4
Selection on pollen donors
by Echium vulgare (Boraginaceae).
Summary
Less than only one quarter of the ovules of Echium vulgare in the field develop into viable seeds, even in the absence of pollen limitation. The loss of ovules could enhance the fitness of the maternal parent, if the less fit embryos are selectively aborted. Two pollination experi- ments were performed to examine the selectivity of maternal parents on self-pollen and different cross-pollen sources.
Pollinated with one pollen genotype per flower, self-pollen was, on average, equally successful in siring seeds as cross-pollen. However, the relative success of self-pollen com- pared to outcross-pollen differed significantly among the maternal parents. These results sug- gest that under certain conditions, selfing can be more advantageous for the number of seeds produced than cross-pollinations. Pollen donors differed significantly in outcrossing success.
The plants that were more successful in selfing were also more successful as pollen donor in outcross-pollinations. A significant interaction between maternal parent and paternal genotype was absent.
Pollinations with a pollen mixture produced selfed and outcrossed seeds in the same ratios as in the single-donor experiment. Overall, only slight differences were found between the single- and mixed-donor experiment.
Pollen tube growth does not show a significant correlation with the success of the pa- rental genotypes in the mixed-donor experiment, indicating that pollen tube growth is not the determining factor controlling the paternity of the seeds. These results are discussed with reference to possible mediating mechanisms.
C. Melser, M.C.J. Rademaker & P.G.L. Klinkhamer 1997. Sexual Plant Reproduction 10:305-312
Introduction
Many species produce far more ovules than seems necessary for the production of their seeds.
In many cases this low seed to ovule ratio is not caused by a lack of compatible pollen (Willson & Burley 1983), although Burd (1994) has argued that effects of pollen limitation have been underestimated. For Echium vulgare, our model plant, the average seed to ovule ratio is only 0.23, although seed set is not pollen limited in natural populations (Klinkhamer et al. 1994). These results imply that a large fraction of the fertilized ovules abort in an early stage. This abortion could serve to increase the female fitness if the fittest offspring are selected to ripen to viable seeds. If this selection is under female control, the problem for the maternal parent is how to determine the offspring quality. Stearns (1992) suggested that em- bryos are in a competitive arena for resources. With such a selective mechanism it is con- ceivable that there is a positive correlation between embryo selection and potential offspring quality later in life. Differences in offspring quality may arise in populations with a relative high and variable inbreeding depression. In such populations, differences are most pronoun- ced among selfed progeny.
Emphasis in literature has been on different mechanisms to prevent self-pollination.
Pre-pollination mechanisms include spatial or temporal separation of sexes. Post-pollination mechanisms to avoid selfing include self-incompatibility (de Nettancourt 1977). With these mechanisms, there is no possibility to discriminate among individuals of the selfed progeny.
Selection among selfed embryos has the advantage that through offspring with a low genetic load, two copies of the genome are passed to the next generation while large investments of resources in offspring with high genetic load can be avoided. Selection against selfing may also lead to cryptic self-incompatibility (Bateman 1956). With cryptic self-incompatibility the self-pollen is able to sire seeds if no cross-pollen is available, but in the presence of cross- pollen, the self-pollen is consistently outperformed and the majority of the seeds will be sired by the cross-pollen. In this way some advantages of both self- and cross-pollination can be combined.
Selection of pollen sources will lead to a deviation in paternity percentages of the seeds from the paternity percentages of the pollen that is originally applied. Such deviations in seed numbers from different pollen sources have been established for a number of species (Raphanus raphanistrum in Mazer et al. 1986, Ellstrand & Marshall 1986, Marshall &
Ellstrand 1988, Marshall 1991, Marshall & Folsom 1992, Campsis radicans in Bertin 1982, 1985 and 1988, Chamaecrista fasciculata in Fenster 1991). However, the ovules of these species are not equally arranged. Discrimination among pollen sources could be influenced by position effects if pollen sources are not randomly distributed among ovule positions. For species with seeds which are not linearly arranged, the impact of the pollen source effect has not yet been estimated. Other studies (Stephenson & Winsor 1986, Casper 1984, 1988) report that selective seed abortion increases offspring quality, without relating this to paternal genotype.
