Our ability to acquire immunity to a disease depends on our immune system learning to recognise foreign molecules—called antigens—that are specific to the disease-causing virus, bacterium or parasite. However, some pathogens, such as the malaria-causing parasite Plasmodium falciparum, get around this defence through a process called antigenic variation. This involves the parasite switching between different antigens over the course of an infection, preventing the host immune system from learning to recognise them and leading to infections that last many weeks or even months.

The main antigen in P. falciparum is a protein called PfEMP1, which is encoded by a family of genes called var (‘variable’). Var genes have evolved to be highly diverse, and different parasites have different repertoires of around 50–60 var genes. This ensures that there are a huge number of distinct variants of the PfEMP1 antigen available within the population, allowing the malaria parasite to maintain long-lasting infections and also to infect the same individuals again and again.

Previous work has shown that the expression of var genes is not random, but it is not clear what determines which genes are expressed at any given time. Now, Noble et al. have performed a detailed investigation of antigenic switching in P. falciparum. Using clonal parasites, they closely monitored the expression of the entire var gene repertoire during many generations of parasite culture. They observed that although different cultures initially expressed distinct var genes, most of them ended up expressing two particular genes—var27 and var29—at high levels, indicating a hard-wired gene ‘activation hierarchy’.

Noble et al. found that whenever the parasites switched antigens, var genes that were centrally located on chromosomes—such as var27 and var29—were more likely to be activated than those at the ends of chromosomes. Moreover, var genes that were highly diverse were more likely to be activated than more conserved genes: this is the first evidence linking var gene evolution with gene activation probabilities. Together, these factors gave rise to the proposed activation hierarchy, which favours genes optimised for immune evasion and aids their continued evolution and diversification. Further work is now needed to identify the molecular mechanisms that control antigenic switching and to determine whether these could represent new therapeutic targets for malaria.

Acquired antibody-mediated immunity against Plasmodium falciparum is directed against the disease-causing, intra-erythrocytic life-stage of the parasite’s life-cycle. Continual exposure to P. falciparum infection can lead to a form of semi-immunity, whereby protection against life-threatening disease is acquired after a few infections only (Gupta et al., 1999) whereas sterile and long-lasting immunity against infection is never achieved through natural exposure (Greenwood et al., 1987; Marsh, 1992; Tran et al., 2013). The process of natural acquired immunity therefore coincides with a transition from acute and severe infection in young children to asymptomatic carriage in older individuals (Marsh and Snow, 1997). At the core of this variable outcome of infection and poor development of immunity lies the major parasite virulence factor and variant antigen, P. falciparum erythrocyte membrane protein 1 (PfEMP1). Members of this family of proteins are expressed on the surface of infected red blood cells and are involved in the binding of parasitised cells to other host cells and tissues (for an overview see Kraemer and Smith, (2006)). The conferred binding phenotypes are believed to enhance parasite survival by avoiding splenic clearance (Boone and Watters, 1995; Buffet et al., 2009) and also contribute to malaria pathology through parasite sequestration in the deep vasculature (Ockenhouse et al., 1991; Pongponratn et al., 1991). Mutually exclusive transcriptional switching between PfEMP1 variants during the course of an infection thus affords the parasite an immune evasion strategy and a means to exploit different host tissues (Roberts et al., 1992; Smith et al., 1995).

Each haploid parasite genome contains approximately 60 members of the var gene family (Gardner et al., 2002) encoding distinct PfEMP1 variants, with a high degree of divergence between the antigenic repertoires of any two parasites. The high diversity of var genes and var gene repertoires is predominantly generated and maintained by frequent recombination and gene conversion events (Deitsch et al., 1999; Ward et al., 1999; Freitas-Junior et al., 2000; Taylor et al., 2000; Kraemer et al., 2007; Frank et al., 2008), which has also resulted in a high degree of mosaicism within these genes (Ward et al., 1999; Taylor et al., 2000; Bull et al., 2008). The modular and highly polymorphic, extracellular portion of PfEMP1 is composed of a variable number of Duffy-binding-like domains (DBL; at least two and up to seven per gene) and cysteine-rich interdomain regions (CIDR; up to two per gene) (Kraemer et al., 2007). Despite the overall diversity of var genes at the sequence level, distinct associations with severe malaria have been identified for several such genes (Salanti et al., 2003; Jensen et al., 2004), with the involvement of var2csa in pregnancy-associated malaria being the clearest example to date (Salanti et al., 2004), as well as associations with particular DBL domains (Warimwe et al., 2009) and combinations of DBL and CIDR (Avril et al., 2012; Claessens et al., 2012; Lavstsen et al., 2012).

Var genes can also be classified according to a conserved upstream promoter sequence into four types termed UpsA, UpsB, UpsC, and UpsE (Lavstsen et al., 2003). Of particular interest are the UpsA-type genes, which appear to have diverged from the other groups (Kraemer et al., 2007; Bull et al., 2008) and have been shown to be upregulated during severe infections in young hosts (Jensen et al., 2004; Bull et al., 2005; Kaestli et al., 2006; Kyriacou et al., 2006; Rottmann et al., 2006; Warimwe et al., 2009). As a possible consequence, acquisition of anti-PfEMP1 immunity appears to follow a particular order, with UpsA variants generally being the first to be broadly recognised in older children (Warimwe et al., 2009; Cham et al., 2010). Nevertheless, clear associations between the expression of particular genes or Ups-groups, host age and severity of disease are still missing and could in fact be strain- or location-specific. Understanding the underlying pattern of antigenic switching between var genes is therefore important to explain not only the mechanisms and dynamics of persistent infections but also the relationship between immune-mediated expression of particular PfEMP1 sub-types and infection outcomes.

Various studies have investigated var gene expression during natural or experimental infections (Peters et al., 2002; Lavstsen et al., 2005; Wunderlich et al., 2005; Blomqvist et al., 2010). However, as the diversity and order of PfEMP1 variants in the human host is influenced by immune responses and other host factors (Kyriacou et al., 2006; Warimwe et al., 2009), attempts to elucidate the underlying patterns of antigenic change and estimate related switching parameters are commonly based on in vitro cultured parasites in the absence of selection pressure. In this setting, Horrocks et al. (Horrocks et al., 2004) found that the rates at which var genes activate or deactivate are non-random, gene-specific and highly dissimilar. It has furthermore been suggested that var genes occupying subtelomeric loci tend to switch off faster than those positioned in central chromosomal regions (Frank et al., 2007), offering a possible explanation for the rapid decline in the transcription of subtelomeric genes in parasites during culture adaptation (Peters et al., 2007; Zhang et al., 2011). More recently, Recker et al. (2011) proposed that var gene transcriptional change involves a highly structured pathway that has evolved in response to a trade-off between the within-host and the between-host level fitness. However, most of these previous studies were limited in analytical depth due to restrictions in the number of variants considered and/or time points at which gene transcription levels were measured, and could therefore describe only small fragments of the whole switching network.

