Variant antigen repertoires in Trypanosoma congolense populations and experimental infections can be profiled from deep sequence data using universal protein motifs

Abstract

African trypanosomes are vector-borne hemoparasites of humans and animals. In the mammal, parasites evade the immune response through antigenic variation. Periodic switching of the variant surface glycoprotein (VSG) coat covering their cell surface allows sequential expansion of serologically distinct parasite clones. Trypanosome genomes contain many hundreds of VSG genes, subject to rapid changes in nucleotide sequence, copy number, and chromosomal position. Thus, analyzing, or even quantifying, VSG diversity over space and time presents an enormous challenge to conventional techniques. Indeed, previous population genomic studies have overlooked this vital aspect of pathogen biology for lack of analytical tools. Here we present a method for analyzing population-scale VSG diversity in Trypanosoma congolense from deep sequencing data. Previously, we suggested that T. congolense VSGs segregate into defined "phylotypes" that do not recombine. In our data set comprising 41 T. congolense genome sequences from across Africa, these phylotypes are universal and exhaustive. Screening sequence contigs with diagnostic protein motifs accurately quantifies relative phylotype frequencies, providing a metric of VSG diversity, called the "variant antigen profile." We applied our metric to VSG expression in the tsetse fly, showing that certain, rare VSG phylotypes may be preferentially expressed in infective, metacyclic-stage parasites. Hence, variant antigen profiling accurately and rapidly determines the T. congolense VSG gene and transcript repertoire from sequence data, without need for manual curation or highly contiguous sequences. It offers a tractable approach to measuring VSG diversity across strains and during infections, which is imperative to understanding the host-parasite interaction at population and individual scales.

Figure 1.

Maximum likelihood (ML) phylogeny of T. congolense VSGs. The phylogeny was estimated from full-length VSG protein sequences of IL3000 (Kenya), IL3674 (The Gambia), and IL3900 (Forest subtype, Burkina Faso) with RAxML (Stamatakis 2014) using a ML method with a WAG+Γ model and 100 bootstrap replicates. The 15 phylotypes identified in IL3000 are color-coded according to the key. Positions of example sequences from IL3674 and IL3900 are indicated according to the key. Labels for the internal nodes of each phylotype (marked by the open squares) are shown on the right. These labels indicate the bootstrap percentages for ML from the complete tree (RAxML), as well as ML (PhyML) (Guindon et al. 2010), ML (MEGA7) (Kumar et al. 2016), neighbor joining (NJ) (Felsenstein 1989), and posterior probabilities (BI) (Huelsenbeck and Ronquist 2001) estimated from a pruned tree containing 147 sequences. Tree is rooted with two T. vivax VSG sequences (Fam23).

Figure 2.

Performance of the protein motif-based variant antigen profile (VAP). (A) Correlation of motif-based and manually curated phylotype frequencies in the T. congolense IL3000 reference genome sequence. Pearson's product moment correlation statistics: R²= 0.88, t₍₁₃₎ = 9.7321, P < 0.001. (B) Correlation of motif-based and manually curated phylotype frequencies in 41 T. congolense strains. Manual VAPs were estimated by counting the top matches from BLASTx (Altschul et al. 1990). Pearson's product moment correlation: R²= 0.64, t₍₅₆₆₎ = 34.39, P < 0.001. Phylotypes are color-coded according to the key.

Figure 3.

Relationships between the VSG repertoire, geography, and population structure in T. congolense. (A) ML phylogeny of T. congolense strains in this study based on whole-genome single-nucleotide polymorphisms (SNPs), estimated with RAxML (Stamatakis 2014) with a GTR + Γ model and 100 bootstrap replicates (branches with bootstrap greater than 70 are shown in bold). Labels “i” to “v” denote examples referred to in the text. Label “i” shows the long phylogenetic distance between T. congolense Savannah and Forest subtypes; “ii” points to the only clade maintaining a geographic signature. Labels “iii,” “iv,” and “v” show examples of lack of concordance between the population history recapitulated by the SNP phylogeny and the VAP, demonstrated by the dendrogram. (B) VAPs for all strains shown as a heatmap of the proportions of 15 universal phylotypes. (C) A dendrogram depicting the relationships among VAPs based on Euclidian distances estimated in R. Gray ribbons link the position of parasite strains in A and C. (D) A bar chart showing the average proportion of each phylotype (mean ± σ) across all strains. Strains are color-coded by provenance according to the key.

Figure 4.

Phylotype variation across the sample cohort. The heatmap represents phylotype variation across the sample cohort expressed as the deviation from the mean. The dendrogram reflects the relationships among the VSG repertoires of each strain. Strains are color-coded by location of collection according to the key. Labels “i” to “ii” denote examples of phylotype variation signatures referred to in the text. Label “i” shows a pattern of underrepresented P1-3 among strains of multiple countries; “ii” shows a pattern of overrepresented P5-6 in Gambian isolates; “iii” shows a pattern of underrepresented P15 common to Forest-subtype isolates.

Figure 5.

VAP applied to mVSG expression in experimentally infected tsetse mouthparts. (A) Transcriptomic VAPs of trypanosomes extracted from tsetse mouthparts. VAPs from the transcriptomes are remarkably similar yet significantly different from the genomic VSG repertoire (Poisson regression, P < 0.001) and the VSG found at telomeric expression sites. Infection 1 represents a sample of 40 pooled mouthparts; infection 2 represents 24 individual mouthparts; infection 3 represents a sample of 131 pooled mouthparts after metacyclic parasite enrichment by anion exchange chromatography. The genomic VAP represents the average profile of 24 sets of 79 VSGs randomly sampled from the genome of Tc1/148. Stacked columns are color-coded by phylotype according to the key. The number of VSG transcripts recovered in each sample infection is noted in the figure. (B) Comparison of average phylotype proportion (adjusted for transcript abundance) in transcriptomic samples presented in A and genomic profiles from a random selection of VSG of Tc1/148 (mean ± σ). Statistical analysis reveals that, in comparison to the genome, P7, P12, and P15 are underrepresented in the transcriptomes (independent t-test, P-value <0.001), while P4, P8, P9, and P11 are significantly overrepresented (independent t-test, P-value <0.001). (C) ML phylogeny of P8 showing 12 distinct loci found across our T. congolense strain genomes (denoted by gray boxes), the position of Tc1/148 P8 transcripts, and those from two previous studies (Eshita et al. [1992] [UniProt ID “M74803.1” and “M74802.1”] and Helm et al. [2009] [“mVSG1”]). Internal nodes are labeled with bootstrap values greater than 70.