The Himalayan uplift and evolution of aquatic biodiversity across Asia: Snowtrout (Cyprininae: Schizothorax) as a test case

markedly, depending upon which ‘sub-genome’ is being accessed when polyploid data are phylogenetically evaluated [34]. Given this, phylogenetic analyses of polyploid species are often limited in practice to either mtDNA (given its haploid nature), or nuclear genes cloned into bacteria and subsequently iso- lated as single-copy (such as RAG-1) [20]. However, the latter are evolutionarily more conser- vative and, for example, cannot be utilized in a TIMETREE analysis [35] as herein. Yet polyploidization in Schizothorax also represents an evolutionary mechanism that may have facilitated its invasion of high-elevation habitat [31]. We thus restricted our study to mtDNA haplotypes and relied upon GENBANK to increase our sample number and geographic representation. Available numbers are often limited, due either to restricted sampling and/or logistic difficulties in accessing those regions where study species occurs. In this context, Schizothorax inhabits regions notoriously difficult to access, both topographically as well as socio-politically. In addition, GENBANK samples often represent output from ongoing investigations and their metadata thus lack precise GPS coordinates required for specific analyses. This, as well as the large variance in number of base pairs avail- able per sample, has constrained our total GENBANK sample number. Many of the above issue can also confound analyses and, to avoid such issues, we rigorously assessed GENBANK meta- data prior to incorporating sequences. Analyses to infer evolutionary history

We employed a four-step approach to infer potential evolutionary scenarios that could explain the biodiversity in Schizothorax across the Himalaya. This involved: (a) Evaluating congruence between taxonomic and genetic designation of samples; (b) Detecting cryptic biodiversity; (c) Characterizing regional and basin-specific genetic lineages; and (d) Quantifying phylogeo- graphic patterns. We employed a Bayesian clustering analysis to identify regional gene pools without a priori assumptions pertaining to location and/or taxonomic classification. We then used these groupings to statistically evaluate genetic variation within and among regions and basins (AMOVA). We generated phylogenetic hypotheses by deriving ML and Bayesian tree topologies, which were statistically and qualitatively compared for congruence (IQ-TREE). We also derived a haplotype network as a means of visualizing the number of stepwise mutations separating distinct sequences and groups. Fossil-calibrated methods were then employed to estimate divergence times for major groupings in the context of regional geologic events. After compiling a dataset of 132 Schizothorax samples using the criteria described above, we derived a dataset that included only variable sites (N = 365bp: DNASP v.6; [36]). We employed a Bayesian approach (GENELAND V.4.0.3: [37, 38]) to infer the number of distinct groups (K) and to designate group membership for each sample. Ten runs of 100,000 MCMC iterations and a thinning interval of 100 were used, with K varying between 1 and 15, but with-

out use of optional geo-referencing, given the absence of such coordinates for most GENBANK sequences. The estimated number of populations (K) was defined using the maximum poste- rior estimate. We conducted an analysis of molecular variance (AMOVA: ARLEQUIN v.3.5.1.2; [39]) to gauge how genetic variation was partitioned statistically among geographic regions (as informed by GENELAND). Here, we defined populations (= distinct gene pools) as being haplo- types within regions [40]. Probabilities were ascertained using 1,000 permutations, and statisti- cal significance was assigned using a Bonferroni-correction. We derived phylogenetic hypotheses using a maximum likelihood (ML) analysis (IQ-TREE v2; [41]), with the best-fitting substitution model being generated via MODELFINDER, selecting flexible rate heterogeneity (e.g., gamma-distributed) across sites [42]. Model space was hierar- chically explored by sequentially adding parameters (e.g., rate categories) until the model failed PLOS ONE Himalayan uplift and aquatic evolution PLOS ONE | https://doi.org/10.1371/journal.pone.0289736 October 24, 2023 8 / 27