Evolutionary genomics of epidemic visceral leishmaniasis in the Indian subcontinent

Abstract

Leishmania donovani causes visceral leishmaniasis (VL), the second most deadly vector-borne parasitic disease. A recent epidemic in the Indian subcontinent (ISC) caused up to 80% of global VL and over 30,000 deaths per year. Resistance against antimonial drugs has probably been a contributing factor in the persistence of this epidemic. Here we use whole genome sequences from 204 clinical isolates to track the evolution and epidemiology of L. donovani from the ISC. We identify independent radiations that have emerged since a bottleneck coincident with 1960s DDT spraying campaigns. A genetically distinct population frequently resistant to antimonials has a two base-pair insertion in the aquaglyceroporin gene LdAQP1 that prevents the transport of trivalent antimonials. We find evidence of genetic exchange between ISC populations, and show that the mutation in LdAQP1 has spread by recombination. Our results reveal the complexity of L. donovani evolution in the ISC in response to drug treatment.

Keywords: epidemiology; evolution; genomics; global health; infectious disease; leishmania donovani; microbiology.

Figure 1.. History and geography of Indian subcontinent L. donovani.

(a) Location of the patients from which the 204 L. donovani genomes were isolated, and of historical Kala-Azar outbreaks. Genetic groups of the parasite isolates are indicated by the colour of the dots representing them, matching those in Figure 2a,c. Sampling dates and locations are summarised in Figure 1—figure supplement 1, and detailed information about each strain including GPS coordinates are given in the source data file. Citations are to historical primary literature reviewed and cited in (Gibson, 1983). Posterior probability distributions of estimated ages for the oldest split in (b) the main population in Bihar and Nepal and (c) the ISC5 group associated with Sb resistance. Dark shading shows estimates under a strict molecular clock, light shading from relaxed molecular clock and lines show relaxed clock results with Bangladeshi and putative hybrid isolates included. (d) Estimated effective population size through time for ISC5 population (green) and the rest of the parasite population (black/grey). Lines show median of posterior distributions, dark and light shading cover 50% and 95% of the posterior density respectively. Dates for all splits on this phylogeny and other results of phylogeographic analysis are shown in Figure 1—figure supplement 2. DOI: http://dx.doi.org/10.7554/eLife.12613.003

Figure 1—figure supplement 1.. Sampling of genetic groups.

Pie charts indicate the number of samples in each year (columns), for each genetic group (rows) coming from each country (grey shading). Horizontal lines connect and surround isolates of each group, with colours matching the groups shown in panels (b) and (c). *8 samples from Bangladesh were all sampled in 2006, and form a distinct population to Nepalese and Indian isolates (ISC2). DOI: http://dx.doi.org/10.7554/eLife.12613.004

Figure 1—figure supplement 2.. Full results of discrete-space, constant population size molecular clock Bayesian phylogeography analysis of core population.

(a) Maximum posterior probability phylogeny, with tips coloured by country of origin for sample (green for India, blue for Nepal), and branches coloured by maximum posterior probability of country reconstructed in discrete phylogeography model. Values on nodes indicate posterior probability of assigned country/colour, with filled circles marking nodes with probability 1. Other panels represent posterior probability distributions for rates of migration (lineage switches per month) from (b) Nepal to India and (c) from India to Nepal. Note the mode (maximum posterior probability estimate) for migration from Nepal is zero, but non-zero migration in the reverse direction is supported. DOI: http://dx.doi.org/10.7554/eLife.12613.005

Figure 2.. Genealogical history of L. donovani from the ISC.

(a) Maximum-likelihood tree based on SNPs called for 191 strains (see Figure 2—figure supplement 1) from the core population in the Indian subcontinent. Samples are coloured by population assignment, with putative hybrid strains not clustered in the main groups in black. Further analysis confirms the hybrid ancestry of some of these isolates (Figure 2—figure supplement 2). (b) Unrooted phylogenetic network of the L. donovani complex based on split decomposition of maximum-likelihood distances between isolates described here, reference genome isolates and two published Sri Lankan isolates (Zhang et al., 2014). (c) Model-based clustering of 191 isolates from the core population reveals six discrete monophyletic groups, and some groups and other samples of less certain ancestry. Coloured bars show the fraction of ancestry per strain assigned to a given cluster, with colours assigned to the population most closely related to each cluster. More detailed population clustering analysis shows largely congruent results (Figure 2—figure supplements 3 and 4). DOI: http://dx.doi.org/10.7554/eLife.12613.006

Figure 2—figure supplement 1.. Flowchart of SNP detection using COCALL.

Overview of the SNP detection method COCALL (COnsensus of SNP CALL). COCALL finds genetic variants that show a concordant signal over five different SNP callers (Cortex, Freebayes, GATK, Samtools Mpileup and Pileup). See supplementary methods for details. DOI: http://dx.doi.org/10.7554/eLife.12613.007

Figure 2—figure supplement 2.. Haplotype networks for core population isolates.

Haplotype networks indicate putative hybrids as isolates with ancestry from multiple distinct populations. Chromosomal haplotype neighbour-joining networks of the phased data for the core population were constructed using the R ape package. Each node represents one haplotype variant for (a) chromosome 32 and (b) chromosome 35, coloured by group. Black lines are network edges and red lines connect haplotype variants from the same isolate for selected isolates where haplotypes appear in different parts of the network (with isolate names shown). Six ungrouped isolates (BHU815/0, BHU764/0cl1, BHU274/0, BHU574cl4, BHU581cl2, BHU572cl3) have mixed ancestry from ISC5 and other groups, and two (BHU744/0 and BHU774/0) have a mix of ISC6/7/8/9/10 haplotypes. No mixing among ISC2/3/4 was evident. DOI: http://dx.doi.org/10.7554/eLife.12613.008

Figure 2—figure supplement 3.. Haplotype similarity for core population isolates.

