Comprehensive Analysis of Salamander Hybridization Suggests a Consistent Relationship between Genetic Distance and Reproductive Isolation across Tetrapods

The first goal was the establishment of a comprehensive list of published salamander hybrids, which did not exist when we began this research (Table 1). Taxonomy is continually revised to provide scientific names that convey accurate information about species boundaries as well as the evolutionary relationships among species. Making these species designations is a rich discipline, with disagreements among taxonomists requiring different levels of divergence or isolation between populations before they are formally named as species (Highton, 1998; Kuchta and Wake, 2016). For our literature review, we used a general lineage concept of species where a species equates to a segment of a population-level evolutionary lineage (de Queiroz, 1998). We chose to include populations far enough along their own evolutionary trajectories that when secondary contact occurred, the populations did not completely admix into a single population. By doing this, we were able to include several instances of hybridization in which the hybridizing populations have not been formally named as species. We included cases where secondary contact occurred because of human-mediated introductions, and we also included laboratory crosses. Hybridization that resulted in polyploidy was excluded to eliminate the confounding variable of increased ploidy levels. Species with evidence of historical, but not ongoing, hybridization were also excluded. This compilation of hybridizing species represents extant salamanders that are known to have had the opportunity to hybridize and done so successfully; there are undoubtedly pairs of lineages that would be able to hybridize, given the opportunity (either with or without human mediation), as well as undetected instances of natural hybridization. The criteria we chose allowed us to compile the maximum amount of information about hybrid salamanders from the literature, as well as make comparisons with existing literature surveys on birds, mammals, and lizards that applied similar criteria (Fitzpatrick, 2004; Jančúchová-Lásková et al., 2015).

Database searches were performed using the terms salamander, newt, hybrid, and contact zone. Databases used were JSTOR, Web of Science, and Wildlife and Ecology Studies Worldwide. Using the same terms, additional sources were found with the search engine Google Scholar. The searches were undertaken from January 2018 to February 2020 and included research published between 1979–2020. Many taxonomic changes occurred during this 40-year period; whenever there was a conflict in species name, the current listing on AmphibiaWeb (https://amphibiaweb.org) was used to resolve the issue.

To check whether or not the reported number of hybrids within each salamander family was a function of the intensity of publication, a general review was also conducted on how well each family of salamanders is represented in the literature (Fig. 1). Using the Web of Science database, searches were performed using terms based on variations on the salamander family names with two to three search terms used per family. Terms used were: Ambystomatidae (ambystomatid and Ambystoma), Amphiumidae (amphiumid and Amphiuma), Cryptobranchidae (cryptobranchid and Cryptobranchus), Dicamptodontidae (dicamptodontid and Dicamptodon), Hynobiidae (hynobiid and Hynobius), Plethodontidae (plethodontid and Plethodon), Proteidae (proteid), Rhyacotritonidae (rhyacotritonid and Rhyacotriton), Salamandridae (salamandrid and Salamandra), and Sirenidae (sirenid). The genus names Proteus and Siren were excluded due to their use in Greek mythology; because the corresponding salamander families are small, missing papers is unlikely. The number of articles per family was standardized by dividing by the number of species in each family based on AmphibiaWeb as of July 2020 (https://amphibiaweb.org). To test whether the reported number of hybrids within each salamander family was related to the intensity of publication, a Kendall correlation coefficient was calculated comparing papers per species and hybrids per species for each family.

The genetic distances across the parental species for each hybridizing salamander pair were estimated using the mitochondrial gene cytochrome b (MT-CYB; Fig. 2A). MT-CYB has long been used as a phylogenetic and phylogeographic marker for salamanders, so there is wide coverage for many different species (Johns and Avise, 1998). MT-CYB has also been used as a proxy for overall genetic distance in summaries of hybridizing pairs of species in other vertebrate groups (Fitzpatrick, 2004; Jančúchová-Lásková et al., 2015). For each parental species, the longest high-quality sequence was downloaded from NCBI GenBank (https://www.ncbi.nlm.nih.gov/genbank/). In some cases, this involved extracting the MT-CYB sequence from a complete mitochondrial genome sequence. When multiple equally long, high-quality sequences were available, one was selected at random. Of the 76 salamander species pairs known to hybridize, MT-CYB sequence data were available for 62 (Table 2). For each hybridizing pair, a pairwise alignment was calculated with default ClustalW settings implemented in MEGA X (Kumar et al., 2018). Each alignment was then trimmed to the first and last overlapping nucleotide position. After trimming, the alignments ranged from 345 to 1,141 base pairs. The genetic distances were estimated with PAUP * Version 4.0a (Swofford, 2002) using the HKY85 + Γ nucleotide substitution model (Hasegawa et al., 1985). In order to maximize the amount of sequence data used, the Γ distribution shape parameter α was estimated using the MEGA X maximum likelihood model selector based on a ClustalW alignment of all 81 salamander MT-CYB sequences used in this study (Kumar et al., 2018).

