Introduction

Plant mating systems incorporate both genetic and ecological factors that affect reproduction and are primary determinants of the level of genetic variation within a population and how this variation is distributed among individuals. They are the result of evolutionary pressures both for and against inbreeding, such that there is a balance between genetic load, fitness and fecundity. Specifically, mating systems are influenced by factors such as plant size and floral display, flowering phenology, pollination syndrome, and pre- and post-zygotic incompatibility (Stephenson, 1981; Goldingay & Carthew, 1998). They can also be affected by population parameters such as population size (Buza et al., 2000), density (Watkins & Levin, 1990) and isolation (Young & Brown, 1999).

Understanding the limitations that mating systems place on demographic, genetic and evolutionary dynamics is of particular interest for species endangered by habitat loss, degradation and fragmentation. In such cases, rapid changes in population size, density, isolation and pollinator guilds may mean that species mating systems are maladapted for their current circumstances. For example, for self-compatible but normally outcrossing species, increased self-fertilization will lead to rapid expression of genetic load and reduced fitness. Alternatively, for self-incompatible species, if reduced population size is accompanied by the loss of S-allele richness then mate availability and fecundity may be limited (DeMauro, 1993; Vekemans et al., 1998). Recently, it has been shown that a combination of such factors is often implicated in reducing the viability of small populations of endangered plants (Oostermeijer, 2000; Young et al., 2000).

This study used controlled pollinations and allozyme markers to investigate the mating system and genetic structure of Grevillea iaspicula McGill. (Proteaceae). Grevillea iaspicula is a species for which population persistence is of critical concern, as the few remaining populations have become fragmented due to land clearing for agriculture and population sizes have declined owing to grazing, flooding and possibly competition with invasive species (Briggs & Leigh, 1990; Butler et al., 1991). Four main questions are addressed: (1) Is G. iaspicula self-compatible? (2) Do mating patterns change with population size and isolation? (3) How much genetic variation is present in G. iaspicula populations? (4) How much genetic differentiation is there among populations?

Materials and methods

Study species

The Australian shrub Grevillea iaspicula (Proteaceae) is listed as nationally endangered and is known from eight populations occurring in remnant eucalypt vegetation on rocky limestone outcrops (ANZECC, 1999). The populations surround the town of Wee Jasper, in south-eastern New South Wales (Fig. 1) and range in size from 20 to several hundred individuals. The vegetation at all but one of these sites has been highly disturbed by the conversion of land for stock grazing. It is predicted that the species will disappear from natural environments within 10–20 years if present land use and other threats, such as grazing and competition with invasive exotics, persist (Briggs & Leigh, 1996).

Figure 1
figure 1

Distribution of Grevillea iaspicula populations around the townships of Wee Jasper and Burrinjuck, New South Wales. Populations are labelled 1–8: 1, Punchbowl; 2, Slyney 1; 3, Slyney 2; 4, Ashley; 5, Carey (inaccessible); 6, Cathles (inaccessible); 7, Dutton’s Bluff; and 8, Burrinjuck.

Full size image

The plants are straggling to erect bushy shrubs, 1.2–2.5 m in height (occasionally up to 4 m) with light green, leathery leaves. They generally have terminal, or occasionally subterminal inflorescences, that are often pendant and always branched. There are about 12 cream-yellow flowers per branch with red protruding styles. Fruits are thin-walled ovoid follicles about 10–15 mm long that split opposite the style to release, generally, two grey-brown arillate seeds about 8 mm in length. McGillivray (1993) provides a full taxonomic treatment.

Given the flower coloration, G. iaspicula is probably bird pollinated (Olde & Marriott, 1995). Field observations have identified five honeyeater species visiting the flowers: Lichenostomus penicillatus (White-plumed Honeyeater), L. leucotis (White-eared Honeyeater), L. chrysops (Yellow-faced Honeyeater), Acanthorhychus tenuirostris (Eastern Spinebill), and Phylidonyris novaehollandiae (New Holland Honeyeater) (personal observations). The European honeybee (Apis mellifera) also visits the species.

