Allozyme and morphological variation in two subspecies of Dryas octopetala (Rosaceae)in Alaska ^†

Study organism

Dryas octopetala L. (Rosaceae) occurs in the mountains of Alaska and western Yukon as two morphologically and ecologically distinct forms. These forms are recognized as subspecies (D. octopetala ssp. octopetala and ssp. alaskensis; Hulte´n, 1959, 1968) or species (D. octopetala and D. alaskensis; Porsild, 1947), based on leaf pubescence and the occurrence of scales or glands on the leaves. Gottlieb (1984) argues that such diagnostic traits are often determined by one or a few genes. These subspecies (as they are treated in this paper) have different suites of morphological and life-history traits, which are genetically determined and which confer selective advantage in their respective habitats (McGraw and Antonovics, 1983; McGraw, 1985a, b, 1987a, b) . Subspecies octopetala is adapted to xeric ridge or fellfield habitats and is identified by the presence of “octopetala” scales on the undersides of densely pubescent leaves (Hulte´n, 1959, 1968; McGraw and Antonovics, 1983). Subspecies alaskensis is adapted to wet tundra and snowbed depressions and is distinguished by the presence of glandular structures on the underside of glabrous leaves (Hulte´n, 1959, 1968; McGraw and Antonovics, 1983). Subspecies alaskensis is endemic to western Yukon and central Alaska and may have arisen during the Pleistocene from the circumpolar species Dryas. o. ssp. octopetala (Hulte´n, 1959). In interior Alaska both subspecies grow in alpine tundra geographically isolated from other tundra areas by unsuitable lowland habitat. The fellfield habitat of ssp. octopetala and snowbed depression habitat of ssp. alaskensis are often closely adjacent (Miller, 1982; McGraw and Antonovics, 1983); thus, these subspecies may grow only a few metres apart. To our knowledge ssp. alaskensis only occurs adjacent to ssp. octopetala, but ssp. octopetala often occurs without an adjacent population of ssp. alaskensis. Flowers are obligately entomophilous on Wrangell Island (Tikhmenev, 1985) and are apparently obligate outcrossers in Alaska, as there is no evidence for apomixis, auto-fertility, or self-fertility (McGraw and Antonovics, 1983). The flowers are visited primarily by Diptera (Elkington, 1971; McGraw and Antonovics, 1983). The species is diploid with 2n = 18 (Federov, 1969; Elkington, 1971; Lo¨ve and Lo¨ve, 1975; Krogulevich and Rostovtseva, 1984). Seeds are wind dispersed (Elkington, 1971), though it is unlikely that wind often carries achenes between isolated tundra areas (Elkington, 1971; McGraw and Antonovics, 1983). The subspecies are interfertile (McGraw and Antonovics, 1983), and morphologically intermediate plants (presumed hybrids) are observed in habitat intermediate between fellfield and snowbed (Hulte´n, 1968). Previous studies have not reported whether putative hybrids are fertile or sterile.

Plant material

Achenes of D. o. ssp. octopetala and ssp. alaskensis were collected from nine locations in interior Alaska (Fig. 1). Six of these locations (Eagle Summit, McCallum Creek, Mount Fairplay, Paxson Mountain, Polychrome Ridge, and Twelvemile Summit) host adjacent populations of ssp. octopetala and alaskensis. Three locations (Lime Mine, Murphy Dome, Wickersham Dome) contain only ssp. octopetala. The four northern populations occur in the White Mountains; Mount Fairplay is part of the Yukon-Tanana Uplands. The remaining four southern populations occur in the Alaska Range. Mountain ranges are isolated by regions of coniferous forest (Viereck et al., 1992, p. 9).

At each location we sampled ∼100 plants of each subspecies, collecting one leaf and one entire infructescence (achenecetum), representing one open-pollinated maternal family. Acheneceta were collected at least 2 m apart to reduce the possibility of collecting more than one achenecetum per genet. Subspecies designations were confirmed in the laboratory: presence of glands on the undersides of leaves indicates ssp. alaskensis, whereas the presence of “octopetala” scales indicates ssp. octopetala (Hulte´n, 1968). Samples for which neither glands nor scales, or both glands and scales, were present (<1%) were not used. Acheneceta were stored at −17°C until needed for electrophoresis or greenhouse experiments.