Heatmap showing the mean expected number of haplotypes shared between pairs of core population isolates. Samples listed on the y-axis are haplotype donors to those on the x-axis. 18,747 phased genotypes at 2397 SNPs sites were computed with Chromopainter v0.0.2 using recombination rates from PHASE for 79 haplotype chunks with c=0.00054 effective chunks. This image confirms six discrete populations ISC2-7 and illustrates complex ancestry in certain samples not belonging to these groups. DOI: http://dx.doi.org/10.7554/eLife.12613.009

Figure 2—figure supplement 4.. Mosaic ancestry patterns in eight putative hybrid L. donovani isolates.

Representative samples from three potentially parental groups (BPK282/0cl4, ISC6; BHU200/0, ISC7; BPK275/0cl18, ISC5) were compared to eight putative hybrid samples (BHU815/0, BHU764/0cl1, BHU274/0, BHU574cl4, BHU581cl2, BHU572cl3, BHU744/0 and BHU774/0). To the left is a maximum-likelihood tree constructed with RAxML showing the evolutionary history of the aligned haplotypes. The table shows a set of SNPs for which ChromoPainter assigned ancestry probability values >0.4 in any of these eight hybrids. Individual SNPs are coloured if the sample had an ancestry probability >0.4: uncoloured ones represent those observed in multiple ISC populations. All isolates have mixed ancestry from the two groups, but four isolates (BHU574cl4, BHU764/0cl1, BHU815/0 and BHU274/0) have haplotypes that appear to have a more complex origin. DOI: http://dx.doi.org/10.7554/eLife.12613.010

Figure 3.. Structural variations in ISC L. donovani.

(a) Stacked barplots per chromosome showing the proportion of ISC strains that are monosomic, disomic, trisomic, tretrasomic or pentasomic for the respective chromosome. A full breakdown of somy per strain is presented in Figure 3—figure supplement 3, and a complete catalogue of other structural variants in Figure 3—figure supplement 1. Violin plots showing the copy number of MAPK1 (b) and H-locus (c) per ISC group, except for ISC1 where these amplicons were absent. These amplicons are intra-chromosomal (Figure 3—figure supplement 2). (d) Tetrameric protein model of the transport protein aquaglyceroporin-1. The C-terminus part that is affected by the 2-nucleotide frameshift found in all ISC5 isolates is shown in magenta. Image was created using PyMOL version 1.50.04 (Schrödinger). DOI: http://dx.doi.org/10.7554/eLife.12613.011

Figure 3—figure supplement 1.. Copy number variants in all 206 genomes.

The position in the genome is shown on the y-axis, while individual isolates are shown on the x-axis. Colours of each copy number variant (CNV) represent the haploid depth variation (D) compared to the median depth for that chromosome (see legend for colour key). When the depth of the majority of the strains is high like the episome in ch23, this appears as a reduced depth in the strains that lack the episome. The length of each CNV is reflected by its length along the y-axis (i.e. thickness of the line). Four major CNVs – gp63, rDNA, an episome in ch23 and the MAPK amplification – are indicated with arrows. Group-specific copy number variants were highlighted with a box and numbered – detailed information about these CNVs are given in the table. The 206 samples included here are 204 ISC samples with L. infantum JPCM5 (MCAN/ES/1998/LLM-877) and L. donovani LV9 (MHOM/ET/1967/HU3) for reference. DOI: http://dx.doi.org/10.7554/eLife.12613.012

Figure 3—figure supplement 2.. Copy number variation by intrachromosomal tandem duplication or extrachromosomal linear amplification in clinical isolate.

(a, c) Chromosomes from L. donovani BPK282/0cl4 (ISC6), BPK380/0 (ISC9) and BPK026/0cl5 (ISC1) were separated by pulsed-field gel electrophoresis (PFGE). (b) The MAPK1-locus and H-locus were detected by southern blot hybridization with probes specific for MAPK1 or HTBF, respectively. Hybridization was only observed in fragments of lengths equal to those of chr36 (~2.5 Mb) and chr23 (~1 Mb) and no additional smaller fragments were observed, indicating the absence of extra-chromosomal amplifications. (d) In contrast, linear extrachromosomal amplification (as evidenced by a second and smaller band) is shown for chromosome 35 in BPK380/0 by hybridization of a probe specific to the LinJ35.4130 gene. DOI: http://dx.doi.org/10.7554/eLife.12613.013

Figure 3—figure supplement 3.. Chromosome number variation in L. donovani in the ISC.

Average number of chromosomes found within each cell culture for each of the 36 chromosomes (y-axis) and each of the 204 L. donovani strains (x-axis). Samples are coloured by population assignment following Figure 1c, with strains not clustered in the main populations shown in white. DOI: http://dx.doi.org/10.7554/eLife.12613.014

Figure 4.. SNP heterozygosity and somy variation in two subclades.

Two subclades show an expansion of polysomic strains from disomic ancestors (below) and an expansion of disomic strains from polysomic ancestors (above). Somy variation per chromosome (1–36; above heatmap) and the total number of heterozygote SNPs (right to heatmap) are shown for each individual strain. DOI: http://dx.doi.org/10.7554/eLife.12613.015