. Penzo, C. Withers-

Martinez, D. a. Baker, and M. J. Blackman. Malaria

parasite cGMP-dependent protein kinase regulates

blood stage merozoite secretory organelle discharge

and egress. PLoS pathogens, 9(5):e1003344, May 2013.

ISSN 1553-7374. doi: 10.1371/journal.ppat.1003344. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3649973&tool=pmcentrez&rendertype=abstract.

[39] A. Combe, D. Giovannini, T. G. Carvalho, S. Spath, B. Bois-

son, C. Loussert, S. Thiberge, C. Lacroix, P. Gueirard, and

R. Ménard. Clonal conditional mutagenesis in malaria para-

sites. Cell host & microbe, 5(4):386–96, Apr. 2009. ISSN 1934-

6069. doi: 10.1016/j.chom.2009.03.008. URL http://www.ncbi.

nlm.nih.gov/pubmed/19380117.

[40] O. E. Cornejo and A. A. Escalante. The origin and age

of Plasmodium vivax. Trends in parasitology, 22(12):558–

63, Dec. 2006. ISSN 1471-4922. doi: 10.1016/j.pt.2006.09.

007. URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=1855252&tool=pmcentrez&rendertype=abstract.

[41] A. F. Cowman and B. S. Crabb. Invasion of red blood cells by

malaria parasites. Cell, 124(4):755–66, Feb. 2006. ISSN 0092-8674.

doi: 10.1016/j.cell.2006.02.006. URL http://www.ncbi.nlm.nih.

gov/pubmed/16497586.

[42] A. F. Cowman, D. Berry, and J. Baum. The cellular and

molecular basis for malaria parasite invasion of the human

red blood cell. The Journal of cell biology, 198(6):961–71, Sept.

2012. ISSN 1540-8140. doi: 10.1083/jcb.201206112. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3444787&tool=pmcentrez&rendertype=abstract.

[43] J. Cox-Singh and R. Culleton. Plasmodium knowlesi: from se-

vere zoonosis to animal model. Trends in parasitology, pages 1–7,

Mar. 2015. ISSN 1471-5007. doi: 10.1016/j.pt.2015.03.003. URL

http://www.ncbi.nlm.nih.gov/pubmed/25837310.

[44] J. Cox-Singh, T. M. E. Davis, K.-S. Lee, S. S. G. Shamsul, A. Ma-

tusop, S. Ratnam, H. a. Rahman, D. J. Conway, and B. Singh.

Plasmodium knowlesi malaria in humans is widely distributed

and potentially life threatening. Clinical infectious diseases : an

ofﬁcial publication of the Infectious Diseases Society of America, 46

(2):165–71, Jan. 2008. ISSN 1537-6591. doi: 10.1086/524888.

URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=2533694&tool=pmcentrez&rendertype=abstract.

[45] C. Crosnier, L. Y. Bustamante, S. J. Bartholdson, A. K.

Bei, M. Theron, M. Uchikawa, S. Mboup, O. Ndir, D. P.

Kwiatkowski, M. T. Duraisingh, J. C. Rayner, and G. J.

references 197

Wright. Basigin is a receptor essential for erythrocyte in-

vasion by Plasmodium falciparum. Nature, 480(7378):534–7,

Dec. 2011. ISSN 1476-4687. doi: 10.1038/nature10606. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3245779&tool=pmcentrez&rendertype=abstract.

[46] R. Culleton, C. Coban, F. Y. Zeyrek, P. Cravo, A. Kaneko,

M. Randrianarivelojosia, V. Andrianaranjaka, S. Kano, A. Farn-

ert, A. P. Arez, P. M. Sharp, R. Carter, and K. Tanabe.

The origins of African Plasmodium vivax; insights from

mitochondrial genome sequencing. PloS one, 6(12):e29137,

Jan. 2011. ISSN 1932-6203. doi: 10.1371/journal.pone.

0029137. URL http://journals.plos.org/plosone/article?

id=10.1371/journal.pone.0029137.

[47] C. Daneshvar, T. M. E. Davis, J. Cox-Singh, M. Z. Rafa’ee,

S. K. Zakaria, P. C. S. Divis, and B. Singh. Clinical and

laboratory features of human Plasmodium knowlesi in-

fection. Clinical infectious diseases : an ofﬁcial publication

of the Infectious Diseases Society of America, 49(6):852–60,

Sept. 2009. ISSN 1537-6591. doi: 10.1086/605439. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=2843824&tool=pmcentrez&rendertype=abstract.

[48] T. F. de Koning-Ward, P. R. Gilson, and B. S. Crabb. Advances

in molecular genetic systems in malaria. Nature reviews. Micro-

biology, 13(6):373–387, May 2015. ISSN 1740-1534. doi: 10.1038/

nrmicro3450. URL http://dx.doi.org/10.1038/nrmicro3450.

[49] K. Deitsch, C. Driskill, and T. Wellems. Transformation

of malaria parasites by the spontaneous uptake and ex-

pression of DNA from human erythrocytes. Nucleic acids

research, 29(3):850–3, Feb. 2001. ISSN 1362-4962. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=30384&tool=pmcentrez&rendertype=abstract.

[50] X. Deng, R. Gujjar, F. El Mazouni, W. Kaminsky, N. A.

Malmquist, E. J. Goldsmith, P. K. Rathod, and M. A. Phillips.

Structural plasticity of malaria dihydroorotate dehydroge-

nase allows selective binding of diverse chemical scaffolds.

The Journal of biological chemistry, 284(39):26999–7009, Sept.

2009. ISSN 1083-351X. doi: 10.1074/jbc.M109.028589. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=2785385&tool=pmcentrez&rendertype=abstract.

[51] M. B. Denis, R. Tsuyuoka, P. Lim, N. Lindegardh, P. Yi, S. N.

Top, D. Socheat, T. Fandeur, A. Annerberg, E. M. Christophel,

and P. Ringwald. Efﬁcacy of artemether-lumefantrine for the

treatment of uncomplicated falciparum malaria in northwest

Cambodia. Tropical medicine & international health : TM &

IH, 11(12):1800–7, Dec. 2006. ISSN 1360-2276. doi: 10.1111/

198 references

j.1365-3156.2006.01739.x. URL http://www.ncbi.nlm.nih.gov/

pubmed/17176344.

[52] A. Dereeper, V. Guignon, G. Blanc, S. Audic, S. Buf-

fet, F. Chevenet, J.-F. Dufayard, S. Guindon, V. Lefort,

M. Lescot, J.-M. Claverie, and O. Gascuel. Phylogeny.fr:

robust phylogenetic analysis for the non-specialist. Nu-

cleic acids research, 36(Web Server issue):W465–9, July

2008. ISSN 1362-4962. doi: 10.1093/nar/gkn180. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=2447785&tool=pmcentrez&rendertype=abstract.

[53] E. Deu, M. J. Leyva, V. E. Albrow, M. J. Rice, J. A. Ell-

man, and M. Bogyo. Functional studies of Plasmodium falci-

parum dipeptidyl aminopeptidase I using small molecule in-

hibitors and active site probes. Chemistry & biology, 17(8):808–

19, Aug. 2010. ISSN 1879-1301. doi: 10.1016/j.chembiol.2010.06.

007. URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=2929396&tool=pmcentrez&rendertype=abstract.

[54] P. C. S. Divis, B. Singh, F. Anderios, S. Hisam, A. Matusop,

C. H. Kocken, S. a. Assefa, C. W. Duffy, and D. J. Conway. Ad-

mixture in Humans of Two Divergent Plasmodium knowlesi

Populations Associated with Different Macaque Host Species.

PLoS pathogens, 11(5):e1004888, May 2015. ISSN 1553-7374. doi:

10.1371/journal.ppat.1004888. URL http://www.ncbi.nlm.nih.

gov/pubmed/26020959.

[55] J. G. Doench, E. Hartenian, D. B. Graham, Z. Tothova, M. Hegde,

I. Smith, M. Sullender, B. L. Ebert, R. J. Xavier, and D. E.

Root. Rational design of highly active sgRNAs for CRISPR-

Cas9-mediated gene inactivation. Nature Biotechnology, 32(12):

1262–7, Sept. 2014. ISSN 1087-0156. doi: 10.1038/nbt.3026. URL

http://www.ncbi.nlm.nih.gov/pubmed/25184501.

[56] L. Duval, M. Fourment, E. Nerrienet, D. Rousset, S. A. Sadeuh,

S. M. Goodman, N. V. Andriaholinirina, M. Randrianarivelo-

josia, R. E. Paul, V. Robert, F. J. Ayala, and F. Ariey. African

apes as reservoirs of Plasmodium falciparum and the origin

and diversiﬁcation of the Laverania subgenus. Proceedings of

the National Academy of Sciences of the United States of Amer-

ica, 107(23):10561–6, June 2010. ISSN 1091-6490. doi: 10.1073/

pnas.1005435107. URL http://www.pnas.org/content/107/23/

10561.

[57] J. A. Dvorak and H. Miller, Louis. Invasion of Erythrocytes by

Malaria Merozoites. Science, 187:5–7, 1975.

[58] A. Ecker, S. B. Pinto, K. W. Baker, F. C. Kafatos, and

R. E. Sinden. Plasmodium berghei: plasmodium perforin-

like protein 5 is required for mosquito midgut invasion in

references 199

Anopheles stephensi. Experimental parasitology, 116(4):504–8,

Aug. 2007. ISSN 0014-4894. doi: 10.1016/j.exppara.2007.01.

015. URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=1916484&tool=pmcentrez&rendertype=abstract.

[59] A. Ecker, E. S. C. Bushell, R. Tewari, and R. E. Sinden. Reverse

genetics screen identiﬁes six proteins important for malaria de-

velopment in the mosquito. Molecular microbiology, 70(1):209–20,

Oct. 2008. ISSN 1365-2958. doi: 10.1111/j.1365-2958.2008.06407.

x. URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=2658712&tool=pmcentrez&rendertype=abstract.

[60] R. C. Edgar. MUSCLE: multiple sequence alignment with high

accuracy and high throughput. Nucleic acids research, 32(5):

1792–7, Jan. 2004. ISSN 1362-4962. doi: 10.1093/nar/gkh340.

URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=390337&tool=pmcentrez&rendertype=abstract.

[61] S. Egarter, N. Andenmatten, A. J. Jackson, J. A. Whitelaw,

G. Pall, J. A. Black, D. J. P. Ferguson, I. Tardieux, A. Mogilner,

and M. Meissner. The toxoplasma Acto-MyoA motor complex

is important but not essential for gliding motility and host cell

invasion. PloS one, 9(3):e91819, Jan. 2014. ISSN 1932-6203. doi:

10.1371/journal.pone.0091819. URL http://journals.plos.

org/plosone/article?id=10.1371/journal.pone.0091819.

[62] A. Farrell, S. Thirugnanam, A. Lorestani, J. D. Dvorin, K. P.

Eidell, D. J. P. Ferguson, B. R. Anderson-White, M. T. Durais-

ingh, G. T. Marth, and M.-J. Gubbels. A DOC2 protein identi-

ﬁed by mutational proﬁling is essential for apicomplexan par-

asite exocytosis. Science (New York, N.Y.), 335(6065):218–21, Jan.

2012. ISSN 1095-9203. doi: 10.1126/science.1210829. URL

http://www.sciencemag.org/content/335/6065/218.long.

[63] S. Fauquenoy, A. Hovasse, P.-J. Sloves, W. Morelle, T. Dilezi-

toko Alayi, T. Dilezitoko Ayali, C. Slomianny, E. Werkmeister,

C. Schaeffer, A. Van Dorsselaer, and S. Tomavo. Unusual N-

glycan structures required for trafﬁcking Toxoplasma gondii

GAP50 to the inner membrane complex regulate host cell en-

try through parasite motility. Molecular & cellular proteomics

: MCP, 10(9):M111.008953, Sept. 2011. ISSN 1535-9484. doi:

10.1074/mcp.M111.008953. URL http://www.mcponline.org/

content/10/9/M111.008953/F9.expansion.

[64] J. Fonager, E. M. Pasini, J. A. M. Braks, O. Klop, J. Ramesar,

E. J. Remarque, I. O. C. M. Vroegrijk, S. G. van Duinen,

A. W. Thomas, S. M. Khan, M. Mann, C. H. M. Kocken,

C. J. Janse, and B. M. D. Franke-Fayard. Reduced CD36-

dependent tissue sequestration of Plasmodium-infected

erythrocytes is detrimental to malaria parasite growth in

vivo. The Journal of experimental medicine, 209(1):93–107, Jan.

200 references

2012. ISSN 1540-9538. doi: 10.1084/jem.20110762. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3260870&tool=pmcentrez&rendertype=abstract.

[65] K. Frénal, C. L. Tay, C. Mueller, E. S. Bushell, Y. Jia, A. Grain-

dorge, O. Billker, J. C. Rayner, and D. Soldati-Favre. Global

analysis of apicomplexan protein S-acyl transferases reveals an

enzyme essential for invasion. Trafﬁc (Copenhagen, Denmark), 14

(8):895–911, Aug. 2013. ISSN 1600-0854. doi: 10.1111/tra.12081.

URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=3813974&tool=pmcentrez&rendertype=abstract.

