| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Article |
Department of Biology, University of New Mexico, Albuquerque, New Mexico USA
Received for publication May 24, 2006. Accepted for publication December 13, 2006.
ABSTRACT
In previous studies of the weedy annual Raphanus sativus we have demonstrated that mating is nonrandom in greenhouse plants, suggesting that sexual selection is possible. To investigate how these greenhouse results might translate to conditions more similar to the field, we manipulated both pollen load size and the number of competing pollen donors on stigmas. While the smallest pollen loads (22 grains per stigma) were small enough to reduce fruit and seed set, seed siring success was unaffected by pollen load size. When the number of competing donors in a mixed pollination was increased to four, the proportion of seeds sired by the pollen donors was the furthest from expectation, suggesting that nonrandom mating increases as the number of donors per pollination increases. There was no significant interaction between pollen load size and number of competitors per pollination. Overall, mating remained nonrandom across all treatments. Thus differential seed paternity is likely to occur in the field as well as in the greenhouse.
Key Words: Brassicaceae • mate choice • nonrandom mating • pollen competition • pollen load size • Raphanus sativus • wild radish
The concept of sexual selection has fascinated evolutionary biologists since its introduction by Charles Darwin (1859
, 1871
). It is a powerful selective force, producing elaborate behaviors and morphologies in animals (e.g., Andersson, 1994
). Thus, the realization that sexual selection might also occur in plants has attracted considerable attention and controversy (e.g., Charnov, 1979
; Willson, 1979
, 1982
; Stephenson and Bertin, 1983
; Willson and Burley, 1983
; Charlesworth et al., 1987
; Grant, 1995
; Delph and Havens, 1998
; Skogsmyr and Lankinen, 2002b
) Part of this controversy is due to the overall complexity of plant mating systems.
Plants are able to sort among mates at a number of levels (Marshall and Folsom, 1991
). (1) Between self and nonself pollen. Because of the negative implications of inbreeding, it is not surprising that physiological self-incompatibility has received the most attention in the literature (e.g., deNettancourt, 1997
; Hinata and Nishio, 1980
; Richards, 1986
). (2) Among relatives and nonrelatives. Plants may also sort among mates in ways that ameliorate reductions in fitness due to mating with individuals that are too closely related (inbreeding depression) (Holsinger, 1991
; McCall et al., 1994
; Cruzan and Barrett, 1996
) or too distantly related (outbreeding depression; e.g., Waser and Price, 1991
). (3) Among compatible, unrelated mates. This level of sorting is of particular interest because it may generate sexual selection in plants by favoring certain pollen donors across a range of maternal plants (Marshall and Folsom, 1991
; Mitchell and Marshall, 1998
).
Two mechanisms have been proposed to explain nonrandom mating between compatible, unrelated plants: pollen competition and mate choice. When there is more pollen available than there are ovules to fertilize, pollen grains may compete for access to ovules (Hormaza and Herrero, 1992
, 1996
; Delph et al., 1998
; Delph and Havens, 1998
; Skogsmyr and Lankinen, 2002a
, b) and the female tissue may effectively choose among mates (Stephenson and Bertin, 1983
; Willson and Burley, 1983
; Waser et al., 1987
; Marshall and Ellstrand, 1988
; Marshall and Folsom, 1991
). For example, analyses of seed paternity (a measure of pollen donor fitness) in wild radish have demonstrated that, mating is nonrandom, and certain pollen donors sire more seeds than others across a wide range of maternal plants (e.g., Marshall, 1991
, 1998
; Marshall and Ellstrand, 1986
, 1988
).
The intensity of both pollen competition and mate choice may depend upon the amount of pollen available (Schlichting et al., 1990
; Bjorkman et al., 1995
; Johannsson and Stephenson, 1997
; Mitchell, 1997
; Neisenbaum, 1999; Marshall et al., 2000
; Skogsmyr and Lankinen, 2002a
, b). That is, when there is abundant pollen, the potential for both pollen competition and mate choice may be high. Alternatively, when pollen is limited, the potential for nonrandom mating is reduced. Because pollinator service may be variable in the field, the opportunity for nonrandom mating may vary. Thus, there has been considerable speculation about how frequently pollen load size in the field might be large enough to allow pollen competition and mate choice (Davis et al., 1987
; Snow, 1990
, 1991
; Mitchell, 1997
; Niesenbaum, 1999
; Marshall et al., 2000
).
Pollen loads in the field may also vary in their composition due to pollinator behavior. A pollinator may visit several plants, depositing and picking up pollen at each plant (Thomson and Plowright, 1980
; Craig, 1989
; Rademaker et al., 1997
; Talavera et al., 2001
). This will create variation in both the amount of pollen and the number of competing donors represented in each pollen load. Because pollen donors likely vary in their competitive ability, the exact composition of a pollen load may affect the opportunity of particular pollen donors to sire seeds. Therefore, it is important to not only consider how much pollen is present, but also to consider the number of competitors represented in a particular pollination event.
