Ethics statement

This study was conducted as part of a public health surveillance activity following a request by the Ministry of Health, Garissa County to provide diagnostic and laboratory support during a protracted leishmaniasis outbreak in Garissa County. The use of these routine surveillance data in research is considered acceptable because of the public health impact, the fact that risks to the subjects are minimized through data deidentification, and the impracticability of seeking retrospective consent (ref to the WHO guideline). The MOH provided the samples for diagnostic support and we recorded the bio-specimens. The approval to provide diagnostic and laboratory support, including whole genome sequencing was approved by the Kenya Medical Research Institute Scientific and Ethics Review Unit (KEMRI SERU, # 4634) and the Walter Reed Army Institute of Research, Human Subject Protection Branch, (WRAIR #2699). Whole genome sequencing of eight samples (S1 Table) conducted at Institute of Tropical Medicine, Antwerp (ITM) using Leishmania target enrichment (SureSelect Sequencing, SuSL-seq) was approved by the ITM IRB (code 45/2024).

Study site and design

The study utilized 286 archived blood samples collected from patients suspected of having visceral leishmaniasis during the 2019–2022 VL outbreak in Garissa and Kitui counties, Kenya. Patients’ blood samples were collected at Garissa County referral hospital (S1 Fig). Consent was not obtained from the patients because this was a public health activity. The risks to the subjects were minimized through data deidentification. This study applied WGS to characterize Leishmania directly from patient blood samples, eliminating the need for parasite isolation and culture.

DNA extraction and screening for Leishmania parasites

DNA was extracted from the archived blood samples using the MagMAX CORE, Nucleic Acid Purification reagents (Thermo Fisher Scientific, USA). For detection of Leishmania parasites, extracted DNA was screened at genus level by quantitative real-time PCR of the arginine permease gene AAP3 (AAP3-qRT-PCR). The following primer sets and probe were used [27]: forward - (AAP3), 5’ GGC GGC GGT ATT ATC TCG AT 3’, reverse - (AAP3), 5’ ACC ACG AGG TAG ATG ACA GAC A 3’ and probe - FAM 5’ ATGTCGGGCATCATC 3’ NFQ. All reactions were conducted in a 10.0μL reaction volume comprising 5.0µL of 2x SensiFAST master mix (Meridian Bioscience USA, https://www.meridianbioscience.com), 0.8µL of each primer at a concentration of 10 μM, 0.4µL of probe at concentration of 10μM, 2.0µL of DNA, and 1.0µL of PCR water (Thermo Fisher Scientific, USA). Each reaction included known positive controls of Leishmania spp (L. donovani and L. major). Non-target controls (PCR water) was used as negative control to track contamination.

Species identification and targeted genotyping

DNA was used to PCR amplify the Hsp70 gene and ITS region using primer sets as published elsewhere [28] with minor modifications. The following primers were used: Hsp70 gene forward primer (F25) 5’ -GGA CGC CAC GAT TKC T- 3’ and reverse primer (R1310) 5’ -CCT GGT TGT TCA GCC ACT C- 3’. For the ITS, the forward primer (LITSR): 5’ -CTG GAT CAT TTT CCG ATG- 3’ and reverse primer (LITSV): 5’ -ACA CTC AGG TCT GTA AAC- 3’ were used. The assays were conducted in 25μL reaction volume comprising: 12.5μL, 2X Mytaq Red Mix (Meridian Bioscience, USA), 0.5μL of each primer, 9.5μL of PCR grade water (Thermo Fisher Scientific, USA), and 2μL of the sample. The following conditions were used: For Hsp70: 94°C for 5 minutes, followed by 45 cycles of 94°C for 30 seconds, 61°C for 1 minute and 72°C for 3 minutes, followed by a final extension step at 72°C for 10 minutes. For ITS: 95°C for 2 minutes followed by 34 cycles of 95°C for 20 seconds, 53°C for 30 seconds, 72°C for 1 min, followed extension at 72°C for 6 min. The amplified products were visualized on a 2% agarose gel stained with 3.0μL GelRed (Biotium Inc., USA). Amplicons were cleaned using AmpureXP beads (Beckman Coulter, USA) following manufacturer’s instructions, and then used to construct sequencing libraries with the Collibri ES DNA Library Prep Kit from Illumina (Thermo Fisher scientific, USA). The individual libraries for both Hsp70 and ITS loci were normalized to 4nM concentration and then pooled separately. The pooled amplicon library was denatured and diluted to a final concentration of 8pM, then spiked with 5% PhiX (Illumina, USA) as a sequencing control and sequenced on the MiSeq platform, using v3 600 cycle kit; 2 × 300 bp paired reads (Illumina, CA, USA).