[66] S. Fyfe, C. Williams, O. J. Mason, and G. J. Pickup. Apophe-

nia, theory of mind and schizotypy: perceiving meaning and

intentionality in randomness. Cortex; a journal devoted to the

study of the nervous system and behavior, 44(10):1316–25, Jan. .

ISSN 0010-9452. doi: 10.1016/j.cortex.2007.07.009. URL http:

//www.ncbi.nlm.nih.gov/pubmed/18635161.

[67] M. R. Galinski and J. W. Barnwell. Plasmodium vivax:

who cares? Malaria journal, 7 Suppl 1:S9, Jan. 2008.

ISSN 1475-2875. doi: 10.1186/1475-2875-7-S1-S9. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=2604873&tool=pmcentrez&rendertype=abstract.

[68] M. R. Galinski, C. C. Medina, P. Ingravallo, and J. W. Barnwell.

A reticulocyte-binding protein complex of plasmodium vivax

merozoites. Cell, 69(7):1213–1226, June 1992. ISSN 00928674.

doi: 10.1016/0092-8674(92)90642-P. URL http://linkinghub.

elsevier.com/retrieve/pii/009286749290642P.

[69] C. R. Garcia, F. Manzi, F. Tediosi, S. L. Hoffman, and E. R.

James. Comparative cost models of a liquid nitrogen vapor

phase (LNVP) cold chain-distributed cryopreserved malaria

vaccine vs. a conventional vaccine. Vaccine, 31(2):380–6,

Jan. 2013. ISSN 1873-2518. doi: 10.1016/j.vaccine.2012.10.

109. URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=3666854&tool=pmcentrez&rendertype=abstract.

[70] M. J. Gardner, N. Hall, E. Fung, O. White, M. Berriman, R. W.

Hyman, J. M. Carlton, A. Pain, K. E. Nelson, S. Bowman, I. T.

Paulsen, K. James, J. A. Eisen, K. Rutherford, S. L. Salzberg,

A. Craig, S. Kyes, M.-S. Chan, V. Nene, S. J. Shallom, B. Suh,

J. Peterson, S. Angiuoli, M. Pertea, J. Allen, J. Selengut, D. Haft,

M. W. Mather, A. B. Vaidya, D. M. A. Martin, A. H. Fair-

lamb, M. J. Fraunholz, D. S. Roos, S. A. Ralph, G. I. McFad-

den, L. M. Cummings, G. M. Subramanian, C. Mungall, J. C.

Venter, D. J. Carucci, S. L. Hoffman, C. Newbold, R. W. Davis,

C. M. Fraser, and B. Barrell. Genome sequence of the human

malaria parasite Plasmodium falciparum. Nature, 419(6906):

498–511, Oct. 2002. ISSN 0028-0836. doi: 10.1038/nature01097.

URL http://dx.doi.org/10.1038/nature01097.

references 201

[71] S. Y. Gerdes, M. D. Scholle, J. W. Campbell, G. Balázsi,

E. Ravasz, M. D. Daugherty, A. L. Somera, N. C. Kyrpides,

I. Anderson, M. S. Gelfand, A. Bhattacharya, V. Kapatral,

M. D’Souza, M. V. Baev, Y. Grechkin, F. Mseeh, M. Y. Fonstein,

R. Overbeek, A.-L. Barabási, Z. N. Oltvai, and A. L. Oster-

man. Experimental determination and system level analysis

of essential genes in Escherichia coli MG1655. Journal of

bacteriology, 185(19):5673–84, Oct. 2003. ISSN 0021-9193. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=193955&tool=pmcentrez&rendertype=abstract.

[72] M. Ghorbal, M. Gorman, C. R. Macpherson, R. M. Martins,

A. Scherf, and J.-J. Lopez-Rubio. Genome editing in the human

malaria parasite Plasmodium falciparum using the CRISPR-

Cas9 system. Nature biotechnology, 32(8):819–821, 2014. ISSN

1546-1696. doi: 10.1038/nbt.2925. URL http://dx.doi.org/10.

1038/nbt.2925.

[73] G. Giaever, A. M. Chu, L. Ni, C. Connelly, L. Riles, S. Véron-

neau, S. Dow, A. Lucau-Danila, K. Anderson, B. André,

A. P. Arkin, A. Astromoff, M. El-Bakkoury, R. Bangham,

R. Benito, S. Brachat, S. Campanaro, M. Curtiss, K. Davis,

A. Deutschbauer, K.-D. Entian, P. Flaherty, F. Foury, D. J.

Garﬁnkel, M. Gerstein, D. Gotte, U. Güldener, J. H. Hege-

mann, S. Hempel, Z. Herman, D. F. Jaramillo, D. E. Kelly, S. L.

Kelly, P. Kötter, D. LaBonte, D. C. Lamb, N. Lan, H. Liang,

H. Liao, L. Liu, C. Luo, M. Lussier, R. Mao, P. Menard,

S. L. Ooi, J. L. Revuelta, C. J. Roberts, M. Rose, P. Ross-

Macdonald, B. Scherens, G. Schimmack, B. Shafer, D. D. Shoe-

maker, S. Sookhai-Mahadeo, R. K. Storms, J. N. Strathern,

G. Valle, M. Voet, G. Volckaert, C.-y. Wang, T. R. Ward, J. Wil-

helmy, E. A. Winzeler, Y. Yang, G. Yen, E. Youngman, K. Yu,

H. Bussey, J. D. Boeke, M. Snyder, P. Philippsen, R. W. Davis,

and M. Johnston. Functional proﬁling of the Saccharomyces

cerevisiae genome. Nature, 418(6896):387–91, July 2002. ISSN

0028-0836. doi: 10.1038/nature00935. URL http://www.ncbi.

nlm.nih.gov/pubmed/12140549.

[74] D. G. Gibson, J. I. Glass, C. Lartigue, V. N. Noskov, R.-

Y. Chuang, M. A. Algire, G. A. Benders, M. G. Montague,

L. Ma, M. M. Moodie, C. Merryman, S. Vashee, R. Krishnaku-

mar, N. Assad-Garcia, C. Andrews-Pfannkoch, E. A. Denisova,

L. Young, Z.-Q. Qi, T. H. Segall-Shapiro, C. H. Calvey, P. P.

Parmar, C. A. Hutchison, H. O. Smith, and J. C. Venter. Cre-

ation of a bacterial cell controlled by a chemically synthe-

sized genome. Science (New York, N.Y.), 329(5987):52–6, July

2010. ISSN 1095-9203. doi: 10.1126/science.1190719. URL

http://www.ncbi.nlm.nih.gov/pubmed/20488990.

[75] P. R. Gilson and B. S. Crabb. Morphology and kinetics of the

three distinct phases of red blood cell invasion by Plasmodium

202 references

falciparum merozoites. International journal for parasitology, 39

(1):91–6, Jan. 2009. ISSN 1879-0135. doi: 10.1016/j.ijpara.2008.

09.007. URL http://www.ncbi.nlm.nih.gov/pubmed/18952091.

[76] D. Giovannini, S. Späth, C. Lacroix, A. Perazzi, D. Bargieri,

V. Lagal, C. Lebugle, A. Combe, S. Thiberge, P. Baldacci,

I. Tardieux, and R. Ménard. Independent roles of apical mem-

brane antigen 1 and rhoptry neck proteins during host cell in-

vasion by apicomplexa. Cell host & microbe, 10(6):591–602, Dec.

2011. ISSN 1934-6069. doi: 10.1016/j.chom.2011.10.012. URL

http://www.ncbi.nlm.nih.gov/pubmed/22177563.

[77] S. Glushakova, V. Lizunov, P. S. Blank, K. Melikov,

G. Humphrey, and J. Zimmerberg. Cytoplasmic free Ca2+ is

essential for multiple steps in malaria parasite egress from in-

fected erythrocytes. Malaria journal, 12(1):41, Jan. 2013. ISSN

1475-2875. doi: 10.1186/1475-2875-12-41. URL http://www.

malariajournal.com/content/12/1/41.

[78] R. Godiska, D. Mead, V. Dhodda, C. Wu, R. Hochstein,

A. Karsi, K. Usdin, A. Entezam, and N. Ravin. Linear plas-

mid vector for cloning of repetitive or unstable sequences

in Escherichia coli. Nucleic acids research, 38(6):e88, Apr.

2010. ISSN 1362-4962. doi: 10.1093/nar/gkp1181. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=2847241&tool=pmcentrez&rendertype=abstract.

[79] A. R. Gomes, E. Bushell, F. Schwach, G. Girling, B. Anar, M. A.

Quail, C. Herd, C. Pfander, K. Modrzynska, J. C. Rayner, and

O. Billker. A Genome-Scale Vector Resource Enables High-

Throughput Reverse Genetic Screening in a Malaria Parasite.

Cell Host & Microbe, Feb. 2015. ISSN 19313128. doi: 10.

1016/j.chom.2015.01.014. URL http://www.cell.com/article/

S1931312815000347/fulltext.

[80] A. R. B. Gomes. High-throughput reverse genetic screening in Plas-

modium berghei using barcode sequencing. PhD thesis, 2014.

[81] C. A. Goodman, P. G. Coleman, and A. J. Mills. Cost-

effectiveness of malaria control in sub-Saharan Africa. Lancet

(London, England), 354(9176):378–85, July 1999. ISSN 0140-6736.

URL http://www.ncbi.nlm.nih.gov/pubmed/10437867.

[82] J. L. Gordon. Characterization of Actin-like Protein 1 (ALP1),

a Novel Actin-related Protein in the Apicomplexan Parasite Toxo-

plasma Gondii. PhD thesis, Washington University, 2007. URL

https://books.google.com/books?id=QBn6UdlaHwkC&pgis=1.

[83] O. A. Gorleku, A.-M. Barns, G. R. Prescott, J. Greaves, and

L. H. Chamberlain. Endoplasmic reticulum localization of

DHHC palmitoyltransferases mediated by lysine-based sorting

signals. The Journal of biological chemistry, 286(45):39573–84, Nov.

references 203

2011. ISSN 1083-351X. doi: 10.1074/jbc.M111.272369. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3234780&tool=pmcentrez&rendertype=abstract.

[84] D. C. Greenbaum, A. Baruch, M. Grainger, Z. Bozdech, K. F.

Medzihradszky, J. Engel, J. DeRisi, A. A. Holder, and M. Bogyo.

A role for the protease falcipain 1 in host cell invasion by the hu-

man malaria parasite. Science (New York, N.Y.), 298(5600):2002–6,

Dec. 2002. ISSN 1095-9203. doi: 10.1126/science.1077426. URL

http://www.ncbi.nlm.nih.gov/pubmed/12471262.

[85] S. Guindon and O. Gascuel. A simple, fast, and accurate algo-

rithm to estimate large phylogenies by maximum likelihood.

Systematic biology, 52(5):696–704, Oct. 2003. ISSN 1063-5157.

URL http://www.ncbi.nlm.nih.gov/pubmed/14530136.

[86] W. E. Gutteridge and P. I. Trigg. Action of pyrimethamine

and related drugs against Plasmodium knowlesi in vitro. Para-

sitology, 62(03):431, Apr. 1971. ISSN 0031-1820. doi: 10.1017/

S0031182000077581. URL http://journals.cambridge.org/

abstract

S0031182000077581.

[87] D. S. Guttery, B. Poulin, A. Ramaprasad, R. J. Wall, D. J. P.

Ferguson, D. Brady, E.-M. Patzewitz, S. Whipple, U. Straschil,

M. H. Wright, A. M. A. H. Mohamed, A. Radhakrishnan, S. T.

Arold, E. W. Tate, A. A. Holder, B. Wickstead, A. Pain, and

R. Tewari. Genome-wide functional analysis of Plasmodium

protein phosphatases reveals key regulators of parasite de-

velopment and differentiation. Cell host & microbe, 16(1):128–

40, July 2014. ISSN 1934-6069. doi: 10.1016/j.chom.2014.05.

020. URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=4094981&tool=pmcentrez&rendertype=abstract.

[88] S. Hasenkamp, K. T. Russell, and P. Horrocks. Com-

parison of the absolute and relative efﬁciencies of

electroporation-based transfection protocols for Plas-

modium falciparum. Malaria journal, 11(1):210, 2012.

ISSN 1475-2875. doi: 10.1186/1475-2875-11-210. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3407700&tool=pmcentrez&rendertype=abstract.

[89] C. H. Hastings. Novel malaria parasite proteins involved in

erythrocyte invasion. (December), 2012.

[90] P. W. Hedrick. Population genetics of malaria re-

sistance in humans. Heredity, 107(4):283–304, Oct.

2011. ISSN 1365-2540. doi: 10.1038/hdy.2011.16. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3182497&tool=pmcentrez&rendertype=abstract.

[91] M. Hirai, M. Arai, S. Kawai, and H. Matsuoka. PbGCbeta is

essential for Plasmodium ookinete motility to invade midgut

204 references

cell and for successful completion of parasite life cycle in

mosquitoes. Journal of biochemistry, 140(5):747–57, Nov. 2006.

ISSN 0021-924X. doi: 10.1093/jb/mvj205. URL http://www.

ncbi.nlm.nih.gov/pubmed/17030505.