In previous studies of wild radish, when large amounts of pollen were deposited on stigmas of greenhouse or field plants, pollen donors sired unequal numbers of seeds (e.g., Marshall and Ellstrand, 1986
, 1988
; Marshall, 1991
, 1998
). However, these studies have generated concern because the number of pollen grains deposited per stigma may have been too large to adequately mimic field conditions. This has prompted us to attempt studies in which seed paternity was measured after deposition of pollen in a range of load sizes.
In our initial attempt, seed paternity was invariant across pollen load sizes from about 40 to about 200 pollen grains per stigma (Marshall et al., 2000
). However, because wild radish flowers have only about six ovules per flower, our smallest pollen loads may not have been small enough to affect mating. Another study suggested that small pollen load size could affect mating patterns, but in this case, maternal condition as well as pollen load size was altered (Shaner and Marshall, 2003
), and a very small total number of pollen donors was used. In addition, only two pollen donors were used per cross, which may not give a full indication of what happens in the field. Field flowers may receive pollen from up to four competitors during a single pollination event (Ellstrand, 1984
). This is an important consideration because some studies have shown that when the number of competitors is increased, offspring vigor (measured as seed size) and number of seeds per fruit increases (Marshall and Ellstrand, 1986
; Marshall, 1991
). However, the effects of multiple pollen donors have been less clear in other studies (Snow, 1990
; Sork and Schemske, 1992
).
To understand the potential for sexual selection to act on plants in the field, it is necessary to consider how both the amount of pollen and the number of competitors in a mix may affect nonrandom mating. For example, if multiple competing pollen donors are present in a single pollination event and the amount of pollen is large, the opportunity for pollen competition and mate choice may be maximized. However, if the number of competitors is high and the pollen load is quite small, random events during pollen deposition may cause the most competitive pollen to be underrepresented. In other words, there may be too few pollen grains from each donor for the most competitive donor to sire the bulk of the seeds. When considering a donor's competitive ability, it is also important to consider how that donor acts in competition with one additional donor or multiple additional donors because its competitive ability may change (Marshall, 1991
).
We approached the challenge of understanding the range of conditions under which nonrandom mating may occur by asking the following questions: (1) Does a reduction in pollen availability affect seed siring success? If so, mating should become more random when pollen is limited. (2) Does the number of potential competitors affect the success of target pollen donors? If so, then siring success might become more random when donors are limited in number (excluding single donor crosses). (3) Is there an interaction between the number of competitors and pollen availability such that the mating pattern is altered? If so, when pollen load sizes are quite small, a reduction in competition may allow inferior competitors to be overrepresented with regard to the total number of seeds sired due to chance variation in pollen load composition.
MATERIALS AND METHODS
The study organism
Raphanus sativus L. (wild radish; Brassicaceae) is an annual weed found in disturbed areas such as roadsides and abandoned fields in California. Wild radish germinates in late winter and produces flowers and fruits from early January until late spring or summer depending upon the yearly moisture. Wild radish is visited by a wide range of pollinators including honeybees, syrphid flies, and lepidopterans (Kay, 1976
; Stanton, 1987
; Stanton and Preston, 1988
). The variety and range of these pollinators are likely to produce variation in pollen load sizes and numbers of competitors in the field.
Wild radish has many characteristics that facilitated this study. The plant has abundant flowers, so many crosses are possible. The life cycle of the plant is relatively short such that mature fruit are available approximately 1 month after pollination. The fruits are multiseeded, so nonrandom mating is possible both within and among fruits. Analysis of mating patterns (seed paternity) is possible because there is sufficient isozyme variability (Ellstrand, 1984
). Finally, wild radish demonstrates sporophytic self-incompatibility, which eliminates the complication of self-fertilization (Hinata and Nishio, 1980
; Pundir et al., 1983
).
Due to the nature of pollinator visitation, it is likely that plants in the field receive mixtures of pollen from between two and four competitors (Ellstrand, 1984
; Marshall and Ellstrand, 1986
), and plants receive pollen loads that average 60 and 120 pollen grains per stigma from single and multiple pollinator visits, respectively. The range of pollen load sizes deposited in the field is from zero to over 300 (Young and Stanton, 1990
). Further, under field conditions, wild radish has been shown to demonstrate multiple paternity within a single fruit (Ellstrand, 1984
). In addition, paternity has been shown to be nonrandom in the field, in greenhouse studies, and in experimental gardens (Marshall and Ellstrand, 1988
; Devlin and Ellstrand, 1990
; Marshall, 1991
, 1998
; Marshall and Fuller, 1994
; Marshall et al., 2000
).
