* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Patterns of Genetic Differentiation in Appalachian Desmognathine
Survey
Document related concepts
Transcript
Patterns of Genetic Differentiation in Appalachian Desmognathine Salamanders S. G. Tllley Studies of allozymic variation In salamanders of the Desmognathus ochrophaeus complex have revealed complex patterns of genetic fragmentation among populations and species. Populations of D. carolinensls within the Black and Great Craggy Mountains of western North Carolina exhibit pronounced variation In allozyme frequencies at five polymorphic loci. Allozyme frequencies at one of these loci vary with elevation. There are no obvious patterns at the remaining loci with respect to elevation, latitude, or longitude, and only a weak suggestion of isolation by distance among populations. On a broader geographic scale, the northernmost form (D. ochrophaeus) exhibits very little Intrademic variation, even over distances approaching 1000 km. Forms with progressively more southern distributions exhibit progressively higher levels of Isolation by distance. Genetic distances among local populations of the more southern forms (D. orestes, D. carolinensls, and D. ocoee) Increase with geographic distances among populations separated by less than 100 km, but level off at greater geographic distances. Genetic distances for Interspecific comparisons also correlate positively with geographic distances, but the relationships are not oriented toward the origins of the scatterplots, indicating additional genetic differentiation that cannot be attributed to Isolation by distance. Patterns of Interspecific differentiation do not relate to either geographic or cladlstlc relationships among these forms. Levels of genetic fragmentation appear higher In southern Appalachian forms of the D. ochrophaeus complex than in two larger, more aquatic congeners: D. montlcola and D. fuscus. Blotic Interactions with these and other low-elevation predators and competitors may have produced genetic fragmentation In the D. ochrophaeus complex. From the Department of Biological Sciences, Smith College, Northampton, MA 01063. This research was supported by the Blakeslee Genetics Research Fund at Smith College and National Science Foundation grant BSR-8508363. Collecting permit* were provided by the States of North Carolina and Tennessee; Plsgah, Nantahala, and Cherokee National Forests; the Great Smoky Mountains National Park; and the Blue Ridge Parkway. Louise Meade, Meredith Mahoney, Robert Merritt, and David Wake provided comments on the manuscript. A large part of the dataset was collected by Smith College students pursuing Independent research projects supported by the Blakeslee Genetics Research Fund, the Howard Hughes Medical Institute, and the Smith College Tomllnson Fund. I thank all these individuals, especially Nancy Karelia, Lenl-Sarah Machlnton, Meredith Mahoney, Lada Soljan, and Emily White, for their contributions. This paper was delivered at a symposium entitled "Genetics of Fragmented Populations" sponsored by the American Genetic Association at the University of Georgia on May 18, 1996. Journal of Heredity 1997:88:305-315; 0022-1503/97/$5.00 Taxa that combine high species richness with ecological diversification can teach us much about how, to paraphrase G. Evelyn Hutchinson, the processes of genetic fragmentation play out in various ecological theaters. Desmognathine salamanders represent such a group. Desmognathines are restricted to middle and eastern North America, attaining their greatest taxonomic diversity (12 currently named species) in the Appalachian mountains. Desmognathines combine morphological conservatism with extensive Intra- and Interdemic variation in color patterns, and some species are extremely difficult or even impossible to distinguish in the field. Allozyme studies have greatly expanded our understanding of species diversity in this group, leading to the resurrection of the cryptic species Desmognathus imitator (Tilley et al. 1978), the description of D. santeetlah (Tllley 1981; Tilley and Schwerdtfeger 1981), and, most recently, the subdivision of D. ochrophaeus into four allopatric, genetically differentiated forms: D. ochrophaeus, D. carolinensis, D. ocoee, and D. orestes (Tilley and Mahoney 1996). Concommitantly our understanding of the diversity of desmognathine life histories (Tllley and Bernardo 1993), behaviors (Arnold et al. 1993; Forester 1977), and ecological relationships (Halrston 1986,1987) has expanded greatly in recent years. We should now be able to at least begin relating patterns of genetic fragmentation to the biogeographic histories and ecological attributes of species, and to thus begin understanding the Interplay of ecological and genetic mechanisms that generate taxonomic diversity in this group of animals. This article deals primarily with patterns of genetic differentiation in the D. ochrophaeus complex. It summarizes and expands on findings published recently by Tilley and Mahoney (1996) to consider genetic differentiation at three levels: among populations of a single species (D. caroli- 305 too complex, however, to be adequately conveyed by any traditional taxonomic treatment. The taxonomic statuses of D. ocoee and D. orestes are especially problematical. Populations assigned to D. ocoee display considerable genetic differentiation, indicating that further taxonomic subdivision may be warranted. There is evidence of intergradation between populations assigned to D. ocoee and D. carolinensis along the Blue Ridge Divide east of Asheville, North Carolina (Tilley and Mahoney 1996), but combining these two forms would create an even more loosely denned "species" than the current D. ocoee. D. orestes consists of two distinct Salamanders of the D. clusters of populations, referred to by Tilochrophaeus Complex ley and Mahoney (1996) as groups B and C. The D. ochrophaeus complex, as defined by Tilley and Mahoney (1996) consists of Members of the D. ochrophaeus complex six species: apalachicolae, carolinensis, im- are all medium-size (to about 65 mm stanitator, ocoee, ochrophaeus, and orestes. D. dard length), semiterrestrial salamanders, apalachicolae is restricted to the Gulf whose preferred habitats are the banks of coastal plain, and D. imitator to the Great first-order streams, seepage areas, wet cliff Smoky and Great Balsam Mountains of faces, and hlgh-elevatlon forests. In the western North Carolina and eastern Ten- forms for which Information on reproducnessee, where it is sympatrlc with D. tive biology Is available (principally ochocoee. The work described here deals with rophaeus, carolinensis, and ocoee), ovipothe remaining four species, which were sition, followed by brooding by the female until recently treated as a single form, D. until the eggs hatch, occurs in the viciniochrophaeus. D. ochrophaeus in its new re- ties of small streams, springs, seepages, stricted sense (Tilley and Mahoney 1996) and wet cliff faces. Females oviposit in ranges from Kentucky to Quebec across brooding chambers under moss on banks the Allegheny and Cumberland Mountains and