Background

Malaria pathogenesis poses a major hindrance to development in many parts of the world with more than 500 million people suffering from the disease and at least one million dying from P. falciparum infection each year [1]. Disease severity has been associated with the accumulation of infected erythrocytes (IEs) in microvasculature of vital organs, such as the brain and placenta [2]. A key protein family involved in IE binding is the antigenic variant P. falciparum Erythrocyte Membrane 1 (PfEMP1) [3–5]. Each parasite genome contains about 60 var genes that encode PfEMP1 proteins [6], which are expressed in a mutually exclusive fashion at the IE surface [7, 8]. Switches in var gene expression allow parasites to evade the host antibody response and sequester at different microvascular sites in the body [9]. Therefore, further definition of the var gene family conservation, the factors regulating variant antigen gene diversification, and the expression of particular var genes during disease will provide critical insights into malaria pathogenesis and aid disease interventions.

Var genes have a two-exon structure [4]. The first exon is large (~3.5 to 9.0 kb) and encodes multiple adhesion domains called the Duffy binding-like (DBL) and cysteine-rich interdomain region (CIDR). The second exon is smaller (~1.0 to 1.5 kb) and codes for a more conserved cytoplasmic tail. Although PfEMP1 sequences are highly diverse, the adhesion domains can be grouped by sequence similarity [10] into seven types of DBL domains (α, α1,β, γ, δ, ε, and x) and four types of CIDR domains (α, α1, β, and γ) that have been used as criteria for dissecting PfEMP1 protein domain structures and binding functions.

The PfEMP1 proteins in the 3D7 genome have been arbitrarily classified into one of seventeen different protein architectural types based upon domain composition [6, 11] and divided into three major (A, B, and C) and two intermediate (B/A and B/C) groups on the basis of 5' upstream (Ups) sequence and chromosomal location [6, 12, 13]. Var group A genes have UpsA flanking sequences and are located in sub-telomeric regions transcribed toward the telomere, while group B consists of telomeric var genes flanked by UpsB sequences that are transcribed toward the centromere, and group C are flanked by UpsC sequences and are located in central chromosomal regions. Group B/A genes are very similar in location and transcriptional orientation to group B genes, but are located further from the telomere following other var genes or pseudogenes. In contrast, group B/C genes have an UpsB-like 5' flanking sequence, but are located in central chromosomal regions. Thus, it has been postulated that groups B/A and B/C represent transitional groups between the major groupings [13].

Inter-isolate comparisons have also revealed the existence of three unusual genes: var1csa, var2csa, and Type 3 var genes, which appear in nearly all parasite isolates [12, 14–19]. These semi-conserved homologs may have important roles in the host-parasite interaction. The PfEMP1 encoded by var2csa binds the placental adhesion receptor, chondroitin sulfate A (CSA), and therefore has a critical role in the pathogenesis of pregnancy associated malaria [20, 21], while no function has yet been ascribed to the proteins encoded by var1csa and Type 3 var.

The genomic organization of var genes may have an important role in var gene evolution. Similar to other variant antigen families, gene recombination or gene conversion between var paralogs may contribute to the rapid evolution of the gene family [22–24]. It has been hypothesized that the frequency of recombination between var genes may depend upon chromosomal location, gene orientation, and homology in the gene flanking sequence. Sequence and binding analysis of 3D7 var genes indicate that groups B and C PfEMP1 proteins bind the primary microvasculature receptor (CD36) while group A PfEMP1 proteins do not [12, 13, 25]. Thus, var gene recombination hierarchies may promote the evolution of PfEMP1 adhesion groups with different patterns of sequestration and disease. A fundamental question is whether the gene organization observed in 3D7 occurs in other parasite isolates and contributes to a general recombination mechanism shaping the variant antigen repertoire.

To investigate evolutionary mechanisms of the var gene family and provide new tools to study the role of PfEMP1 proteins in mediating cytoadhesion, we have sequenced var genes from isolate IT4/25/4 (IT4), which has maintained the ability to cytoadhere after in vitro adaptation [26–29], and compared these genes to the var repertories of the 3D7 genome reference isolate and of the HB3 isolate, for which sequence contigs were recently made available (Plasmodium falciparum HB3 Sequencing Project, Broad Institute of Harvard and MIT [30]). Although there are currently relatively few reports, isolate HB3 also maintains cytoadherence in culture [31, 32] and is therefore a useful addition to comparative var analyses.