Electrophoresis

Achenes were placed on wet paper towels under constant illumination to effect germination. Only one seedling from each maternal family was chosen for electrophoresis when the first two true leaves developed at 4–6 wk. Whole seedlings were homogenized in 30–50 μL of extraction buffer. A tris-citrate buffer described by Cheliak and Pitel (1984) was used to grind samples for isocitrate dehydrogenase (IDH) visualization. Wendel and Weeden's (1989) tris-HCl grinding buffer number 3 for “tissues with high levels of interfering substances” was used for other enzymes. Whole-plant extracts were loaded into sample wells for cellulose acetate gels (Super Z-12 applicator kit, Helena Laboratories, Beaumont, Texas) or absorbed onto 0.3 × 1.5 cm filter paper wicks for starch gels.

IDH was resolved on 76 × 76 mm cellulose acetate gel plates (Titan III Zip Zone, Helena Laboratories, Beaumont, Texas) in the tris-borate-EDTA pH 8.6 electrode buffer number 10 of Soltis et al. (1983), using the techniques and enzyme staining recipe outlined by Hebert and Beaton (1989). Gels were run at 200 VDC for 20 min at 4°C. Glyceraldehyde-3-phosphate dehydrogenase, NAD-dependent form ([NAD]G3PDH) and NADP-dependent form ([NADP]G3PDH), glucose-6-phosphate dehydrogenase (G6PDH), malate dehydrogenase (MDH), phosphogluconate dehydrogenase (PGD), and shikimate dehydrogenase (SKD) were resolved on 12% starch gels using the tris-citrate pH 7.0 gel and electrode buffer system of Meizel and Markert (1967) as described in Wendel and Weeden (1989). Gels were run at 50 mA for 6 h at 4°C. Protocols for enzyme staining and techniques for gel preparation, loading, and slicing followed Wendel and Weeden (1989). Cellulose acetate and starch gels were scored when bands showed clearly, usually within 0.5 h of staining. For those enzyme systems with more than one presumed locus present on a gel, the most anodal locus was designated as number 1. Alleles at a locus were designated f for the fastest moving (most anodal migration), m for medium, or s for slowest moving (most cathodal position).

Allele frequencies

For each population (subspecies within a location), genotype frequencies were inferred from counts of electrophoretic phenotypes. A, the mean number of alleles per locus calculated across all nine resolved loci, P, the percentage of loci that exhibited more than one allele, H_O, average heterozygosity based on direct counts, and H_E, the average proportion of heterozygotes based upon Hardy-Weinberg expectations, were calculated (BIOSYS step VARIAB; Swofford and Selander, 1981). These statistics were calculated once for each population in the study and again for each subspecies, with samples pooled without regard for collection location.

Genetic differences between subspecies and among populations were examined by four types of analyses. First, heterogeneity of allele frequencies was examined using a contingency table analysis and χ² test (BIOSYS step Swofford and Selander, 1981). Second, genetic structure was described for means across polymorphic loci using the inbreeding coefficients (F statistics) of Wright (1978; BIOSYS step FSTAT; Swofford and Selander, 1981). Third, Roger's (1972) genetic distance (D) was calculated for all pairs of populations (BIOSYS step SIMDIS; Swofford and Selander, 1981). Fourth, an exact randomization test was performed to determine whether allelic differences between adjacent populations of ssp. octopetala and ssp. alaskensis are less (or greater) than allelic differences between geographically isolated populations of these two subspecies. We measured heterogeneity between adjacent populations using χ² (Workman and Niswander, 1970), which, unlike other distance measures, is sensitive to sample size. Mean χ² of the six pairs was used to summarize overall heterogeneity among adjacent populations of the two subspecies. Then, ssp. alaskensis populations (e.g., 1, 2, … 6) were paired with ssp. octopetala populations (e.g., A, B, … F) in all possible (6! = 720) combinations (e.g., 1A 2B 3C … 6F, 1B 2A 3C … 6F, 1C 2A 3B … 6F, and so forth). For each of the 720 pairings, mean χ² was calculated.