In this study, we used a novel approach to provide the first detailed characterisation of antigenic switching in the HB3 isolate of P. falciparum. We analysed pooled var gene transcription data from verified quantitative real time PCR measurements of a diverse set of clonal parasite cultures using a statistically rigorous method of parameter estimation. This revealed a global hierarchy in var gene activation favouring a highly diverse set of genes in central chromosomal locations. Our results further suggest a role of active gene transcription in the generation of antigenic diversity and will have important implications for understanding the age- and exposure-dependent pathology of P. falciparum malaria.

11 subclones of the lab-adapted HB3 isolate of P. falciparum were obtained by serial dilution of the parent culture. The clones were cultivated and transcript levels of 38 var genes measured by quantitative PCR, first at 29–33 parasite generations (8–9 weeks) after cloning and then at several time points over the following 60–79 generations in each culture (Table 1); a detailed description can be found in the ‘Materials and methods’ section. These 38 genes comprised almost all known functional HB3 var genes (Kraemer et al., 2007; Rowe et al., 2009) and one partial gene. The HB3 parent culture was also observed at seven time points. At the first observation, each of the 11 clones predominantly transcribed one of the nine different var genes, which were most likely the genes transcribed at the time of cloning. Importantly, these nine genes—hereafter referred to as ‘starter genes’—were representative of the var repertoire diversity in terms of chromosomal location and Ups type.

Var gene transcription dynamics

The proportional transcript levels of the different starter genes declined at notably different rates over the course of the experiment (Figure 1, Figure 1—figure supplements 1 and 2). For example, the centrally located, UpsC starter gene var29 in culture #8 accounted for more than 90% of the total transcript level at the first measurement and remained the dominant variant over the following 60 generations. In contrast, the transcript level of the var35 starter gene of culture #11 (subtelomeric, UpsA promoter type) was already below 50% of the total transcript level at the point of the first observation (approximately 30 generations post cloning) and continued to decline during the time course of culture, to be eventually replaced by other genes as dominant transcripts. Similar patterns were seen in cultures #3 and #7, which both had centrally located, UpsC starter genes, whereas the remaining five starter genes showed intermediate rates of decline.

Figure 1 with 2 supplements see all

Download asset Open asset

Of particular interest was the apparent convergence of most cultures towards high transcription of var27 and/or var29 (Figure 1, Figure 1—figure supplements 1 and 2), which are both situated in central chromosomal location but have different promoter types (UpsB and UpsC, respectively). In light of the very different initial conditions, in terms of var gene transcription levels at which these cultures were established, this indicates that antigenic switching in P. falciparum might be governed by an activation hierarchy.

To fully understand the patterns of antigenic switching underlying the observed transcriptional changes in our parasite cultures, we estimated relevant switch parameters using a previously described Markov chain Monte Carlo (MCMC) method (Noble and Recker, 2012) (‘Materials and methods’). We have previously proposed that var gene switching can be fully described dynamically by two sets of parameters (‘Materials and methods’ and [Recker et al., 2011]). First, each gene i has an off-rate, ${\omega }_i$, which is the per generation probability of switching from active to silenced. Second, each gene has a switch bias, ${\beta }_{ij}$, which is the probability that when gene i switches off, gene j becomes activated. We applied our algorithm to the full 38-gene data set and separately to a reduced set of the 16 most transcribed genes, and both models generally gave good and similar fits between the data and predicted transcript levels (Figure 1—figure supplements 1 and 2), with deviations resembling stochastic measurement errors.

Our method provided estimated switch parameters for the entire var repertoire, which are illustrated in Figure 2 for the 16 most highly transcribed genes. Due to higher signal to noise ratios, parameter estimates of the starter genes (indicated in red) were more accurate and resulted in much narrower credible intervals, as indicated by less fuzzy rings that represent respective switch probabilities. As clearly demonstrated in Figure 2, there was wide distribution in switch biases and off-rates, and we next analysed for specific associations between these parameters and other gene characteristics.

Figure 2 with 1 supplement see all

Download asset Open asset

Figure 2—source data 1

Estimated switch biases and 95% credible intervals (CI).

Mean switch biases from gene i (row) to gene j (column) for the 16 most transcribed var genes, together with the 95% credible intervals.

https://doi.org/10.7554/eLife.01074.008

Download elife-01074-fig2-data1-v1.xls

Off-rate estimates suggest a role of chromosomal location, not promoter type

Estimated off-rates ranged from 0.3% to more than 5% per generation, with a mean of approximately 2.9% (Table 2). Notably, the two most transcribed genes, var29 and var27 (both centrally located on chromosome 4), had exceptionally low off-rates of approximately 0.3% and 0.8%, respectively, whereas the other starter genes switched off at rates between 1.4% and 3.5% per generation. In general we found off-rates of centrally located genes to be significantly lower than of genes in subtelomeric regions (Welch’s t = 3.3, p=0.002), as shown in Figure 3. This result remained significant ($p=0.035$) when we accounted for uncertainty in our estimates (‘Materials and methods’).