Plant selection
The seeds used in this study were collected from three small, natural populations at the Agricultural Operations Facility of the University of California at Riverside, Riverside, California, USA. We used seeds from all three populations to increase the probability of finding several plants with useful isozyme phenotypes. Fruits were collected from a total of 100 field plants, 30–35 per population. We haphazardly selected one fruit from each of these plants. Two seeds were planted from each of these fruits. We also planted seeds from four additional plants. These plants were from maternal lineages known to express a rare phenotype of phosphoglucose isomerase (PGI). These maternal lineages were developed from plants that originated from the same southern California locations as our other seeds. The seeds were planted in a 2 : 1 : 1 sand : peat moss : perlite mixture in 3.79-L pots. The plants were watered once daily until they were approximately 5 wk old and were then watered three times daily via an automated dripper system. Once the seedlings had their first true leaves, their PGI phenotype at a single locus was determined by starch gel electrophoresis. PGI is a dimeric enzyme with three common and one rare phenotype in most Southern California populations (Ellstrand, 1984
). Note that we refer to PGI phenotypes because isozymes are proteins, not DNA. However, the different phenotypes of an isozyme are commonly referred to by the alleles responsible for the phenotype. After electrophoresis, those seedlings that were heterozygous for PGI were immediately eliminated. Then, where two seedlings remained per pot, we thinned to one seedling per pot. Plants were fertilized with Peters 20-20-20 fertilizer (Scotts Co., Marysville, Ohio, USA) beginning with the emergence of their first true leaves at approximately 4 wk. All plants that served as pollen donors were then fertilized twice weekly with their respective fertilizer treatment. The maternal plants were also fertilized twice weekly for the first two replicates of pollinations. This amount was then increased to three times weekly for the duration of the experiment. In addition, plants received micronutrients (Minor L, Albuquerque Chemical, Albuquerque, New Mexico, USA) once per wk.
When plants began flowering, crosses were performed in order to select plants with a wide range of compatibility. To determine compatibility, we performed experimental pollinations by swiping the anther of a potential competitor across the stigmatic surface of a potential maternal plant. After 10 d, these pollinations were scored. If the fruit was setting seed, the plants were considered compatible. If the fruit had aborted, the cross was repeated. If this cross resulted in an aborted fruit a second time, the plants involved in that cross were considered incompatible. Nine plants were selected to act as the maternal plants. These plants were chosen based on compatibility with the competitors, homozygous PGI phenotype, health, and flower morphology. The flower morphology was especially important because flowers with elongated, large stigmas were easier to pollinate and did not require emasculation.
Eleven plants were selected to act as competitors. Of these, one subset consisted of eight "target" competitors that were homozygous for phenotype 2 of PGI. Three additional plants were chosen to act as "standard competitors" of the target competitors in all applicable pollinations. These plants each had a unique PGI phenotype (44, 33, and 11, respectively) (see Treatments). In addition to PGI phenotypes, plants were chosen based on compatibility with the maternal plants and health.
Treatments
One of the aims of this study was to investigate how mating patterns change when competitors interact with pollen from varying numbers of other competitors. To achieve this aim, we manipulated the composition of pollen loads. Pollen from each of the 12 target competitors was mixed with pollen from zero competitors, one standard competitor, two standard competitors, or three standard competitors. Each type of cross was repeated on all maternal plants. For example, if the pollination called for pollen from competitor 3, then each maternal plant received pollen from target competitor 3 alone, pollen from target competitor 3 + i, pollen from target competitor 3 + i + j, and finally pollen from target competitor 3 + i + j + k, where competitors i, j, and k were the standard competitors. In this way, each competitor had the opportunity to sire seeds in four different competitive regimes. Each type of cross was performed at all pollen load sizes (as described later in this section).
The plant designated to be the first, second, or third standard competitor was the same throughout the experiment with one exception. After the second replicate of pollinations was complete, one of the standard competitors died. For the third and final replicate of pollinations, this standard competitor, which was homozygous for the phenotype 3 of PGI and was used in three- and four-competitor crosses, was replaced with a full-sibling that shared the same PGI phenotype. There were no significant interaction effects of competitor number x replicate on good seeds per fruit (F1,337 = 0.89, P < 0.89), fruit fate (X2 = 4.33, P < 0.63, df = 6), or seed siring ability (F4,826 = 0.86, P < 0.49), so the data were not analyzed separately for the third replicate.
Another aim of this study was to assess whether mating patterns vary when pollen availability changes. To explore this question, we varied the amount of pollen applied to stigmas. The three pollen load treatments included a high, medium, and low treatment. All these pollination types were performed with 2-pound fishing line (Maxima America, Santa Ana, California, USA). Pollen was collected from the appropriate donors by collecting two flowers per donor and tapping those flowers 10 times in a circular pattern in a glass petri dish. The pollen was then further mixed using forceps. We applied the high pollen load, by dragging fishing line through the mixed pollen in the petri dish and then dragging the fishing line across the stigma of the pollen recipient. We applied the medium pollen load, by tapping the end of the fishing line twice in the mixed pollen in the petri dish and then tapping the fishing line twice on the recipient stigma. Finally, we applied the low pollen load by tapping the end of the fishing line once in the pollen mixture in the petri dish and then gently tapping the line once on the recipient stigma. We cleaned the fishing line with 90% ethanol after each pollination. The three load sizes were used in crosses with all the combinations of pollen described.
To ensure that the pollen loads delivered were consistent in size, test pollinations were performed at the end of each replicate. These test pollinations were performed in the manner described, and then the recipient stigma was removed and stained with Alexander's stain (Alexander, 1980
). The number of pollen grains on the stigmatic surface was then counted. Our pollination techniques produced pollen load sizes that were significantly different from one another (F2,33 = 71.85, P < 0.0001). The low pollen load size was 22.1 ± 6.0 pollen grains (mean ± SE), the medium pollen load size was 41.4 ± 5.4 pollen grains, and the high pollen load size was 105.94 ± 26.0 pollen grains.