Greenhouse study

A subset of the populations assayed for allozyme variation was examined for morphological variation (Fig. 1). Twenty-five maternal families were randomly selected from ssp. alaskensis populations at Polychrome Ridge and Twelvemile Summit and from ssp. octopetala populations at Eagle Summit, Murphy Dome, Polychrome Ridge, and Twelvemile Summit. Eight achenes from each of these 150 families were randomly placed in seed trays and germinated at room temperature under constant illumination. Seedlings were potted in standard potting mix within 24 h of germination and randomly assigned positions on the greenhouse bench. Seedlings were watered as needed and fertilized weekly with a 20:20:20 NPK solution. After 31 wk, 436 surviving individuals were scored for two characters used to discriminate subspecies: presence of glands and presence of scales. Seven quantitative traits not used as taxonomic discriminators were measured: height of plant from soil surface to highest point; maximum rosette diameter; number of teeth per centimetre on the margin of largest leaf; length of largest leaf blade; maximum width of largest leaf with revolute margins in natural configuration; maximum width of largest leaf with revolute margins flattened and spread; and width between deepest incisions on the margin of largest leaf. Presence or absence of scent was assessed for each plant in a double-blind test.

Morphometric analyses

Each quantitative variable was examined for normality and homogeneity of variance. Plant height, plant diameter, and number of teeth on leaf margins were log transformed. Leaf dimensions were square-root transformed. Statistical testing of differences between subspecies and among populations was performed using mixed-model analysis of variance (PROC MIXED; SAS, 1997). Subspecies and populations nested within subspecies were treated as fixed effects; family was treated as a random effect nested within population. As a descriptive tool, we used Model II nested analysis of variance (PROC NESTED; SAS, 1990) to partition phenotypic variance between subspecies, among populations within subspecies, among families within populations, and within families. Canonical discriminant analysis (PROC CANDISC; SAS, 1990) was performed on standardized measurements to determine which suite of morphological traits maximizes the multivariate distance among populations. A matrix of generalized-squared distances between population centroids (D²) was created (PROC DISCRIM; SAS, 1990).

Breeding studies

To determine the potential for gene flow between subspecies of D. octopetala, we generated hybrid plants and backcrossed these hybrids with plants of both subspecies. Hybrid seed was produced in the field by transferring pollen from flowers of ssp. octopetala to bagged flowers of ssp. alaskensis and from flowers of alaskensis to bagged flowers of ssp. ocotopetala. We collected hybrid seed and seed produced by open pollination of both subspecies and grew the seed in a greenhouse for several months. We performed all possible crosses (including self-pollinations) among the following: (1) plants produced by open pollination of ssp. octopetala in the field, (2) plants produced by open pollination of ssp. alaskensis in the field, (3) hybrid plants with ssp. octopetala maternal parents, and (4) hybrid plants with ssp. alaskensis maternal parents.

Allozyme variation

Twelve electromorphs, presumably encoded by nine enzyme loci, were resolved. Twenty-two enzyme systems were rejected because of poor resolution, lack of interpretability, or low repeatability. Seven loci were monomorphic in all populations examined, whereas two loci, SKD and PGD-2, were polymorphic (Table 1). All alleles were found in both subspecies.

Mean number of alleles per locus within populations ranged from 1.0 to 1.3, with each subspecies having a mean of 1.3. Percentage polymorphic loci within populations ranged from 0.0 to 22.2. Mean heterozygosity ranged from 0.0 to 0.066 within populations (Table 1). Mean heterozygosity for ssp. alaskensis populations was 0.043. Mean heterozygosity for ssp. octopetala populations was 0.019. Computation of F statistics requires defining subpopulations relative to a total population. Because historical and contemporary patterns of gene flow within and among isolated populations of ssp. alaskensis and ssp. octopetala are uncertain, subdivision can be modeled in two ways (Table 2). Model A treats subspecies as separate taxa; collection locations are subpopulations of either ssp. alaskensis or ssp. octopetala. Model B treats the two subspecies as one taxon and focuses on subdivision between subspecies within a collection location. The two models describe different hierarchical structures and alternative hypotheses about reproductive isolation and gene flow. Inbreeding coefficients are reported as means across polymorphic loci. Mean F_IS values for four of six comparisons between subspecies and for both comparisons among locations were positive, indicating fewer heterozygotes within subpopulations than expected. F_ST ranged from 0.0 between subspecies at Eagle Summit to 0.056 between subspecies at Polychrome Ridge. Mean F_ST among geographically isolated populations (Model A) was 0.050; mean between-subspecies divergence was 0.044 (Model B). Roger's D ranged from 0.0 to 0.047 with a mean of 0.012 for 36 pairs of ssp. octopetala populations, a mean of 0.019 for 15 ssp. alaskensis pairs, and a mean of 0.019 for six pairs of both subspecies at a location.