Table 2

Gene	Location	Domains*	Ups	Off-rate (95% CI)	Activation bias (95% CI)
var29	Central	4	C	0.3% (0.1, 0.5)	27% (22, 32)
var27	Central	4	B	0.8% (0.5, 1.2)	23% (19, 28)
var35	Central	4	C	3.3% (2.2, 4.7)	8.0% (5.8, 11)
var28	Central	4	C	2.2% (1.4, 3.1)	6.9% (5.2, 9.1)
var32	Central	4	C	4.7% (2.5, 5.9)	2.5% (1.5, 3.3)
var36	Central	4	C	1.8% (0.2, 4.2)	2.5% (1.5, 4.1)
var31	Central	4	C	1.4% (0.9, 2.0)	2.2% (1.7, 2.9)
var34	Central	6	C	1.3% (0.1, 3.4)	2.1% (1.4, 3.6)
var13	Subtelomeric	4	B	1.9% (1.5, 2.4)	2.1% (1.6, 2.8)
var30	Central	4	C	1.9% (1.3, 2.7)	1.9% (1.4, 2.5)
var10	Subtelomeric	5	B	3.4% (1.1, 5.7)	1.7% (0.94, 2.7)
var17	Central	6	B	3.0% (0.8, 5.6)	1.7% (0.90, 2.8)
var24	Central	6	B	1.9% (0.1, 5.2)	1.6% (0.85, 3.1)
var7	Subtelomeric	7	B	2.2% (0.3, 5.1)	1.5% (0.83, 2.7)
var19	Central	4	B	1.3% (0.1, 3.2)	1.4% (0.93, 2.5)
var26	Central	4	B	1.5% (0.2, 3.4)	1.4% (0.90, 2.3)
var25	Central	6	B	2.7% (0.4, 5.6)	1.2% (0.60, 2.1)
var1csa	Subtelomeric	8	A	5.5% (4.4, 6.0)	1.2% (0.89, 1.5)
var33	Central	4	C	2.7% (0.5, 5.8)	1.1% (0.57, 1.9)
var16	Subtelomeric	4	B	2.8% (1.0, 5.3)	0.91% (0.52, 1.5)
var4	Subtelomeric	8	A	4.3% (2.1, 5.9)	0.86% (0.52, 1.3)
var14	Subtelomeric	4	B	4.0% (2.2, 5.9)	0.73% (0.45, 1.1)
var22	Central	7	B	1.9% (0.2, 5.1)	0.72% (0.42, 1.3)
var21	Central	7	B	4.7% (2.7, 5.9)	0.70% (0.46, 0.94)
var8	Subtelomeric	7	B	3.7% (1.8, 5.8)	0.66% (0.39, 1.0)
var11	Subtelomeric	5	B	1.5% (0.9, 2.2)	0.60% (0.45, 0.80)
var18	Subtelomeric	4	B	5.2% (3.8, 6.0)	0.60% (0.42, 0.79)
var5	Subtelomeric	6	A	3.5% (2.7, 4.3)	0.53% (0.39, 0.69)
var39p†	Subtelomeric	2	B	4.7% (2.7, 6.0)	0.48% (0.30, 0.66)
var50Ψ‡	Central	6	C	4.4% (2.0, 5.9)	0.46% (0.26, 0.67)
var23	Central	6	B	0.6% (0.0, 2.0)	0.38% (0.26, 0.55)
var9	Subtelomeric	6	B	2.8% (0.7, 5.3)	0.36% (0.20, 0.58)
var2csa	Subtelomeric	6	E	4.7% (2.3, 5.9)	0.28% (0.16, 0.40)
var2	Subtelomeric	6	A	1.4% (0.2, 4.2)	0.27% (0.17, 0.50)
var12	Subtelomeric	4	B	3.8% (1.5, 5.8)	0.26% (0.15, 0.40)
var20	Subtelomeric	4	B	4.1% (1.8, 5.9)	0.25% (0.14, 0.37)
var1	Subtelomeric	7	A	5.5% (4.2, 6.0)	0.024% (0.018, 0.031)
var6	Central	8	A	2.9% (0.4, 5.9)	0.018% (0.0086, 0.032)

1. *

   Number of encoded DBL (Duffy binding-like) and CIDR (Cys rich inter-domain region).

2. †

   Partial gene.

3. ‡

   Psuedogene.

Figure 3

Download asset Open asset

Previous studies reported that UpsA var genes from patient isolates were down-regulated faster than other var genes during adaptation to culture (Peters et al., 2007; Zhang et al., 2011). The HB3 repertoire contains seven UpsA genes, all but one of which are subtelomeric. We compared the off-rates of the subtelomeric UpsA genes with those of other subtelomeric genes and found no significant difference (Welch’s t = 0.65, p=0.54; accounting for uncertainty p=0.55), as shown in Figure 3. Among centrally located genes we compared UpsB with UpsC and again found no significant difference (Figure 3; Welch’s t = 0.44, p=0.44; accounting for uncertainty p=0.57). This lends support to a previous study suggesting that the rate at which transcriptionally active var genes become silenced is less dependent on promoter type than on chromosomal location (Frank et al., 2007).

Var gene switching is governed by an activation hierarchy

In each of our cultures, gene transcription levels spanned several orders of magnitude, and only a minority of genes were ever highly transcribed during the experiment. As mentioned earlier, despite the diversity of the starter genes, the relative transcript levels of the activated genes appeared to converge towards similar levels across all cultures (Figure 1, Figure 1—figure supplements 1 and 2). To test the degree of similarity in gene activation, we excluded the nine starter genes and ranked the rest of the repertoire according to average transcript levels at later time points (more than 65 generations) for each culture. Pair-wise comparison showed that the ranking was highly correlated between the cultures (Spearman’s ρ > 0.8, p=10⁻⁵ for all pairs), suggesting that the outcome of antigenic switching in P. falciparum, or at least within the HB3 isolate, is largely governed by an activation hierarchy that is independent of the currently activated gene.

Consistent with this hypothesis, the estimated switch biases between genes, ${\beta }_{ij}$, were remarkably similar, so that ${\beta }_{ij}\approx {\beta }_{kj}$ for each i, j and k. (Figure 2, Figure 2—source data 1, Figure 2—figure supplement 1). All genes were found to have high switch biases towards the same few genes, which explain the observed convergence in var transcription in our cultures. As expected, the highest biases across the repertoire were found towards var29 and var27, with some genes also frequently switching towards var35 and/or var28 (both UpsC-types in central chromosomal location).

To better describe the observed activation hierarchy, we used our MCMC method to estimate the average switch bias towards each gene, which we hereafter refer to as the gene’s activation bias. That is, the activation bias of gene j is defined as the average of the switch biases ${\beta }_{ij}$ for all deactivating genes i. It can be understood as a gene specific, per switch activation probability and is therefore distinct from the previously defined on-rate (Horrocks et al., 2004), in that it is invariant to transcript levels. As shown in Table 2, there was a clear hierarchy in gene activation biases spanning over three orders of magnitude. The ranking also appeared to be non-random so that, for example, most genes within central chromosomal location were situated in the top half of the table. Analysing for possible associations with genetic attributes confirmed that activation biases of centrally located var genes are significantly higher than those of subtelomeric genes (Figure 4, Welch’s t = 2.8, p=0.009; accounting for uncertainty p=0.011). In fact, of the highest four activation biases, three belonged to neighbouring genes on chromosome 4: var29, var27 and var28. Importantly, no significant and independent effect of promoter type was found, which again points towards a strong influence of chromosomal location on var gene activation.

