Ethics statement

The rearing of animals was regulated by Czech legislation (Act No 246/1992 Coll.). All housing, feeding and experimental procedures were conducted under protocol 90/2013 approved by Biology Centre, Czech Academy of Sciences and Central Commission for Animal Welfare of the Czech Republic.

Whole genome sequencing

Trypanosomes (T. evansi strain STIB805) were purified from mice by DEAE (DE52) cellulose [23]. Total DNA was extracted as described elsewhere [24]. Briefly, the cells were lysed using SDS, and incubated with proteinase K and RNase. DNA was harvested after phenol extraction and ethanol precipitation. Four runs of single-end 454 sequencing plus 2 runs of paired-end 454 sequencing were obtained using GS FLX(+) System following the manufacturer's instruction (Roche) and generated 1,904,327 reads (225,826 paired end, 1,678,501 single end) [25]. Approximately 10 µg of genomic DNA was sheared by nebulization into desired fragments sizes (∼400 bp for single-end 454, ∼3 kb for paired-end) and adaptor oligos ligated to create the library for sequencing. Additional sequence data was obtained by shearing genomic DNA to ∼200–300 bp fragments sizes for sequencing on an Illumina GAIIx producing 19,701,740 tags with an ordered read length of 76mers. Illumina reads for T. b. brucei strains TREU 927/4 and Lister 427 (provided by the Wellcome Trust Sanger Institute, Hinxton, UK) were downloaded from the European Nucleotide Archive (accession nos. ERX009953 and ERX008998).

Sequence assembly and SNP calling

Genome assemblies and identification of sequence polymorphisms (SNPs and indels) were carried out with CLC Genomics Workbench (CLC bio). Reads for T. evansi STIB805, T. b. brucei TREU 927/4 and T. b. brucei Lister 427 were mapped against the T. brucei TREU 927/4 version 4 reference (Tb927) using the following mapping parameters: global alignment, similarity fraction  = 0.9, length fraction  = 0.5, insertion cost  = 3, deletion cost  = 3, mismatch cost  = 2. De novo assemblies for each strain were created using either all reads or those binned during the reference-based assemblies, using the following parameters: automatic word size  =  yes, bubble size  = 50, similarity fraction  = 0.9, length fraction  = 0.5, deletion cost  = 3, insertion cost  = 3, mismatch cost  = 2. De novo contigs were aligned to Tb927 using standalone NCBI BLAST version 2.2.25 and Artemis Comparison Tool release 12.0 [26]. For RPKM (reads per kilo base per million) analysis, aligned STIB805 and TREU 927/4 reads were assigned to sequential 1-kb bins along the length of the Tb927 reference. For each chromosome, the log2 ratio of binned reads from the two read-sets was calculated for each bin, and normalized to a median log2 ratio of 0 by offsetting all values by the median log2 ratio. SNPs and indels were called with CLC bio using the following parameters: minimum coverage  = 10, maximum coverage  = 100, minimum variant frequency  = 30%, minimum central quality  = 20, minimum average quality  = 15.

Gene prediction and annotation

Coding sequence prediction of T. evansi genome was done using a combination of de novo gene prediction approach and reference based gene transfers. The Rapid Annotation Transfer Tool (RATT) was used to transfer the gene boundaries and functional annotations from Tb927 onto T. evansi chromosomes (target hereafter) [27]. To prevent the transfer of paralogs to the same target region, thus resulting in multiple overlapping/duplicate gene calls, the annotation transfer was performed pairwise between one chromosome from Tb927 and the corresponding chromosome from the target. This resulted in the transfer of more than 96.4% (9427/9776) of genes from Tb927_v4 onto TevSTIB805ra. Though we observed extensive conservation of synteny between Tb927 and target genomes, this approach would likely have missed genes in targets that were shuffled by chromosomal fission and fusion that occurred during evolution. We performed another RATT transfer considering all Tb927 chromosomes and one target chromosome at a time. We then parsed out genes that were predicted uniquely by this approach (i.e. from regions where the initial RATT transfer failed to predict any genes). Combining these two approaches allowed us to predict genes that were shuffled across the chromosomes while avoiding predictions that overlap each other. We then used an in-house consensual de novo gene prediction suite called 'AutoMagi' to predict protein coding genes from T. evansi [28]. AutoMagi internally uses three gene prediction algorithms (genescan, testcode, codonusage) and predicts a consensus gene model out of individual gene predictions. The codon usage table required by AutoMagi was generated by ‘cusp’ using the RATT output of T. evansi STIB805 assembly [29]. We then used an in-house built prolog system to combine the de novo gene predictions and reference-based gene predictions. The genes that are unique to RATT predictions were automatically included in the final set. Genes that are unique to AutoMagi were compared with the NCBI non-redundant (NR) database (accessed 23 February 2011) using 'BLASTp' (default criteria: EXPECT = −10, WORD SIZE = −3, MATRIX_NAME =  BLOSUM62, GAP COST =  Existence:11 Extension:1) algorithm. All of the 'BLASTp' results were reviewed manually, and genes meeting the following two criteria were retained: (a) an E-value of 5e-6 or lower to the matched NR sequence; (b) coding strand identical to nearest neighbors on either side (i.e. on same poly-cistronic unit). This manual curation removed 165 putative CDS from making into the finalized TevSTIB8805ra.

The gene calls that overlapped between RATT and AutoMagi were subdivided into the following 4 groups. 1) Identical: overlapping exactly, 2) AM_subsetof_RATT: AutoMagi prediction is entirely contained within RATT prediction, 3) RATT_subsetof_AM: RATT prediction is entirely contained within AutoMagi prediction 4) StaggeredOverlap: both predictions overlap in a staggered fashion. In the first two cases (Identical & AM_subsetof_RATT), AutoMagi gene calls were ignored and RATT models were retained. In the third case (RATT_subsetof_AM) coordinates from AutoMagi were combined with annotation information from the RATT model. Genes in the fourth category (Staggered) were subjected to a thorough manual review process. The review process involved (i) 'BLASTp' search against NCBI's NR database to identify coordinates for queries and subject; (ii) a ClustalW multiple sequence alignment of both candidate genes with their potential homolog(s) from T. brucei. Manual review of BLAST and ClustalW was performed in each case to decide either to split/merge/choose one of the AutoMagi/RATT predictions. These newly derived coordinates were then combined with the annotation information from the RATT model. GeneIDs of the final set of protein coding genes were unified and ordered from left to right end of the chromosome.