[92] G. Hu, A. Cabrera, M. Kono, S. Mok, B. K. Chaal, S. Haase,

K. Engelberg, S. Cheemadan, T. Spielmann, P. R. Preiser, T.-W.

Gilberger, and Z. Bozdech. Transcriptional proﬁling of growth

perturbations of the human malaria parasite Plasmodium falci-

parum. Nature biotechnology, 28(1):91–8, Jan. 2010. ISSN 1546-

1696. doi: 10.1038/nbt.1597. URL http://www.ncbi.nlm.nih.

gov/pubmed/20037583.

[93] R. Idro, K. Marsh, C. C. John, and C. R. J. Newton. Cere-

bral malaria: Mechanisms of brain injury and strategies for im-

proved neurocognitive outcome. Pediatric Research, 68(4):267–

274, 2010. ISSN 00313998. doi: 10.1203/PDR.0b013e3181eee738.

[94] Illumina. Speciﬁcations for HiSeq 2500. URL

http://www.illumina.com/systems/hiseq

2500

1500/

performance

specifications.html.

[95] D. Jacot, N. Tosseti, A. Graindorge, and D. Soldati-Favre. Motil-

ity, invasion and egress critically depend on an universal con-

nector that links the actomyosin system with adhesins. In Molec-

ular Parasitology Meeting, 2015.

[96] R. Jambou, F. El-Assaad, V. Combes, and G. E. Grau. In

vitro culture of Plasmodium berghei-ANKA maintains in-

fectivity of mouse erythrocytes inducing cerebral malaria.

Malaria journal, 10(1):346, Jan. 2011. ISSN 1475-2875. doi:

10.1186/1475-2875-10-346. URL http://www.malariajournal.

com/content/10/1/346.

[97] C. J. Janse, B. Franke-Fayard, G. R. Mair, J. Ramesar, C. Thiel,

S. Engelmann, K. Matuschewski, G. J. van Gemert, R. W.

Sauerwein, and A. P. Waters. High efﬁciency transfec-

tion of Plasmodium berghei facilitates novel selection pro-

cedures. Molecular and biochemical parasitology, 145(1):60–70,

Jan. 2006. ISSN 0166-6851. doi: 10.1016/j.molbiopara.2005.09.

007. URL http://www.sciencedirect.com/science/article/

pii/S0166685105002859.

[98] H. Jeong, S. P. Mason, A. L. Barabási, and Z. N. Oltvai. Lethality

and centrality in protein networks. Nature, 411(6833):41–2, May

2001. ISSN 0028-0836. doi: 10.1038/35075138. URL http://dx.

doi.org/10.1038/35075138.

[99] M. L. Jones, E. L. Kitson, and J. C. Rayner. Plasmodium fal-

ciparum erythrocyte invasion: A conserved myosin associated

complex. Molecular and Biochemical Parasitology, 147(1):74–84,

2006. ISSN 01666851. doi: 10.1016/j.molbiopara.2006.01.009.

references 205

[100] S. Jongwutiwes, P. Buppan, R. Kosuvin, S. Seethamchai,

U. Pattanawong, J. Sirichaisinthop, and C. Putaporntip.

Plasmodium knowlesi Malaria in humans and macaques,

Thailand. Emerging infectious diseases, 17(10):1799–806, Oct.

2011. ISSN 1080-6059. doi: 10.3201/eid1710.110349. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3310673&tool=pmcentrez&rendertype=abstract.

[101] D. A. Joy, X. Feng, J. Mu, T. Furuya, K. Chotivanich, A. U.

Krettli, M. Ho, A. Wang, N. J. White, E. Suh, P. Beerli, and

X.-z. Su. Early origin and recent expansion of Plasmodium

falciparum. Science (New York, N.Y.), 300(5617):318–21, Apr.

2003. ISSN 1095-9203. doi: 10.1126/science.1081449. URL http:

//www.sciencemag.org/content/300/5617/318.abstract.

[102] B. F. C. Kafsack, N. Rovira-Graells, T. G. Clark, C. Bancells,

V. M. Crowley, S. G. Campino, A. E. Williams, L. G. Drought,

D. P. Kwiatkowski, D. A. Baker, A. Cortés, and M. Llinás. A

transcriptional switch underlies commitment to sexual devel-

opment in malaria parasites. Nature, 507(7491):248–52, Mar.

2014. ISSN 1476-4687. doi: 10.1038/nature12920. URL http:

//dx.doi.org/10.1038/nature12920.

[103] I. Kaneko, S. Iwanaga, T. Kato, I. Kobayashi, and M. Yuda.

Genome-Wide Identiﬁcation of the Target Genes of AP2-

O, a Plasmodium AP2-Family Transcription Factor. PLoS

pathogens, 11(5):e1004905, May 2015. ISSN 1553-7374. doi: 10.

1371/journal.ppat.1004905. URL http://journals.plos.org/

plospathogens/article?id=10.1371/journal.ppat.1004905.

[104] S. H. I. Kappe, A. M. Vaughan, J. A. Boddey, and A. F. Cowman.

That Was Then But This Is Now : Malaria Research in the Time

of an. Science, 328, 2010.

[105] S. M. Khan, H. Kroeze, B. Franke-Fayard, and C. J. Janse. Stan-

dardization in generating and reporting genetically modiﬁed

rodent malaria parasites: the RMgmDB database. Methods in

molecular biology (Clifton, N.J.), 923:139–50, Jan. 2013. ISSN 1940-

6029. doi: 10.1007/978-1-62703-026-7\_9. URL http://www.

ncbi.nlm.nih.gov/pubmed/22990775.

[106] S. Khuri and S. Wuchty. Essentiality and centrality in protein in-

teraction networks revisited. BMC bioinformatics, 16(1):109, Jan.

2015. ISSN 1471-2105. doi: 10.1186/s12859-015-0536-x. URL

http://www.biomedcentral.com/1471-2105/16/109.

[107] L. A. Kirkman, E. A. Lawrence, and K. W. Deitsch.

Malaria parasites utilize both homologous recombina-

tion and alternative end joining pathways to maintain

genome integrity. Nucleic acids research, 42(1):370–9, Jan.

2014. ISSN 1362-4962. doi: 10.1093/nar/gkt881. URL

206 references

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3874194&tool=pmcentrez&rendertype=abstract.

[108] R. M. Knowles and B. Das Gupta. A study of monkey-malaria

and its experimental transmission to man. Ind Med Gaz, 67:301–

320, 1932.

[109] C. H. M. Kocken, H. Ozwara, A. V. D. Wel, A. L. Beetsma, J. M.

Mwenda, and A. W. Thomas. Plasmodium knowlesi Provides

a Rapid In Vitro and In Vivo Transfection System That Enables

Double-Crossover Gene Knockout Studies. 70(2):655–660, 2002.

doi: 10.1128/IAI.70.2.655.

[110] M. Kono, S. Herrmann, N. B. Loughran, A. Cabrera, K. En-

gelberg, C. Lehmann, D. Sinha, B. Prinz, U. Ruch, V. Heus-

sler, T. Spielmann, J. Parkinson, and T. W. Gilberger. Evolu-

tion and architecture of the inner membrane complex in asex-

ual and sexual stages of the malaria parasite. Molecular biology

and evolution, 29(9):2113–32, Sept. 2012. ISSN 1537-1719. doi:

10.1093/molbev/mss081. URL http://www.ncbi.nlm.nih.gov/

pubmed/22389454.

[111] I. Kozarewa, Z. Ning, M. A. Quail, M. J. Sanders, M. Berri-

man, and D. J. Turner. Ampliﬁcation-free Illumina sequencing-

library preparation facilitates improved mapping and assem-

bly of (G+C)-biased genomes. Nature methods, 6(4):291–5, Apr.

2009. ISSN 1548-7105. doi: 10.1038/nmeth.1311. URL http:

//dx.doi.org/10.1038/nmeth.1311.

[112] U. Krzych, S. Zarling, and A. Pichugin. Memory T cells

maintain protracted protection against malaria. Immunology

letters, 161(2):189–95, Oct. 2014. ISSN 1879-0542. doi: 10.

1016/j.imlet.2014.03.011. URL http://www.ncbi.nlm.nih.gov/

pubmed/24709142.

[113] I. Kursula, P. Kursula, M. Ganter, S. Panjikar, K. Matuschewski,

and H. Schüler. Structural basis for parasite-speciﬁc functions

of the divergent proﬁlin of Plasmodium falciparum. Structure

(London, England : 1993), 16(11):1638–48, Nov. 2008. ISSN 0969-

2126. doi: 10.1016/j.str.2008.09.008. URL http://www.ncbi.nlm.

nih.gov/pubmed/19000816.

[114] D. P. Kwiatkowski. How malaria has affected the human

genome and what human genetics can teach us about

malaria. American journal of human genetics, 77(2):171–92,

Aug. 2005. ISSN 0002-9297. doi: 10.1086/432519. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=1224522&tool=pmcentrez&rendertype=abstract.

[115] a. H. Lee, L. S. Symington, and D. a. Fidock. DNA Repair Mech-

anisms and Their Biological Roles in the Malaria Parasite Plas-

modium falciparum. Microbiology and Molecular Biology Reviews,

references 207

78(3):469–486, Sept. 2014. ISSN 1092-2172. doi: 10.1128/MMBR.

00059-13. URL http://mmbr.asm.org/cgi/doi/10.1128/MMBR.

00059-13.

[116] K.-S. Lee, P. C. S. Divis, S. K. Zakaria, A. Matusop, R. A.

Julin, D. J. Conway, J. Cox-Singh, and B. Singh. Plasmodium

knowlesi: reservoir hosts and tracking the emergence in

humans and macaques. PLoS pathogens, 7(4):e1002015, Apr.

2011. ISSN 1553-7374. doi: 10.1371/journal.ppat.1002015. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3072369&tool=pmcentrez&rendertype=abstract.

[117] Leiden. Rodent Malaria Genetic Modiﬁcation database.

[118] L. Li, C. J. Stoeckert, and D. S. Roos. OrthoMCL: identiﬁcation

of ortholog groups for eukaryotic genomes. Genome research,

13(9):2178–89, Sept. 2003. ISSN 1088-9051. doi: 10.1101/gr.

1224503. URL http://genome.cshlp.org/content/13/9/2178.

full.

[119] M. R. Lieber. The mechanism of double-strand DNA break

repair by the nonhomologous DNA end-joining pathway.

Annual review of biochemistry, 79:181–211, Jan. 2010. ISSN

1545-4509. doi: 10.1146/annurev.biochem.052308.093131. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3079308&tool=pmcentrez&rendertype=abstract.

[120] C. Lim, E. Hansen, T. M. DeSimone, Y. Moreno, K. Junker,

A. Bei, C. Brugnara, C. O. Buckee, and M. T. Duraisingh.

Expansion of host cellular niche can drive adaptation of a

zoonotic malaria parasite to humans. Nature communications,

4:1638, Jan. 2013. ISSN 2041-1723. doi: 10.1038/ncomms2612.

URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=3762474&tool=pmcentrez&rendertype=abstract.

[121] C. Lim, J. Goldberg, A. Griggs, C. Gruring, D. Neafsey, and

M. Duraisingh. Changes in Plasmodium knowlesi invasin lig-

and genes are associatied with adaptation to human red blood

cells. In Molecular Parasitology Meeting, 2015.

[122] J. Limenitakis and D. Soldati-Favre. Functional genetics in Api-

complexa: Potentials and limits. FEBS Letters, 585(11):1579–

1588, 2011. ISSN 00145793. doi: 10.1016/j.febslet.2011.05.002.

URL http://dx.doi.org/10.1016/j.febslet.2011.05.002.

[123] J. Lin. Generation of genetically attenuated blood-stage malaria par-

asites : characterizing growth and virulence in a rodent model of

malaria. PhD thesis, 2013.

[124] J. W. Lin, P. Meireles, M. Prudêncio, S. Engelmann, T. Annoura,

M. Sajid, S. Chevalley-Maurel, J. Ramesar, C. Nahar, C. M. C.

Avramut, A. J. Koster, K. Matuschewski, A. P. Waters, C. J. Janse,

208 references

G. R. Mair, and S. M. Khan. Loss-of-function analyses deﬁnes vi-

tal and redundant functions of the Plasmodium rhomboid pro-

tease family. Molecular Microbiology, 88(2):318–338, 2013. ISSN

0950382X. doi: 10.1111/mmi.12187.

[125] S. E. Lindner, K. E. Swearingen, A. Harupa, A. M. Vaughan,

P. Sinnis, R. L. Moritz, and S. H. I. Kappe. Total and Putative

Surface Proteomics of Malaria Parasite Salivary Gland Sporo-

zoites. Molecular & Cellular Proteomics, 12(5):1127–1143, Jan.

2013. ISSN 1535-9476. doi: 10.1074/mcp.M112.024505. URL

http://www.mcponline.org/content/12/5/1127.full.