Figure 4

Download asset Open asset

Additionally, we considered whether gene architecture had an effect on the observed activation biases, as recent research suggests that gene length, in terms of the number of binding domains, is another important var gene characteristic besides chromosomal location and promoter type (Buckee and Recker, 2012). Grouping genes depending on whether they encoded only four domains (short genes) or more than four domains (long genes) showed that activation biases differed significantly depending on gene length (Figure 4, F = 9.7, p=0.004) as well as chromosomal location (F = 8.1, p=0.007) and the interaction of these two factors (F = 5.0, p=0.032). As such, central var genes encoding only four binding domains (also known as Type 1 var genes [Gardner et al., 2002]) had on average a 10 times higher activation bias than the rest of the repertoire. Results were similar when we accounted for uncertainty in our estimates (p=0.005, p=0.0096, p=0.037, respectively).

Var gene switching is influenced by the deactivating gene

We next tested explicitly whether the currently active gene has an influence on the direction of switching, that is if in general ${\beta }_{ij}\ne {\beta }_{kj}$. We used an optimisation algorithm (‘Materials and methods’) to fit two alternative models to the data for the 16 most highly transcribed var genes: (i) a general model including off-rates and individual switch biases (240 independent parameters), and (ii) a simpler model, in which all sets of switch biases were required to be identical, that is where ${\beta }_{ij}={\beta }_j$ for all i and j (31 independent parameters). A likelihood ratio test indicated that the model with individual switch biases was significantly more likely than the second model (F = 1.84, p<0.0001), confirming that antigenic switching in P. falciparum is not only controlled by an activation hierarchy but also influenced by the currently active gene. We again tested for associations with chromosomal location and promoter type (‘Materials and methods’), which indicated that switch biases were higher between var genes with matching chromosomal locations than between differently located genes (Figure 5; Welch’s t = 2.2, p=0.029), whereas no significant association was found between switch biases and matching Ups types or gene lengths. This result needs to be confirmed with larger data sets, however, as statistical significance was lost when we accounted for uncertainty in our parameter estimates.

Figure 5

Download asset Open asset

Var gene sequences suggest a role of activation bias in antigenic evolution

Var gene sequences can be described using a set of 628 conserved homology blocks, with an average length of 19 amino acids (Rask et al., 2010). Using homology blocks found only within the DBL1 and CIDR1 domains, which are present in all HB3 var genes except var1csa, var2csaA, var2csaB and the partial gene var39, we constructed a var gene homology network, in which edges between genes are defined and weighted by the number of shared homology blocks (Figure 6—figure supplement 1). We then calculated the centrality of each gene within this network. This is a distance measure between all pairs of nodes (i.e., genes) and can therefore be used as a proxy for the relatedness of a gene to the rest of the repertoire. The inclusion of network centrality as an explanatory variable in the previous ANOVA model confirmed that the degree by which var genes share sequence blocks is positively correlated with activation bias, independently of chromosomal location and gene length (Figure 6A; F = 8.8, p=0.006 for the model comparison; accounting for uncertainty p=0.006).

Figure 6 with 1 supplement see all

Download asset Open asset

We hypothesised that this could indicate a possible role of activation bias, or gene transcription, in var gene sequence evolution. Under the assumption that gene transcription facilitates recombination we would expect more frequently activated genes to share more sequence blocks with other genes, simply because they more often act as recombination sites. To test this we explicitly simulated gene conversion, in which pairs of genes were chosen at random, weighted by their activation bias and homology blocks copied between them (‘Materials and methods). As expected, we found a strong positive correlation between activation bias and our measure of relatedness in terms of network centrality (Figure 6B).

Under the same assumption we would also expect genes with higher activation biases to be overall more diverse. Indeed, using the average sequence conservation of all domains as a measure of population level diversity (Buckee and Recker, 2012), we again found this to be significantly and positively correlated, independently of chromosomal location and gene length (Figure 6C; F = 13.6, p=0.0009 for the model comparison; accounting for uncertainty p=0.0006), suggesting a role of active gene transcription in generating antigenic diversity in P. falciparum malaria.

Several studies have noted that var genes located near the centres of chromosomes tend to be more highly transcribed in vitro than those in subtelomeric location (Frank et al., 2007; Peters et al., 2007; Enderes et al., 2011; Zhang et al., 2011; Fastman et al., 2012), and it has been suggested that this might be due to low deactivation rates (Frank et al., 2007). In this study, we found that high transcription levels are confined to a subset of central genes that encode only four PfEMP1 binding domains (termed Type 1 var genes [Gardner et al., 2002]) and, importantly, are the result of high activation biases rather than low off-rates. That is, even though some of the most highly transcribed genes had very low off-rates, the total sum of other genes switching off with high bias towards these genes significantly outweighs the effect of off-rates. This is consistent with our previous study using artificial transcription data, which also showed that off-rates, at least within a biologically plausible range, can have only a modest effect on the total transcript levels within parasite populations (Noble and Recker, 2012).

Reanalysis of data from a recent study on the NF54 parasite strain by Fastman et al. (2012) supports these findings. Consistent with our results for HB3, centrally located var genes encoding only four binding domains were more highly transcribed than the rest of the repertoire (Welch’s t = 2.3, p=0.02), with PF3D7_0809100, PF3D7_0421300 and PF3D7_0800100 being the most highly transcribed genes. According to our model, these are predicted to have high activation bias, and indeed, PF3D7_0809100 has previously been reported as the most transcribed var gene in NF54 bulk culture, whereas PF3D7_0421300 and PF3D7_0800100 were found to be the two genes most commonly activated by var gene switching in previous studies (Dzikowski et al., 2006; Frank et al., 2007; Enderes et al., 2011). Also in agreement with our results, NF54 gene transcription levels were positively correlated with relatedness to the rest of the repertoire in terms of shared DBL1 and CIDR1 homology blocks (Spearman’s ρ = 0.3, p=0.044).

Our results further indicate that the direction of transcriptional switching between var genes depends not only on intrinsic activation biases but also on the particular preferences of each deactivating gene. Even though these variations were found to have only a modest effect on in vitro switching, they are expected to play a much greater role during infection, where factors such as immune responses and binding affinities can significantly alter the var gene expression profiles. Furthermore, the combination of an activation hierarchy and these specific variations could explain our previously proposed source-sink model underlying antigenic switching in P. falciparum (Recker et al., 2011), in which genes with high in-degrees, i.e., those most commonly switched to, correspond to the genes here identified as having high activation biases. Intriguingly, also, our results suggest that switch biases may be stronger between genes in similar chromosomal locations: central genes prefer to switch to other central genes, and subtelomeric genes prefer to switch to other subtelomeric genes. More data are needed to fully establish the factors influencing variation in switch biases, however, especially for less activated genes, where parameter estimations are more prone to uncertainties.