logs, in situations where the emerging and Plateaus west of the Ridge and Valley larvae can readily enter shallow surface Physiographic Province. The remaining water. Larvae typically hatch In late sumforms (maps in Figs. 5 and 6) are para- mer and transform the following spring. patrically distributed from north to south Females reach sexual maturity in 4-5 in the southern Blue Ridge Physiographic years, males in 3-4 years, with maturity Province: D. orestes in southwest Virginia, being delayed with increasing elevation northwest North Carolina, and northeast (Tilley and Bernardo 1993). Adult surviTennessee; D. carolinensis in the Unaka, vorship is relatively high for a small verBald, Black, and Newfound Mountains tebrate (Tilley 1980). north of the Pigeon River in eastern TenThe lives of these salamanders are intinessee and western North Carolina; and D. mately connected with aquatic sites asocoee in the ranges south of the Pigeon sociated with first-order streams and their River, including the Great Smoky, Great headwaters. These sites provide habitats Balsam, Cowee, Nantahala, and Unicoi for larvae, brooding habitats for females, Mountains and the highlands along the and overwintering sites for all life-history Blue Ridge Divide into northwest Georgia, stages. At high elevations, individuals with an isolated group of populations in from populations centered around seepthe Cumberland Plateau of northeast Ala- age areas in the heads of coves can disbama. perse into surrounding woodlands Tilley and Mahoney (1996) based their throughout even the warmest months. At decision to taxonomically subdivide D. low elevations, individuals seem much ochrophaeus on an evolutionary species more restricted to the vicinities of small concept. Each of the four forms that they seepage areas, and especially to wet cliff recognized (D. carolinensis, D. ochro- faces. However, aquatic habitats at low elphaeus, D. ocoee, and D. orestes) seemed evations usually support dense populato represent a collection of populations on tions of D. quadramaculatus and D. montia distinct evolutionary trajectory. The pat- cola. These two more aquatic species atterns of differentiation in this group are tain larger body sizes and prey on memnensis) within a restricted geographic region, among populations of single species throughout their geographic ranges, and among populations of different species of the D. ochrophaeus complex. It also adds comparative information on two congeners, D. fuscus and D. monticola, to show how patterns of genetic fragmentation may reflect the diverse ecological attributes of these species and members of the D. ochrophaeus complex. Finally, it attempts to relate these patterns to other attributes of the biology and biogeographic history of these animals. 3 0 6 The Journal of Heredity 1997:88(4) bers of the D. ochrophaeus complex (Halrston 1986, 1987). Even in habitats where they are abundant and widely distributed, local demes may be much more isolated than they appear, due to several ecological and behavioral attributes that may constrain dispersal abilities and gene exchange among local populations. Factors Restricting Dispersal Along Streams Since most of the larval stage occurs during the winter months, there is little larval growth or opportunity for dispersal of larvae from the vicinities of the oviposition sites. Larvae are often observed in mere surface films at the very edges of seeps and on wet cliff faces. Downstream dispersal is probably restricted by small size, relatively weak swimming ability, preference for very shallow surface water, and the short duration of the larval stage. Downstream larval "drift" is probably much less important in the ochrophaeus complex than in species with larger stream-dwelling larvae. The streams of Appalachia are hazardous dispersal corridors for members of the ochrophaeus complex. They are inhabited by one to four species of large plethodontids that are known to feed on smaller salamanders: D. quadramaculatus, D. monticola, D. fuscus, and Gyrinophilus porphyriticus. The terrestrial tendencies of the ochrophaeus complex and other small Desmognathus have been interpreted as adaptations that minimize predation by these larger, more aquatic salamanders (Hairston 1986, 1987; Tilley 1968). Experimental manipulations of D. monticola densities have been shown to influence densities of D. ocoee (Hairston 1986). Factors Restricting Dispersal Across Land Female D. ocoee are known to exhibit phllopatry with respect to brooding sites (Forester 1977) in successive years. Thus, while females may disperse into the forest, particularly in years in which they do not brood, they may repeatedly oviposit in their natal stream headwaters. Mating occurs in the spring when gravid females occur in the vicinities of oviposition sites, and in the fall when both sexes congregate to overwinter In those same sites. Male philopatry has not been studied, but males that disperse into terrestrial habitats during the summer may return to and mate in their natal stream headwaters in the fall. The manner in which the geographic null model for such analyses is an array of populations in which genetic differentiaLatitude Longitude tion occurs in response to mutation at Min. Sec. Locality Elevation (m) Deg. Sec. County Deg. Mln. rate u and is retarded by gene exchange 1 11 42 29 1,554 35 82 at rate m. In such a system, it can be McDowell 16 2 Yancey 933 35 44 40 82 12 50 shown that the relationship between Nei's 1,676 Yancey 3 35 43 42 82 16 49 genetic distance and geographic distance McDowell 4 707 35 45 23 82 9 11 5 McDowell 488 35 44 11 82 7 9 eventually reaches an equilibrium, at Yancey 1,951 6 35 44 21 82 17 8 which that relationship is expected to be 1,987 Yancey 7 35 44 5 82 17 10 linear with zero Intercept and slope pro1,451 Yancey 8 35 42 6 82 15 56 Yancey 1,231 9 35 42 34 82 14 58 portional to V2u/m, where u and m are Yancey 1,067 10 35 43 18 82 14 57 the mutation and migration rates, respecYancey 11 969 35 44 2 82 15 6 tively (Nei 1972). Deviations from this simYancey 1,170 12 35 44 26 82 11 49 McDowell 1,183 13 35 44 31 82 11 24 ple relationship can illuminate the proYancey 14 829 35 52 27 82 18 40 cesses and history of genetic fragmentaBuncombe 1,609 15 35 49 15 82 21 49 Buncombe 1,247 16 35 48 45 82 21 26 tion (Good DA, unpublished manuscript; Buncombe 17 866 35 47 60 82 21 37 Good and Wake 1992, 1993). Buncombe 18 878 35 43 42 82 24 33 These deviations might take two general Buncombe 1,463 19 35 41 60 82 24 12 forms. Deviations from linearity indicate that opposing forces that promote differentiation and cohesion have not yet distributions of D. ochrophaeus, D. orestes, carolinensis in a topographically and eco- achieved an equilibrium. A relationship D. carolinensis, and D. ocoee replace each logically heterogeneous area; among pop- between genetic distance and geographic other parapatrically from north to south, ulations from throughout the ranges of D. distance requires two things: sufficient across about 10 degrees of latitude, carolinensis, D. ochrophaeus, D. ocoee, and time and geographic isolation to produce should