Interspecific seed discounting and the fertility cost of hybridization in an endangered species


  • Kevin S. Burgess,

    1. Department of Botany, University of Guelph, Guelph, Ontario N1G 2W1, Canada;
    2. Present address: Department of Ecology & Evolutionary Biology, University of Toronto, 25 Willcocks Street, Toronto, Ontario M5S 3B2, Canada
    Search for more papers by this author
  • Martin Morgan,

    1. School of Biological Sciences, Washington State University, PO Box 644236, Pullman, WA 99164-4236, USA;
    Search for more papers by this author
  • Brian C. Husband

    1. Department of Botany, University of Guelph, Guelph, Ontario N1G 2W1, Canada;
    Search for more papers by this author

Author for correspondence:
Kevin S. Burgess
Tel: +1 416 9460007
Fax: +1 416 9785878


  • • Hybrid fertilizations can have negative demographic effects on taxa by usurping ovules that would otherwise give rise to nonhybrid offspring. The consequent reduction in conspecific matings may be exaggerated in rare taxa and constitutes a fertility cost that has rarely been quantified.
  • • Here, the effect of interspecific mating was estimated on the fecundity of locally rare red mulberry (Morus rubra), which hybridizes with introduced white mulberry (Morus alba) and red ¥ white hybrids. First, the asymmetry in pollen production among red, white and hybrid mulberry in two sympatric populations was quantified. The fertility cost of hybridization was then assessed experimentally by estimating seed production and rates of interspecific mating in red mulberry trees from plots where white and hybrid mulberry trees were selectively removed.
  • • On average, the percentage of mulberry pollen per plot produced by red mulberry (8%) was significantly lower than the mean for white and hybrid mulberry combined (92%). Experimentally removing white and hybrid mulberry increased the siring fertility of red mulberry by 14% but produced no change in seed set.
  • • Results indicate that seeds of red mulberry, ordinarily sired by conspecific pollen, are being discounted through fertilization of ovules by heterospecific pollen, which may contribute to local decline of red mulberry.


Hybridization can have multiple consequences for the regulation of biological diversity, one of which is the local extinction or extirpation of species (Ellstrand, 1992; Levin et al., 1996; Rhymer & Simberloff, 1996; Arnold, 1997). Extirpation may result when interspecific mating directly reduces the number of conspecific offspring produced (a form of demographic swamping) and facilitates the incorporation of one parent's genome into that of another (genetic assimilation) (Levin et al., 1996; Huxel, 1999; Haygood et al., 2003). These processes are rarely symmetrical and, all else being equal, will occur in proportion to the relative abundance of the hybridizing taxa. As a result, species in small populations may be particularly vulnerable to the direct impacts of hybridization (Rieseberg, 1991; Levin et al., 1996; Rhymer & Simberloff, 1996).

Hybridization can impose a fertility cost on parental taxa by reducing the formation of conspecific offspring as an outcome of competition among pollen tubes for fertilizations. This cost may affect the abundance of parental taxa when interspecific fertilizations usurp ovules that would otherwise have generated conspecific offspring (Levin et al., 1996; Wolf et al., 2001; Buerkle et al., 2003; Haygood et al., 2003). A comparable process, referred to as seed discounting, has been described for intraspecific plant mating, in which cross-fertilization is pre-empted by self-fertilization (Lloyd, 1992; Herlihy & Eckert, 2002). In the same way, interspecific seed discounting can occur through hybridization. This process will operate even if hybrids are not viable (Ellstrand, 1992; Ellstrand & Elam, 1993; Haygood et al., 2003), providing that ovules (and hence seeds) are no longer available to conspecific pollen. If viable hybrids are produced, they can further enhance the effect of discounting by increasing the numerical disadvantage of the rare taxon (Levin et al., 1996). Interspecific seed discounting redirects resources and seeds that would otherwise be destined for conspecific offspring, and ultimately can drive rare taxa to increasingly lower frequencies through a positive feedback process (Potts et al., 2003).