Despite recent advances in elucidating the genetics and epigenetics of var gene activation and silencing, not much is known about how mutually exclusive switching between the 60 var genes is controlled at the molecular level. The discovery of intrinsic var gene activation biases and their associations with chromosomal locations and other genetic features could therefore have important implications for future studies. These factors are clearly correlated, however, and it is uncertain how they are linked by cause and effect. For example, in a recent study it was found that gene length is positively correlated with sequence conservation (Buckee and Recker, 2012), and long var genes are more often positioned in subtelomeric regions. Nevertheless, the correlation between activation bias and relatedness, which was found to be independent of chromosomal location, gene length and upstream promoter, suggests a role of gene transcription in facilitating recombination and therefore in generating and maintaining antigenic diversity in P. falciparum. Transcription-associated recombination has been observed in other eukaryotes and may result from increased accessibility to recombination proteins or from stalled replication forks (Prado and Aguilera, 2005; Gottipati and Helleday, 2009). Importantly, for P. falciparum this phenomenon could help to explain why different groups of var genes within the repertoire display very different levels of diversity within the population (Buckee et al., 2009; Buckee and Recker, 2012).

It is further possible that some var genes are under balancing selection, whereby the relative sequence conservation of longer genes could be due to optimised functionality, for example in terms of binding avidity and/or affinity. This assumption would also help to reconcile apparent contradictions between our observations and var gene transcription in vivo. That is, various in vivo studies have reported UpsA var genes to be over-expressed during infections in individuals with little previous exposure (Jensen et al., 2004; Bull et al., 2005; Lavstsen et al., 2005; Kyriacou et al., 2006; Warimwe et al., 2009), despite their low activation rates in vitro. Each var repertoire consists of a set of relatively long and conserved genes, such as those predominantly found with in the UpsA group, and another, short and diverse set of genes, such as those predominantly found within the UpsB and UpsC groups (Buckee and Recker, 2012). Under the assumption that more conserved genes have an in vivo survival advantage, possibly due to higher binding affinities (Avril et al., 2012; Claessens et al., 2012; Lavstsen et al., 2012), these variants will dominate early infections in naive hosts, regardless of activation rates. Their lower sequence diversity, however, will then lead to a more rapid acquisition of protective immunity, causing a transition towards the expression of more diverse variants, facilitated by their inherent activation biases. In light of this we would also speculate that the proposed order in which individuals acquire immunity (anti-UpsA followed by anti-UpsB/C), as well as the proposed associations between the expression of UpsA var genes and severe disease outcome, especially in young children, is less determined by the Ups group itself and is more a consequence of within-host selection of long, conserved genes, which are often but not exclusively found within the UpsA group.

Together, these results demonstrate how selection pressures operating on multiple ecological scales have shaped the phenotypic plasticity embodied within the antigenic repertoire of P. falciparum. Within this setup, var genes and the molecular mechanisms that underlie their sequential activation and silencing have co-evolved to allow the parasite to express the most advantageous phenotype in response to its current environment.

1. 1. Avril M
   2. Tripathi AK
   3. Brazier AJ
   4. Andisi C
   5. Janes JH
   6. Soma VL

   et al. (2012) A restricted subset of var genes mediates adherence of Plasmodium falciparum-infected erythrocytes to brain endothelial cells

   Proc Natl Acad Sci USA 109:E1782–E1790.

   https://doi.org/10.1073/pnas.1120534109

   • Google Scholar
2. 1. Blomqvist K
   2. Normark J
   3. Nilsson D
   4. Ribacke U
   5. Orikiriza J
   6. Trillkott P

   et al. (2010) var gene transcription dynamics in Plasmodium falciparum patient isolates

   Mol Biochem Parasitol 170:74–83.

   https://doi.org/10.1016/j.molbiopara.2009.12.002

   • Google Scholar
3. 1. Boone KE
   2. Watters DA

   (1995) The incidence of malaria after splenectomy in Papua New Guinea

   BMJ 311:1273.

   https://doi.org/10.1136/bmj.311.7015.1273

   • Google Scholar
4. 1. Buckee CO
   2. Recker M

   (2012) Evolution of the multi-domain structures of virulence genes in the human malaria parasite, Plasmodium falciparum

   PLOS Comput Biol 8:e1002451.

   https://doi.org/10.1371/journal.pcbi.1002451

   • Google Scholar
5. 1. Buckee CO
   2. Bull PC
   3. Gupta S

   (2009) Inferring malaria parasite population structure from serological networks

   Proc Biol Sci 276:477–485.

   https://doi.org/10.1098/rspb.2008.1122

   • Google Scholar
6. 1. Buffet PA
   2. Safeukui I
   3. Milon G
   4. Mercereau-Puijalon O
   5. David PH

   (2009) Retention of erythrocytes in the spleen: a double-edged process in human malaria

   Curr Opin Hematol 16:157–164.

   https://doi.org/10.1097/MOH.0b013e32832a1d4b

   • Google Scholar
7. 1. Bull PC
   2. Berriman M
   3. Kyes S
   4. Quail MA
   5. Hall N
   6. Kortok MM

   et al. (2005) Plasmodium falciparum variant surface antigen expression patterns during malaria

   PLOS Pathog 1:202–213.

   https://doi.org/10.1371/journal.ppat.0010026

   • Google Scholar
8. 1. Bull PC
   2. Buckee CO
   3. Kyes S
   4. Kortok MM
   5. Thathy V
   6. Guyah B

   et al. (2008) Plasmodium falciparum antigenic variation. Mapping mosaic var gene sequences onto a network of shared, highly polymorphic sequence blocks

   Mol Microbiol 68:1519–1534.

   https://doi.org/10.1111/j.1365-2958.2008.06248.x

   • Google Scholar
9. 1. Cham GK
   2. Turner L
   3. Kurtis JD
   4. Mutabingwa T
   5. Fried M

   et al. (2010) Hierarchical, domain type-specific acquisition of antibodies to Plasmodium falciparum erythrocyte membrane protein 1 in Tanzanian children

   Infect Immun 78:4653–4659.

