Recently, the Vallarta Mud Turtle (Kinosternon vogti) was described based on morphological characteristics (López-Luna et al. 2018). The Vallarta Mud Turtle showed noticeable differences from other turtle species in the Pacific region, especially in the proportions of plastron and carapace scutes, body size, and the presence of a large yellow rostral shield in males (López-Luna et al. 2018). The suite of morphological traits exhibited by K. vogti clearly differentiates it from other species within the genus Kinosternon. However, as Uetz et al. (2019) commented, “The diagnosis of this species is a bit dubious as the measurements are not summarized graphically. There is also no (phylo-) genetic analysis of relationships to other members of the genus.” Therefore, the question remains as to whether DNA data support or confirm this proposed new species within the genus Kinosternon. Herein, we use DNA data and 3 statistical methods to evaluate whether the morphological differentiation of the Vallarta Mud Turtle K. vogti corresponds to genetic differentiation and supports the existence of a new operational taxonomic unit (OTU), likely at the species level.

METHODS

Sampling and Laboratory Methods. — We examined the 2 paratypes (CUC-REPTILARIO KC1 and CUC-REPTILARIO KC2) described by López-Luna et al. (2018), which were collected in July 2012 near Río Pitillal in Puerto Vallarta, Jalisco, Mexico. The turtles were kept in outdoor enclosures at the Reptilario Cipactli (under the permit SEMARNAT-UMA-EA-0035-JAL). Genomic DNA was isolated from tissue samples using the Qiagen DNeasy Blood and Tissue Extraction Kit following the manufacturer's protocols. Two mitochondrial genes were amplified: cytochrome oxidase I (COI) and cytochrome b (Cyt b). COI was amplified using the Meyer (2003) primer pair dgLCO1490 (5′-GGTCAACAAATCATAAAGAYATYGG-3′) and dgHCO219 (5′-TAAACTTCAGGGTGACCAAARAAYCA-3′) and Palumbi et al. (1991) GLUDGE (5′-TGATCTTGAARAAC-CAYCGTTG-3′) and CB3H Palumbi (1996; 5′-GGCAAATAGGAARTATCA-3′) were used to amplify Cyt b. DNA was amplified using the Qiagen Taq DNA Polymerase kit (Qiagen, Valencia, CA). The final concentration of each reagent in the reaction mix for amplification was 1×Taq buffer, 1.5 mM MgCl₂, 0.2 mM each dNTP, 10 µM of each primer, and 0.04 U/µl Taq polymerase. The following thermocycler program was used: initial denaturation at 94°C for 180 sec, denaturation at 94°C for 30 sec, annealing at 42°C for 30 sec, extension at 72°C for 30 sec. This sequence was repeated for 30 cycles, followed by the final extension at 72°C for 7 min. Polymerase chain reaction (PCR) products were electrophoresed in 1.5% low-melting-point agarose gels using a Tris-acetate buffer (pH 7.8) containing 3 µl of GelRed® Nucleic Acid Gel Stain. The PCRs were cleaned with 2 µl EXO-SapIT. Two microliters of purified PCR product was used as a template in a 10-ml cycle sequencing reaction using a BigDye 3.1 terminator cycle sequencing kit (Applied Biosystems, Foster City, CA). Each PCR product was sequenced in both directions using the same primers used for the PCR amplification. After cycle sequencing, samples were run on an ABI 3100 capillary sequencer (Applied Biosystems) following the manufacturer's protocol. Chromatograms were reviewed, assembled, and aligned using Sequencher version 4.5 (Gene Codes, Ann Arbor, MI). DNA sequences were translated into amino acids and reviewed in MacClade version 4.1 (Maddison and Maddison 2005). Kinosternon vogti sequences were deposited in GenBank (see Supplemental Table S1 for accession numbers; all supplemental material is available at https://doi.org/10.2744/CCB-1387.1.s1).

Analytical Methods. — Preliminary data exploration was performed to compare the COI and the Cyt b sequences generated in this study from paratype specimens of K. vogti against the National Center for Biotechnology Information (NCBI) nucleotide database. The NCBI BLAST search yielded 39 COI and 100 Cyt b sequences with a similarity of 88.3% to 95% to our COI and Cyt b sequences (NCBI 2019). A subset of 40 sequences of these high similarity NCBI BLAST sequences, 20 sequences for each gene, were included in subsequent analyses. FASTA files were created for each gene separately and for the concatenated sequences with the 40 NCBI BLAST sequences representing 13 Kinosternon species and 3 Sternotherus species. FASTA files also included the 2 sequences from paratype specimens of K. vogti generated in this study. Combined sequences were aligned using MEGA X (Kumar et al. 2018). Nucleotide composition of the Cyt b and COI genes was examined with MEGA X.