Whole genome sequencing by target enrichment

In this proof-of-principle study, eight samples with high parasitemia as measured by genus-specific PCR (Ct values between 26–33) were selected for WGS. For SuSL-seq, two further criteria were added: the total DNA concentration and a measure of the % of Leishmania DNA in the sample: a minimum of 10 ng of total DNA and 0.006% of Leishmania DNA are required [23]. Total DNA was measured by qubit fluorometer (Thermo Fisher Scientific, USA) and 21–50ng used for library preparation. The quantification of Leishmania DNA in the library sample was performed by qPCR as described by Domagalska et al. [23]: Leishmania DNA in the samples ranged from 0.0064 to 0.1415%.

DNA fragmentation was performed enzymatically using the Agilent SureSelect Enzymatic Fragmentation Kit (Agilent, Santa Clara, USA). Libraries were prepared using the SureSelect XT HS Target Enrichment System for Illumina (Agilent, Santa Clara, USA). The probe design used in this study was based on a previously published SureSelect design [24]. Briefly, to generate the custom Agilent SureSelect (SuSL) capture array, repetitive and low-complexity regions of the reference genome (Leishmania donovani BPK282A1) were masked prior to probe selection. Candidate 120-mer probes were designed at 1x tiling density to minimize overlap, avoiding regions with ambiguous bases (Ns) and excessive sequence redundancy. To ensure high specificity, all probes were aligned using BLAST against the genome of Homo sapiens; any probes showing significant homology to these genomes were excluded to prevent off-target hybridization. The final design (Agilent SuSL design ID S3377046) comprised approximately 318,000 probes, collectively targeting ~29.99 Mb of the reference genome. Adaptor-ligated libraries were prepared using 11 cycles in pre-capture PCR. Libraries were hybridized with the custom probes at a dilution 1:10 and captured with Dynabeads MyOne Streptavidin T1 magnetic beads (Thermo Fisher Scientific, Waltham, USA). After washing steps, the DNA captured by streptavidin beads was amplified by PCR and purified with AMPure XP beads. The quantity and quality of the libraries were assessed on a TapeStation using High Sensitivity D1000 ScreenTape (Agilent Technologies, Santa Clara, USA). Sequencing was conducted on the Illumina NovaSeq platform using 2x150 bp paired reads at GenomeScan (Netherlands), for which 59,22–78,49 million raw reads were obtained.

Analysis of data (targeted sequencing)

AAP3-qRT-PCR results for positive samples at genus level were tabulated as either negative or positive based on the set cycle threshold (Ct) values. For the Hsp70 and ITS sequences, demultiplexed sequence reads from the Miseq were processed using the ngs_mapper (https://github.com/VDBWRAIR/ngs_mapper) pipeline which performed read trimming, reference mapping and consensus sequence generation. The consensus sequences were checked manually to resolve errors using Geneious prime 2022.2.1 (GraphPad LLC, MA, USA) and https://igv.org. Blastn v2.15.0 was used for homology analysis of the Hsp70 (n = 86) and ITS (n = 79) sequences. Reference sequences obtained from GenBank (NC_018255.1 for Hsp70 gene and NC_018254 for ITS region) were used to guide the homology analysis. To establish the phylogenetic relationship between Kenyan Leishmania spp and those from other regions, a comprehensive dataset of curated, annotated, and published Leishmania sequences were retrieved from NCBI GenBank (Hsp70, n = 67 and ITS, n = 65) (S2 Table) and aligned using the MAFFT v7.490; algorithm: E-INS-I in Geneious prime v.2022.2.1 (Auckland, New Zealand) plugin.