Here we describe a study on the number and the weight of the seeds, selectively sired by different pollen sources of Echium vulgare (L), a monocarpic, self-compatible perennial of calcareous grasslands and dune vegetations. Specifically, we investigated the impact of self- pollination and different cross-pollinations on seed numbers per flower and seed weight. The
experiments addressed the following questions: is there a difference in the number and the weight of the seeds sired by different pollen sources, (1) between self-pollinations and cross- pollinations or (2) among different cross-pollinations, and (3) is there a relation between the success of the self-pollinations and the success as a pollen donor in cross-pollinations. If there appears to be a difference in the number of seeds sired by different pollen sources, another question can be addressed: (4) Can these differences in number of seeds sired by different pollen donors be explained by differences in number of pollen tubes or early pollen tube growth or are they more likely to result from competition among embryos?
Material and methods
Plant species
Echium vulgare (L) is a rosette-forming monocarpic perennial. From the main flowering stem, cymes diverge at which flowers develop sequentially in time (Nicholls 1987). Each day new flowers open at each cyme. Flowers are hermaphrodite with five anthers and four ovules.
The four ovules are arranged in a square. Flowers are protandrous: first the anthers present the ripe pollen in the male phase. Hereafter the style elongates and the two lobes of the stigma diverge and become receptive to pollination. Although protandry and herkogamy reduce self- pollination within one flower, selfing by geitonogamy can still occur because flowers in the male and female phase are present on one plant simultaneously. In our study area male- steriles, individuals producing yellow pollen, occur at a frequency of 7% (Klinkhamer et al.
1994). Plants (presumably different genotypes) were collected in natural populations of E.
vulgare in the dunes of Meyendel, near The Hague, the Netherlands. All individuals were collected more than 100 m. apart. The different plants were propagated vegetatively in growth chambers and the replicates were used in the experiments. Day and night temperature were respectively 20 and 15˚C ± 1˚C. and relative humidity ranged between 60 and 85%.
The average seed yield per flower per plant of E. vulgare in our study area does not exceed 1.5 seeds per flower, although four ovules are present. Pollen limitation is not the determining factor for the low seed set per flower because additional hand pollination in the field did not increase the seed set per flower (Klinkhamer et al. 1994). Moreover, more than 4 pollen tubes were present in each style in the field with a modal amount of 9 (n=30, unpubl.
data).
Single-donor experiment
To examine the effect of selection on different pollen sources after pollination, we pollinated flowers with pure pollen of different pollen donors in a growth chamber and counted the resulting number of seeds per flower.
Ten different genotypes of E. vulgare were used in a complete diallel design. One plant of each genotype received different pollinations with pure pollen from each genotype, at 19 to 22 flowers, in total circa 200 treated flowers per maternal genotype. Flowers were emas- culated with forceps before the style elongated. Emasculation eliminates the possible effects of variation in degree of herkogamy between genotypes on the selfing rate. Pollen of at least two replicates of each genotype was used for the pollinations. The pollen was applied within three to four hours after collection. Flowers were pollinated by rubbing the pollen firmly on to
the lobes of the stigma with the end of a toothpick, which was covered with parafilm. With this method of hand pollination, 90% of the flowers received at least 5 pollen grains on the stigma (counted under the light microscope, n=175, unpubl. data). The treated flowers were marked with a small drop of paint to identify the applied pollen donor. All plants were circulated within the growth chamber to avoid environmental effects. All pollinations were carried out within a period of 30 days. After pollination of the last flowers used in the experiment, the later opening flowers were pollinated with a random sample of pollen for at least three days, so that the seeds of the last flowers in the experiment also developed in the presence of younger seeds. Approximately three weeks after the last pollinations, the numbers of developed seeds per flower were counted and the seeds were weighed.
The number of seeds per flower was tested for effects of the position of the cyme along the main flowering stem, the position of the flower along the cyme, the date of polli- nation, the maternal genotype, the pollination type (either self- or cross-pollination), the inter- action between maternal genotype and pollination type, the cross-pollen source and the inter- action between maternal genotype and cross-pollen source. The analysis of the binomial distributed data of the seed numbers used a GLM procedure with a logit link function (McCullagh & Nelder 1989, SAS Institute 1993). The effects of later entered factors in the analysis are adjusted for the effects of the earlier entered factors (SAS PROC GENMOD, type I). For significant factors, genotypes are compared with Miller's multi-stage procedure (Haccou & Meelis 1992). The position of the cyme along the main flowering stem appeared to have no effect and was excluded from the final analysis. The weight of the seeds was normally distributed and tested for effects of the same factors with a GLM procedure.