The magnitude of the fertility cost experienced by a rare taxon will depend on a number of factors, including physical proximity to congeners; asymmetries in pollen production caused by differential abundance, flower production or gametophyte production; and the degree to which interspecific matings replace conspecific offspring (interspecific seed discounting). The conditions that favor a fertility cost are not always fully met in naturally hybridizing populations. With respect to interspecific seed discounting, it is important to note that hybrid fertilizations will not necessarily replace conspecific matings in the rare taxon, especially when there are uncommitted ovules available (i.e. when there is pollen limitation) and when ovules are partially sheltered by reproductive barriers (Wolf et al., 2001). As a result, the presence of hybridizing taxa may increase total seed production but not decrease the absolute number of conspecific offspring. Because interspecific seed discounting will not always be proportional to the frequency of hybridization, as is often assumed, it should be tested empirically by manipulating hybridization rates and observing the impact on seed production. We are aware of no such test of the fertility cost of hybridization (but see Buggs & Pannell, 2006).

In this study we estimated the fertility cost of hybridization in red mulberry (Morus rubra), an understory tree that is considered endangered in Canada. At the northern edge of its geographical range in southern Ontario, red mulberry is restricted to six populations (Ambrose & Kirk, 2004). In four of these locations, it co-occurs with white mulberry (Morus alba), which was originally introduced to North America from Asia in the 1600s and subsequently became naturalized (Gleason, 1952). Hybrids, likely from backcross generations, have been detected in all sympatric populations (Burgess et al., 2005). Notably, hybrid abundance ranges from 43 to 67% across populations and approx. 68% of all hybrids are introgressed toward white mulberry (Burgess et al., 2005). Interspecific seed discounting as a result of hybridization is especially likely in red mulberry, as the species is: in the minority within each sympatric population; dioecious, so there is no opportunity to ‘shelter ovules’ through selfing; and wind pollinated and phenologically overlapping with white mulberry, so there are likely few pre-mating barriers to hybridization. We had two specific objectives in this study: to estimate the frequency of red and white + hybrid (i.e. all nonred) mulberry in the pollen pool as a measure of the numerical disadvantage experienced by red mulberry; and to estimate the magnitude of interspecific seed discounting resulting from hybrid pollen. Interspecific seed discounting was evaluated by comparing the incidence of hybridization and seed set in plots with white and hybrid mulberry present (‘unmanipulated’) and plots from which they were experimentally removed (‘culled’). If red mulberry experiences a mating disadvantage, pollen production should be higher for heterospecific individuals (white + hybrid mulberry) than for red mulberry, and heterospecific pollen should replace red mulberry as sires.

Materials and Methods

Study sites

We established six 50-m-diameter plots, each centered on a focal female red mulberry (Morus rubra L., Moraceae) tree in each of two sympatric populations: Point Pelee National Park (Universal Transverse Mercator (UTM) coordinates, North American Datum (NAD) 83: 17 374500 N, 4646000 E) and Fish Point Provincial Reserve, Pelee Island (UTM: 17 360812 N, 4621309 E). Each plot was paired with a second plot, no closer than 25 m, and separated from other pairs by a minimum of 80 m.

Pollen production

We estimated the frequencies of viable pollen produced by red mulberry, white mulberry (Morus alba L.), and hybrid mulberry as well as by white and hybrid mulberry combined (i.e. all nonred mulberry pollen; hereafter, referred to as white + hybrid mulberry). For each taxon, pollen density was estimated as T × M, where T is the frequency of individuals per 50-m-diameter plot and M is the number of viable pollen grains produced per tree.

Frequency of individuals (T Excluding the central red mulberry tree, we counted the numbers of red, white and hybrid individuals of reproductive size (diameter at 1.35 m > 3 cm) in each plot. The three taxa were identified using a combination of morphological (leaf area, shape and abaxial trichome density) and genetic (randomly amplified polymorphic DNA (RAPD): nine species-specific and 43 polymorphic fragments) markers (Burgess et al., 2005; K. S. Burgess & B. C. Husband, unpublished data).