The Hsp70 and ITS sequences were first analyzed separately and then concatenated to improve the phylogenetic resolution of individual genes. The concatenated sequences (n = 64) were aligned with previously published Leishmania sequences (n = 36) (S2 Table). The individual and concatenated genes were used to infer Maximum-likelihood (ML) trees using the best fitting nucleotide substitution model as determined by ModelFinder [29] within IQ-tree v2.2.26 (https://pubmed.ncbi.nlm.nih.gov/28481363/). Branch support was determined using both the ultrafast bootstrap approximation (UFboot) with 1,000 replicates and the SH-like approximate likelihood ratio test (SH-aLRT) also with 1,000 bootstrap replicates. Visualization and annotation of the resulting phylogenetic trees was done using Figtree version 1.4.2 (http://tree.bio.ed.ac.uk/software/Figtree). Taxa with unusually long branches had their consensus sequences re-evaluated for chimeras using UCHIME v4.2.40 against the QIIME eukaryotic database (release 2025-02-19), and the non-chimeric reads were used for consensus generation. The consensus sequence reads were then mapped to the L. donovani reference genome using the ngs_mapper pipeline. A Majority Rule Consensus was applied, meaning the base with the highest read support at each position was selected, ensuring the sequence represents the truly dominant allele. Positions not meeting sufficient coverage were masked with ‘N’s, guaranteeing a high confidence phylogenetic sequence.

Analysis of data (whole genome sequencing)

For WGS by target enrichment, demultiplexed paired end reads from the NovaSeq6000 platform were assessed for quality using FastQC v0.14.1 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and FastQ Screen (https://www.bioinformatics.babraham.ac.uk/projects/fastq_screen/). In addition to the genomes sequenced in the context of this work, we also added a set of representative sequences of the different clusters as defined in Franssen et al. [21] as available under the NCBI BioProject number PRJEB2600, PRJEB2724, PRJEB8947 and PRJEB2115 (S2 Table).

Reads were aligned to the L. donovani BPK282 reference genome [30] using BWA (v0.7.17) [31] with a seed length of 50. Only properly paired reads with a mapping quality >30 were retained using SAMtools [32]. Duplicate reads were removed using Picard (v2.22.4) with the RemoveDuplicates function. SNP calling followed the GATK (v4.1.4.1) best practices workflow: (i) HaplotypeCaller was used to generate GVCF files for each sample; (ii) individual GVCFs were merged using CombineGVCFs; (iii) genotyping was performed with GenotypeGVCFs; and (iv) SNP and INDEL filtering was applied using SelectVariants and VariantFiltration, with filtering criteria based on GATK best practices (QD > 2, QUAL > 30, SOR < 3, FS < 60, MQ > 40, MQRankSum > -12.5, ReadPosRankSum > -8.0 [33]. Variant regions associated with known drug resistance markers were extracted using BCFtools version 1.22 [34]. Impact of those mutations was assessed using the SnpEff software version 5.0c [35], and visualized with the heatmap package in R. Overlap of SNPs between two strains was determined using the vcfR [36] (v1.13.0) and adegenet (v2.1.8) [37] with visualization performed using pheatmap (v1.0.12).

For phylogenetic analysis, bi-allelic SNPs were extracted from VCF files using BCFtools version 1.22 and converted to Phylip format with the vcf2phylip.py script (https://github.com/edgardomortiz/vcf2phylip). Phylogenetic trees were inferred using RAxML, version 1.2. [38] under the GTR + G model, using L. aethiopica L100 as an outgroup. Tree visualization was performed with ggtree (v 3.14) [39]. To construct unrooted phylogenetic networks, bi-allelic SNPs were converted to FASTA format using the vcf2fasta.py script (https://github.com/FreBio/mytools/tree/master) and analyzed in SplitsTree4, version 6.4.13 using the NeighborNet algorithm [40].

For downstream population genomic analysis, SNP pruning for linkage disequilibrium (LD) was performed using PLINK 2.0 (--indep-pairwise 50 10 0.5) [41], with a threshold of r² = 0.5, which was selected to retain sufficient SNPs for population structure inference while minimizing the impact of tightly linked variants. This relatively permissive threshold was chosen because of the moderate LD decay observed in previous Leishmania genomic epidemiology studies [42–44] and to maintain representation across the genome despite local variation in recombination rates. ADMIXTURE v.1.3.0 [45] was used to assess ancestry in both WGS and SureSelect samples, excluding clonal individuals within each phylogenetic group. The number of populations (K) was evaluated from 4 to 11 using a five-fold cross-validation procedure. Population structure inferred by ADMIXTURE was visualized with ggplot2 in R.