Phenetic analysis was performed by the Neighbor-Joining method (Saitou and Nei 1987) with uncorrected p-distance as implemented in MEGA X (Kumar et al. 2018). A phylogeny was inferred using the maximum likelihood (ML) criterion (Felsenstein 1981, 2004) using Randomized Axelerated Maximum Likelihood (RAxML; Stamatakis 2014; Kozlov et al. 2019) as well as a Bayesian Markov chain Monte Carlo (MCMC) approach (Rannala and Yang 1996) using MrBayes version 3.2.7a (Ronquist and Huelsenbeck 2003; Ronquist et al. 2012). For the Bayesian analysis, we ran parallel MCMCs with 4 metropolis-coupled chains each for 5,000,000 generations, sampling trees every 1,000 generations. Sampled trees were used to construct a 50% majority rule consensus tree, in which marginal posterior probabilities of each clade were estimated from the clade's proportional representation among the post burn-in samples. We considered a posterior probability of ≥ 95% as significant support for a given clade. For the ML and Bayesian analyses, we partitioned the data by gene and selected the best-fit models of DNA sequence evolution for each data partition using jModeltest version 2.1.3 (Guindon and Gascuel 2003; Darriba et al. 2012). Clade support was also evaluated using rapid 200 bootstrap heuristics in RAxML. Congruence among topologies was evaluated by the Shimodaira-Hasegawa test in PAUP 4 (Swofford 2002). All topologies were generated using Midpoint rooted trees.

Species definition and species delimitation remain controversial. Separated species may be difficult to distinguish by using only one set of data or one type of analysis. To corroborate the taxonomic level of K. vogti obtained by morphological and phylogenetic analyzes, we used an operational approach to species classification, which classifies groups of closely related individuals as species based on similarities (Doyen and Slobodchikoff 1974). We delimited the operative taxonomic units (OTUs) using 3 methods: the CD-HIT-EST DNA clustering algorithms (Li and Godzik 2006), the Automatic Barcoding Gap Discovery (ABGD; Puillandre et al. 2012), and Bayesian Poisson Tree Processes (bPTP; Zhang et al. 2013). We ran CD-HIT-EST on the CD-HIT online server (http://weizhongli-lab.org/cd-hit/servers.php) using the default parameters. ABGD was run on the ABGD web server (https://bioinfo.mnhn.fr/abi/public/abgd/abgdweb.html) with a minimum distance (pmin) of 0.0025 and maximum distance (pmax) of 0.194, X (relative gap width) set to 0.9, Nb bins set to 23, and simple distance. After multiple trials, we got the X value corresponding to the best sensitivity of the method to gap width and the Nb value, which is the number of groups possible, conserving the distance between sequences from different groups larger than the gap distance. Using default parameters, the program bPTP was implemented at https://species.h-its.org/.

RESULTS

From the 2 paratype specimens, PCR amplification and sequencing produced sequence lengths of 590 base pairs (bp) for COI and 647 bp for Cyt b. Most variable and informative sites were found in the third codon position. Base frequencies were homogeneous across taxa (χ²₉₉ = 29.39; p = 1.00). No sequences exhibited premature stop codons when translated into amino acid sequences. The analysis of character status was done in MEGA X (Table S2).

BLAST Analysis. — With mismatch alignment fraction between any pair of sequences of 0.15, we obtained 22 sequences of each gene (20 sequences retrieved from GenBank and the 2 sequences generated in this study; Table S1) with genetic alignment scores above 90%. From those sequences, we generated a concatenated aligned FASTA file in MEGA X for subsequent analyses.

Distance Estimation. — Sequences for COI and Cyt b genes from all 23 samples were concatenated and evolutionary distances (in number of base differences per site) were computed in MEGA X (Kumar et al. 2018) using the p-distance method (Nei and Kumar 2000). First, second, third, and noncoding codon positions were included and all positions containing gaps and missing data were eliminated (complete deletion option).

The mean (± standard error) genetic p-divergence over all sequence pairs was 0.083 ± 0.00 substitutions per site, obtained by a bootstrap procedure of 1000 replicates (Kumar et al. 2018; Table S3). The number of base differences per site from between sequences ranged from 0.006 to 0.124. The average number of base differences per site between K. vogti and all other sequences included in the analysis was 0.083 ± 0.009. The clades closest to K. vogti were K. hirtipes (p-distance 0.057 ± 0.00) and K. scorpioides (p-distance 0.058 ± 0.00; Table S3).