profoundly affect the genetic struc- D. orestes; and among those forms and two genetic fragmentation, and sufficient time tures of these forms. We encounter D. more aquatic desmognathines: D. monti- and gene flow to generate higher levels of ocoee in the foothills of the Georgia Blue cola and D. fuscus. similarity among geographically proximal Ridge and D. ochrophaeus in the New York demes than among more distant ones. If Catskills in the same sorts of habitats oc- F Statistics levels of gene flow are sufficiently low, cupied by the complex throughout its Fsr, or the "fixation index" (Wright 1965) then there will be no discemable relationrange, but patterns of dispersion and gene measures the genetic effect of population ship between genetic and geographic disflow In Georgia and New York may be very subdivision as the proportional reduction tance. In an array of semi-isolated demes different. We might also expect patterns of in overall average heterozygosity owing to whose genetic structures are initially idengenetic differentiation in the ochrophaeus variance in allele frequencies among sub- tical, the effects of gene flow will initially complex to be very different from those in populations. Values of F^ were calculated be evident at the smallest geographic disother related desmognathines, such as D. using BIOSYS-1 version 1.7 (Swofford and tances, and the relationship between gemonticola and D. fuscus, that inhabit larger Selander 1981). These values were aver- netic and geographic distance will initially streams that can provide dispersal corri- aged across populations of each species be asymptotic. Over time, genetic differdors. (as well as D. orestes groups B and C) and entiation raises the asymptote while gene across loci judged polymorphic by the flow increases the geographic distance at 95% criterion (mean frequency of the most which the asymptote Is reached, until the Materials and Methods common variant averaged across popula- relationship becomes linear over the entions of a group or species ^95%). The Data tire range of geographic distances. Tilley and Mahoney (1996) reported on alThe second type of deviation from the lozymic variation across 22 loci in 52 pop- Measures of Heterozygosity null model concerns the y intercept of the BIOSYS-1 was employed to calculate mean ulations of the form previously regarded relationship, that is, the genetic distance as D. ochrophaeus, which they subdivided heterozygoslties expected at Hardy-Weln- expected between two demes separated berg equilibrium, using Nei's (1978) unbi- by zero geographic distance. This interinto D. carolinensis, D. ochrophaeus, D. ocoee, and D. orestes. Information on col- ased method. Mean heterozygosities were cept is expected to be zero within a single averaged across the populations of each lection sites and laboratory procedures Is array of populations that share a simple provided by Tilley and Mahoney (1996). In species and of D. orestes groups B and C. history of isolation by distance. However, addition, my students and I have accuthe array of populations might include mulated data on microgeographic varia- Relationships Between Genetic and multiple groups descended from ancestion at a subset of these loci among D. car- Geographic Distance tors that underwent an initial period of difolinensis populations in the vicinity of the ferentiation, accompanied by differentiation Good DA (unpublished manuscript) and Black and Great Craggy Mountains of west- Good and Wake (1992, 1993) have shown of their constituent demes in response to ern North Carolina (collection localities in isolation by distance. The y intercept for that the relationships between Nei's Table 1), and on the entire set of loci in comparisons between populations of two (1978) standardized genetic distance and several populations of D. fuscus and D. genetically differentiated subgroups of geographic distance among populations monticola. We can thus examine genetic can provide insight into patterns of differ- populations should lie above the origin. fragmentation among local demes of D. entiation within and among species. The The intercept's value in this case repre- Table 1. Collecting locallUe* In the Black and Great Craggy Mountain* Tilley • Genetic Differentiation in Appalachian Desmognaihine Salamanders 307 sents the additional genetic differentiation accomplished by factors other than isolation by distance. These might include a history in which populations ancestral to the subgroups differentiated in allopatry or current nongeographic barriers to gene exchange. The statistical significance of regression parameters in these analyses cannot be tested by traditional methods because a given population enters into several comparisons and the data points are thus not statistically independent (Dietz 1983; Manly 1986). Nevertheless, the broad patterns evident In scatterplots of genetic versus geographic distance deserve our attention. Where regression lines are calculated in some of the analyses below, they are provided for visual purposes only. Mlcrogeographic Variation Among D. carollnensis Populations in the Vicinity of the Black and Great Craggy Mountains The Black and Great Craggy Mountains diverge from and then roughly parallel the Blue Ridge (Tennessee Valley) divide east of Asheville, North Carolina (Figure la). The two ranges collectively form a continuous ridgeline that undulates from northeast to southwest across about 15 minutes of latitude between Micaville (Yancey County, North Carolina) and Swannanoa (Buncombe County, North Carolina). For nearly 30 km of this distance the ridgecrest never drops below 1500 m above sea level. The highest portion of the ridgeline, referred to as the Black Mountains, rises above 2000 m for a distance of 20 km and includes Mt. Mitchell (2037 m), the highest peak in North America east of the Black Hills. D. carolinensis occurs throughout this topographically complex region, from the lowest elevations (about 500 m) to the summit of Mt. Mitchell (Tilley 1973). Our sampling localities span essentially this entire vertical range from locality 5 at 488 m at the foot of the Blue Ridge Divide northwest of Marion (McDowell County, North Carolina) to locality 7 on Clingman's Peak, 1987 m, on the crest of the Black Mountains (Table 1, Figure la). We also attempted to distribute our sampling among the three major mountain ranges (Great Craggy, Black, and Blue Ridge) and drainages (Ivy Creek, Cane River, South Toe River, and Catawba River) of the region. Localities 1, 2, 4, 5, and 8-13 lie along the Blue Ridge Divide running from the point where it joins the Black Mountains near 3 0 8 The Journal of Heredity 1997.88(4) Pinnacle Mountain (locality 1) northeastward on opposite sides of the divide in the drainages of the South Toe and Catawba Rivers. Localities 3, 6, and 7 lie on the main (eastern) ridgecrest of the J-shaped Black Mountain range, localities 15-17 on the lower western ridge, locality 19 on the Great Craggy Mountains, locality 18 below the crest of the Craggies In Dillingham Creek drainage, and locality 14 at a low elevation (829 m) along the Cane River south of Burnsville (Yancey County, North Carolina). Our two most distant sample populations (localities 5 and 18) are separated geographically by 26 km. D. carolinensis populations In this region are strongly polymorphic for five of the 22 loci in our standard set: creatine Unase (CK EC 2.7.3.2), glycerol-3-phosphate dehydrogenase (G3PDH, EC 1.1.1.8), lsocitrate dehydrogenase-1 (IDH-1, EC 1.1.1.42), L-lactate dehydrogenase-2 (LDH-2, EC 1.1.1.27 ), and leucyl-glycyl-glyclne peptidase (PEP, EC 3.4.