Genome publication

Entire genome and all the predicted genes are publicly available at TriTrypDB (http://tritrypdb.org/tritrypdb/). The fastq files containing the Illumina read data for T. evansi STIB805 are available at: http://www.ebi.ac.uk/ena/data/view/ERA000101.

Pulse-field gel electrophoresis

Bloodstream parasites at 2×10⁸ cells/ml were purified using DEAE cellulose (DE52) chromatography, and subsequently used to prepare chromosome blocks as previous described [30]. DNA from T. b. evansi and T. b. brucei cells was embedded in low-melt agarose blocks (final concentration of 5×10⁷ cells/ml) according to [31], and was resolved using a 1% Megabase agarose (Bio-Rad) gel with 0.5X TBE buffer in the CHEF-DRIII system (Bio-Rad). S. cerevisiae DNA was used as a size marker. Pulse field gel electrophoresis (PFGE) was run at 14°C under the following conditions: switch time A increased from 28.6 s to 228 s for 24 hrs, followed by switch time B, with increase from 28.6 s to 1,000 s for another 24 hrs. The angle was set to 120° and voltage gradient to 3 V/cm. The PFGE gel was stained with ethidium bromide after the run.

VSG sequence retrieval and alignment

VSG sequences were extracted from the genome assembly using hidden markov models (HMM) and HMMER 3.0 [32]. HMMs were constructed for a- and b-type VSG respectively using multiple sequence alignments of T. brucei TREU 927/4 and T. b. gambiense DAL972 sequences [33]. All open reading frames>100 bp were marked up and the predicted amino acid sequences were searched for matches to either HMM using HMMER 3.0. Significant matches were checked manually to ensure that each VSG was complete and gene boundaries were correct. Intact VSG were extracted and aligned approximately using Clustal X [34]. The aligned sequences were combined with existing alignments of T. brucei TREU 927/4 a- and b-type VSG [35] and modified by eye.

VSG phylogenetic analysis

C-terminal domains were trimmed (due to recombination these present an inconsistent phylogenetic signal), resulting in a- and b-VSG alignments of 470 and 492 characters, respectively. Neighbour-joining trees were estimated for each amino acid sequence alignment using PHYLIP v3.6, with a JTT rate matrix and 100 non-parametric bootstrap replicates. A frequency distribution of species-specific clade sizes, (i.e. three T. evansi VSG clustered together with a T. brucei TREU 927/4 sequence as the sister lineage has a clade size of 3) was calculated to express the degree of intercalation of sequences from the two strains.

VSG ortholog divergence

Putative VSG orthologs were extracted from the phylogenies. In situations where single VSG from T. evansi STIB805 and T. brucei TREU 927/4 were sister taxa, supported by a bootstrap value>95, these genes were interpreted as orthologs. The rates of synonymous and non-synonymous nucleotide substitutions per site were estimated for these orthologous pairs. The ratio of these rates (ω) was estimated using Ka/Ks Calculator v1.2 [36] using GY and MS methods. For comparison, this was repeated for 151 pairs of non-VSG orthologs chosen at random.

Phylogenetic analyses

The dihydrolipoamide dehydrogenase (LipDH) CDS (Tb927.11.16730) was PCR-amplified from total parasite DNA using primers 5′-ATA AAG CTT ATG TTC CGT CGC TGC-3′ (forward) and 5′-ATA AGA TCT TTA GAA GTT GAT TGT TTT GG-3′ (reverse) and Phusion polymerase (New England Biolabs, NEB). In cases where direct sequencing of the amplicon revealed heterozygosity, sequence information for individual alleles was obtained after cloning. After removal of the primer sequences, LipDH CDSs were aligned using ClustalX [37]. Phylogenies were reconstructed using the program MrBayes [38] via the Topali platform [39], implementing the GTR substitution model and a discrete gamma rate distribution model with four rate categories (to account for rate heterogeneity among sites) as the most appropriate nucleotide substitution model. Four independent MCMC chains of 1×10⁶ generations were sampled every 100^th generation. A 50% majority rule consensus tree was derived after the first 25% of trees were discarded as burn-in. The ATP synthase γ subunit sequence (Tb927.10.180) was PCR-amplified from total parasite DNA with primers 5′-GCG GAA TTC GAA GCA GAT GAC ACC TAA-3′ (forward) and 5′-GCG GAA GAC CTT GCT GCG GAG CCA CTC T-3′ or 5′-GGC GAC ATT CAA CTT CAT-3′ (reverse) and Phusion polymerase (NEB). The sequence was determined by direct amplicon sequencing or, in cases of heterozygosity, after cloning. A partial 812-bp sequence of cytochrome oxidase 1 (COX1) was obtained by PCR and used for phylogenetic analysis as previously described [40]. Briefly, we assessed phylogenetic relationships among T. equiperdum and T. brucei isolates using a haplotype network constructed using the statistical parsimony approach implemented in TCS v. 1.21 [41]. Subnetworks were created with 95% confidence limit and then unconnected subnetworks>10 mutations apart were connected by relaxing the confidence limit. To verify that the haplotypes containing T. equiperdum isolates were on phylogenetically distinct branches, we estimated a phylogenetic tree using the Bayesian approach implemented in MrBayes v. 3.2 [42]. PartitionFinder v. 1.0.1 [43] determined the Hasegawa, Kishino and Yano nucleotide substitution [44] with invariant sites (HKY+I) without partitioning by codon positions as the most appropriate model for the MrBayes analysis. Microsatellite genotyping was carried out exactly as described previously [40]. Briefly, isolates were typed for eight microsatellite markers [45]. Principal component analysis (PCA) was performed in R using the package adegenet, as described [40].

Pulse-field gel electrophoresis (PFGE)

PFGE was performed to visualize the pattern of chromosomes in the akinetoplastic T. evansi STIB805 strain. Multiple bands corresponding to megabase chromosomes ranging in size from ∼1 to ∼5 Mb could be visualized (Fig. 1). The most noticeable differences in the chromosomal pattern in comparison to T. brucei are the five intermediate chromosome bands that range from ∼300 kb to ∼800 kb. Additionally, two size groups of minichromosomes (∼100 kb and ∼200 kb) were observed in T. evansi STIB805. Such a degree of variability is within the range observed among strains of T. b. brucei [46].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Pulse-field gel electrophoresis comparing chromosomes of T. evansi STIB805 (Tev) and T. b. brucei TREU 927/4 (927).

While the sizes of megabase chromosomes are largely similar, differences in the intermediate and minichromosomes (825 kbp and smaller) are evident between T. evansi and T. brucei. Although T. brucei chromosomes I-XI were not unambiguously identified, labels to the left of the gel indicate bands consistent with the expected T. brucei chromosome sizes [46], as well as intermediate and minichromosomes. The signal at the top of the gel is from the well, as indicated.

https://doi.org/10.1371/journal.pntd.0003404.g001

T. evansi reference-based genome

Genomic DNA isolated from T. evansi STIB805 was subjected to both 454 and Illumina sequencing, generating a combined total of 21,445,221 reads. Using the T. brucei TREU 927/4 genome sequence [22] (version 4; from here onwards abbreviated Tb927 for convenience) as a scaffold, a reference-based assembly of these T. evansi reads was generated. This assembly, called TevSTIB805ra, incorporates 17,856,165 reads and has an average coverage of 57.2 reads per nucleotide across the entire assembly and of 48.4 reads per nucleotide for the ‘core regions’ [i.e. regions excluding telomeres, subtelomeres, and internal regions that consist of repetitive coding sequences such as VSGs, expression site-associated genes (ESAGs) or retroposon hot spot genes (RHS)].

An assembly approach based on the genome of a closely related reference strain allows the reliable identification of homologous gene pairs, of any differences that might exist between these sequences, and of genes that are missing or highly diverged in the genome of interest. This approach has limited power in identifying structural differences between genomes and, by definition, cannot identify sequences that are present in the genome of interest but not in the reference genome. For that reason we have supplemented the reference-based approach with the analysis of contigs that were assemble de novo. As detailed below, this allowed the confirmation of Tb927 genes that are absent in in T. evansi STIB805 and the identification of candidate genes that might be present in the latter but absent in the former. However, genome rearrangements that are not associated with differences in gene content will have been missed and the T. evansi STIB805 genome as published on TriTrypDB may indicate gene synteny where it does not exist.

Annotation of the TevSTIB805ra T. evansi genome was performed using a combination of RATT and AutoMagi, followed by manual curation. RATT (Rapid Annotation Transfer Tool; [27]) identified likely orthologs in the T. brucei reference genome and transferred their functional annotations to the T. evansi genome, while AutoMagi predicted genes de novo using the consensus of three gene prediction algorithms (genescan, testcode, codonusage) [28]. A total of 10,110 CDS were identified, with 9368 CDS annotated as T. brucei orthologs by RATT and subsequent manual inspection (Table 1, columns C, E and F; S1 Table), and 742 CDS uniquely predicted by AutoMagi (Table 1, column D; S2 Table). Thus, 92.7% of the identified T. evansi CDS have an identified ortholog in T. brucei, while 7.3% were uniquely detected by de novo gene prediction. Analysis of the 742 AutoMagi CDS by BLAST searching of the GenBank non-redundant database revealed at least one hit (E-value ≤5e-6) for each of these CDS to a gene from a Trypanosoma species (T. b. gambiense 568 CDS; T. b. brucei 168 CDS; T. equiperdum, T. vivax, or T. congolense 6 CDS). The most common annotations among these BLAST hits were for ‘hypothetical unlikely’ (415 CDS) or other hypothetical (150 CDS) sequences (S2 Table). Orthologous sequences had not been annotated in the Tb927 reference, presumably due to more stringent criteria for gene calling. Thus, most if not all of the de novo predicted genes in TevSTIB805ra are also present in T. brucei.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Chromosomal breakdown of CDS found in TevSTIB805ra compared to Tb927 reference.

https://doi.org/10.1371/journal.pntd.0003404.t001

Of the 9368 T. evansi CDS identified as T. brucei orthologs, 8421 CDS were classified as non-repetitive genes; the remaining 947 CDS were VSG, ESAG, RHS, or duplicate sequences. A comparison of the 8421 non-repetitive CDS between T. brucei and T. evansi revealed 7970 (94.6%) had a nucleotide identity of 95% or greater, 320 (3.8%) had a nucleotide identity between 70–95%, and 131 (1.6%) had a nucleotide identity less than 70% (Fig. 2; S1 Fig.). After RATT annotation transfer, a total of 503 (5.1%) Tb927 GeneIDs did not have identified orthologs in the TevSTIB805ra annotated genome (Table 1, column G; S2 Table). The majority of these (406) represent repetitive genes (e.g. VSG, ESAG, and RHS) or ‘hypothetical unlikely’ genes, which were not analyzed further. Five CDS correspond to predicted pseudogenes. T. evansi STIB805 orthologs for 49 Tb927 CDS were shown to be wholly or partially missing in TevSTIB805 reads that mapped to these gene loci (see below). T. evansi homologs for the remaining 43 Tb927 GeneIDs were ‘missed’ by RATT and AutoMagi, but were subsequently identified by manual examination of the T. evansi sequence. Thus, the majority of T. brucei CDS have extremely similar T. evansi orthologs, and very few T. brucei CDS were not found in T. evansi. This result is consistent with a very close phylogenetic relationship between these parasites.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Frequency distribution of nucleotide identity between orthologous T. evansi vs. T. brucei coding sequences, including pseudogenes.

This graph represents 8290 CDS with the highest identities out of 8421 total non-repetitive sequences analyzed.

https://doi.org/10.1371/journal.pntd.0003404.g002

The initial RATT approach only identified T. evansi CDS with an annotated Tb927 homolog in the syntenic position. To identify potential CDS in TevSTIB805ra that are syntenic to unannotated sequences in Tb927, and are homologous to annotated CDS in other (non-syntenic) locations, RATT was performed a second time by comparing each TevSTIB805ra chromosome to the entire Tb927 genome. This approach complemented de novo gene calling by AutoMAGI and identified 112 CDS in TevSTIB805ra (Table 1, columns E and F; S2 Table). The majority of these CDS were annotated as hypothetical proteins (64 CDS) or repetitive sequences such as VSG, ESAG, and RHS (39 CDS). For almost all loci, manual inspection revealed that either (i) a syntenic CDS existed in Tb927 that was unannotated or (ii) a syntenic, annotated CDS did exist in Tb927 but the presence of a highly similar sequence elsewhere caused a RATT artefact. An exception was TevSTIB805.6.770, where the syntenic sequence in Tb927 was found to be disrupted by frame-shifts, but in T. b. gambiense DAL972 to be annotated as Tbg972.6.420 (hypothetical protein, unlikely)

Genes identified in T. evansi by de novo assembly

Two approaches were used to find genes in the T. evansi genome that may not be present in the T. brucei genome. Firstly, reads from T. evansi that did not match Tb927 were de novo assembled to make TevSTIB805dn and putative CDS called with AutoMagi (Table 2 and S2 Table). The 22 CDS with a homolog in Tb927 that was not a repetitive or hypothetical gene included multiple copies of pseudogenes for putative UDP-Gal/UDP-GlcNAc-dependent glycosyltransferase and lytic factor resistance-like protein, an adenosine transporter, calmodulin, a DNA topoisomerase, a leucine-rich repeat protein, ribulose-phosphate 3-epimerase, major surface protease B, and a galactokinase pseudogene. 36 of the 202 CDS that did not get a hit against Tb927 had a BLASTx hit in the NCBI non-redundant database: 23 CDS matched hypothetical Trypanosoma spp. and 9 appeared to be contaminants from other - mostly bacterial - organisms. The remaining 4 BLAST hits were for T. vivax RNA-dependent DNA polymerase and T. evansi VSGs (2 hits) and diminazene-resistance-associated protein.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Genes identified in T. evansi de novo assemblies.

https://doi.org/10.1371/journal.pntd.0003404.t002

Because Tb927 was created using a traditional sequencing approach and only covers the 11 large chromosomes [22], sequences found in the de novo assembly of T. evansi deep sequencing reads might not be unique to this strain or species, but rather reflect a difference in sequencing methodology or stem from intermediate-sized or mini-chromosomes. To address this, Illumina reads from sequencing T. brucei TREU 927/4 that also did not match Tb927 were de novo assembled into Tb927dn. Comparison of Tb927dn to TevSTIB805dn showed that 45.6% of the binned T. evansi reads matched to Tb927dn. The remaining T. evansi reads were then assembled into TevSTIB805dn_sub and putative CDS called with AutoMagi (Table 2 and S2 Table). Of the 84 BLASTp matches to non-repetitive genes, 71 are annotated as hypothetical proteins, 2 are lytic factor resistance-like proteins, 1 is a putative transporter, and 10 are annotated as putative (or pseudogene) UDP-Gal or UDP-GlcNAc-dependent glycosyltransferase. Glycosyltransferases in T. brucei are frequently found in subtelomeric regions that are difficult to assemble due to repetitive sequences, and are known to be highly variable among trypanosome strains [33], [47]. Of the 117 TevSTIB805dn_sub CDS for which no match was identified in Tb927, 18 had a BLASTx hit in the NCBI non-redundant database: 9 CDS matched hypothetical genes from various Trypanosoma species and 1 matched a diminazene resistance-associated protein that was previously suggested to convey diminazene aceturate (Berenil) resistance to certain strains of T. evansi [48]. Thus, gene prediction found very few CDS in de novo T. evansi assemblies that are not present in T. brucei, and analysis of these CDS revealed differences reminiscent of those observed among strains of the same species.

When T. evansi STIB805 de novo contigs were filtered for minimum coverage of at least 5x, the number of genes without obvious homologs in Tb927 was reduced considerably (Table 2). To test the potential absence of these genes in T. b. brucei more rigorously, short read data sets for strains TREU 927/4 and Lister 427 were searched for matches to the 22 candidates from TevSTIB805dn and the 6 candidates from TevSTIB805dn_sub. Only seven CDS candidates remained where the sequence was either entirely absent from both T. b. brucei datasets, or did not contain an undisrupted ORF (S2 Table, highlighted). Whether these candidates are indeed functional CDS, and whether they are generally absent from T. brucei ssp. and present in T. evansi, and therefore are of potentially diagnostic value, requires further investigation. A non-redundant list of the ORFs that did not have a BLASTx hit in NCBI (200 ORFs in total) is provided as S1 Data File.

CDS identified in T. brucei that are absent in T. evansi

A total of 49 CDS found in Tb927 were not identified in TevSTIB805ra. These loci were analyzed individually by (i) specifically searching for matching T. evansi STIB805 reads (S2 Table); (ii) comparing reads per kilobase per million (RPKM) coverage plots of these regions for T. brucei and T. evansi; (iii) aligning contigs of a full T. evansi STIB805 de novo assembly to the respective regions in Tb927. These analyses confirmed the absence or disruption of several of these CDS in T. evansi STIB805 (including the iron/ascorbate oxidoreductase loci, the Tb927.9.7950/Tb927.9.7960 repeat, the Tb927.4.3200-.3270 region, both Tb927.8.490 and Tb927.8.500, and the Tb927.8.7300-.7330 region), as illustrated in S2–S9 Fig., with the procyclin loci described in detail below. These cases have in common that the loci in question show considerable variation among Trypanozoon strains and CDS appear to be absent in T. b. brucei Lister 427 and/or T. b. gambiense DAL972 as well [33]. The differences observed in T. evansi STIB805 compared to Tb927 therefore are unlikely to be relevant for kDNA loss or tsetse-independent transmission.

Procyclin loci

In T. evansi STIB805, the procyclin loci either lack or have disrupted versions of several CDS found in T. brucei. GPEET and EP procyclins and associated genes are encoded in loci on chromosomes 6 and 10, respectively, in various T. brucei strains [49]. Procyclin proteins are GPI-anchored coat glycoproteins that are expressed exclusively in the procyclic insect form of T. brucei, and they have been hypothesized to be involved in protection against tsetse fly midgut hydrolases [50]. Experiments in T. brucei have shown that knocking out all of the procyclin genes (Null mutants of GPEET and EP3 on chromosome 6; EP1 and EP2 on chromosome 10) causes no growth defect in vitro and permits completion of the entire life cycle, but causes a selective disadvantage during co-infection with wild type cells in the tsetse fly midgut [51]. These procyclin loci also contain procyclin-associated genes and a gene related to expression site associated gene 2 (PAG3 and GRESAG2 on chromosome 6; PAG1, 2, 2*, 4, and 5 on chromosome 10) [49]. The functions of the PAG proteins and GRESAG2 are unknown; although transcripts of PAG1–3 have been shown to increase during differentiation to procyclic forms, published experiments using cell lines with all PAG genes knocked out reported no obvious abnormal phenotypes in vitro or in vivo [49]. In multiple T. brucei strains, the chromosome 10 procyclin loci are heterozygous, with one chromosome containing EP1/EP2/PAG1/PAG5/PAG2*/PAG4 and the other chromosome containing EP1/EP2/PAG2/PAG4 (PAG2 being a fusion of the 5′ part of PAG1 and the 3′ part of PAG2*) [49], [52]–[56]. In T. evansi STIB805, chromosome 10 appears to be homozygous, with only the EP1/EP2/PAG2/PAG4 locus present, and the associated absence of the segment containing the 3′ part of PAG1, PAG5 and the 5′ part of PAG2* (Fig. 3). The full STIB805 de novo assembly contained a single 14.9 kb contig corresponding to the EP1/EP2/PAG2/PAG4 locus (S2 Fig.). Also on chromosome 10, the EP2 in T. evansi contains a stretch of 12 divergent amino acids in the domain N-terminal to the EP repeat; this region is highly conserved in T. brucei [57], [58]. Although the function of this domain is unknown, these 12 amino acids are found in all sequenced T. brucei genomes with very few variations. The chromosome 6 procyclin locus contains a triplication of three genes (EP3/PAG3/GRESAG2) in T. brucei TREU 927/4, with GPEET present only in front of the last unit (Fig. 3). Copy numbers of these genes may vary among T. brucei strains, as Southern analysis showed that these are single copy genes in the AnTat1.1 strain [49]. In TevSTIB805ra, coverage of this locus is much reduced (Fig. 3, S3 Fig.), indicating either divergence, reduced copy number, or both. The locus did not assemble into a single contig in the de novo approach and because of the repeat nature of this locus in Tb927, assigning sequence differences to particular gene copies was not possible. Nonetheless, several non-synonymous mutations in EP3 are evident, and the PAG3 and GRESAG2 genes have frameshifts and deletions. In contrast, analysis of a subset of other T. evansi orthologs to CDS shown to be upregulated in PF relative to BF T. brucei (Tb927.1.2310, Tb927.1.2350, Tb927.1.2560, Tb927.1.580, Tb927.10.10260, Tb927.10.10950, Tb927.10.4570, Tb927.11.16130, Tb927.11.8200, Tb927.4.1800, Tb927.4.1860, Tb927.4.4730, Tb927.5.1710, Tb927.5.2260, Tb927.6.510, Tb927.8.5260, Tb927.4.4730, Tb927.8.8300, Tb927.9.15110, and Tb927.9.8420) [59] detected no notable changes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Schematic of reads from T. b. brucei TREU 927/4 and T. evansi STIB805 mapped on to the procyclin loci of T. brucei.

CDS are denoted by yellow arrows, with Tb927 GeneIDs above and gene names in parentheses. Reads that could be uniquely mapped to the reference are colored blue (intact paired reads), green (single forward reads) or red (single reverse reads). Reads that could be mapped to more than one position in the Tb927 reference were placed randomly and are colored yellow. Red asterisks denote mutated or absent CDS in T. evansi. Both PAG5 and PAG2* are missing in T. evansi, and mutations are present in all copies of EP2, PAG3, and GRESAG2.

https://doi.org/10.1371/journal.pntd.0003404.g003

Analysis of T. brucei mitoproteome orthologs in T. evansi

Any Tb927 gene encoding a product that is exclusive involved in (i) maintenance or expression of kDNA, (ii) life cycle progression, or (iii) insect-stage specific energy metabolism, e.g. oxidative phosphorylation, is redundant in T. evansi. Although many of these genes encode mitochondrial proteins, numerous other mitochondrial activities are expected to remain essential even in an akinetoplastic bloodstream trypanosome (see Discussion). Extensive proteomic analyses have identified 978 mitochondrial proteins in T. brucei [60]–[63]. A search of TevSTIB805ra and TevSTIB805dn identified orthologous gene sequences for all of these proteins (S6 Table). A detailed examination of a subset of genes representing categories i-iii above (with the exception of the F₁ subunits of respiratory complex V, which remain essential in T. evansi [19]) suggests that the vast majority, if not all of these genes remain functional (see S6 Table for details). To obtain further evidence for or against mutational decay in these genes, we calculated Ka, Ks, and Ka/Ks for the following orthologous gene pairs in Tb927 and T. evansi: 208 CDS associated with kDNA expression or function; 942 unambiguous orthologous gene pairs for mitochondrial proteins; and 6331 unambiguous orthologous gene pairs corresponding to non-VSG and non-mitoproteome annotations (Table 3). These orthologous gene pairs were selected using greater stringency (reciprocal best matches in BLASTp searches with a minimum e-value threshold of 1×10⁻⁴) than the analysis shown in Fig. 2, to ensure that paralogous gene pairs were excluded (these would possibly increase Ka and Ks estimates and skew the distribution; S6 Table). We found no evidence of relaxed selection for T. evansi CDS associated with kDNA expression or function or the mitoproteome as a whole. Both Ka and Ks were lower for CDS associated with mitochondrial expression, and Ka/Ks was also lower relative to the genomic background, indicating that purifying selection is generally stronger among mitoproteome CDS, perhaps suggesting a higher proportion of essential genes compared to the control set. Furthermore, we made the same comparison between Tb927 and T. b. gambiense DAL972 (both of which retain functional kDNA) and the values for Ka, Ks and Ka/Ks show no significant differences (by t-test) with the values for T. evansi.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Mean average Ka/Ks estimates for orthologous CDS pairs between Tb927 and either T. evansi STIB805 (Tev) or T. b. gambiense DAL972 (Tbg).

https://doi.org/10.1371/journal.pntd.0003404.t003

VSG repertoire

The predicted protein structures of VSGs in T. evansi STIB805 conform to the canonical structures in T. b. brucei and T. b. gambiense, and include all five recognized N-terminal sub-types (N1-5). In T. evansi STIB805 we identified 525 a-type VSGs (i.e. N-1-3 and 5; S2 Data File) and 505 b-type VSGs (i.e. N4; S3 Data File); of these 453 (86%) and 451 (89%), respectively, are full-length. Given that the assembly of subtelomeric and mini-chromosomal regions is fragmentary, these numbers may underestimate the real number of VSGs, although the total is comparable with both T. b. brucei TREU 927/4 [22] and T. b. gambiense DAL972 sequences [33]. The VSG repertoire is largely conserved between the T. b. brucei TREU 927/4 reference and T. evansi STIB805, as the VSGs are interspersed among each other in neighbour-joining molecular cladograms (S10 Fig.). Another indication of the similarity of the VSG repertoire is that very few clades of strain-specific VSG are larger than 2–3 (Fig. 4a and b). Large clades of VSG from a single genome would suggest divergence of VSG repertoire through gene duplication. Hence, 376 and 384 ( = 85.1%) T. evansi STIB805 and T. b. brucei TREU 927/4 a-VSGs, respectively, are most closely related to an ortholog in the other strain, or are paraphyletic to a clade of such sequences (i.e. strain-specific clade size  = 1). Only 40 T. evansi STIB805 VSGs and 51 T. b. brucei TREU 927/4 VSGs form a clade with a paralog from the same genome (i.e. strain-specific clade size  = 2), suggesting a single gene duplication since the strains separated. The same patterns occur with b-VSGs. Orthology between VSGs does not mean that they are unaffected by recombination, only that enough sequence homology persists for two orthologs to cluster together. Indeed, among 151 putative a-VSG orthologs 33 (21.8%) have dissimilar C-terminal types, while among 112 b-VSG orthologs 32 (28.6%) showed similar evidence for recombination having occurred since these two strains split from their common ancestor. Thus, analysis of the VSG repertoire reveals no evidence for the evolution of specific VSG gene clusters or subfamilies in T. evansi, similar to what had been observed for T. b. gambiense [33].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Diversity of a-type and b-type VSG between T. b. brucei TREU 927/4 and T. evansi STIB805 compared.

Histograms showing a-type VSG (A.) or b-type VSG (B.) distributions of strain-specific clade size in T. b. brucei (black bars) and T. evansi (red bars) as defined by the phylogeny (see S10 Fig.). Frequency distributions of a-type VSG (C.) or b-type VSG (D.) synonymous (Ks) and non-synonymous (Ka) substitution rates per site, and the ω (Ka/Ks) for orthologous pairs of VSG (a-type n = 151; b-type n = 112), as defined by the phylogeny, in relation to values for unambiguous non-VSG orthologous pairs (n = 6331).

https://doi.org/10.1371/journal.pntd.0003404.g004

VSG sequence evolution

This comparison of orthologous VSGs from T. evansi STIB805 and T. b. brucei TREU 927/4 provides an overview of the molecular evolutionary forces incident on the VSG archive. Fig. 4c and d show that substitution rates among orthologous VSGs are typically higher than the background defined by comparison to the 6331 non-VSG orthologous gene pairs previously used in the mitoproteome analysis above. The mean average rate for VSG is significantly greater than other genes (p<0.05) and comparison of Ka/Ks (ω) explains this. The distribution of ω is left-skewed for non-VSG genes, which reflects the strong purifying selection on non-synonymous substitutions that typifies most genes that perform essential functions. The distribution for VSG is normal, indicating that purifying selection is much weaker on average, although still in effect. This difference in the distribution of ω is significant (p<0.005) and shows that VSGs evolve in a more neutral fashion than other genes, resulting in higher substitution rates. However, there is no more evidence for positive selection among VSGs than for 'normal' genes, which reflects our current understanding that VSG sequence evolution in T. brucei is driven predominantly by the diversifying effects of recombination [35], rather than directional selection driven by host immune responses.

RoTat 1.2

Consistent with earlier reports [64] that have suggested RoTat 1.2 (NCBI accession AF317914) as a predominant and diagnostic T. evansi VSG, we identified a closely related ORF in STIB805. The first 1250 bp of a 1431-bp ORF on a 3.9-kb de novo contig are nearly identical to the published RoTat1.2 sequence, while the 200 bp at the 3′ end are nearly identical to Tb927.8.240, annotated as ‘VSG, degenerate’ in TriTrypDB (S11 Fig.). Thus, the gene in STIB805 had most likely obtained a different C-terminal region through recombination, an important mechanism in VSG evolution [35]. We did not identify any close homologs to the 5′ region of RoTat1.2 in the publicly available T. b. brucei TREU 927/4, T. b. brucei Lister 427 or T. b. gambiense DAL972 datasets.

Single nucleotide polymorphisms (SNPs) between T. evansi and T. brucei

A comparison of the TevSTIB805ra to the Tb927 reference genome revealed 354,809 single nucleotide polymorphisms, 68,938 of which were non-synonymous in identified CDS. 32,892 of these non-synonymous SNPs were found in 904 CDS classified as repetitive (e.g. RHS, VSG, ESAG), while 36,046 were found in 6630 CDS from non-repetitive genes (S3 Table). This number is slightly lower than the ∼45,000 non-synonymous SNPs reported for T. b. gambiense orthologs of non-repetitive T. b. brucei TREU 927/4 genes [33]. Additionally, there were 45,269 short indels (insertions/deletions of up to eight nucleotides), of which 5544 were in CDS regions. Of these, 4963 (89.5%) were not an indel of mod 3, and would therefore be expected to cause a frameshift. 3247 (58.6%) of the indels were found in 656 CDS classified as repetitive, while 2297 (41.4%) were found in 1091 CDS from non-repetitive genes (S4 Table). The latter consisted of 1154 insertions, 1000 deletions, and 143 that were complex (a combination of insertion, deletion and/or SNP). The Tb927 reference contains information for one allele only for any heterozygous locus. We therefore also identified SNPs and indels for TREU 927/4 using publicly available Illumina reads and the same criteria that were applied for STIB805. We identified 26,522 non-synonymous SNPs and 1630 indels. Of these, 814 and 58, respectively, were homozygous, thus reflecting discrepancies between the original Sanger sequencing data and the Illumina reads. The cause of these discrepancies, which could arise from a number of potential sources, including DNA isolation from different cultures of the same strain, is unknown. Only 1267 non-synonymous allele variations (SNP or indel) in CDSs (repetitive or non-repetitive) were shared by STIB805 and TREU 927/4, 8710 variations affected the same position, but were different, and 89,043 affected a different position. All non-synonymous SNPs and indels identified in STIB805 and TREU 927/4 are compared in S5 Table, listed by chromosome and gene. In summary, while numerous small sequence variations exist between T. evansi STIB805 and the T. b. brucei reference genome, these are comparable in number to differences between the T. b. brucei and T. b. gambiense subspecies.

Phylogenetic analysis with dihydrolipoamide dehydrogenase (LipDH) and ATP synthase subunit γ as genetic markers

The phylogenetic relationship of various T. evansi and T. equiperdum isolates to each other and to isolates of T. brucei subspecies is controversial. We compared T. evansi STIB805 CDSs with their T. b. brucei TREU 927/4 orthologs as retrieved from Tb927 in order to identify a genetic marker that would likely be informative for the elucidation of the phylogenetic origins of T. evansi and T. equiperdum, and for determining their evolutionary relationships with established T. brucei sub-species. Selection criteria included (i) sufficient SNPs over a length of 1–3 kb (to minimize number of internal primers necessary for amplification and sequencing), (ii) lack of paralogs, and (iii) minimal alignment gaps. One promising candidate, the LipDH gene (Tb927.11.16730), a shared component of four mitochondrial multi-enzyme complexes [65], was selected for sequence analysis. Both LipDH CDS alleles were amplified and sequenced from a total of 40 isolates from various geographical locations (13 T. b. brucei, 3 T. b. gambiense group 1, 4 T. b. rhodesiense, 5 T. equiperdum, 15 T. evansi; S7 Table). Since mutations in subunit γ of the mitochondrial ATP synthase complex were recently identified as important factors in the viability of T. evansi and T. equiperdum, the sequence for both γ alleles was also determined for all isolates.

32 unique LipDH haplotypes were identified. The majority of strains (36/41) had two different alleles and four strains appeared to be homozygous at this locus, all of which were T. b. brucei (although it cannot be ruled out that the primers used were selective for one allele in these cases). Phylogenetic analysis (Fig. 5) showed all but three of the haplotype sequences fell into one of five major clusters with strong support (posterior probabilities ≥0.9), which we refer to as clades V, W, X, Y and Z. Importantly, 14 different haplotypes are derived from the T. evansi/T. equiperdum isolates, which are found in three different clades as well as outside of the clades. This feature is incompatible with monophyly of T. evansi or T. equiperdum. Some subspecies did have relatively restricted phylogenetic diversity, for example T. b. gambiense type 1 had only five closely related haplotypes and certain clades contained sequences from a restricted number of the subspecies; e.g. clade X only contained T. b. rhodesiense and T. b. brucei; clade Y included only T. b. gambiense type 1 and T. b. brucei. This may reflect relatively more recent origins of certain populations, bottleneck events and/or sampling bias, and is in line with most previous work ([40] and references therein).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Bayesian phylogeny of Trypanozoon isolates based on the dihydrolipoamide dehydrogenase gene (LipDH; Tb927.11.16730).

Panel A shows a mid-point rooted tree based on an alignment of 32 unique LipDH haplotypes, assembled from sequences derived from 13 T. b. brucei (Tbb), 3 T. b. gambiense type 1 (Tbg1), 4 T. b. rhodesiense (Tbr), 15 T. evansi (Tev) and 5 T. equiperdum (Teq) samples. Scale units for the phylogeny are substitutions per site. The chart illustrates the distribution of each haplotype among samples from each Trypanozoon taxon. Phylogenetic analysis grouped all but three of the haplotype sequences into one of five major clusters with strong support (posterior probabilities ≥0.9), which are referred to as clades V, W, X, Y and Z. Eight unique Tev/Teq genotypes were found, as summarized in panel B. Discounting minor sequence differences (1–2 SNPs) these were reduced to four major genotypes based on the position of haplotypes in the tree, which mirrored the type of mutation (A281Δ, M282L, A273P, WT) in the C-termini of the ATP synthase subunit γ in these the samples.

https://doi.org/10.1371/journal.pntd.0003404.g005

Eight unique T. evansi/T. equiperdum genotypes (haplotype pairs) were found which, by disregarding minor sequence differences, were reduced to four major genotypes: W/Z, V/V, V/Z, or atypical (Fig. 5; S7 Table). These LipDH genotypes correlated very well with the four ATP synthase γ genotypes found in these strains (Fig. 5B; see also S7 Table).

The largest group (Group 1), consisting of 13 T. evansi isolates (including STIB805) and 2 T. equiperdum isolates, is characterized by LipDH genotype W/Z and γ genotype A281Δ/WT. The single exception, Tev48, had two Z haplotypes (Hap24+Hap29) distinguished by two SNPs, which may be a result of loss of heterozygosity (LOH) followed by two mutations, or it may represent a cloning artefact. Non-T. evansi/T. equiperdum strains that had sequences within these clades originated from both East and West Africa in the case of clade Z, and the T. b. gambiense samples from Ivory Coast in the case of clade W. The remaining T. evansi/T. equiperdum strains were split into three further groups according to the LipDH/ATP synthase γ genotypes as follows. Group 2 is characterized by LipDH V/Z and γ genotype A273P/A273P. The two T. equiperdum strains carrying this genotype (Teq21 and Teq22) were heterozygous for two alleles derived from well-separated clades: one ‘V’ allele (Hap17 or Hap18, 1 SNP difference) and one ‘Z’ allele (Hap14 or Hap15, 1 SNP difference). The divergence between the V and Z alleles (11 or 12 discriminating SNPs), suggests that these strains have evolved out of a T. b. brucei background that had relatively old and distinct LipDH alleles, possibly via a recombination event. Interestingly, the most closely related non-T. evansi/T. equiperdum alleles are found in current East African populations as well as in three T. b. brucei from Ivory Coast. Group 3 is characterized by LipDH V/V and γ genotype WT/WT. The one T. equiperdum strain with a WT γ genotype (Teq23 =  STIB841; probably synonymous with the OVI strain [5]) was heterozygous with two similar ‘V’ clade alleles that are closely related to the V alleles in Teq21, Teq22, Tbr02 (Kenya) and Tbb08 (Zambia) (S7 Table). Group 4 is characterized by atypical LipDH and γ genotype M282L/WT. This genotype was found in a single T. evansi strain (Tev42  =  KETRI 2479) with Hap27+Hap28 and could not be reliably grouped with any other strains.

An interesting aspect of the T. evansi/T. equiperdum genotypes is that most strains possess two divergent alleles that are more similar to those found in other, non-Tev/Teq strains than they are to each other. These non-Tev/Teq strains are themselves heterozygous, with second alleles that are clearly distinct from either of the Tev/Teq alleles. A clear example is the most common Teq/Tev LipDH genotype, W/Z, that is carried by 13 out of 15 T. evansi analyzed and 2 out of 5 T. equiperdum analyzed. This finding is incompatible with allelic divergence due to long-term clonal evolution, rather it shows that recombination between ancestral parasites occurred either at or immediately prior to the origin of this genotype which has subsequently undergone clonal expansion in the absence of the possibility of sexual exchange in the tsetse vector. Three of the four T. b. rhodesiense strains (Tbr01, 03, 04) also exhibit alleles split across different clades as does Tbb06. Interestingly, all Tbb except Tbb06 have alleles from the same clade, possibly indicative of inbreeding.

Cytochrome oxidase 1 (COX1)

The COX1 gene, encoded in the mitochondrial maxicircle DNA, was recently used as a highly informative marker to investigate the phylogeography of T. brucei subspecies [40]. Although most, if not all, T. evansi isolates lack a maxicircle, the presence of at least one T. equiperdum isolate with at least a partial maxicircle in three of the above groups prompted us to determine the COX1 haplotypes for these strains and to investigate their relationship with the Trypanozoon isolates from the earlier study (S7 Table). A maximum parsimony network analysis strongly supported the notion of three independent evolutionary origins for these three groups (Fig. 6). T. equiperdum STIB818 (Teq24; COX1 haplotype 23) links Group 1 with COX1 clade A, which is composed of isolates of all three T. brucei subspecies found across all of sub-Saharan Africa [40]. Although Groups 2 and 3, which share a related ‘V’ LipDH genotype, are both linked to COX1 clade C (composed of T. b. brucei and T. b. rhodesiense isolates from eastern and southern Africa), they are well separated within this clade, indicating independent evolutionary origins. T. equiperdum STIB841 (Teq23) shares both its COX1 haplotype 14 and its LipDH haplotype 19 with Tbr02 from Zambia and Tbb08 from Kenya, both members of the Kiboko group, suggesting relatively recent common ancestry. Note that Tev42 (KETRI2479), the lone representative of the ‘atypical’ LipDH genotype, is not represented in this network since this isolate lacks a maxicircle and therefore the COX1 gene.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. Maximum parsimony haplotype network showing phylogenetic relationships among Trypanozoon haplotypes using cytochrome oxidase 1 (COX1).

Relationships among lineages of T. b. brucei (blue), T. b. rhodesiense (red), T. b. gambiense group 1 (dark green) and T. b. gambiense group 2 (light green) are shown in circles as in reference [40], with haplotypes new to this study shown as octagons, including two new haplotypes for T. equiperdum isolates (pink). The size of each circle or octagon is proportional to the frequency with which a particular haplotype was identified. Numbers in the circles and octagons correspond to haplotype ID. Empty circles indicate haplotypes that are inferred to exist but were not found. The light blue boxes correspond to the T. brucei clades defined in [40]. Haplotypes for samples new to this study are as follows (from left to right): 8 =  Tbg14; 5 =  Tbr01/Tbr04; 20 =  Tbb49/Tbb50; 21 =  Tbb51; 3 =  Tbb20; 2 =  Tbb38; 23 =  Teq24; 14 =  Tbr02/Teq23; 22 =  Teq21/Teq22.

https://doi.org/10.1371/journal.pntd.0003404.g006

Principal component analysis (PCA) of microsatellite markers

Incorporation of microsatellite data for a subset of T. evansi/T. equiperdum isolates into an established PCA network [40] also gave results that are inconsistent with monophyly of either species (Fig. 7). Most T. evansi isolates, together with Teq24 (STIB818), formed a cluster (grey circle) related to, but somewhat distinct from, non-Kiboko T. b. brucei (light blue circle) and T. b. rhodesiense (red circle). The single exception among T. evansi was again Tev42 (KETRI2479), which localized near the centre of the non-Kiboko cluster. Teq21 (BoTat1.1) was also more related to non-Kiboko T. b. brucei, but relatively distant from Tev42. The PCA analysis, consistent with the COX1 and LipDH data, suggested a relatively close evolutionary relationship of T. equiperdum STIB841/OVI (Teq23) with the Kiboko group of T. b. brucei (dark blue circle).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 7. Evaluation of the genetic differentiation between isolates of T. evansi and T. equiperdum and subspecies of T. brucei using principal component analysis (PCA) of microsatellite data.

PCA was performed in R using the package adegenet. Within subspecies of T. brucei, the differentiation between temporally and geographically cohesive subgroups was estimated using DEST and calculated with the program smogd. Points representing individual genotypes are connected by a line to the centroid of an ellipse, which circumscribes a region encompassing 95% of the variance observed within six trypanosome subgroups identified: major T. evansi and T. equiperdum group (grey and pink), T. b. rhodesiense (red), T. b. brucei Kiboko (dark blue), T. b. brucei non-Kiboko (light blue), T. b. gambiense group 1 (dark green), T. b. gambiense group 2 (light green). T. evansi and T. equiperdum isolates that fell outside the major group are shown as black data points. The wide distribution of T. evansi and T. equiperdum isolates among distinct subgroups strongly supports multiple independent origins for these dyskinetoplastic strains. The first two principal components (PC1 and PC2) explain 29.2% and 8.4% of the total variance in the data, respectively. Data for isolates other than T. evansi and T. equiperdum had been published previously [40].

https://doi.org/10.1371/journal.pntd.0003404.g007

The results of the phylogenetic analyses are summarized in Table 4, along with the ATPase subunit γ genotypes and, where known, dominant minicircle class and RoTat 1.2 VSG genotype. Combined, these markers suggest that the T. evansi/T.equiperdum isolates investigated in our study can be arranged into four distinct groups of independent evolutionary origin (see Discussion)

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Proposed grouping of T. evansi/T. equiperdum isolates.

https://doi.org/10.1371/journal.pntd.0003404.t004

Conclusions

Taken together, the genome analysis and accompanying phylogenetic studies presented in this work revealed important insights into the biology and evolution of T. evansi STIB805 and dyskinetoplastic trypanosomes in general. Their re-classification as subspecies of T. brucei, i. e. T. b. evansi and T. b. equiperdum, seems clearly justified considering the vast similarities observed at the genome level and the lack of monophyly confirmed by phylogenetic analyses. Important questions that remain include the molecular basis of tsetse-independent transmission and the evolutionary timescale of the appearance of the T. evansi/T. equiperdum groups. The genome data presented here provide an important tool for future studies aimed at resolving these questions.