Phylogenetic Results. — The best models of evolution were HKY+I+G for COI and TN93+G for Cyt b. Our three phylogenetic analyses produced similar topologies. The ML and Bayesian trees (Fig. 1) were identical and differed from the neighbor-joining tree on the placement of K. alamosae and the relationship between K. stejnegeri (for taxonomic status see Turtle Taxonomy Working Group 2017) and the K. durangoense and K. chimalhuaca, K. hirtipes, K. integrum, K. scorpioides, K. vogti clades. The Shimodaira-Hasegawa test showed no significant differences among the ML and Bayesian consensus trees (p > 0.05; Fig. 1). In general, tree topologies from RAxML and Bayesian reconstruction methods were well-resolved, with the majority (80%) of nodes having high support (bootstrap support > 70% in RAxML and posterior probability > 95%). Kinosternon vogti was most closely related to the clade formed by K. hirtipes and K. chimalhuaca (bootstrap support = 89% and posterior probability = 1).

View Figure

• Download as Powerpoint
• Download Figure

Midpoint rooted topologies showed 3 major well-supported clades. One clade included K. acutum, K. dunni, K. angustipons, and K. leucostomum (referred to as Cryptochelys by Iverson et al. 2013); a second clade included all remaining representatives of the genus Kinosternon (including K. vogti); and a third clade included all 3 representatives of the genus Sternotherus.

OTU delimitation using CD-HIT-EST, ABGD, and bPTP recognized 13, 19, and 21 OTUs, respectively (Fig. 1). CD-HIT-EST clustered together the following 4 pairs of species: K. subrubrum and K. flavescens, K. stejnegeri and K. durangoense, K. hirtipes and K. chimalhuaca, and K. integrum and K. scorpioides. The American Board of Genetic Counseling (ABGC) recognized all taxa included in the analysis as separate OTUs, except for K. integrum 1, K. integrum 2, and K. integrum 3, which were grouped into a single OTU. Although bPTP recognized all taxa included in the analysis as valid OTUs (21 clusters), posterior probabilities were found significantly lower than Bayesian posterior probabilities, ranging from 0.271 to 0.715, except for K. alamosae, which had a support value of 0.977.

DISCUSSION

All our results confirmed the status of K. vogti as a distinct OTU. The most striking result from our analysis was the phylogenetic position of K. vogti. Morphologically, K. vogti is most similar to K. angustipons, K. dunni, and K. herrerai (the last of which was not included in this analysis), based on a narrow and weak kinetic posterior plastron and a short bridge. Available life-history information indicates that kinosternid species with a small plastron are primarily aquatic (Bramble et al. 1984); thus, the morphological similarity among K. vogti, K. angustipons, K. dunni, and K. herrerai indicates that they have convergently evolved, possibly in response to environmental pressures that promote similarity in species even though they do not exhibit a direct phylogenetic relationship (Iverson 1991; Iverson et al. 2013; Spinks et al. 2014; present study). Kinosternon vogti formed a monophyletic group with K. hirtipes and K. chimalhuaca, which at the same time were recognized all as separate OTUs by CD-HIT-EST, ABGC, and bPTP analyzes. On the other hand, the description of a new species of mud turtle, K. cora, in the lowlands of the Western Pacific coast (Webb 2001; Loc-Barragán et al. 2020) and our results could suggest that an ancestor between K. chimalhuaca and K. vogti (most likely related to K. hirtipes) colonized the lowlands along the Pacific coast from the Central Mexican plateau independently of K. integrum, which presumably invaded the Pacific lowlands earlier than the ancestor of K. vogti and K. chimalhuaca (J. Iverson, pers. comm., 2020). Although we recovered the 3 major clades of phylogeny proposed by Iverson et al. (2013; Cryptochelys, Sternotherus, and Kinosternon), the placement of K. leucostomum and K. acutum, S. carinatus, S. depressus, and S. odoratus, and K. alamosae was incongruent with Iverson et al. (2013). Moreover, our results strongly differed from Spinks et al. (2014). We attribute these inconsistencies to reduced taxon sampling and a lack of highly variable nuclear molecular markers in our study. The incongruities between our phylogeny and those presented by Iverson et al. (2013) and Spinks et al. (2014) do not obscure the evidence obtained in our study, which supports the hypothesis that Kinosternon vogti is an OTU and likely a distinct evolutionary lineage.

Future work using more exhaustive molecular sampling in combination with morphological data would improve our understanding of the evolutionary history and environmental variations within kinosternid turtles. Furthermore, information about the effects of population demography and population genetic structure in kinosternid species remains somewhat opaque (Macip-Ríos et al. 2011); therefore, future work on this topic deserves urgent attention for management and conservation purposes. The Vallarta Mud Turtle (K. vogti) was recently proposed to be Critically Endangered by the International Union for Conservation of Nature (Rhodin et al. 2018). Based on the criteria used by the Mexican government to determine species at risk (Sánchez et al. 2007), K. vogti was recently included in the Mexican Official Norm as Endangered (NOM-059-SEMARNAT; https://www.dof.gob.mx/nota_detalle.php?codigo=5578808&fecha=14/11/2019). In addition, the demographic and life history information of K. vogti is not only important as reference information to assess population dynamics, but could provide, along with additional genetic information, a baseline to design a conservation program for this species.