   https://doi.org/10.1128/IAI.00593-10

   • Google Scholar
10. 1. Claessens A
    2. Adams Y
    3. Ghumra A
    4. Lindergard G
    5. Buchan CC
    6. Andisi C

    et al. (2012) A subset of group a-like var genes encodes the malaria parasite ligands for binding to human brain endothelial cells

    Proc Natl Acad Sci USA 109:E1772–E1781.

    https://doi.org/10.1073/pnas.1120461109

    • Google Scholar
11. 1. Deitsch KW
    2. del Pinal A
    3. Wellems TE

    (1999) Intra-cluster recombination and var transcription switches in the antigenic variation of Plasmodium falciparum

    Mol Biochem Parasitol 101:107–116.

    https://doi.org/10.1016/S0166-6851(99)00062-6

    • Google Scholar
12. 1. Dzikowski R
    2. Frank M
    3. Deitsch K

    (2006) Mutually exclusive expression of virulence genes by malaria parasites is regulated independently of antigen production

    PLOS Pathog 2:e22.

    https://doi.org/10.1371/journal.ppat.0020022

    • Google Scholar
13. 1. Enderes C
    2. Kombila D
    3. Dal-Bianco M
    4. Dzikowski R
    5. Kremsner P
    6. Frank M

    (2011) Var gene promoter activation in clonal Plasmodium falciparum isolates follows a hierarchy and suggests a conserved switching program that is independent of genetic background

    J Infect Dis 204:1620–1631.

    https://doi.org/10.1093/infdis/jir594

    • Google Scholar
14. 1. Fastman Y
    2. Noble R
    3. Recker M
    4. Dzikowski R

    (2012) Erasing the epigenetic memory and beginning to switch–the onset of antigenic switching of var genes in Plasmodium falciparum

    PLOS ONE 7:e34168.

    https://doi.org/10.1371/journal.pone.0034168

    • Google Scholar
15. 1. Frank M
    2. Dzikowski R
    3. Amulic B
    4. Deitsch K

    (2007) Variable switching rates of malaria virulence genes are associated with chromosomal position

    Mol Microbiol 64:1486–1498.

    https://doi.org/10.1111/j.1365-2958.2007.05736.x

    • Google Scholar
16. 1. Frank M
    2. Kirkman L
    3. Costantini D
    4. Sanyal S
    5. Lavazec C
    6. Templeton TJ

    et al. (2008) Frequent recombination events generate diversity within the multi-copy variant antigen gene families of Plasmodium falciparum

    Int J Parasitol 38:1099–1109.

    https://doi.org/10.1016/j.ijpara.2008.01.010

    • Google Scholar
17. 1. Freitas-Junior LH
    2. Bottius E
    3. Pirrit LA
    4. Deitsch KW
    5. Scheidig C
    6. Guinet F

    et al. (2000) Frequent ectopic recombination of virulence factor genes in telomeric chromosome clusters of P. falciparum

    Nature 407:1018–1022.

    https://doi.org/10.1038/35039531

    • Google Scholar
18. 1. Gardner MJ
    2. Hall N
    3. Fung E
    4. White O
    5. Berriman M
    6. Hyman RW

    et al. (2002) Genome sequence of the human malaria parasite Plasmodium falciparum

    Nature 419:498–511.

    https://doi.org/10.1038/nature01097

    • Google Scholar
19. 1. Gottipati P
    2. Helleday T

    (2009) Transcription-associated recombination in eukaryotes: link between transcription, replication and recombination

    Mutagenesis 24:203–210.

    https://doi.org/10.1093/mutage/gen072

    • Google Scholar
20. 1. Greenwood BM
    2. Bradley AK
    3. Greenwood AM
    4. Byass P
    5. Jammeh K
    6. Marsh K

    et al. (1987) Mortality and morbidity from malaria among children in a rural area of the Gambia, West Africa

    Trans R Soc Trop Med Hyg 81:478–486.

    https://doi.org/10.1016/0035-9203(87)90170-2

    • Google Scholar
21. 1. Gupta S
    2. Snow RW
    3. Donnelly CA
    4. Marsh K
    5. Newbold C

    (1999) Immunity to non-cerebral severe malaria is acquired after one or two infections

    Nat Med 5:340–343.

    https://doi.org/10.1038/6560

    • Google Scholar
22. 1. Horrocks P
    2. Pinches R
    3. Christodoulou Z
    4. Kyes SA
    5. Newbold CI

    (2004) Variable var transition rates underlie antigenic variation in malaria

    Proc Natl Acad Sci USA 101:11129–11134.

    https://doi.org/10.1073/pnas.0402347101

    • Google Scholar
23. 1. Jensen AT
    2. Magistrado P
    3. Sharp S
    4. Joergensen L
    5. Lavstsen T

    et al. (2004) Plasmodium falciparum associated with severe childhood malaria preferentially expresses PfEMP1 encoded by group a var genes

    J Exp Med 199:1179–1190.

    https://doi.org/10.1084/jem.20040274

    • Google Scholar
24. 1. Kaestli M
    2. Cockburn IA
    3. Cortes A
    4. Baea K
    5. Rowe JA
    6. Beck HP

    (2006) Virulence of malaria is associated with differential expression of Plasmodium falciparum var gene subgroups in a case-control study

    J Infect Dis 193:1567–1574.

    https://doi.org/10.1086/503776

    • Google Scholar
25. 1. Kraemer SM
    2. Kyes SA
    3. Aggarwal G
    4. Springer AL
    5. Nelson SO
    6. Christodoulou Z

    et al. (2007) Patterns of gene recombination shape var gene repertoires in Plasmodium falciparum: comparisons of geographically diverse isolates

    BMC Genomics 8:45.

    https://doi.org/10.1186/1471-2164-8-45

    • Google Scholar
26. 1. Kraemer SM
    2. Smith JD

    (2006) A family affair: var genes, PfEMP1 binding, and malaria disease

    Curr Opin Microbiol 9:374–380.

    https://doi.org/10.1016/j.mib.2006.06.006

    • Google Scholar
27. 1. Kyes S
    2. Christodoulou Z
    3. Pinches R
    4. Kriek N
    5. Horrocks P
    6. Newbold C

    (2007) Plasmodium falciparum var gene expression is developmentally controlled at the level of RNA polymerase II-mediated transcription initiation

    Mol Microbiol 63:1237–1247.

    https://doi.org/10.1111/j.1365-2958.2007.05587.x

    • Google Scholar
28. 1. Kyes S
    2. Pinches R
    3. Newbold C

    (2000) A simple RNA analysis method shows var and rif multigene family expression patterns in Plasmodium falciparum

