|Home | About | Journals | Submit | Contact Us | Français|
When conserving rare plant species, managers are often faced with small and/or isolated populations displaying low levels of sexual reproduction and genetic variation. One option for reinvigorating these populations is the introduction of genetic material from other sites, but in some cases fitness may be reduced as a result of outbreeding depression. Here the pollination biology of the rare shrub Grevillea repens is studied across its natural range and reproductive responses following cross-pollination among populations are examined to determine factors that may be limiting sexual reproduction and the potential for genetic rescue.
Pollen manipulation treatments (self-, autogamous self-, cross- and open pollination) were applied to flowers to examine the breeding system and fruit and seed production in five populations of G. repens. Pollen production, presentation and viability were investigated and interpopulation crosses of increasing genetic distance performed among the populations.
The study species is self-incompatible and displayed very low natural seed set over two seasons, due partly to low pollen viability in one of the populations. Within-population crossing increased fruit and seed production at some sites, indicating pollinator limitation. Interpopulation crosses further increased reproductive output in one population, suggesting mate limitation, and for this site there was a positive relationship between genetic distance among populations and the size of genetic rescue benefits. However, in other populations there was a decrease in fruit and seed set with increasing genetic distance.
The results highlight that management strategies involving interpopulation crosses can improve reproductive output in small, isolated populations of rare plants, but guidelines need to be developed on a population by population basis.
Small and isolated populations of rare plants often display reduced fitness relative to larger populations (Allee effects). This may be expressed as reduced reproductive output, seedling performance or pollen viability (e.g. Willis, 1993; Ågren, 1996; Fischer et al., 2003). Several genetic and ecological factors may be involved. These include the expression of inbreeding depression as a result of increased mating among close relatives (Fischer et al., 2003; Willi et al., 2005), increased genetic drift load (Willi et al., 2005; but see Betancourt, 2007), varied animal pollinator presence and foraging behaviour (Ågren, 1996; Aguilar et al., 2006), and mate limitation in self-incompatible species (DeMauro, 1993; Young et al., 2000). In addition, populations that become reliant on clonal reproduction as a means of reproductive assurance may accumulate genetic load that is unable to be shed (Müller's ratchet), leading to reduced fertility (Richards, 1986; Eckert, 2002). These factors combined with an increased exposure to demographic and environmental stochasticity are likely to increase the risk of extinction of small isolated populations (Frankham et al., 2002; Frankham, 2005).
The introduction of new material to genetically depauperate plant populations is an approach that may alleviate such problems. This management practice has the potential to increase offspring fitness and reproductive output in inbred populations by heterosis effects (e.g. Hufford and Mazer, 2003; Bossuyt, 2007; Willi et al., 2007), increase the number of compatible partners in populations of self-incompatible species through augmented S-allele diversity (DeMauro, 1993; Pickup and Young, 2008), increase the adaptive potential of the population to changing environmental pressures by increasing genetic variation (Reed and Frankham, 2003; Willi et al., 2006) and reduce the impact of stochastic factors by increasing the effective population size (Frankham et al., 2002).
While there are many potential benefits of this conservation management strategy, there is much contention about the inherent risks of introducing genetic material into populations (Fischer and Matthies, 1997; Edmands, 2007). Besides the possible negative consequences of unintentionally combining material from populations of differing chromosome number or ploidy (Murray and Young, 2001), hybridization between divergent populations of some taxa can lead to outbreeding depression expressed as reduced offspring fitness (e.g. Lynch, 1991; Hufford and Mazer, 2003). Predicting under which circumstances this response will occur can be complex (Dudash and Fenster, 2000; Willi and Van Buskirk, 2005; Edmands, 2007); however, it may be likely after crossing between populations adapted to differing local conditions (ecotypes) in species displaying limited dispersal capability (Frankham et al., 2002). Two of the possible, but non-exclusive, mechanisms for outbreeding depression are the dilution of locally adapted genotypes and the breakdown of co-adapted gene complexes (Dudash and Fenster, 2000; Frankham et al., 2002; Hufford and Mazer, 2003; Edmands, 2007). These mechanisms may reduce fitness of resultant hybrid offspring in the F1 or later generations (Frankham et al., 2002; Becker et al., 2006). Pre- or early post-zygotic genetic incompatibilities may also be involved in outbreeding depression; inhibited pollen tube growth or reduced seed set can occur after crosses among distant plants (Waser and Price, 1993; Montalvo and Ellstrand, 2001; Pélabon et al., 2005).