-.-). All these loci show pronounced variation In allozyme frequencies in this restricted geographic region (Figure lb-f). F^ values average 0.219 for the five loci (Table 2). The only regular pattern occurs at the CK locus (Figures lb and 2), which exhibits the highest F^ value (0.580) and where allozyme frequencies vary strongly from west to east (Figure lb) and with elevation (Figure 2). This pattern is most pronounced among sample populations 1-7, between the crest of the Black Mountains to the foot of the Blue Ridge near Lake Tahoma. The Ckb variant that characterizes the highest elevation populations is otherwise a rare allozyme in the units of the ochrophaeus complex that we have studied (Table 3 in Tilley and Mahoney 1996). It is known to occur In only two other populations of D. ocoee: as the predominant variant (P = .75) in the Pink Beds population, Transylvania County, North Carolina, 1024 m; and as a rare variant (P = .036) at Rough Butt Bald, Jackson County, North Carolina, 1676 m. Interestingly, the same allozyme (or one with the same mobility) also distinguishes D. santeetlah, a high-elevation form, from low-elevation populations of D. fuscus (Tilley 1988). None of the remaining loci exhibit any sort of geographic pattern of variation (Figure lc-l). The same allozymes occur throughout the sampling regions at frequencies that vary discordantly among the loci and that sometimes differ strongly among sampling localities separated by only a few kilometers. Figure 3 shows how genetic distances vary with respect to geographic and vertical distances for comparisons among populations in the Black and Great Craggy Mountains, across the five loci that are strongly polymorphic in these populations. The dataset excludes populations 3, 9, 12, and 15, for which data are lacking for one or more of the loci. These genetic distances cannot be compared to those presented below, which are based on the entire set of 22 loci. Comparisons between populations that are strongly differentiated at CK (represented by filled circles in Figure 3) generally produce larger genetic distances, which increase with elevational distance as a consequence of the elevational cline at that locus (Figure 2). Populations 5 and 7 (separated by 15.1 km and 1500 vertical m) exhibit the greatest genetic distance (0.60) on the graphs. Geographic distance alone accounts for very little of the variation in genetic distance. Macrogeographlc Variation Within and Among Forms of the D. ochrophaeus Complex Intraspeciflc Differentiation The Fjy values for the four forms of the D. ochrophaeus complex treated here are shown in Table 2 and Figure 4. Levels of differentiation among these forms lie well within the range of FOT values given by Larson (1984, Table IV) and Larson et al. (1984, Table 2), who also based their values on averages across loci judged polymorphic according to the 95% criterion. Of the four species, D. carolinensis, D. ochrophaeus, and D. orestes exhibit similar F^ values in the middle of the distribution, while a notably higher proportion of the genetic variation in D. ocoee occurs at the interdemic level. Larson et al. (1984) concluded from their analysis of fCT statistics in species of plethodontid salamanders that gene flow is probably too attenuated to act as a "cohesive force" in these amphibians. The same can be said for these members of the D. ochrophaeus complex, in which levels of genetic differentiation probably reflect historical patterns rather than current levels of gene exchange. D. ochrophaeus. Populations of this form exhibit very low levels of average heterozygosity across the 22 loci (mean H = 0.05) and only six of these loci are polymorphic according to the 95% criterion (Table 2). Genetic distances calculated across all 22 loci indicate very low levels of differentiation over distances of as much as 1000 km (Figure 5a). However, when only the six polymorphic loci are BTT3CT 82°22-3Cr 82-2230" 82"15' STTXT . d) IDH-1 3 b) . e) LDH-2 CK cr w-^it/Jg, 17 18 19 . C) G3PDH -f) Figure 1. Variation in allozyme frequencies at five highly polymorphic loci among populations of D carolinensis in the Black Mountains and vicinity. Buncombe. McDowell, and Yancey Counties, North Carolina, (a) Sampling localities, numbered as in Table 1. (b)-{f) Variation in allozyme frequencies. Frequencies are proportional to shaded areas: allelic variants are distinguished by different shading patterns In a few cases commonest and very rare variants have been pooled. Tilley • QenetJC Differentiation in Appalachian Desmognathine Salamanders 3 0 9 Table 2. Levels of genetic variation and average ocoee, and D. orestes Species D. carolinensis: Throughout range Black and Great Craggy Mountains No. of populations O 0.8 o Mean no. of polymorphic Mean loci heterozygoslty 02 9 13 0.131 0.469 19 9 16 14 4 10 6 16 16 10 15 0.055 0.097 0.139 0.121 0.146 0.420 0.644 0.437 0.170 0.352 Numbers of polymorphic loci are averaged across populations. Heterozygotles were calculated using Nel's (1978) unbiased method and averaged across loci and populations. Fn values are averaged across populations for loci In which the mean frequency of the most common variant was ^95%. considered, the mean F^ value for this form (0.420, Table 2) is similar to those for D. carolinensis and D. orestes (0.469 and 0.437, respectively; Table 2 and Figure 4). Genetic differentiation in these forms nevertheless involves more loci and has occurred over much more restricted geographic areas. D. orestes. Patterns of differentiation in this, the next form to the south (Figure 5b), are complicated by the fact that D. orestes consists of two distinct groups of populations (groups B and C, respectively, of Tilley and Mahoney 1996). Group B is apparently restricted to the Roan and Iron Mountain highlands along the North Carolina-Tennessee state line, and lowland populations occur in the vicinity of Johnson City, Tennessee. In that region it forms a wedge between group C D. orestes to the north and D. carolinensis to the south. Group B constituted the most highly differentiated form studied by Tilley and Mahoney (1996), appearing as the most basal group of populations on their neighborjoining and parsimony trees. Nevertheless, Tilley and Mahoney (1996) presented evidence of intergradation between group B and the remaining D. orestes populations (their group C). Differentiation between groups B and C contributes to the magnitude of the overall F^ for D. orestes populations (0.437), as indicated by the lower values for groups B (0.170) and C (0.352; Table 2). Population comparisons within the group C D. orestes (Figure 