Syst. Biol. 41(3):273-291, 1992 EVOLUTIONARY RELATIONSHIPS WITHIN THE ENSATINA ESCHSCHOLTZII COMPLEX CONFIRM THE RING SPECIES INTERPRETATION CRAIG MORITZ,1 CHRISTOPHER J. SCHNEIDER,2 AND DAVID B. WAKE2 ^Department of Zoology, University of Queensland, St. Lucia, Queensland 4072, Australia 2Museum of Vertebrate Zoology, University of California, Berkeley, California 94720, USA Downloaded from Abstract.—Sequences (644-681 bp) from the mitochondrial cytochrome b gene were obtained for 24 individuals representing the geographic range and morphological diversity of the poly- typic salamander ring species Ensatina eschscholtzii. These data were used to estimate the phy- logeny of components of the ring to test the biogeographic scenario underlying current inter- pretations of speciation in this complex. The analysis revealed high levels of nucleotide variation among subspecies. Strong subdivision was evident within the subspecies platensis and oregonensis. The phylogenetic hypothesis of minimum length that is best supported by the data contains http://sysbio.oxfordjournals.org/ one monophyletic group that includes populations from the southern Sierra Nevada and moun- tains of southern California (croceater, klauberi, and southern platensis) and another that includes populations of the southern and central coastal regions (xanthoptica and eschscholtzii). Samples of oregonensis were typically basal, but their precise branching order was unstable. Both oregonensis and platensis were paraphyletic, with several disparate lineages in oregonensis and a strong north- south dichotomy in platensis. The data were incompatible with a biogeographic model that required all subspecies to be monophyletic but were compatible with slightly modified predic- tions of a model assuming stepwise colonizations from north to south down the Sierra Nevada and independently down the coastal ranges. These features provide strong support for the biogeographic scenario central to the interpretation of Ensatina eschscholtzii as a ring species. Division of this complex into separate species on the basis of the observed patterns of monophyly at North Carolina State University on March 25, 2014 for mitochondrial DNA (mtDNA) is unwarranted because further sampling could reveal addi- tional instances of paraphyly across subspecies and, more generally, because mtDNA alone should not be used to infer species boundaries. [DNA sequences; mitochondrial DNA; cytochrome b; ring species; phylogenetics; geographic variation; speciation; Caudata.] Salamanders of the Ensatina eschscholtzii Stebbins (1949) recognized seven sub- complex provide a particularly clear ex- species distributed in the mountains en- ample of a ring species (Stebbins, 1949). circling the central valley of California (Fig. Ring species are circularly arranged poly- 1). Unblotched subspecies occur on the typic taxa with gradual transitions be- coastal margins and replace each other from tween adjacent components but abrupt southern {eschscholtzii), through central changes where the terminal races come into (xanthoptica), to northern California (ore- contact. They are supposed to exemplify gonensis and picta) into Oregon and Wash- speciation through gradual divergence ington. Blotched subspecies are distributed among large population units (Rensch, down the inland side of the valley on the 1929; Mayr, 1963) and, thus, the linkage mesic western slopes ©f the Sierra Nevada, between microevolutionary processes and The subspecies platensis ranges widely speciation. This view contrasts with recent through the Sierra but is replaced at the models of speciation that emphasize ge- southern end by populations of croceater. netic changes in small peripheral popula- The third blotched form, klauberi, is dis- tions (Mayr, 1982; White, 1982; Carson and junct and occurs as isolated populations in Templeton, 1984). It is also contrary to the the mountains of southern California. The opinion that the process of speciation is blotched and unblotched forms are in con- qualitatively distinct from intraspecific mi- tact in three places: across the top of the croevolution (e.g., Goldschmidt, 1933; ring, in the central Sierran foothills where Gould and Eldredge, 1977). a disjunct series of populations of xanthop- 273 274 SYSTEMATIC BIOLOGY VOL. 41 115° less continuous, particularly in the south- ern part of the range. The Sierran popu- lations of xanthoptica were thought to be recently derived from the coastal popula- tions via a transvalley leak during the more mesic periods of the Pleistocene. Based on this model of historical bio- geography, Stebbins proposed that the southward colonizations were each accom- panied by differentiation such that inter- breeding was precluded when the two lin- eages came into contact