A randomization test showed that allele frequencies in adjacent populations of the two subspecies are as different as random pairs of populations. Mean heterogeneity for adjacent populations of ssp. alaskensis and ssp. octopetala was 10.7. Mean heterogeneity for randomly paired populations ranged from 6.7 to 17.7. Adjacent populations ranked 218 out of the 720 possible pairings of populations, providing little evidence that genetic heterogeneity between adjacent populations is less (or greater) than heterogeneity between randomly paired populations (P = 0.3).

Morphological variation

Mixed-model analysis of variance showed highly significant differences (P < 0.0001) between subspecies in all traits measured except leaf width at incisions. Trait means differed between subspecies by as much as 50% (Table 3). Significant differences among populations were observed for six traits (P < 0.05) but not for number of incisions per centimetre (P > 0.8). Because plants were raised in a randomized design we attribute these phenotyic differences to genetic differences between subspecies and among populations.

Among-family variance arises because of nuclear genetic, maternal genetic, and maternal environmental variation among families and sets an upper limit for the narrow-sense heritability of traits. Restricted maximum-likelihood estimates of among-family variance in six traits were significantly greater than zero (P < 0.006), indicating contributions of one or more of these three sources of phenotypic variance. No significant among-family variance was observed in leaf margin incisions per centimetre (P > 0.5), indicating little or no within-population genetic variance in this trait.

Application of a random-effects model to these data provides a description of the relative magnitude of each statistical effect (Table 3). Most of the variance in traits was observed within populations, but of that portion of the variance not observed within populations, most occurred between subspecies rather than among populations.

Canonical discriminate analysis determines the linear combination of morphological characters that best distinguishes populations. The first canonical axis (CAN1) separated populations of tall plants with long, deeply incised leaves from populations of short plants with short, shallowly incised leaves (Table 4). Most of the variance in CAN1 lay between subspecies (Table 3) even though CAN1 was calculated to maximize distance among populations.

Generalized-squared distance (D²) provides a measure of the genetic distance between pairs of populations based on morphological characters (Fig. 2). D² ranged from 0.4 to 17.5. Genetic distances between population pairs composed of different subspecies (mean D² = 8.9) were greater than between pairs of ssp. octopetala populations (mean D² = 2.1) or pairs of ssp. alaskensis populations (mean D² = 2.0).

A total of 14% of all progeny had gland and scale characteristics different than their maternal parent (Table 5). Scent was closely associated with the presence of glands, but not all plants with glands had a detectable scent (Table 6).

Breeding studies revealed no genetic barriers to gene flow between subspecies. Viable seed resulted from crosses between hybrid plants and parental subspecies in 17 of 20 hand-pollinations. Other crosses confirmed interfertility between subspecies and self-incompatability as reported in earlier studies; eight of ten crosses between subspecies produced seeds, and self-pollination of seven bagged flowers failed to produce seed.