Pollen production per tree (M We calculated the mean number of viable pollen grains produced per red, white and hybrid mulberry tree as the product of mean catkin number per tree and number of viable pollen grains per catkin. Male catkin production was estimated for 10 randomly selected red, white, and hybrid mulberry trees at Point Pelee. Sample sizes were limited by the scarcity of reproductive red mulberry trees (Burgess et al., 2005). For each tree, we counted the number of catkins along four 1-m segments of four primary branches (facing different cardinal directions), averaged them, and then multiplied by the mean number of 1-m segments per branch and the number of primary branches per tree. We estimated the number of viable pollen grains per catkin for the same individuals. Nine mature catkins were collected from the four primary branches and the number of anthers per catkin was counted and averaged. One undehisced mature anther was sampled per catkin and transferred into 150 µl of 70% ethanol, crushed and vortexed to release pollen grains. Pollen was stained for 15–24 h in 20 µl of 0.1% acidified aniline blue and vortexed, and a subsample (0.9 µl) was placed on a hemocytometer cell counter (Improved Neubauer Bright line, Hausser Scientific, Horsham, PA, USA). Pollen grains within a 3 mm × 3 mm grid were counted and characterized as either ‘viable’ (darkly stained, full pollen grains) or ‘nonviable’ (faintly stained, unstained or deformed pollen grains). Counts were averaged for the nine replicates for each tree, recalculated as the number of pollen grains per anther and multiplied by the number of anthers per catkin. Mean catkin number per individual, mean number of viable pollen grains per catkin, and mean number of viable pollen grains per tree were calculated for red, white, and hybrid mulberry, as well as for white + hybrid mulberry.

Analysis  We examined variation in pollen density and its components using two-way ANOVA, with taxon (fixed effect), population (fixed effect), and taxon × population as sources of variation. Mean number of individuals of each taxon per plot (T), mean catkin number per individual, mean number of viable pollen grains per catkin, mean number of viable pollen grains per tree (M), and pollen density per taxonomic class (T × M) were all log-transformed to meet assumptions of normality and homoscedasticity. Taxon means were compared using Tukey's HSD. Populations were also compared with respect to the relative frequency of white + hybrid mulberry in the same way. All results are reported as back-transformed means.

Interspecific seed discounting

Experimental design  We used the 12 50-m-diameter experimental plots at Point Pelee National Park and Fish Point Provincial Reserve described in the section ‘Study sites’ above. Two treatments were randomly assigned to each pair of plots: removal of all white and hybrid mulberry trees (‘culled’); and no removal of white and hybrid mulberry trees (‘unmanipulated’). In the culled plots, all red mulberry trees were left untouched. Thirty fruits were randomly harvested in three successive years from each central red mulberry tree.

Response variables  To estimate seed set, we wetted each fruit for 24 h, removed the pulp, and counted the number of filled seeds per fruit. While it is possible that the total number of fruit may differ in the presence of hybridization, we measured the total number of seeds per fruit to provide a minimum estimate of discounting in sympatric populations. The effect of culling on the number of seeds per fruit across the three sampling periods (1999, 2000 and 2001) was assessed using repeated-measures manova with treatment, population and treatment × population interaction as sources of variation.