5b, filled circles) all yield genetic distances below 0.2, with a very modest tendency for genetic 3 1 0 The Journal of Heredity 1997.88(4) 0.0 400 0580 0.170 0.152 0.079 0.114 0.219 distance to increase with geographic distance. Adding the four group B populations to the analysis steepens this relationship by superimposing a series of points across which the relationship between geographic and genetic distance rises much more steeply, to genetic distances of about 0.4. Six of these additional points near the origin of the graph represent comparisons among the group B populations; the remainder all represent comparisons between groups B and C. The overall relationship between genetic and geographic distance trends toward the origin of the graph in Figure 5b. Interdemic comparisons within D. orestes as a whole thus reflect, in striking contrast to the much more wide-ranging D. ochrophaeus, isolation by distance. There is, however, an interesting substructure to this general pattern. Much of the trend owes to a strong tendency for group B and group C D. orestes populations to become more genetically differentiated as they grow more geographically distant. Group C D. orestes populations are genetically quite homogeneous with only a weak tendency for levels of differentiation to increase with geographic distance. D. carolinensis. The distribution of this form extends only about 100 km from its contact with D. orestes to the Newfound Mountains southwest of Asheville, North Carolina. Nevertheless, comparisons among populations of this form yield a mean F^ of 0.469 and suggest Isolation by distance (Figure 5c), with genetic distance rising from 0 to about 0.3 as populations grow more geographically separated. O o O 0.6 0.4 CK G3PDH 1DH-1 LDH-2 PEP Mean D. ochrophaeus D. ocoee D. orestes Group B Group C 1.0 values among D. carolinensis, D, ochrophaeas, D. O O o <a o °o o -K3 900 1400 KD 1900 Elevation (m) Figure 2. Frequency of the Oallele (Figure lb) plotted against elevation among populations of D. carolinensis In the Black Mountains and vicinity. D. ocoee. This form occupies the most southerly and topographically complex region of any of the species treated here: the southwestern Blue Ridge Physiographic Province from the Great Smoky and Great Balsam Mountains southward. D. ocoee populations formed the loosest cluster In the study by Tilley and Mahoney (1996), with some pairs of populations in different mountain ranges displaying genetic distances as great as those between different species of Desmognathus across the same set of 22 loci. This form displays the greatest fjT value (0.644) of the four species (Table 2). Populations of D. ocoee show a tendency to become more differentiated over the first 100 km of geographic separation, but at higher geographic distances genetic distance appears to level off at about 0.3-0.4 (Figure 5d). All but two of the points beyond geographic distances of more than 150 km involve comparisons with the Alabama populations, but suggestion of an asymptotic relationship persists even with these points omitted. Figure 5e summarizes the patterns suggested by the scatterplots in Figure 5a-d. Levels of genetic fragmentation appear to increase from north to south. Patterns suggesting Isolation by distance characterize differentiation within each of the southern Appalachian forms, in very distinct contrast to the situation in the most northern form, D. ochrophaeus. Furthermore, in the northernmost member of the southern Appalachian complex, group C D. orestes, the isolation-by-distance pattern appears distinctly weaker than In the two more southerly forms. In D. carolinensis and D. ocoee, and in comparisons involving the most southerly (form "B") D. orestes populations, genetic distance rises with geographic distance so that populations separated by 100 km typically differ by a Nei distance of about 0.3. There is no evidence, in any species of the complex, of isolation by distance among populations separated by more than 100 km. a) 0.7 0.6 0.5 t 0.4 0.3 o °» o o 0.2 rt. 0.1 o 8 10 to Q o (D CD n O 20 30 Geographic distance (km) b) 0.7 0.6 0.5 \ 0.4 0.3 \ 0.2 0.1 %% o . . o &_ 8 o OCP6)' 500 1000 1500 Elevational distance (m) Figure 3. Genetic distance plotted against (a) geographic distance and (b) elevatlonal distance for comparisons among populations of D. carolmensis In the Black Mountains and vicinity. Genetic distances are calculated across five loci polymorphic In these populations, and are not comparable to those shown in Figures 5-7, which are based on 22 loci. Differentiation Among Forms of the D. ochrophaeus Complex Figure 6 shows the relationships between genetic and geographic distances for comparisons involving populations of different species. Comparisons involving D. ochrophaeus populations and those of the three other species (Figure 6a-c) yield genetic distances of about 0.3, on average. Since the most northern (New York) and southern (Kentucky) D. ochrophaeus popula- tions are so genetically similar to one another (Figure 5a), there is no relationship between genetic and geographic distance for either of the three categories of Interspecies comparisons. The remaining three relationships between genetic and geographic distance for interspecies comparisons (Figure 6d-f) are all strongly positive. For all the interspecific comparisons Involving D. orestes, D. carolinensis, and D. ocoee, positive re- lationships between genetic and geographic distance are evident even at geographic distances exceeding 100 km In contrast to the pattern of intraspecific differentiation within D. ocoee (Figure 5d). The relationship between genetic and geographic distance for comparisons of D. orestes to D. carolinensis (Figure 6d) is particularly revealing. The sample includes populations of these two parapatric forms that are separated by only a few kilometers, and even these geographically adjacent populations differ by genetic distances of about 0.3. This is approximately the same genetic distance that separates D. ochrophaeus from the other three species. The relationship between genetic and geographic distance for comparisons of D. orestes to D. carolinensis thus does not trend toward the origin of the graph. Its upward displacement indicates that populations of these two forms are more genetically differentiated than we would predict from geographic distances alone, and that isolation by distance does not entirely account for the pattern of genetic fragmentation. In the remaining two classes of interspecific comparisons, D. orestes versus D. ocoee (Figure 6e) and D. carolinensis versus D. ocoee (Figure 6f), it is less clear whether isolation by distance alone falls to account for the patterns of differentiation. D. orestes and D. ocoee are allopatric, and the contact zone between D. carolinensis and D. ocoee is currently under study. Thus we lack pairs of geographically adjacent populations for either of these two interspecific comparisons. A further complication is that the Alabama D. ocoee populations are consistently less differentiated from either D. orestes or D. carolinensis than would be predicted from geographic distances. If comparisons involving the Alabama populations are excluded, the least-squares regression lines fit to the remaining comparisons (involving only the D. ocoee populations in the southern Blue Ridge Physiographic Province) intersect the genetic distance axes between 0.1 and 0.2. Since the points in the scatterplots are not statistically independent, we cannot determine whether these y intercepts differ significantly from zero. It does appear that D. orestes and D. carolinensis are more genetically dissimilar than D. orestes and D. ocoee or D. carolinensis and D. ocoee, when genetic distances are "corrected" for geographic distances among populations. Tiltey • Genetic Differentiation in Appalachian Desmognathine Salamanders 3 1 1 4 • ft 3req 3 Itl 00.10 0.20 o o Q> o Q O •• • ^m 0.00 CO IK? !