[126] W. Liu, Y. Li, G. H. Learn, R. S. Rudicell, J. D. Robertson,

B. F. Keele, J.-B. N. Ndjango, C. M. Sanz, D. B. Morgan,

S. Locatelli, M. K. Gonder, P. J. Kranzusch, P. D. Walsh,

E. Delaporte, E. Mpoudi-Ngole, A. V. Georgiev, M. N. Muller,

G. M. Shaw, M. Peeters, P. M. Sharp, J. C. Rayner, and

B. H. Hahn. Origin of the human malaria parasite Plas-

modium falciparum in gorillas. Nature, 467(7314):420–425,

Sept. 2010. ISSN 0028-0836. doi: 10.1038/nature09442. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=2997044&tool=pmcentrez&rendertype=abstract.

[127] W. Liu, Y. Li, K. S. Shaw, G. H. Learn, L. J. Plenderleith, J. A.

Malenke, S. A. Sundararaman, M. A. Ramirez, P. A. Crystal,

A. G. Smith, F. Bibollet-Ruche, A. Ayouba, S. Locatelli, A. Este-

ban, F. Mouacha, E. Guichet, C. Butel, S. Ahuka-Mundeke, B.-

I. Inogwabini, J.-B. N. Ndjango, S. Speede, C. M. Sanz, D. B.

Morgan, M. K. Gonder, P. J. Kranzusch, P. D. Walsh, A. V.

Georgiev, M. N. Muller, A. K. Piel, F. A. Stewart, M. L. Wil-

son, A. E. Pusey, L. Cui, Z. Wang, A. Färnert, C. J. Suther-

land, D. Nolder, J. A. Hart, T. B. Hart, P. Bertolani, A. Gillis,

M. LeBreton, B. Tafon, J. Kiyang, C. F. Djoko, B. S. Schnei-

der, N. D. Wolfe, E. Mpoudi-Ngole, E. Delaporte, R. Carter,

R. L. Culleton, G. M. Shaw, J. C. Rayner, M. Peeters, B. H.

Hahn, and P. M. Sharp. African origin of the malaria par-

asite Plasmodium vivax. Nature communications, 5:3346, Jan.

2014. ISSN 2041-1723. doi: 10.1038/ncomms4346. URL http:

//europepmc.org/articles/PMC4089193.

[128] M. J. López-Barragán, J. Lemieux, M. Quiñones, K. C.

Williamson, A. Molina-Cruz, K. Cui, C. Barillas-Mury,

K. Zhao, and X.-z. Su. Directional gene expression and

antisense transcripts in sexual and asexual stages of Plas-

modium falciparum. BMC genomics, 12:587, Jan. 2011.

ISSN 1471-2164. doi: 10.1186/1471-2164-12-587. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3266614&tool=pmcentrez&rendertype=abstract.

[129] A. Lorestani, L. Sheiner, K. Yang, S. D. Robertson, N. Sahoo,

C. F. Brooks, D. J. P. Ferguson, B. Striepen, and M.-J. Gubbels.

references 209

A Toxoplasma MORN1 null mutant undergoes repeated divi-

sions but is defective in basal assembly, apicoplast division and

cytokinesis. PloS one, 5(8):e12302, Jan. 2010. ISSN 1932-6203. doi:

10.1371/journal.pone.0012302. URL http://journals.plos.

org/plosone/article?id=10.1371/journal.pone.0012302.

[130] L. J. Mackey, A. Hochmann, C. H. June, C. E. Contr-

eras, and P. H. Lambert. Immunopathological aspects

of Plasmodium berghei infection in ﬁve strains of mice.

II. Immunopathology of cerebral and other tissue le-

sions during the infection. Clinical and experimental im-

munology, 42(3):412–20, Dec. 1980. ISSN 0009-9104. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=1537166&tool=pmcentrez&rendertype=abstract.

[131] M. Madamet. Malaria Prophylaxis Failure with Doxycycline,

Central African Republic, 2014. Emerging Infectious Disease jour-

nal, 21(8), 2015. URL http://wwwnc.cdc.gov/eid/article/21/

8/15-0524

article.

[132] B. Mahajan, D. Jani, R. Chattopadhyay, R. Nagarkatti, H. Zheng,

V. Majam, W. Weiss, S. Kumar, and D. Rathore. Identiﬁcation,

cloning, expression, and characterization of the gene for Plas-

modium knowlesi surface protein containing an altered throm-

bospondin repeat domain. Infection and immunity, 73(9):5402–9,

Sept. 2005. ISSN 0019-9567. doi: 10.1128/IAI.73.9.5402-5409.

2005. URL http://iai.asm.org/content/73/9/5402.full.

[133] P. Mali, K. M. Esvelt, and G. M. Church. Cas9 as a versatile

tool for engineering biology. Nature methods, 10(10):957–63, Oct.

2013. ISSN 1548-7105. doi: 10.1038/nmeth.2649. URL http:

//dx.doi.org/10.1038/nmeth.2649.

[134] B. Malleret, A. Li, R. Zhang, K. S. W. Tan, R. Suwanarusk,

C. Claser, J. S. Cho, E. G. L. Koh, C. S. Chu, S. Pukrit-

tayakamee, M. L. Ng, F. Ginhoux, L. G. Ng, C. T. Lim, F. Nos-

ten, G. Snounou, L. Rénia, and B. Russell. Plasmodium vi-

vax: restricted tropism and rapid remodeling of CD71-positive

reticulocytes. Blood, 125(8):1314–24, Feb. 2015. ISSN 1528-

0020. doi: 10.1182/blood-2014-08-596015. URL http://www.

bloodjournal.org/content/125/8/1314.abstract.

[135] C. B. Mamoun, I. Y. Gluzman, S. Goyard, S. M. Bev-

erley, and D. E. Goldberg. A set of independent se-

lectable markers for transfection of the human malaria

parasite Plasmodium falciparum. Proceedings of the Na-

tional Academy of Sciences of the United States of Amer-

ica, 96(15):8716–20, July 1999. ISSN 0027-8424. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=17582&tool=pmcentrez&rendertype=abstract.

210 references

[136] M. Mani, K. Kandavelou, F. J. Dy, S. Durai, and S. Chan-

drasegaran. Design, engineering, and characterization of zinc

ﬁnger nucleases. Biochemical and Biophysical Research Communi-

cations, 335(2):447–457, 2005. ISSN 0006291X. doi: 10.1016/j.

bbrc.2005.07.089.

[137] L. Martín-Jaular, A. Elizalde-Torrent, R. Thomson-Luque,

M. Ferrer, J. C. Segovia, E. Herreros-Aviles, C. Fernández-

Becerra, and H. A. del Portillo. Reticulocyte-prone malaria par-

asites predominantly invade CD71hi immature cells: implica-

tions for the development of an in vitro culture for Plasmod-

ium vivax. Malaria Journal, 12(1):434, 2013. ISSN 1475-2875. doi:

10.1186/1475-2875-12-434. URL http://www.malariajournal.

com/content/12/1/434.

[138] M. T. McIntosh, A. Vaid, H. D. Hosgood, J. Vijay, A. Bhat-

tacharya, M. H. Sahani, P. Baevova, K. A. Joiner, and P. Sharma.

Trafﬁc to the malaria parasite food vacuole: a novel path-

way involving a phosphatidylinositol 3-phosphate-binding pro-

tein. The Journal of biological chemistry, 282(15):11499–508, Apr.

2007. ISSN 0021-9258. doi: 10.1074/jbc.M610974200. URL

http://www.jbc.org/content/282/15/11499.

[139] D. Ménard, C. Barnadas, C. Bouchier, C. Henry-Halldin, L. R.

Gray, A. Ratsimbasoa, V. Thonier, J.-F. Carod, O. Domarle,

Y. Colin, O. Bertrand, J. Picot, C. L. King, B. T. Grimberg,

O. Mercereau-Puijalon, and P. A. Zimmerman. Plasmodium

vivax clinical malaria is commonly observed in Duffy-negative

Malagasy people. Proceedings of the National Academy of

Sciences of the United States of America, 107(13):5967–71, Mar.

2010. ISSN 1091-6490. doi: 10.1073/pnas.0912496107. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=2851935&tool=pmcentrez&rendertype=abstract.

[140] E. V. S. Meyer, A. A. Semenya, D. M. N. Okenu, R. Anton, L. H.

Bannister, J. W. Barnwell, and M. R. Galinski. The reticulocyte

binding-like proteins of P. knowlesi locate to the micronemes

of merozoites and deﬁne two new members of this invasion

ligand family. Mol Biochem Parasitol, 165(2):111–121, 2009. doi:

10.1016/j.molbiopara.2009.01.012.The.

[141] L. H. Miller, S. J. Mason, D. F. Clyde, and M. H. McGinniss.

The Resistance Factor to Plasmodium vivax in Blacks. New

England Journal of Medicine, 295(6):302–304, Aug. 1976. ISSN

0028-4793. doi: 10.1056/NEJM197608052950602. URL http:

//www.ncbi.nlm.nih.gov/pubmed/778616.

[142] R. W. Moon, J. Hall, F. Rangkuti, Y. S. Ho, N. Almond, G. H.

Mitchell, A. Pain, A. a. Holder, and M. J. Blackman. Adapta-

tion of the genetically tractable malaria pathogen Plasmodium

references 211

knowlesi to continuous culture in human erythrocytes. Pro-

ceedings of the National Academy of Sciences of the United States of

America, 110(2):531–6, Jan. 2013. ISSN 1091-6490. doi: 10.1073/

pnas.1216457110. URL http://www.ncbi.nlm.nih.gov/pubmed/

23267069.

[143] S. Moonah, N. G. Sanders, J. K. Persichetti, and

D. J. Sullivan. Erythrocyte lysis and Xenopus lae-

vis oocyte rupture by recombinant Plasmodium falci-

parum hemolysin III. Eukaryotic cell, 13(10):1337–45, Oct.

2014. ISSN 1535-9786. doi: 10.1128/EC.00088-14. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=4187644&tool=pmcentrez&rendertype=abstract.

[144] R. R. Moraes Barros, J. Straimer, J. M. Sa, R. E. Salzman, V. A.

Melendez-Muniz, J. Mu, D. A. Fidock, and T. E. Wellems. Edit-

ing the Plasmodium vivax genome, using zinc-ﬁnger nucleases.

The Journal of infectious diseases, 211(1):125–9, Jan. 2015. ISSN

1537-6613. doi: 10.1093/infdis/jiu423. URL http://www.ncbi.

nlm.nih.gov/pubmed/25081932.

[145] E. J. Mui, G. A. Schiehser, W. K. Milhous, H. Hsu, C. W.

Roberts, M. Kirisits, S. Muench, D. Rice, J. P. Dubey, J. W. Fow-

ble, P. K. Rathod, S. F. Queener, S. R. Liu, D. P. Jacobus, and

R. McLeod. Novel triazine JPC-2067-B inhibits Toxoplasma

gondii in vitro and in vivo. PLoS neglected tropical diseases, 2

(3):e190, Jan. 2008. ISSN 1935-2735. doi: 10.1371/journal.pntd.

0000190. URL http://journals.plos.org/plosntds/article?

id=10.1371/journal.pntd.0000190.

[146] C. Nusslein-Volhard. The identiﬁcation of genes controlling de-

velopment in ﬂies and ﬁshes. In Nobel Lecture, 1995.

[147] R. A. O’Donnell, L. H. Freitas-Junior, P. R. Preiser, D. H.

Williamson, M. Duraisingh, T. F. McElwain, A. Scherf, A. F.

Cowman, and B. S. Crabb. A genetic screen for improved plas-

mid segregation reveals a role for Rep20 in the interaction of

Plasmodium falciparum chromosomes. The EMBO journal, 21

(5):1231–9, Mar. 2002. ISSN 0261-4189. doi: 10.1093/emboj/

21.5.1231. URL http://emboj.embopress.org/content/21/5/

1231.abstract.

[148] R. A. O’Donnell, F. Hackett, S. A. Howell, M. Treeck, N. Struck,

Z. Krnajski, C. Withers-Martinez, T. W. Gilberger, and M. J.

Blackman. Intramembrane proteolysis mediates shedding of

a key adhesin during erythrocyte invasion by the malaria par-

asite. The Journal of cell biology, 174(7):1023–33, Sept. 2006.

ISSN 0021-9525. doi: 10.1083/jcb.200604136. URL http://jcb.

rupress.org/content/174/7/1023.

[149] W. P. O’Meara, J. N. Mangeni, R. Steketee, and B. Greenwood.

Changes in the burden of malaria in sub-Saharan Africa. The

212 references

Lancet. Infectious diseases, 10(8):545–55, Aug. 2010. ISSN 1474-

4457. doi: 10.1016/S1473-3099(10)70096-7. URL http://www.

ncbi.nlm.nih.gov/pubmed/20637696.

[150] T. D. Otto, U. Böhme, A. P. Jackson, M. Hunt, B. Franke-Fayard,

W. A. M. Hoeijmakers, A. A. Religa, L. Robertson, M. Sanders,

S. A. Ogun, D. Cunningham, A. Erhart, O. Billker, S. M. Khan,

H. G. Stunnenberg, J. Langhorne, A. A. Holder, A. P. Waters,

C. I. Newbold, A. Pain, M. Berriman, and C. J. Janse. A com-

prehensive evaluation of rodent malaria parasite genomes and

gene expression. BMC biology, 12:86, Jan. 2014. ISSN 1741-7007.

doi: 10.1186/s12915-014-0086-0. URL http://europepmc.org/

articles/PMC4242472/?report=abstract.