    Mol Biochem parasitol 105:311–315.

    https://doi.org/10.1016/S0166-6851(99)00193-0

    • Google Scholar
29. 1. Kyriacou HM
    2. Stone GN
    3. Challis RJ
    4. Raza A
    5. Lyke KE
    6. Thera MA

    et al. (2006) Differential var gene transcription in Plasmodium falciparum isolates from patients with cerebral malaria compared to hyperparasitaemia

    Mol Biochem Parasitol 150:211–218.

    https://doi.org/10.1016/j.molbiopara.2006.08.005

    • Google Scholar
30. 1. Lambros C
    2. Vanderberg JP

    (1979) Synchronization of Plasmodium falciparum erythrocytic stages in culture

    J Parasitol 65:418–420.

    https://doi.org/10.2307/3280287

    • Google Scholar
31. 1. Lavstsen T
    2. Magistrado P
    3. Hermsen CC
    4. Salanti A
    5. Jensen AT
    6. Sauerwein R

    et al. (2005) Expression of Plasmodium falciparum erythrocyte membrane protein 1 in experimentally infected humans

    Malar J 4:21.

    https://doi.org/10.1186/1475-2875-4-21

    • Google Scholar
32. 1. Lavstsen T
    2. Salanti A
    3. Jensen ATR
    4. Arnot DE
    5. Theander TG

    (2003) Sub-grouping of Plasmodium falciparum 3D7 var genes based on sequence analysis of coding and non-coding regions

    Malar J 2:27.

    https://doi.org/10.1186/1475-2875-2-27

    • Google Scholar
33. 1. Lavstsen T
    2. Turner L
    3. Saguti F
    4. Magistrado P
    5. Rask TS
    6. Jespersen JS

    et al. (2012)

    Plasmodium falciparum erythrocyte membrane protein 1 domain cassettes 8 and 13 are associated with severe malaria in children

    E1791–E1800, Proc Natl Acad Sci USA, 109, 10.1073/pnas.1120455109.

    • Google Scholar
34. 1. Marsh K

    (1992) Malaria–a neglected disease?

    Parasitology 104:S53–S69.

    https://doi.org/10.1017/S0031182000075247

    • Google Scholar
35. 1. Marsh K
    2. Snow RW

    (1997) Host-parasite interaction and morbidity in malaria endemic areas

    Philos Trans R Soc Lond B Biol Sci 352:1385–1394.

    https://doi.org/10.1098/rstb.1997.0124

    • Google Scholar
36. 1. Newman MEJ

    (2001) Scientific collaboration networks. II. shortest paths, weighted networks, and centrality

    Phys Rev E Stat Nonlin Soft Matter Phys 64:016132.

    https://doi.org/10.1103/PhysRevE.64.016132

    • Google Scholar
37. 1. Noble R
    2. Recker M

    (2012) A statistically rigorous method for determining antigenic switching networks

    PLOS ONE 7:e39335.

    https://doi.org/10.1371/journal.pone.0039335

    • Google Scholar
38. 1. Ockenhouse CF
    2. Ho M
    3. Tandon NN
    4. Vanseventer GA
    5. Shaw S
    6. White NJ

    et al. (1991) Molecular-basis of sequestration in severe and uncomplicated Plasmodium falciparum malaria - differential adhesion of infected erythrocytes to CD36 and ICAM-1

    J Infect Dis 164:163–169.

    https://doi.org/10.1093/infdis/164.1.163

    • Google Scholar
39. 1. Opsahl T

    (2009) Structure and evolution of weighted networks

    104–122, Structure and evolution of weighted networks, http://toreopsahl.com/publications/thesis/.

    http://toreopsahl.com/publications/thesis/

    • Google Scholar
40. 1. Peters J
    2. Fowler E
    3. Gatton M
    4. Chen N
    5. Saul A
    6. Cheng Q

    (2002) High diversity and rapid changeover of expressed var genes during the acute phase of Plasmodium falciparum infections in human volunteers

    Proc Natl Acad Sci USA 99:10 689–10 694.

    https://doi.org/10.1073/pnas.162349899

    • Google Scholar
41. 1. Peters JM
    2. Fowler EV
    3. Krause DR
    4. Cheng Q
    5. Gatton ML

    (2007) Differential changes in Plasmodium falciparum var transcription during adaptation to culture

    J Infect Dis 195:748–755.

    https://doi.org/10.1086/511436

    • Google Scholar
42. 1. Pfaffl MW

    (2001) A new mathematical model for relative quantification in real-time RT-PCR

    Nucleic Acids Res 29:e45.

    https://doi.org/10.1093/nar/29.9.e45

    • Google Scholar
43. 1. Pongponratn E
    2. Riganti M
    3. Punpoowong B
    4. Aikawa M

    (1991)

    Microvascular sequestration of parasitized erythrocytes in human falciparum-malaria - a pathological study

    Am J Trop Med Hyg 44:168–175.

    • Google Scholar
44. 1. Prado F
    2. Aguilera A

    (2005) Impairment of replication fork progression mediates RNA polII transcription-associated recombination

    EMBO J 24:1267–1276.

    https://doi.org/10.1038/sj.emboj.7600602

    • Google Scholar
45. 1. Rask TS
    2. Hansen DA
    3. Theander TG
    4. Pedersen AG
    5. Lavstsen T

    (2010) Plasmodium falciparum erythrocyte membrane protein 1 diversity in seven genomes - divide and conquer

    Plos Comput Biology 6:e1000933.

    https://doi.org/10.1371/journal.pcbi.1000933

    • Google Scholar
46. 1. Recker M
    2. Buckee CO
    3. Serazin A
    4. Kyes S
    5. Pinches R
    6. Christodoulou Z

    et al. (2011) Antigenic variation in Plasmodium falciparum malaria involves a highly structured switching pattern

    Plos Pathog 7:e1001306.

    https://doi.org/10.1371/journal.ppat.1001306

    • Google Scholar
47. 1. Roberts DJ
    2. Craig AG
    3. Berendt AR
    4. Pinches R
    5. Nash G
    6. Marsh K

    et al. (1992) Rapid switching to multiple antigenic and adhesive phenotypes in malaria

    Nature 357:689–692.

    https://doi.org/10.1038/357689a0

    • Google Scholar
48. 1. Rottmann M
    2. Lavstsen T
    3. Mugasa JP
    4. Kaestli M
    5. Jensen AT
    6. Muller D

    et al. (2006) Differential expression of var gene groups is associated with morbidity caused by Plasmodium falciparum infection in Tanzanian children