Crossing design
One of the primary goals of this study was to vary features of pollen deposition in such a way that the greenhouse pollinations would mimic those that occur in the field. To achieve this aim, all the treatments described were performed simultaneously across all maternal plants and replicated three times (Fig. 1). These pollinations were performed in a random order within replicates, and the random order of crosses differed for each plant and each replicate. All pollinations for a replicate were completed for each maternal plant before the next replicate of pollinations was started. However, some of the maternal plants deteriorated to such an extent that not all pollinations were performed. Four maternal plants received pollinations from all three replicates (plants A, J, K, and R). Two full replicates of pollinations were performed on maternal plant M. Maternal plants D, H, and I did not receive two full replicates of pollinations. The duration of each replicate was no more than 2 months, and all three replicates were completed in 5 months. A total of 2028 crosses was performed.
|
Ten days after each pollination the fruits were checked and recorded as either aborted or filling. If the fruit had aborted, the pollination was redone and checked again. If the fruit aborted a second time, the cross was not repeated. Upon maturation, fruits were collected and opened. The seeds were counted and catalogued as either good seeds (fully filled, round, and mature) or aborted seeds (small, shriveled, and black). Finally, paternity was determined for each seed by starch gel electrophoresis (Ellstrand, 1984
). A total of 8592 seeds was genotyped.
Data analysis
Fruit set was compared among treatments with logistic regressions (PROC LOGISTIC, using the program SAS, version 8.2, 2002, SAS Institute, Cary, North Carolina, USA). We asked whether the frequencies of aborted and full fruits were different among maternal plants, target competitors, pollen load sizes, number of competitors, and their two-way interactions.
To ask how pollen load size and competition affected seed set, we asked whether the number of good seeds per fruit changed with pollen load size and the number of competitors. Number of good seeds per fruit was the dependent variable in an ANOVA where maternal plant, pollen load size, number of competitors, and their interactions were the independent variables. The individual target competitors were not used as independent variables because fruits were often multiply sired. We chose to use fixed effects ANOVAs because both the maternal plants and the competitors were specifically selected based upon a range of criteria (e.g., PGI phenotype, compatibility, health, and flower morphology). While a mixed model ANOVA with maternal plant, competitor, and their interactions might have allowed for broader interpretation, this was not appropriate due to our selection criteria. Results from fixed effects and mixed models were indistinguishable in all but one case.
Seed paternity was compared across target competitors at the fruit level. Because crosses were performed only one time for each fruit, data taken for each seed in a fruit might not be independent. Thus, we compared the proportion of seeds sired within a fruit by target competitors across the independent variables. This comparison was performed in an ANOVA (PROC GLM, SAS, version 8.2). The arcsine square root transformation of the proportion of seeds sired per fruit was the dependent variable, while maternal plant, target competitor, number of competitors present in a pollination, pollen load size, and all of their interactions were the independent variables. Additionally, because the amount of pollen produced by an individual competitor might affect siring ability, we used the mean pollen production per competitor in each replicate as a covariate in the model.
RESULTS
Fruit and seed production
Fruit set increased significantly as the number of competing pollen donors per pollination increased (Table 1, Fig. 2). Note that the number of competitors is 0 if only pollen from the target donor was used, 1 if pollen from the target donor was mixed with pollen from one standard competitor, etc. Fruit set also differed significantly among flowers pollinated with different pollen load sizes (Table 1, Fig. 2). The pattern was that higher pollen load size produced greater fruit set, except in the case with only one competitor. In that case, fruit set leveled off in the medium and low pollen load sizes. There was no significant interaction between pollen load size and number of competitors (Table 1). Fruit set also varied significantly among maternal plants (Table 1).
|
|
|
|
|
|
Seed paternity and number of competing pollen donors
If mating was random, we expected each target donor to sire 50% of the seeds per fruit in a one-competitor cross, 33% of the seeds per fruit in a two-competitor cross, and 25% of the seeds per fruit in a three-competitor cross. The proportion of seeds sired per fruit by target donors was 39%, 34%, and 17% for the one-, two-, and three-competitor crosses, respectively. Thus, the number of competitors had a significant effect on the proportion of seeds sired per fruit by target donors (Table 3, Fig. 5). Target donors usually sired fewer seeds than expected in the one-competitor and three-competitor crosses, and about as many seeds as expected in the two-competitor crosses (Table 4, Fig. 5).
|
|
Interaction effects of pollen load size and number of competitors
The interaction of pollen load size and number of competitors did not have a significant effect on fruit set (Table 1), the number of good seeds per fruit (Table 2), or the proportion of seeds sired per fruit by target pollen donors (Table 3).