For the exploration of possible genomic signatures of drug resistance, we selected 11 loci that were previously shown to be involved in L. donovani resistance to drugs used in clinical practice. Relevant regions [46–50] were subset using BCFtools, version 1.22 [34]. Only SNPs having an effect at the protein level (missense and non-sense mutations) were retained. Visualization of the heatmap was performed using the pheatmap function in R.

Screening of Leishmania parasites

The demographics associated with the Leishmania samples is shown in S3 Table. Out of the 286 blood samples tested by AAP3-qRT-PCR, 128 (45%) were positive for Leishmania at the genus level. There was a 3:1 male:female infection ratio probably because of gender roles in this nomadic community. Men spend prolonged periods herding livestock where sand flies thrive. Additionally, seasonal migrations with livestock further expose men to diverse sand fly habitats, while women, who primarily engage in household activities, encounter less risk [13].

Phylogenetic trees from individual genes

After quality control and assembly, we obtained consensus sequences of sizes 1,292 bp for Hsp70 and 1,060 bp for ITS loci. Of the 128 samples positive for Leishmania, 86 generated usable sequences for Hsp70 gene and 79 for the ITS region (S4 Table). Homology analysis of the Hsp70 (S2 Fig) and ITS (S3 Fig) sequences revealed monophyletic grouping of study samples that clustered with the L. donovani species complex obtained from the NCBI GenBank.

The closest relatives of the study sequences inferred from individual genes differed, with Hsp70 (S2 Fig) showing phylogenetic proximity to countries north of Kenya (Sudan, Ethiopia) and ITS (S3 Fig) showing phylogenetic proximity to India, France, Spain, Morocco, Sri-Lanka, Sudan and Ethiopia. Given the different phylogenetic histories of Hsp70 and ITS genes, we concatenated 64 study genomes against similar genomes available from the GenBank. In the concatenated tree, the study genomes clustered with the L. donovani complex in a group with L. donovani BPK282A1 (Fig 1). Like in S3 Fig., the study sample Kenya/GSA-164/2020 remained an outlier in the concatenated tree (Fig 1, shown in red star) and branched with the main L. major cluster, in a well-supported branch (bootstrap support of 63%). Because the Hsp70 sequence for GSA-164 showed no substantial difference from other L. donovani strains in this study (S2 Fig), while the ITS sequences were highly divergent (S3 Fig), we verified that the observed divergence was not due to introduction of chimera.

Download:

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

Fig 1. Maximum-likelihood phylogenetic tree inferred from an alignment of concatenated study Hsp70 and ITS sequences (black) and select Leishmania sequences retrieved from NCBI GenBank (colored text).

Red star shows an outlier study sample (Kenya/GSA-164/2020) that branched with the main L. major cluster. The yellow stars shows eight of the current study samples that were selected for targeted whole genome sequencing (see Fig 2). Branch support was estimated using SH-like likelihood ratio test and are indicated as numbers. The scale bar represents the number of substitutions per site.

https://doi.org/10.1371/journal.pntd.0013144.g001

Direct Leishmania genome sequencing in clinical samples after target enrichment

Based on parasite load, eight blood samples with Ct values of <33 were selected (S5 Table) as pilots for direct targeted sequencing by SuSL-seq. Genome sequencing was successful for all the samples, with percentage of genome coverage > 5x ranging between 21.30% and 88.39% (average 67.97%). Summary of sequencing performances statistics can be found in S5 Table. Phylogenetic analysis of the eight genomes together with published genomes of L. donovani (including groups Ldon1–5, following classification by Franssen et al. [21], L. infantum and the three dermotropic species (L. tropica, L. major and L. aethiopica) [28] allowed us to identify the Leishmania clades in these 8 Kenyan samples. They clearly branched in the L. donovani s.s. cluster, separately from L. infantum and the dermotropic species (Fig 2). The eight Kenyan samples formed two distinct clades. Five of these samples clustered within the Ldon5 group, which includes isolates from Kenya and Ethiopia collected between 1954 (LRC-L53) and 2009 (AM560WTI) [21]; they were called K-Ldon5 (see Fig 2 red taxa). The remaining three samples clustered together close to three ‘other Ldon’ samples (LRV-L740 from Israel, GE, and LEM3472 from Sudan) in a distinct intermediate position in the rooted phylogenetic tree, between the Ldon3 group (Sudan/Ethiopia) and Ldon4 group (Iraq) (Fig 2); they were called K-Ldon3/Ldon4 (see Fig 2 purple taxa). In the reticulated network (S4 Fig), the three K-Ldon3/Ldon4 samples (purple taxa) branched separately from ‘other Ldon’ samples, suggesting an independent evolutionary origin. Interestingly, both K-Ldon5 and K-Ldon3/Ldon4 were present in both Garissa (Balambala and Kitui (Mwingi North) counties (see S1 Fig and S1 Table).