We estimated the incidence of conspecific (sired by red mulberry) and interspecific (sired by white and hybrid mulberry) fertilizations for seeds from the central trees in unmanipulated and culled plots using randomly amplified polymorphic DNA (RAPD) analysis. Approximately 187 fruit (= 15.6 fruit per tree) were randomly selected from the fruits collected in each year, wetted for 24 h and cleaned. We sowed approx. 3526 seeds ( = 293 seeds per tree) per yearly cohort onto moist filter paper in Petri dishes (maximum of 25 seeds per dish) and incubated them in a growth chamber for 2 wk (12 h light at 26°C and 12 h dark at 22°C). We then transplanted seedlings into 10-cm pots containing perlite, turface, and ProMix (Premier Horticulture Ltd, Dorval, QC, Canada) in 1 : 1 : 6 proportions and grew them in a glasshouse (16 h daylight and 8 h dark at 21°C). Leaf material was collected from approx. 30 progeny from each red mulberry tree (approx. 10 from each cohort) for DNA extraction. Total genomic DNA was isolated from approx. 80 mg of frozen leaf material using Qiagen Dneasy plant mini extraction kits (Qiagen, Mississauga, ON, Canada) yielding approx. 30 ng µl−1 of DNA. All progeny from unmanipulated and culled plots were screened for six RAPD fragments (primer #28 (CCGGCCTTAA); Nucleic Acid–Protein Service Unit, Biotechnology Laboratory, University of British Columbia, Vancouver, BC, Canada); three independent loci previously identified as diagnostic of red, white and hybrid mulberry and three polymorphic but informative loci (Burgess et al., 2005). Mean frequency differentials for the three polymorphic loci across our study sites were 0.38, 0.09 and 0.28. Amplification and visualization of RAPD fragments followed the protocols of Burgess et al. (2005).

The paternity of red mulberry offspring was estimated using a maximum likelihood model (program available upon request from M. Morgan). This model partitions offspring paternity amongst discrete groups of individuals. Specifically, it estimates the likelihood that a particular offspring from a known maternal lineage is sired by a certain paternal class, given all possible paternal genotypes in the local paternity pool. To do this, the model incorporates information on the relative frequency of each paternal class into a maximum likelihood estimation procedure. The model makes the following assumptions.

  • 1All markers are dominant.
  • 2Each maternal haplotype is known without error.
  • 3Sufficient offspring are assayed in each family to correctly infer (diploid) maternal genotype.
  • 4Sufficient paternal types are assayed to estimate allele frequencies at each locus.

The model calculates the likelihood of offspring genotype i, given known maternal genotype and known frequency of paternal allele X, where X indexes types alba, rubra and hybrid. The likelihood of offspring genotype i, given known maternal parent and paternal allele frequency in type X, is written as L(i | X). The fraction of paternity assigned to type X is


This is the ‘standardized fertility index’ of the ith offspring.

We estimated allele frequencies for each donor class in each population from previous estimates of RAPD marker frequencies for red, white and hybrid mulberry (Burgess et al., 2005) and the frequency of donor classes in the pollen pool (from the section ‘Pollen production’ above). To estimate fertility rates for each paternal taxon, standardized fertility index scores, which range from 0 to 1, were averaged over progeny for each unmanipulated and culled plot.

We tested for variation in fertility of red vs nonred (white + hybrid) mulberry (and red vs white vs hybrid) among populations and treatments (and all interactions) using logistic regression. Because there was a significant taxon × population × treatment interaction, we then compared the fertility values between red and white + hybrid mulberry (and red, white and hybrid mulberry separately) in unmanipulated vs culled plots for each population separately using 2 × 2 and 2 × 3 contingency analyses. All statistical analyses were performed using jmp® statistical software, version 5.0 (SAS Institute, 2002).


Pollen production

The frequency of trees per plot differed among taxa and between populations, but there was no taxon × population interaction (Table 1). Overall, the number of trees per plot averaged 0.9 (6% of all mulberry) for red mulberry, 3.7 (25%) for white mulberry and 10.5 (69%) for hybrids (Table 2). The mean frequency was higher at Fish Point than at Point Pelee (Table 2), but red mulberry was significantly less common than white + hybrid mulberry (combined) in both populations. Red mulberry was less abundant than hybrids in both populations, while hybrids were more frequent than white and red mulberry at Fish Point, but did not differ in frequency from white mulberry at Point Pelee (Table 2).

