All dogs were enrolled with informed consent from their caretakers, and protocols followed were approved by the University of Iowa Institutional Animal Care and Use Committee. This AAALAC International–accredited institution follows the requirements for the US National Institutes of Health Office of Laboratory Animal Welfare Assurances and operates under the 2015 reprint of the Public Health service Policy on Humane Care and Use of Laboratory Animals.

The 7 L. infantum samples from US hunting dogs used in this study were identified during a retrospective cohort study of L. infantum infection in US hunting dogs (26,27,30). To identify Leishmania-infected dogs, an active surveillance cohort of 4 large (>50 dogs each) kennels was established from 3 different states in the midwestern United States during 2007–2017. Licensed veterinarians collected 1–5 mL whole blood and serum samples from all dogs at these kennels. Dogs were considered infected if they were positive by quantitative PCR detecting Leishmania-specific DNA and had Leishmania-specific antibodies (31). Parasites from the buffy coat of Leishmania-positive dogs were cultured in both Schneider and HOMEM media overnight at 26°C then placed onto agar slants and incubated for 3–4 weeks and observed daily for growth. Parasite cultures include 1 sibling pair (foxymo_01, foxymo_02); remaining dogs all have different grandparents. Because of the frequent exchange of hunting dogs among kennels and states, within 2 generations the ancestors of the sampled dogs came from 12 kennels and 9 different US states (Georgia, Illinois, Iowa, Kansas, Minnesota, Missouri, New Jersey, New York, and Virginia) that included the primary US locations for hunting hound breeding.

We used QIAamp DNA Blood Mini Kit (QIAGEN, https://www.giagen.com) according to manufacturer specifications to isolate DNA directly from primary parasite cultures. We thawed parasite cultures, counted, and placed 1 million parasites into Trizol Reagent (ThermoFisher Scientific, https://www.thermofisher.com) and extracted according to manufacturer specifications. We assessed quality and quantity of isolated DNA by using NanoDrop 2000 (ThermoFisher Scientific).

We sheared DNA into 400–600-bp fragments by focused ultrasonication using the Covaris Adaptive Focused Acoustics technology (Covaris, https://www.covaris.com). We performed 2 methods of DNA sequencing, depending on the amount of DNA supplied, by using the NEBNext DNA Library Prep kit (New England BioLabs, https://www.neb.com). For volumes <500 ng, we amplified libraries by using KAPA HiFI DNA polymerase (Kapa Biosystems, https://kapabiosystems.com) and generated 100-bp paired-end reads on the Illumina HiSeq 2000 (Illumina, https://www.illumina.com). For volumes >500 ng, we generated amplification-free libraries and obtained 150-bp paired-end reads on the Illumina HiSeq X10 (Illumina). We performed sequencing following manufacturers’ standard protocols.

We analyzed the genomic data of 7 L. infantum US hound isolates with an additional 92 publicly available L. infantum isolates sampled from a global distribution (Appendix 1). For all samples, we subjected newly generated and downloaded fastq files to identical analysis pipelines. We trimmed reads using Trimmomatic version 0.39 (http://www.usadellab.org/cms/?page=trimmomatic) (parameters “ILLUMINACLIP:PE_adaptors.fa:2:30:10 TRAILING:15 SLIDINGWINDOW:4:15 MINLEN:50”) and mapped them against the reference genome of JPCM5 v45 (https://tritrypdb.org) with BWA version 0.7.17 (bwa mem -M option) (32). Single-nucleotide polymorphisms (SNPs) were called using GATK version 4.1.2.0 (33): HaplotypeCaller was used with parameters “-ERC GVCF–annotate-with-num-discovered-alleles–sample-ploidy 2” to generate gvcf files for each sample, then combined using “GenomicsDBImport” and genotyped with “GenotypeGVCFs.” Calls were filtered with “VariantFiltration” (filters: “QD<2.0, MQ<50.0, FS>20.0, SOR>2.5, BaseQRankSum<-3.1, ClippingRankSum<-3.1, MQRankSum<-3.1, ReadPosRankSum<-3.1and DP<6”) and only polymorphic SNPs retained. We removed SNPs with >20% missing calls across samples, reducing the total number of SNPs from 43,528 to 43,336.

We performed phylogenetic reconstruction by using distance-based and maximum-likelihood methods on genome-wide genotype calls. For the distance-based approach, we calculated pairwise Nei D distances and reconstructed trees by the neighbor-joining method using the R packages StAMPP version 1.6.1 (34) and ape version 5.4. We based bootstrap values on 100 replicates. For maximum-likelihood phylogenies, we converted the vcf file to fasta format with IUPAC codes using bcftools consensus. We estimated 1,000-bootstrap maximum-likelihood phylogenies by using RAxML-NG version 0.8.1-c1 (35) and the GTJC model that captures changes between heterozygous and homozygous states.