DISCUSSION
Plants are able to sort among mates at various levels including against self pollen, among compatible mates, and against pollen from different species (deNettancourt, 1997
; Hinata and Nishio, 1980
; Richards, 1986
; Holsinger, 1991
; Marshall and Folsom, 1991
; Waser and Price, 1991
; McCall et al., 1994
; Cruzan and Barrett, 1996
). This results in nonrandom mating, which may have consequences for maternal plant fitness, offspring fitness, and population genetic structure. Here, we were interested in the ability of plants to sort among compatible mates. This level of sorting is important, because it creates the potential for sexual selection (Marshall and Folsom, 1991
; Mitchell and Marshall, 1998
; Skogsmyr and Lankinen, 2002a
, b). However, studies that examine the outcome of mating at this level have been subject to controversy. We addressed one problem, the translation of results from greenhouse studies to field conditions. Plants in the field likely experience variation in size and composition of pollen loads (Snow, 1994
; Skogsmyr and Lankinen, 2002a
, b). We considered whether these two factors affected mating patterns of wild radish.
Pollen load size
Initially, we considered the effects of variation in the amount of pollen deposited on a receptive stigma. The statistically distinct pollen load sizes used in this study are similar to those used in previous studies of wild radish (Shaner and Marshall, 2003
; Marshall et al., unpublished manuscript), and they proved to be biologically relevant in two ways. First, fruit set declined in the low pollen load size. Perhaps the small pollen load sizes produced fewer and/or poorer quality of fertilized seeds, leading to selective abortion of fruits. While fertilization and abortion were not measured in this study, this interpretation is consistent with previous studies of wild radish (Marshall et al., 2000
; Shaner and Marshall, 2003
), Lotus corniculatus (Stephenson and Winsor, 1986
; Stephenson et al., 1988
), Thalictrum pubescence (Davis, 2004
), and Mirabilis jalapa (Niesenbaum, 1999
). Second, the number of good seeds per fruit was reduced in the low pollen load size, suggesting that not all ovules were fertilized or embryos of poor quality were aborted in the smaller pollen load sizes (Tables 1 and 2). Therefore, it is likely that the smaller pollen load sizes generated reductions in pollen competition.
We predicted that as pollen load size decreased, mating would become more random. However, seed paternity was very similar across pollen load sizes (Table 3, Fig. 4). This is consistent with some previous studies of wild radish (Marshall et al., 2000
; Marshall et al., unpublished manuscript). However, in another study with wild radish, where some maternal plants were exposed to water stress, mating approached random in low pollen load sizes (Shaner and Marshall, 2003
).
Although pollen load size did not have a significant effect on the proportion of seeds sired in a fruit, there were significant differences in seed siring ability across target donors (Table 3, Fig. 4), and some of these differences were quite large. For example, target donor number 6 sired 12% of the seeds when its pollen was mixed with standard competitors, while target donor number 2 sired 39% of the seeds in similar crosses. These kinds of differences in seed siring ability are consistent with previous studies of wild radish (Marshall and Ellstrand, 1986
; Marshall, 1991
, 1998
) and Cucurbita pepo (Stephenson et al., 1988
; Richardson and Stephenson, 1992
).
Number of competitors
We next asked how the number of competitors in a pollination affected mating patterns. We predicted that as the number of competitors decreased, so would the opportunity for competition and mate choice and that this would in turn lead to more random seed paternity. In one-competitor crosses, the target donors sired 78% as many seeds as expected, while in the three-competitor crosses, the target donors sired 68% as many seeds as expected (Fig. 5). Thus, the deviation from expectation was proportionally larger in three-competitor crosses than in one-competitor crosses. This is consistent with our initial prediction. However, the average paternity of target donors in two-competitor crosses was very near to the predicted value.
Part of the difference among one-, two- and three- competitor crosses may be explained by the ability of each of the standard competitors to sire seeds. The first standard competitor (used in one- to three-competitor crosses) sired 73% of seeds in one-competitor crosses (the expectation was 50%; 
2 = 209, df = 1, P < 0.0001), the second standard competitor (used in two- and three-competitor crosses) sired 24% of seeds in two-competitor crosses (the expectation was 33%; 2 = 35.2, df = 1, P < 0.001) and the third standard competitor (used only in three-competitor crosses) sired 58% of seeds in three-competitor crosses (the expectation was 25%; 2 = 438, df = 1, P < 0.0001). In other words, the second standard competitor was a relatively weak competitor, while the first and third standard competitors were strong competitors.
The rate of fruit set increased significantly as the number of competitors increased. This suggests that there was less abortion of fruit when there was greater opportunity for pollen competition and mate choice (Table 1, Fig. 2).
Pollen load size x number of competitors interaction
Because it is likely that the number of competitors and the pollen load size vary simultaneously in the field, we next asked whether these variables had an interaction effect on the mating pattern. We predicted that when the number of competitors is high but the pollen load size is small, mating could approach random, but we found no significant interaction effect on seed siring success (Table 3).