In order to test how representative MT-CYB genetic distances are in describing the overall genetic divergence between species, the substitution rate of mitochondrial genes versus nuclear genes was compared (Fig. 2B). This was accomplished by dividing the genetic distance of the mitochondrial gene MT-CYB by the genetic distance of the nuclear gene recombination activating 1 (RAG1) for the parental species that hybridize (Table 2). RAG1 was selected since the gene has widespread use in phylogenetic studies and was available for the most taxa (Chiari et al., 2009). For every parental species available, the longest high-quality RAG1 sequence was downloaded from GenBank. When multiple equally long, high-quality sequences were available, one was selected at random. RAG1 genetic distances between parental species that hybridize were then estimated using the same methods as for MT-CYB genetic distances.

The MT-CYB genetic distances between hybridizing parental species of salamanders were compared to those of birds, mammals, and lizards (i.e., squamate reptiles excluding snakes). These tetrapod groups were selected because they span part of the range of vertebrate metabolic rates (Pough, 1980; White el al., 2006; Anderson and Gillooly, 2018) and because of the existence of published summaries of known hybrids that include natural, human-introduced, and lab-crossed hybridization events (Fitzpatrick, 2004; Jančúchová-Lásková et al., 2015). A hybrid toad summary was excluded because it focused only on experimental crosses (Blair, 1972; Malone and Fontenot, 2008). In some tetrapod groups, many closely related species hybridize, or a single species hybridizes several times. This creates a potential bias when looking at many pairwise comparisons as a single taxon becomes overrepresented, and the results become dependent on a few select taxa (Fitzpatrick, 2004). Previously published work in mammals and birds resolved this issue by removing repeated taxa to remove nonindependence (Fitzpatrick, 2004), so the list of hybrids for salamanders and lizards was also pared down to make results comparable. For salamanders, the dataset was pared down to a single pair that can hybridize per genus; all genera are monophyletic. When a genus had multiple hybridizing species pairs, the pair with the greatest MT-CYB genetic distance was used. The list of hybrids for lizards was also pared down to the largest MT-CYB genetic distance per genus based on previously published work (Jančúchová-Lásková et al., 2015). MT-CYB genetic distances between the parental species for each hybridizing species pair were calculated for birds, mammals, and lizards using the same methods as for salamanders (Fig. 3, Table 2). For each of the tetrapod groups, a different Γ distribution shape parameter α was estimated using all MT-CYB sequences included in this study (46 birds, 50 mammals, and 40 lizards; Kumar et al., 2018). A Kruskal-Wallis test was then used to determine if the genetic distance across parental species that hybridize differs among the tetrapod groups. Effect size was calculated and reported in terms of η².

In order to test how representative MT-CYB genetic distances are in describing the overall genetic divergence between species, the substitution rate of mitochondrial genes relative to the substitution rate of nuclear genes was compared across the four vertebrate clades (Fig. 4). This was accomplished by dividing the genetic distance of the mitochondrial gene MT-CYB by the genetic distance of the nuclear gene RAG1 for the parental species that hybridize (Table 2). RAG1 genetic distances between the parental species for each hybridizing species pair were calculated for birds, mammals, and lizards using the same methods as for salamanders. A Kruskal-Wallis test was used to determine if the ratio of MT-CYB/RAG1 genetic distances differed among the four tetrapod groups. A Kruskal-Wallis test was also used to determine if RAG1 genetic distance differed among the four tetrapod groups. Effect size was calculated for both tests and reported in terms of η².