We preprocessed genome-wide SNPs for admixture analysis version 1.3.0 (36) only with plink version 1.90 changing the vcf format into ped and map format and removing SNPs with a missing fraction of >0.05 and variants closer to each other than 2,000-bp with the arguments “–geno” and “–bp-space.” We ran admixture for values of K from 1 to 20 and optimal numbers of groups (K) were chosen on the basis of lowest cross-validation error (Appendix 2 Figure 1). Because there was no clear number of K at which the cross-validation error plateaued, we present analyses with the smallest K at first sign of plateauing of the error and 2 larger Ks with smaller errors.

We used 2 molecular clock approaches. The first method was a simple clock model using PATHd8 (37) for all RAxML-NG bootstrap trees, constraining the root of the non-US New World clade to 537 years ago. The second method was a Bayesian approach that used BEAST version 1.10.4 (https://beast.community) to enable flexible modeling of rate variation with standard substitution models, a narrow uniform prior of 536.9–537.1 years for the New World clade and leaf heights set to the year of collection (Appendix 1), or constrained to 2005–2007 for samples from (39) and to 1900–2020 for the sample ‘DOG_STRAIN’ of unknown sampling date (38). New World and US hound clades were constrained to be monophyletic, and Bayesian Markov Chain Monte Carlo analysis was initialized with the RAxML-NG phylogeny for concatenated chromosomes. The substitution model was Hasegawa-Kishino-Yano with a 4-category gamma distribution of rate variation across sites. Results are based on 8 independent Bayesian Markov Chain Monte Carlo chains of 10 million generations, 1 million generations burn-in, and convergence checked using Tracer version 1.7.1 (https://beast.community/tracer). We accepted analyses if 6 out of 8 chains were at similar likelihoods for 2 million generations. Remaining parameters were defaults from Beauti version 1.10.4. Only results for both strict and uncorrelated gamma-distributed clocks converged and are shown.

We grouped parasite samples according to geographic origin and isolated host type (Table). Groups were characterized by their number of segregating SNPs, inbreeding coefficients, and linkage decay with distance. We performed analysis in R (R Foundation for Statistical Computing, https://www.r-project.org) with the exception of R² estimates, which we estimated as genotype correlations with vcftools version 0.1.16 (41) and parameters “–geno-r2” and “–interchrom-geno-r2.” We used genotype correlations because haplotypes cannot be accurately phased for our small population sets. We calculated the inbreeding coefficient F based on the formula F = 1 –((c_AB/N)/(2 × f_A × f_B)), where c_AB represents the heterozygote count, N the group size, and f_A and f_B the frequency of alleles A and B.

We estimated sequencing coverage on the basis of sample-specific mapped bam files. For each sample, indels were determined and indel realignment was performed with the GATK version 3.6 (33) tools “RealignerTargetCreator” and “IndelRealigner.” Quality filtering and duplicate removal was done with samtools version 1.3 using the parameters “-F 1024 -f 0x0002 -F 0x0004 -F 0x0008.” Coverage was estimated with bedtools version 2.17.0 (42) genomecov and parameters “-d -split.” For each sample, the median coverage per chromosome was assumed to represent the diploid state, so chromosome somy = (chromosome_coverage/median_coverage) × 2. Allele frequencies for isolate-specific SNPs were estimated on the basis of previous bam files and quality filtered with samtools “-q 20 -f 0x0002 -F 0x0004 -F 0x0008.” Coverage by genomic position was obtained with samtools mpileup “-d 3500 -B -Q 10” and transformed into sync format with mpileup2sync “–min-qual 20” (43).

To assess the geographic origins of L. infantum parasites within US hunting dogs, we generated whole-genome sequence data for 7 L. infantum isolates from outbred hounds from 4 kennels in the midwestern United States and an ancestry tracing back to kennels in 9 US states within 2 generations with haploid coverage ranging from 29 to 78 (median 69). We compared these samples with 92 previously published L. infantum genome sequences of other strains from other global populations (38,39,44) (Appendix 1).

We constructed distance-based and ML phylogenies from whole-genome SNP variants to compare L. infantum genomes from US dogs to samples from L. infantum–endemic regions of South America and the Old World. Parasites from US hounds were monophyletic, part of the L. infantum MON-1 clade (38,45), and clearly distinct from L. infantum isolates from South America (Figure 1; Appendix 2 Figure 2). These factors suggest independent introduction to the New World. The genetically closest parasite samples were from southern Europe, but the exact origin was ambiguous. Distance-based methods suggested 4 samples from France as genetically most closely related to US isolates (Figure 1; Figure 2, panel A). The ML phylogeny placed US parasites close to a more widespread group of MON-1 parasites (Figure 2, panel B).