    Infect Immun 74:3904–3911.

    https://doi.org/10.1128/IAI.02073-05

    • Google Scholar
49. 1. Rowe JA
    2. Claessens A
    3. Corrigan RA
    4. Arman M

    (2009) Adhesion of Plasmodium falciparum-infected erythrocytes to human cells: molecular mechanisms and therapeutic implications

    Expert Rev Mol Med 11:e16.

    https://doi.org/10.1017/S1462399409001082

    • Google Scholar
50. 1. Salanti A
    2. Dahlback M
    3. Turner L
    4. Nielsen MA
    5. Barfod L
    6. Magistrado P

    et al. (2004) Evidence for the involvement of VAR2CSA in pregnancy-associated malaria

    J Exp Med 200:1197–1203.

    https://doi.org/10.1084/jem.20041579

    • Google Scholar
51. 1. Salanti A
    2. Staalsoe T
    3. Lavstsen T
    4. Jensen ATR
    5. Sowa MPK
    6. Arnot DE

    et al. (2003) Selective upregulation of a single distinctly structured var gene in chondroitin sulphate A-adhering Plasmodium falciparum involved in pregnancy-associated malaria

    Mol Microbiol 49:179–191.

    https://doi.org/10.1046/j.1365-2958.2003.03570.x

    • Google Scholar
52. 1. Smith JD
    2. Chitnis CE
    3. Craig AG
    4. Roberts DJ
    5. Hudson-Taylor DE
    6. Peterson DS

    et al. (1995) Switches in expression of Plasmodium falciparum var genes correlate with changes in antigenic and cytoadherent phenotypes of infected erythrocytes

    Cell 82:101–110.

    https://doi.org/10.1016/0092-8674(95)90056-X

    • Google Scholar
53. 1. Su XZ
    2. Heatwole VM
    3. Wertheimer SP
    4. Guinet F
    5. Herrfeldt JA
    6. Peterson DS

    et al. (1995) The large diverse gene family var encodes proteins involved in cytoadherence and antigenic variation of Plasmodium falciparum-infected erythrocytes

    Cell 82:89–100.

    https://doi.org/10.1016/0092-8674(95)90055-1

    • Google Scholar
54. 1. Taylor HM
    2. Kyes SA
    3. Newbold CI

    (2000) Var gene diversity in Plasmodium falciparum is generated by frequent recombination events

    Mol Biochem Parasitol 110:391–397.

    https://doi.org/10.1016/S0166-6851(00)00286-3

    • Google Scholar
55. 1. Trager W
    2. Jensen JB

    (1976) Human malaria parasites in continuous culture

    Science 193:673–675.

    https://doi.org/10.1126/science.781840

    • Google Scholar
56. 1. Tran TM
    2. Li S
    3. Doumbo S
    4. Doumtabe D
    5. Huang CY
    6. Dia S

    et al. (2013) An intensive longitudinal cohort study of Malian children and adults reveals no evidence of acquired immunity to Plasmodium falciparum infection

    Clin Infect Dis 57:40–47.

    https://doi.org/10.1093/cid/cit174

    • Google Scholar
57. 1. Untergasser A
    2. Nijveen H
    3. Rao X
    4. Bisseling T
    5. Geurts R
    6. Leunissen JAM

    (2007) Primer3Plus, an enhanced web interface to Primer3

    Nucleic Acids Res 35:W71–W74.

    https://doi.org/10.1093/nar/gkm306

    • Google Scholar
58. 1. Ward CP
    2. Clottey GT
    3. Dorris M
    4. Ji DD
    5. Arnot DE

    (1999) Analysis of Plasmodium falciparum PfEMP-1/var genes suggests that recombination rearranges constrained sequences

    Mol Biochem Parasitol 102:167–177.

    https://doi.org/10.1016/S0166-6851(99)00106-1

    • Google Scholar
59. 1. Warimwe GM
    2. Keane TM
    3. Fegan G
    4. Musyoki JN
    5. Newton CR
    6. Pain A

    et al. (2009) Plasmodium falciparum var gene expression is modified by host immunity

    Proc Natl Acad Sci USA 106:21 801–21 806.

    https://doi.org/10.1073/pnas.0907590106

    • Google Scholar
60. 1. Wunderlich G
    2. Alves FP
    3. Golnitz U
    4. Tada MS
    5. Camargo EF
    6. Pereira-da Silva LH

    (2005) Rapid turnover of Plasmodium falciparum var gene transcripts and genotypes during natural non-symptomatic infections

    Rev Inst Med Trop Sao Paulo 47:195–201.

    https://doi.org/10.1590/S0036-46652005000400004

    • Google Scholar
61. 1. Zhang Q
    2. Zhang Y
    3. Huang Y
    4. Xue X
    5. Yan H
    6. Sun X

    et al. (2011) From in vivo to in vitro: dynamic analysis of Plasmodium falciparum var gene expression patterns of patient isolates during adaptation to culture

    PLOS ONE 6:e20591.

    https://doi.org/10.1371/journal.pone.0020591

    • Google Scholar

Author details

1. Robert Noble

   Department of Zoology, University of Oxford, Oxford, United Kingdom

   Contribution

   RN, Analysis and interpretation of data, Drafting or revising the article

   Contributed equally with

   Zóe Christodoulou

   Competing interests

   The authors declare that no competing interests exist.

2. Zóe Christodoulou

   Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Oxford, United Kingdom

   Contribution

   ZC, Acquisition of data, Analysis and interpretation of data

   Contributed equally with

   Robert Noble

   Competing interests

   The authors declare that no competing interests exist.

3. Sue Kyes

   Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Oxford, United Kingdom

   Contribution

   SK, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

4. Robert Pinches

   Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Oxford, United Kingdom

   Contribution

   RP, Acquisition of data, Analysis and interpretation of data

   Competing interests

   The authors declare that no competing interests exist.

5. Chris I Newbold

   Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Oxford, United Kingdom

   Contribution

   CIN, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   chris.newbold@imm.ox.ac.uk

   Competing interests

   The authors declare that no competing interests exist.

6. Mario Recker

   Department of Zoology, University of Oxford, Oxford, United Kingdom

   Contribution

   MR, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   mario.recker@zoo.ox.ac.uk

   Competing interests

   The authors declare that no competing interests exist.

• 1,677

  views

• 192

  downloads

• 25

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.