Table 1.  Summary of anova for five components of pollen production in red (Morus rubra), hybrid and white (Morus alba) mulberry in southern Canada
VariableSource of variationF (red, white + hybrid)F (red, white, hybrid)
  1. Each anova tested for effects of taxon, population and taxon × location interaction and for differences between red and white + hybrid mulberry combined and for all taxa separately. Degrees of freedom for each respective source of variation were 1, 1, 1 for red and white + hybrid (error d.f. = 26 (20 for tree frequency per plot)), and 2, 1, 2 for red, white and hybrid (error d.f. = 24 (30 for tree frequency per plot)). *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Tree frequency per plotTaxon28.44***19.47***
Taxon × population3.232.74
Number of catkins per treeTaxon0.514.63*
Taxon × population3.042.53
Number of pollen grains per catkinTaxon3.254.60*
Taxon × population2.011.35
Total number of pollen grains per treeTaxon0.074.73*
Taxon × population1.511.29
Pollen production per plotTaxon6.20*14.32***
Taxon × population0.690.79
Table 2.  Mean frequency of red (Morus rubra), hybrid and white (Morus alba) mulberry trees per plot (T) in two sympatric populations in southern Ontario, Canada
PopulationWhite + hybridF (red, white + hybrid)RedHybridWhiteF (red, white + hybrid)
  1. Mean values are shown, with the standard error in square brackets and the proportion in round brackets.

  2. anovas of taxon differences were compared for red vs white + hybrid mulberry combined and for all three classes separately. Means with different superscript letters are significantly different based on Tukey's HSD test. Degrees of freedom for each respective source of variation were 1, 1, 3 for white + hybrid (error d.f. = 10 (20 for all sites)), and 2, 2, 5 for red, white and hybrids (error d.f. = 15 (30 for all sites)). *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Point Pelee9.1A10.78**0.4B6.5A2.6AB7.18**
[2.69] [0.16][1.82][0.88] 
(0.97) (0.03)(0.71)(0.26) 
Fish Point19.2A17.9**1.3B14.5A4.7B12.55***
[4.17] [0.56][3.10][1.07] 
(0.90) (0.10)(0.68)(0.22) 
All sites14.2A12.16***0.9B10.5A3.7B10.46***
[2.8] [0.31][2.09][0.73] 
(0.94) (0.06)(0.69)(0.25) 

The remaining components of pollen production differed significantly among taxa but not between populations, and none of the taxa × population interactions was significant (Table 1). Catkin production per red mulberry tree ( = 16 128) did not differ significantly from the mean of white + hybrid mulberry ( = 14 210) (Table 3) or from the separate means of white and hybrid mulberry (Table 3). Pollen production per catkin for red mulberry ( = 5.2 × 106) did not differ from the mean of white + hybrid mulberry (4.0 × 106) or the mean of hybrids alone; white mulberry produced the lowest mean number of pollen grains per catkin, significantly lower than that of red mulberry (Table 3). The number of pollen grains per tree for red mulberry ( = 8.4 × 1010) did not differ significantly from the mean of white + hybrid mulberry ( = 7.0 × 1010) or the separate means of white and hybrid mulberry; the latter taxon had the highest number of pollen grains per tree (Table 3). Mean pollen production per plot for red mulberry ( = 4.74 × 1010 pollen grains) was significantly lower than for white + hybrid mulberry ( = 5.85 × 1011 pollen grains) and for hybrids alone, but was not significantly different from that for white mulberry, which was intermediate (Table 3). The overall proportions of red and white + hybrid mulberry pollen within the pollen pool were 0.08 and 0.92, respectively.

Table 3.  Means (with standard errors) for four components of pollen production in red (Morus rubra), hybrid and white (Morus alba) mulberry from southern Ontario, Canada
ComponentRedHybridWhiteWhite + hybrid
  1. Means with the same superscript letter are not statistically different using a Tukey's HSD test.