To further investigate parasites’ relatedness, we performed admixture analysis, which was consistent with the phylogenetic results. We applied cross validation, a standard approach in admixture to determine an optimal number of populations (K) that best explains the relatedness between samples. Because this process did not identify a single optimal K (Appendix 2 Figure 1), we considered more than one K (Figure 1; Appendix 2 Figure 2). We concentrated our analysis on 83 core samples consisting of samples from the United States and other samples from the MON-1 clade (Figures 1, 2). For K = 4 populations, US hound parasites were placed together with all remaining samples from Europe and single samples from Israel and Morocco (Figure 1). For K = 6 and K = 15, US samples formed a separate group, only inferred to share ancestry with one sample from Italy and one from Morocco for K = 6. A similar pattern was present within the total set of 99 samples. For K = 7, US and 2 parasites from France grouped together, and for K = 11, US samples only shared substantial variation with 1 sample from Italy (Appendix 2 Figure 2, panel A), which together suggested a clear origin from Mediterranean Europe but no clear country of origin.

We dated the independent introduction of US hound parasites by using 2 different molecular clock approaches, relying on previously estimated introduction of L. infantum parasites into the New World ≈500 years ago (46). The first analysis using our maximum-likelihood phylogeny estimated the mean date of divergence between US parasites and relatives from Europe as 1897 (95% CI 1873–1917), whereas 2 Bayesian approaches produced estimates of 1938 (strict clock, 95% highest posterior density CI 1910–1965) and 1889 (relaxed clock, 95% CI 1689–1991) (Figure 3). Estimates across a range of approaches thus suggest that US hound parasites were introduced much more recently than L. infantum parasites were introduced to South America.

The genetic variation in a population should reflect its reproductive biology. We thus compared variation in US hound parasites with L. infantum populations isolated from dogs in areas where vector transmission occurs and with populations isolated from humans or a mixture of both hosts in other parts of the world (Table). Within-population diversity of the US hound parasites was intermediate between the high diversity of populations from the Old World and the low diversity of parasites from different regions within Brazil (Figure 4). For most populations, the number of polymorphic sites increased with sample size, indicating that increasing numbers of rare variants were detected with larger sample sizes. This sample size–based increase was minimal in the US hound parasite population, suggesting a large proportion of shared variation among these isolates.

To explore this shared variation further, we directly estimated population heterozygosity through the inbreeding coefficient F and the fraction of population-specific polymorphic heterozygous SNP sites (Figure 5; Appendix 2 Figure 3). The inbreeding coefficient was significantly different between populations (Kruskal-Wallis test, χ² = 2843.1, df = 12; p<0.001), and the US hound parasite population had exceptionally low F values compared with all other populations (Dunn test, adjusted; p<0.001) (Figure 5). This difference was largely caused by 79% of all polymorphic sites within US hound–derived parasites sharing the same heterozygous genotype across all 7 sampled hound isolates. This extreme excess of shared heterozygosity is present across all chromosomes and is in strong contrast to the remaining populations. Absolute numbers of heterozygous sites in the US samples were higher than in other populations (Table; Appendix 2 Figure 4, panel A). This difference could be caused by either the accumulation of mutations during a period of clonal evolution shared by these samples or a hybrid origin of the founder strain of our US samples between 2 closely related L. infantum populations (Appendix 2 Figure 4, panel B), because clonal propagation would maintain any heterozygosity.

If L. infantum parasite transmission in US hunting dogs occurs solely through vertical transmission, we would expect genomic signatures of sexual reproduction to be absent because sexual reproduction is thought to be limited to the vector stage (18). Sexual reproduction returns proportions of heterozygous and homozygous variants to the Hardy-Weinberg equilibrium. We propose that the observed extreme excess of shared heterozygous sites in US hound parasites is possible because these parasites evolve clonally for many generations with no mechanism to reduce the number of heterozygous sites through sexual reproduction. To test this proposition, we investigated whether genetic linkage between pairs of SNPs reduces as the distance between loci increases, which would be expected if recombination is occurring. Almost all global L. infantum populations showed this expected decay in linkage within chromosomes, except US hound–derived parasites and 2 populations from Brazil (Figure 6). The 2 populations from Brazil had too few polymorphic sites to reliably assess linkage patterns. The US hound parasites also had relatively few sites for analysis, because unphased shared heterozygous sites cannot be used for linkage estimation. However, the remaining loci showed no evidence of linkage decay with genetic distance. Pairs of variants on different chromosomes showed very similar linkage to within-chromosome comparisons (Figure 6). This finding indicates that evidence for meiotic recombination in the US dog L. infantum population is lacking.

Leishmania populations frequently show variation in copy number of individual chromosomes with frequent aneuploidy turnover even within a clonal population (mosaic aneuploidy). Aneuploidy variation between US isolates was largely limited to one third of the chromosomes and variation did not correlate to chromosome-specific heterozygosity, which should have been reduced if aneuploidy turnover was high (Figure 7; Appendix 2 Figure 5). Although this estimate of aneuploidy variation through mean ploidy profiles between isolates is conservative, it supports initial findings that aneuploidy turnover might be greater in cultured promastigotes versus intra-host amastigotes (47,48).