Artificial and Natural Hybridization in Schiedea and Alsinidendron (Caryophyllaceae: Alsinoideae): the Importance of Phylogeny, Genetic Divergence, Breeding System, and Population Size

Artificial hybridizations of Schiedea and Alsinidendron (Caryophyllaceae: Alsinoideae), genera endemic to the Hawaiian Islands, were used to study the effects of phylogenetic relationship, genetic divergence, breeding system, and population size on the ability of species to cross and produce fertile F1 hybrids. Using 17 taxa, all crosses attempted produced vigorous F1 hybrids, although pollen fertility of hybrids varied substantially. Fertility of F1 hybrids was positively related to genetic identity measured using allozyme variability. Fertility of hybrids was not predicted by phylogenetic relationships, estimated from morphological and molecular data. Hybrids between species with dimorphic breeding systems had significantly higher pollen fertility than other combinations, although this effect was not significant when population size was controlled in the analysis. The association of dimorphism and ability to produce fertile F1 hybrids is probably indirect: dimorphic species of Schiedea occur in large populations and have high genetic identities. Although artificial hybrids are easily produced in the greenhouse, natural hybridization in Schiedea and Alsinidendron is limited, apparently because instances of sympatry are rare and autogamy is often found in one of the two species occurring sympatrically. The ability of species to hybridize and produce fertile offspring is often interpreted as indicative of close evolutionary relationship, with high fertility of hybrids resulting from genetic similarity among parental species (Clausen et al. 1939; Stebbins 1950; Raven 1977; Grant 1981; Coyne and Orr 1989, 1997). The ability to produce fertile hybrids has been used to infer relationships among plants for over a century (Raven 1980). From a phylogenetic perspective, the ability to hybridize has been hypothesized as a plesiomorphic or ancestral trait (Vickery 1978; Rosen 1979; Funk 1985; Smith 1992). Acquisition of a post zygotic barrier to hybridization represents an evolutionarily derived trait resulting from the accumulation of genetic differences. Subsequent speciation events may result in clades of closely related, interfertile species, until new barriers to hybridization evolve within these clades. Eventually, lineages may consist of clades of interfertile species, isolated by hybrid sterility from other clades. Post-zygotic crossing barriers may not evolve in clades, and there are many genera, and even families of flowering plants, where few barriers to hybridization are known (Solbrig 1970; Grant 1981; Hodges 1997). In cases of sympatry, the occurrence of post-zygotic isolating mechanisms associated with reduced fitness of hybrids might result in selection for pre-zygotic barriers to hybridization, because parents that do not produce any hybrids would have the highest fitness. In lineages where barriers to hybridization occur, barriers might follow patterns related to phylogeny, or instead be related to overall genetic similarity (e.g., phenetic patterns based on allozyme similarity; Nei 1978), which often differ substantially from phylogenetic patterns. The amount of genetic similarity retained by diverging populations would be affected by many factors, including population size and breeding system. In species with large populations characterized by outbreeding, fixation or loss of genetic variation would be less common than in small or inbreeding populations. Retention of genetic similarity in large populations is more likely and should be associated with potentially greater interfertility. In contrast, in small populations genomic divergence associated with random processes and inbreeding should result in reduced potential for production of fertile hybrids. To address these predictions about the potential effects of phylogeny and overall genetic similarity on hybrid fertility, we studied artificial hybrids in 572 [Volume 26 SYSTEMATIC BOTANY Schiedea and Alsinidendron (Caryophyllaceae: Alsinoideae), two closely related Hawaiian genera that demonstrate extraordinary variation in breeding systems as well as population size. A central goal of the study was to determine whether sterility barriers between species in this lineage follow patterns related to phylogeny, or are better explained by factors associated with population size and breeding system. A second goal was to examine patterns of sympatry and putative cases of natural hybridization in Schiedea and Alsinidendron, and the role hybridization may have played in the evolution of this diverse lineage. MATERIALS AND METHODS Study Organisms. Schiedea and Alsinidendron (Caryophyllaceae: subfam. Alsinoideae), represent