Number of catkins per tree16 128ab24 643a4157b14 210a
(10 968)(6794)(1097)(4066)
Number of pollen grains per catkin5.2 × 106 a4.8 × 106 ab3.2 × 106 b4.0 × 106 a
(6.6 × 105)(6.9 × 105)(4.4 × 105)(4.4 × 105)
Number of pollen grains per tree8.4 × 1010 ab1.3 × 1011 b1.4 × 1010 a7.0 × 1010 a
(5.4 × 1010)(4.3 × 1010)(4.2 × 109)(2.5 × 1010)
Number of pollen grains per plot4.7 × 1010 a1.1 × 1012 b4.9 × 1010 a5.9 × 1011 b
(2.5 × 1010)(3.1 × 1011)(3.4 × 1010)(2.0 × 1011)

Interspecific seed discounting

Estimates of pollen fertility (which approximates siring rate) for red, white, and hybrid mulberry were based on genetic analyses of an average of 23 progeny per unmanipulated plot (total N = 143) and 19 progeny per culled plot (total N = 115 progeny). The paternal taxa × treatment interaction in the logistic regression (Table 4) revealed that, across the two populations, the proportion of seeds from the central red mulberry that were sired by red mulberry (vs white + hybrid combined) was weakly dependent on treatment (χ2 = 3.35, P = 0.067; d.f. = 1). Red mulberry sired 23% of its own seeds in unmanipulated plots (hence white + hybrid mulberry sired 77%), whereas culling significantly increased this value by 14% (to 37%) (Fig. 1). When hybrid and white mulberry fertilities were examined separately, white mulberry siring success was found to have significantly increased from 18 to 35% as a result of culling, whereas hybrid siring success was reduced from 59% in unmanipulated plots to 28% in culled plots (χ2 = 9.05, P = 0.011; d.f. = 2).

Table 4.  Logistic regression analysis of mulberry paternal fertility
Source of variationχ2 (red, white + hybrid)Pχ2 (red, white, hybrid)P
  1. χ2 and P values are given for fertility associated with taxon, population (Fish Point vs Point Pelee), and treatment (culled vs unmanipulated). The effect of taxon was based on two paternal classes (red and white + hybrid) and three paternal classes (red, white and hybrid). The treatment × taxon interaction measures the dependence of red mulberry fertility (vs heterospecific fertility) on removal treatment. The treatment × taxon × population interaction reflects the extent to which this effect depends on population.

Treatment × taxon3.350.0679.0470.011
Treatment × population0.110.7420.0180.895
Taxon × population5.150.02413.8710.001
Taxon × population × treatment5.400.0205.1820.075
Figure 1.

Percentage of conspecific mating and the number of seeds per fruit on the central red mulberry tree in unmanipluted and culled plots for two sympatric populations in southern Ontario. The fertility of red mulberry increased significantly by 14% in culled plots (χ2 = 6.29, P < 0.05) with no change in seed set (F1,7 = 0.51, P > 0.1).

The effect of culling on the fertility of red mulberry (vs white + hybrid) was dependent on population (taxon × population × treatment interaction: χ2 = 5.40, P = 0.020; d.f. = 1; Table 4). Culling had a significant effect at Point Pelee (2 × 2 contingency analysis: χ2 = 26.43, P < 0.0001; d.f. = 1) but not at Fish Point (2 × 2 contingency analysis: χ2 = 0.61, P > 0.1; d.f. = 1). At Point Pelee, culling caused an increase in red mulberry fertility of 33.8% (from 2.7 to 36.5%). No seeds were sired by white mulberry; hence, the reduction in siring success of white + hybrid was attributable solely to the effects of culling on hybrid fertility. At Fish Point, there was a significant effect of culling when hybrid and white mulberry were treated as separate paternal classes (2 × 3 contingency analysis: χ2 = 8.81, P < 0.05; d.f. = 2), reflecting the fact that culling decreased hybrid fertility by 14% but had no effect on fertility of white mulberry.

In the manova, seed production per fruit by red mulberry in unmanipulated plots ( = 20.6 seeds per fruit) did not differ significantly from that in culled plots ( = 24.6 seeds per fruit) (F1,7 = 0.51, P > 0.1). Furthermore, the effects of population on seed production (F1,7 = 0.41, P > 0.1) and the treatment × population interaction (F1,7 = 0.75, P > 0.1) were not significant.