A list of the salamanders that have heteromorphic sex chromosomes was generated to see if the presence of heteromorphic sex chromosomes impacted hybridization. Salamanders with heteromorphic sex chromosomes were identified by using two published reviews and a recently published amphibian karyotype database (Hillis and Green, 1990; Evans et al., 2012; Perkins et al., 2019). In the available literature, it is not always apparent whether a salamander species is known to have homomorphic sex chromosomes or if the species was merely never confirmed as having heteromorphic sex chromosomes (Perkins et al., 2019). To further clarify, additional unpublished information on salamanders that are known to have homomorphic sex chromosomes was obtained (S. Sessions, pers. comm., 19 June 2019). MT-CYB genetic distances between parental species with heteromorphic sex chromosomes that hybridize were compared to the genetic distances between parental species with homomorphic sex chromosomes that hybridize using a t-test (Fig. 5). A comparison was also made between the overall percentage of salamanders known to hybridize and the percentage of salamanders with heteromorphic sex chromosomes known to hybridize using a one-sided exact test of goodness-of-fit. The overall results were similar whether the analyses were performed using only the species confirmed to have homomorphic sex chromosomes (based on personal communication) or using the larger dataset that assumed that all salamanders that have not been reported to have heteromorphic sex chromosomes have homomorphic sex chromosomes.

Our literature review found 81 pairs of genetically distinct salamander lineages that hybridize in nature (Table 1). These lineages represent 12.1% of named salamander species and seven out of ten salamander families (Fig. 1). The seven families represented include Plethondontidae (42 species), Salamandridae (29 species), Ambystomatidae (9 species), Hynobiidae (5 species), Cryptobranchidae (2 species), Dicamptodontidae (2 species), and Proteidae (2 species; Fig. 1). Many pairs that hybridize come from Plethondontidae and Salamandridae, consistent with these being the two largest families of salamanders (486 and 125 species, respectively; https://amphibiaweb.org). The third largest salamander family (Hynobiidae with 82 species) contains proportionally fewer hybrids with only three recorded pairs that can hybridize (Fig. 1). The families Amphiumidae, Rhyacotritonidae, and Sirenidae have no recorded cases of hybrids. These families of salamanders contain very few species with the largest of the three, Sirenidae, only containing five species. There is no case of hybridization occurring between species of different genera. Overall, these results are consistent with those of another recently published study (Patton et al., 2020).

We identified a total of 32,499 papers published using the salamander family related terms. Out of the large salamander families, the family with the highest number of papers per species, Ambystomatidae, did not have the highest number of reported hybrids (Fig. 1). However, Ambystoma mexicanum is a contributor to the high papers-per-species count because it is a model lab animal and makes up a quarter (2,694) of the Web of Science results for Ambystomatidae (Fig. 1; Voss et al., 2010). Two of the families with the highest numbers of species have the lowest intensity of publication: Hynobiidae (21 papers per species) and Plethondontidae (13 papers per species). There are proportionally more papers published for families with fewer species: Amphiumidae (270 papers per species), Cryptobranchidae (252 papers per species), Proteidae (249.5 papers per species), Dicamptodontidae (107 papers per species), Sirenidae (79 papers per species), and Rhyacotritonidae (44 papers per species). There was no significant correlation between the number of papers per species and the number of salamanders found to hybridize per family (Kendall's rank correlation P = 0.236). We acknowledge that studies of salamanders address far more biological questions than only those that are relevant to hybrid identification. Further analyses based on refined literature searches that target only the most pertinent disciplines (e.g., phylogeography, systematics) would provide a stronger test of the relationship between research effort and hybrid identification.

For the 62 parental pairs of salamanders that have published MT-CYB sequence data, the mean genetic distance was 0.117 (Table 2). Lissotriton vulgaris X Lissotriton helveticus had the largest genetic distance of 0.360, while Plethodon teyahalee X Plethodon shermani had the smallest distance of 0.003 (Table 2). For the 33 parental pairs of salamanders that have published RAG1 data, the mean genetic distance was 0.00657 (range 0.000683 to 0.0306; Table 2). The genetic distance ratio of the mitochondrial gene MT-CYB over the nuclear gene RAG1 between salamanders that hybridize was highly variable (Fig. 2B). The ratio ranged from 0.6x to 162x with most species pairs having a higher mitochondrial genetic distance (median ratio = 21.1).