[151] a. Pain, U. Böhme, a. E. Berry, K. Mungall, R. D. Finn, a. P.

Jackson, T. Mourier, J. Mistry, E. M. Pasini, M. a. Aslett,

S. Balasubrammaniam, K. Borgwardt, K. Brooks, C. Carret,

T. J. Carver, I. Cherevach, T. Chillingworth, T. G. Clark, M. R.

Galinski, N. Hall, D. Harper, D. Harris, H. Hauser, A. Ivens,

C. S. Janssen, T. Keane, N. Larke, S. Lapp, M. Marti, S. Moule,

I. M. Meyer, D. Ormond, N. Peters, M. Sanders, S. Sanders,

T. J. Sargeant, M. Simmonds, F. Smith, R. Squares, S. Thurston,

a. R. Tivey, D. Walker, B. White, E. Zuiderwijk, C. Churcher,

M. a. Quail, a. F. Cowman, C. M. R. Turner, M. a. Rajandream,

C. H. M. Kocken, a. W. Thomas, C. I. Newbold, B. G. Barrell,

and M. Berriman. The genome of the simian and human

malaria parasite Plasmodium knowlesi. Nature, 455(7214):

799–803, Oct. 2008. ISSN 1476-4687. doi: 10.1038/nature07306.

URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=2656934&tool=pmcentrez&rendertype=abstract.

[152] I. Pal-Bhowmick, J. Andersen, P. Srinivasan, D. L. Narum,

J. Bosch, and L. H. Miller. Binding of aldolase and

glyceraldehyde-3-phosphate dehydrogenase to the cyto-

plasmic tails of Plasmodium falciparum merozoite duffy

binding-like and reticulocyte homology ligands. mBio, 3(5),

Jan. 2012. ISSN 2150-7511. doi: 10.1128/mBio.00292-12. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3448169&tool=pmcentrez&rendertype=abstract.

[153] K. C. Pandey, N. Singh, S. Arastu-Kapur, M. Bogyo,

and P. J. Rosenthal. Falstatin, a cysteine protease in-

hibitor of Plasmodium falciparum, facilitates erythro-

cyte invasion. PLoS pathogens, 2(11):e117, Nov. 2006.

ISSN 1553-7374. doi: 10.1371/journal.ppat.0020117. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=1630708&tool=pmcentrez&rendertype=abstract.

[154] C. Pfander, B. Anar, F. Schwach, T. D. Otto, M. Brochet, K. Volk-

mann, M. A. Quail, A. Pain, B. Rosen, W. Skarnes, J. C. Rayner,

and O. Billker. A scalable pipeline for highly effective genetic

references 213

modiﬁcation of a malaria parasite. Nature methods, 8(12):4–10,

2011. doi: 10.1038/nMeth.1742.

[155] C. Pfander, B. Anar, F. Schwach, T. D. Otto, M. Brochet, K. Volk-

mann, M. A. Quail, A. Pain, B. Rosen, W. Skarnes, J. C. Rayner,

and O. Billker. A scalable pipeline for highly effective genetic

modiﬁcation of a malaria parasite. Nature methods, 8(12):1078–

82, Dec. 2011. ISSN 1548-7105. doi: 10.1038/nmeth.1742. URL

http://dx.doi.org/10.1038/nmeth.1742.

[156] M. M. Pinheiro, M. A. Ahmed, S. B. Millar, T. Sanderson, T. D.

Otto, W. C. Lu, S. Krishna, J. C. Rayner, and J. Cox-Singh.

Plasmodium knowlesi genome sequences from clinical isolates

reveal extensive genomic dimorphism. PloS one, 10(4):e0121303,

Jan. 2015. ISSN 1932-6203. doi: 10.1371/journal.pone.0121303.

URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=4382175&tool=pmcentrez&rendertype=abstract.

[157] PlasmoDB. PlasmoDB: An integrative database of the

Plasmodium falciparum genome. Tools for accessing and

analyzing ﬁnished and unﬁnished sequence data. The

Plasmodium Genome Database Collaborative. Nucleic

acids research, 29(1):66–9, Jan. 2001. ISSN 1362-4962. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=29846&tool=pmcentrez&rendertype=abstract.

[158] R. N. Price, E. Tjitra, C. a. Guerra, S. Yeung, N. J. White,

and N. M. Anstey. Vivax malaria: neglected and not

benign. The American journal of tropical medicine and hy-

giene, 77(6 Suppl):79–87, Dec. 2007. ISSN 0002-9637. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=2653940&tool=pmcentrez&rendertype=abstract.

[159] M. Prudêncio, A. Rodriguez, and M. M. Mota. The silent path

to thousands of merozoites: the Plasmodium liver stage. Na-

ture Reviews Microbiology, 4(11):849–856, Nov. 2006. ISSN 1740-

1526. doi: 10.1038/nrmicro1529. URL http://www.nature.com/

doifinder/10.1038/nrmicro1529.

[160] D. Quammem. Spillover: Animal Infections and the Next Human

Pandemic. 2012.

[161] N. B. Quashie, H. P. de Koning, and L. C. Ranford-Cartwright.

An improved and highly sensitive microﬂuorimetric method

for assessing susceptibility of Plasmodium falciparum to an-

timalarial drugs in vitro. Malaria journal, 5(1):95, Jan. 2006.

ISSN 1475-2875. doi: 10.1186/1475-2875-5-95. URL http://www.

malariajournal.com/content/5/1/95.

[162] J. C. Rayner, E. Vargas-serrato, C. S. Huber, M. R. Galinski, and

J. W. Barnwell. A Plasmodium falciparum Homologue of Plas-

modium vivax Reticulocyte Binding Protein ( PvRBP1 ) Deﬁnes

214 references

a Trypsin-resistant Erythrocyte Invasion Pathway. Journal of Ex-

perimental Medicine, 194(11), 2001.

[163] K. G. L. Roch, Y. Zhou, P. L. Blair, M. Grainger, J. K. Moch, J. D.

Haynes, P. D. Vega, A. A. Holder, S. Batalov, D. J. Carucci, and

E. A. Winzeler. Discovery of Gene Function by. 301(September):

1503–1508, 2003.

[164] C. V. Rooyen and G. R. Pile. Observations on infection by plas-

modium knowlesi ( ape malaria) in the treatment of general

paralysis of the insane. pages 662–666, 1935.

[165] M. Rougemont, M. V. Saanen, R. Sahli, H. P. Hinrikson, J. Bille,

and K. Jaton. Detection of Four Plasmodium Species in Blood

from Humans by 18S rRNA Gene Subunit-Based and Species-

Speciﬁc Real-Time PCR Assays. 42(12):5636–5643, 2004. doi:

10.1128/JCM.42.12.5636.

[166] K. J. Roux, D. I. Kim, M. Raida, and B. Burke. A promiscuous

biotin ligase fusion protein identiﬁes proximal and interacting

proteins in mammalian cells. The Journal of cell biology, 196(6):

801–10, Mar. 2012. ISSN 1540-8140. doi: 10.1083/jcb.201112098.

URL http://jcb.rupress.org/content/196/6/801.full.

[167] N. Roy, S. Bhattacharyya, S. Chakrabarty, S. Laskar, S. M. Babu,

and M. K. Bhattacharyya. Dominant negative mutant of Plas-

modium Rad51 causes reduced parasite burden in host by abro-

gating DNA double-strand break repair. Molecular Microbiology,

94(2):353–366, 2014. ISSN 0950382X. doi: 10.1111/mmi.12762.

URL http://doi.wiley.com/10.1111/mmi.12762.

[168] S. C. T. P. RTS. Efﬁcacy and safety of RTS,S/AS01 malaria

vaccine with or without a booster dose in infants and chil-

dren in Africa: ﬁnal results of a phase 3, individually ran-

domised, controlled trial. The Lancet, 386(9988):31–45, Apr. 2015.

ISSN 01406736. doi: 10.1016/S0140-6736(15)60721-8. URL http:

//www.thelancet.com/article/S0140673615607218/fulltext.

[169] N. Rudin, E. Sugarman, and J. E. Haber. Genetic

and physical analysis of double-strand break repair

and recombination in Saccharomyces cerevisiae. Ge-

netics, 122(3):519–34, July 1989. ISSN 0016-6731. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=1203726&tool=pmcentrez&rendertype=abstract.

[170] A. Ruecker, M. Shea, F. Hackett, C. Suarez, E. M. A. Hirst, K. Mi-

lutinovic, C. Withers-Martinez, and M. J. Blackman. Proteolytic

activation of the essential parasitophorous vacuole cysteine pro-

tease SERA6 accompanies malaria parasite egress from its host

erythrocyte. The Journal of biological chemistry, 287(45):37949–63,

Nov. 2012. ISSN 1083-351X. doi: 10.1074/jbc.M112.400820. URL

references 215

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3488066&tool=pmcentrez&rendertype=abstract.

[171] P. Rumore, B. Muralidhar, M. Lin, C. Lai, and C. R.

Steinman. Haemodialysis as a model for studying

endogenous plasma DNA: oligonucleosome-like struc-

ture and clearance. Clinical and experimental immunol-

ogy, 90(1):56–62, Oct. 1992. ISSN 0009-9104. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=1554541&tool=pmcentrez&rendertype=abstract.

[172] R. A. Seder, L.-J. Chang, M. E. Enama, K. L. Zephir, U. N. Sar-

war, I. J. Gordon, L. A. Holman, E. R. James, P. F. Billingsley,

A. Gunasekera, A. Richman, S. Chakravarty, A. Manoj, S. Vel-

murugan, M. Li, A. J. Ruben, T. Li, A. G. Eappen, R. E. Stafford,

S. H. Plummer, C. S. Hendel, L. Novik, P. J. M. Costner, F. H.

Mendoza, J. G. Saunders, M. C. Nason, J. H. Richardson, J. Mur-

phy, S. A. Davidson, T. L. Richie, M. Sedegah, A. Sutamihardja,

G. A. Fahle, K. E. Lyke, M. B. Laurens, M. Roederer, K. Tewari,

J. E. Epstein, B. K. L. Sim, J. E. Ledgerwood, B. S. Graham, and

S. L. Hoffman. Protection against malaria by intravenous im-

munization with a nonreplicating sporozoite vaccine. Science

(New York, N.Y.), 341(6152):1359–65, Sept. 2013. ISSN 1095-9203.

doi: 10.1126/science.1241800. URL http://www.ncbi.nlm.nih.

gov/pubmed/23929949.

[173] B. Shen and L. D. Sibley. Toxoplasma aldolase is required for

metabolism but dispensable for host-cell invasion. Proceedings

of the National Academy of Sciences, 111(9):3567–3572, Mar.

2014. ISSN 0027-8424. doi: 10.1073/pnas.1315156111. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3948255&tool=pmcentrez&rendertype=abstract.

[174] P. S. Sijwali, K. Kato, K. B. Seydel, J. Gut, J. Lehman, M. Klemba,

D. E. Goldberg, L. H. Miller, and P. J. Rosenthal. Plasmodium

falciparum cysteine protease falcipain-1 is not essential in

erythrocytic stage malaria parasites. Proceedings of the National

Academy of Sciences of the United States of America, 101(23):8721–6,

June 2004. ISSN 0027-8424. doi: 10.1073/pnas.0402738101. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=423262&tool=pmcentrez&rendertype=abstract.

[175] J. C. Silva, A. Egan, R. Friedman, J. B. Munro, J. M. Carlton,

and A. L. Hughes. Genome sequences reveal divergence times

of malaria parasite lineages. Parasitology, 138(13):1737–49, Nov.

2011. ISSN 1469-8161. doi: 10.1017/S0031182010001575. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3081533&tool=pmcentrez&rendertype=abstract.

[176] B. Singh, L. Kim Sung, A. Matusop, A. Radhakrishnan, S. S. G.

Shamsul, J. Cox-Singh, A. Thomas, and D. J. Conway. A large

216 references

focus of naturally acquired Plasmodium knowlesi infections in

human beings. Lancet, 363(9414):1017–24, Mar. 2004. ISSN 1474-

547X. doi: 10.1016/S0140-6736(04)15836-4. URL http://www.

ncbi.nlm.nih.gov/pubmed/15051281.

[177] S. Singh, M. M. Alam, I. Pal-Bhowmick, J. A. Brzostowski,

and C. E. Chitnis. Distinct external signals trigger sequen-

tial release of apical organelles during erythrocyte invasion

by malaria parasites. PLoS pathogens, 6(2):e1000746, Feb.

2010. ISSN 1553-7374. doi: 10.1371/journal.ppat.1000746.

URL http://journals.plos.org/plospathogens/article?id=

10.1371/journal.ppat.1000746.

[178] A. Sinha, K. R. Hughes, K. K. Modrzynska, T. D. Otto, C. Pfan-

der, N. J. Dickens, A. A. Religa, E. Bushell, A. L. Graham,

R. Cameron, B. F. C. Kafsack, A. E. Williams, M. Llinás, M. Ber-

riman, O. Billker, and A. P. Waters. A cascade of DNA-binding

proteins for sexual commitment and development in Plasmod-

ium. Nature, 507(7491):253–7, Mar. 2014. ISSN 1476-4687.

doi: 10.1038/nature12970. URL http://dx.doi.org/10.1038/

nature12970.