• 2 • O O cara //n If 0.30 0.40 050 ^^^^B ; 1IH| • • I • • • • 060 0.70 0.80 FST Figure 4. Histogram showing distributions of F^ value* for plethodontld salamanders. Shaded squares represent values given by Larson (1984, Table IV). Squares with filled circles indicate species Inhabiting middle and eastern North America; empty circles Indicate the species studied here. Comparisons With Other Desmognathines D. monticola and D. fuscus are abundant species along low-elevation streams in the Appalachians. Both species are more aquatic than members of the D. ochrophaeus complex, seldom venturing more than a few meters from flowing water. The low-elevation regions that evidently act as barriers to dispersal in the D. ochrophaeus complex may actually function as dispersal corridors for D. monticola and D. fuscus. It is thus of interest to compare patterns of genetic variation in these forms to those in the D. ochrophaeus complex. We have accumulated allozyme data for four populations of D. monticola (Cocke County, Tennessee, n = 29; Polk County, Tennessee, n = 28; Washington County, Tennessee, n = 16; Macon County, North Carolina, n = 20) and three populations of D. fuscus (Grayson County, Virginia, n = 7; Scioto County, Ohio, n = 16; Franklin County, Massachusetts, n = 18). These and the D. ochrophaeus complex data are based on the same 22 loci. The relationships between genetic and geographic distance for intraspecific comparisons in our samples of D. monticola and D. fuscus are shown in Figure 7. When adjusted for geographic distances among populations, levels of intraspecific differentiation in D. htscus and D. monticola are quite low compared to those in D. carolinensis, D. ocoee, or D. orestes, but not to those in D. ochrophaeus. Our four D. monticola populations are all from the southwestern Blue Ridge Physiographic Province. These samples are nearly genetically identical to one another, in contrast to samples of D. carolinensis, D. ocoee, or D. orestes in the same geographic region and separated by 3 1 2 The Journal of Heredity 1997.88(4) similar geographic distances (Figure 5be). The highest genetic distances for intraspecific comparisons in D. fuscus or D. monticola correspond to the greatest geographic distances: at about 1000 km of geographic separation, Massachusetts D. fuscus differs from Ohio and Virginia D. fuscus by genetic distances of 0.14 and 0.25, respectively. Karlln and Guttman (1986) reported a very similar range of genetic distances (0.0-0.2 versus our 0.09-0.22) for comparisons among northern populations of D. fuscus over a similar latitudinal range, employing a set of 21 loci that had 14 allozyme systems in common with ours. Thus D. fuscus populations apparently require 1000 km of geographic separation to achieve the same degree of genetic differentiation (D « 0.2) that is typically exhibited by D. carolinensis, D. ocoee, or D. orestes populations separated by less than 100 km. In contrast, this level of differentiation is approximately twice the average genetic distance between D. ochrophaeus populations separated by the same geographic distance (Figure 5a). History, Biogeography, and Genetic Fragmentation Patterns of genetic differentiation within and among Appalachian units of the D. ochrophaeus complex are quite intricate. It may be premature to attempt to develop a historical hypothesis that attempts to account for all of them, but we can identify patterns that such a hypothesis must eventually explain: 1. Evidence for genetic differentiation over very short distances among populations of D. carolinensis. 2. The lack of differentiation among pop- ulations of the most widely distributed form, D. ochrophaeus. 3. A general tendency for levels of genetic differentiation to increase from north to south within the D. ochrophaeus complex. 4. The lack of evidence for a relationship between genetic and geographic distance among D. ocoee populations separated by more than 100 km. 5. The positive relationships between genetic and geographic distance for interspecific comparisons Involving D. orestes, D. carolinensis, and D. ocoee, even for populations separated by more than 100 km. 6. Lower degrees of differentiation, when corrected for geographic distances among populations, for D. ocoee versus D. orestes and D. ocoee versus D. carolinensis than for any of the other interspecific comparisons. 7. Higher degrees of Intraspecific differentiation, when corrected for geographic distances, in D. carolinensis, D. ocoee, and D. orestes than in D. monticola or D. fuscus. Taken as a whole, the data for D. carolinensis populations in the Black and Great Craggy Mountains indicate that within this restricted, topographically diverse region, these populations exhibit considerable local differentiation in allozyme frequencies. The tendency for allozyme frequencies to vary with elevation at CK suggests that selection may account for some of this differentiation, as Tilley (1973) and Bernardo (1993) suggested that it did for elevational variation in life-history characteristics of D. carolinensis and D. ocoee populations. It must also be true that levels of gene flow among these populations are extremely low. Any model of genetic differentiation within this complex must accommodate the contrast between D. carolinensis demes that exhibit differentiation within a single high-elevation region that contains no evident barriers to dispersal and D. ochrophaeus, which exhibits exceedingly low levels of differentiation among local populations as distant as 1000 km from one another. D. ochrophaeus has clearly expanded its range northward since the last (Wisconsin) episode of continental glaciation. Its northernmost populations lie about 500 km north of the southernmost extent of Wisconsin glaciation, which was attained about 21,000 BP (Peltier 1994). Salamanders of the genus Plethodon that have invaded formerly glaciated territory also exhibit high levels of genetic uniformity (Highton et al. 1989; Highton and Webster 1976; Larson 1984; ochrophaeus carolinensis CO O 0 200 400 600 800 1000 orestes 0 _i\ 100 ocoee •within "form C" 200 300 • comparisons involving Alabama populatons o 100 200 300 0 100 200 300 O "55 e) 0) o ocoee 0.3- 0.1ochrophaeus 100 Geographic distance (km) Figure 5. Relationships between geographic and genetic distances for