the fifth or sixth largest adaptive radiation in the Hawaiian Islands, and together the genera are considered monophyletic because of specialized extensions of the nectaries (Wagner et al. 1995; Weller et al. 1995). The 29 species of Schiedea and 4 species of Alsinidendron occupy a wide range of habitats and possess diverse morphology and breeding systems (Wagner et al. 1995; Weller et al. 1995; Sakai et al. 1997; Wagner and Herbst 1999). Species occurring in wet or mesic forests are large-leaved hermaphroditic vines, herbs, or shrubs. Species occupying dry habitats are often narrow-leaved woody shrubs that are typically gynodioecious (i.e., female and hermaphroditic individuals present in populations), subdioecious (female, male, and hermaphroditic individuals in populations), or dioecious (female and male individuals in populations); gynodioecy, subdioecy, and dioecy are collectively termed dimorphic breeding systems. All dimorphic species in this lineage are wind pollinated (Weller et al., 1998). Phylogenetic analysis based on morphological traits indicates that there are four major clades within the endemic Hawaiian Alsinoideae (Fig. 1; Wagner et al. 1995; Weller et al. 1995); results from cpDNA studies, which yield less resolved trees, also support the existence of three of the four clades (Soltis et al. 1996). Allozyme studies indicate that high genetic identities are positively associated with large population size (Weller et al. 1996). Artificial Hybridization. Crosses of Schiedea and Alsinidendron were made in the greenhouse under pollinator-free conditions. Plants were grown from seeds or cuttings collected in the field. Sixteen of twenty-nine species of Schiedea and one species of Alsinidendron were used in crosses. Of the 136 non-reciprocal combinations possible for 17 taxa, 98 different, usually non-reciprocal hybrid combinations were made. Sixteen species of Schiedea and Alsinidendron were not crossed because species were extinct (two species), considered extinct at the time (three species), were discovered and described as new to science after the crosses were completed (two species), or were not in cultivation (nine cases, including five species that were not recognized as distinct at the time the crosses were made). For crosses between sexually dimorphic species and hermaphroditic species, female plants were generally used as seed parents to avoid the necessity of emasculating hermaphroditic flowers. When both species pairs were hermaphroditic, flowers of individuals used as female parents were emasculated on the first day of anthesis, prior to anther dehiscence. Pollinations were carried out two to three days later, after stigmas had become receptive. Recently dehisced anthers were used to coat all stigmas with pollen. Wind-pollinated species were isolated in different areas of the greenhouse to prevent contamination. Because females were used whenever possible as seed parents, very few reciprocal crosses were made. In the first of two sets of crosses, we investigated the effects of male parentage on seed production, seed germination, and pollen fertility of F1 hybrids. For these crosses we used the same female individuals as parents in crosses to males or hermaphrodites of different species as well as individuals of the same species. Two individuals each of S. globosa and S. kealiae, and one individual of S. adamantis were used as female parents. Seed production was recorded for an average of 25 capsules per cross. Seeds were germinated in the greenhouse following a required six-month period of dormancy. We planted eight to nine two-inch pots with 25 seeds per pot and scored germination after all seedlings had emerged. Essentially no mortality occurred prior to transplanting, and survival was not analyzed separately from germination. For each plant used as a seed parent, we made intraspecific crosses within populations for comparison to the effects of interspecific hybridization. To gain a more complete understanding of patterns of F1 hybrid sterility, we used many more species in the second set of crosses. These crosses were made at different times than those used to detect the effect of male parentage and no attempt was made to compare seed production or percent germination between the two sets of crosses. 