Whilst there have been numerous studies examining the effects of cross-pollination between plants in natural populations of varied geographic separation, relatively few have coupled pollination treatments with measures of genetic distance (see Trame et al., 1995; Montalvo and Ellstrand, 2001; Pélabon et al., 2005; Raabova et al., 2007; Willi et al., 2007). Geographic distance between populations may often be a surrogate for genetic divergence under isolation by distance, but there is not always a strong relationship (e.g. Montalvo and Ellstrand, 2001; Mable and Adam, 2007). The interpretation of results from cross-pollination between populations may therefore be clearer when accompanied by measures of genetic differentiation.
In addition, the genetic background and breeding system of the population in question are likely to influence the effects of outcrossing. Aside from the effects of mate limitation, populations of normally outbreeding plant species that are inbred are likely to display greater heterosis effects after outcrossing than similar sized populations with greater genetic diversity. However, small populations of species adapted to regular selfing are likely to be purged of deleterious recessive alleles or employ mechanisms to maintain heterozygosity, and therefore may not always display this response (Dudash and Fenster, 2000; Heiser and Shaw, 2006). To complicate matters, populations of normally obligate outbreeding (self-incompatible) species may experience strong selection pressures for self-compatibility where pollinator activity is scarce or there are few compatible mates (Busch, 2005). Thus the breeding system, and consequently the effect of outcrossing, may vary across the geographic range of a species.
The genus Grevillea (Proteaceae) includes numerous species restricted to small, isolated populations often as a result of limited historical distributions augmented by more recent human-related fragmentation (e.g. Rossetto et al., 1995; Hoebee and Young, 2001; England et al., 2002). While low natural seed set relative to flower production is often observed in Grevillea populations (Hermanutz et al., 1998), in many cases the biology of the species in question may provide a buffer against the effects of low reproductive output (e.g. long life spans, long flowering seasons and soil seed banks). However, in some populations, low seed production may be compounded by the negative effects of small population size and/or isolation, particularly in those that are self-incompatible. For example, using a model to assess population viability, Hoebee et al. (2008) predicted that persistence times for G. iaspicula are negatively affected by current low establishment rates, but that this effect is compounded when allelic diversity at the incompatibility locus is also restricted. In such cases, interpopulation cross-pollination and/or the introduction of plant material from other sites may provide a way of increasing long-term population viability, but this must be balanced against potential outbreeding depression.
The pollination biology of the rare shrub, Grevillea repens, was investigated in five populations with varied genetic diversity and size across its natural range (Fig. 1). At each site, reproductive responses (fruit initiation, fruit maturation and seed set) were also compared after performing crosses using pollen either sourced locally or from other populations of varying genetic distances. It was predicted that reproductive output in the smallest and/or inbred populations would increase with genetic distance, while in genetically diverse populations there would be negligible or negative reproductive effects. The following questions were specifically addressed. (a) Does the breeding system of G. repens vary across its range? (b) Is sexual reproductive output in small, isolated populations reduced by pollinator or mate limitation or by low pollen fertility when compared with larger populations? (c) Can reproductive output in the study populations be increased with interpopulation gene flow? (d) Is there a relationship between genetic distance of populations involved in a cross and reproductive output?
Grevillea repens (Proteaceae) is a prostrate, woody shrub to 3 m diameter or more found in montane dry eucalypt forests in the Central Highlands of Victoria, Australia (Makinson, 2000). The species is restricted to populations in three disjunct regions (Fig. 1) which have been historically isolated (Holmes et al., 2008). Plants in the western and western central regions appear to regenerate by seed or from buds in lignotubers after disturbance events such as fire. Limited clonal reproduction by root-suckering has also been confirmed in some eastern populations (Holmes et al., 2008). Voucher specimens of plants from these three regions are lodged at MELU.
The species has maroon to green hermaphrodite flowers that open sequentially in secund (‘toothbrush’) conflorescences up to 80 mm long. An average of 22·8 (s.e. ±0·6) flowers and buds per conflorescence were produced over five populations in the 2006 flowering season (G. D. Holmes, unpubl. data). As is typical for Grevillea flowers (Collins and Rebelo, 1987), four epitepalous anthers deposit pollen onto a specialized apical region of the style (the pollen presenter) before the style is fully released from the perianth at anthesis. This pollen presenter is also the site of the stigmatic region, and self-pollen must be dislodged before interfloral pollination can occur. Flowers are likely to be adapted for bird pollination; however, low visitation rates precluded detailed observation of foraging behaviour in the current study. Successful fertilization may result in the formation of follicular fruit containing up to two seeds that, when released, appear to be dispersed by gravity, ants or soil movement.
Pollen manipulation experiments were undertaken in natural populations to assess variation in its mating system and its capacity for fruit and seed production. Each treatment (geitonogamous and autogamous self-pollination, cross-pollination and open pollination as described below) was applied to single conflorescences on each of the replicate plants to allow paired comparisons. Plants treated were a minimum of 5 m apart, had at least five developing conflorescences and showed no signs of senescence. Population details and replicate numbers are given in Table 1. In season 1 (spring–summer 2005–2006), within-population treatments were applied at PG and FP, while in season two (2006–2007) these treatments were repeated and performed at an additional three populations (ST, AL and MF). Interpopulation crosses were also performed in season 2. Selected sites are relatively large but of varied size and are positioned away from potential human interference. In a parallel study these populations were found to have moderately high genetic divergence (FST = 0·272) and varied levels of diversity (Table 1). One small population (MF) was inbred and the largest (AL) had low genetic diversity (Holmes et al., 2008).
To assess self-compatibility, a developing conflorescence on each replicate plant at FP and PG was selected, and unsuitable flowers (those damaged, old or newly opened with disturbed pollen loads on presenters) were removed. Flowers about to open were ‘tripped’ by gently prying apart the tepals. Pollen was transferred among flowers of the same conflorescence (geitonogamy) using a fingertip or the bent tip of a dissecting needle. The conflorescence was then enclosed in a fine-mesh nylon bag to exclude external pollen sources. When necessary, extra suitably developed flowers in the conflorescence were treated after 1–3 d. An average of 8·0 flowers was treated per conflorescence with individual flowers pollinated only once. The styles of any untreated flowers were subsequently cut to prevent autogamous fertilization. For several plants in the FP population little pollen was seen on the pollen presenters at anthesis. In these cases, extra pollen was scraped from anthers and applied to the stigmatic region with a dissecting needle. Pollinating tools were washed in 70 % ethanol between all treatments. The capacity for fruit and seed production after within-flower self-pollination (autogamy) in the absence of floral visitation was assessed by bagging a developing conflorescence on each plant after removing unsuitable flowers. An average of 25·9 flowers was included per conflorescence.
To determine the cross-pollination response, a conflorescence on each plant was selected and unsuitable flowers removed. Extra flowers for treatment were ‘tripped’ and self-pollen removed from the pollen presenters with a cotton bud. Exposed stigmas were cross-pollinated using a pollen presenter sourced from a plant at least 20 m distant. One paternal plant was used for each replicate (overall, 14 and 11 plants were used as pollen sources at FP and PG, respectively). The conflorescence was then bagged. When necessary, the treatment was performed on extra flowers after 1–3 d using pollen from the same sources and the styles of untreated flowers subsequently cut. An average of 7·6 flowers was treated per conflorescence.
The effectiveness of floral visitors at transferring pollen between compatible flowers was determined with an open pollination treatment. A conflorescence was selected and left exposed to visitation after damaged or senescing flowers were removed. An average of 23·5 flowers was included per conflorescence. To retain developing fruits, each conflorescence was bagged when the tepals on the youngest flowers began to senesce. This treatment was positioned away from bagged pollination treatments to minimize potential disturbance of floral visitors. To test for disturbance, single conflorescences were selected on ten extra plants in both populations and monitored for fruit production. No significant differences in production were found (data not presented).
In the second season, the four treatments described above were applied to plants at five sites (FP, PG, MF, ST and AL). The numbers of plants flowering and conflorescences per plant were reduced in relation to the previous season at FP and PG. This may reflect reduced rainfall in the September–December period of season 2 relative to season 1 (25 % less in the eastern region and 75 % less in the western region; Bureau of Meteorology, 2008). As a result, fewer plants were treated (Table 1). Application of the pollination treatments commenced at the sites over a 4 week period in mid spring. The autogamy treatment was not applied at FP due to a shortage of conflorescences. An average of 11·4 flowers was treated per conflorescence for self- and cross-pollination treatments, and 21·9 and 20·7 flowers included for autogamous self- and open pollination, respectively. While the stigmas of flowers may not have been fully receptive when cross- and self-pollination treatments were performed at anthesis, the results from season 1 indicated that pollen applied at this stage was captured by the stigmas and maintained viability until full receptivity commenced. This approach was therefore repeated for treatments performed in season 2.
To identify if the reproductive output of G. repens populations was limited by mate availability or inbreeding depression, cross-pollinations were performed in season 2 using populations of increasing genetic divergence. Results from these crosses were compared with within-population cross-pollination treatments described above. At each of the five study sites, flowers on single conflorescences of ten plants were pollinated with pollen from plants from another population. Each of the ten replicate crosses involved one male and one female, and pollen source plants were only used in one cross. Pollen exclusion bags were then applied. This was repeated with pollen from another two populations including at least one inter-regional site (Table 2). Maternal plants with at least three conflorescences were used, and an average of 10·5 flowers treated per conflorescence. Only six paternal plants were used for the MF × FP and AL × MF treatments due to a lack of suitable flowers. These treatments were usually applied after intrapopulation treatments had been initiated.
In both seasons, treated plants were revisited after 4 weeks to determine the proportion of initiated fruit (flowers attached or recently abscissed with ovaries ≥1·5 mm wide) and again after 8 weeks for mature fruit. Sites were then visited weekly and any pollen exclusion bags containing open follicles collected. The number of filled seed was recorded, and seeds were weighed 3–4 d after collection. Unfilled seeds (usually ≤10 mg) were considered unviable. The proportion of filled seed relative to ovule number in treated flowers was determined. As the overall seed production was low, comparisons of seed weight are not presented and seedling fitness not assessed.
To estimate pollen viability, a conflorescence was collected from each of ten plants per population. For each of two pre-anthesis flowers per conflorescence, freshly dehisced pollen was brushed from the pollen presenter onto a glass slide and stained in a 0·5 % 2,3,5-triphenyl tetrazolium chloride (TTC) solution with 15 % sucrose before incubation for 36–48 h (based on Gross and Caddy 2006). For three of the ten plants from FP for which little pollen was presented, anthers were excised then macerated in TTC. The proportion of positively stained pollen grains was determined in each of five random fields of view (average: 44·2 pollen grains per view) at 100× magnification. Data from the two flowers examined per conflorescence were pooled, and the mean percentage viability determined for each plant and population. Pollen killed by heating flowers at 90 °C for 2 h was used as a control.
Pollen production per flower was estimated for each population using a method modified from Dafni (1992). For each of the ten conflorescence used for pollen viability tests, an undehisced anther was excised from a flower and macerated in a tube containing 150 µL of methylene blue stain. After vortexing, two 1 µL aliquots of solution were pipetted onto a microscope slide. This process was repeated five times. The number of pollen grains in each aliquot was counted, and grains per flower estimated by multiplying mean grains per aliquot by total stain volume and anther number. The amount of pollen on the pollen presenters of newly opened flowers from FP, MF and ST was also determined using single flowers from each of five conflorescences also used to test pollen viability. Pollen was washed from pollen presenters into methylene blue, and the number of grains per flower estimated as above.
Proportional fruit and seed set data could not be transformed to a normal distribution; therefore, non-parametric analyses were performed. To examine differences in reproductive output among treatments, Kruskal–Wallis tests were used for between-season and population comparisons, and Wilcoxon signed-rank tests for pairwise within-population treatments. Seasonal variation in the proportion of developing fruit aborted between initiation and maturation after cross-pollination was tested for the PG population using a Kruskal–Wallis test. Replicates where no fruit was initiated (three in both seasons) were excluded from the fruit abortion analysis.
The self-compatibility index (the ratio of proportional selfed to crossed fruit) was determined at the fruit initiation stage as per Hermanutz et al. (1998) for each population over both seasons. The ratio was determined for each plant, and the mean calculated for each population and the species. For this index 0 = fully self-incompatible and 1 = fully self-compatible. Replicates including zero data points for cross-pollination were excluded.
Ordered differences in proportional fruit and seed set after interpopulation crosses of increasing genetic distance were tested using non-parametric Jonckheere–Terpstra tests. When the order of increasing genetic and geographical distances between populations involved in crosses did not concur, Jonckheere–Terpstra tests were also performed on geographic distance.
Proportional pollen viability data could not be normalized with standard transformations; therefore, a Kruskal–Wallis test was undertaken to compare populations. Estimated counts of pollen production per flower among populations and the amount of pollen ‘presented’ at anthesis were investigated using analysis of variance (ANOVA). When significant, post hoc comparisons between population means were undertaken with Tukey's HSD tests. All statistical analyses were performed using the program SPSS® v15·0.
Grevillea repens produced very few fruit or seeds after self-pollination over all populations and both seasons. Of the 3683 flowers included in the unassisted (autogamy) and assisted self-pollination treatments over two seasons, 23 fruits were initiated but only two seeds formed (both at PG in season 1). The self-compatibility index for the species using pooled data from each plant over both seasons for all populations was 0·026 ± 0·011. In season 1, cross-pollination resulted in a higher proportion of fruit maturation relative to self-pollination in both populations (PG, Z = −4·109, P < 0·01; FP, Z = −2·388, P < 0·05; Fig. 2A). The two sites differed in the proportion of fruit set after cross-pollination (χ2 = 18·553, d.f. = 1, P < 0·01). In season 2, cross-pollination resulted in no mature fruit in the MF population and no seed in either eastern population. However, more mature fruit resulted after cross-pollination than self-pollination at AL (Z = −2·670, P < 0·01), PG (Z = −2·524, P < 0·05) and ST (Z = −2·027, P < 0·05; Fig. 2B).
At FP, mature fruit production resulting from cross-pollination in season 2 was 85·9 % lower than in season 1; however, this difference was not significant (χ2 = 1·533, d.f. = 1, P = 0·216) due to the overall low numbers of fruit produced. At the PG site there was a 69·3 % reduction in mature fruit production in the second season relative to the first (χ2 = 12·012, d.f. = 1, P < 0·01; Fig. 2A, B). Fruit often aborted between initiation and maturation for cross-pollination treatments at all sites (see Supplementary Information available online). In season 1, losses were 41·5 and 30·8 % for FP and PG, respectively, while in season 2 the corresponding figures were 75·2 and 61·9 %. The proportion of aborting fruit differed between seasons at PG (χ2 = 5·806, d.f. = 1, P < 0·05). Seasonal variation in fruit abortion at the FP site was not tested as very few fruits were initiated in both seasons.
In season 1, the proportion of fruits and seeds set for open pollination and autogamy was similar in both populations (Fig. 2A and Supplementary Information available online). Cross-pollination increased fruit set when compared with open pollination at both FP (Z = −2·395, P < 0·05) and PG (Z = −4·109, P < 0·01; Fig. 2A). In season 2, significantly more mature fruit were produced in open pollination treatments relative to autogamy for only the ST population (Z = −2·201, P < 0·05). Open pollination resulted in seed set in only the ST and PG populations, while cross-pollination resulted in significantly higher mature fruit production than open pollination in the AL (Z = −2·603, P < 0·01) and PG (Z = −2·547, P < 0·05) populations (Fig. 2B). Detailed results of fruit initiation, fruit maturation and seed production from all within-population treatments over both seasons are available online (see Supplementary Information).
In three of the five populations, cross-pollination and resultant fruit/seed set showed either a positive or negative association with genetic distance (Fig. 3). From the western region, the large ST population showed no significant trend for increasing or decreasing fruit or seed set with increasing distance of the cross (Fig. 3A). This contrasts with the small PG population for which JT tests indicated significant decreases in initiated fruit (P < 0·01), mature fruit (P < 0·05) and seed set (P < 0·05) with increasing distance (Fig. 3B). A similar pattern was observed for the large AL population for mature fruit (P < 0·05) and seed production (P < 0·05) (Fig. 3C). In the eastern region, the inbred MF population displayed significant increases in fruit initiation (P < 0·05), fruit maturation (P < 0·01) and seed set (P < 0·01) with increasing genetic distance of crosses (Fig. 3D). For this population, the proportion of ovules developing to seed was increased from 0 % after within-population crosses to 7·33 % after cross-pollination with the most genetically distant population. In the small FP population, there was higher (although not statistically significant) fruit initiation with increasing genetic distance (Fig. 3E). No seed was produced after any pollination treatment in this population during season 2.
In some cases, genetic and geographic distance among populations did not concur (PG, FP and MF crosses). When analyses of reproductive output compared with geographical distance of crosses were performed, the trend of decreasing mature fruit set at PG was not significant (P = 0·063). In total, six of the eight JT tests performed for reproductive output vs. geographic distance showed reduced significance compared with when genetic distances were employed (data not presented).
Pollen viability differed significantly among populations (χ2 = 27·121, d.f. = 4, P < 0·001; Fig. 4A). Mean viability over all populations was 71·1 % and was highest in the largest western population (ST; 90·8 %) and lowest in the small eastern population (FP; 30·1 %). At the latter site, viability was as low as 1·8 % for one plant and <8 % in four of the ten plants assessed. Mean percentage pollen viability for the three (non-clonal) western/western central populations was 85·5 % and in the two eastern (clonal/sexual) populations was 49·6 %.
The number of pollen grains produced per flower differed significantly among populations (F1,45 = 9·153, P < 0·001; Fig. 4B). The mean pollen–ovule ratio for the five populations was 2852:1 (s.e. ± 126) and ranged between 2061:1 for PG and 3639:1 for ST. There was no significant difference among populations in the amount of pollen presented at anthesis per flower (FP, 1982 ± 743; MF, 3414 ± 209; ST, 4292 ± 568; F2,12 = 2·221, P = 0·151); however, two of the five flowers tested from the FP population were estimated to have ≤10 pollen grains on their pollen presenters. This is compared with >7000 presented grains for one plant from the ST population.
The pollination study suggested that G. repens is an obligate-outcrosser in the five populations examined and was consistently so over two seasons at the FP and PG populations. The finding is supported by its high pollen–ovule ratio which falls within the range generally found for xenogamous (outbreeding) hermaphrodite species adapted to animal pollination (Cruden, 2000). A system of gametophytic self-incompatibility has been suggested to operate in both G. robusta and G. iaspicula (Kalinganire et al., 2000; Hoebee and Young, 2001); however, the relative contributions of an incompatibility system and early-acting inbreeding depression to low fruit initiation after self-pollination in G. repens is unclear. An assessment of self pollen-tube growth within the styles of the study species similar to that of Kalinganire et al. (2000) may help clarify this uncertainty. Breeding systems for populations of Grevillea have been reported to range from fully self-compatible (Ayre et al., 1994; Hermanutz et al., 1998; Smith and Gross, 2002) to self-incompatible (Hermanutz et al., 1998; Kalinganire et al., 2000; Hoebee and Young, 2001), with many displaying preferential outcrossing but the ability to self (Hermanutz et al., 1998; Collins et al., 2008).
Small or geographically peripheral populations of self-incompatible plants may experience strong selective pressures for self-compatibility as a result of unreliable pollinator services and/or mate limitation (Busch, 2005). Significant spatial variation in pollen compatibility relationships was not observed among populations of G. repens. Self-pollination did however result in a very low number of initiated fruits in some populations and two seeds in the PG population in season 1. While this seed set occurred only after artificial self-pollination, autogamy also resulted in at least some mature fruit production in the first season, indicating that interflower pollen movement is not necessary for self-fertilization in G. repens. Nevertheless, the possibility of contamination with compatible pollen, while unlikely, cannot be completely discounted in these instances without genetic confirmation through paternity analyses.
While increased autonomous selfing may confer reproductive assurance in some species when natural pollinator activity is infrequent (Kalisz et al., 2004), F1 offspring resulting from self-fertilization in normally outcrossing species may display low fitness due to inbreeding depression and be strongly selected against (e.g. Heliyanto et al., 2005). Selection against such low fitness offspring may limit the rate of spread of any self-compatible mutants in natural populations of G. repens. Reproductive assurance through the ability to regenerate vegetatively may further reduce selection for self-compatibility in eastern populations in G. repens; in the genus Solanum, a positive correlation has been found between the ability of a species to reproduce clonally and the occurrence of self-incompatibility (Vallejo-Marín and O'Brien, 2007).
While the development of clonality may provide reproductive assurance in the absence of pollinators in small isolated populations of G. repens, it also has the potential to limit sexual reproduction and adaptive potential in this self-incompatible species. In the mixed clonal/sexual eastern populations, the number of compatible mates is likely to be lower than in comparably sized sexual populations. In addition to potential problems with mate limitation for self-incompatible species (Wilcock and Jennings, 1999) and an increased level of selfing in self-compatible species, populations of plants with the ability for clonal reproduction are more likely to accumulate mutations normally purged from those reproducing sexually, which may lead to reduced fecundity (Richards, 1986; Eckert, 2002). This may be the case in the clonal/sexual FP population where plants displaying low pollen viability and presentation (possibly due to triploidy; Holmes et al., 2008) can persist by means of clonality. Low fecundity or pollen presentation has also been identified in several clonally reproducing populations of Grevillea species (e.g. G. infecunda, G. renwickiana and G. rhizomatosa) (Makinson, 2000; Kimpton et al., 2002; Gross and Caddy, 2006).
Within-population cross-pollination generally resulted in higher mature fruit and seed set than for the open pollination treatments. Additionally, mature fruit production in conflorescences exposed to pollen vectors was significantly greater than autogamy treatments only in the ST population, suggesting low effective pollinator activity. Similarly low natural levels of fruit production relative to flower number were observed in open-pollinated flowers in two populations of the self-incompatible insect-pollinated G. sphacelata (0·8–4·1 %; Richardson et al., 2000) and several other Grevillea species (e.g. Hermanutz et al., 1998). While crossing between G. repens plants at least 20 m apart may have caused increased cross-compatibility and/or heterosis, the increase in reproduction suggests that pollinator limitation may have contributed to low natural seed set at some sites.
During the course of the study, visitation of flowers by red wattlebirds (Anthochaera carnunculata) and eastern spinebills (Acanthorhynchus tenuirostris) of the family Meliphagidae was only rarely seen, and only observed at the large ST and AL populations. Several invertebrate species (including the introduced honeybee, Apis mellifera) were seen feeding on nectar in all populations, but these species rarely contacted the pollen presenters and were less likely to effect cross-pollination. Visitation of flowers by honeybees may have a negative impact on fruit and seed production or outcrossing rates due to competition for nectar resources in populations of some bird-adapted Grevillea species (e.g. Vaughton, 1996); however, the effect on reproduction in G. repens is unknown.
Interpopulation cross-pollination increased seed set relative to within-population treatments at MF. This small population is inbred (Holmes et al., 2008), and the increased reproductive output may have been the result of heterosis effects or a release from mate limitation. This contrasts with all other populations for which there was no significant inbreeding. Tallmon et al. (2004) suggested that inbred populations are more likely to display greater genetic rescue benefits; this may apply to G. repens. While S-allele-related mate limitation may have an effect on reproductive output at MF, it did not necessarily limit fruit and seed production in the other populations. At the equally small PG population, 22 of the 25 plants crossed using within-population pollen in season 1 formed at least one mature fruit, and reproductive output was not increased by using interpopulation pollen. These results suggest that mate limitation may only apply to populations that are considerably smaller and more isolated or that are more inbred than those at PG.
However, as seed set did not approach 100 % of flowers hand cross-pollinated in any population, other factors besides mate limitation (and low pollinator activity) also affected reproductive output in G. repens. While it is possible that low seed set after interpopulation crosses may have been partly related to shared S-allele profiles, the varied reproductive responses observed among populations and the significant genetic divergence among populations (Holmes et al., 2008) reduce the likelihood of this explanation. Resource limitation has been proposed as a major cause of constrained fruit and seed set for several Grevillea species (Hermanutz et al., 1998; Collins et al., 2008). Seasonal variation in water availability is likely to influence reproductive output strongly in G. repens; natural seed set may be greater and the response after interpopulation crosses varied in years of higher rainfall.
Interpopulation pollen failed to increase seed production above zero at the small FP population. The presence of near sterile (possibly triploid) plants amongst fertile plants within the population compounded by resource limitation may have limited the rescue of reproductive output at this site. This population may experience mate limitation related to differing levels of ploidy and/or fertility. Gross and Caddy (2006) also found that seed set could not be recovered after the introduction of pollen to a population of clonally reproducing G. rhizomatosa, and hypothesized that this was the result of the accumulation of genetic load and associated sexual sterility.
Trends of either decreasing or increasing fruit or seed production with increasing genetic distance of a cross were observed at three of the five G. repens populations studied. In the inbred MF population, there was a pattern of gradual increase in sexual reproductive output with increasing genetic distance of the paternal plants. However, this trend may have been influenced by the inadvertent use of low fertility pollen in some crosses with plants from the FP population. The increase in reproductive output with genetic distance at MF contrasts with the opposite pattern for the AL and PG populations and no pattern at ST and FP. In MF, the benefits of increased heterozygosity in progeny after crossing between populations and/or the release from mate limitation may outweigh negative effects of genetic dissimilarity.
Two commonly cited causes of reduced offspring fitness after distant crosses are the dilution of locally adapted genotypes and the breakdown of co-adapted gene complexes (Dudash and Fenster, 2000; Hufford and Mazer, 2003; Edmands, 2007). While these mechanisms may lead to the expression of outbreeding depression in the F1 generation (Frankham et al., 2002), negative fitness effects are predominantly experienced from the F2 stage (Dudash and Fenster, 2000). While fitness traits of offspring were not tested in the current study due to low seed set, the observed reproductive response for the AL and PG populations suggests pre- or early post-zygotic genetic incongruence of increasing magnitude between increasingly divergent populations. Montalvo and Ellstrand (2001) also found a significant negative relationship between seed set and genetic distance in Lotus scoparius. These authors speculated that reduced crossing ability may have been caused by chromosome structural differences, or intragene interactions such as break up of favourable additive × additive epistasis or underdominance. Similar genetic processes may be affecting G. repens interpopulation crosses. Pélabon et al. (2005) also found reduced seed set and mass with increasing genetic distance in Dalechampia scandens. Differences in the expression of outbreeding depression on seed set in this species were affected by the direction of a cross, suggesting negative cytoplasmic × nuclear gene interactions or interaction between the maternal and progeny genotype. To explore interactions such as these in G. repens it would be interesting to apply pollen from single male plants to multiple maternal plants (or vice versa) and/or to perform full reciprocal crosses. Pre-fertilization barriers may also reduce the success of distant crosses. For example, Waser and Price (1993) found that pollen sourced from 1 and 100 m distance of maternal plants performed worse in the stylar tissue of Delphinium nelsonii than pollen from intermediate distances.
In summary, it is proposed that low effective pollinator activity and resource limitation decreased reproductive output in many populations of G. repens during the study period and that mate limitation related to low S-allele diversity may have influenced seed set in one small, inbred population. In some populations, low sexual reproductive output is likely to be compounded by varied ploidy and/or pollen fertility among plants maintained by clonal reproduction. The introduction of pollen from other populations successfully rescued seed set in an inbred population and in such cases may be an effective conservation management strategy. As suggested by Hoebee et al. (2008) for G. iaspicula, supplementing the gene pool of G. repens by the introduction of seed and/or propagated plant material whilst increasing establishment rates may achieve this over the long term. However, this approach may only be suitable in inbred or mate-limited populations at risk of extinction in the short term; reduced reproductive output was observed in other populations and the effect of possible outbreeding depression in subsequent generations was not assessed. Additionally, in G. repens populations that are not inbred, natural seed set, while low, may be adequate to maintain population viability for this perennial shrub if periodic opportunities for seedling recruitment occur. These results highlight that genetic rescue management strategies for threatened plants cannot be generalized across populations, and may need to be coupled with an understanding of genetic and ecological parameters.
Supplementary information is available online at www.aob.oxfordjournals.org and contains fruit and seed production data resulting from within-population pollination treatments over two seasons, the number of replicate plants and the number of flowers treated.
We thank the staff from Parks Victoria and DSE for allowing access to sites, Caroline Gross for advice on pollen viability testing, Adrian and Penelope Holmes for fieldwork assistance, and Susan Hoebee and two anonymous reviewers for providing constructive comments on the manuscript. This work was supported by the Cybec Foundation; Royal Botanic Gardens Melbourne; and the Centre for Environmental Stress and Adaptation Research. A.A.H. was supported by a Federation Fellowship from the Australian Research Council. Research was undertaken under Department of Sustainability and Environment permits 10003055 and 05/1/09/11/06.