2 MATERIALS AND METHODS
In total, 24 specimens previously identified as S. collaris andS. chinensis were collected from China (Fig. 1), then sequenced and analyzed. Additional samples from the two species and their congeners, were included and their sequences were retrieved from GenBank (Table 1). Scaphiodontophis annulatus was chosen as the outgroup based on previous work (Pyron et al., 2013).
Total DNA was extracted from 85% alcohol-preserved liver or muscle tissues using M5 HiPer Universal DNA Mini Kit (Mei5 Biotechnology Co., Ltd., Beijing, China) following the manufacturer’s protocols. Two mitochondrial gene fragments, cytochrome b (cyt.b ) and NADH subunit 2 (ND2), and nuclear gene oocyte maturation factor mos (c-mos), were amplified by polymerase chain reaction (PCR) using primers L14910/H16064 (Burbrink, Lawson, & Slowinski, 2000), ND2L49/ND2H50 (Alfaro & Arnold, 2001), and S77/S78 (Lawson, Slowinski, Crother, & Burbrink, 2005). Cycling parameters were identical to those described in the above studies. Prior to sequencing, PCR products were purified using various commercial kits. The double-stranded product was sequenced by a commercial company (GENEWIZ Company, Suzhou, China).
The sequences were manually edited using SeqMan in Lasergene v15.1 (DNASTAR Inc., USA), aligned using Muscle with default settings, and quality-checked using MEGA v7.0 (Kumar, Nei, Dudley, & Tamura, 2008; Tamura, Stecher, Peterson, Filipski, & Kumar, 2013). Phylogenetic analyses were performed based on the two mitochondrial DNA (mtDNA) fragments using Bayesian inference (BI) and maximum-likelihood (ML). BI was executed in MrBayes v3.2.2 (Ronquist et al., 2012) using evolutionary models selected in PartitionFinder v2.1.1 under Bayesian information criterion (BIC) (Lanfear, Calcott, Ho, & Guindon, 2012). All searches were performed with three independent runs, each initiating a random tree. Each run consisted of four Markov chains (three heated and one cold chain), with 2 × 106 generations, sampling every 1 000 generations, and 25% of initial samples discarded as burn-in. Convergence was assessed by examining effective sample size (ESS) (> 200) and likelihood plots through time in Tracer v1.7 (Rambaut, Drummond, Xie, Baele, & Suchard, 2018). The resulting trees were combined to calculate posterior probabilities (PP) for each node in a 50% majority-rule consensus tree. The ML trees were constructed using RAxML v7.2.6 (Stamatakis, 2006) with the GTMMAGTRCAT model under the same partitioning scheme as BI analysis. Branch support was assessed using 1 000 non-parametric bootstrap (BS) topological replicates.
We constructed a haplotype network to depict interspecific/clade relationships based on c-mos sequences. Analyses were executed using PopART v1.7 (Leigh & Bryant, 2015) with the parameter epsilon set to 0.
Pairwise genetic distances (p -distances) between species or mtDNA clades were also calculated using Mega v7.0 (Kumar et al., 2008; Tamura et al., 2013).