After paring down the salamanders and lizards to the pair that can hybridize with the highest genetic distance per genus, the across-tetrapod comparison included 23 pairs of hybrids for birds, 20 pairs for lizards, 25 pairs for mammals, and 18 pairs for salamanders (Table 2). The median MT-CYB genetic distances for the groups were: lizards (0.181), birds (0.167), salamanders (0.160), and mammals (0.152; Fig. 3). There were no significant differences in distributions of genetic distance among the four groups (χ² = 1.5922, P = 0.66, η² = –0.0172). Across all tetrapods, birds had both the lowest and highest genetic distances with Regulus calendula X Regulus satrapa (0.002) and Gallus gallus X Alectura lathami (0.366). Using the pared down list of hybrids, RAG1 sequences were available for 12 pairs of hybrids for birds, 10 pairs for lizards, 10 pairs for mammals, and 13 pairs for salamanders (Table 2). The median RAG1 genetic distances for the four groups were: birds (0.0151), mammals (0.00722), salamanders (0.00651), and lizards (0.00600). There were no significant differences among the four tetrapod groups (χ² = 5.8748, P = 0.12; η² = 0.0701). The median MT-CYB/RAG1 genetic distance ratio for the four groups were: lizards (23.4), salamanders (20.7), mammals (17.6), and birds (10.6; Fig. 4). There were no significant differences among the four tetrapod groups (χ² = 6.0734, P = 0.11; η² = 0.075).

Forty-eight species of salamanders in seven families have been discovered with heteromorphic sex chromosomes (Ambystomatidae, Hynobiidae, Plethodontidae, Proteidae, Salamandridae, Sirenidae, and Proteidae; Table 3). Salamander families are not characterized by a single type of sex chromosome; different members of Salamandridae and Plethodontidae have ZZ/ZW and XX/XY systems. Out of the salamanders with known heteromorphic sex chromosomes, 12/48 are known to hybridize with a different species (Table 3). There is no difference between the percentage of salamanders that hybridize with known heteromorphic sex chromosomes and the known overall salamander hybridization rate (P = 0.99). Also, the genetic distances between parental salamanders that hybridize with known heteromorphic sex chromosomes (median = 0.109) and parental salamanders that hybridize with homomorphic sex chromosomes (median = 0.152) are not significantly different (P = 0.53; Fig. 5). Based on these analyses, there is no indication that the presence of heteromorphic sex chromosomes hinders hybridization in salamanders.

Overall, we found that hybridization is widespread throughout the salamander clade, occurring in both the biggest and smallest families (Fig. 1). The largest four families (Ambystomatidae, Hynobiidae, Plethodontidae, and Salamandridae) account for nearly 96% of all salamander species and account for over 91% of the hybrids (Fig. 1). With respect to geographic location, known hybrids are widespread in the Northern temperate zone, occurring in North America, Europe, and Asia. In contrast, there is only one reported case of salamanders hybridizing in the tropics, occurring between Bolitoglossa franklini and B. lincolni (Table 1). Bolitoglossa is the largest genus of salamanders (134 species), and it occurs exclusively in the tropics. More generally, close to half of all salamander diversity occurs in the tropics, suggesting that the low reported number of tropical hybrids is not a result of low tropical species diversity (https://amphibiaweb.org). If the low reported number of hybrids is an accurate reflection of the underlying biological reality, it may reflect general differences in mechanisms underlying the origin and maintenance of species in temperate versus tropical areas worthy of further investigation (Kozak and Wiens, 2010). However, low reported numbers of hybrids could also mean that additional cases have yet to be discovered and described. Additional research will allow discrimination between these two possibilities.