Allozyme diversity

The overall amount of allozyme variation found in Dryas octopetala in interior Alaska is very low compared with amounts reported for plant species with similar life-history characteristics. Although we were limited to nine interpretable loci (most studies examine 12–20 loci; Hamrick and Godt, 1990), both subspecies of D. octopetala have fewer alleles per locus, less than one-half the proportion of polymorphic loci, and less than one-half the expected heterozygosity reported as the average for long-lived woody perennials, narrowly distributed species, boreal-temperate plants, animal-pollinated or wind-dispersed species (see Hamrick and Godt, 1990). The probability of selecting two or fewer polymorphic loci in a random sample of nine loci from a species with 50% polymorphic loci is 0.10, indicating that there is a reasonable chance our low estimate of genetic variation in D. octopetala is an artifact of the few loci sampled. Nevertheless, low genetic variability in this species is entirely consistent with Hulte´n's (1959) proposal that Alaskan Dryas species survived the last glaciation isolated from other Dryas populations in the nonglaciated refugia of central Alaska, a period of possibly small population size and high genetic drift. Alternatively, low genetic variation may be a characteristic of the Rosaceae (e.g., Neviusia alabamensis—Freiley, 1994; Geum radiatum—Godt, Johnson, and Hamrick, 1996, Cercocarpus traskiae—Rieseberg and Gerber, 1995), a family for which few allozyme data are available (see Hamrick, Linhart, and Mitton, 1979; Loveless and Hamrick, 1984).

The distribution of allozymes across study sites indicates surprisingly little differentiation among populations of this species. Although we found statistically significant heterogeneity in allele frequencies among populations, values of F_ST in this study fall among the lowest to be found in a review of 124 plant species (Loveless and Hamrick, 1984). Several processes may account for the surprisingly low differentiation observed. Divergence of populations by genetic drift may be slow because of the large size of populations and because individual genotypes may persist for more than 100 yr (Kihlman, 1890 [cited in Porsild, 1947]). Gene flow among populations would counter divergence, but the possibility of significant gene flow between collection sites seems remote; collection sites were distributed among three isolated mountain ranges and separated by large distances. Alternatively, low differentiation among populations within subspecies may reflect relatively recent origins. Pollen records (Ritchie and Cwynar, 1982) indicate that following the last glaciation the range of Alaskan Dryas species expanded from glacial refugia in central Alaska and achieved a widespread continuous distribution (Hulte´n, 1959) 12 000–9000 yr BP. More recently, in response to climate amelioration, trees and shrubs invaded lowland areas, and Alaskan Dryas became isolated in alpine areas (Ritchie and Cwynar, 1982). Finally, balancing selection at allozyme loci also could counter the diversifying effects of genetic drift (Berger, 1975; Hedrick, 1986).

The similarity at allozyme loci between subspecies contrasts with the pattern of genetic differentiation often found between ecotypes and between morphologically distinct subspecies. Previous studies generally have found clear differences at allozyme loci that parallel phenotypic differences among taxa (e.g., Grant and Mitton, 1977; Crawford, 1985; Galen, Shore, and Deyoe, 1991; Hedre´n, 1996). Furthermore, Mopper et al. (1991) suggest that such genetic differences can arise in response to selection in a relatively brief period of time. Exceptions to the general pattern are seen among subspecies where divergence was recent and rapid (Crawford, 1985), where gene flow countered by selection for adaptive traits swamps differences at allozyme loci (Aitken and Libby, 1994), or where no distinct subspecies/ecotypes are present (i.e., surveys of geographic variation within species). In these cases, morphologically distinct taxa often display few if any differences at allozyme loci.

We found no evidence of gene flow between subspecies of D. octopetala. F statistics revealed differences between subspecies comparable to differences among populations isolated by hundreds of kilometres of unsuitable habitat, and a randomization test indicated no significant similarity in allele frequencies between adjacent populations of ssp. octopetala and ssp. alaskensis. This result suggests strong reproductive isolation between subspecies and is consistent with two observations: (1) McGraw and Antonovics (1983) found that although subspecies at Eagle Summit are interfertile, differences in time of flowering reduces the frequency of cross-pollination between subspecies to <1%, and (2) at sites where subspecies co-occur we found <1% of flowering plants combined traits used to distinguish subspecies. These results support studies that treat subspecies as distinct taxa. Conflicting with this interpretation are reports of the widespread occurrence of putative hybrids (Hulte´n, 1959, 1968; McGraw, 1987a). Indeed, among field-collected seed we found significant numbers of plants (14%) from families that combine the traits (scales and glands; Hulte´n, 1968) used to distinquish subspecies.