[179] S. Sinha, B. Medhi, and R. Sehgal. Challenges of drug-

resistant malaria. Parasite (Paris, France), 21:61, Jan. 2014.

ISSN 1776-1042. doi: 10.1051/parasite/2014059. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=4234044&tool=pmcentrez&rendertype=abstract.

[180] M. E. Sinka, M. J. Bangs, S. Manguin, T. Chareonviriyaphap,

A. P. Patil, W. H. Temperley, P. W. Gething, I. R. F. Elyazar, C. W.

Kabaria, R. E. Harbach, and S. I. Hay. The dominant Anopheles

vectors of human malaria in the Asia-Paciﬁc region: occurrence

data, distribution maps and bionomic précis. Parasites & vectors,

4(1):89, Jan. 2011. ISSN 1756-3305. doi: 10.1186/1756-3305-4-89.

URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=3127851&tool=pmcentrez&rendertype=abstract.

[181] L. Solyakov, J. Halbert, M. M. Alam, J.-P. Semblat, D. Dorin-

Semblat, L. Reininger, A. R. Bottrill, S. Mistry, A. Abdi, C. Fen-

nell, Z. Holland, C. Demarta, Y. Bouza, A. Sicard, M.-P. Nivez,

S. Eschenlauer, T. Lama, D. C. Thomas, P. Sharma, S. Agar-

wal, S. Kern, G. Pradel, M. Graciotti, A. B. Tobin, and C. Do-

erig. Global kinomic and phospho-proteomic analyses of the

human malaria parasite Plasmodium falciparum. Nature com-

munications, 2:565, Jan. 2011. ISSN 2041-1723. doi: 10.1038/

ncomms1558. URL http://dx.doi.org/10.1038/ncomms1558.

[182] M. D. Spring, J. T. Lin, J. E. Manning, P. Vanachayangkul,

S. Somethy, R. Bun, Y. Se, S. Chann, M. Ittiverakul, P. Sia-

ngam, W. Kuntawunginn, M. Arsanok, N. Buathong, S. Chao-

rattanakawee, P. Gosi, W. Ta-aksorn, N. Chanarat, S. Sun-

drakes, N. Kong, T. K. Heng, S. Nou, P. Teja-isavadharm,

references 217

S. Pichyangkul, S. T. Phann, S. Balasubramanian, J. J. Juliano,

S. R. Meshnick, C. M. Chour, S. Prom, C. A. Lanteri, C. Lon,

and D. L. Saunders. Dihydroartemisinin-piperaquine failure as-

sociated with a triple mutant including kelch13 C580Y in Cam-

bodia: an observational cohort study. The Lancet. Infectious dis-

eases, 15(6):683–91, June 2015. ISSN 1474-4457. doi: 10.1016/

S1473-3099(15)70049-6. URL http://www.ncbi.nlm.nih.gov/

pubmed/25877962.

[183] A. Srivastava, S. Singh, S. Dhawan, M. Mahmood Alam,

A. Mohmmed, and C. E. Chitnis. Localization of apical sushi

protein in Plasmodium falciparum merozoites. Molecular and

biochemical parasitology, 174(1):66–9, Nov. 2010. ISSN 1872-9428.

doi: 10.1016/j.molbiopara.2010.06.003. URL http://www.ncbi.

nlm.nih.gov/pubmed/20540969.

[184] J. Straimer, M. C. S. Lee, A. H. Lee, B. Zeitler, A. E. Williams,

J. R. Pearl, L. Zhang, E. J. Rebar, P. D. Gregory, M. Llinás, F. D.

Urnov, and D. a. Fidock. Site-Speciﬁc Editing of the Plasmod-

ium falciparum Genome Using Engineered Zinc-Finger Nucle-

ases. Nature methods, 9(10):993–998, 2013. doi: 10.1038/nmeth.

2143.Site-Speciﬁc.

[185] A. Sturm, R. Amino, C. van de Sand, T. Regen, S. Retzlaff,

A. Rennenberg, A. Krueger, J.-M. Pollok, R. Menard, and V. T.

Heussler. Manipulation of host hepatocytes by the malaria

parasite for delivery into liver sinusoids. Science (New York,

N.Y.), 313(5791):1287–90, Sept. 2006. ISSN 1095-9203. doi:

10.1126/science.1129720. URL http://www.ncbi.nlm.nih.gov/

pubmed/16888102.

[186] C. J. Sutherland, N. Tanomsing, D. Nolder, M. Oguike, C. Jenni-

son, S. Pukrittayakamee, C. Dolecek, T. T. Hien, V. E. do Rosário,

A. P. Arez, J. a. Pinto, P. Michon, A. a. Escalante, F. Nosten,

M. Burke, R. Lee, M. Blaze, T. D. Otto, J. W. Barnwell, A. Pain,

J. Williams, N. J. White, N. P. J. Day, G. Snounou, P. J. Lockhart,

P. L. Chiodini, M. Imwong, and S. D. Polley. Two nonrecombin-

ing sympatric forms of the human malaria parasite Plasmod-

ium ovale occur globally. The Journal of infectious diseases, 201

(10):1544–50, May 2010. ISSN 1537-6613. doi: 10.1086/652240.

URL http://www.ncbi.nlm.nih.gov/pubmed/20380562.

[187] T. Taechalertpaisarn, C. Crosnier, S. J. Bartholdson, A. N. Hod-

der, J. Thompson, L. Y. Bustamante, D. W. Wilson, P. R. Sanders,

G. J. Wright, J. C. Rayner, A. F. Cowman, P. R. Gilson, and B. S.

Crabb. Biochemical and functional analysis of two Plasmodium

falciparum blood-stage 6-cys proteins: P12 and P41. PloS one, 7

(7):e41937, Jan. 2012. ISSN 1932-6203. doi: 10.1371/journal.pone.

0041937. URL http://journals.plos.org/plosone/article?

id=10.1371/journal.pone.0041937.

218 references

[188] C. L. Tay. The targets and role of palmitoylation in Plasmodium

parasites. PhD thesis, 2013.

[189] R. Tewari, U. Straschil, A. Bateman, U. Böhme, I. Cherevach,

P. Gong, A. Pain, and O. Billker. The systematic functional

analysis of Plasmodium protein kinases identiﬁes essential reg-

ulators of mosquito transmission. Cell host & microbe, 8(4):377–

87, Oct. 2010. ISSN 1934-6069. doi: 10.1016/j.chom.2010.09.

006. URL http://www.pubmedcentral.nih.gov/articlerender.

fcgi?artid=2977076&tool=pmcentrez&rendertype=abstract.

[190] M. Theron, R. L. Hesketh, S. Subramanian, and J. C. Rayner.

An adaptable two-color ﬂow cytometric assay to quantitate

the invasion of erythrocytes by Plasmodium falciparum

parasites. Cytometry. Part A : the journal of the Interna-

tional Society for Analytical Cytology, 77(11):1067–74, Nov.

2010. ISSN 1552-4930. doi: 10.1002/cyto.a.20972. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3047707&tool=pmcentrez&rendertype=abstract.

[191] J. Thompson, R. E. Cooke, S. Moore, L. F. Anderson, C. J.

Janse, and A. P. Waters. PTRAMP; a conserved Plasmod-

ium thrombospondin-related apical merozoite protein. Molec-

ular and biochemical parasitology, 134(2):225–32, Apr. 2004. ISSN

0166-6851. doi: 10.1016/j.molbiopara.2003.12.003. URL http:

//www.ncbi.nlm.nih.gov/pubmed/15003842.

[192] J. Thurston. The action of antimalarial drugs in mice infected

with Plasmodium berghei. British journal of pharmacology and

chemotherapy, 5(3):409–16, Sept. 1950. ISSN 0366-0826. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=1509934&tool=pmcentrez&rendertype=abstract.

[193] C. J. Tonkin, G. G. van Dooren, T. P. Spurck, N. S. Struck,

R. T. Good, E. Handman, A. F. Cowman, and G. I. McFad-

den. Localization of organellar proteins in Plasmodium fal-

ciparum using a novel set of transfection vectors and a new

immunoﬂuorescence ﬁxation method. Molecular and biochemi-

cal parasitology, 137(1):13–21, Sept. 2004. ISSN 0166-6851. doi:

10.1016/j.molbiopara.2004.05.009. URL http://www.ncbi.nlm.

nih.gov/pubmed/15279947.

[194] W. Trager and J. B. Jensen. Human malaria parasites in continu-

ous culture. Science (New York, N.Y.), 193(4254):673–5, Aug. 1976.

ISSN 0036-8075. URL http://www.ncbi.nlm.nih.gov/pubmed/

781840.

[195] A. Z. Tremp, F. S. Al-Khattaf, and J. T. Dessens. Distinct

temporal recruitment of Plasmodium alveolins to the subpel-

licular network. Parasitology research, 113(11):4177–88, Nov.

2014. ISSN 1432-1955. doi: 10.1007/s00436-014-4093-4. URL

references 219

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=4200347&tool=pmcentrez&rendertype=abstract.

[196] M. Tufet-Bayona, C. J. Janse, S. M. Khan, A. P. Waters, R. E.

Sinden, and B. Franke-Fayard. Localisation and timing of ex-

pression of putative Plasmodium berghei rhoptry proteins in

merozoites and sporozoites. Molecular and biochemical parasitol-

ogy, 166(1):22–31, July 2009. ISSN 1872-9428. doi: 10.1016/j.

molbiopara.2009.02.009. URL http://www.ncbi.nlm.nih.gov/

pubmed/19428669.

[197] P. Uzureau, J.-C. Barale, C. J. Janse, A. P. Waters, and C. B.

Breton. Gene targeting demonstrates that the Plasmodium

berghei subtilisin PbSUB2 is essential for red cell invasion

and reveals spontaneous genetic recombination events. Cellu-

lar microbiology, 6(1):65–78, Jan. 2004. ISSN 1462-5814. URL

http://www.ncbi.nlm.nih.gov/pubmed/14678331.

[198] M. R. van Dijk, B. C. L. van Schaijk, S. M. Khan, M. W.

van Dooren, J. Ramesar, S. Kaczanowski, G.-J. van Gemert,

H. Kroeze, H. G. Stunnenberg, W. M. Eling, R. W. Sauer-

wein, A. P. Waters, and C. J. Janse. Three members of

the 6-cys protein family of Plasmodium play a role in

gamete fertility. PLoS pathogens, 6(4):e1000853, Apr. 2010.

ISSN 1553-7374. doi: 10.1371/journal.ppat.1000853. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=2851734&tool=pmcentrez&rendertype=abstract.

[199] A. v.d Wel, C. H. Kocken, T. C. Pronk, B. Franke-Fayard, and

A. W. Thomas. New selectable markers and single crossover

integration for the highly versatile Plasmodium knowlesi trans-

fection system. Molecular and Biochemical Parasitology, 134(1):

97–104, Mar. 2004. ISSN 01666851. doi: 10.1016/j.molbiopara.

2003.10.019. URL http://www.sciencedirect.com/science/

article/pii/S0166685103003311.

[200] S. Vinayak, M. T. Alam, T. Mixson-Hayden, A. M. McCollum,

R. Sem, N. K. Shah, P. Lim, S. Muth, W. O. Rogers, T. Fandeur,

J. W. Barnwell, A. A. Escalante, C. Wongsrichanalai, F. Ariey,

S. R. Meshnick, and V. Udhayakumar. Origin and evolu-

tion of sulfadoxine resistant Plasmodium falciparum. PLoS

pathogens, 6(3):e1000830, Mar. 2010. ISSN 1553-7374. doi:

10.1371/journal.ppat.1000830. URL http://dx.plos.org/10.

1371/journal.ppat.1000830.

[201] J. C. Wagner, R. J. Platt, S. J. Goldﬂess, F. Zhang, and J. C. Niles.

Efﬁcient CRISPR-Cas9-mediated genome editing in Plasmod-

ium falciparum. Nature Methods, 11(August):1–6, 2014. ISSN

1548-7091. doi: 10.1038/nmeth.3063. URL http://www.nature.

com/doifinder/10.1038/nmeth.3063.

220 references

[202] Y. Wang, L. Y. Geer, C. Chappey, J. A. Kans, and S. H. Bryant.

Cn3D: sequence and structure views for Entrez. Trends in bio-

chemical sciences, 25(6):300–2, June 2000. ISSN 0968-0004. URL

http://www.ncbi.nlm.nih.gov/pubmed/10838572.

[203] Wel. New selectable markers and single crossover integration

for the highly versatile Plasmodium knowlesi transfection sys-

tem. - PubMed - NCBI. URL http://www.ncbi.nlm.nih.gov/

pubmed/14747147.

[204] S. P. Wertheimer and A. W. Barnwell. Plasmodium vivax inter-

action with the Human Duffy Blood Group Glycoprotein : Iden-

tiﬁcation of a Parasite Receptor-like Protein sera against the

Fya Duffy determinant and proteolytic cleavage of the Duffy

determi- nants from erythrocytes to inhibit P . ( . 350:340–350,

1989.

[205] T. William, H. a. Rahman, J. Jelip, M. Y. Ibrahim, J. Menon, M. J.

