DISCUSSION

Variation at floral regulatory gene loci: The APETALA3 and PISTILLATA loci of A. thaliana play central roles in floral organ development (Jacket al. 1992; Goto and Meyerowitz 1994). Assessment of the levels of variation at AP3 and PI, as well as previous work on the floral meristem identity gene CAULIFLOWER (Purugganan and Suddith 1998), indicates that these developmental regulatory genes possess appreciable levels of nucleotide variation. The overall estimates of intraspecific variation at these two organ identity genes (π = 0.005–0.0065) are comparable to the estimate for CAULIFLOWER (π = 0.0070). The levels of diversity of these three regulatory genes are also similar to nucleotide variation estimates for two structural loci (ChiA and Adh) in Arabidopsis (see Table 3).

Population subdivision and expansion of A. thaliana: The extent and patterning of variation in loci are governed by the interplay of selective and demographic processes that shape the diversity of genes and genetic pathways. Evolutionary forces that affect the entire genome leave similar patterns across different unlinked loci and provide information on population history, structure, and dynamics. The specific details of population structuring in this plant species are crucial to understanding the nature of the evolutionary forces that shape diversity at these floral regulatory loci.

A recent study investigated variation in three nuclear loci within and between Arabidopsis populations using restriction site variation (Bergelsonet al. 1998). Comparison of this study with sequence data on AP3 and PI as well as three previously studied, unlinked loci–the CAL (Purugganan and Suddith 1998), ChiA (Kamabeet al. 1997), and Adh (Hanfstinglet al. 1994; Innanet al. 1996) genes–permits us to evaluate common patterns of variation that presumably arise from their shared evolutionary histories within the A. thaliana genome. Table 3 summarizes three features characteristic of most Arabidopsis genes that have been thus far analyzed: (1) excess of low-frequency nucleotide polymorphisms, (2) excess of within-species replacement site variation, and (3) evidence for allelic dimorphism.

For both the AP3 and PI floral regulatory genes, the Tajima and Fu and Li tests reveal significantly negative D estimates, indicating a bias toward rare polymorphisms at these loci. A significant excess of singleton changes is also observed for ChiA (Kamabeet al. 1997), while CAL shows negative D* statistics that are significant at the 10% level (P < 0.057 in the Fu and Li test; Purugganan and Suddith 1998). Excess in low-frequency alleles can arise from hitchhiking effects that accompany selective sweeps, which reduce linked sequence variation at specific loci (Aquadro 1997). The regulatory genes AP3 and CAL (Purugganan and Suddith 1998), however, as well as the structural loci ChiA (Kamabeet al. 1997) and Adh (Innanet al. 1996), all possess intermediate frequency polymorphisms that differentiate allele classes; this variation would presumably have been eliminated by genetic hitchhiking during species-wide fixation of selected sites.

The excess in low-frequency polymorphisms may be explained by the background selection hypothesis (Charlesworthet al. 1993), which suggests that rare variants may be present in population samples as a result of linkage to deleterious mutations. This effect should be marked in regions of low recombination or in primarily selfing species such as Arabidopsis. The background selection hypothesis, however, does not explain the excess in within-species replacement polymorphisms common to most of the Arabidopsis nuclear loci that have been examined (see Table 3).

Population subdivision and/or recent population expansion are mechanisms that may explain, however, both the excess in low-frequency allelic variation for Arabidopsis nuclear loci and the excess of within-species replacement polymorphisms (Aquadro 1997). Despite occasional cross-pollination, low outcrossing rates in A. thaliana may have resulted in population subdivision across the species range as a result of limited gene flow between inbred subpopulations (Allardet al. 1968; Charlesworthet al. 1997; Nordborg 1997). Inbreeding plant populations generally exhibit low within-population heterozygosities, given the low effective population sizes of primarily selfing organisms (Allardet al. 1968). Reduced levels of within-population variation have been observed for allozyme (Abbott and Gomes 1989), microsatellite (Todokoroet al. 1996; Kuittinenet al. 1997), and mitochondrial and nuclear restriction site studies (Bergelsonet al. 1998) in Arabidopsis populations. Inbreeding plants, however, also display elevated levels of between-population differentiation as alternate genotypes are fixed in different populations by both drift and selection (Allardet al. 1968). The excess of low-frequency polymorphisms observed in this species-wide survey of AP3 and PI nucleotide variation in A. thaliana may result from overdispersed sampling of distinct, inbred populations.