Download:

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

Fig 2. Rooted phylogenetic tree based on genome-wide bi-allelic SNPs inferred using RAxML, with L. aethiopica L147 as the outgroup.

Most relevant bootstrap values are indicated in blue. The color code represents clustering as proposed by Franssen et al. [21] with samples from this study distributed across two clusters, highlighted in red and purple (“KenyaLeish”). The taxa in blue (three samples) indicate ‘Other-Ldon’ hybrid genotypes identified by Franssen et al. [21].

https://doi.org/10.1371/journal.pntd.0013144.g002

Given the intermediate phylogenetic position of the three Kenyan samples, we assessed if this genotype could have been generated by recombination between the previously reported L. donovani groups. Analysis of population structure was done in ADMIXTURE for L. donovani based on 61,186 genome-wide bi-allelic SNPs. After LD pruning, a total of 4050 SNPs remained and were used as input for ADMIXTURE analysis. We evaluated K = 2–20 using fivefold cross-validation. ADMIXTURE version 1.3.0 computes point estimates only, without standard errors, as noted in the log output. To assess model stability, we repeated each K value across 10 independent runs with different random seeds and obtained consistent ancestry proportions among replicates. The mean CV error values for each K are visualized including the standard error in S5 Fig, showing that CV error reached a minimum with K-values ranging between 4 and 9. Mixed ancestry (Ldon3 and Ldon4) is shown for isolates LRC-L740, LEM3472 and GE as reported earlier [21], at K = 5, 7 and 8 and the same pattern was observed in the three K-Ldon3/Ldon4 samples at K = 5 and 7 (Fig 3). In contrast, K-Ldon5 samples showed a single ancestral origin. Presence of mixed ancestry signatures in a sample could be reflecting events of genetic exchange as well as polyclonality (mixture of different genotypes). By plotting the alternative allele frequency along the 36 chromosomes of admixed sample GSA047, we found the signatures of genetic exchange, i.e., alternation of stretches of heterozygosity and homozygosity and average allele frequency of 50/50 in heterozygous stretches (see S6 Fig).

Download:

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

Fig 3. Admixture analysis (K between 4 and 9, see S5 Fig); the samples in red and purple correspond to the 8 samples from Kenya (this study); each bar represents a sample, with colors corresponding to ancestry components. Samples in blue indicate the genotypes with mixed ancestry reported by Franssen et al. [21].

https://doi.org/10.1371/journal.pntd.0013144.g003

Finally, a scan of genomic signatures of resistance was undertaken on K-Ldon5, the three K-Ldon3/Ldon4 samples and a series of reference sequences. We first compiled a list of 11 genes reported to be involved in resistance to Antimony (Sb), Amphotericin B (AmB) and Miltefosine (MIL) [46–50] and analyzed SNPs having an effect at the protein level (missense and non-sense mutations). Overall, 46 SNPs were identified in the different genes that were analyzed. Of the 46 SNPs, 16 homozygous mutations related with resistance to Sb (MRPA), AmB (C5D) and MIL (MSL) were shared among the eight Kenyan samples and all Ldon5 reference genomes and LRC-L740 (S7 Fig). LiMT, the transporter of Miltefosine showed mutations in all Kenyan samples, but patterns differed between K-Ldon5 and K-Ldon3/Ldon4 samples: respectively, positions 486 and 183 and positions 248 and 495. In addition, one of the K-Ldon3/Ldon4 samples showed a mutation in AQP1 gene (see S7 Fig).