Conclusions
This study had some important limitations. First, several plants in this study deteriorated over time because the crosses were conducted over a period of 5 mo. This may affect pollen quality of target donors and the patterns of fruit abortion, the number of good seeds per fruit, and mating patterns on maternal plants (Marshall and Ellstrand, 1988
; Marshall and Diggle, 2001
). This would increase error variance in our statistical models and reduce the chance of observing significant results. Second, because of this deterioration, the crossing design was not completed as planned. This reduced our sample size for scoring seed paternity and thus, our power to detect significant effects. However, the effects of pollen load size on seed paternity never approached significance, so we doubt that there was a strong effect of reduced sample size on our conclusions. Third, for plants in the field, the additional competitors in mixed pollinations would not be a single set of standard competitors. The potential combinations of competitors are much more numerous. However, it is important to note the similarity of these results to those of a previous study of wild radish that used many more combinations of competitors (Marshall, 1991
). Finally, several studies of pollen load size investigated offspring fitness as a consequence of pollen competition, which was not done in this study (Davis et al., 1987
; Marshall and Whittaker, 1989
; Bjorkman et al., 1995; Quesada et al., 1996
; Johannsson and Stephenson, 1997
; Mitchell, 1997
; Neisenbaum, 1999; Davis, 2004
). This is an important distinction, because the ability of a pollen donor to sire more seeds per fruit may or may not correlate with offspring vigor. We are currently investigating the relationship between pollen load size and offspring growth for a different set of wild radish plants (Marshall and Shaner, unpublished manuscript).
These data, as well as others for wild radish suggest that pollen donors and maternal plants have different opportunities to affect the outcome of mating. Maternal plants have the opportunity to act from pollen receipt through fertilization. For example, the maternal plant may act early by influencing the outcome of pollen germination, pollen tube growth, and fertilization of ovules. Later selective actions by maternal plants may include abortion of seeds or fruits that have been sired by poor donors (e.g., Marshall and Folsom, 1991
). In this experiment both fruit set and seeds per fruit were affected by pollen load size, suggesting the action of maternal choice (Tables 1, 2, Figs. 2, 3). Because wild radish plants produce many flowers and cannot fill fruit at all of them, the cost of aborting some seeds and fruits may be quite small (Stephenson and Winsor, 1986
). It may be that abortion of seeds and fruits after pollination with very small pollen loads was selective. If so, that could produce similar seed paternity across the pollen load sizes. It is intriguing that the one time when we found a detectable change in seed paternity with small pollen load size was in a study where maternal plants were under considerable stress (Shaner and Marshall, 2003
).
Microgametophytes, on the other hand, have no additional opportunities to mate once they arrive on stigmas. In other words, the pollen grains cannot move to a more favorable plant or to a stigma where competition might be reduced. The environment available to pollen grains might vary depending on the number and type of other competitive pollen grains present and the condition of the maternal plant. This is important, because when maternal plants are in poor condition, they appear less able to influence mating (Marshall and Diggle, 2001
; Shaner and Marshall, 2003
). Some pollen donors may be able to take advantage of the variation in pollen load size and competitive regime. For example, donor 9 sired significantly more seeds in the low pollen load treatments (Fig. 4). In other experiments one of every six to 10 competitors has a unique response to changes in competitive circumstances (Marshall, 1991
; Shaner and Marshall, 2003
; Marshall et al., unpublished manuscript). If the qualities of pollen that allow for these strategies are different from those that allow success under strongly selective conditions there may be a variety of mating niches. This variety may allow variation in the ability of pollen donors to persist in natural populations.
Clearly, it is difficult to replicate all the diversity of field conditions in a greenhouse study. However, by combining some of the key factors that might influence both the number and the quality of seeds sired, we can further understand the conditions that allow nonrandom mating. Overall, despite variations in pollen load size, number of competitors, and plant condition, seed paternity was still nonrandom. This suggests that the potential for sexual selection to act in wild radish persists across a wide range of conditions in the field.
FOOTNOTES
1 The authors thank J. J. Avritt for help in the greenhouse; A. Harbison, J. Salaz, and S. Martinez for work both in the greenhouse and in the lab; E. Leve for help in the lab; J. Reynolds, E. L. Charnov, W. T. Pockman, and K. Johnson for invaluable discussions and helpful comments. This work was supported by NSF Grant DEB 9981796. 
2 Author for correspondence (marieken{at}unm.edu
)
LITERATURE CITED
Alexander M. P.. 1980. A versatile stain for pollen, fungi, yeast, and bacteria. Stain Technology 55: 13-18.[ISI][Medline]
Andersson M.. 1994. Sexual selection Princeton University Press, Princeton, New Jersey, USA.
Bjorkman T. Samimy C. Pearson K.. 1995. Variation in pollen performance among plants of Fagopyrum esculentum. Euphytica 82: 235-240.[CrossRef][ISI]
Charlesworth D. Schemski D. W. Sork V. L.. 1987. The evolution of plant reproductive characters: sexual vs. natural selection. In S. C. Stearns [ed.] Evolution of sex 317-355 Birkhauser, Basel, Switzerland.
Charnov E. L.. 1979. Simultaneous hermaphroditism and sexual selection. Proceedings of the National Academy of Sciences, USA 76: 2480-2484.
Craig J. L.. 1989. Seed set in Phormium: interactive effects of pollinator behavior, pollen carryover, and pollen source. Oecologia 81: 1-5.[CrossRef][ISI]
Cruzan M. B. Barrett S. C. H.. 1996. Postpollination mechanisms influencing mating patterns and fecundity: an example from Eichhornia paniculata. American Naturalist 147: 576-598.[CrossRef][ISI]
Darwin C.. 1859. On the origin of species John Murray, London, England.
Darwin C.. 1871. The descent of man, and selection in relation to sex John Murray, London, England.