again in the Downloaded from mountains of southern California. The key observation (Stebbins, 1949:506) that sug- gested the gradual accumulation of repro- E3 Picta ductive isolation was that MM °r09 E23 esch progressive genetical divergence with progres- http://sysbio.oxfordjournals.org/ sively increasing reproductive incompatibility is klau observed from north to south between the coastal croc and interior series of races—from free interbreed- plat ing in northern California, through occasional hy- xan bridization in the Sierra in central California, to sympatry with reproductive isolation in southern California. Additional collecting and analyses (Steb- bins, 1957; Brown and Stebbins, 1964; Brown, 1974; Wake et al., 1986, 1989) have at North Carolina State University on March 25, 2014 35°- supported the basic interpretation. An analysis of allozyme variation (Wake and Yanev, 1986) revealed strong differ- entiation throughout the complex. Some 200 km subspecies were remarkably diverse for al- lozymes (e.g., platensis, oregonensis), where- FIGURE 1. Map of western USA showing the dis- tribution of localities from which Ensatina mtDNAs as others (e.g., xanthoptica, eschscholztii) were sequenced. The ranges of the subspecies are showed little differentiation. The presence defined on the basis of information in Stebbins (1949) of fixed differences between local popu- and extensive recent collections (Wake, unpubl. data). lations of the same subspecies was contrary Abbreviations, localities, and MVZ catalog numbers to the view (Dobzhansky, 1958) that the are in the Appendix. continuity of the ring was the result of ongoing gene flow. Given such strong geo- tica is in contact with platensis, and in the graphic structuring, the initial allozyme mountains of southern California where study did not sample sufficient popula- klauberi and eschscholtzii meet at four dis- tions to determine the discreteness of sub- junct sites. species or their relationships. Wake and Stebbins (1949) proposed that the species Yanev (1986:713), however, suggested that originally occurred in the Douglas fir and the blotched and unblotched forms "might redwood forests of southern Oregon and represent separate adaptive responses to northern California and that during peri- provincial selection pressures rather than ods of greater humidity, salamanders dis- discrete historical entities or incipient spe- persed down the coastal and inland ranges cies." to meet again in southern California. With In this paper, we used sequences from increasing aridity, the populations became mitochondrial DNA (mtDNA) to estimate 1992 RELATIONSHIPS WITHIN A RING SPECIES 275 phylogenetic relationships among com- Plethodon elongatus were selected as out- ponents of the ring and to evaluate the groups on the basis of other studies (Wake, historical biogeographic scenario basic to 1963; Larson et al., 1981) that suggested Stebbins's (1949) evolutionary model. Spe- these species as members of the sister clade cifically, we evaluated the phylogenetic of Ensatina, itself a monotypic genus. The predictions from two alternative scenarios: pilot study revealed appropriate levels of (1) Stebbins's model, which suggests in- variation (up to 12% observed sequence di- dependent stepwise colonizations down vergence within Ensatina), although phy- the coastal and inland ranges from north- logenetic analysis failed to resolve the ern Californian populations, with a recent branching order at the base of the tree. transvalley leak of xanthoptica, and (2) a In the second phase, we added nine more model based on the premise that the cur- samples of Ensatina to replicate represen- Downloaded from rent distribution of subspecies is the result tation of some subspecies and to attempt of secondary contacts between subspecies to split long branches, a potential cause of that had diverged considerably in geo- instability or error in parsimony analysis graphic isolation (cf. Stebbins, 1949:503). (Felsenstein, 1978; Swofford and Olsen, We compared mtDNAs to infer popu- 1990). We also extended the sequences by lation history. The mtDNA "gene tree" 303 bp in an attempt to increase resolution http://sysbio.oxfordjournals.org/ should reasonably reflect population his- at the base of the phylogeny. The final data tory where local populations are recipro- set included 24 individuals from 22 local- cally monophyletic for their mtDNAs and ities (Fig. 1) and one each from the out- diversity within populations is much less groups, Aneides and Plethodon. In the third than that among populations (Avise, 1989). phase of sampling, we screened multiple These conditions should be met where lo- samples from three populations for diag- cal populations have been small and have nostic restriction fragment length