Grigg, T. W. Yeo, N. M. Anstey, and B. E. Barber. Increasing

Incidence of Plasmodium knowlesi Malaria following Control

of P. falciparum and P. vivax Malaria in Sabah, Malaysia. PLoS

neglected tropical diseases, 7(1):e2026, Jan. 2013. ISSN 1935-2735.

doi: 10.1371/journal.pntd.0002026. URL http://www.ncbi.nlm.

nih.gov/pubmed/23359830.

[206] C. C. Wirth and G. Pradel. Molecular mechanisms of host

cell egress by malaria parasites. International journal of medi-

cal microbiology : IJMM, 302(4-5):172–8, Oct. 2012. ISSN 1618-

0607. doi: 10.1016/j.ijmm.2012.07.003. URL http://www.ncbi.

nlm.nih.gov/pubmed/22951233.

[207] G. J. Wright and J. C. Rayner. Plasmodium falciparum

erythrocyte invasion: combining function with immune

evasion. PLoS pathogens, 10(3):e1003943, Mar. 2014. ISSN

1553-7374. doi: 10.1371/journal.ppat.1003943. URL

http://www.pubmedcentral.nih.gov/articlerender.fcgi?

artid=3961354&tool=pmcentrez&rendertype=abstract.

[208] Y. Wu, C. D. Sifri, H. H. Lei, X. Z. Su, and T. E. Wellems. Trans-

fection of Plasmodium falciparum within human red blood

cells. Proceedings of the National Academy of Sciences, 92(4):973–

977, Feb. 1995. ISSN 0027-8424. doi: 10.1073/pnas.92.4.973. URL

http://www.pnas.org/content/92/4/973.abstract.

[209] S. Yeoh, R. A. O’Donnell, K. Koussis, A. R. Dluzewski, K. H.

Ansell, S. A. Osborne, F. Hackett, C. Withers-Martinez, G. H.

Mitchell, L. H. Bannister, J. S. Bryans, C. A. Kettleborough,

and M. J. Blackman. Subcellular discharge of a serine pro-

tease mediates release of invasive malaria parasites from host

erythrocytes. Cell, 131(6):1072–83, Dec. 2007. ISSN 0092-8674.

doi: 10.1016/j.cell.2007.10.049. URL http://www.ncbi.nlm.nih.

gov/pubmed/18083098.

references 221

[210] M. Yuda, S. Iwanaga, S. Shigenobu, G. R. Mair, C. J. Janse,

A. P. Waters, T. Kato, and I. Kaneko. Identiﬁcation of a tran-

scription factor in the mosquito-invasive stage of malaria par-

asites. Molecular microbiology, 71(6):1402–14, Mar. 2009. ISSN

1365-2958. doi: 10.1111/j.1365-2958.2009.06609.x. URL http:

//www.ncbi.nlm.nih.gov/pubmed/19220746.

[211] N. A. Yusuf, J. L. Green, R. J. Wall, E. Knuepfer, R. W. Moon,

C. Schulte-Huxel, R. R. Stanway, S. R. Martin, S. A. Howell,

C. H. Douse, E. Cota, E. W. Tate, R. Tewari, and A. A. Holder.

The Plasmodium Class XIV Myosin, MyoB, Has a Distinct Sub-

cellular Location in Invasive and Motile Stages of the Malaria

Parasite and an Unusual Light Chain. The Journal of biological

chemistry, 290(19):12147–64, May 2015. ISSN 1083-351X. doi:

10.1074/jbc.M115.637694. URL http://www.jbc.org/content/

290/19/12147.long.

[212] C. Zhang, B. Xiao, Y. Jiang, Y. Zhao, Z. Li, H. Gao, Y. Ling,

J. Wei, S. Li, M. Lu, X.-Z. Su, H. Cui, and J. Yuan. Efﬁ-

cient Editing of Malaria Parasite Genome Using the CRISPR/-

Cas9 System. mBio, 5(4):1–9, 2014. ISSN 2150-7511. doi:

10.1128/mBio.01414-14. URL http://www.ncbi.nlm.nih.gov/

pubmed/24987097.

[213] Y. Zhang, F. Buchholz, J. P. Muyrers, and A. F. Stewart. A

new logic for DNA engineering using recombination in Es-

cherichia coli. Nature genetics, 20(2):123–8, Oct. 1998. ISSN 1061-

4036. doi: 10.1038/2417. URL http://www.ncbi.nlm.nih.gov/

pubmed/9771703.

[214] E. S. Zuccala, A. M. Gout, C. Dekiwadia, D. S. Marapana, F. An-

grisano, L. Turnbull, D. T. Riglar, K. L. Rogers, C. B. Whitchurch,

S. A. Ralph, T. P. Speed, and J. Baum. Subcompartmental-

isation of proteins in the rhoptries correlates with ordered

events of erythrocyte invasion by the blood stage malaria par-

asite. PloS one, 7(9):e46160, Jan. 2012. ISSN 1932-6203. doi: 10.

1371/journal.pone.0046160. URL http://journals.plos.org/

plosone/article?id=10.1371/journal.pone.0046160#s4.

Part IV

A P P E N D I X

TAB L E S

a.1 plasmogem vectors used in this work

gene vector gene vector

PBANKA_010260 PbGEM-327571 PBANKA_100170 PbGEM-292664

PBANKA_010290 PbGEM-005512 PBANKA_100240 PbGEM-034599

PBANKA_010630 PbGEM-327723 PBANKA_100250 PbGEM-292784

PBANKA_011160 PbGEM-267395 PBANKA_100360 PbGEM-264676

PBANKA_020460 PbGEM-082095 PBANKA_100530 PbGEM-333507

PBANKA_021450 PbGEM-269163 PBANKA_100550 PbGEM-243822

PBANKA_030110 PbGEM-269371 PBANKA_100920 PbGEM-293808

PBANKA_030450 PbGEM-328427 PBANKA_101040 PbGEM-333691

PBANKA_030670 PbGEM-009687 PBANKA_101910 PbGEM-263388

PBANKA_030950 PbGEM-270402 PBANKA_102010 PbGEM-096670

PBANKA_031140 PbGEM-111762 PBANKA_102760 PbGEM-296182

PBANKA_031480 PbGEM-010745 PBANKA_103170 PbGEM-334339

PBANKA_031660 PbGEM-328723 PBANKA_103210 PbGEM-334355

PBANKA_040180 PbGEM-230686 PBANKA_103540 PbGEM-039414

PBANKA_040640 PbGEM-328923 PBANKA_103970 PbGEM-297582

PBANKA_040690 PbGEM-271986 PBANKA_110140 PbGEM-334659

PBANKA_041600 PbGEM-085062 PBANKA_110330 PbGEM-298173

PBANKA_041780 PbGEM-273282 PBANKA_110650 PbGEM-098459

PBANKA_050180 PbGEM-273642 PBANKA_110690 PbGEM-098495

PBANKA_050410 PbGEM-273986 PBANKA_110760 PbGEM-265060

PBANKA_050690 PbGEM-230990 PBANKA_111530 PbGEM-121282

PBANKA_050720 PbGEM-274330 PBANKA_112010 PbGEM-335251

PBANKA_051140 PbGEM-231569 PBANKA_112390 PbGEM-247136

PBANKA_051390 PbGEM-329643 PBANKA_112410 PbGEM-300723

PBANKA_051520 PbGEM-086151 PBANKA_112810 PbGEM-099883

PBANKA_051535 PbGEM-232086 PBANKA_113780 PbGEM-100665

PBANKA_052170 PbGEM-329843 PBANKA_113880 PbGEM-335867

PBANKA_061970 PbGEM-330355 PBANKA_120180 PbGEM-303714

PBANKA_062240 PbGEM-088353 PBANKA_120200 PbGEM-046708

PBANKA_070200 PbGEM-330539 PBANKA_120800 PbGEM-101938

PBANKA_071270 PbGEM-330859 PBANKA_121060 PbGEM-304938

225

226 references

gene vector gene vector

PBANKA_071310 PbGEM-330875 PBANKA_121180 PbGEM-336371

PBANKA_071660 PbGEM-282090 PBANKA_121440 PbGEM-336467

PBANKA_072070 PbGEM-331075 PBANKA_122440 PbGEM-120990

PBANKA_072150 PbGEM-237247 PBANKA_122540 PbGEM-050063

PBANKA_080130 PbGEM-331195 PBANKA_122670 PbGEM-265652

PBANKA_080220 PbGEM-023655 PBANKA_122740 PbGEM-050376

PBANKA_080250 PbGEM-283042 PBANKA_122760 PbGEM-252519

PBANKA_080450 PbGEM-283290 PBANKA_123730 PbGEM-103704

PBANKA_080960 PbGEM-237098 PBANKA_124060 PbGEM-103910

PBANKA_081510 PbGEM-284458 PBANKA_130340 PbGEM-310399

PBANKA_081700 PbGEM-238732 PBANKA_130430 PbGEM-337499

PBANKA_082160 PbGEM-285114 PBANKA_130520 PbGEM-053796

PBANKA_082490 PbGEM-026600 PBANKA_131280 PbGEM-337723

PBANKA_082870 PbGEM-027131 PBANKA_131420 PbGEM-311741

PBANKA_083100 PbGEM-259972 PBANKA_131540 PbGEM-265356

PBANKA_083290 PbGEM-286410 PBANKA_133270 PbGEM-250003

PBANKA_083300 PbGEM-286426 PBANKA_133460 PbGEM-265364

PBANKA_083560 PbGEM-028140 PBANKA_133580 PbGEM-058012

PBANKA_090260 PbGEM-332267 PBANKA_133660 PbGEM-058133

PBANKA_090380 PbGEM-111794 PBANKA_133890 PbGEM-107133

PBANKA_090590 PbGEM-260788 PBANKA_134490 PbGEM-338675

PBANKA_090610 PbGEM-332331 PBANKA_134640 PbGEM-316019

PBANKA_090840 PbGEM-288058 PBANKA_134700 PbGEM-338747

PBANKA_090860 PbGEM-288074 PBANKA_135570 PbGEM-317163

PBANKA_091030 PbGEM-029684 PBANKA_135610 PbGEM-317179

PBANKA_091150 PbGEM-332491 PBANKA_135740 PbGEM-317307

PBANKA_091170 PbGEM-029891 PBANKA_136440 PbGEM-109041

PBANKA_091240 PbGEM-332523 PBANKA_140920 PbGEM-319715

PBANKA_091510 PbGEM-332595 PBANKA_141200 PbGEM-339691

PBANKA_091940 PbGEM-332699 PBANKA_141380 PbGEM-320219

PBANKA_092090 PbGEM-240365 PBANKA_141980 PbGEM-256665

PBANKA_092170 PbGEM-332739 PBANKA_142310 PbGEM-065482

PBANKA_092250 PbGEM-289713 PBANKA_143360 PbGEM-322619

PBANKA_092510 PbGEM-289985 PBANKA_143910 PbGEM-323355

PBANKA_092560 PbGEM-290017 PBANKA_144290 PbGEM-340475

PBANKA_092940 PbGEM-290680 PBANKA_144720 PbGEM-340635

PBANKA_092970 PbGEM-238659 PBANKA_145700 PbGEM-264860

PBANKA_093190 PbGEM-290936 PBANKA_145870 PbGEM-325915

references 227

gene vector gene vector

PBANKA_093200 PbGEM-333019 PBANKA_145910 PbGEM-258884

PBANKA_093210 PbGEM-239370 PBANKA_145950 PbGEM-325987

PBANKA_093690 PbGEM-240525 PBANKA_146060 PbGEM-071005

PBANKA_100010 PbGEM-333323

228 references

Table S2: DNA oligos used in these studies. Note that other oligos for Plas-

moGEM vectors are not listed since these are displayed on the

PlasmoGEM website.