Hulte´n (1959) suggests that ssp. alaskensis may have arisen from a ssp. octopetala-like ancestor during the Pleistocene in the isolation of unglaciated refugia in Alaska and Yukon. Reduced genetic variability in derived taxa relative to progenitor taxa has been widely reported (e.g., Gottlieb, 1973, 1974; Baskauf, McCauley, and Eickmeier, 1994; Pleasants and Wendel, 1989; Sherman-Broyles et al., 1992; Purdy, Bayer, and Macdonald, 1994; Purdy and Bayer, 1995, 1996), yet we found no evidence supporting this relationship in D. octopetala. However, the pattern of allozyme variation is consistent with Hulte´n's (1959) hypothesis that contemporary Alaskan populations of both subspecies originated relatively recently from glacial refugia. Crawford (1983, 1985) reviews several allozyme studies that also report divergence among subspecies comparable to divergence among isolated populations within subspecies. He points out that divergence at allozyme loci reflects time since isolation of populations/subspecies and argues that isolation of subspecies may often take place on the same time scale as isolation of populations within subspecies. A similar result was obtained by Godt and Hamrick (1998) who also found little evidence of progenitor-derivative relationship between subspecies of Sarracenia rubra.

Morphological differentiation

In contrast to the allozyme data, subspecies of D. octopetala showed significant genetic differences in morphological traits. Although much of the variation in morphological traits occurred within populations, we found significant differences between subspecies for almost every trait measured. These differences are consistent with McGraw's (1985a, b, 1987b) extensive investigations of morphological, ecological, and growth differences between subspecies of D. octopetala at Eagle Summit. We conclude that differences between subspecies at Eagle Summit reported by McGraw extend to other sites in central Alaska as well.

In contrast to the clear differences between subspecies, morphological variation among populations within subspecies paralleled the allozyme data and indicated little differentiation among populations. While differences among populations were statistically significant, these differences generally accounted for <5% of the variance observed, suggesting that within a subspecies different populations display the same range of variation in morphological traits. The occurrence of one suite of traits in ssp. alaskensis and another suite of traits in ssp. octopetala suggests that the evolution of these traits is linked. Rather than evolving independently, evolutionary change in one trait parallels evolutionary change in a host of other traits. This may occur for three reasons (Armbruster and Schwaegerle, 1996): (1) the morphological features of ssp. alaskensis were fixed at the time of its origin and have not evolved since that time; (2) traits are genetically correlated due to linkage and/or pleiotropy, so that natural selection acting on one trait brings about a correlated response in other traits; or, (3) natural selection acts on traits in a correlated manner so that one combination of traits is favored in one environment and another combination of traits is favored in another environment. This third scenario is reminiscent of Wright's shifting-balance model, which presumes epistatic interaction among fitness traits, such that natural selection favors certain combinations of traits and eliminates other combinations. This third scenario also follows McGraw and Antonovics (1983) and McGraw (1987a) who concluded that selection maintains differences between subspecies. They found strong selection against alien transplants in both snowbed and fellfield habitats and reported that intermediates showed no advantage over native subspecies in either habitat or in intermediate habitat.

The high degree of fertility between subspecies and between hybrids and parental subspecies adds further support to the role of selection in maintaining subspecies differences. The absence of gene flow between subspecies suggested by the allozyme data and the lack of genetic barriers to interbreeding indicates that selection against hybrids eliminates the possibility of significant gene flow between subspecies.

Conclusions

Our results together with biogeographic and paleoecological data suggest relatively recent origins for Alaskan populations of D. octopetala ssp. octopetala and ssp. alaskensis. Divergence between subspecies at allozyme loci appears to have occurred on the same time scale as divergence among populations within subspecies. Lack of morphological differentiation among populations within each subspecies further supports a recent origin for these populations. Morphological differences between subspecies follow earlier studies that indicate natural selection is responsible for subspecies differences, and the high degree of interfertility between ssp. octopetala, ssp. alaskensis, and their hybrids strengthens claims that natural selection against intermediates maintains complete or nearly complete reproductive isolation between subspecies.