Culling of hybrid and white mulberry from 50-m-diameter plots reduced the overall proportion of hybrid seeds produced by focal red mulberry trees, but it had no effect on seed production per fruit. Therefore, we conclude that when hybrid and white mulberry are present they sire seeds at the expense of those sired by red mulberry pollen; hence, there is evidence for interspecific seed discounting. Furthermore, interspecific seed discounting was complete in that the increase in the hybridization rate of 14% (34% at Point Pelee alone) in the presence of white and hybrid mulberry was matched by an equivalent reduction in the number of red mulberry offspring. Extrapolating to natural populations (i.e. unmanipulated plots), where 77% of offspring produced by red mulberry are hybrids, heterospecific seed discounting represents a significant cost to the production of conspecific offspring by red mulberry when it grows in association with white mulberry.

Although there is a close correspondence between the increase in hybridization rate and the reduced production of conspecific offspring in red mulberry, this is not necessarily the case in other populations or species. If uncommitted ovules are available (i.e. when there is pollen limitation) or when ovules are partially sheltered by reproductive barriers (Wolf et al., 2001), competition for ovules may be restricted to only a portion of the ovary, and thus hybrid fertilizations may be limited to only a subset of the ovules currently fertilized by conspecific pollen. That interspecific seed discounting is complete in red mulberry indicates that seed production is not pollen-limited and red mulberry is mostly interfertile with white and hybrid mulberry. Additional measures of the fertility cost of hybridization are needed to determine the circumstances under which interspecific seed discounting will be important.

Culling had a significant effect on red mulberry fertility in one of the two populations used in this study. The absence of an effect at Fish Point may be related to differences in the relative frequency of red mulberry between populations. If white and hybrid mulberry are less abundant at Fish Point, the removal treatment may have a relatively small effect on the frequency of heterospecific pollen in the pollen pool. This hypothesis is supported by a broad survey of mulberry in Canada, in which hybrid and white mulberry occurred at lower relative frequencies at Fish Point than at Point Pelee (Burgess et al., 2005). However, in this study we found the absolute frequency of hybrid and white mulberry to be higher at Fish Point and the relative frequency (0.06) to be similar to that at Point Pelee (0.04). This discrepancy in results may be related to differences in the intensity and spatial scale of sampling in the two studies. Therefore, the lack of effect of culling at Fish Point remains ambiguous. Regardless of the cause, however, the absence of a shift in fertility at Fish Point precludes any conclusions about interspecific seed discounting at this location, as discounting can be detected only when there is variation in the relative siring rates of conspecific and heterospecific taxa.

The significant effect of culling hybrid and white mulberry within 25 m of the central red mulberry at Point Pelee underscores the importance of local trees as pollen donors in mulberry. Previous studies have shown that, in wind-pollinated trees, most pollen travels > 20 m and originates from an area between 1 and 10 000 km2 (Levin, 1988; Ellstrand, 2003). Some work has shown that pollen dispersal distances in other wind-pollinated species can be limited depending on local wind conditions (Meagher et al., 2003; Bacles et al., 2005). However, most research has focused on trees in open habitats or forest canopies. Less is known about understory wind-pollinated trees, such as red mulberry. Experimental removal of sources of heterospecific pollen at different spatial scales will be a valuable method for measuring the impact of local sources of heterospecific pollen in other species.