Differentiated, inbred populations of low effective population size would also explain the widespread occurrence of excess intraspecific replacement polymorphisms in nuclear loci of A. thaliana. The McDonald-Kreitman tests indicate that both AP3 and PI, as well as CAL (Purugganan and Suddith 1998) and ChiA (Kamabeet al. 1997), have significantly high levels of intraspecific replacement polymorphisms compared to levels of interspecific divergence. The nearly neutral theory suggests that the intensity of purifying selection is relaxed and slightly deleterious mutations may be fixed at a higher rate in small populations (Ohta 1992). If the ecotypes sampled represent distinct, inbred populations of small Ne, then this would suggest that high levels of observed species-wide intraspecific replacement polymorphisms result from slightly deleterious mutations that are confined to individual ecotype populations.

Recent, rapid population expansion (Cummings and Clegg 1998) could also account for the excess in low-frequency and replacement site polymorphisms observed in A. thaliana. It is likely that both population subdivision and expansion have played roles in structuring the molecular diversity within this selfing plant species. Because population subdivision is associated with selfing plant species, one may expect that some of the features observed in A. thaliana may be shared by other inbreeding plant groups. There are indications that at least some of the features of nucleotide variation found in Arabidopsis characterize nuclear gene diversity in other selfing plants as well. The A. gemmifera Adh gene, for example, also displays a significant excess of within-species low-frequency polymorphisms and intraspecific replacement variation (Miyashitaet al. 1996). Population subdivision in this selfing species has also been invoked to explain the patterning of diversity observed at this locus. The Adh locus from the self-fertilizing H. vulgare also displays a significant excess of within-species replacement polymorphisms and a greater (though not significant) frequency of singleton variation (Cummings and Clegg 1998).

Dimorphic variation in Arabidopsis nuclear genes: Four of the five Arabidopsis genes examined display varying degrees of allelic dimorphism (Innanet al. 1996; Kamabeet al. 1997; Purugganan and Suddith 1998). The observed dimorphism for most genes is based on only a few sites; in the AP3 gene, for example, class A and B alleles are differentiated by only four linked polymorphisms in intron 4. The widespread occurrence of allelic dimorphism in Arabidopsis nuclear loci has been ascribed to either introgression from related species or the breakdown of barriers between two previously isolated populations (Hanfstinglet al. 1994; Kamabeet al. 1997).

It is unclear what mechanisms have contributed to the continued maintenance of polymorphisms that differentiate allelic classes in A. thaliana nuclear genes. Association analysis reveals that the four polymorphisms that differentiate the AP3 allele classes are in complete linkage disequilibrium with each other. In CAL, the polymorphic sites that differentiate two allelic classes are spread out over a 260-bp region and include a distinctive replacement polymorphism that may account for phenotypic variation in floral homeotic function. Linkage among sites, coupled with the low outcrossing rates for A. thaliana, may have led to the persistence of these sites in linkage disequilibrium.

Selective forces may also operate to maintain these linked polymorphisms. The background selection model suggests that balancing selection in selfing plants such as A. thaliana would lead to linkage among neutral variants and selected sites (Nordborget al. 1996). A similar pattern should also be observed under local selection in subdivided populations (Charlesworthet al. 1997). Given their location within introns, however, the selective advantages of the linked changes responsible for the dimorphism at AP3 are not immediately apparent. Studies on evolution of intron sites suggest that some selective pressures related to the formation of pre-mRNA hairpin structures may be involved in gene regulation (Kirbyet al. 1995; Parschet al. 1997). RNA secondary structure analysis indicates that a hairpin associated with these AP3 polymorphic sites may possess different melting temperatures in the two allele classes (M. D. Purugganan, unpublished observations). Moreover, molecular studies in the related MADS-box floral homeotic gene AGAMOUS reveal that regulatory sites within introns may modulate expression of this regulatory gene (Sieburth and Meyerowitz 1997). No comparable studies for AP3 have been reported, although it is possible that cis-acting regulatory sequences may also be present in the introns of this floral homeotic gene.

Evolutionary dynamics of functional domains within APETALA3 and PISTILLATA: The distribution of replacement variation between distinct functional domains of APETALA3 and PISTILLATA suggests that different regions of these regulatory proteins experience contrasting evolutionary pressures. These differing evolutionary forces between domains are reflected in the distribution of replacement and synonymous changes between gene regions (Templeton 1996). On the basis of a comparison of the patterning of coding region variation among several genes, the expectation for A. thaliana is (1) an excess of within-species replacement polymorphisms as a result of an increased tolerance for slightly deleterious mutations in small, inbred populations, and (2) a lower ratio of interspecific replacement-to-synonymous differences arising from purifying selection that prevents fixation of polymorphisms across species.

Contingency analyses can incorporate functional information and partition the mutations between distinct gene domains (Riechmann and Meyerowitz 1997). We can examine whether the pattern of replacement to synonymous variation is similar between the core and noncore domains of A. thaliana AP3 and PI genes (see Table 2). The patterns of replacement and synonymous differences between A. thaliana and A. lyrata indicate that both core and noncore domains of APETALA3 are constrained against amino acid changes. In these two domains, levels of replacement differences are lower than synonymous ones, suggesting that purifying selection prevents the fixation of amino acid polymorphisms.

This trend is reversed, however, in the distribution of within-species polymorphisms. The noncore region of AP3 has 15 replacement and only 5 synonymous polymorphisms, which suggests that this domain can tolerate a high level of possibly slightly deleterious amino acid changes. Contingency tests on levels of replacement and synonymous changes in the AP3 noncore region reveal the expected (standard pattern) of significant excess in intraspecific replacement polymorphisms. The AP3 core region, however, shows a lower level of replacement to synonymous polymorphisms (5 replacements to 4 synonymous); this ratio is not significantly different from observed interspecies differences. Given the low overall numbers of polymorphisms and differences in the AP3 core region, the contingency test applied may have restricted power. Otherwise, this test does indicate that the core region of AP3, which performs central functions in this floral regulatory protein, is less tolerant of replacement mutations than the noncore domain.

This contrast in evolutionary dynamics between core and noncore regions is also exhibited by the PISTILLATA protein-coding region. Contingency testing suggests that in PI it is the core region that is behaving according to the expected standard pattern–high levels of intraspecific replacement polymorphisms and reduced levels of interspecific amino acid change. The noncore region, however, deviates from this expected pattern and indicates that a large number of replacement substitutions have been fixed in the divergence between the A. thaliana and A. lyrata PI locus. The difference between the pattern of interspecific divergence in core and noncore regions is significant (P = 0.042 using Fisher's exact test).

The interspecific amino acid changes in the PI noncore region are centered primarily in the C-terminal region, which previous molecular evolutionary analyses have indicated is the most rapidly evolving structural domain of plant MADS-box proteins (Puruggananet al. 1995). This high level of amino acid change could arise from a relaxation in the functional constraint at PISTILLATA after the separation of these two Brassicaceae species or adaptive fixation of selectively favored amino acid changes. AP3 and PI proteins form a heterodimer necessary for nuclear localization and DNA-binding (Riechmannet al. 1996), and the differences in the evolutionary dynamic between domains may arise as a result of some degree of functional complementation between proteins. It should be noted that there are no dramatic differences in the floral morphologies between A. thaliana and A. lyrata, although the petals of the latter are about twice as large (3–4 mm in A. thaliana vs. 7–10 mm in A. lyrata). The growth trajectories of petals and stamens should also differ between the two taxa as the selfing syndrome evolved in A. thaliana.

The molecular population genetics of the floral homeotic genes APETALA3 and PISTILLATA provide insights into both the population structure of A. thaliana and the dynamics of regulatory gene variation. Population structuring could account for the excess of both singleton variation and replacement polymorphisms in species-wide surveys, and the subdivision of A. thaliana populations determines, in part, the fate of mutational variation at these regulatory genes. The pattern of variation indicates that functional domains evolve differently from one another within these two floral regulatory genes, which suggests that members of the floral developmental pathway are subject to distinct evolutionary forces.