All three parasites, IT4, HB3, and 3D7, have been cloned in vitro and represent single parasite genotypes. The IT4 parasite was originally isolated from Brazil [33], but is known to have undergone accidental cross-contamination at an early stage of its history after in vitro adaptation [34]. The HB3 clone was derived from the Honduras I/CDC isolate [35] and the NF54 parent to the 3D7 clone was isolated from an individual who lived near an airport in Amsterdam and never left the Netherlands [36]. Based upon genotyping and parasite population studies, IT4 groups with Asian isolates, 3D7 groups with African isolates, and HB3 represents Central America [37].

Despite progress in understanding the mechanisms of cytoadhesion and antigenic variation of PfEMP1, limited information about the factors regulating variant antigen diversification and the extent of repertoire overlap between parasite isolates exists. Most studies have relied on small var sequence "tags" amplified from the first DBL domain in PfEMP1 proteins [19, 23, 38–48]. The studies presented here represent the first comprehensive analyses of var genes across multiple parasite isolates. These comparisons reveal general principles of var gene organization that have become established across geographically diverse parasite isolates and provide powerful tools to study the cytoadherent and immunogenic properties of PfEMP1 proteins.

Results

The var gene repertoires from the IT4 and HB3 isolates

Conservation in the var gene 5' and 3' gene-flanking regions, the semi-conserved exon 2, and other domains allowed us to design a series of primers (Additional file 4: Table S1) and extend IT4 var tags that we had previously sequenced from the PfEMP1 DBLα, β, γ, and δ domains [49]. We sequenced 28 full-length var genes and 10 partial genes from the IT4 isolate [GenBank:EF158071-EF158105], in addition to the 10 full-length var genes that have been previously characterized (Figure 1). These genes represent all but 11 of the 59 IT4 sequence tags identified from other studies (Tables S2 and S3)[42, 49–51]. In order to estimate the proportion of IT4 var genes represented by these sequences, we searched the 1× coverage IT4 genome sequence at the Wellcome Trust Sanger Institute [52] for additional var sequences. Out of 949 reads with sequence similarity to the first exon of any known var genes, only ~15% do not overlap with our data set. Assembly of these reads shows that most of the non-overlapping reads represent small sequence fragments no larger than a single read and three partial gene fragments of 3–4 kb (data not shown). Thus, the var gene repertoire presented here includes partial or complete sequence for most IT4 var genes. Eight var genes were mapped to specific chromosomes using pulsed-field gel electrophoresis and Southern analysis (Figure 1, data not shown) and in some cases, intrachromosomal location (central versus sub-telomeric) was identified based on Apa I restriction fragment length [53]. The chromosomal locations of a further 13 var genes were based on previously published data [42].

Figure 1
figure 1

Schematic representation of the IT4 var gene repertoire. Gene names, Ups sequence type, domain architecture, chromosomal location, transcription orientation, and binding functions are listed. IT4 var genes are primarily assigned to different groups on the basis of 5' flanking sequence (Ups type) and chromosomal location when known. PfEMP1 proteins are comprised of multiple domains termed N-termimal segment (NTS), Duffy binding-like (DBL), cysteine-rich interdomain region (CIDR), C2, transmembrane (TM), and acidic terminal segment (ATS or exon2) which have been classified by sequence criteria into different types. The PfEMP1 proteins in the 3D7 clone were arbitrarily classified into 17 different protein architectural types on the basis of domain composition [6]. Types 18–25 (bolded) are unique to IT4. Chromosome locations are indicated as T, ST, SST: first, second, and third var genes from the telomere respectively. C: internal var genes. t: transcribed towards telomere, c: transcribed towards centromere. The chromosomal location of var2csa was determined in [78]. Accession numbers for newly sequenced genes are EF158071-EF158105.

Full size image

Analysis of the HB3 sequence contigs obtained from the 10× coverage genome sequence at the Broad Institute[30] identified 52 var genes that contain a DBLα domain as well as two var2csa homologs; 39 of the 54 var genes are full-length, 9 are incomplete and six are pseudogenes containing stops or frame-shifts (Figure 2). Examination of 5' and 3' flanking sequences (see Materials and Methods) enabled us to predict the chromosomal location of most genes (Figure 2), although in a number of cases they could not be assigned to specific chromosome ends. These predictions assume that recombination has not changed the arrangement of chromosome ends in HB3 and 3D7.

Figure 2
figure 2

Schematic representation of HB3 var genes. Genes are organized as in figure 1 and grouped according to 5' flanking sequence (Ups type) and chromosomal location. Partial (p) and pseudogenes (Ψ) are labeled. Bolded domain structure types are unique to the HB3 parasite line. Binding properties have not been mapped to HB3 PfEMP1 proteins.

Full size image

Comparisons of the IT4 and HB3 PfEMP1 protein domain architecture revealed representatives of most classes previously described in 3D7 (Additional file 1) [6] plus fourteen new types (Figures 1 &2). Of the 31 domain architectures, most types contain only a single representative per isolate and only seven (1, 5, 7, 8, 11, 13, and 17) are found in all three isolates. While five (2, 6, 9, 14, and 16) are found only in 3D7, eight (18–25) are unique to IT4, and six (26–31) to HB3 (Table 1). Moreover, the distribution of var genes among the shared domain architecture classes differs substantially between isolates. More than half (40/62) of the 3D7 var genes have a Type 1 architecture, but this type of var gene is rarer in IT4 (12/48) and HB3 (20/54). Conversely, IT4 contains six Type 11 var genes compared to only one in 3D7 and HB3; while HB3 contains six representatives of Type 27, which is not present in either 3D7 or IT4. The differential abundance of individual PfEMP1 types between parasite isolates and the presence of new PfEMP1 types in isolates IT4 and HB3 indicate considerable inter-strain plasticity in the variant antigen repertoire. Despite these differences in gene repertoires, the previously described tandem domain associations (DBLα-CIDR1, DBLβ-c2, and DBLδ-CIDR non-α types)[6, 10] are consistently preserved, indicating the potential structural and functional significance of these domain relationships.

Table 1 Var gene chromosomal locations and domain architectures across isolates.
Full size table

While seven protein architectural types are shared among the three isolates, most var genes have overall amino acid sequence identities of < 50% in individual domains (Additional file 2), even those within the same architectural type. However, three var genes (var1csa, var2csa, and Type 3 var) are highly conserved at the sequence level, with > 75% identity over multiple domains. Partial gene sequence tags for these three var genes have been amplified from many parasite isolates indicating their unusual conservation for the var gene family [19]. However, these isolate-transcendent members can have different copy numbers between parasite isolates. For instance, while 3D7 has three copies of the Type 3 var, we could amplify only one copy in IT4 and did not find any copies in the genomic sequence from HB3. In addition, HB3 contains two var2csa copies rather than one copy as in 3D7 and IT4. Although var1csa is present in all three parasites, it is a truncated pseudogene in 3D7 (first exon) and the second exon has a frameshift in the HB3 allele. Also, var1csa is believed to be truncated in many field isolates [18] and has a distinct gene transcription pattern from other var genes [54]. Therefore it may have a different biological role than other var genes.

Sequence comparison of 1.5–2.0 kb of 5' flanking sequence from the 3D7 var genes has defined five upstream types; UpsA, B, C, D, and E types [6, 11–13]. Phylogenetic analysis of 500 bp of 5' flanking sequences from the IT4, HB3, and 3D7 var genes revealed that IT4 and HB3 have similar sequence groupings as 3D7 (Figure 3). While we found similar classes as in previous studies, we have sub-divided UpsB into four sub-groups (B1–B4) and UpsC into two sub-groups (C1 and C2). This study also revealed that UpsD is very similar to UpsA (Figure 3), and that these categories can be more accurately referred to as UpsA1 (formerly UpsA) and UpsA2 (formerly UpsD). Notably, the proportion of var genes in each Ups type is similar between isolates (Figures 1, 2, S1).

Figure 3
figure 3

Phylogenetic comparison of var gene flanking regions from IT4, HB3, and 3D7 parasite isolates. A neighbor-joining tree was generated based upon 500 bp of 5' gene flanking sequence. Upstream groupings (Ups groups) with bootstrap support out of 1000 replicates are color shaded and labeled. Gene names have been removed from the figure for simplification.

Full size image

The 3D7 var gene repertoire has been previously categorized into three major (A, B, and C) and two intermediate (B/A and B/C) groups on the basis of Ups sequence and chromosomal location [6, 12, 13]. The HB3 and IT4 var genes can be similarly assigned to the three major groups on the basis of Ups sequence (Figure 3), but differences in chromosomal location between isolates argue for a modification of the sub-groupings. For example, the HB3 repertoire contains one UpsA1-associated var gene (HB3var6) that is in a central chromosomal cluster rather than the typical sub-telomeric location (Figure 4). Therefore, to allow for the future addition of 'atypical' genes, we have developed a naming system based upon var gene location and Ups sequence type. The Ups types (A1-2, B1-B4, C1-2, and E) when known are listed first followed by a chromosome location reference. T, ST, and SST refer to the first, second, and third var genes from the telomere respectively and C refers to central var genes. For example, we have now separated var group A into sub-groups A1C, A1ST, and A1SST to represent central and sub-telomeric var genes, respectively. Similarly, group B is divided to represent var genes with corresponding central (B1C, B2C, etc.) or telomeric (B1T, B2T, etc.) locations. Members of the B/A group previously defined by Lavstsen et al. [13] are now classified as B1ST, B2ST, etc. denoting both the 5' upstream type plus a distinct sub-telomeric chromosomal location which follows other var genes or pseudogenes [6].

Figure 4
figure 4

Chromosomal distribution of var genes in the 3D7 and HB3 parasite isolates. Var genes are color shaded according to 5' gene flanking Ups type (U indicates unknown) and labeled according to protein architecture. The chromosomal locations were predicted for 36 of the 54 HB3 PfEMP1 proteins based upon gene flanking sequence and comparison to the 3D7 reference genome (see methods). Arrows without an outline indicate pseudogenes.

Full size image

As observed previously [55], the var gene chromosomal location was highly predictive of 5' gene flanking sequence (Figure 4). For instance, nearly all centromere-transcribed var genes in the telomeric location were UpsB1 type (Figure 4). In contrast, members of the "transitional" B/C var group located in central chromosomal locations, associate with any of the 5' flanking sequences UpsB1-4. Interestingly, while HB3 contains a copy of the semi-conserved var1csa gene (domain architecture Type 17) with the expected UpsA2 sequence (formerly UpsD), the HB3 isolate is unique in having a second distinct PfEMP1 protein associated with the UpsA2 sequence (HB3var4, domain architecture Type 10). Thus, we have classified both within group A2ST. The highly conserved, and sequence divergent, var2csa remains in a separate Ups group (ET, EST, or ESST).

Although the general chromosomal distribution of var genes in sub-telomeric regions or central regions on chromosomes 4, 6, 7, 8, and 12 are similar between the three isolates, the genes themselves are not conserved with the exception of var1csa, var2csa and Type 3 var. Significantly,var genes in the same chromosomal location from the three isolates differ in both sequence and protein architecture (Figure 4, Additional file 4: Table S4). Furthermore, the order of var Ups types in central var gene clusters differ between isolates (Figure 4). These differences between isolates are evidence of gene recombination that has occurred within the coding and gene-flanking regions.

Var gene recombination

To study the genetic relationship of different var genes, we performed repertoire-wide nucleotide and amino acid sequence comparisons using a number of different approaches and visualization tools (see Materials and Methods). The Artemis Comparison Tool (ACT) [56] was used to visualize regions of similarity identified by reciprocal BLASTN searches of var exon1 nucleotide sequences. Using criteria of a word size of 90nt, > 90% identity we observed the gene duplication in 3D7 (PFD1235w and MAL8P1.207) and identified one gene duplication in HB3 (var2csaA and var2csaB). These gene pairs are nearly identical over their entire lengths (Figure 5 & Additional file 4: Table S5). These analyses also visualize the semi-conserved var genes (var1csa, var2csa and Type 3 var) identified above, which have multiple regions of high sequence similarity (> 90%) between isolates (Additional file 4: Table S5).

Figure 5
figure 5

ACT nucleotide comparison of var gene repertoires. Concatamers of var gene exon1 sequences were arranged sequentially by Ups type: UpsE, A, C, B2-4, and B1, (colored as indicated) with one genome per horizontal line. The isolate-transcendent var genes var2csa, var1csa and Type 3 var are positioned at the left end of the concatemer. BLASTN was performed with word length set at 90 nucleotides (filter for low complexity removed). The comparisons were viewed in ACT with a window size of 120 nucleotides, at minimum 90% identity to show segments of similarity between var genes. Diagonal bands connecting individual var genes are colored according to percent identity, shown in the scale diagram (inset). Band width corresponds to region of sequence identity. Var names are listed, in order of appearance, in Additional file 4: Table S7.

Full size image

ACT comparisons also identified several instances of partial sequence similarity (greater than 500 bp) between two var genes of the same isolate. Selected examples are shown (Table 2), illustrating the segmental nature of sequence similarity between genes, with only part of each sequence showing a high degree of sequence similarity to the other partner(s). The 3D7 repertoire contains four examples of such "chimeric" gene pairs, the current set of IT4 var genes has eight, and the HB3 repertoire has six (Table 2). In some cases (e.g. PFD0995c/PFD1000c/PFD1005c in 3D7), a var gene appears to be a "true" chimera of two different var genes (Figure 6), while in other cases the chimeras represent partial duplications between two var genes.

Table 2 High scoring BLASTn matches within and between isolates.
Full size table
Figure 6
figure 6

Examples of chimeric genes in the 3D7 and IT4 parasite isolates. Identical or nearly identical regions are indicated by brackets.

Full size image

Despite geographic separation of 3D7, IT4, and 3D7, similar examples of segmental sequence similarity greater than 500 bp can be seen between var genes of the different isolates; with five examples between 3D7 and HB3, three between HB3 and IT4 and one between 3D7 and IT4 (Table 2). Remarkably, one of the inter-isolate gene pairs, HB3var23 and PFL1950w, are both the first var genes in a central cluster on chromosome 12 (Figure 4). These two genes have nearly identical and highly distinctive UpsB4 type 5' flanking sequences (Figure 3) and share approximately 1000 bp of coding region identity (Table 2), but otherwise have diverged from one another. This region of similarity identifies a recombination event that likely predates the continental separation of P. falciparum isolates.

These analyses also demonstrate that most var genes share little sequence identity suggesting that the var genes have diverged extensively between parasite isolates and have undergone segmental recombination (Figure 5). However, the patterns of sequence identities are not random in that similarities preferentially occur between members of the same Ups group (Table 2). For instance, UpsA1 var genes are 7.3× more likely to share similarity with other UpsA1 genes than with different Ups groups and UpsB2-4 var genes are 8.6× more likely to share similarity within the UpsB2-4 group (Additional file 4: Table S6). The same trend holds for gene similarities involving smaller gene segments 90 nucleotides and up (Additional file 4: Table S6). An exception is central var genes, which contain "mixed" chimeras of UpsB and UpsC-associated var genes (Figure 6), suggesting both groups of central chromosome var genes are recombining.

To detect patterns of protein similarities, we conducted "repertoire-wide" dot-plot analyses using concatamers of var exon1 sequences ordered by isolate and 5' flanking sequence type. These analyses are designed to detect small windows of sequence similarity (80% amino acid identity, 30 amino acid window length) between PfEMP1 amino acid sequences and clearly show that UpsA PfEMP1 proteins share less similarity with UpsB and UpsC proteins than with other UpsA proteins (Figure 7). Conversely, UpsB and UpsC proteins are indistinguishable in terms of their degree of sequence identity with each other. This analysis combined with the analyses of individual domains (Additional file 2) shows approximately as much overall repertoire similarity within as between these geographically diverse strains.

Figure 7
figure 7

Dotplot comparisons of PfEMP1 protein coding sequence. The extracellular binding region of PfEMP1 proteins are organized by parasite isolate and 5' Ups sequence type. Uncl (unclassified) refers to sequences in which the Ups sequences have not been determined. Dot plot parameters include a window length of 30 amino acids and percent identity of 80% or greater.

Full size image

To identify the regions of similarity between PfEMP1 proteins, the dotplot matches were plotted along the length of individual proteins. Overall, the DBL1 domains in PfEMP1 proteins tend to have the most similarity between proteins, although there are regions of similarity in some CIDR domains (Additional file 3). Most of the similarity between PfEMP1 proteins, including between the B and C groups, is associated with semi-conserved homology blocks in DBL domains (Additional file 3). These homology blocks correspond to structural elements in solved structures [57, 58]. These analyses also clearly illustrate that var2csa and Type 3 proteins share almost no identity with other PfEMP1 proteins. Curiously, the rosetting-associated IT4var60 protein is not related to other UpsA proteins over most of its length (data not shown). However, unlike the semi-conserved Type 3 var or var2csa, HB3 and 3D7 do not have a var60 homolog. Although this result suggests that IT4var60 is not recombining with other var genes, more study is needed to determine its conservation in the parasite population. Taken together, these sequence comparisons support the hypothesis that var genes have differentiated into separately recombining groups that may be important to the evolution of the structure and function of PfEMP1 proteins.

Discussion

While the sequencing of the 3D7 genome has contributed greatly in determining PfEMP1 functions and genetic diversity [reviewed in [11]], the associations of var gene repertoires within and between parasite isolates and the factors regulating variant antigen diversification remain largely unknown. To gain understanding into the evolutionary mechanisms shaping the variant antigen repertoire we present here the nearly complete var repertoire of a cytoadhesive laboratory isolate, IT4/25/5, and compare it to the 3D7 genome reference isolate and the recently sequenced HB3 genome.

Despite the enormous diversity of these genes, several features of the var gene family are conserved across isolates including var groupings based upon central or telomeric chromosome location and 5' flanking sequence that may have an important role in the evolution and function of var genes. It has been hypothesized that an original ancestral var gene was duplicated and diverged into the three main var groups (A, B, and C) and subsequently into additional transitional groups [13]. This interpretation is supported by our analyses showing similar categories of var genes in all three parasite isolates.

Based upon sequence comparisons, B and C groups are more similar, even though these genes tend to occupy different chromosomal locations at sub-telomeric and central chromosomal regions, respectively. However, the regions of similarity are predominantly associated with semiconserved homology blocks that are predicted to form the structural scaffolding for the DBL adhesion domains. Conversely, the group A genes differ greatly from the B and C groups while the coding region of the three isolate transcendent var genes, var1csa, var2csa, and Type 3 var genes have unique features and are different from all other var genes. However, these isolate transcendent var genes are more related to the UpsA group in that they are sub-telomeric, transcribed towards the telomere, and have 5' gene flanking regions that most resemble the UpsA type.

Repertoire-wide sequence comparisons show that most gene similarities occur between genes within the same var group, particularly for gene segments larger than 500 bp. An exception is central var clusters, which contain both UpsB and UpsC-associated var genes. While the functional significance of these different 5' promoter types is not completely understood [59], these two sets of central var genes appear to be recombining with each other. Taken together, these analyses suggest that var gene recombination preferentially occurs within var groups, with the exception of the semi-conserved var homologs that appear to recombine on their own. Further var gene comparisons of parasites undergoing more frequent recombination in nature or parasite crosses will be of interest to determine the relative frequency of intra- versus inter-group gene recombination. These findings provide insight into the mechanisms that generate antigenic diversity in P. falciparum through gene recombination hierarchies, and may have parallels in other variant antigen gene families in Plasmodium and other organisms.

The mechanisms of gene recombination/conversion are not well studied in P. falciparum. Sequence comparisons and restriction fragment length polymorphism analysis of parasite crosses and population studies suggest that both small (~100–200 nt) and larger recombination events contribute to var gene evolution [19, 22, 23]. Here, we observe that chimeric junction sites are often not "clean" breakpoints and have smaller sections of 90–95% identity 500 bp upstream and/or downstream of the central homologous region (data not shown). This feature may relate to a mechanism of recombination. Control of var gene expression has been connected to silence-inducing regulators of gene expression (e.g. Silent Information Regulator protein 2, SIR2) and chromatin packaging [60, 61]. Recent studies have shown a possible link between factors involved in transcription regulation (including SIR2) and recombination (reviewed in [62, 63]). It is interesting to speculate that in P. falciparum, factors that are silencing/controlling var gene expression may also be involved in the recombination and gene conversion mechanisms.

From a study of 3D7 var genes expressed after antibody selection, it has been hypothesized that group A may contain common antigenic types that are responsible for severe disease [64]. Although the duplicated 3D7 UpsA var genes, PFD1235w and MAL8P1.207, have been proposed as a fourth isolate-transcendent variant, termed var4, our analyses do not support this conclusion since a var4 homolog was not found in IT4 or HB3, although the HB3 isolate contained a match over a portion of the gene. In addition, var4-like gene fragments were not common in a global survey of parasite isolates using gene-specific primers [19]. Instead, this observation may represent one of a number of between-genome var chimeras which are not present in all parasite genomes. More study is required to determine which segments of var4 are maintained in the parasite population and the extent to which the same segments are shared by different parasite isolates.

More generally, with the exception of the Type 3 var genes and var1csa, the UpsA-associated var genes are not highly conserved between the three isolates. This observation reinforces findings of high genetic diversity of UpsA-associated DBLα tags from a global collection of parasite isolates [19]. Various factors may influence the stability of large var gene segments across a parasite population, including malaria endemicity, the frequency of mixed infections, or functional selection on that gene segment for binding. The diversity of the UpsA var genes suggests that antibody cross-reactivity between different parasite isolates does not necessarily imply the presence of isolate-transcendent var genes, but may be due to cross-reacting antibody epitopes on different PfEMP1 sequences. Although the possibility that a subset of var genes may be associated with severe malaria remains, these genes may not be as conserved across parasite isolates as the pregnancy malaria vaccine candidate var2csa.

The concept of a recombination hierarchy has implications for the evolution of parasite virulence and disease investigation. The conservation of var groupings across isolates raises the possibility that var groups may be diverging and/or evolving in characteristic patterns. For instance, group A var genes, with the exception of the Type 3 var genes, encode larger proteins with more complex domain compositions and have different protein head structures (DBL1-CIDR1 domains) from other var groups (Figures 1 &2). In contrast, the relatively small Type 1 proteins, which consist of four adhesion domains, are not associated with group A in any of the isolates. It has been suggested that immune selection can cause polymorphic antigens to self-organize into sets of non-overlapping variants within a population [65]. Increased frequency of inter-locus recombination or gene conversion may also act as a homogenizing force leading to the functional and structural specialization of different gene groups [25]. Interestingly, the proportion of small to larger PfEMP1 proteins and the distribution of PfEMP1 architectural types differed between isolates. Given the different selective pressures for binding and immune evasion, it may be to the parasite's advantage to have different sets of recombining genes [66, 67]. These sets might include genes optimized to promote rapid parasite growth and transmission in the non-immune host, diversified genes that promote parasite transmission and persistence of infection in the face of higher levels of host immunity or organ-specific variants that expand parasite tropism to new host tissues, such as the placenta.

Unlike many isolates that have been adapted to in vitro cultivation, the IT4 genotype stably maintains the cytoadherent phenotype and therefore has become the primary model for this virulence determinant. CD36 binding, intercellular adhesion molecule 1 (ICAM-1) binding, and infected erythrocyte rosetting, or the binding of infected erythrocytes to uninfected erythrocytes, have been shown to reside in multiple different IT4 PfEMP1 proteins [for review, see [68]]. Although proteins that bind the same host receptor frequently use the same type of binding domain [11], the overall protein architectures are highly distinct. For instance, three ICAM-1 binding PfEMP1 proteins (A4tres, A4var, and IT-ICAM-1) all use DBLβ c2 domains but have different domain structure types and two rosetting PfEMP1 proteins, R29var [51] and FCRS1.2var1 [51, 69] also have different domain structure types and bind different receptors on the erythrocyte surface. Our study completes the sequences for three additional IT4 var genes upregulated in rosetting parasite clones, which were previously identified by only their DBLα tag sequences (IT4var1, IT4var27, and IT4var60) [70]. Overall, the rosetting var protein structures are not highly related (Figure 1), although the DBL1α1 domain is similar between the UpsA-linked R29var and IT4var60 predicted proteins, which may be significant because DBL1α1 is an erythrocyte binding region in R29var [51]. In contrast, the other three rosetting-linked var genes (IT4var27, IT4var1, and FCRS1.2var1) have DBLα domains instead of DBLα1 domains and associate with Ups B, C, or unknown Ups sequence. While these comparisons suggest that rosetting PfEMP1 proteins are not restricted to particular var groups, further study is needed to determine whether rosetting proteins in the same var group use a similar constellation of erythrocyte receptors.

Conclusion

A detailed understanding of the molecular mechanisms responsible for malaria pathogenesis is lacking, partly because of the complexity of the var gene family and the inability to model cytoadhesion with most culture-adapted laboratory isolates. In this study, we determined the var gene repertoires from the IT4 and HB3 isolates and provide evidence for a recombination hierarchy that shapes the evolution of the PfEMP1 virulence determinant. Furthermore, determination of the nearly complete var gene repertoire from the cytoadhesive IT4 parasite genotype, which has been adapted to both grow in the laboratory and infect new world monkeys, provides a unique capability to model cytoadhesion and immune acquisition in vitro and in vivo. Future binding and expression studies with cytoadhesive laboratory isolates, such as IT4 and HB3, in conjunction with analyses of the fully sequenced genomes will allow us to classify PfEMP1 proteins into biologically meaningful subsets and greatly accelerate understanding into malaria pathogenesis and immune evasion.

Methods

Parasites

Var genes were cloned from genomic DNA of the A4 clonal line. The A4 clone was originally derived by micromanipulation from P. falciparum isolate IT4/25/5 [29]. The IT4/25/5 isolate is one of several isolates including FVO, FCR3, and Palo Alto that appear to have a common genetic origin due to a laboratory cross-contamination event [34].

Long PCR amplification of var gene sequences

Larger var gene PCR products were amplified from genomic DNA using previously described techniques [49]. PCR primers (Additional file 4: Table S1) were designed to the different types of var gene flanking sequence or the relatively conserved var exon 2. These primers were paired with gene-specific primers from small sequence tags that had been amplified from internal domains in IT4 PfEMP1 proteins [49]. PCR reactions were done using TaKaRa LA Taq™ polymerase (Fisher) following the manufacturer's recommendations and supplied buffer. 50 ul reactions containing 50 ng template, 1× buffer, 0.4 mM dNTPs (each), 2.5 mM MgCl2, 0.5 μM primers, and 2.5 U enzyme were run in a DNA Engine Dyad™ Peltier Thermal Cycler from MJ Research. PCR conditions were 1 cycle of 94°C for 1 min followed by 35 cycles of 98°C for 1 min, primer annealing temperature for 1 min, and an extension temperature of 62–68°C for 8–18 min. Primer annealing temperatures were 0–5 degrees lower the TMs listed in Additional file 4: Table S1. Sequencing was performed on PCR products that were directly hydosheared and cloned into a sequencing vector or were first cloned into the pCR®4-TOPO vector from Invitrogen before hydoshearing and subcloning into the sequencing vector. Sequences were assembled using the PHRED/PHRAP/CONSED software suite [71]. To confirm that recombination had not occurred during the PCR reaction or bacterial cloning, specific oligos were designed along the length of var genes and independent PCR reactions were performed on genomic DNA.

Var gene chromosome assignation by pulsed field gel

IT4/FCR3 parasites were suspended in agarose blocks, then the intact chromosomes were size-fractionated on pulsed-field gels as described [72]. Gels were depurinated for 10 min in 0.25 M HCl, rinsed, then capillary transferred to Hybond N+ (GE/Amersham) in 0.4 M NaOH. Blots were hybridized at 50°C with DBLα tag probes corresponding to var genes IT4var1,IT4var5, IT4var25, IT4var27,IT4var 29,IT4var33, IT4var60 and A4Tres, as detailed previously [73]. Chromosome-central location of IT4var1 and IT4var27 was confirmed by ApaI digestion and size separation on pulsed-field gels, with hybridization to these tags at 60°C. Subtelomeric var genes lie on relatively short ApaI fragments [53], and these two genes are on large (> 400kbp) ApaI fragments.

Extraction of var gene sequences from public genome project information

Var genes were identified in HB3 contigs downloaded from the Plasmodium falciparum HB3 sequencing project, Broad Institute of Harvard and MIT [30]. Contig assembly 1 was used, which contains approximately 10× genomic coverage. To identify var genes, sequences were searched for a common DBLα motif, DIGDI, using Artemis genome viewer (Rutherford et al 2000). Var genes retrieved in this manner were confirmed by comparison to results from BLASTN with the full DBL1α sequence of PFA0005w, at the Broad Institute malaria website. HB3 homologs for var2csa were identified by BLASTN search at the same website, using the 3D7 allele. HB3 pseudovar genes were not confirmed by reamplification, but had approximately the same level of sequence coverage as other var genes (8–10×). Where possible, the var gene chromosomal context was also noted, using %GC content graphs to locate both telomeres and GC-rich DNA elements, which are short sequences associated only with central var genes [74]. The predicted HB3 chromosomal assignments were based upon comparison to the 3D7 isolate using the NUCmer program in MUMmer to identify sequence identities in var gene flanking regions [75]. Sequence data for P. falciparum 3D7 var genes and unassembled IT4 var reads were obtained from The Sanger Institute website [52]. Sequencing of P. falciparum IT4 is a component of the BioMalPar Consortium. For the ACT comparisons, two new 3D7 var genes are included, MAL7P1.212 and MAL8P1.220, which appear in the latest annotation [52]. These are both the most common type, Type 1, with UpsB type promoters, bringing the total number of this type in 3D7 up to 40 (out of 61 var genes).

Sequence analysis

PfEMP1 domain classification was performed according to previous criteria [10]. Neighbor-joining trees for all of the domains and flanking regions were generated using ClustalX for multiple alignments and PAUP*4.0b10 (* Phylogenetic Analysis Using Parsimony and other methods) [76]. Bootstrap analysis was performed with 1000 replicates. Gap opening and gap extension penalties of 5.0 and 0.05 or 10 and 0.1 were used for amino acid and DNA alignments, respectively. Percentage sequence identities of DBL, CIDR, and C2 domains were calculated using the algorithm in DNAStar MEGALIGN, version 5.0 based upon a ClustalW alignment. Means and standard deviations of these percentages were calculated and plotted in Excel.

Dotplot analysis was performed on concatemers of the variable extracellular region (exon1) sequences ordered by isolate and Ups type using the programs Megalign and DSGene. A percent identity matrix was used for all parameters tested (window length and percent identity threshold). To visualize the alignment results at the level of individual proteins, the output alignments from the dotplot were collected, and for each alignment, the aa positions of the alignment and Ups category of the pair of proteins were determined. For each aa position, the number of "alignment hits" from each Ups category was counted and plotted along the length of individual proteins. Microsoft Excel was used to generate plots of the number of hits from 3D7 PfEMP1 proteins of the different Ups types along the length of individual proteins for both IT4 and 3D7 PfEMP1 proteins. Based upon the distribution of var genes in the 3D7 isolate, the maximum number of hits at individual amino acid positions for genes of each promoter type is UpsA (9), UpsB (22), UpsC (13), UpsA2 (formerly UpsD; 1), UpsE (1), and UpsB2-4 (13, based upon previous published 2000 bp tree) [6, 12].

The Artemis Comparison Tool (ACT) was used to view exon1 for IT4, 3D7 and HB3 var genes. For each genome, a concatemer of var exon1 nucleotide sequences (from the start ATG to the splice donor site) was created. Sequences were organized by Ups group, and a string of 30 N's placed between each exon1 pair to clarify gene borders. Local BLASTN, (word length 90 nucleotides, filter for low complexity removed) was performed in all possible combinations between IT, 3D7 and HB3 var exon1 concatemers. These comparisons were then viewed with ACT [56, 77], with different window sizes (90 to 510 nucleotides), 90% minimum identity, and self-matches removed.