Controlled pollinations

In Grevillea species, pollen adheres to modified style-ends (pollen presenters) following anther dehiscence. As a result, many species are protandrous (Carolin, 1961; Johnson & Briggs, 1975). Therefore, artificial manipulation of flowers is relatively easy as pollen can be removed from the pollen presenter and subsequently applied to the stigmatic papillae once receptive. Inflorescences of G. iaspicula may contain up to 10 branches (McGillivray, 1993), with flowers on a branch generally opening after flowering on the preceding branch has finished. Ovaries of fertilized flowers swell within five to seven days after pollination, whereas unfertilized flowers wither and fall from the rachis at this stage (Harriss & Whelan, 1993). Fruit development takes between six and 14 weeks, depending on environmental conditions.

Pollinations were conducted from August–October 1998 and seed were harvested in January 1999. The experimental design consisted of two populations (Slyney 1 and 2), five plants per population and five treatments applied at the inflorescence level (see Table 1 for a description of these treatments). Each treatment was replicated five times within a plant. Within each inflorescence, all flowers from one, two or three branches were pollinated; other branches were removed, resulting in an average of 20 flowers manipulated per inflorescence. The total number of inflorescences in the experiment was 250. Following seed set, both seed and fruit were collected in mesh bags.

Table 1 Pollination treatments applied to inflorescences of Grevillea iaspicula. Inflorescences in closed treatments were ‘bagged’ prior to opening and subsequently to exclude all floral visitors. All pollen manipulation was conducted using a paintbrush. Outcross treatments were crossed using pollen from two other plants within the population
Full size table

The effects of pollination treatment, population and maternal plant were assessed in relation to fruit and seed set, using an analysis of variance (ANOVA), with pollination treatment as a fixed effect and population and maternal plant as random effects, using the SAS General Linear Models Procedure for uneven sample sizes (PROC GLM) at P < 0.05 (SAS Institute, 1990).

Fruit and seed set

During the 1998 fruiting season, between two and 19 inflorescences with swollen ovules were bagged on 7–15 plants from each of five populations. One, population (Burrinjuck) was excluded because the combined effects of drought and grazing severely limited its reproductive output and two other populations (Carey and Cathles) were excluded due to inaccessibility. In January 1999, after seed were shed, bags were collected and the number of mature fruit, the number of filled seed and the number of flower scars per inflorescence were recorded. Differences in fruit per flower and filled seed per flower were analysed as a nested ANOVA, with population and maternal plant as random effects, using the SAS PROC GLM at P < 0.05 (SAS Institute, 1990). All populations had damage to the mesh bags used for seed collecting as a result of bird predation. Effects of this predation were analysed as a nested ANOVA with predation (bag torn or intact) as the fixed effect and population and maternal plant as random effects, using SAS PROC GLM as above.

Allozyme markers

Seed were collected from up to eight inflorescences (depending on availability) from six to eight adult plants from the same five populations of G. iaspicula as were assessed for fruit and seed set. Allozyme assays using starch gel electrophoresis were conducted on 8–15 seed sampled as evenly as possible across inflorescences. Seed were allowed to imbibe for 12 h, seed coats were removed and individual seed were ground in five drops of pH 9.0 borate buffer containing 1 mg mL–1 dithiothreitol (DTT) and 20 mg mL–1 polyvinylpyrrolidone (PVP, mol. wt.=40 000). The resulting solution was absorbed onto 6 × 4 mm filter paper wicks and inserted into horizontal 9.6% starch gels, with morpholine citrate buffer. The gels were run at 48 mA for 5 h and were then sliced horizontally into three 2 mm slices. Each slice was stained for a different enzyme. The enzyme systems examined were shikimate dehydrogenase (SDH, E.C. 1.1.1.25), malate dehydrogenase (MDH, E.C. 1.1.1.37), phosphoglucomutase (PGM, E.C. 5.4.2.2), triose-phosphate isomerase (TPI, E.C. 5.3.1.1), isocitrate dehydrogenase (IDH, E.C. 1.1.1.41) and phosphogluconate dehydrogenase (PGD, E.C. 1.1.1.43). In total, eight putative loci were examined.

Estimation of mating system and genetic diversity parameters

Estimates for multilocus and single locus outcrossing rates (tm and ts, respectively), correlation of outcrossed paternity (rp; Ritland, 1989) and maternal fixation indices (FIS) were calculated from allozyme data using MLTR version 0.9 (Ritland, 1994). Standard errors for these estimates were based on 100 bootstraps with resampling among maternal plants. The divergence between estimates of allele frequencies in the pollen pool and their corresponding estimates in the maternal pool was estimated using the Pollak estimator (Fk; Waples, 1989) following Young & Brown (1999).

Polymorphism (P), allelic richness (A), expected and observed heterozygosity (He and Ho, respectively) were estimated for all populations. Genetic divergence among populations was measured using Wright’s (1951) FST. Nei’s (1972) genetic distance was also calculated among all possible pairs of populations. A Mantel test (Mantel, 1967) was conducted to assess correlation between the genetic and geographical distances among populations. All diversity and differentiation statistics were calculated using POPGENE (Yeh et al., 1998). Gene flow among populations, as measured by Nm (the number of migrants per generation), was estimated using Slatkin & Barton’s (1989) modification to Slatkin’s (1985) private alleles method for a sample size of 50 individuals.

Results

Controlled pollinations

Both fruit and seed set differed significantly among the pollination treatments (F4,36=50.19, P < 0.0001 for fruit and F4,36=8.57, P < 0.0001 for seed). Open and outcross treatments were comparable and exhibited significantly higher fruit and seed set than the other three treatments for which seed set was uniformly close to zero (Fig. 2).

Figure 2
figure 2

Mean fruit (solid bars) and seed (open bars) set per Grevillea iaspicula inflorescence under different pollination treatments. Errors bar represents one standard error.

Full size image

Populations did not differ significantly for either fruit set or seed set; nor was there any population- by-treatment interaction. There were significant effects of maternal plant for fruit set (F9,36=2.21, P < 0.05) but not for seed set.

Fruit and seed set

There was no significant difference in open-pollinated fruit and seed set among the five populations studied. In general, fruit and seed production per flower were low (Table 2). All populations experienced high levels of damage to seed collecting bags (range of 0.56–0.98) and damaged bags contained significantly fewer seed than undamaged ones (F1,79=49.92, P < 0.0001). However, the damage did not significantly affect the number of mature fruit collected. This is reflected in the higher fruit numbers, as opposed to seed. There were significant among-plant differences in both fruit and seed set for some populations (Table 2).

Table 2 Levels of seed and fruit set for G. iaspicula populations
Full size table

Genetic variation and differentiation

There was 100% polymorphism across the eight allozyme loci assayed although the degree of polymorphism within populations differed (P=50–100%; Table 3). Allelic richness was moderate to high (Table 3) for all populations regardless of size, ranging from A=1.63–2.50 alleles per locus, with the lowest value observed in the Dutton’s Bluff population. Additionally, both observed and expected heterozygosities were high (mean Ho=0.30 and mean He=0.27; Table 3), though again Dutton’s Bluff was the least variable in this respect.

Table 3 Genetic variation and differentiation statistics for Grevillea iaspicula populations based on eight allozyme loci
Full size table

Wright’s FST was high at 0.204 (SE=0.040). Genetic distances among populations were also generally high, with a minimum of D=0.04 (between Punchbowl and Slyney 2) and a maximum of D=0.32 (between Slyney 1 and Ashley). Average genetic distances of the populations from all other populations are presented in Table 3. The estimate of interpopulation gene flow, based on the frequencies of nine private alleles, was very low (Nm=0.136). There was no significant relationship between geographical and genetic distances between populations.

Mating system parameters

Multilocus outcrossing rates were consistently high for the five populations assayed (tm=0.96–1.00) regardless of population size (Table 4). Values of tm − ts were all low (Table 4); this is indicative of low levels of biparental inbreeding. Maternal and progeny fixation indices were uniformly low (Table 4).

Table 4 Mating system parameters of Grevillea iaspicula populations: multilocus outcrossing rate (tm), an indicator of current levels of biparental inbreeding (tm − ts), levels of correlated paternity (rp), neighbourhood size (1/rp; Sun & Ritland, 1998), inbreeding coefficients for both the maternal and progeny cohorts (FIS), and the divergence of allele frequencies between the pollen and maternal pools (Fk)
Full size table

The correlations of outcrossed paternity estimates were all significantly greater than zero (rp=0.31–0.54; Table 4) suggesting that a large proportion of open-pollinated seed are full-sibs. Neighbourhood sizes based on 1/rp (Sun & Ritland, 1998) averaged 2–3 plants. Overall, Fk estimates were relatively low, suggesting little deviation between allele frequencies in the pollen pool and their corresponding frequencies in the maternal pool, though there was some variation among the populations.

Discussion

Grevillea species have highly variable mating systems. For example, G. linearifolia is predominantly self-incompatible but G. longifolia is self-compatible (Hermanutz et al., 1998). In this study, the lack of both fruit and seed set from self-pollinations and the uniformly high multilocus outcrossing rates observed indicate that G. iaspicula is self-incompatible. Gametophytic incompatibility is the only form of self-incompatibility that has so far been reported in the Proteaceae (Kalinganire et al., 2000) and families generally exhibit a single form of incompatibility (Matton et al., 1994). Therefore, it is likely that gametophytic incompatibility is also the system operating in G. iaspicula.

The observation of low maternal fixation indices for all G. iaspicula populations suggests that plants within the populations are not inbred. The negative fixation values for progeny may be indicative of disassortative mating that might be driven by the incompatibility system. However, the large standard errors associated with the progeny estimates suggest they should be interpreted cautiously.

The estimates of correlated paternity were substantially higher than those observed for many other species. For example, the extent of correlated paternity between flowers on the same plant of Mimulus guttatus was rp=0.20 (Ritland, 1989), rp=0.26 within fruits of Eucalyptus rameliana (Sampson, 1998), rp=0.20 for Daviesia mimosoides and rp=0.25 for D. suaveolens (Young & Brown, 1998) and rp=0.19 for Centaurea solstitialis, although levels up to rp=0.64 were reported in some populations (Sun & Ritland, 1998). However, they were similar to estimates from another bird pollinated Proteaceae, Lambertia orbifolia, for which rp=0.39 (Coates & Hamley, 1999).

High levels of correlated paternity and low derived estimates of neighbourhood size may arise in two ways. First, they could be due to mating between pairs or small groups of plants. This has also been suggested for L. orbifolia where bird movements are generally less than five metres and restricted to only two or three mature plants (Coates & Hamley, 1999). Alternatively, there may be a few males within populations that are particularly successful at siring seed. These mating scenarios have very different implications for effective population size. However, since low Fk estimates are also observed in G. iaspicula populations, the former explanation is more likely in this instance (Young & Brown, 1999). In other words, since G. iaspicula populations show little allozyme divergence between pollen and maternal pools, it is unlikely that a single plant is, or few plants are, dominating paternity.

For G. iaspicula, seed set is pollinator dependent, since the exclusion of pollinators from inflorescences (by bagging) resulted in negligible seed set. Moreover, since levels of seed set were comparable between the open and outcross pollination treatments, it is unlikely that either pollen availability or pollinator service is limiting in these two populations (Slyney 1 and 2). Given that there were no interpopulation differences in fruit or seed set it seems likely that this is generally the case and that low mate availability, suggested by high rp, has not yet translated into reproductive failure. The significant interplant variation in fruit and seed set that was observed for several of the populations might reflect differences arising from several factors that are known to affect plant mating systems, such as plant size and isolation, floral display and incompatibility genotype.

The very low levels of fruit and seed set per flower in G. iaspicula is a phenomenon that is apparently common within the Proteaceae (Ayre & Whelan, 1989; Harriss & Whelan, 1993). For this study, fruit set is likely to be a more reliable measure of reproductive output as it was less affected by damage to mesh bags associated with seed predation by parrots. However, as flowering can continue for up to six months in most Grevillea species, with seed setting continuously during this period (Olde & Marriott, 1995), it is plausible that during such an extended reproductive season plants can produce large quantities of viable seed regardless of low flower: fruit ratios.

The genetic diversity estimates of percentage polymorphic loci, average number of alleles per locus and observed and expected heterozygosities for G. iaspicula are high in comparison to such estimates for other proteaceous species and other taxa with similar life-history characteristics (Hamrick & Godt, 1990). For example, in other woody proteaceous species, He=0.11 for G. robusta (Harwood et al., 1997), He=0.16 for Hakea carinata (Starr & Carthew, 1998), expectations for Lomatia polymorpha and L. tinctoria are both He=0.17, while L. tasmanica has no detectable variation at all (Lynch et al., 1998). Ignoring L. tasmanica, as it is believed to be clonal (Lynch et al., 1998), the other species are apparently genetically more depauperate than the endangered G. iaspicula. In part, this may be due to both the breeding system and history of the species. For example, H. carinata populations are substantially selfing, and small and isolated populations appear to be a long-term evolutionary condition in this species (Starr & Carthew, 1998). On the other hand, G. robusta is a predominant outcrosser but some plants do produce a small proportion of self-fertilized seed (Harwood et al., 1992) and these apparently low levels of selfing may still be sufficient to affect heterozygosity levels within G. robusta populations. Indeed, FIS estimates for G. robusta (FIS=0.121–0.152) indicate significant levels of inbreeding (Harwood et al., 1992, 1997).

It appears that current population size is unrelated to the level of genetic variation within G. iaspicula populations. This may be a sampling effect caused by sampling only a few populations that do not differ greatly in size; however, the populations chosen do represent the majority of all G. iaspicula populations and include the current size range. The lack of relationship may also be because Grevillea species are relatively long-lived (generally 10–30 years; Olde & Marriott, 1995) and the populations may have only been recently reduced in size, so that insufficient generations have passed for any effect on allozyme diversity to be detected. Therefore, the extant stand may still represent a greater part of the natural genetic variation. Additionally, the likely presence of a soil-stored seed bank may help to buffer the effects of population decline, in terms of genetic variation, at least for the short term (Levin, 1990).

Interpopulation genetic distances are large, considering that the populations are often separated by only a few kilometres. Similarly large genetic distances have been found between populations of another proteaceous shrub, Hakea carinata (Starr & Carthew, 1998), though this was associated with a relationship between geographical distance and genetic differentiation across a range of about 500 km. Large genetic distances between populations are indicative of genetic isolation and are possibly due to founder effects, differential selection or limited gene flow among populations. Slatkin’s (1985) rare alleles procedure indicated that there was indeed limited gene flow between G. iaspicula populations.

In summary, populations of Grevillea iaspicula show significant interpopulation genetic differentiation that is associated with low levels of gene flow among them. Mating within populations appears to be similarly restricted, with neighbourhood sizes of 2–3 plants, though populations are not yet experiencing reproductive limitation. Despite the apparent restrictions on mating and immigration, genetic diversity both at the individual and population level is high. Maintenance of heterozygosity in this species is perhaps not surprising, as it may well be a function of disassortative mating arising from the self-incompatibility system. The high allelic richness is more difficult to explain. One possibility is that there is a substantial seed bank that is acting as a temporal reservoir of genetic variation, providing small populations with diversity that was present in previous generations. Alternatively, though not exclusively, current populations may have only recently become small and therefore too few generations have passed for significant loss of alleles to have occurred due to genetic drift.