The primrose path to heterostyly


The repeated evolution of heterostyly in 28 or more families of flowering plants (Ganders, 1979; Barrett, 2002) has elicited considerable research attention. Perhaps no family has sustained the interest of researchers in heterostyly for so long, or provided the subject to research on it with such frequency, as the primroses (Primula) and their relatives in Primulaceae (e.g. Darwin, 1862, 1877; Richards, 2002). The special section on heterostyly in this issue grew out of a symposium that we organized for the 2005 International Botanical Congress in Vienna (, entitled ‘Integrating phylogenetic, molecular genetic, and ecologic approaches to understanding the evolution and function of heterostyly in Primulaceae’. In this commentary, we place the new results reported here into the broader context of research on heterostyly and discuss new ways to integrate these approaches.

‘Inference that a polymorphism seen in a group of species has a common origin using traditional ancestral state inference is a prediction that can be tested by inferring the allele trees of genes responsible for the phenotype in the linkage group.’

In heterostylous plants, a genetic polymorphism produces two (distyly) or three (tristyly) floral types (morphs) among individuals of a population. The floral morphs have their anthers and stigma(s) positioned at reciprocal heights (a type of reciprocal herkogamy; Webb & Lloyd, 1986). This reciprocal herkogamy is often accompanied by an incompatibility system that makes intermorph crosses more successful than intramorph (and thus self-) crosses (reviewed in Barrett & Cruzan, 1994). Recent reviews (Ganders, 1979; Lloyd & Webb, 1992; Barrett, 2002) consider the reciprocal herkogamy necessary and sufficient to describe a group as heterostylous, and we adopt this criterion. However, Darwin (1877, p. 3) considered the presence of both reciprocal herkogamy and an incompatibility system necessary to define a group as heterostylous, and this resonated throughout the literature on heterostyly for much of its history (Lloyd & Webb, 1992). This is a result, in part, of the assumption that the intramorph incompatibility preceded the origin of reciprocal herkogamy (e.g. Charlesworth & Charlesworth, 1979) or that the two arose in concert (Mather & de Winton, 1941), and thus both phenomena could be usefully discussed with a single term. In this issue of New Phytologist, Pérez-Barrales et al. (pp. 553–567) document the absence of an intramorph incompatibility system in species of Narcissus sect. Apodanthi that are heterostylous or have a style-length dimorphism (without an anther–height dimorphism). Also in this issue of New Phytologist, Hodgins & Barrett (pp. 569–580) explore the implications of reciprocal herkogamy in the absence of an intramorph (but presence of a self-) incompatibility system for morph ratios in populations of N. triandrus.

In many instances of heterostyly, there are additional genetic polymorphisms that cosegregate with anther and stigma position (e.g. polymorphisms in pollen size and stigma papillae length). These might reinforce intramorph incompatibility and/or make intermorph pollen transfer more efficient (Lloyd & Webb, 1992). In this issue, two new polymorphisms of this type are documented. Armbruster et al. (this issue; pp. 581–590) provide evidence that a three-dimensional reciprocity arose in a heterostylous ancestor of Linum suffruticosum and that this reciprocity has ramifications for the deposition of pollen on, and receipt of pollen from, different positions on the bodies of some types of pollen vectors. Webster & Gilmartin (this issue; pp. 591–603) describe the occurrence of a larger corolla tube mouth in the morph with anthers high and stigma low in the tube (‘thrum’) than in the other morph (‘pin’) in the distylous P. vulgaris. They also elucidate the combination of character states (e.g. style cell size and number) that produce reciprocal herkogamy in P. vulgaris, providing important phenotypic data for genotype/phenotype mapping in this model species.