comparisons among populations of (a) D. ochrophaeus, (b) D. orestes, (c) D. carolinensis, and (d) D. ocoee. (e) Curves, fit by eye, summarizing the trends In (aM<0- Larson et al. 1984). They and members of the D. ochrophaeus complex confront us with a biogeographic paradox: their genetic fragmentation in the south, particularly the southern Appalachians, requires very low levels of gene exchange, yet they have colonized vast areas of formerly glaciated terrain since the retreat of Wisconsin ice sheets. The curves in Figure 5e resemble the series that would represent the relationships between genetic and geographic distance in populations that have either been undergoing isolation by distance for increasing lengths of time or have been differentiating for the same period but under different levels of gene flow. It may be impossible to disentangle these two mechanisms In this complex, because both levels of gene flow and time available for differentiation probably vary with latitude in this complex of species. Group C D. orestes may represent a complex of populations that has been differentiating longer than D. ochrophaeus, or which are more isolated from one another in the topographically complex southwestern Blue Ridge Physiographic Province, or both. The still greater tendency for genetic distance to accumulate with geographic distance in D. carolinensis and D. ocoee would be expected in these still more southerly complexes of populations. The roughly asymptotic relationship between geographic and genetic distance for comparisons among D. ocoee populations is to be expected in an array of populations that has not had time to achieve the equilibrial linear relationship predicted under simple isolation by distance. Viewing the D. ocoee pattern as a stage in progression toward an equilibrium may be misleading. There can presently be no gene exchange at all between the Alabama populations and the remainder of D. ocoee. Even among more continuously distributed populations, at geographic distances beyond 100 km, the time required for the equilibria! relationship to be attained may exceed the time scales of major climatic and biogeographic shifts. Turning to the interspecific comparisons, the model proposed by Good DA (unpublished manuscript) accounts for the positive relationships between genetic and geographic distance seen In Figure 6d-f as manifestations of past gene exchange between these forms, and the upward displacements of the relationships as the consequences of additional differentiation during subsequent periods of genetic isolation. If these positive, nona- Tffley • Genetic Differentiation in Appalachian Desmognathine Salamanders 3 1 3 i l0.3 3 0.2 CD S • Desmognathuifuscia O Desmognathus monttcota 01 o.o 200 400 600 800 1000 Geographic distance (km) Figure 7. Relationships between geographic and genetic distances In D. monticola and D fuscus. a) ochrophaeus vs. orestes r\\ orestes vs. carolinensis 1.0 i 0.8 0.6 0.4 0.2 0.0 500 8 CO w b) 1.00.8 06 0.4 02 0.0 0 c) 1.0 T 08 0.6 0.4 0.2 00 0 1000 1500 0 ochrophaeus vs. carolinensis ^\ e) 500 1000 1500 ochrophaeus vs. ocoee 0 r\ 200 600 400 orestes vs. ocoee I comparisons involving Alabama populations 200 400 600 carolinensis vs. ocoee S"i*a V * i • comparisons involving Alabama populations CDo 500 1000 1500 0 200 400 600 Geographic distance (km) Figure 6. Relationships between geographic and genetic distances lor between-specles population comparisons, between (a) D. ochrophaeus and D orestes, (b) D carolwensis, and (c) D ocoee; between (d) D. orestes and D. carolinensis and (e) D. ocoee; and between (f) D. carolinensis and D ocoee. Least-squares regression lines are shown for visual comparisons only. Regression calculations for (e) and (0 exclude the Alabama populations. Shading patterns as In Figure 5. D. orestes and D. carolinensis are less differentiated from the Alabama populations of D. ocoee than from less distant populations of that form. Tilley and Mahoney (1996) suggested that the geographic distributions of these species may have been quite different in the past. Any historical blogeographic scenario that explains the current patterns of genetic differentiation would be at odds with the evolutionary relationships suggested by parsimony analyses of the allozyme data. The shortest tree obtained by Tilley and Mahoney (1996), rooted with respect to D. imitator, rendered D. orestes paraphyletic, with group B populations basal and group C populations terminal. D. carolinensis (group D in Tilley and Mahoney 1996) was likewise paraphyletic, with respect to group C D. orestes, on this tree. A tree that portrayed D. carolinensis as monophyletic was only two steps longer. The phylogenetic relationships suggested by this tree are shown in Figure 8. Of the relationships suggested by this phylogeny, only the relatively basal position of D. ochrophaeus agrees with the patterns shown in Figure 6, which portray that form as about equally differentiated from D. carolinensis, D. orestes, and D. ocoee. It seems prudent to refrain from further speculation on the evolutionary and biogeographic history of this complex until additional data, such as nucleotide sequences and information on imitator Group B orestes ochrophaeus ocoee symptotic relationships were established at some time in the past when the three species were genetically contiguous, levels of gene flow at that time were evidently higher than they presently are among populations of D. ocoee, which show no evidence of isolation by distance among populations separated by more than 100 km. 3 1 4 The Journal of Heredity 1997:88(4) Levels of differentiation among the three species in the southern Blue Ridge Physiographic Province do not agree with the current geographic relationships among these forms. D. orestes is clearly more differentiated from D. carolinensis, its neighbor to the south, than from the more distant D. ocoee. Furthermore, both carolinensis Group C orestes Figure 8. Phyolgenetlc relationships suggested by cladlstlc analysis of allozyme data, rooted with D. imitator designated as the outgroup and constrained to portray all of the population groups recognized by Tilley and Mahoney (1996) as monophyletic- morphological characters, are brought to bear on the issue of phylogenetic relationships. Patterns of intraspecific genetic differentiation in the southern Blue Ridge desmognathines seem consistent with the ecological attributes of these species. The genetically homogeneous D. monticola is an abundant, stream-bank dwelling, lowland form for which there are probably few barriers to dispersal. The genetically much more heterogeneous D. carolinensis, D. ocoee, and D. orestes are montane forms whose lives seem centered on the headwaters of first-order streams. These habitat requirements may very well reflect competition with and predation by D. monticola and D. quadramaculatus, another low-elevation, streambank form (Hairston 1986, 1987). Interspecific interactions may, in fact, have a long history of involvement in genetic differentiation in the D. ochrophaeus complex. If the forms of this complex are simply adapted to