Near-neighbour optimal outcrossing in the bird-pollinated Anigozanthos manglesii

Abstract

Background and Aims

In plants, the spatial and genetic distance between mates can influence reproductive success and offspring fitness. Negative fitness consequences associated with the extremes of inbreeding and outbreeding suggest that there will be an intermediate optimal outcrossing distance (OOD), the scale and drivers of which remain poorly understood. In the bird-pollinated Anigozanthos manglesii (Haemodoraceae) we tested (1) for the presence of within-population OOD, (2) over what scale it occurs, and (3) for OOD under biologically realistic scenarios of multi-donor deposition associated with pollination by nectar-feeding birds.

Methods

We measured the impact of mate distance (spatial and genetic) on seed set, fruit size, seed mass, seed viability and germination success following hand pollination from (1) single donors across 0 m (self), <1 m, 1–3 m, 7–15 m and 50 m, and (2) a mix of eight donors. Microsatellite loci were used to quantify spatial genetic structure and test for the presence of an OOD by paternity assignment after multi-donor deposition.

Key results

Inter-mate distance had a significant impact on single-donor reproductive success, with selfed and nearest-neighbour (<1 m) pollination resulting in only ~50 seeds per fruit, lower overall germination success and slower germination. Seed set was greatest for inter-mate distance of 1–3 m (148 seeds per fruit), thereafter plateauing at ~100 seeds per fruit. Lower seed set following nearest-neighbour mating was associated with significant spatial genetic autocorrelation at this scale. Paternal success following pollination with multiple sires showed a significantly negative association with increasing distance between mates.

Conclusions

Collectively, single- and multi-donor pollinations indicated evidence for a near-neighbour OOD within A. manglesii. A survey of the literature suggests that within-population OOD may be more characteristic of plants pollinated by birds than those pollinated by insects.

INTRODUCTION

The sessile nature of plants and the dispersal kernels of their seed and pollen means that neighbours are typically more closely related than more distant plants (Heywood, 1991; Vekemans and Hardy, 2004; Ellstrand, 2014). This spatial genetic structure means pollination, even across moderate distances, can encompass a large range in genetic distances between mates (Waser and Price, 1989). At one extreme, inbreeding between neighbouring plants can result in inbreeding depression, with a reduction in components of fitness such as seed set, germination, seedling growth and survival (Charlesworth and Charlesworth, 1987; Waser, 1993; Dudash and Fenster, 2001). At the other extreme, mating between genetically divergent plants can result in outbreeding depression, potentially arising from a breakdown of coadapted gene complexes due to maladaptation or genetic incompatibilities, or the disruption of environment-specific adaptations (Price and Waser, 1979; Bateson, 1980; Waser, 1993; Schierup and Christiansen, 1996; Dudash and Fenster, 2000; Edmands, 2007). A logical consequence of these extremes in mating is that an optimal outcrossing distance (OOD) exists somewhere in between. At the OOD, reproductive success is maximized through a reduction in the detrimental effects of inbreeding and outbreeding depression, termed ‘Fritz Müller’s law’ (Müller, 1883; Bateson, 1978; Price and Waser, 1979).

While outbreeding depression is often seen after mating between species, data in support of this phenomenon within or between populations of a taxon are less common (Frankham et al., 2011). Evidence for OOD has been found after single-donor hand-pollination experiments between plants at large spatial scales (e.g. Montalvo and Ellstrand, 2001; Galloway and Etterson, 2005; Pélabon et al., 2005), as well as at small spatial scales of only a few metres within populations (e.g. Price and Waser, 1979; Waser and Price, 1989, 1994; Quilichini et al., 2001; Grindeland, 2008). An OOD of short to intermediate geographical distances is predicted to occur primarily in plant species with highly restricted pollen and/or seed dispersal (Price and Waser, 1979). Ultimately, the spatial scale across which this phenomenon might occur remains poorly understood, as does the role of pollination vector, seed dispersal strategy and general life-history parameters.

For an estimated 90 % of plant species (Ollerton et al., 2011), pollen dispersal patterns are predominantly determined by the behaviour of animal pollinators. Typically, an animal will display an optimal foraging strategy that maximizes energy gain relative to energy expended (Charnov, 1976). For pollinators, this is usually achieved by foraging between nearby flowers (Waser, 1982). For the plants they pollinate, such behaviour often results in self- and/or near-neighbour pollination. Therefore, pollen is typically dispersed in a leptokurtic pattern, with most pollen grains dispersed close to the pollen source and fewer grains dispersed far away from the source (Levin and Kerster, 1974; Proctor et al., 1996). While both invertebrate and vertebrate pollinators tend to follow this near-neighbour foraging pattern, realized pollen dispersal may be influenced by other factors. Pollinator mobility, as well as inter- and intra-species interactions may disrupt foraging patterns (Ford et al., 1979; Orians and Milewski, 2007; Hasegawa et al., 2015; Krauss et al., 2017). Additionally, pollinator grooming can influence the extent of pollen carryover, altering pollen dispersal patterns (Holmquist et al., 2012). Due to pronounced variation in animal behaviour and morphology, different groups of pollinators are predicted to display a range of pollen dispersal patterns (Krauss et al., 2017).