Paths to heterostyly

Two models for the origin of heterostyly (defined by the particular form of reciprocal herkogamy described above) and the often co-occurring intramorph incompatibility system dominate today. The selfing avoidance model (Charlesworth & Charlesworth, 1979) proposes the origin of an incompatibility system with two mating types in an ancestor with nonherkogamy (stigmas and anthers at same height) to reduce inbreeding, followed by the origin of reciprocal herkogamy. The pollen transfer model (Darwin, 1877; Lloyd & Webb, 1992) proposes the origin of reciprocal herkogamy in an approach herkogamous ancestor (with stigmas above anthers) to make cross-pollination more efficient, with the later origin of an intramorph incompatibility system. The ancestral suites of states assumed to exist before the origin of each type of polymorphism is an important difference between the two models (Lloyd & Webb, 1992). Ancestral state inference on phylogenies for groups with heterostylous members has become a common approach used to address the ordering of origins of polymorphisms (Kohn et al., 1996) and to determine the relationship between the origins of polymorphisms and other plant features or life history traits (Barrett & Graham, 1997).

In this issue of New Phytologist, ancestral state inference is used to address the origin of heterostyly in two families for the first time, and it is used with an impressive collection of new reproductive data in a third (Pérez-Barrales et al.). Armbruster et al. conclude that heterostyly arose multiple times in the genus Linum, based on a preliminary phylogeny of the genus (15 spp. sampled out of c. 200; Mabberley, 1993). Mast et al. (this issue; pp. 605–616) infer the most recent common ancestor of Primula to have been reciprocal herkogamous (of the type associated with heterostyly), with multiple subsequent shifts to monomorphy, based on a phylogeny that includes a third of the c. 430 species in the genus. They conclude that two sections in Primula thought to contain monomorphic, nonherkogamous descendants of lineages that diverged before the origin of reciprocal herkogamy in the group (sections Proliferae and Sphondylia) are instead descended from ancestors with reciprocal herkogamy and are not a useful starting point for modeling the origin of heterostyly in the group (e.g. as performed by Al Wadi & Richards, 1993). Pérez-Barrales et al. demonstrate on the phylogeny that a shift from style dimorphism to distyly in Narcissus sect. Apodanthi is not preceded or succeeded by a shift to an intramorph incompatibility system from the pre-existing self-incompatibility system.

New methods for inferring the path to heterostyly

We suggest that future integration of phylogenetic analyses into the assessment of allelic diversity in genes involved in heterostyly will provide unique insights into its origin. In Primula, the suite of dimorphic floral features and the intermorph incompatibility system are thought to be controlled by at least three tightly linked genes (Dowrick, 1956). The style length, stigmatic papillae length and female mating type are thought to be controlled by G, the pollen size and male mating type by P (or two separate genes Pp and Pm, respectively; Kurian & Richards, 1997) and the anther height by A. ‘Pin’ plants are homozygous recessive for these genes (gpa/gpa); ‘thrum’ plants are heterozygous (GPA/gpa; Lewis & Jones, 1992). Manfield et al. (2005) recently characterized an 8.8-kb DNA genomic clone that appears to be sitting close to the PA end of the linkage group, and McCubbin et al. (2006) found 11 classes of differentially expressed genes in pins and thrums that do not appear to be linked to GPA. In this issue Shore et al. (pp. 539–551) discuss similarities in the genetics underlying heterostyly in Primulaceae and Turneraceae, and describe the recent progress on similar molecular genetic characterizations underway in Turnera.

We propose that our current genetic and phylogenetic understanding implies that the dominant and recessive alleles for each gene in the heterostyly linkage group in Primula and other groups with multiple, closely related heterostylous members (e.g. Turnera), will prove to be ‘deep persistent polymorphisms’ (Igic et al., 2006). These are polymorphisms that coalesce deeper in the organismal phylogeny than the origin of a single species harboring them (Fig. 1). Inference that a polymorphism seen in a group of species has a common origin using traditional ancestral state inference (e.g. reciprocal herkogamy in Primula; Mast et al.) is a prediction that can be tested by inferring the allele trees of genes responsible for the phenotype in the linkage group. The dominant alleles of an underlying gene in the linkage group sampled across the phylogeny should be more closely related to one another than any one is to the recessive alleles in the same species, and the converse is expected for the recessive alleles (Fig. 1a,b). In addition, if dominant and recessive alleles of other genes in the heterostyly linkage group prove to coalesce at different depths in the phylogeny (Fig. 1a,c), it will provide a new source of evidence for the ordering of steps in the build-up of polymorphisms associated with heterostyly (Fig. 1e). This has proven to be a powerful approach for determining the common origin of the incompatibility system in Solanaceae, where self-compatible and self-incompatible species are interdigitated on the phylogeny and traditional ancestral state inference has given ambiguous results (Igic et al., 2006).

Figure 1.

Heterostyly and associated polymorphisms as deep persistent polymorphisms in genes in the heterostyly linkage group. Organismal phylogenies for a group of heterostylous species (species A–E) and an outgroup (OUT) are provided in outline in (a), (c) and (e), with allele phylogenies for hypothetical genes in the heterostyly linkage group (H and I) given within the outlines in (a) and (c), respectively. The inferred allele phylogenies for genes H and I are given in (b) and (d), respectively. The point of coalescence of dominant and recessive alleles in genes of the heterostyly linkage group can be used to order the origins of polymorphisms (e), further clarifying the evolutionary path to heterostyly.

‘… correlation between polyploidy and homostyly might be more a function of hybridization than of polyploidization per se, because the higher rates of recombination associated with the merging of highly differentiated genomes, as in allopolyploids, would favor the disruption of the heterostyly linkage group.

Paths leading away from heterostyly

Comparative studies on the function of polymorphisms, individually and in combination, rely on variation in the presence/absence of floral polymorphisms among closely related species, among populations of a single species, or in experimental manipulations. Such comparative studies benefit from an understanding of evolutionary correlates to both the gain (as discussed above), and the loss, of polymorphisms. In this issue of New Phytologist, Pérez-Barrales et al. demonstrate that the loss of self-incompatiblity in Narcissus section Apodanthi is correlated with the shift from a class of pollen vectors (solitary bees), which carries pollen loads sufficient to fertilize the mean number of ovules in a flower on localized portions of their body, to those that do not (syrphids and lepidopterans). Hodgins & Barrett report both theoretical and empirical results, which suggest that interactions between mating patterns and female fertility lead to loss of the mid-level style height morph from some populations of the tristylous N. triandrus.

Guggisberg et al. (this issue; pp. 617–632) examine the relationship between the complete loss of polymorphisms (the production of a homostylous population or species, presumably by recombination of the heterostyly linkage group and fixation of the gPA alleles) and polyploidization in Primula sect. Aleuritia subsect. Aleuritia. They find that the switch from distyly to homostyly is significantly correlated with the switch from diploidy to polyploidy, with the exception of the distylous polyploid P. borealis. The polyploid homostyles of Primula sect. Aleuritia subsect. Aleuritia tend to occur at higher latitudes than diploid distylous species (Richards, 2002). Expanding upon the secondary contact model of Stebbins (1984), Kelso (1992) proposed that glacial advancement during the Quaternary climatic cycles would have fragmented diploid, heterostylous populations of arctic primroses, which, following a period of differentiation, came into secondary contact giving origin to hybrid, polyploid populations. Shifts of pollinators faunas associated with oscillating climatic conditions would have favored the establishment of the self-fertile, polyploid homostyles. The phylogenetic and haplotype network results of Guggisberg et al. essentially confirm the predictions of the secondary contact model, and they conclude that selection for reproductive assurance probably favored the independent origin and establishment of allopolyploid homostyles at the margins of retreating ice sheets, possibly in response to mate and pollen limitation.

In contrast, Shore et al. conclude that there is no clear relationship between polyploidy and homostyly in Turneraceae. In this regard, it is interesting to note that, while all homostyles appear to be allopolyploid in the T. ulmifolia complex, the distyles are either diploid or autotetraploid (Barrett & Shore, 1987). This suggests that the correlation between polyploidy and homostyly might be more a function of hybridization than of polyploidization per se, because the higher rates of recombination associated with the merging of highly differentiated genomes, as in allopolyploids, would favor the disruption of the heterostyly linkage group (Guggisberg et al., 2006). Further cytological and molecular genetic studies into genomic changes that typically follow distinct types of polyploidization (autopolyploidization, segmental allopolyploidization and allopolyploidization), and a more complete understanding of the degree of differentiation between the parental genomes that contributed to the formation of polyploid homostyles, will help to clarify the roles of recombinational and epigenetic changes associated with their origin.