cooler, moister, environments, their present geographic fragmentation could be interpreted as a postPleistocene event. Highton (1995) proposed that speciation events in the P. glutinosus complex predate the Pleistocene, attributing them to vicariance events during Pliocene episodes of glaciation and aridity. His criterion for species recognition was a Net genetic distance of at least 0.15, which is considerably less than the genetic distances we have obtained among units of the D. ochrophaeus complex. Furthermore, our set of loci excludes "fast-evolving" transferrin, albumin, and esterase loci that Highton employed. Had we employed these loci, we probably would have obtained even greater genetic distances. Unless rates of molecular evolution are unusually high in these salamanders, levels of differentiation within and among D. carolinensis, D. ocoee, and D. orestes appear far too high to have accumulated only since the last episode of continental glaciation. The beginnings of genetic fragmentation in the D. ochrophaeus complex appear to be at least as ancient as the fragmentation of the P. glutinosus complex, and I suggest that they may have different origins. Plethodon are completely terrestrial animals that require relatively mesic conditions for reproduction and foraging. Highton's (1995) proposal that arid climates have imposed geographic isolation among populations in cool, moist refugia seems perfectly reasonable. Mem- bers of the D. ochrophaeus complex, however, require stream headwaters and seepage areas for opposition, brooding, and larval development. Their dispersal across lowland areas may be constrained by biotic interactions with the larger, more aquatic species of desmognathines that inhabit higher-order streams. These interactions have apparently generated associations between small body size and terrestriality (Halrston 1986, 1987; Tilley 1968; Tilley and Bernardo 1993) in desmognathines. Tendencies toward genetic fragmentation may be an additional consequence of adaptation to terrestrial environments in these salamanders. D. monticola is the most abundant, ubiquitous occupant of low-elevation streams in the southern Appalachians. It may be significant that the genetic distances between D. monticola and members of the D. ochrophaeus complex average about 0.41, while those among units of the complex average 0.33 (Tilley and Mahoney 1996, Table 7). These genetic distances suggest that diversification of the D. ochrophaeus complex followed, perhaps not too distantly, the appearance of a large-bodied, stream-dwelling desmognathine that came to be exceedingly abundant in low-elevation habitats in the southern Appalachians. The occurrence of large predatory salamanders in low elevation habitats may have been the main barrier to gene exchange among populations of the D. ochrophaeus complex, while ice caps and spruce-fir forests have expanded and contracted, for as long as these species have coexisted In the southern Appalachians. References Arnold SJ, Reagan NL, and Verrell PA, 1993. Reproductive isolation and speciation in piethodontid salamanders. Herpetologlca 49:216-228. Bernardo J, 1993. Experimental analysis of allocation In two divergent, natural salamander populations. Am Nat 14314-38. Dletz EJ, 1983. Permutation tests for association between two distance matrices. Syst Zool 32:21-26. Forester DC, 1977. Comments on the female reproductive cycle and phllopatry by Desmognalhus ochrophaeus (Amphibia, Urodela, Plethodontldae). J Herpetol 11:311-316. Good DA and Wake DB, 1992. Geographic variation and speciation In the torrent salamanders of the genus Rhyacotriton (Caudata: Rhyacotritonldae). Unlv Calif Publ Zool 126:1-91. Good DA and Wake DB, 1993. Systematic studies of the Costa Rican moss salamanders, genus Notonilon, with descriptions of three new species. Herpetol Monogr 7: 131-159. Halrston NG, 1986. Species packing In Desmognathus salamanders: experimental demonstration of predation and competition. Am Nat 127:266-291. Halrston NG, 1987. Community ecology and salamander guilds. Cambridge: Cambridge University Press. Highton R, 1995. Speciation In eastern North American salamanders of the genus Plethodon. Annu Rev Ecol Syst 26:579-600. Highton R, Maha GC, and Maxson LR, 1989. Biochemical evolution In the slimy salamanders of the Plethodon glutinosus complex In the eastern United States. Illinois Blol Monogr 57:1-153. Highton R and Webster TP, 1976. Geographic protein variation and divergence in populations of the salamander Plethodon cinereus. Evolution 3(h33-45. Kariin A and Guttman S, 1986. Systematlcs and geographic Isozyme variation In the piethodontid salamander Desmognathus fuscus (Rafinesque). Herpetologlca 42:282-301. Larson A, 1984. Neontotoglcal Inferences of evolutionary pattern and process In the salamander family Plethodontldae. Evol Blol 17:119-217. Larson A, Wake DB, and Yanev KP, 1984. Measuring gene flow among populations having high levels of genetic fragmentation. Genetics 106:293-308. Manly BFJ, 1986. Randomization and regression methods for testing for associations with geographic, environmental and biological distances between populations. Res Popul Ecol 28:201-218. Nel M, 1972. Genetic distance between populations. Am Nat 106:283-292. Nel M, 1978. Estimation of average heterozygoslty and genetic distance from a small number of individuals. Genetics 89:583-590 Peltier WR, 1994. Ice age paleotopography. Science 265: 195-201. Swofford DL and Selander RB, 1981. Blosys-1: a FORTRAN computer program for the comprehensive analysis of electrophoretlc data In population genetics and systematlcs. J Hered 72:281-283. Tilley SG, 1968. Size-fecundity relationships and their evolutionary Implications In five desmognathine salamanders. Evolution 22:806-816. Tilley SG, 1973. Ufe histories and natural selection In populations of the salamander Desmognathus ochrophaeus. Ecology 54:3-17. Tilley SG, 1980. Ufe histories and comparative demography of two salamander populations. Copela 1980: 806-821. Tilley SG, 1981. A new species of Desmognathus (Amphibia: Caudata: Plethodontldae) from the southern Appalachian Mountains. Occas Pap Mus Zool Unlv Mich 695:1-23. Tilley SG, 1988. Hybridization between two species of Desmognathus (Amphibia: Caudata: Plethodontldae) In the Great Smoky Mountains. Herpetol Monogr 2:27-39 Tilley SG and Bernardo J, 1993. Ufe history evolution In piethodontid salamanders. Herpetologlca 49:154163. Tilley SG and Mahoney MJ, 1996. Patterns of genetic differentiation in salamanders of the Desmognalhus ochrophaeus complex (Amphibia: Plethodontldae). Herpetol Monogr 10:1-42. Tilley SG, Merritt RB, Wu B, and Highton R, 1978. Genetic differentiation In salamanders of the Desmognathus ochrophaeus complex. Evolution 32:93-115. Tilley SG and Schwerdtleger PM, 1981. Electrophoretlc variation In Appalachian populations of the Desmognathus hocus complex (Plethodontidae). Copela 1981: 109-119. Wright S, 1965. The Interpretation of population structure by F-statlstlcs with special regard to systems of mating. Evolution 19-395-420. THley • Genetic Differentiation in Appalachian Desmognathine Salamanders 3 1 5