Davis L. E.. 2004. Natural levels of pollination intensity and effects of pollen loads on offspring quality in females of Thalictrum pubescens (Ranunculaceae). Plant Systematics and Evolution 244: 45-54.[CrossRef][ISI]
Davis L. E. Stephenson A. G. Winsor J. A.. 1987. Pollen competition improves performance and reproductive output of the common zucchini squash under field conditions. Journal of the American Society for the Advancement of Horticultural Science 112: 712-716.
Delph L. F. Havens K.. 1998. Pollen competition in flowering plants. In T. R. Birkhead and A. P. Moller [eds.] Sperm competition and sexual selection 55-90 Academic Press, San Diego, California, USA.
Delph L. F. Weining C. Sullivan K.. 1998. Why fast growing pollen tubes give rise to vigorous progeny: the test of a new mechanism. Proceedings of the Royal Society of London, B, Biological Sciences 265: 935-939.[CrossRef]
deNettancourt D.. 1997. Incompatibility in angiosperms. Sexual Plant Reproduction 10: 185-199.[CrossRef][ISI]
Devlin B. Ellstrand N. C.. 1990. Male and female fertility variation in wild radish, a hermaphrodite. American Naturalist 136: 87-107.[CrossRef][ISI]
Ellstrand N. C.. 1984. Multiple paternity within the fruits of wild radish Raphanus sativus. American Naturalist 123: 819-828.[CrossRef][ISI]
Grant V.. 1995. Sexual selection in plants: pros and cons. Proceedings of the National Academy of Sciences, USA 92: 1247-1250.
Hinata K. Nishio T.. 1980. Self-incompatibility in crucifers. In S. Tsunoda, K. Hinata, and C. Gomez-Campo [eds.] Brassica crops and wild allies: biology and breeding 223-234 Japanese Scientific Societies, Tokyo, Japan.
Holsinger K. E.. 1991. Inbreeding depression and the evolution of plant mating systems. Trends in Ecology and Evolution 6: 307-308.[CrossRef]
Hormaza J. I. Herrero M.. 1992. Pollen selection. Theoretical and Applied Genetics 83: 663-672.[ISI]
Hormaza J. I. Herrero M.. 1996. Dynamics of pollen tube growth under different competitive regimes. Sexual Plant Reproduction 10: 236-240.[CrossRef][ISI]
Johansson M. H. Stephenson A. G.. 1997. Effects of pollination intensity on the vigor of the sporophytic and gametophytic generation of Cucurbita texana. Sexual Plant Reproduction 10: 236-240.[CrossRef][ISI]
Kay Q. O. N.. 1976. Preferential pollination of yellow-flowered morphs of Raphanus raphanistrum by Pieris and Eristalis spp. Nature 261: 230-232.[CrossRef]
McCall C. Waller D. M. Mitchell-Olds T.. 1994. Effects of serial inbreeding of fitness components in Impatiens capensis. Evolution 48: 818-827.[CrossRef][ISI]
Marshall D. L.. 1991. Nonrandom mating in wild radish: variation in pollen donor success and effects of multiple paternity among one- to six-donor pollinations. American Journal of Botany 78: 1404-1418.[CrossRef][ISI]
Marshall D. L.. 1998. Pollen donor performance can be consistent across maternal plants in wild radish (Raphanus sativus, Brassicaceae): a necessary condition for the action of sexual selection. American Journal of Botany 85: 1389-1397.
Marshall D. L. Avritt J. J. Shaner M. Saunders R. L.. 2000. Effects of pollen load size and composition on pollen donor performance in wild radish, Raphanus sativus (Brassicaceae). American Journal of Botany 87: 1619-1627.
Marshall D. L. Diggle P. K.. 2001. Mechanisms of differential pollen donor performance in wild radish, Raphanus sativus (Brassicaceae). American Journal of Botany 88: 252-257.
Marshall D. L. Ellstrand N. C.. 1986. Sexual selection in Raphanus sativus: experimental data on the nonrandom fertilization, maternal choice, and consequences of multiple paternity. American Naturalist 127: 446-461.[CrossRef][ISI]
Marshall D. L. Ellstrand N. C.. 1988. Effective mate choice in wild radish: evidence for selective seed abortion and its mechanism. American Naturalist 131: 736-759.
Marshall D. L. Folsom M. W.. 1991. Mate choice in plants: an anatomical to population perspective. Annual Review of Ecology and Systematics 22: 37-63.[Medline]
Marshall D. L. Fuller O. S.. 1994. Does nonrandom mating among wild radish plants occur in the field as well as in the greenhouse?. American Journal of Botany 81: 439-445.[CrossRef][ISI]
Marshall D. L. Whittaker K. L.. 1989. Effects of pollen donor identity on offspring quality in wild radish Raphanus sativus. American Journal of Botany 76: 1081-1088.[CrossRef][ISI]
Mitchell R. J.. 1997. Effects of pollination intensity on Lesquerella fendleri seed set variation among plants. Oecologia 190: 382-388.
Mitchell R. J. Marshall D. L.. 1998. Nonrandom mating and sexual selection in a desert mustard: an experimental approach. American Journal of Botany 85: 48-55.[Abstract]
Niesenbaum R. A.. 1999. The effects of pollen load size and donor diversity on pollen performance, selective abortion, and progeny vigor in Mirabilis jalapa (Nyctaginaceae). American Journal of Botany 86: 261-268.