In unmanipulated plots, 77% of red mulberry seeds were sired by white and hybrid mulberry. This rate is lower than the combined frequency of white and hybrid individuals in the vicinity of red mulberry trees (94%) and the proportion of nonred taxa in the pollen pool (92%). At the same time, 77% hybrid seed is higher than the frequency of hybrids observed in adult populations (58%; Burgess et al., 2005). The discrepancies between the frequency of hybrids in seed and that in the pollen and adult stages are consistent with moderate selection against heterospecific pollen and hybrid offspring during pollination (Honig et al., 1992) and seedling establishment, respectively. Selection against hybrid offspring in the field has been widely reported for other species (Day & Schluter, 1995; Arnold et al., 1999, but see Arnold & Hodges, 1995; Arnold et al., 2001; Tiffin et al., 2001). In a transplant experiment with mulberry seedlings, we found evidence of hybrid breakdown in hybrids produced by red mulberry (but not in hybrids from white mulberry mothers; Burgess & Husband, 2004, 2006). However, no corroborative data are available on the competitive ability of heterospecific mulberry pollen.

The fertility cost experienced by red mulberry as a result of hybridization is attributable mostly to mating with hybrids (59% of red mulberry seed), rather than with white mulberry (18% of seed). This pattern is likely related to the high frequency of hybrids in experimental plots, although we cannot exclude the possibility that hybrid pollen has a superior siring ability. Regardless, the result highlights the potential role of hybrids in extirpation of native taxa. Levin et al. (1996) argued that demographic swamping could arise during mating when hybrid seeds are inviable, as long as ovules otherwise destined for conspecific mating are available to heterospecific fertilizations. However, the impact of swamping may be exaggerated further if hybrids can establish and mate, as they can further reduce the relative frequency of conspecific pollen in the pollen pool, and increase competition for ovules (Ellstrand, 1992; Buerkle et al., 2003). Hybrids may further affect the abundance of red mulberry during establishment, if they compete for establishment sites and resources (Huxel, 1999). In fact, it has been argued in previous studies that the competitive ability of rare taxa may be more important than the siring ability of hybrids to the persistence of rare taxa when mating is random and hybridizing taxa are in close proximity (Burke & Arnold, 2001; Wolf et al., 2001).

What are the consequences of interspecific seed discounting for hybridizing taxa such as red mulberry? Because interspecific seed discounting replaces conspecific offspring with heterospecific offspring, the frequencies of parental genotypes should decline, especially if population growth is limited by fecundity. Moreover, successful establishment and reproduction of heterospecific taxa will simply accelerate this trend as it increases the numerical disadvantage of the rare taxon. Interspecific seed discounting is complete in red mulberry and therefore will likely contribute to local extirpation. In addition, evidence from reciprocal transplant studies suggests that this decline may be reinforced by relatively limited establishment and growth by red mulberry seedlings (Burgess & Husband, 2006).

The expected decline of a rare taxon, as mediated by hybridization, may have significance for conservation if the goal is to preserve that species with its historical genetic composition (Ellstrand, 1992; Levin et al., 1996; Huxel, 1999). For red mulberry, interspecific seed discounting poses a serious concern, as conservation goals for this species are generally not accepting of introgressed genotypes (Ambrose & Kirk, 2004). At the level of the gene, however, hybridization may have some conservation benefits. By producing hybrid genotypes with relatively high fitness, hybridization may help to sustain some red mulberry genes that are introgressed in the genetic background of white mulberry. Ultimately, the priorities for conserving biological diversity will determine the significance of hybridization for genome and gene conservation.


We thank Royal Botanical Gardens, Niagara Peninsula Conservation Authority, Niagara Parks Commission, Rondeau Provincial Park, Point Pelee National Park and Fish Point Provincial Nature Reserve for access to field sites and field assistance. We also thank the Ontario Ministry of Natural Resources (A. Woodliffe, D. Kirk and M. Thomson) and the National Red Mulberry Recovery Team for assistance in setting up large-scale culling experiments at Point Pelee and Fish Point, and R. Beavers, A. Bauman, A. Crawford, C. Leduc and S. Spender for assistance with data collection. Financial assistance was provided by the Ontario Ministry of Natural Resources, The Endangered Species at Risk Fund (World Wildlife Fund Canada, Canadian Wildlife Service), Canadian Forest Service, and a Canada Research Chair and Natural Sciences and Engineering Research Council of Canada Grant to BCH.