Vertebrate pollination, especially by birds, is common in many parts of the Southern Hemisphere, often in old climatically buffered infertile landscapes (OCBILs) (Hopper, 2009; Hopper et al., 2016; Krauss et al., 2017). In the temperate OCBILs of the Southern Hemisphere, bird pollination is dominated by honeyeaters (Meliphagidae), sunbirds (Nectariniidae) and lorikeets (Psittaculidae, subfamily Loriinae). Honeyeaters and sunbirds tend to display aggressive territorial behaviours that interrupt foraging bouts (Wolf, 1978; Ford, 1979; Ford and Paton, 1982; Ford, 2001; Padyšáková et al., 2017), often resulting in departure from near-neighbour mating (Krauss et al., 2009; Llorens et al., 2012; Frick et al., 2014). Further, pollen transport by birds can result in higher levels of pollen carryover than pollen transport by bees (Castellanos et al., 2003). Birds lack the grooming behaviour of bees, which is thought to reduce the magnitude of pollen carryover (Robertson, 1992; Castellanos et al., 2003; Holmquist et al., 2012; Krauss et al., 2017). In bird-pollinated plants, both departure from predominantly near-neighbour pollination and higher levels of pollen carryover are likely to increase genetic diversity and conserve heterozygosity. This may be one reason bird pollination is prevalent in older landscapes, where plant populations are often more naturally fragmented [the James effect (Hopper, 2009)].

Multiple paternity within fruits usually results in high levels of genetic variation among offspring, (Karron et al., 2012; Rhodes et al., 2017), increasing the likelihood that some offspring will be adapted to a variable environment (Karron and Marshall, 1990, 1993). High levels of pollen carryover can also increase competition amongst pollen grains (Winsor et al., 2000; Skogsmyr and Lankinen, 2002), or provide greater opportunities for direct female choice (Skogsmyr and Lankinen, 2000; Pannell and Labouche, 2013). Conversely, pollen carryover may also increase the likelihood of pollen being provided from individuals that are extremely closely related or extremely distantly related (Price and Waser, 1979; Waser and Price, 1989, 1994; Waser et al., 2000), potentially realizing the negative effects associated with inbreeding and outbreeding depression.

Evidence for multiple paternity has been found in many plant species, and is thought to be widespread in outcrossing species (Campbell, 1998; Burkhardt et al., 2009; Pannell and Labouche, 2013). Despite this, to our knowledge, experimentation has not yet been used to investigate optimal outcrossing under more biologically realistic multi-donor pollination. Here, for the first time, we aimed to address this gap in knowledge by performing both single- and multi-donor hand pollination. Our experiments were undertaken with Anigozanthos manglesii (Haemodoraceae), a preferentially outcrossing bird-pollinated wildflower from the Southwest Australian Floristic Region (SWAFR). Plants of A. manglesii produce up to 20 flowers, with large floral features that can be easily manipulated for hand pollination. Due to the gravity-dependent seed dispersal of A. manglesii (Hopper, 1993), we hypothesize that (1) there will be strong fine-scale spatial genetic structure, and (2) inter-mate fitness will be associated with this spatial genetic structure, due to the effects of inbreeding depression after nearest-neighbour mating and outbreeding depression after mating between genetically distant mates. If hypotheses 1 and 2 are supported, we hypothesize (3) the existence of a within-population OOD after both single- and multi-donor hand pollination.

MATERIALS AND METHODS

Study species and site

Anigozanthos manglesii, the red and green kangaroo paw (Fig. 1), is endemic to the SWAFR and characterized by large showy, red and green inflorescences, on stems between 30 and 80 cm tall (Hopper, 1993). There are two subspecies of A. manglesii, the southern A. manglesii subsp. manglesii and the northern A. manglesii subsp. quadrans (Hopper, 1987, 1993). Both subspecies produce copious amounts of sugary nectar [up to 250 µL per flower per day (Hopper, 1993)]. Flowers are visited primarily by nectar-feeding birds (Hopper and Burbidge, 1978), as well as by the introduced European honeybee (Apis mellifera) and potentially scansorial mammals. Up to 20 flowers open progressively over several days, beginning from the bottom of the inflorescence. Seed set and germination success tend to be higher in basal fruits (Tieu et al., 2001). Dried fruits dehisce during early summer and the small, hard seeds drop through wind-shake or other agency directly below the maternal plant (Hopper, 1993). Populations separated by as little as 15 km have been shown to be genetically differentiated (Bussell et al., 2006).

Fig. 1.

Anigozanthos manglesii flowering head with a New Holland Honeyeater. Photograph by S. D. Hopper.

Experiments were conducted in a population of A. manglesii subsp. manglesii in the bushland area of Kings Park, an urban remnant of ~300 ha within the Perth metropolitan area (31°57′20.5″ S, 115°50′27.9″ E). Although found naturally throughout bushland areas of the reserve, additional A. manglesii of unknown provenance were introduced into parts of the park through broadcast seeding, though this practice ended in the early 1990s (Kings Park and Botanic Garden, 1995, R. Glowacki, Botanic Gardens and Parks Authority, Perth, Western Australia pers. comm.).

Spatial genetic structure

To determine the spatial genetic structure of the population, DNA was extracted from leaf samples taken from all maternal plants and their donors using a modified Carlson extraction method. Amplification of ten microsatellite markers (Supplementary Data Table S1) and genotyping on an ABI 3500 Genetic Analyzer (Life Technologies) proceeded as per Ayre et al. (2017). Scoring of microsatellite alleles was undertaken in Geneious version 7.1 (Biomatters). Spatial genetic structure was assessed by a Mantel test of correlation between genetic and spatial distances, as well as spatial autocorrelation analysis (SAA) in Genalex 6.5 (Peakall and Smouse, 2006, 2012). Spatial autocorrelation analysis was performed using the Single Pop function. Distance classes were chosen to correlate with the hand-pollination treatments described below, and repeated at 1-, 2- and 5-m intervals. We used 9999 permutations with a bootstrap re-sampling procedure run 10 000 times to estimate the 95 % confidence interval.

Single-donor hand pollination

To test for the reproductive consequences of intra-population mate distance, we enclosed 72 inflorescences in a fine mesh and wire cage that prevented access by all potential pollinators. In spring 2015 (23 September to 20 October) and 2016 (7 September to 4 October) all open flowers across 36 experimental plants (nine per treatment) were emasculated and donor pollen was applied by hand. In 2015, single-donor pollen was applied from either the same plant (self-pollination), a donor plant from 1–3 m (neighbour) or from a donor plant between 20 and 55 m (distant). After preliminary analysis, distance classes were refined to include three extra pollen treatments in 2016: no pollen (flowers emasculated), <1 m (nearest neighbour) and 7–15 m (intermediate). The 20–55 m (distant) treatment was repeated to control for year-to-year variation. In all cases, stigmas were coated with pollen grains harvested from respective donor plants in the previous 30 min. Anthers were removed with tweezers and stored in 1.5-mL Eppendorf tubes between collection and pollination, and pollen was provided to stigmas directly off the collected anther. An average of 6.7 flowers per experimental plant were pollinated: <1 m, 60 flowers across nine plants; 1–3 m, 52 flowers, nine plants; 7–15 m, 70 flowers, nine plants; 20–55 m, 36 flowers, seven plants (two lost via predation) and 62 flowers, nine plants in 2015 and 2016 respectively. A total of 105 flowers across nine plants were self-pollinated.

Multiple-donor hand pollination

To assess the influence of mate distance on reproductive success under the more realistic scenario of multiple pollen donor deposition, pollen mixes were provided to the stigmas of 100 flowers across nine maternal plants. In 2015, eight donor plants located between 0.6 and 60 m from experimental maternal plants were chosen from within the study population (Supplementary Data Fig. 1). While all the distance classes used in the single-donor experiments are represented, the majority (63 %) of pollen donors in the mixed pollen treatment occurred between 20 and 50 m away from maternal plants.

Pollen mixes were created by combining one anther from each donor plant in an Eppendorf tube. A toothpick was used to dislodge pollen from anthers. Maternal flowers were emasculated and donor pollen was applied until the stigmas were visibly covered in pollen. Each flower received a single application of pollen from all eight donors. A total of 100 flowers were pollinated across the nine maternal plants.

Data collection and analysis

The location of all experimental plants was mapped each year using a differential global positioning system (dGPS), and a spatial distance matrix for pollen donors and recipients was created. Fruits were harvested between late December and early January each year. Fruit size was measured (length and width), and the number of seeds per fruit was counted. Due to the high number of seeds produced (total of 35 441 seeds across the entire experiment), all seeds in the self-pollination treatment and 10 % of seeds from all other experimental plants were X-rayed (MX-20 X-ray cabinet; Faxitron, Tucson, AZ, USA) to assess viability. Seeds with a visible endosperm were considered viable and those without a visible endosperm as inviable (e.g. Kamra, 1964; Gagliardi and Marcos-Filho, 2011). Viable seeds were grouped by fruit and weighed, and the average individual seed mass was calculated for each fruit. Fifteen viable seeds, from three fruits per plant, were heat-shocked at 100 °C for 3 h in an oven to break dormancy (Tieu et al., 2001). To induce germination, seeds were plated on 5 % agar and placed in a 15 °C incubator with 12 h of light and 12 h of dark. Germination success and vigour (overall percentage germinated and time to 50 % germination, respectively) were scored over 4 weeks. Seeds were considered to have germinated when the radicle was observed to be a third of the length of the seed itself.

The fitness consequences of mate distance were assessed using fruit volume [$\mathrm{f}\mathrm{r}\mathrm{u}\mathrm{i}\mathrm{t}\mathrm{l}\mathrm{e}\mathrm{n}\mathrm{g}\mathrm{t}\mathrm{h}\times \mathrm{\pi }{(\frac{\mathrm{w}\mathrm{i}\mathrm{d}\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{w}\mathrm{i}\mathrm{d}\mathrm{t}\mathrm{h}}{2})}^2$], the number of seeds per fruit, seed mass, seed viability and germination success. As all data sets violated the assumption of normality and could not be transformed to fit a normal distribution, the impacts of treatments were determined using the non-parametric Kruskal–Wallis rank sum test, followed by a post hoc Dunn test (R Development Core Team, 2016; Dinno, 2017).

Seedling paternity analysis

Seedlings were genotyped with the previously described ten microsatellite markers to (1) confirm the paternity of seeds following self-pollination, and (2) determine the individual siring success of donor plants after mixed pollination. Paternity was assessed for 45 putatively selfed and 270 multi-donor seedlings (ten per fruit, three fruits per plant) using CERVUS (Kalinowski et al., 2007). The simulation parameters for CERVUS to assign paternity to the most likely sire with a known level of statistical confidence were: 10 000 cycles of simulation, eight candidate fathers, 1.0 as the proportion of the population sampled and 0.01 as the proportion of mistyped loci. Confidence levels calculated using the LOD score (the natural log of the overall likelihood ratio) were 80 % for relaxed confidence and 95 % for strict confidence of assignment of paternity to offspring. The percentage of seeds fathered by each paternal donor per maternal plant was calculated. A pairwise test of proportions was used to test whether there was variation in the proportion of seeds fathered by different pollen donors using RVAideMemorie (Hervé, 2018).

RESULTS

Spatial genetic structure

The Mantel test showed no overall significant relationship between genetic and spatial distance within the study population (P = 0.260). However, spatial autocorrelation analysis detected significant spatial genetic structure in the smallest distance class of 0–1 m (r = 0.189, P = 0.006), with a stabilizing profile for distances up to ~2.5 m (Fig. 2). At all distances exceeding 2.5 m, there was no significant autocorrelation between pairwise multilocus genotypes

Fig. 2.

Correlogram showing the genetic correlation coefficient r as a function of distance for A. manglesii distance class sizes of 1, 5, 15 and 50 m. Grey lines show the 95 % confidence interval around the null hypothesis of a random distribution of genotypes. Error bars indicate the 95 % confidence of r as determined by bootstrapping.

Effect of inter-mate distance: single donors

Year of pollination had no effect on the number of seeds per fruit, seed viability, overall germination success and time to 50 % germination, as assessed for the 20–55 m treatment (Fig. 3). However, fruit volume (χ² = 71.78, P = 0.007) and seed mass (χ² = 46.59, P = 0.005) were both significantly lower in 2015 than 2016.

Fig. 3.

Tukey’s boxplots showing the effect of distance between mates after single and multi-donor hand pollination on A. manglesii. (A) Fruit size. (B) Number of viable seeds per fruit. (C) Mass of viable seeds. (D) Percentage viability. (E) Total germination success. (F) Days to 50 % germination. Letters indicate Dunn test significance groupings (P < 0.05). The box represents the upper, median and lower quartiles and the whiskers indicate the 1.5× interquartile range; dots represent outliers.

Only 26 % of self-pollinated flowers produced fruit containing seeds. Fruits originating from self-pollination were the smallest and contained the lowest number of seeds (Fig. 3). Paternity assignment with CERVUS confirmed that all seeds collected from selfed treatments were the product of self-pollination, with no non-maternal alleles expressed in any of the progeny.

Although there was no overall significant correlation between the average number of seeds produced and the spatial distance between mates (r = 0.05, P = 0.78), there was a significant difference between distance classes (P < 2.2e−16). Nearest-neighbour pollination produced an average of 50 seeds per fruit, followed by maximum seed production (148 seeds per fruit) at 1–3 m, and all further distances plateauing at ~100 seeds per fruit (Fig. 3B). All outcrossed flowers produced fruit with seeds, and >97 % of seeds were viable for all treatments (Fig. 3D). Inter-mate distance had an overall negative effect on seed mass (Fig. 3C) (P = 8.77e−11, r = −0.416). The number of seeds produced per fruit and fruit size were positively correlated (P < 2.2e−16, r = 0.685).

The seeds of nearest-neighbour mating (<1 m) had significantly lower germination success than all other non-selfed distances (Fig. 3E). Nearest-neighbour and selfed crosses also had a longer time to 50 % germination than all other crosses (Fig. 3F), with seeds from self-pollinated flowers taking on average two extra days and those of nearest neighbour crosses an extra day.

By combining the fitness metrics described above, an estimate of the number of germinating seeds per fruit can be produced. Comparing this new metric with the best-performing mate distance (1–3 m) shows a reduction in germinating seeds per fruit. Flowers produced 3 % of the optimum seed per fruit after self-pollination, 29 % of the optimum in the nearest-neighbour (<1 m) treatment and 65–69 % of the optimum in the intermediate (7–15 m), distant (20–55 m) and mixed treatments.

Effect of inter-mate distance: multiple donors

Fruit, seed set and germination success.

After mixed-donor crosses, flowers produced an average of 98 seeds per fruit, significantly less than the optimal 148 seeds per fruit (P = 3.492e−05) for the nearest-neighbour (1–3 m) treatment, and similar to the ~100 seeds per fruit of intermediate and distant crosses (Fig. 3).

Paternal success.

Out of 270 progeny, paternity was successfully assigned to all but 12 seeds. Within each fruit, between six and eight sires were identified, with individual paternal donors siring 6–20 % of all offspring within a fruit (Fig. 4). In general, donor success was strongly correlated to the average pairwise genetic distance between parents (Fig. 4), excluding donor number 6 (P = 0.0032, r² = 0.85). If we include donor number 6, there was a non-significant trend (P = 0.065, r² = 0.46). There was significant variation between donor plant siring success (χ² = 18.6443, d.f. = 7, P = 1.58e−0.3). Donors 6, 1 and 2 had the highest success (siring 19, 20 and 17 % of seedlings, respectively), while donors 7 and 8 had the lowest success (6 and 7 %, respectively).

Fig. 4.

Percentage of A. manglesii offspring fathered by each paternal donor after pollen from eight donors (D1–D8) was provided simultaneously. Paternal donors are ordered based on the average genetic distance between the donor and maternal plants. Error bars indicate standard error and letters show the grouping by Dunn test significance (P < 0.05). Individual pairwise genetic distances were calculated in Genalex 6.5 (Peakall and Smouse, 2006, 2012).

Paternal success of individual donors decreased with increasing spatial distance between mates, with a significant weak negative correlation between spatial distance and paternity success (r² = −0.28, P = 0.026) (Fig. 5A), and genetic distance and paternity success (r² = −0.303, P = 0.023) (Fig. 5B). However, the majority of pollen used in the mixed treatment (63.3 %) was transferred 20–50 m between the donor and maternal plants. When comparing pollen donor success with the distance categories used in single-donor pollination, there was no difference between siring success and the number of crosses at that distance. Two percent of pollen was transferred <1 m, and 2.1 % of all seed were sired by plants within this distance class. Likewise, 5 % of pollen was moved <3 m (5.8 % of sires, P = 0.7), 23 % <15 m (28.7 % of sires, P = 0.075) and 64 % <50 m (63.3 % of sires, P = 0.64).

Fig. 5.

Effect of (A) spatial distance (r = 0.28, P = 0.026) and (B) pairwise genetic distance (r = −0.30, P = 0.023) between parents on the number of A. manglesii seeds fathered by each pollen donor after pollen from eight donors was provided simultaneously. Individual pairwise genetic distances were calculated in Genalex 6.5 (Peakall and Smouse, 2006, 2012).

DISCUSSION

Inter-mate distance had a significant effect on reproductive fitness in an A. manglesii population, with an optimum detected at 1–3 m following controlled hand pollinations. Reduced fitness for mates within 1 m reflects inbreeding depression associated with significant spatial genetic autocorrelation at this scale. The strength of this inbreeding depression was substantial, with seed set following mating between nearest neighbours 30 % that of the OOD. Beyond an inter-mate distance of 3 m, seed set following single-donor pollination was ~65 % of the optimum. However, single-donor pollination is probably unrepresentative of pollination in wild populations. Under such conditions, there is likely to be extensive pollen carryover due to the behaviour of the highly mobile nectar-feeding birds that pollinate A. manglesii, leading to multi-donor pollen deposition onto stigmas. Under the more realistic scenario of hand pollination with multiple donors, siring success declined with increasing inter-mate distance (both geographical and genetic). Consequently, multi-donor pollinations support the conclusion from single-donor pollinations that an OOD exists for near neighbours. It also highlights the role of competitive interactions within the pistil that appear to drive the negative association between reproductive success and inter-mate distance at a surprisingly fine spatial scale.

Within-population optimal outcrossing after multi-donor pollination

To our knowledge, very few studies have tested for an association between inter-mate genetic or spatial distance and reproductive success following multi-donor pollination. In Persoonia mollis (Proteaceae), applying a 15-donor pollen mix to the flowers of the same 15 plants revealed that the most successful mate pairs had an intermediate genetic dissimilarity, suggesting a fine-scale intermediate OOD (Krauss, 2000). Further assessment of these results found that the individual fitness of pollen donors was more important in determining siring success than the distance between mates (Krauss, 2000). Here, for A. manglesii, we propose that the genetic and spatial distance between mate pairs is the primary driver behind the success of pollen donors. Siring success decreased with increasing spatial and genetic distance between mates, suggesting that near-neighbour mating gives a paternal donor a competitive advantage compared with more distant crosses. This is likely being driven by female choice, interactions between pollen and the pistil, and competitive interactions between pollen grains from multiple donors.

Mechanisms that allow the pistil of a flower to recognize and reject pollen from closely related genetic individuals are widespread across plant species (Hiscock and Allen, 2008). The rejection of pollen can occur pre-fertilization, where it is driven by a complex series of chemical and cellular interactions between pollen and stigma, resulting in reduced pollen grain germination, or pollen tube growth (e.g. Silva and Goring, 2001; Souto et al., 2002; Hiscock and Allen, 2008). We know this can be influenced by distance between parents, with 10-m crosses of Delphinium nelsonii resulting in 20 % more pollen tubes than any other pollen cross (Waser and Price, 1993). There are no pre-fertilization barriers in A. manglesii after self-pollen deposition, with pollen tubes from self-pollen grains successfully reaching the ovules (Sukhvibul, 1991). Instead, self-incompatibility occurs post-fertilization, with most self-fertilized ovules aborted (Sukhvibul, 1991). In seed plants generally, abortion of inbred embryos occurs during early seed development, with considerable evidence indicating that plants abort seeds selectively (Bookman, 1983; Lee, 1984; Casper, 1988; Rocha and Stephenson, 1990; Niesenbaum, 1999). Here, both pre- and post-fertilization barriers may also be reducing seed production after crosses >3 m. Due to maternal epigenetic factors and paternal gene expression, mating between distantly related individuals can lead to lower or slower pollen germination, stunted or slower pollen tube growth, and non-random seed abortion, though these effects are typically seen between taxa as a form of reproductive isolation (e.g. Lopez et al., 2000; Kirkbride et al., 2015; Swanson et al., 2016).

Lower overall seed set after multi-donor pollination compared with OOD single-donor pollination appears to be a result of non-random seed abortion of suboptimal mates (associated with distance), with multi-donor pollination resulting in 35 % fewer seeds per fruit than the observed maximum. Selective abortion has been quantified in other species using supplementary hand pollination and anther removal, where the application of outcross pollen with self-pollen results in lower seed set than seed set following pollination with only outcross pollen (Marshall, 1988; Waser and Price, 1991; Aizen and Harder, 2007). The removal of anthers from immature flowers can also raise overall seed set per fruit (de Jong et al., 1992). Abortion rates can be influenced by the timing of pollen application, with simultaneous application of self and outcross pollen sometimes lowering set (Waser and Price, 1991). In some species, self-pollen appears to influence seed set only if applied before outcross pollen, suggesting that when they are provided together outcross pollen is outcompeting self-pollen (Galen et al., 1989).

Although the genetic distance between mates was associated with individual siring success, there was a secondary effect of the siring success of individual paternal donors (for example, see donor 6 in Fig. 4). These differences between paternal donors could be partly explained by variation in the absolute number of pollen grains present on an anther and hence applied to a stigma (as we applied one anther from each pollen donor to our mix) or differences in pollen grain viability and vigour. Most A. manglesii anthers produce >95 % viable pollen, although individuals with <80 % viable have been documented (Hopper, 1977). Anigozanthos manglesii plants have long styles up to 4 cm, providing ample opportunity for pollen competition as pollen tubes race towards ovules (Lankinen and Skogsmyr, 2001). This may allow individuals with faster pollen tube growth and/or greater compatibility with the maternal tissue to father a greater percentage of offspring. Pollen tube growth rate has been correlated with the number of seeds sired by individual plants (Snow and Spira, 1991; Lankinen and Skogsmyr, 2001), but is yet to be investigated in detail in A. manglesii. The presence of multiple fertilized ovules (up to 280) within the fruit of A. manglesii provides a further opportunity for mate–mate competition and differential pollen donor success through selective abortion of developing embryos.

Within-population optimal outcrossing after single-donor pollination

The clear pattern of within-population optimal outcrossing after single-donor hand pollination was associated with spatial genetic structure within this population, where mate pairs <2.5 m apart were genetically more related than mate pairs >2.5 m apart. This fine-scale genetic structure suggests that the historical introduction of A. manglesii seeds until the 1990s into parts of the Kings Park bushland (R. Glowacki, Botanic Gardens and Parks Authority, Perth, Western Australia pers. comm.) has no impact on natural population structure at our study site. Instead, it likely reflects the natural patterns of seed dispersal via gravity, while beyond 2.5 m the spatial genetic structure is being eroded by the highly mobile behaviour of bird pollinators, a situation also surmised for other populations of bird-pollinated plants (Williams and Waser, 1999; Krauss et al., 2009).

The spatial genetic structure evident in A. manglesii, coupled with reduced seed set and germination success in self- and nearest-neighbour-pollinated flowers, indicates that genetic factors are responsible for decreased progeny performance following inbreeding. Lower total germination and germination vigour (measured as time to 50 % germination) were seen after mating among nearest neighbours (<1 m), despite all sown seeds having a visible endosperm and being heat-shocked to break dormancy. No difference in germination success or seedling vigour was seen after crosses >1 m, showing that while outbreeding depression only influences seed production, the impact of inbreeding depression extends to early-stage seedling performance. These impacts may extend into later growth stages, with earlier germinants sometimes showing a competitive advantage over later germinants, though this has been primarily tested between species (Seiwa, 1997; Weitbrecht et al., 2011; Baskin and Baskin, 2015; Vaughn and Young, 2015). Nevertheless, germination success has been found to be negatively correlated with mating between closely related individuals in other species (Baskin and Baskin, 2015).

While there was a ~35 % reduction in the number of seeds per fruit after single-donor crosses of >3 m, this reduction is not as severe as the ~70 % reduction after mating between nearest neighbours. Consequently, for seed set in A. manglesii the strength of inbreeding depression is at least twice that of outbreeding depression (relative to the OOD). All crosses between mates >3 m apart produced ~100 seeds per fruit, a similar amount to that found following single-donor interpopulation crosses of up to 400 km (Hopper, 1980). This suggests that at a landscape scale there is a negligible impact of inter-mate distance on reproductive success in A. manglesii, at least following single-donor pollination. Given the high seed set after pollination at any distance >1 m, high levels of multiple paternity after mixed pollen deposition and uniformly high germination success, frequent mating beyond the OOD is unlikely to have an overall negative fitness impact on A. manglesii persistence.

Strategies favouring outcrossing are predicted to be common in the flora of the SWAFR, where OCBILs are common (Hopper and Gioia, 2004; Hopper, 2009) and plant populations are naturally patchy within the landscape. Hypotheses as to why outcrossing is favoured focus on the James effect (Hopper, 2009) – natural selection for genetic, cytogenetic or phenotypic adaptations (such as floral traits that encourage bird visitation and hence pollination) leading to frequent outcrossing and pollen movement across long distances to conserve (and even enhance) heterozygosity in the face of inbreeding due to small population size (Hopper and Burbidge, 1986; Krauss et al., 2017). This heterosis has been demonstrated in another bird-pollinated Southwest Australian species, Banksia ilicifolia, where seed set after crossing between plants 20 km apart was greater than that following local crossing (Heliyanto et al., 2005). Pollination is expected to be especially important in achieving outbreeding as many OCBIL species lack the ability to disperse seeds widely (Hopper, 2009).

Pollinator strategy and optimal outcrossing

Of the 22 species we are aware of where tests have been undertaken for within-population optimal outcrossing, eight are predominately bird-pollinated, 13 are insect-pollinated, one is wind-pollinated and one is self-pollinated (Table 1). Among these studies, five of eight (62.5 %) primarily bird-pollinated species exhibited within-population optimal outcrossing compared with four of 13 (30.8 %) primarily insect-pollinated species. Like A. manglesii, most of the species investigated are herbs with limited seed dispersal, suggesting that life form is not contributing to the difference. Instead, pollination strategy could play an important role in the prevalence of within-population optimal outcrossing. Of the other bird-pollinated studies, two hummingbird-pollinated species showed an OOD distance of between 1 and 10 m (Price and Waser, 1979; Waser and Price, 1989), the honeyeater-pollinated Grevillea mucronulata showed an OOD at 30–55 m (Forrest et al., 2011), while Grevillea repens showed decreasing fruit and seed set with increasing pollination distances in multiple populations, and the opposite in another (Holmes et al., 2008). The conflicting results between populations for G. repens was attributed to the impact of population genetic structure, with the population that showed evidence of inbreeding depression benefiting from wide outcrossing. This highlights the importance of understanding within-population genetic variation and the spatial-genetic structure of populations (Holmes et al., 2008).

Table 1.

Species where within-population optimal outcrossing has been assessed after experimental single-donor hand pollination

Species	Life form	Seed dispersal	Scale assessed	Fitness assessed	Within-population OOD	Pollinators	References
Delphinium nelsonii	Herb SC	Within 1 m	Self,1, 3, 10, 30, 100, 1000 m	Seed set per flower Seedling survival, growth rate	3–10 m: Highest seed set, highest survival and faster production of flowers	Hummingbirds Bumblebees	Price and Waser, 1979; Waser and Price, 1994
Castilleja miniata	Herb SI	Unreported	Self, 0.5, 2, 10, 30 m (1979) and self, 0.5, 1, 5, 10, 30 m (1980)	Seeds per pod	2 m (1979 and combined): Seeds per pod (no OOD detected in 1980)	Hummingbirds	Lertzman, 1981
Blandfordia nobilis	Herb SI	Unreported	0–5, 10, 50, 200 m	Seed set per unparasitized carpel	Not detected: No difference between treatments	Honeyeaters	Zimmerman and Pyke, 1988
Espeletia schultzii	Shrub SI	Within 2 m	1, 3, 6, 10, 15, 20, 50, 150, 500 m, 10 km, 78 km (variation between sites and years)	Proportion of filled achenes, proportion of aborted seeds, seedling survivorship	Not detected: Lower success after near-neighbour crosses	Hummingbirds Bumblebees (Berry and Calvo, 1989)	Sobrevila, 1988
Ipomopsis aggregata	Herb SI	Within 1 m	0, 1, 10, 100 m	Seed set, seedling survival, flowering time,	10 m: Higher seed set, higher seedling survival, greater chance of flowering/earlier flowering	Hummingbirds	Waser and Price, 1989
Impatiens capensis	Herb SC	Ballistically; within 20 cm	2, 10, 50 m	Seed mass, seed maturation, emergence, seedling height	29 m (estimate): Height of offspring at 1 month	Hummingbirds Bumblebees (Travers et al., 2003)	McCall et al., 1991
Anchusa crispa	Herb SC	Ants; within 75 cm (Quilichini and Debussche, 2000)	Self, 5–10, 10–20, 20–30 m	Achenes per fruit, germination, survival, cymes per plant, flowers per cyme	Not detected: Selfed-pollinated plants have lower number of achenes per fruit and germination but higher survival, cymes per plant and flowers per cyme	Hummingbirds Insects	Quilichini et al., 2001
Grevillea mucronulata	Shrub SI	Gravity	Self, 1–4 m, 30–50 m, 5 km	Initiation, seed set, seed size, seed mass, germination time, seedling height, leaf area, number of leaves	30–50 m: Highest seed set, seed size, germination, leaf area, number of leaves	Honeyeaters	Forrest et al., 2011
Picea glauca	Tree	<200 m	0–3200 m (including self and nearest neighbour)	Seed set, germination, seedling growth,	Not detected: Lower viability and seedling height after self-pollination	Wind-pollinated	Coles and Fowler, 1979
Scleranthus annuus	Herb SC	Gravity	Self, 6, 12, 25, 50, 100 m	Seedling stamen arrangements and fertility	6–12 m: Highest number of stamens, more staminodes developing into fertile stamens	Self (occasionally ants)	Svensson, 1988
Polemonium viscosum	Herb SI	Unreported	Randomly paired up to 60 m	Number of seeds, seed mass	Not detected: Distance has no impact	Bumblebees	Newport, 1989
Costus allenii	Herb SC	Unreported	Mean interparental distances of 60, 200, 264 and 318 m; included multiple pollen donors	Seed production, seed weight, percentage germination, vegetative growth	Average 264 m: Highest seed production and percent germination	Euglossine bees (Schemske, 1981)	Schemske and Pautler, 1984
Phlox drummondii	Herb SI	Explosive force of fruit	Self, random up to 35 m	Seed abortion	Not detected: Self-pollination increases seed abortion	Lepidopterans	Levin, 1984
Amianthium muscaetoxicum	Herb near SI	Unreported	5, 15, 60 m	Fruit production, seeds per fruit, seeds per plant, seed mass	Not detected: Lower fruit production and seeds per fruit after near-neighbour crosses in two of three populations	Beetles (Travis, 1984)	Redmond et al., 1989
Gentiana pneumonanthe	Herb SC	Wind; within a few metres	Self, 1–50 cm, 50–100 cm, 1–2 m, 2–5 m, 5–10 m, 10–20 m, >20 m and 800 m	Seed set and seed mass, germination, seedling mass, adult mass, total relative fitness	Not detected: Lower fitness after self-pollination	Bumblebees	Oostermeijer et al., 1995
Agave schottii	Shrub SC	Within a few metres	1, 10, 100, 2500 m	Fruit set, seed set, seed mass, germination, combined relative fitness	10 m: Seed set combined relative fitness	Bumblebees and sphinx moths (Schaffer and Schaffer, 1977)	Trame et al., 1995
Gentianella germanica	Herb SC	Wind; <130 cm (Verkaar et al., 1983)	Self, 1 m, 10 m, 25 km	Seeds per fruit, seed mass, germination rate, number of leaves, rosette diameter, cumulative leaf length, survival of seedlings (over 175 d period)	10 m: Faster germination, larger number of leaves, rosette size	Diptera and small bees	Fischer and Matthies, 1997
Trillium erectum	Herb SC	Ant dispersal; 0.02–10 m	1, 10, 100, 1500 m	Fruit and seed production, seed: ovule ratio	Not detected: Lower seed production after self-pollination	Insect	Irwin, 2001
Trillium grandiflorum	Herb SC	Ant dispersal; 0.02–10m	1, 10, 100, 1500 m	Fruit and seed production, seed:ovule ratio	Not detected: Lower seed production after self-pollination	Insect	Irwin, 2001
Alstroemeria aurea	Herb SC	Within a few metres	1, 10, 100 m	Germinated pollen grains, number of pollen tubes,	Not detected: Pollen tube number increases with distance	Bumblebees	Souto et al., 2002
Digitalis purpurea	Herb SC	Ballistically; within a few metres	Self, 1, 6, 30 m	Seed set, seed mass, germination, seedling survival, seedling growth, time to flowering, flower number	6 m: Highest cumulative fitness (seed germination, juvenile survival, flower number)	Bumblebees	Grindeland, 2008
Grevillea sphacelata	Shrub SI	Unreported	<5 m, 200 m, 3–4 km away	Fruit development, fruit set, seeds per fruit, seed mass, germination success, seedling survival, seeding growth, seedling photosynthesis	Not detected: Distance had negligible impact	Insects (primarily Apis mellifera)	Ayre and O’Brien, 2013
Phlox hirsuta	Herb SI	Gravity; within 1 m	1, 10, 100 m, 6740 m	Seed set, seed abortion, seed mass, emergence, radical length, cotyledon emergence, number of leaves, number of flowers	Not detected: No impact within population, shorter radicles and less cotyledon emergence after between-population crosses	Insects	Ruane et al., 2015

SI, self-incompatible; SC, self-compatible.

While the fine-scale genetic structure of bird-pollinated species is not often reported, two North American bird-pollinated species display similar spatial genetic structure to A. manglesii. The hummingbird-pollinated Ipomopsis aggregata (Table 1) displays significant spatial autocorrelation at <5 m (Campbell and Dooley, 1992), while the insect- and hummingbird-pollinated Delphinium nuttallianum displays structure at <10 m (Williams and Waser, 1999). The presence of fine-scale genetic structuring and small genetic neighbourhoods may partially explain the presence of within-population OOD.

A higher prevalence of within-population OOD in bird-pollinated species is unexpected. Inbreeding depression in most of the reported species can be explained by short-distance seed dispersal leading to fine-scale spatial genetic structuring and biparental inbreeding. Outbreeding depression at the small scales observed is much harder to explain, particularly in bird-pollinated species, with pollination by birds often leading to extensive pollen dispersal within and sometimes between plant populations (Krauss et al., 2009; Llorens et al., 2012; Frick et al., 2014; Bezemer et al., 2016), which erodes fine-scale spatial genetic structure (Krauss et al., 2009). The apparent prevalence of both outbreeding and inbreeding depression may be partially explained by short-distance seed dispersal, with overlapping genetic neighbourhoods increasing the chance of mating between highly unrelated individuals occurring after crosses over short distances. While more studies are needed to test the generality of our observations, and the mechanisms behind it, an explanation of higher prevalence of OOD in bird-pollinated species could be that high pollen carryover and therefore mixed-donor pollination creates more opportunities for pollen competition and female choice (both pre- and post-fertilization) within the long styles that are typical of bird-pollinated plants (Hanley et al., 2009). This may give bird-pollinated species a greater capacity for competitive interactions among pollen donors and between pollen donors and the maternal plant. More studies that combine hand pollination with an understanding of fine-scale spatial genetic structure may show that optimal outcrossing within populations is more common in species with a combination of restricted seed dispersal, wide pollen dispersal, deposition of pollen from multiple donors on stigmas, and long styles. Implementing more realistic scenarios of multi-donor pollination is critical in revealing more details of the complex relationship between inter-mate distance and fitness in plant populations.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Figure S1: map of nine maternal plants and eight pollen donors used for mixed-pollination treatment. Table S1: number and frequency of alleles in this study.

FUNDING

This work was supported by an Australian Research Council Discovery Grant to S. Hopper, S. Krauss and R. Phillips (DP 140103357), an Australian Government Research Training Program Scholarship and a Holsworth Wildlife Research Endowment Equity Trustees Charitable Foundation grant to B. Ayre. Professor S. Hopper was supported by a Discovery Outstanding Researcher Award (attached to DP 140103357) and by grants from the Great Southern Development Commission and the Jack Family Trust.