Pundir N. S. Abbas R. F. Al-Attar A. A. A.. 1983. Pollen germination and pollen tube growth in Raphanus sativus L. following self- and cross-pollinations. Phyton 43: 127-130.[ISI]
Quesada M. Winsor J. A. Stephenson A. G.. 1996. Effects of pollen competition on the reproductive performance in cucurbit hybrids (Cucurbitaceae): F1 and backcross generations. Canadian Journal of Botany 74: 1113-1118.[ISI]
Rademaker M. C. J. DeJong T. J. Klinkhammer P. G. L.. 1997. Pollen dynamics of bumble-bee visitation on Echium vulgare. Functional Ecology 11: 554-563.[CrossRef][ISI]
Richards A. J.. 1986. Plant breeding systems George Allen and Unwin, London, UK.
Richardson T. E. Stephenson A. G.. 1992. Effects of parentage, prior fruit set and pollen load on fruit and seed production in Campanula americana L. Evolution 46: 1731-1739.[CrossRef][ISI]
Schlichting C. D. Smalli L. Stephenson A. G. Winsor J. A.. 1990. Pollen loads and progeny vigor: the next generation. Evolution 44: 1358-1372.[CrossRef][ISI]
Shaner M. G. M. Marshall D. L.. 2003. Under how wide a set of conditions will nonrandom mating occur in wild radish (Raphanus sativus) L.?. American Journal of Botany 90: 1604-1611.
Skogsmyr I. Lankinen A.. 2002a. Pollen competitive ability: the effect of proportion in two-donor crosses. Evolutionary Ecology Research 4: 687-700.[ISI]
Skogsmyr I. Lankinen A.. 2002b. Sexual selection: an evolutionary force in plants?. Biological Reviews 77: 537-562.[Medline]
Snow A. A.. 1990. Effects of pollen-load size and number of donors on sporophyte fitness in wild radish (Raphanus raphanistrum). American Naturalist 136: 742-758.[CrossRef][ISI]
Snow A. A.. 1991. Effects of pollen-load size on sporophyte competitive ability in two Epilobium species. American Midland Naturalist 125: 348-355.[CrossRef][ISI]
Snow A. A.. 1994. Postpollination selection and male fitness in plants. American Naturalist 144: 69-83.[CrossRef]
Sork V. L. Schemske D. W.. 1992. Fitness consequences of mixed-donor pollen loads in the annual legume Chamaecrista fasciculate. American Journal of Botany 79: 508-515.[CrossRef][ISI]
Stanton M. L.. 1987. Reproductive biology of petal color variants in wild populations of Raphanus sativus L. I. Pollinator response to color morphs. American Journal of Botany 74: 178-187.[CrossRef][ISI]
Stanton M. L. Preston R. E.. 1988. Ecological correlates of petal size variation in wild radish, Raphanus sativus (Brassicaceae). American Journal of Botany 75: 528-539.[CrossRef][ISI]
Stephenson A. G. Bertin R. J.. 1983. Male competition, female choice, and sexual selection in plants. In L. Real [ed.] Pollination biology 109-149 Academic Press, Orlando, Florida, USA.
Stephenson A. G. Johnson R. S. Winsor J. A.. 1988. Effects of competition on growth of Lotus corniculatus L. seedlings produced by random and natural patterns of fruit abortion. American Midland Naturalist 120: 102-107.[CrossRef][ISI]
Stephenson A. G. Winsor J. A.. 1986. Lotus corniculatus regulates offspring quality through selective fruit abortion. Evolution 40: 453-458.[CrossRef][ISI]
Talavera S. Bastida F. Ortiz P. L. Arista M.. 2001. Pollinator attendance and reproductive success in Cistus libanotis L. (Cistaceae). International Journal of Plant Sciences 162: 343-352.[CrossRef]
Thompson J. D. Plowright R. C.. 1980. Pollen carryover, nectar rewards, and pollinator behavior with special reference to Diervilla lonicera. Oecologia 46: 68-74.[CrossRef][ISI]
Waser N. M. Price M. V.. 1991. Outcrossing distance effects in Delphinium nelsonii pollen loads, pollen tubes, and seed set. Ecology 72: 171-179.[CrossRef][ISI]
Waser N. M. Price M. V. Montalvo A. M. Gray R. N.. 1987. Female mate choice in a perennial herbaceous wildflower, Delphinium nelsonii. Evolutionary Trends in Plants 1: 29-33.
Willson M. F.. 1979. Sexual selection in plants. American Naturalist 113: 777-790.[CrossRef][ISI]
Willson M. F.. 1982. Sexual selection and dicliny in angiosperms. American Naturalist 119: 579-583.[CrossRef][ISI]
Willson M. F. Burley N.. 1983. Mate choice in plants Princeton University Press, Princeton, New Jersey, USA.
Young H. J. Stanton M. L.. 1990. Influences of floral variation on pollen removal and seed production in wild radish. Ecology 71: 536-547.[CrossRef][ISI]
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK |
