The evolutionary dynamics of variant antigen genes in Babesia reveal a history of genomic innovation underlying host–parasite interaction

Babesia spp. are tick-borne, intraerythrocytic hemoparasites that use antigenic variation to resist host immunity, through sequential modification of the parasite-derived variant erythrocyte surface antigen (VESA) expressed on the infected red blood cell surface. We identified the genomic processes driving antigenic diversity in genes encoding VESA (ves1) through comparative analysis within and between three Babesia species, (B. bigemina, B. divergens and B. bovis). Ves1 structure diverges rapidly after speciation, notably through the evolution of shortened forms (ves2) from 5′ ends of canonical ves1 genes. Phylogenetic analyses show that ves1 genes are transposed between loci routinely, whereas ves2 genes are not. Similarly, analysis of sequence mosaicism shows that recombination drives variation in ves1 sequences, but less so for ves2, indicating the adoption of different mechanisms for variation of the two families. Proteomic analysis of the B. bigemina PR isolate shows that two dominant VESA1 proteins are expressed in the population, whereas numerous VESA2 proteins are co-expressed, consistent with differential transcriptional regulation of each family. Hence, VESA2 proteins are abundant and previously unrecognized elements of Babesia biology, with evolutionary dynamics consistently different to those of VESA1, suggesting that their functions are distinct.


INTRODUCTION
Antigenic variation in pathogens is an adaptation to protective host immunity. It describes the serial replacement of variant antigens situated on the pathogen surface during an infection, causing the negation of immune responses mounted to the preceding antigen. Ultimately, the host cannot gain effective immunity and so suffers persistent infection and recurrent disease. Antigenic variation is a tangible expression of the coevolutionary arms race between hosts and pathogens and has evolved on several occasions in diverse lineages (1). This includes the etiological agents of prominent diseases of humans and animals, such as malaria (2), trypanosomiasis (3,4), influenza (5), bacterial meningitis (6) and anaplasmosis (7). Since antigenic variation is fundamental to circumventing the immune response, it plays a central role in pathogenesis and virulence and is one of the main obstacles to developing effective vaccines (8)(9)(10).
This study investigates antigenic variation in Babesia spp., apicomplexan hemoparasites that cause a tick-borne disease in animals and occasionally in people (11). Babesiosis has a global distribution, and while found predominantly in tropical and sub-tropical regions, it frequently reaches far into temperate regions (12). Bovine babesiosis has a significant negative impact on livestock productivity in both developed and developing countries. While most infections are asymptomatic, mild infections can cause fever and diarrhoea while severe infections, of the type caused by Babesia bovis in cattle, lead to acute haemolytic anaemia and may trigger a cerebral disease that is likely inflammatory in origin and almost invariably terminal. Here we examine two additional species that cause babesiosis in livestock, Babesia bigemina (13) and Babesia divergens (14). These three species are members of the Babesia sensu stricto clade and, of the three, B. bigemina and B. divergens are most closely related (15). Babesiosis in humans is uncommon; however, B. divergens infections are increasingly reported in immuno-compromized or splenectomized patients, often causing death (16). Babesia microti is another human parasite but will not be considered here since, despite its generic name and the presence of three ves-like genes in its genome (17), it is relatively distant from Babesia sensu stricto in phylogenetic terms and offers limited value as an out-group in this study (15).
Babesia spp. infections begin with invasion of host red blood cells and continue through cycles of intraerythrocytic re-invasion. Secreted parasite proteins are present on the surface of infected red blood cells (IRBC). Variation in these epitopes first indicated that serological responses to B. bovis infections are isolate-specific and do not provide lasting immunity (18). Antigenic variation was directly inferred from successive waves of gain and loss of immuno-reactivity to the major expressed IRBC surface antigen, and changes in antigen size, over the course of a clonal infection (19). The proteins responsible for these responses in B. bovis are the heterodimeric variant erythrocyte surface antigen (VESA), encoded by the multi-copy ves1α and ves1β gene families in B. bovis (20,21). ves1 genes are expressed from specific sites known as loci of active transcription (LAT) that typically contain a ves1α gene and a ves1β gene in divergent orientation (22). ves1 gene expression is regulated by bidirectional promoters within the LAT intergenic region (23) and involves localized chromatin remodeling (24). This precise regulatory mechanism is thought to ensure monoallelic expression, in which only the 'active' ves1 gene within the LAT is expressed while all other loci are silenced; and recent analysis of ves1 complementary deoxyribonucleic acid (cDNA) in B. bovis C9.1 supported this view (25).
The ves1 gene repertoire of B. bovis has been established from its complete genome sequence (26). This family contains 72 ves1α, 43 ves1β and four atypical genes that do not conform to either structure; 66 ves1 (55%) are arranged in divergent orientation, i.e. potential LATs. These loci are distributed throughout the chromosomes, and overall, 75% of ves1 genes are found in the interstitial chromosome regions, rather than showing any positional bias towards subtelomeres. LATs also tend to contain a third locus, the small open reading frame (smORF) genes (26), whose function is unknown. Ves1α and ves1β genes have different gene architectures; canonical ves1α genes consist of three exons and two short introns; ves1β genes vary widely in intron number and length, with up to 11 introns (20,22,26). Due to the relative simplicity of the ves1α gene structure and lack of structural variation between paralogs relative to ves1β, it was suggested that ves1α may have originated through the retrotranscription of an incompletely spliced ves1β gene transcript (22).
The ability of pathogens to employ antigenic variation in the long-term depends on the generation of novel variant antigens through evolutionary change. Genomic research in two models of antigenic variation, var genes in the malaria parasite Plasmodium falciparum (27)(28)(29) and vsg genes in the African trypanosome Trypanosoma brucei (30,31), have begun to explain how variant antigens change over evolutionary time, and how their dynamics differ from 'ordinary' genes (32). By taking a similar approach, we can examine how antigenic diversity is generated in Babesia. The evolution of variant antigen genes is rapid; gene duplication is more frequent and sequence divergence faster relative to the genomic background (27,33). Rapid change is due in part to frequent gene conversion, which is a strategy independently evolved by many pathogens to promote antigenic diversity (34). Phylogenetic comparisons of both var and vsg show that rapid change results in species-specific gene repertoires and a lack of orthology [i.e. genes from different species form mutually exclusive clades (31,35)], as well as structural differentiation of the repertoire for life stage or disease stage-specific expression (36,37), and the derivation of novel, invariant genes that appear to acquire new functions (30,31).
Our understanding of ves1 gene diversity and antigenic variation is largely based on the analysis of B. bovis. In this study, our aim is to characterize the variant antigen repertoires in multiple Babesia species using comparative genomics and, to identify how variant antigens originated, how repertoires in other Babesia spp. vary from the model based on B. bovis, and to examine the evolutionary processes that generate antigenic diversity on a population scale. We have sequenced the genomes of several strains of B. bigemina and B. divergens. Our results show that, while ves1 genes are widespread, primary structure and genomic repertoire varies substantially, such that features in B. bovis like heterodimerism and organization into LATs, may not occur in other species. This study reveals a history of constant innovation in Babesia genomes with respect to genes implicated in antigenic variation, and, in so doing, we have discovered that novel ves-like gene families, thought to be secreted, have been independently derived from canonical ves1 genes in different species. The expression profiles and evolutionary dynamics of these novel ves-like genes differ from canonical ves1 genes, suggesting that they are exposed to fundamentally different functional constraints.

Parasite isolation
The B. bovis C9.1 clonal line is a derivative of the Mexico isolate. Its derivation from the MO7 clonal line has been described before (19). The B. bigemina BOND clone C7-1 was derived from an Australian isolate in 1991 by the Queensland Department of Agriculture, Fisheries and Forestry. It was cloned by limiting dilution in in vitro culture (38) and then stored under liquid nitrogen until 2004 when it was removed and inoculated into a splenectomized Bos taurus calf. Parasites were first detected five days after inoculation of the calf and 2 l of peripheral blood was collected after a further two days. Parasites were purified from erythrocytes and white blood cells using a saponin lysis method (39) and then DNA was extracted using phenolchloroform. The B. bigemina JG29 and Puerto Rico (PR) isolates were kind gifts from T.F. McElwain and were maintained in in vitro culture for isolation of nucleic acids and proteins. The B. bigemina BbiS3P strain was isolated from an 18 months old Braford heifer, chronically infected, in the Salta province, Argentina, in March 2005. The isolation was achieved in vitro, by cultivation of heifer's erythrocytes previously washed with buffer VYM. The B. divergens 1802A strain was isolated in May 1988 from a cow in the Le Cher department of France with acute babesiosis. The B. divergens strain Rouen1987 (40) was a kind gift of H. Vial, Montpellier.

Cell culture
The B. bovis C9.1 clonal line, and B. bigemina JG29 and PR isolate parasites were cultivated in vitro under microaerophilous stationary phase conditions as described (41), with slight modifications (42). Closed herd bovine blood donors maintained at the University of Florida were used in all work with these strains. All animal procedures were approved by the University of Florida Institutional Animal Care and Use Committee. The B. bigemina BbiS3P strain was also cultivated in vitro and stored frozen in liquid nitrogen. A culture expansion to achieve ≥50% of parasitized erythrocytes was accomplished to purify merozoites for gDNA extraction. The B. divergens 1802A strain was obtained by infecting a single gerbil with the cow blood. In vitro culture was initiated from this isolate. After initiation, we cloned the culture by the limiting dilution technique to avoid producing DNA from a mixed parasite population. One clone was selected and expanded in vitro immediately. The B. divergens Rouen1987 was cultivated in human red blood cells, blood group A+ in 10% human serum of blood group A+. Cultivation was in a 5% O 2 , 5% CO 2 atmosphere.

Sample preparation
The B. bovis C9.1 and B. bigemina JG29 and PR parasites were grown in erythrocytes depleted of contaminating DNA and bovine white blood cells by use of Whatman CF-11 cellulose (43). B. bigemina PR-infected erythrocytes were enriched to near homogeneity using Percoll gradients (44). Genomic DNA was isolated as described (45) with spooling at the final step. For protein analysis, cells were washed two times with phosphate-buffered saline, the supernatants were removed and the cells were flash-frozen on dry ice. B. bigemina BbiS3P free merozoites were isolated from stroma by using Percoll gradients. Merozoite parasites were pelleted and pellets were lysed at 58 • C for 1 h in lysis buffer (0.05 M Tris-HCl pH 8.0, 0.1 M EDTA, 0.1 M NaCl, 2% SDS) with 160 g of proteinase K (Invitrogen R ). gDNA was extracted with 1 vol of phenol/chloroform/isoamyl alcohol (Invitrogen R ), precipitated with ice-cold isopropyl alcohol and washed once with 75% ice-cold ethanol. Pellets were suspended in 50 l distilled water and were kept at −20 • C until use. Genomic DNA was prepared for B. divergens 1802A and Rouen1987 from cell lysates of monoclonal cultures using commercial kits (Promega and Qiagen, respectively).

Genome sequencing and annotation
The B. bigemina BOND genome was sequenced using capillary technology to 8X coverage and assembled with Phrap. The genome was manually improved in Gap4. Sequencing errors were corrected using a polymerase chain reaction (PCR) free Illumina library and the iCORN algorithm (46). The B. divergens 1802A strain was sequenced using the 454 Titanium platform combining reads from a genomic fragment paired-end library with 3 and 8 kb mate-pair libraries and assembled using Newbler v2.6. Sequencing errors were corrected using additional SOLiD v3 reads. Protein-coding genome annotation for B. bigemina BOND and B. divergens 1802A were generated by taking a consensus of several ab initio gene prediction programs. Annotation from B. bovis (26) was used to train Augustus (47), glimmerHMM (48) and SNAP (49). Each of these trained gene finders were then run on the target genomes and this annotation was combined in each case using Jigsaw (50). The annotation was improved manually. Three further B. bigemina genomes (PR, BbiS3P and JG29), the B. bovis C9.1 genome and the B. divergens Rouen1987 genome were sequenced using the Illumina GA platform. The libraries were prepared with the PCR-free protocol and sequenced to 76 read pairs with an insert size of around 350 bp. Illumina reads were assembled with velvet (51). The genomes were improved using the Post Assembly Genome Improvement Toolkit (PAGIT) pipeline (52), applying the contig ordering against the BOND reference using the program ABACUS (53) and the gap-filling step (54). For the B. bigemina genome a further gap-filling was performed, using gap-filler (55). Contigs smaller than 1 kb were ignored. Two methods were used to annotate these draft genome sequences and the outcome merged. First, annotation was transferred from the reference genomes (i.e. B. bovis T2Bo, B. divergens BOND and B. divergens 1802A) to additional strains using the Rapid Annotation Transfer Tool (RATT) (56). Genes models with errors were excluded. Second, Augustus (47) was trained with the complete reference gene set to perform ab initio annotation.

Transcriptomic analysis
Strand-specific messenger ribonucleic acid (mRNA) sequencing libraries were prepared from total RNA of B. bigemina PR strain (two biological replicates) using TruSeq stranded mRNA Sample Prep Kit LT (Illumina) according to manufacturer's instructions. Briefly, polyA+ mRNA was purified from total RNA using oligo-dT dynabead selection. First strand cDNA was synthesized using randomly primed oligos followed by second strand synthesis where dUTPs were incorporated to achieve strand-specificity. The cDNA was adapter-ligated and the libraries amplified by PCR. Libraries were sequenced in Illumina Miseq with paired-end 150 bp read chemistry. Strand-specific RNA-seq paired-end reads were mapped onto the B. bigemina PR genome with TopHat2 (57) with options '-library-type = fr-firststranded' and '-no-novel-juncs'. Genes in B. bigemina PR orthologous to B. bigemina BOND ves genes were identified from B. bigemina PR gene annotations and their corresponding normalized transcript abundances (FPKM values) were quantified from aligned reads using Cuffdiff2 (58). In case of multiple orthologs in PR for a particular BOND ves gene, the orthology with highest FPKM was selected.

Sample preparation for proteomics
Protein from B. bigemina PR cell lysates was dispensed into low protein-binding microcentrifuge tubes (Sarstedt, Leicester, UK) and made up to 160 l by addition of 25 mM ammonium bicarbonate. The proteins were denatured using 10 l of 1% (w/v) RapiGest TM (Waters MS Technologies, Manchester, UK) in 25 mm ammonium bicarbonate followed by three cycles of freeze-thaw and two cycles of 10 min sonication in water bath. Sample was then incubated at 80 • C for 10 min and reduced with 3 mM dithiothreitol (Sigma-Aldrich, Dorset, UK) at 60 • C for 10 min then alkylated with 9 mM iodoacetamide (Sigma-Aldrich, Dorset, UK) at room temperature for 30 min in the dark. Proteomic grade trypsin (Sigma-Aldrich, Dorset, UK) was added at a protein: trypsin ratio of 50:1 and samples incubated at 37 • C overnight. In a parallel experiment aiming to maximize protein separation, protein samples were also solubilized and resolved on a NuPAGE R Novex R 4-12% Bis-Tris Gel (Life Technologies Ltd., Paisley, UK) following manufacture's instruction. Four gel slices were excised and each digested with trypsin (Sigma-Aldrich, Dorset, UK).

Mass spectrometry
Peptide mixtures from both in solution digestion and 1Dsodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) were analysed by on-line nanoflow liquid chromatography using the nanoACQUITY-nLC system (Waters MS Technologies, Manchester, UK) coupled to an LTQ-Orbitrap Velos (ThermoFisher Scientific, Bremen, Germany) mass spectrometer equipped with the manufacturer's nanospray ion source. The analytical column (nanoACQUITY UPLCTM BEH130 C18 15 cm × 75 m, 1.7 m capillary column) was maintained at 35 • C and a flow-rate of 300 nl/min. The gradient consisted of 3-40% acetonitrile in 0.1% formic acid for 90 min then a ramp of 40-85% acetonitrile in 0.1% formic acid for 3 min. Full scan MS spectra (m/z range 300-2000) were acquired by the Orbitrap at a resolution of 30 000. Analysis was performed in data-dependent mode. The top 20 most intense ions from MS1 scan (full MS) were selected for tandem MS by collision induced dissociation and all product spectra were acquired in the LTQ ion trap. Ion trap and Orbitrap maximal injection times were set to 50 and 500 ms, respectively.

Proteomic analysis
Thermo RAW files were imported into Progenesis LC-MS (version 4.1, Nonlinear Dynamics, UK). Runs were time aligned using default settings and using an auto selected run as reference. Peaks were picked by the software and filtered to include only peaks with a charge state of between +2 and +6. Peptide intensities were normalized against the reference run by Progenesis LC-MS and these intensities are used to highlight differences in protein expression between control and treated samples with supporting statistical analysis (ANOVA and q-values) calculated by the Progenesis LC-MS software. Spectral data were transformed to mgf files with Progenesis LC-MS and exported for peptide identification using the Mascot (version 2.3.02, Matrix Science) search engine. Tandem MS data were searched against the predicted protein set of the B. bigemina BOND reference genome sequence. Search parameters were as follows: precursor mass tolerance set to 10 ppm and fragment mass tolerance set to 0.8 Da. One missed tryptic cleavage was permitted. Carbamidomethylation (cysteine) was set as a fixed modification and oxidation (methionine) set as a variable modification. Mascot search results were further processed using the machine learning algorithm Percolator. The false discovery rate was <1%. Individual ion scores >13 indicated identity or extensive homology (P < 0.05). Results were imported into Progenesis LC-MS as .xml files. At least two unique peptides were required for reporting proteins that were differentially expressed. Results from in solution digestion and gel slices of 1D-SDS PAGE are grouped using 'Combine analysed fractions' function in Progenesis LC-MS where statistical analysis are updated.

Annotation of ves-like genes
We searched for ves-like and smORF-like genes among B. bigemina and B. divergens translated open reading frames using hidden Markov models (HMM) built using HMMER v3.0 (http://hmmer.janelia.org/) from sequence alignments of BbovVes1α, BbovVes1β and BbovSmORF. Once this had revealed native ves1 sequences in B. bigemina and B. divergens (though not smORF), these sequences were used in turn to create new hidden Markov models (HMMs) for a second search, which identified ves2 in each case. Lastly, tBLASTx and BLASTn were applied to the B. bigemina and B. divergens genome sequences using native ves1 or ves2 sequences to identify any unannotated copies. ves1 structures are highly mutable, and therefore it is inadvisable to simply rely on annotation transfer and sequence homology with known genes in B. bovis, which may not adequately capture ves-like genes in other genomes. Coding sequences for the ves-like genes analysed in this study have been deposited on the Wellcome Trust Sanger Institute FTP site (ftp://ftp.sanger.ac.uk/pub/pathogens/Babesia/).

Comparative genomics
We used OrthoMCL v2.0.9 (59,60) to examine speciesspecific genes and gene families with species disparities in copy number. Protein sets from B. bovis T2Bo, B. bigemina BOND and B. divergens 1802A were clustered using Or-thoMCL, set to maximize cluster size (i.e. minimized sequence identity requirements for clustering). From these data we calculated the number of genes shared by all three species in single copy ('conserved') or as part of gene families with variable copy number ('semi-conserved'). We also extracted the number of ves1 homologs and ves1-like shortened forms (i.e. ves2) in each genome. Finally, we calculated the number of species-specific genes in single or multiple copies. In comparisons of chromosomal rearrangements across species, and in comparisons of conserved ves1/ves2 loci across strains, we used the Artemis Comparison Tool [ACT (61)] to visualize conservation in gene order.

Phylogenetic analysis
Translated nucleotide sequences for each sub-family were aligned in ClustalW (62) and then manually edited in BioEdit v7.1.3. (63). A phylogeny of all ves1-like genes was estimated by selecting representative samples of BbovVes1␣/1␤/1␥ , BbigVes1a/1b and BdivVes1 sequences and creating an 840 character DNA sequence alignment of the conserved C-terminal domain (corresponding to the transmembrane and cytoplasmic domains). The phylogeny was estimated using a GTR+G model in PHYML (64) with 100 non-parametric bootstraps and in MrBayes (65,66) under these settings: Nruns = 4, Ngen = 5 000 000, samplefreq = 500 and default prior distribution. Phylogenies of all sequences were estimated for each sub-family (combining reference and strain sequences) using a GTR+G model in PHYML. Bootstrap proportions were attempted but these were low, as is typical for large alignments of such hypervariable genes. Phylogenies of reference and strain sequences at conserved positions were estimated using the same approach for each species, for use in co-phylogenetic analyses (see below).

Co-phylogenetic analysis
Phylogenetic reconciliation is a method for resolving topological disparities between two phylogenetic trees that are expected to have the same topology and was initially developed for the comparison of parasite phylogenies with that of their hosts, and of gene family trees with species phylogenies (67,68). Disparities are resolved by positing evolutionary events from a general model of host-parasite or gene family coevolution (i.e. codivergence, duplication, loss and host switching). The statistical significance of topological congruence between two associated trees is typically assessed using permutation tests. Taking each reference genome and one other strain, we identified all ves-like loci that were conserved in position and contained a sequence in both genomes. After extracting these sequences, we estimated two phylogenies for each ves-like sub-family, one containing 'reference' sequences and another 'strain' sequences. Since these genes occurred in conserved positions we can assume that they evolved before the separation of reference and strain genomes, and therefore the two phylogenies should look exactly the same in the absence of any transposition between loci after separation (i.e. recombination). The significance of phylogenetic congruence was assessed for each sub-family by permuting the strain tree in Jane 4 (69).

Recombination analysis
Phylogenetic incompatibility describes the presence of multiple phylogenetic signals within a single sequence alignment and is the historical signature of recombination. The pair-wise homoplasy index (PHI) detects incompatibility between sites and is robust in the presence of rate heterogeneity (70), which might otherwise simulate the effects of recombination. P < 0.05 for PHI indicates significant incompatibility between sites within an alignment, consistent with recombination. For each ves-like gene family, PHI was calculated for 1000 sequence quartets selected at random from multiple alignments of full-length nucleotide sequences. Sequence triplets were attempted but failed to produce sufficient genetic variation for a viable test in most cases. The proportion of quartet alignments showing significant phylogenetic incompatibility (P pi ) was calculated. A second method for detecting sequence mosaicism was implemented using 3seq (71). 3seq carries out exhaustive comparisons of sequence triplets in a multiple alignment of nucleotide sequences to identify mosaics and returns a P-value for each triplet adjusted by a Dunn-Sidak correction for the multiple comparisons made in that run. P pi was again calculated from the proportion of sequence triplets with P < 0.05.

Relative structure and content of Babesia genome sequences
For a comparative analysis of the ves gene repertoire in Babesia, we have produced high-quality, draft genome sequences for an additional strain of Babesia bovis (C9.1), four strains of B. bigemina (BOND, PR, BbiS3P and JG29) and two strains of B. divergens (1802A and Rouen1987). Descriptions of these strains and genome sequences are given in Table 1. Variation in genome size, gene number and the proportion of coding sequences are related to the quality of sequence assemblies and sequence contiguity, which is lower in sequences produced from short reads only. The larger number of coding sequences in B. bigemina relative to other species is due to unique gene duplications of conserved gene families and to a greater number of species-specific sequences encoding hypothetical proteins. Nonetheless, clustering analysis of B. bovis T2Bo coding sequences combined with corresponding data from B. bigemina BOND and B. divergens 1802A shows that gene content is consistent between species and that variation in surface antigens occurs against a largely conserved genomic background. Figure 1 shows that coding sequences seen in all species represent 78-88% of genes; 68-81% of all coding sequences show oneto-one correspondence (i.e. perfect orthology). Therefore, species-specific genes represent between 12.2 and 21.5% of genes in these species, with B. bigemina displaying the highest proportion of unique features. These are maximum estimates since they include predicted protein sequences that failed to cluster, and so potential mis-annotated sequences could be designated as species-specific genes. Ves gene homologs comprise a large proportion of these species-specific genes: 4.7% of all coding sequences in B. bovis T2Bo, 8% in B. bigemina BOND and 11.2% in B. divergens 1802A. While the remaining unique sequences are not homologous to ves1, they are often predicted to be expressed at the cell surface. This prediction is based upon their overall structural similarities with the B. bovis VESA1 polypeptides, which are placed on the erythrocyte surface (42), including a wellconserved C-terminal end and predicted transmembrane domain. This indicates that the most dynamic features of these genomes are associated with the host-pathogen interface. nullnull

Ves-like loci coincide with chromosomal breakpoints
By mapping sequence scaffolds to the B. bovis genome, we determined that B. bigemina BOND has four chromosomes, as suggested previously (70); however, these are not co-linear with B. bovis chromosomes. Sequence alignment based on conserved gene order demonstrates that segmental inversion and chromosomal rearrangement have been very frequent since these Babesia species separated. Supplementary Figure S1A/B aligns various sequence blocks from the B. bovis T2Bo and B. divergens 1802A genomes that correspond to full-length chromosomes in B. bigemina BOND. The positions of coding sequences homologous to B. bovis ves1 genes are marked and these coincide with chromosomal breakpoints. When we inspect breakpoints in colinearity more closely, as shown in Figure 2, ves-like genes occur at homologous positions in all three genomes, but it is clear that these homologs have quite different size and orientation (Figure 2a). Other multi-copy genes (shaded yellow) also congregate at similar positions (Figure 2b), reminiscent of the proximity of smORF to ves1 in B. bovis T2Bo (22). In addition, and somewhat anomalously, B. bigemina BOND has 28 ves1b loci on chromosome 3 at positions not associated with genome rearrangements, which likely represent lineage-specific insertion events (Figure 2c). The presence of ves-like genes at corresponding positions in different Babesia species may be ancestral, (i.e. these genes are orthologs, descended directly from a progenitor at the same position in the ancestral genome), or it may be mechanistic, (i.e. these positions are prone to rearrangement in all species and ves-like genes, which are not orthologous, independently transpose to them as regions in which purifying selection is weak).

Ves-like sequences diverge rapidly after speciation
Ves1 genes are among the fastest-evolving genes in Babesia genomes; in comparisons between species, they display about 25% amino acid identity and 75% of ves1 genes occur in the top 25% of all genes when ranked by sequence divergence, which accounts for the skew towards low identity in frequency distributions of sequence divergence across the whole genome (see Supplementary Figure S2). Rapid divergence of VESA proteins is also evident in comparisons of genomic repertoire. The B. bovis T2Bo genome contains two families of ves1 genes (ves1α and ves1β), as well as the smORF gene family that are found interspersed among ves1 clusters. Through comparison with other genome sequences using BLAST and HMMER, we have identified ves-like genes in both B. bigemina BOND and B. divergens 1802A, although with important differences, as shown in Figure 3. Hereafter, we will prefix gene names with species labels, e.g. In addition to BbigVes1a and BbigVes1b, there is a third gene family in B. bigemina that we refer to as BbigVes2 (N = 116). These genes were identified due to their homology with the 5 -most 500 bp of BbigVes1b and they encode VESA-like proteins lacking the C-terminal domain of the canonical VESA1 or any GPI-anchor signal. These pre-  dicted proteins do not possess a PEXEL motif, which is required by some proteins in Plasmodium for secretion (73), but they do have a predicted signal peptide and, without any obvious means of membrane attachment, we suggest that they are secreted.  In B. divergens 1802A, we observe a single gene family homologous to ves1 (N = 202) that encode predicted proteins with a canonical VESA C-terminal domains for membrane attachment. This family includes two forms shown in Figure  3, one short and another longer due to a cysteine-rich, low complexity region towards the N-terminus that is absent from the short forms; otherwise these two forms are closely related and do not form monophyletic groups in phylogenetic trees (i.e. long and short forms are paraphyletic; see Figure 4). These genes are arranged in divergent orientation, suggestive of an LAT, in only three instances, although the abundance of gaps in the B. divergens assembly substantially limits our ability to quantify genomic context. The impression is, however, that ves-like genes in B. divergens are typically arranged in tandem at sub-telomeric loci, but rarely on opposing strands like the LAT. As with B. bigemina, we found no smORF homologs in B. divergens but we did identify ves-like genes that are homologous to the 5 region of ves1, encoding predicted proteins without transmembrane attachment. In B. divergens these genes separate into two families [referred to as BdivVes2a (N = 95) and BdivVes2b (N = 62)], both of which are reciprocally monophyletic with respect to ves1, and therefore represent distinct lineages rather than partial ves1 pseudogenes (see Figure 4).
With the discovery of ves2 in the B. bigemina and B. divergens genomes, it is intuitive to suggest that ves2 are analogous to smORF in B. bovis. There is no compelling evidence for homology between smORF and ves-like genes, and so it seems most likely that smORF has evolved in B. bovis independently to perform a role analogous to ves2. We chose to include smORF in our subsequent analyses because, as a multi-copy family of secreted proteins, showing clonal variation (74) and spatial association with ves1, smORF clearly share similar circumstances to BbigVes2 and BdivVes2. We checked the B. bovis genomes for genuine orthologs to ves2 and discovered a tandem gene pair (BBOV III002570/2580), which encode homologs to BbigVes2 in both B. bovis strains. Hence, these genes may be the sole remaining representatives of a BbovVes2 family now largely lost, perhaps replaced by smORF.
Ves1 genes share a 3 region that encodes the conserved VESA C-terminus. This is the only region for which a multiple sequence alignment can be made across all species. A maximum likelihood phylogeny was estimated from the 840-character sequence alignment using PHYML (see Figure 4a). The phylogeny indicates that, while the ancestor of these Babesia species possessed a ves1 gene family, the dimorphism evident in B. bovis (BbovVes1α/1β) and B. bigemina (BbigVes1a/1b) is not ancestral and has independent origins in these lineages. Given the large genetic distances between these paralogous families within the same species, this indicates very substantial structural innovation postspeciation on multiple occasions. If we compare the branch lengths between BbovVes1α/1β (0.5 substitutions per site) and BbigVes1a/1b (1.5), the paralogs in B. bovis appear to have diverged much less. This is consistent with the hypothesis that BbovVes1α evolved from BbovVes1β after the origin of B. bovis through reverse transcription, i.e. integration of an incompletely spliced BbovVes1β transcript, which was originally proposed to account for the lack of introns in BbovVes1α (22) . Furthermore, we have identified a third ves1 sequence type in the B. bovis T2Bo genome not previously reported (BbovVes1γ ; N = 3). Figure 4a shows that the branch length between BbovVes1γ and BbovVes1α/1β (1.5) is consistent with the distance between paralogous families in B. bigemina (1.46), suggesting that BbovVes1α may have substituted BbovVes1γ as one half of the heterodimer.
The relationships of ves2 are not shown in Figure 4a because ves2 lack the conserved C-terminal region of the canonical forms. Therefore, we generated a network based on pair-wise FASTA scores (Figure 4b), exhaustively comparing full-length VESA1 and VESA2 sequences from all species, to examine the suggestion that ves2 genes independently evolved from ancestral ves1 sequences. Canonical VESA1 sequences cluster together at the centre of the network, while protein sequences corresponding to BbigVes2 and BdivVes2a/2b share no connections. This is consistent with ves1 in all species having a single common origin, as implied in Figure 4a, while ves2 have separate ori- to the conserved C-terminal domain of VESA1 only (840 characters). A GTR+ model was applied. Support for principal nodes is indicated by nonparametric bootstraps and posterior probabilities from a Bayesian analysis using the same model. (B) Sequence similarity network based on FASTA scores generated from pair-wise comparisons of VESA1 and VESA2 amino acid sequences and generated using BioLayout Express v3.0. Individual sequences are represented by spheres, shaded by gene family, connected by lines that represent sequence homology. The network was organized such that spheres edge length is minimized and spheres are positioned nearest to their closest relatives. A lower threshold has been applied to exclude poor sequence matches, leaving only the strongest similarities as determined by FASTA. SMORF sequences were included, but no FASTA scores exceeded the threshold. Ves-like gene families are labelled as described in the text; a single Babesia bovis sequence that clusters close to BbigVes2 (BBOV III002580) is shown with a red circle.
gins. BdivVes2a/2b unambiguously cluster most closely to BbigVes1a, suggesting that they are derived from an ancestor of BbigVes1a now lost or unrecognizable in B. divergens. BbigVes2 clusters most closely to BbigVes1b and BbovVes1α, suggesting that it was derived from an ancestor of these two gene families. The rare BbovVes2 sequences introduced above are positioned intermediately between BbigVes2 and BbovVes1β in Figure 4b (circled in red), suggesting that may share the same origin as BbigVes2. Therefore, given that a ves1 structure is present in all species (and, indeed, in B. microti (17)), but a common ves2 structure is not widespread, we may infer that ves2 in B. bigemina and B. divergens are not orthologous, and that they have been derived from ves1 (and not vice versa) on separate occasions.

Multiple VESA2 proteins may be expressed with greater abundance than VESA1
Genome comparison indicates that ves-like genes lacking the conserved 3 region have been independently derived from the canonical ves1 genes during Babesia evolution. Sequence comparison indicates that these ves2 genes are not non-functional ves1 pseudogenes but, in fact, encode highly abundant and polymorphic proteins. To explore the biological differences between ves1 and ves2, we carried out proteomic analyses of global gene expression in B. bigemina PR. We identified 1777 peptides corresponding to 366 predicted proteins in the B. bigemina BOND reference genome (i.e. 8.2% of the predicted proteome; see Supplementary  Table S1). Figure 5 shows the abundance of VESA1 and VESA2 protein relative to each other, and to all other proteins identified in our data set. Most importantly, it shows that of the 15 different VESA2 proteins that are expressed, seven are more abundant than the most abundant VESA1 protein; in general VESA2 are significantly more abundant than VESA1 (P < 0.001; randomization test). The distribution of abundance between proteins is also subtly different. We observe two co-dominant VESA1B proteins (BBBOND 0303890 and BBBOND 0305310) and 17 other copies expressed at a low level (i.e. between 0.08 and 17.7% of the abundance of the dominant protein). This observation could be consistent with either the biallelic expression of a mixed VESA1B homodimer (we did not detect expression of VESA1A) or monoallelic expression of VESA1b monomers in the mixed population. The results also suggest that either multiple VESA2 are expressed simultaneously or there is rapid switching among ves2 genes. B. bigemina PR is not a clonal line, and our data cannot distinguish these possibilities. In contrast to VESA1, the most abundant VESA2 protein (BBBOND 0302650) showed much less conspicuous dominance over other expressed VESA2 (see Figure  5). Hence, VESA2 expression may be less regulated than VESA1, although what we observed constitutes only 23% of the ves2 repertoire in B. bigemina PR. These results confirm that multiple ves2 genes are competent to be expressed and are not simply degenerate ves1 gene copies.
The apparent lack of VESA1A expression is puzzling; it is unlikely to be due to technical insensitivity since we are able to detect a variety of VESA1B and VESA2 peptides at even very low abundance. To check this result, we carried out RNA-seq analysis of ves-like transcripts and this corroborated the proteomic data in showing little evidence for ves1a expression (Supplementary Figure S3). It is possible that, if expressed individually and monoallelically, there may be in vitro selection over many generations against parasites expressing VESA1A, e.g. by adhesiveness towards polystyrene, removing these parasites from the population. This observation deserves further investigation.

Transposition of ves genes is evident from intraspecies variation
It is apparent from the species comparisons above that ves1 and ves2 genes are highly labile in terms of their molecular sequences and their genomic repertoire. To achieve such substantial divergence, we must predict that ves genes are frequently transposed to new genomic positions over ecological or individual timescales. To observe transposition directly we sequenced the genomes of additional strains of B. bovis, B. bigemina and B. divergens from diverse locations (see Table 1) and determined the conservation of ves loci at the intraspecific level. Table 2 describes the considerable variation within species in the presence and absence of ves loci. For example, between 65.3 and 69.4% of BbigVes1a loci and between 78.2 and 80.8% of BbigVes2 loci are present in multiple B. bigemina strains. However, phylogenetic analysis of ves gene repertoires from multiple strains produced no evidence for strain-specific gene family expansions that might suggest structural differentiation of ves repertoires within a species (data not shown). Thus, while ves gene repertoire among divergent B. bigemina strains overlaps structurally and phylogenetically, a sizable minority of ves1 genes are absent from conserved genomic positions in any given instance, indicating that gene transposition is much more frequent than either gene duplication or the evolution of novel sequences through amino acid substitution.
Moreover, even when genes are present at the same locus in multiple strains, these are not always orthologous; indeed, Table 2 shows that the majority of ves1 in B. bovis and B. bigemina are not orthologous when conserved in position. Conversely, 75% of conserved smORF loci contain orthologous sequences, and likewise 79.5-85.1% of BbigVes2 and 86.3% of BdivVes2 are orthologous. Supplementary Figure S4 illustrates this distinction for B. bigemina ves1 and ves2 loci across four strains, contrasting the complete conservation of a ves2 locus with three ves1 loci for which micro-homology has broken down to some extent. While the orthologous sequences at the ves2 and upstream ves1 loci cluster together in the global gene family phylogenies, sequences from the downstream ves1 loci, which are not orthologous, cluster apart.
From the results above, widespread transposition of ves sequences is apparent. To assess its significance we adopted a co-phylogenetic approach, as described in Supplementary Figure S5. Taking pairs of strains in turn, we estimated maximum likelihood phylogenies for ves1 and ves2 genes that occurred at conserved positions, excluding ves genes at positions that were unique to a single strain. If transposition has not occurred, these phylogenies should be identical, or at least display variations not significantly greater than expected by systematic error, because the genes would be orthologous and diversified in the common ances-  Note. When a comparison was not possible due to sequence gaps in the B. bovis and B. bigemina genomes (mostly affecting sub-telomeric regions), this was recorded as 'No assembly'. This was not recorded for B. divergens, for which most ves-like genes are located on unscaffolded contigs, and percentage values for this species refer to only those loci that were confirmed in the same genomic context in both 1802A and Rouen1987 genome sequences.
tor of the two strains concerned. If strain phylogenies are significantly different, then this means that, despite being found at conserved positions, gene sequences have changed since the strains originated, which we interpret as evidence for transposition through gene conversion. It is likely that the individual genes had at one time been the active locus for ves1α/a and/or ves1β/b transcription (LAT). While the LAT, the genes would have undergone repetitive events of segmental gene conversion with formation of extensive mosaics, as previously described (22). Congruence between strain phylogenies was assessed for each ves-like subfamily using permutation tests executed in the programme Jane 4 (69) and judged to be significant if P < 0.01. For each panel in Figure 6, the observed 'cost' of reconciling incongruence between strain phylogenies is marked by a vertical red line; adjacent to this is a frequency distribution showing the costs of reconciliation when the phylogenies are randomized. The distance between the observed and randomized costs is a reflection of the congruence in the trees, where they overlap congruence is likely to be significant. Contrary to the null expectation, Figure 6 shows that ves1 phylogenies were incongruent, (i.e. shared a level of topological similarity no greater than that due to chance), for B. bovis (BbovVes1a, P = 0.315; BbovVes1b, P = 0.264), B. bigemina (BbigVes1a, P = 0.055; BbigVes1b, P = 0.014) and B. divergens (P = 0.1). In contrast, strain phylogenies for BbigVes2, BdivVes2a and BdivVes2b were significantly congruent (P < 0.001). Thus, when ves1 genes are present at the same genomic position in different strains they almost invariably have different sequences, whereas when ves2 genes are conserved in position, they are also often conserved in sequence (i.e. are orthologous). In the absence of recombination after the strains diverge, such trees should have the same topology. Significance of topological congruence is assessed through phylogenetic reconciliation using the programme Jane 4, whereby evolutionary events are posited to explain topological disparities between the trees. Each histogram shows the frequency distribution of event costs for 100 randomized trees generated by permuting the reference strain phylogeny, compared to the observed event cost (vertical dashed line). Where observed and randomized event costs overlap, this indicates that there is no significant agreement between the trees, which we interpret as evidence for recombination. P-values represent the probability of obtaining the observed cost in randomized co-phylogenies (i.e. of observed tree similarity being due to chance), and are mean averages taken over all cost combinations.

ves1 and ves2 genes differ in their exposure to recombination
The transposition of ves1 and, to a lesser extent, ves2 sequences between loci is consistent with the idea that ectopic gene conversion of ves sequences is frequent and important in driving antigenic variation (22). Sequence mosaicism has been observed in BbovVes1 genes and is interpreted as evidence for historical recombination (22). We sought to compare the role of recombination between ves1 in different species, and between ves1 and ves2 in the same genomes, by estimating the frequency of sequence mosaicism using phylogenetic methods. Evidence for recombination was identified in the form of phylogenetic incompatibility (PI) among characters, when sampling alignments of full-length, veslike sequences. Two programs, PhiPack (70) and 3seq (71), were used to assess the significance of PI in multiple alignments sampled from all full-length sequences (see methods). The metric P pi records the proportion of alignments that returned significant PI (i.e. P < 0.05) for any given subfamily. As Figure 7 shows, ves1 genes in all three species displayed significant PI. Using 3seq, P pi tended to be in the 0.1-0.25 range, although BdivVes1b sequences displayed PI much more frequently (P pi = 0.42). Using PhiPack, P pi was between 0.88-0.99 for the same gene families. The difference between methods likely results from the use of four sequences in PhiPack analyses (one more than in 3seq analyses), which increases the chances of observing PI. The increased P pi in BdivVes1b was not observed when using PhiPack.
A clear difference in P pi is observed between ves1 and ves2 subfamilies using both programs. In co-phylogenetic analyses, BbigVes2 and BdivVes2b showed little evidence of transposition or in situ gene conversion. This is corroborated in Figure 7, where P pi is 0.001 and 0.033 respectively using 3seq. SmORF and BdivVes2a co-phylogenies were more equivocal but these two subfamilies also offer less evidence for recombination than among ves1 sequences (P pi = 0.007 and 0.128 respectively). When we compare the actual P-values associated with 3seq analyses of all sequence triplets (Figure 8), these are significantly lower for each ves2/smORF subfamily relative to ves1 genes in the same genome. The exception here is BdivVes2a, which has a comparable P pi to BdivVes1 but a lower mean P-value. In summary, those sub-families showing greater congruence in co-phylogeny analyses tend to show significantly less PI in analysis of sequence mosaics, and so ves1 and ves2 subfamilies display a markedly different evolutionary dynamic that is consistent across species. Although the function(s) of these proteins are not yet known, they are unlikely to be exposed on the infected erythrocyte surface. Immune selection pressure may have been greatly reduced following their derivation from ves1 genes as a result, and they therefore show greater co-phylogeny.
The position of recombination breakpoints within veslike sequence alignments could indicate low-complexity structural motifs that facilitate recombination by promoting strand annealing between homologous sequences (32). The most likely 5 and 3 breakpoints for recombinant tracts detected by 3seq can be plotted onto the sequence alignments for each ves-like gene family ( Supplementary Figure S6). While there is no consistent pattern between families and sometimes no preference is shown at all, it appears that some breakpoints may be associated with repetitive regions that are unique to each ves gene family. For example, in the case of BbovVes1α, the 5 breakpoint between 800 and 900 bp is immediately preceded by sequences encoding an alanine-rich repeat region and another domain containing cysteine-based repeats. The 3 breakpoint between 2700 and 2800 bp is immediately preceded by sequences encoding the largest glycine/lysine-rich repeat region in the molecule. Similarly, BbigVes1a has a dominant breakpoint between 1250 and 1500 bp that immediately precedes sequences encoding a low-complexity region dominated by serine, threonine and proline residues. However, such regions are common in these molecules and so these associations will require further validation.