Oligo Sequence

arg91 TCGGCATTCCTGCTGAACCGCTCTTCCGAT

CTGTAATTCGTGCGCGTCAG

arg97 ACACTCTTTCCCTACACGACGCTCTTCCGA

TCTCCTTCAATTTCGATGGGTAC

1306400_del CGTGCAGCAAAGGAATCAGT

1306400_del AATCGACAAGGCTCCCATGA

0941100_del GGAGGGGTAAACAAGTGGGA

0941100_del CGTATGACTCGGACGGACAT

1207600_del CCCGCTATGCTCTCCAAAAC

1207600_del GGACGAAGGGGAGTGGATAG

0918700_del GGAGCGATGAGAAGAGAGCT

0918700_del TCGAATGCCCACCTTCTCTT

pkg_del ctatcgcccacctgcatttc

pkg_del tccgttcactggtaccgtac

OligoDNAJfw attgCCAGGCACGGGCACTTCCAG

OligoDNAJrev aaacCACCATATAGCGAACAAGTGGGAC

TGL58int TCGCCCACGTAAACTGCGCA

TGL59int CCCACCCACCCTCTTGCAGG

TGL60int AATTCGGTGTAGGGCGGCGC

TGL61int CGGTGAGGGCGCAGAAGTGG

TGL62int TGCCCACCTGAAGAACAGTGAGC

TGL63int CCAGGCACGGGCACTTCCAG

TGL64int GCCTACTGAGCGACCGCGTC

TGL65int TGCGCGTGGTCTTCTCACGG

TGL66int TCTGTGCAGGCGAGCATGGC

pkhsp70L tggagtgtagccatccgaat

pkhsp70R atatgtgtactccccagcgc

dhfrL ttcgctaaactgcatcgtcg

dhfrR tgggtgattcatggcttcct

TPR458 TTTTAAGTGTAGTTAATTCATCAAATAGCATGCCTGC

AGGTCGACTCTAGAGGATCC-

GAAAATGGGTAAAAAAAAAAAAAAAA-

GAAAAGAGAACATG

TPR459 agttgataacggactagccttattttaacttgctaTTT Ctagctc-

taaaacagGTCTTCtcGAAGACccCAATAATAT-

ACTGTAACTCAGAATA

S U P P L E M E N TA RY F I G U R E S

Figure S1: AP2-O expression levels in P. falciparum homolog by stage [163]

229

230 references

12,332 bp

Figure S2: Map of Crispr/Cas9 vector expressing guide RNA, from the U6

promoter, yDHODH resistance marker and Cas9 protein. BbsI

sites are digested for insertion of the guide.

references 231

Slow Redundant Essential

Essential Redundant Slow

Figure S3: This is a Cohen-Friendly association plot demonstrating the rela-

tionship between the phenotypes on either end of a PlasmoINT

connection. The width of each cell represents the expected size of

the cell and the height the signed difference between observed

and expected. The data for this ﬁgure is combined from the

barseq experiment and PhenoPlasm.

R AW B A R S E Q D ATA

The following pages display, for the sake of completeness, the raw

barcode-ratios for barseq experiments. In due course these data will

also be available at http://plasmogem.sanger.ac.uk in an interactive

format.

233

PBANKA_031480

PBANKA_040820

PBANKA_060550

PBANKA_082870

PBANKA_112430

PBANKA_124430

PBANKA_131480

PBANKA_134170

PBANKA_136210

PBANKA_140160

PBANKA_142540

PBANKA_142880

0.000

0.005

0.010

0.015

0.020

0.025

0.00

0.01

0.02

0.03

0.04

0.05

0.0000

0.0005

0.0010

0.0015

0.0020

0.0000

0.0025

0.0050

0.0075

0.003

0.006

0.009

0.07

0.09

0.11

0.000

0.002

0.004

0.006

0.008

0.010

0.012

0.0000

0.0005

0.0010

0.0015

0.00

0.05

0.10

0.15

0.09

0.11

0.13

0.000

0.001

0.002

0.003

4 5 6 7 8 4 5 6 7 8 4 5 6 7 8

Day

Barcode ratio

Mouse

234

PBANKA_051820

PBANKA_052140

PBANKA_052240

PBANKA_093240

PBANKA_101290

PBANKA_101330

PBANKA_103510

PBANKA_103780

PBANKA_110520

PBANKA_114490

PBANKA_123730

PBANKA_141580

0.000

0.003

0.006

0.009

0.000

0.001

0.002

0.003

0.004

0.05

0.06

0.07

0.08

0.09

0.000

0.001

0.002

0.003

0.000

0.002

0.004

0.000

0.002

0.004

0.000

0.001

0.002

0.003

0.000

0.005

0.010

0.000

0.002

0.004

0.006

0.008

0.0000

0.0025

0.0050

0.0075

0.0100

0.0125

0.0000

0.0005

0.0010

0.0015

0.0000

0.0025

0.0050

0.0075

0.0100

4 5 6 7 8 4 5 6 7 8 4 5 6 7 8

Day

Barcode ratio

Mouse

235

PBANKA_020460

PBANKA_051500

PBANKA_060530

PBANKA_092520

PBANKA_110420

PBANKA_111190

PBANKA_111470

PBANKA_113780

PBANKA_113870

PBANKA_133760

PBANKA_144040

PBANKA_146060

0.000

0.002

0.004

0.006

0.02

0.04

0.06

0.000

0.001

0.002

0.003

0.004

0.000

0.002

0.004

0.006

0.008

0.01

0.02

0.03

0.010

0.015

0.020

0.025

0.05

0.06

0.07

0.00

0.01

0.02

0.03

0.04

0.05

0.005

0.010

0.015

0.020

0.025

0.000

0.001

0.002

0.003

0.004

0.03

0.04

0.05

0.06

4 5 6 7 8 4 5 6 7 8 4 5 6 7 8

Day

Barcode ratio

Mouse

236

PBANKA_030670

PBANKA_031420

PBANKA_051110

PBANKA_051160

PBANKA_051520

PBANKA_052170

PBANKA_061910

PBANKA_080340

PBANKA_093980

PBANKA_101520

PBANKA_121780

PBANKA_130920

0e+00

1e−04

2e−04

3e−04

4e−04

5e−04

0.000

0.005

0.010

0.015

0.01

0.02

0.03

0.04

0.0000

0.0005

0.0010

0.0015

0.0020

0.000

0.001

0.002

0.003

0.004

0.000

0.005

0.010

0.015

0.000

0.001

0.002

0.003

0.004

0.010

0.015

0.020

0.025

0.030

0.05

0.06

0.07

0.08

0.000

0.002

0.004

0.006

0e+00

3e−04

6e−04

9e−04

0.12

0.13

0.14

0.15

0.16

4 5 6 7 8 4 5 6 7 8 4 5 6 7 8

Day

Barcode ratio

Mouse

237

PBANKA_000440

PBANKA_040180

PBANKA_071750

PBANKA_080090

PBANKA_082520

PBANKA_110690

PBANKA_112010

PBANKA_113880

PBANKA_124460

PBANKA_130380

PBANKA_135570

PBANKA_144610

0.00000

0.00025

0.00050

0.00075

0.000

0.002

0.004

0.006

0.005

0.006

0.007

0.008

0.009

0.010

0.000

0.004

0.008

0.012

0.004

0.006

0.008

0.010

0.000

0.001

0.002

0.003

0.004

0.005

0.012

0.016

0.020

0.000

0.005

0.010

0.015

0.01

0.02

0.03

0.04

0.0000

0.0005

0.0010

0.0015

0.0020

0.0025

0.0050

0.0075

0.0100

0.0000

0.0005

0.0010

0.0015

0.0020

4 5 6 7 8 4 5 6 7 8 4 5 6 7 8

Day

Barcode ratio

Mouse

238

PBANKA_040710

PBANKA_050720

PBANKA_061360

PBANKA_080450

PBANKA_090610

PBANKA_090930

PBANKA_091890

PBANKA_100250

PBANKA_130640

PBANKA_131420

PBANKA_141910

PBANKA_144720

0.0025

0.0050

0.0075

0.0025

0.0050

0.0075

0.015

0.020

0.025

0.030

0.005

0.007

0.009

0.011

0.005

0.006

0.007

0.008

0.009

0.000

0.002

0.004

0.006

0.008

0.010

0.012

0.014

0.0100

0.0125

0.03

0.04

0.05

0.06

0.0000

0.0005

0.0010

0.0015

0.0020

0.0100

0.0125

0.0150

0.0175

0.000

0.001

0.002

0.003

0.004

4 5 6 7 8 4 5 6 7 8 4 5 6 7 8

Day

Barcode ratio

Mouse

239

PBANKA_011160

PBANKA_050180

PBANKA_050690

PBANKA_081510

PBANKA_092160

PBANKA_092940

PBANKA_121060

PBANKA_121330

PBANKA_133040

PBANKA_134000

PBANKA_143270

PBANKA_146350

0.000

0.001

0.002

0.0000

0.0025

0.0050

0.0075

0.0100

0.0125

0.002

0.004

0.006

0.008

0.005

0.010

0.0000

0.0025

0.0050

0.0075

0.0100

0.0025

0.0050

0.0075

0.0100

0.0125

0.000

0.005

0.010

0.006

0.007

0.008

0.009

0.010

0.0000

0.0025

0.0050

0.0075

0.0100

0.009

0.012

0.015

0.018

0.004

0.008

0.012

0.016

0.000

0.001

0.002

0.003

0.004

0.005

4 5 6 7 8 4 5 6 7 8 4 5 6 7 8

Day

Barcode ratio

Mouse

240

PBANKA_061510

PBANKA_071660

PBANKA_081240

PBANKA_083300

PBANKA_083380

PBANKA_093190

PBANKA_093200

PBANKA_120350

PBANKA_131280

PBANKA_135740

PBANKA_144160

PBANKA_145950

0.04

0.06

0.08

0.10

0.12

0.004

0.005

0.006

0.001

0.002

0.003

0.004

0.04

0.05

0.06

0.07

0.000

0.002

0.004

0.006

0.0000

0.0025

0.0050

0.0075

0.0100

0e+00

2e−04

4e−04

6e−04

8e−04

0.005

0.010

0.015

0.000

0.001

0.002

0.003

0.010

0.011

0.012

0.013

0.014

0.024

0.028

0.032

0.036

0.0000

0.0025

0.0050

0.0075

4 5 6 7 8 4 5 6 7 8 4 5 6 7 8

Day

Barcode ratio

Mouse

241

PBANKA_061580

PBANKA_081760

PBANKA_082160

PBANKA_090260

PBANKA_092510

PBANKA_093690

PBANKA_103170

PBANKA_111980

PBANKA_122980

PBANKA_132890

PBANKA_135610

PBANKA_140440

0.001

0.002

0.003

0.000

0.002

0.004

0.006

0.008

0.000

0.001

0.002

0.003

0.004

0.000

0.002

0.004

0.006

0.000

0.002

0.004

0.006

0.000

0.002

0.004

0.006

0.008

0.0175

0.0200

0.0225

0.000

0.001

0.002

0.015

0.020

0.025

0.030

0.000

0.003

0.006

0.009

0.012

0.006

0.008

0.010

0.000

0.001

0.002

0.003

4 5 6 7 8 4 5 6 7 8 4 5 6 7 8

Day

Barcode ratio

Mouse

242

PBANKA_021450

PBANKA_040280

PBANKA_040640

PBANKA_040840

PBANKA_091150

PBANKA_091510

PBANKA_092090

PBANKA_092900

PBANKA_112140

PBANKA_113980

PBANKA_121970

PBANKA_142000

0.0000

0.0025

0.0050

0.0075

0.0100

0.000

0.001

0.002

0.003

0.000

0.001

0.002

0.003

0.004

0.005

0.002

0.003

0.004

0.005

0.006

0.008

0.010

0.015

0.020

0.025

0.030

0.000

0.002

0.004

0.006

0.008

0.0025

0.0050

0.0075

0.005

0.010

0.015

0.005

0.007

0.009

0.011

0.015

0.020

0.025

0.030

0.0005

0.0010

0.0015

0.0020

0.0025

4 5 6 7 8 4 5 6 7 8 4 5 6 7 8

Day

Barcode ratio

Mouse

243

PBANKA_030110

PBANKA_050410

PBANKA_051360

PBANKA_051535

PBANKA_062250

PBANKA_091840

PBANKA_111160

PBANKA_111620

PBANKA_121440

PBANKA_142090

PBANKA_142790

PBANKA_144100

0.004

0.006

0.008

0.010

0.012

0.02

0.03

0.04

0.05

0.000

0.001

0.002

0.003

0.005

0.007

0.009

0.011

0.013

0.020

0.025

0.030

0.012

0.016

0.020

0.024

0.000

0.001

0.002

0.003

0.004

0.000

0.002

0.004

0.006

0.008

0.004

0.005

0.006

0.007

0.008

0.009

0.003

0.004

0.005

0.006

0.0000

0.0025

0.0050

0.0075

0.0100

4 5 6 7 8 4 5 6 7 8 4 5 6 7 8

Day

Barcode ratio

Mouse

244

PBANKA_041700

PBANKA_070200

PBANKA_082410

PBANKA_092560

PBANKA_103010

PBANKA_103440

PBANKA_112870

PBANKA_130430

PBANKA_131540

PBANKA_142720

PBANKA_144110

PBANKA_145870

0.000

0.001

0.002

0.003

0.004

0.005

0.002

0.003

0.004

0.005

0.000

0.001

0.002

0.06

0.07

0.08

0.09

0.000

0.001

0.002

0.003

0.004

0.005

0.010

0.015

0.020

0.003

0.004

0.005

0.006

0.0100

0.0125

0.0150

0.000

0.001

0.002

0.003

0.000

0.001

0.002

0.003

0.006

0.007

0.008

0.009

0.000

0.005

0.010

4 5 6 7 8 4 5 6 7 8 4 5 6 7 8

Day

Barcode ratio

Mouse

245

p230p−tag

PBANKA_061040

PBANKA_081430

PBANKA_092250

PBANKA_100920

PBANKA_112390

PBANKA_121180

PBANKA_122760

PBANKA_142220

0e+00

5e−04

1e−03

0.000

0.002

0.004

0.006

0.008

0.012

0.016

0.020

0.000

0.001

0.002

0.003

0.001

0.002

0.003

0.01

0.02

0.03

0.030

0.035

0.040

0.0000

0.0025

0.0050

0.0075

0.0100

0.03

0.04

4 5 6 7 8 4 5 6 7 8 4 5 6 7 8

Day

Barcode ratio

Mouse

246