2001] 573 WELLER ET AL.: SCHIEDEA AND ALSINIDENDRON HYBRIDS FIG. 1. Phylogeny of Schiedea and Alsinidendron (Weller et al. 1995), based on morphological data. Breeding system transitions are shown on branches of phylogeny, which represents one of six equally parsimonious trees and differs only minimally from the strict consensus. Molecular data support results shown here, although the phylogeny is less resolved (Soltis et al. 1996). Unique transitions are indicated by solid bars, homoplasious transitions by open triangles, and reversals by open bars (for bootstrap values of the strict consensus tree and details on character mapping, see Weller et al. 1995). Taxonomy reflects updates outlined by Wagner and Herbst (1999; the occurrence of S. helleri2 in the phylogeny reflects errors in coding characters in Wagner et al. [1995] and Weller et al. [1995]). Not all currently recognized species are shown in the phylogeny. Breeding systems (BS), population size (Pop), and island distributions (Island) indicated to right of species (hermaphroditic 5 H, gynodioecious 5 G, subdioecious 5 SD, dioecious 5 D; L 5 populations with more than 100 individuals, S 5 populations with less than 30 individuals [see Weller et al., 1996 for details of how population size was assessed]; Hawai‘i 5 H, Lana‘i 5 L, Maui 5 M, Moloka‘i 5 Mo, Nihoa 5 N, Kaua‘i 5 K, O‘ahu 5 O). Asterisks by population size indicate species included in crossing program. 574 [Volume 26 SYSTEMATIC BOTANY For both sets of crosses, we raised approximately 30 progeny from each cross to maturity. Pollen fertility was measured for five F1 hybrids, chosen haphazardly from each group of progeny. One anther from each of five different flowers from each plant was placed on a separate slide with 0.1 % aniline blue in lactophenol (1:1:1:1 lactic acid, phenol, glycerin, and water by weight); 200 pollen grains per anther were scored for fertility. Fully expanded pollen grains filled to the cell wall with darkly staining cytoplasm were considered fertile. Shrunken pollen grains lacking cytoplasm or with reduced quantities of cytoplasm were scored as infertile. Average fertility for each individual was used in statistical analyses, where individuals served as replicates for the effects of hybridization on pollen fertility. Temporal variation in pollen fertility was compared for some F1 hybrids and parents by measuring the pollen fertility of the same individuals at intervals ranging from six months to one year. Data Analysis. Analysis of variance was used to detect the effects of clade (the S. adamantis, S. globosa, S. membranacea, and S. nuttallii clades; Fig. 1) on fertility of F1 hybrids among species. Arcsinetransformed pollen fertility was the dependent variable in the ANOVA. Crosses were largely restricted to clades dominated by dimorphic species (the S. adamantis and S. globosa clades; Weller et al. 1995) due to the relatively small number of crosses between hermaphroditic species that were attempted. Because population size of parental species might affect hybrid fertility, these analyses were also carried out for species occurring in large populations only. Populations consisting of more than 100 individuals were classified as large; those with fewer than 30 individuals were considered small (Weller et al. 1996). An analysis using species occurring in small populations could not be performed because of limited numbers of crosses. All analyses were carried out using SAS (SAS Institute 1989). When necessary, significance levels were adjusted using the sequential Bonferroni method for multiple contrasts (Sokal and Rohlf, 1995). Because not all contrasts in analysis of variance are phylogenetically independent, results should be interpreted with caution. Fertility of F1 hybrids was also interpreted with respect to Nei’s (1978) unbiased genetic identities, derived from allozyme data (Weller et al. 1996). Arcsine-transformed pollen fertility of F1 hybrids was regressed against genetic identity. For this analysis pollen fertility was averaged for crosses involving the same parental species. The data were corrected for phylogenetic relationships using Felsenstein’s (1985) method (cf. Coyne and Orr 1989, 1997). The effects of breeding system on fertility of F1 hybrids between species were also investigated using ANOVA. The four categories of crosses used in the analysis were dimorphic 3 dimorphic crosses, dimorphic 3 hermaphroditic crosses, hermaphroditic 3 dimorphic crosses, and hermaphroditic 3 hermaphroditic crosses. The hermaphroditic category refers to those species where populations consist only of hermaphroditic individuals. Effects of Geographic Separation. The potential effects of geographic separation on fertility of F1 hybrids were investigated by coding crosses according to whether parental species occur on the same island, adjacent islands (e.g., Kaua‘i and O‘ahu), or are separated by one or two islands (e.g., Kaua‘i and Maui or Kaua‘i and the island of Hawai‘i). Analysis of variance was used to detect differences in arcsine-transformed pollen fertility. Natural Hybridization. Potential natural hybridization was detected through morphological intermediacy of putative hybrids in the field and in collections from more than 30 herbaria. The potential occurrence of hybrids among progeny of species occurring sympatrically was also investigated.