DISCUSSION
In common with other pathogens, the most labile features of Babesia genomes are those genes encoding surface antigens. Ves1 genes, and the shortened forms derived from them, display the greatest divergence in species comparisons and the most polymorphism in strain comparisons. These genes are also found consistently in genomic regions otherwise depleted of coding sequences and prone to rearrangement. At each level of inspection, our analyses have emphasized that ves1-like sequences are transposed between Figure 8. Comparison of P-values in tests for phylogenetic incompatibility (using 3seq). Mean P-value is shown for each ves1 and ves2 gene family ± one standard deviation. The values are converted to a negative natural log scale for ease of comparison. Gene families are shaded as previously. The results of pair-wise t-tests between all P-values returned in each analysis are indicated (n.s. = not significant). Note that most P values for BdivVes1b tests were zero, although a minority returned non-zero (but highly significant) log values. This circumstance is responsible for the apparently large standard deviation around this mean. In fact, all BdivVes1b sequence triplets returned highly significant tests for phylogenetic incompatibility. *** P < 0.001, * P < 0.05. locations: the variation in presence or absence of individual loci, the non-orthology of sequences occupying homologous positions in different species and strains, the incongruence of ves1 co-phylogenies, and the abundant evidence for recombination between ves1 sequences. In each respect, ves2 and smORF sequences are different; they are more consistently conserved in position, their co-phylogenies have greater congruence and their sequence alignments present significantly less evidence for sequence mosaicism.
Therefore, there is a fundamental difference in evolutionary dynamic between the canonical ves1, the evolution of which is perhaps determined by its particular requirement for transposition, and ves2, which evolves in a manner more consistent with a basic birth-death model, albeit with substantially more gene duplication than most gene families.
The genomic basis for immune evasion in B. bovis (19,20,22) is, in several respects, analogous to the expression of PfEMP1 from var genes in the malaria parasite Plasmodium falciparum (2), to the Variant Surface Glycoprotein (vsg) in the African trypanosome Trypanosoma brucei (75); and to the Major Surface Glycoproteins (msg) in the ascomycete Pneumocystis carinii (76). The convergent evolution of immune evasion functions among these proteins and their regulatory milieux in diverse organisms has been recognized (77)(78)(79)(80) and perhaps this reflects their shared experience of a relatively conservative vertebrate immune response. Typically, an abundant gene family encoding a repetitive major surface protein is sited in labile regions of the genome; we have seen how ves-like genes are typically distributed at the most dynamic sites in the Babesia genome. Expression of the surface antigen is commonly restricted to a specific locus, while most or all other gene copies are silenced. Available evidence suggests this is the case for ves1 genes in B. bovis (25) and perhaps ves genes among Babesia spp. in general (this study). Periodic switching of a new vari-ant antigen gene, in whole or in part, into the expression site from silent loci, or epigenetic transcriptional switching among members of the multigene family, then results in antigenic variation on the parasite (or IRBC) surface.
The common role of gene conversion in antigenic variability across diverse organisms has also been highlighted (32,34,81). By detecting the signature of past recombination in multiple sequence alignments, recombination has been identified as a principal evolutionary force affecting gene family structure of var genes in P. falciparum (82)(83)(84), vsg in T. brucei (30,31,85) and msg in P. carinii (86). However, recombination of var genes appears to be meitoic, and gene conversion is not a demonstrated mechanism during antigenic variation within an individual host. Repetition within genes or in non-coding flanking regions is thought to facilitate recombination, suggesting that genome structure is adapted for increasing the mutability of variant antigen genes (4,87). Analogous motifs may exist in Babesia; the well-conserved C-terminal domains of ves1 genes could plausibly act as annealing points during cross-over, while our recombination analyses have located breakpoints to repetitive parts of ves1 genes. If so, they are unique to a given species, since they fall within regions that are not widely conserved among species.
What is recognized less is that these contingency gene families have similar phylogenetic patterns, i.e. variant antigen gene repertoires in related species do not overlap. There is no orthology between the ves1 in different Babesia spp. (Figure 4), just as there are no orthologous vsg in different African trypanosome genomes (31,88), orthologous var in comparisons of P. falciparum and P. reichenowi (35), or indeed among P. falciparum strains (82). The mutual exclusivity of repertoires in these situations indicates rapid gene turnover; the frequent gain and loss of gene copies after speciation, resulting in the substitution of shared, ancestral characters by unique, derived ones. Hence, the rapid loss of phylogenetic diversity is a sampling effect; after speciation, species-specific sequence types are constantly created while ancestral types are a limited number, but all can be 'overwritten' by gene conversion. This likely explains the lack of cysteine-rich CKRD and VDCS domains encoded by B. bovis ves genes in those of B. bigemina and B. divergens. Moreover, a lack of these domains may explain the lack of cytoadhesive behaviour in B. bigemina and B. divergens, whereas B. bovis uses VESA1 as a highly variable cytoadhesion ligand (89). In addition to immune evasion, antigenic variation of VESA1 could provide for stochastic sampling of a very large molecular space on the endothelial cell surface for a complementary receptor.
Another feature of the rapid turnover of ves-like structures after speciation is the evolution of ves2 from the Nterminal regions of ves1, which we believe has occurred on independent occasions. Although the functions of ves2 and smORF are open questions, there are good reasons for believing that they are distinct from ves1. First, we have shown that ves1 and ves2/smORF behave differently over evolutionary timescales. Second, they have radically different protein structures, and whereas VESA1 is membranebound in B. bovis (42) and we presume in B. bigemina and B. divergens, VESA2 and SMORF are strongly predicted to be secreted. Third, the two sequence classes have different expression profiles in B. bigemina PR, suggesting the employment of unlinked and perhaps different regulatory strategies.
We can speculate on ves2 function through comparison with related Apicomplexan parasites. In Plasmodium spp., secretion is thought to be instrumental in remodeling host erythrocytes after the parasite has invaded. Altering erythrocyte structure is necessary for parasite metabolism and key survival strategies such as sequestration and antigenic variation (90)(91)(92)(93). The secreted proteomes of Plasmodium spp. contain not only key effectors, for example, the PfEMP-1 protein in P. falciparum, but also diverse chaperones and cofactors thought to facilitate the expression of key effectors (94,95). So, we can imagine VESA2 being secreted into the cytoplasm of the IRBC as part of cocktail of parasite proteins that alter host cell structure. This might be in support of VESA1 expression in the IRBC plasma membrane directly, or to antagonize host proteins with the effect of making the IRBC more amenable to parasite survival, as happens when Theileria spp., a sister genus to Babesia, infect white blood cells (96)(97). Alternatively, or perhaps additionally, since VESA2 proteins are expressed abundantly, they may serve to obfuscate the immune response. This could occur by creating a 'smokescreen' effect, inducing focus of dendritic cells upon antigens that are inaccessible and largely insensitive targets, analogous to the secretion of subtelomere-associated variant surface proteins (SVSP) into the host cytoplasm by Theileria spp. It has been suggested that presentation of SVSP peptides by class 1 MHC distracts the immune response from exposed antigens on the parasite surface (97) and, similarly, VESA2 may promote antibodies that have no reactivity to VESA1 (although without presentation of MHC, since Babesia spp. only infect erythrocytes). It remains to be seen if VESA2 are im-munogenic, and if antibodies to VESA1 and VESA2 are cross-reactive.
In this study, we have confirmed that the genomic basis for antigenic variation in B. bovis is conserved throughout the genus, but with substantial, species-specific divergence in protein structure and gene repertoire. In fact, while Babesia genomes are broadly conserved in other respects, the evolution of ves1 genes is a record of constant change and rapid turnover, which we suggest reflects the enduring challenge of host-parasite interactions over millions of years. In addition, ves-like genes with shortened predicted proteins have evolved from canonical ves1 on independent occasions. These have distinct expression profiles (in B. bigemina at least) and evolutionary dynamics, suggesting a distinct function, though probably still involved in antigenic variation, or at the host-parasite interface. From previous observations of the B. bovis genome, it is thought that recombination between ves1 gene copies could be crucial both to the mechanism of antigenic variation during infection, and to the generation of antigenic diversity in parasite populations (22,26). These results confirm the principal role of recombination across the genus and support the emerging view that genomic architecture facilitates recombination to both promote switching and generates antigenic diversity (98). As with var and vsg, the roles of recombination, rapid gene turnover and structural innovation in the evolution of variant antigens is affirmed in ves1 phylogeny. The association of this phylogenetic pattern with variant antigen function in diverse parasite genomes is testament to the convergence in both structure and mechanism by diverse pathogens to combat vertebrate immunity.