Fig. 5: Metric multidimensional scaling, simulated mating, and tree-to-graph conversion suggest admixture and hybridization between Del and NonDel L. infantum groups.

a Metric multidimensional scaling separates New and Old World (NW and OW) isolates on two axes of variation (goodness-of-fit = 0.40). NonDel isolates from Mato Grosso do Sul (MS) and Del isolates from Rio Grande do Norte (RN, see asterisk) and Mato Grosso (MT, see double-asterisk) position at opposite ends of axis 1, the primary axis of divergence within and between NW populations. HTZ isolates occur at intermediate positions (see pink circles) between these dissimilar groups. Other isolates with such intermediate positions are labeled and may also represent mating events between dissimilar groups. Gray, white, and cyan fill colors, respectively, indicate NonDel, Del, and MIX read-depth profiles found in the NW. Circles for OW (NonDel) isolates are green. Five outlier isolates are excluded as in Fig. 4. b Neighbor-joining positions of simulated hybrids (blue font, left tree) correspond to those of observed HTZ isolates (pink font, right tree) from MT. Hybrids were simulated in two steps. Random 50% haplotype contributions were first drawn from Del and NonDel isolates observed in MT and MS. The resultant offspring genotypes were then either let diversify through random mutation or subjected to a second round of Mendelian recombination as before. The same tree topology resulted in each of 100 simulation replicates. Trees are midpoint-rooted as opposed to outgroup-rooted as in Fig. 4. c Given that mating can create non-treelike divergence patterns within species, TreeMix⁶⁶ was used to search iteratively for up to five migration edges that improve the fit of a maximum-likelihood tree built based on Gaussian approximation of genetic drift among isolates from MT, MS, RN, and OW groups. This input tree (black edges) suggests dichotomous differentiation into MT/RN and MS/OW clades and has a log-likelihood of 84.9206. Tree-to-graph conversion by addition of a migration edge from MS to MT increases log-likelihood to 84.9775. No other edges further increase the fit of the input tree. A four-population test⁷⁶ also supports post-split admixture between MS and MT or RN, because differences in allele frequencies between MT and RN isolates correlate with those within the other population pair (F₄-statistic = 5 × 10⁻⁵, Z-score = 3.51).