Both the mitochondrial genetic distances (estimated with MT-CYB) and the nuclear genetic distances (estimated with RAG1) between parental species of salamanders that hybridize spanned two orders of magnitude (0.003 to 0.360 and 0.0007 to 0.020, respectively). The ratio of genetic distances between these two genes provides an estimate of the difference in substitution rates of the two genomes. In vertebrates, the mitochondrial genome typically has a higher substitution rate than the nuclear genome (Brown et al., 1979). However, this ratio is variable across genes and species in all of the major vertebrate clades including amphibians, where the majority of the estimates are from frogs (Allio et al., 2017). Our salamander estimates (0.6 to 162, median = 21.1) are consistent with existing estimates for other vertebrate taxa, suggesting similar relative evolutionary rates between the two genomes. Because the MT-CYB and RAG1 sequences used are short (345–1,141 bp and 556–1,510 bp, respectively) and represent single loci, some of the variation in genetic distance estimates reflects sampling error.

The ratio between mitochondrial and nuclear genetic distances can also be used to identify candidate examples where loci have introgressed from one parental species to another. In our dataset, potential cases of mitochondrial introgression would show a ratio much less than the median (i.e., <<21.1) These potential cases include Salamandrina perspicillata X Salamandrina terdigitata (MT-CYB /RAG1 ratio = 0.6) and Eurycea neotenes X Eurycea tridentifera (MT-CYB /RAG1 ratio = 0.9; Table 2). In phylogeographic studies, hybridization between Salamandrina perspicillata and S. terdigitata was identified to have limited mitochondrial introgression (Mattoccia et al., 2011). Additional cline-based analysis of mitochondrial and nuclear introgression across the hybrid zones of salamanders with extreme mitochondrial/nuclear genetic distance ratios would be informative.

Within each tetrapod group, the genetic distance between parental species that hybridize is highly variable, but we found no significant differences in genetic distances among the groups. In addition, there were no significant differences in the ratio of mitochondrial to nuclear genetic distances among the groups. Taken together, these results suggest that speciating lineages of salamanders retain the ability to hybridize at similar overall genetic divergence levels as speciating lineages in other tetrapod groups. Thus, the differences in metabolic rates across tetrapods do not appear to translate into predictable patterns of overall genetic divergence and incompatible loci in hybrids. This pattern may reflect a similar evolved match between metabolic “supply and demand” (i.e., ATP use and ATP synthesis; Darveau et al., 2002) in all tetrapod clades, irrespective of overall metabolic rates, such that increased metabolic costs in hybrids have equal detrimental effects on fitness. Cline analyses of OXPHOS-specific loci, combined with OXPHOS functional data in parental species and hybrids, would allow more rigorous testing of this hypothesis.

The presence of heteromorphic sex chromosomes does not appear to deter salamanders from hybridizing (Fig. 3). This pattern suggests that, despite evolutionary lability in amphibian sex chromosomes (Hillis and Green, 1990; Evans et al., 2012; Sessions et al., 2016), the underlying genetic sex-determining factors might be sufficiently conserved to not be a barrier to reproduction. Consistent with this explanation, for some salamanders, the difference between the heteromorphic sex chromosomes is extremely small (Schartl et al., 2016; Keinath et al., 2018), showing that the selective pressures might be similar on each chromosome or that the region of suppressed recombination on the X or Z chromosomes is small (unlike in birds or mammals; Schartl et al., 2016). Additionally, evidence that cryptobranchid salamander homomorphic sex chromosomes have been conserved for nearly 60 million years (Hime et al., 2019) suggests that the underlying genes for determining sex may be highly conserved. Finally, we note that the relationship between sex chromosome morphology and reproductive isolation is complex; in some cases, even homomorphic sex chromosomes show a large X/Z effect (e.g., Hyla arborea X H. orientalis, Dufresnes et al., 2016).

In conclusion, this comprehensive analysis of salamander hybrids revealed hybrids throughout the clade with a wide range of genetic divergences between parental species. Despite being outliers among tetrapods in metabolic rate and sex chromosome diversity, both of which were predicted to impact the relationship between genetic divergence and reproductive isolation, our analyses suggest that salamanders hybridize across similar levels of genetic divergence to the other examined tetrapod groups. This study further demonstrates the power of comparative studies of hybridization as an approach to understanding the process of speciation.

We acknowledge the members of Scott Melander's master's thesis committee: D. Sloan, W. Black IV, and G. Ebel, as well as members of the Sloan and Mueller labs, for their time and advice as this research grew and developed into its final product. We thank S. Sessions for providing unpublished data and guidance. Colorado State University provided financial support.