Maternal success (Mi) was calculated as
10
10 ,
1 ,
,
= j= i ij j i
i
F S M
in which Si,j is the total number of seeds produced by maternal genotype i and sired by pater- nal genotype j, Fi,j is the number of flowers, pollinated on maternal genotype i with pollen of genotype j. Paternal success (Pj) was calculated as
9
10 1
, ,
,
= ji= i j j i
j
F S
P for i≠j
in which Si,j is the total number of seeds produced by maternal genotype i and sired by pater- nal genotype j. Fi,j is the number of flowers, pollinated on maternal genotypes i with pollen of genotype j. Inbreeding depression (δ) was calculated as 1-Mi,i/Mi,j, in which Mi,i is the mean number of seeds per flower after selfing produced by maternal genotype i, and Mi,j (for i≠j) is the average number of seeds per flower produced after outcross-pollination by maternal geno- type i, averaged over all 9 outcross paternal genotypes j. A negative value of δ signifies a lar- ger production of seeds after self-pollinations compared to cross-pollinations. Dominance (h), or the degree to which deleterious alleles are expressed in the heterozygote, was calculated as the coefficient from the regression of Mi,j on Mi,i+Mj,j (Johnston & Schoen 1995). A value of 0.5 indicates complete additivity, a value of 0 indicates complete dominance, and values be-
tween 0 and 0.5 represent partial dominance of the alleles with the highest fitness. Irrespective of the underlying mechanisms, each value less than 0.5 signifies inbreeding depression.
Mixed-donor experiment
To mimic the natural field situation, in which at least three different pollen sources are present on the stigma (Rademaker, unpublished data), we did a second pollination experiment. In this experiment we pollinated flowers in a complete diallel design with a mixture of pollen of three different pollen genotypes.
Three different clones out of the ten which were used in the single-donor experiment were selected on the basis of the availability of flowering individuals at the start of both expe- riments. Three different replicates of each genotype were used for the pollinations. The flowers used for the pollinations were all at the middle of a cyme. Each paternal genotype contributed five anthers of one flower to the pollen mixture that was applied. For each polli- nation a new mixture was made. For the different genotypes, 4 to 28 flowers were used for counting the numbers of pollen grains per flower. After staining the nuclei of the pollen with DAPI (de Laat et al. 1987), pollen were counted with a flow cytometer (CA-II, Partec). With this method, the viable pollen grains are counted. These counts were used to estimate the ratio at which pollen of the three genotypes contributed to the pollen mixture. This ratio is equal to the expected ratio at which these three genotypes sire seeds in the offspring if no selection takes place. The expected number of seeds produced by mother i, sired by father j equals:
i j
j
i S
p p p S p
3 2 1
, = + +
ε
in which Si is the total number of seeds produced by maternal parent i and p1, p2 and p3 denote the number of pollen grains per anther of the three genotypes. The number of seeds sired per paternal genotype was tested for deviations from the expected number of seeds with a G-test (Sokal and Rohlf 1995).
Approximately 3 weeks after the last pollinations, the ripe seeds were collected and sown in a growth chamber. The paternities of 28 to 30 offspring per maternal genotype were analyzed with the use of Random Amplified Polymorphic DNA (RAPD) (Williams et al.
1990). DNA was isolated of samples of the leaves according to Cheung et al. (1993) with the addition of 2% PVP to the extraction buffer. In the PCR reaction the primers OPF4, OPF7, OPF9, OPF11, OPF12, OPF13, OPF16 and OPF18 (Operon Technologies) were used. For each possible paternal genotype we scored the presence or absence of at least six unique bands (either homo- or heterozygote). If for both outcross-pollen sources all unique bands were absent, we concluded that a seed was selfed. In this way, the chance of incorrectly classifying a seed as selfed is smaller than 2% (1/2)6.
Comparing the single- and mixed donor experiment
If the single- and the mixed donor experiment give different results, this would indicate that pollen- or early pollen-tube competition affect selection on pollen source or that competition between developing embryos is more intense within than between flowers. The two experi- ments were compared with a G-test. We derived the expected values from the single donor experiment and used the results of the mixed donor experiment as observed values. We multi- plied the number of seeds per flower of the single donor experiment sired by paternal parent j
with the pollen ratio of paternal genotype j in the pollen mixture to obtain the total expected number of seeds per flower in this mixed donor experiment. This expected number of seeds per flower was multiplied by the number of flowers used for each parental combination in the mixed donor experiment to obtain the expected number of seeds produced for each parental combination. The expected number of seeds in the mixed donor experiment produced by maternal parent i, sired by paternal parent j equals:
m j i j
s j i
s j i m j
i F
p p p
p F
S S , ,
3 2 1 , ,
, , ,
, = + +
ε
in which the indices s and m refer to respectively the single- and mixed donor experiment,
Si,j,s / Fi,j,s is the number of seeds per flower of the single donor experiment produced by
maternal parent i and sired by paternal parent j. p1, p2 and p3 denote the number of pollen grains per anther of the three genotypes, and Fi,j,m is the number of flowers in the mixed donor experiment used for the parental combination of maternal parent i and paternal parent j.
Pollen germination and pollen-tube growth
To examine the effect of pollen germination and pollen-tube competition on the non-random siring of seeds, the number and length of the pollen tubes were recorded. We used all nine pa- rental combinations from the mixed donor experiment, and collected the styles (n=69) five hours after pollination with pollen loads from one single donor.
After a time interval of five hours, the fastest pollen tubes were recorded to have grown one third of the style length. The stigmas were fixed in ethanol with acetic acid (4:1) for one hour and stored in 70% ethanol for later observation. Pollen tubes were stained with aniline blue according to Martin (1959). The ripeness of the stigma, the number of pollen grains in the holes of the stigma and the number and length of the pollen tubes were recorded.
Unripe stigmas were discarded from the statistical analysis. Pollen was collected of each genotype and over 100 grains per genotype were viewed under a microscope to record the percentage of collapsed pollen. Collapsed pollen proved to be non-viable when stained (Melser, unpubl. data) according to Alexander (1980).
The number of the pollen tubes in the style were normally distributed and were analyzed statistically with an ANOVA (type III). Pollen tube growth per hour was normally distributed after a square-root transformation and were analyzed statistically with an ANOVA (type III).
Results
Single-donor experiment
Averaged over all mothers, the mean number of seeds per flower was 0.43 (se = 0.017). The number of seeds per flower was strongly influenced by the position of the flower along the cyme and by the date of pollination (Table 1). Generally, late flowers produced fewer seeds than early flowers, although the growth chamber conditions were constant. Maternal parents differed in the number of seeds per flower. There was no main effect of pollination type (self- or cross-pollination). This means that averaged over all maternal parents, the pollination with either self- or cross-pollen did not significantly affect the number of seeds per flower
(average 0.41; se=0.057 and 0.42; se=0.018, respectively). However, the response to pollina- tion type differed significantly among maternal parents, as indicated by the interaction term (Table 1 and Fig. 1). In some maternal parents self-pollen was more successful than outcross- pollen, while in others the reverse was true. The inbreeding depression of the different motherplants ranged from 0.754 to 0.625 with a mean value of 0.125 (sd = 0.424). Paternal effects were also present: one pollen donor (D) sired significantly fewer seeds per flower in outcross-pollinations compared to the other pollen donors (p<0.05; Table 1; Fig. 2). No differences in pollen viability were present for the different pollen donors, the percentage of inviable pollen grains ranged only from 0 to 1%. The low siring success of pollen of genotype D cannot be explained by inviable pollen. In addition, genotype D was not exceptionally unsuccessful after self-pollination (Fig. 1), which indicates that not all of the pollen was inviable. There was no significant interaction between the motherplants and cross-pollen donor (Table 1), which means that the success of an outcross donor did not vary among motherplants. The genotypes with a negative inbreeding depression performed also well as pollen donor in outcross pollinations (Fig. 3). The genotypes with a positive inbreeding depression showed large variation in their success as pollen donor. The dominance coefficient h for the number of seeds per flower was 0.255.
The average mass of a seed was 2.58 g (se = 0.025, n = 671). The mass of the seeds was strongly influenced by the position of the cyme (seeds at lower cymes were heavier), the date of pollination (seeds produced later were smaller) and the maternal genotype (Table 2).
Selfed seeds were on average of the same mass as seeds from cross-pollination: there was no main effect of pollination type on seed mass, nor was there a significant interaction between maternal parent and pollination type. However, the selfed seeds from the maternal parents which produce relatively few selfed seeds are heavier than the outcrossed seeds from the same maternal parent (Fig. 4). Paternal effects on seed mass were absent: both the main effect of
0 0.5 1.0 1.5 2.0
a b c d e f g h i j
Number of seeds per flower
self-pollination cross-pollination
Maternal genotype Figure 1.
The average number of seeds per flower (s.e.) produced after self- and cross-pollinations with pollen of one paternal genotype per pollinated flower in the single-donor experiment. For GLM analysis see Table 1.
cross-pollen donors and the interaction between maternal and paternal genotype were non- significant (Table 2). There was no correlation between the total number of seeds produced and the seed mass (r=0.083; P>0.05).
Mixed-donor experiment
Pollen counts per anther revealed an unequal contribution of the different pollen sources to the mixture: pollen-donor F contributed 21.3%, donor B 48.1% and donor C 30.6% (F=12.68;
df=2; p=0.0003).
After pollination with a pollen mixture of the three pollen genotypes, significantly fewer seeds were sired by self-pollen compared to cross-pollen (Table 3). Among the three cross-pollen genotypes there were no significant differences in the number of seeds sired (Table 4).
Table 1.
GLM analysis of number of seeds per flower in the single-donor experiment with a logit link function (GENMOD procedure, SAS Institute 1985), with position of the flower along the cyme, date of pollination, maternal effect, pollination type as self- or cross-pollination and paternal effect as main effects.
Source of variation df F Position of flower 3 13.1***
Date of pollination 4 44.6***
Maternal effect 9 12.3***
Pollination type 1 0.1 Maternal effect x pollination type 9 2.2*
Paternal effect 9 2.5**
Maternal effect x paternal effect 71 1.1
*P<0.05; **P<0.01; ***P<0.001.
Table 2.
GLM analysis of the mass of the seeds from the single-donor experiment, with position of the cyme along the stem, the position of the flower along the cyme, date of pollination, maternal effect, pollination type as either self- or cross-pollination and paternal effect as main effects.
Source of variation df F Position of cyme 3 13.2***
Position of flower 3 0.56 Date of pollination 4 66.43***
Maternal effect 9 58.80***
Pollination type 1 0.02 Maternal effect x pollination type 9 1.45
Paternal effect 9 1.49
Maternal effect x paternal effect 71 1.30
***P<0.001.
0 0.25 0.50 0.75
a b c d e f i g h j
Paternal genotype
*
Number of seeds sired per flower
Figure 2:
Paternal success as the average number of seeds sired (s.e.) per flower per paternal genotype at 9 maternal genotypes after cross-pollinations with pollen of one paternal genotype per pollinated flower in the single-donor experiment. *P<0.05, Miller's multi-stage procedure.
Figure 3.
The relation between the inbreeding depression (δ, see Material and Methods) in the stage of seed set, and the success as outcross-pollen donor per genotype in the single-donor experiment.
Maternal parents with δ>1 were significant less successful as outcross-pollen donor compared to maternal parents with δ<1. (P<0.05; Wilcoxon).
Figure 4.
The relation between the ratio of the number of the selfed and crossed seeds, and the ratio of the mass of the selfed and crossed seeds per maternal parent. y = -0.125 x + 1.16; r = 0.7118 (P<0.05).
Table 3.
Number of observed and expected seeds after self- and cross-pollinations from the mixed-donor experiment. The expected number of seeds is adjusted for the different percentages of each pollen genotype in the pollen mixture.
Maternal genotype
F B C
Obs Exp Obs Exp Obs Exp
Self-pollination 6 5.96 5 13.95 7 9.18 Cross-pollination 22 22.04 24 15.05 23 20.82
G = 12.92; df = 3; p = 0.005
Table 4.
Number of observed and expected seeds after different cross-pollinations from the mixed-donor experiment. The expected number of seeds is adjusted for the different percentages of each pollen genotype in the pollen mixture.
Paternal genotype
F B C
Maternal genotype Obs Exp Obs Exp Obs Exp
F - 11 13.45 11 8.55
B 12 9.85 - 12 14.15
C 9 7.06 14 15.95 -
G = 2.62; df =3; p = 0.438.
Table 5.
Comparing the mixed- and single-donor experiment: number of observed and expected seeds after self- and cross-pollinations from the three selected maternal genotypes in the single donor experiment. The expected number of seeds is adjusted for the different number of flowers used for the different parental combinations in the single- donor experiment, and the different percentages of each pollen genotype in the pollen mixture in the mixed-donor experiment.
Maternal genotype
F B C
Obs Exp Obs Exp Obs Exp
Self-pollination 6 8.81 3 7.41 4 6.20 Cross-pollination 20 17.19 20 15.59 14 11.80
G = 7.26; df = 3; p = 0.064.
Comparing the mixed- and single-donor experiment
With respect to the effects of selfing, the results of the mixed-donor experiment seem to deviate at first sight from the general results of the single donor experiment. In the single- donor experiment self-pollinations and cross-pollinations produced on average the same number of seeds. In the mixed-donor experiment the cross-pollinations resulted in more seeds compared to the self-pollinations. However, if the data of the single-donor experiment are analyzed with only the three selected genotypes of the mixed-donor experiment included, the very same trend is shown: more seeds after cross-fertilizations compared to self-fertilizations, but only at the margin of significance (Table 5). The non-significance may be due to a smaller number of seeds derived from the single-donor experiment compared to the mixed-donor experiment. The inbreeding coëfficients are comparable for both experiments: for the mixed- donor experiment δ equals 0.478 and for the single-donor experiment with only the three selected genotypes δ equals 0.407.
As in the mixed-donor experiment, in the single-donor experiment the three selected genotypes showed no difference in the number of seeds among different cross-pollen genotypes (G=5.29; df=3; p=0.160).
Overall, a marginal, but significant, difference between the two experiments was revealed (G=13.52; df=6; p=0.036), due to differences among the cross-fertilizations.
Pollen germination and pollen tube growth
As mentioned, pollen donors of the three genotypes used in the mixed-donor experiment showed no difference in the percentage of collapsed pollen, which ranged only from 0 to 1%.
The number of pollen tubes present in the style was affected only by maternal genotype (F=4.80; df=2; p=0.012). Neither pollen donor nor the interaction between maternal genotype and pollen donor caused significant differences in the number of pollen tubes after 5 hours.
The pollen tube growth rate during the first 5 hours was affected by maternal geno- type, pollen donor, and the interaction between maternal genotype and pollen donor as well (maternal genotype: F=11.1; df=2; p<0.0001; pollen donor: F=3.97; df=2; p=0.021; inter- action: F=5.31; df=4; p=0.0005). However, the parental combinations which sired relatively many seeds in the mixed donor experiment did not have high growth rates of the pollen tubes (Fig. 5).
Figure 5.
The relation between the pollen tube growth rate after 5 hours and the observed total number of seeds per parental combination in the mixed-donor experiment, adjusted for the percentage of paternal pollen in the pollen mixture.
Discussion
Flowers of Echium vulgare are protandrous and the style is longer than the stamens. Both qualities tend to avoid self-pollinations within a flower, but neither avoids geitonogamy. If there is any selective advantage in avoiding selfing within one flower, we would have expec- ted to find a lower production of seeds through geitonogamous selfing compared to outcros- sing. Surprisingly, averaged over all motherplants of E. vulgare, self-pollinations did not yield significantly fewer seeds compared to cross-pollinations. However, the interaction between maternal parent and pollination type revealed the presence of a selective mechanism. On some motherplants selfing produced relatively most seeds, while in others cross-pollination produced relatively most seeds. In literature, emphasis has been on selection in one direction, favouring cross-pollinations at the cost of self-pollinations (Aquilegia caerulea in Montalvo 1992, Erythronium grandiflorum in Rigney 1995). In an experiment with one single donor per flower we used Cynoglossum officinale (Boraginaceae) and found a similar interaction between maternal parent and pollination type (Chapter 6). This indicates that the occurrence of genotypes favouring selfing over outcrossing might be a more general phenomenon.
Without selection, individuals with a high genetic load could produce a large fraction of offspring which performs poorly in reproduction in later life (Seavey & Bawa 1986, Husband & Schemske 1996). For such plants it is advantageous if less fit embryos are aborted at a very early stage to avoid any further investment in such embryos if other higher quality embryos can be nurtured. Without a genetic load, investment in selfed seeds passes two copies of the plant's genome to the next generation. Therefore individuals with less than 50%
inbreeding depression in total lifetime gain a selective advantage, favouring selfed seeds over outcross seeds. The results of our experiment suggest substantial variation in genetic load among individuals of E. vulgare. Variation in genetic load can be caused by different selfing histories in the different lineages. However, according to Schultz and Willis (1995), mutations to deleterious additive or recessive alleles generally produce far more variation among individuals in inbreeding depression than variation in history of inbreeding. The presence of purely additive mutations can produce large variation in inbreeding depression, even though, on average, selfing is equally successful as outcrossing (Schultz & Willis 1995). Individuals with a large number of additive deleterious mutations show inbreeding depression, while at the other extreme, in individuals with a lower than average number of additive deleterious mutations, selfing is more successful. In this line of reasoning, genotypes which carry additive deleterious alleles, perform poorly both after selfing and as a cross-pollen donor. Genotypes which carry only recessive deleterious alleles, perform relatively poorly after selfing, but sire a normal amount of seed, in outcross pollinations. In Figure 3 we plotted the measured paternal success as an outcross-pollen donor against the inbreeding depression. Individuals with a negative inbreeding depression sire significantly more seeds in outcross pollinations than individuals with a positive inbreeding depression. This indicates that these genotypes have few deleterious alleles. Individuals with a positive inbreeding depression showed large variation in success in outcross pollinations, which indicates that in our population, offspring fitness is determined by a combination of additive and recessive deleterious alleles. This is also illustrated by the dominance factor h with a value of 0.255: intermediate between com- plete recessivity (h=0) and complete additivity (h=0.5). However, the impact of the recessive
deleterious alleles is apparently not large enough to reduce the average seed set significantly after self-pollination compared to cross-pollination. Other Boraginaceae species (Amsinckia gloriosa and A. spectabilis) with an extremely high selfing rate (>0.997) had minimum esti- mated mutation rates of 0.24 to 0.84 per sporophyte genome per generation and showed negative effects of inbreeding (Johnston & Schoen 1995). The estimated levels of dominance for the involved alleles, ranged from recessivity (h indistinguishable from zero) for A. spec- tabilis to incomplete recessivity (mean values for h 0.28 and 0.35) for A. gloriosa.
The results of our experiment are compatible with the notion of selection through competition among embryos. It is conceivable that in a selective arena, embryos with a high genetic load are less competitive for resources and suffer a higher abortion rate. Competition for resources alone, however, cannot explain our results. If the competition among embryos is the only mediating factor for abortion rates, we expect that in plants that produce more seeds after selfing than after cross-pollination (δ<0), the selfed seeds are also bigger than the outcrossed ones. However, the selfed seeds on motherplants which produce relatively many selfed seeds weighed less than the outcrossed seeds. A competitive arena for the embryos is thus unlikely to be the only mediating factor.
In literature, next to the effects of inbreeding depression, also outbreeding depression has been reported. In Delphinium nelsonii, crosses between individuals collected more than 100 m apart had fewer pollen tubes at the entrance of the ovaries compared to crosses between individuals which were collected at intermediate distances (Waser and Price 1993).
In Echium vulgare no interaction between maternal parent and cross-pollen donor has been found. We did not collect genotypes at intermediate distances: all maternal genotypes have been collected at least 100 meters apart. Most of the pollen of E. vulgare is not transported over a distance more than 50 m (Rademaker, unpublished data) and seeds fall close to the plants. We assume that the individuals are not close related. Therefore we could not exclude the possibility that, on average, selfing leads to a reduced seed set if compared with outcros- sing between plants separated by intermediate distances.
The difference in seed production between self- and cross-pollinations is not enhanced if the pollen is applied as a pollen mixture compared to one single donor per flower. The use of the two different pollination methods revealed only marginal differences (due to the contri- bution of only the cross-pollen donors) between the single- and mixed-donor experiment. This indicates that the competition between embryos within one flower does not differ substan- tially from the competition between embryos in different flowers. The small differences could not be explained by the number of pollen tubes in the style, nor by the early pollen tube growth rate. A negative correlation between early pollen tube growth and late pollen tube growth has been shown for maize (Sari-Gorla et al. 1995). If a similar negative correlation holds true for E. vulgare, the competition between pollen tubes at a later stage could explain the differences between the two experiments in siring success of the cross-pollen donors.
Some selection on pollen sources has been shown in E. vulgare. If there is a relation between the quality of the offspring and the abortion rate, then an initial surplus of embryos enables the maternal parent to select the offspring with the highest fitness. However, the quality of the discarded embryos is unknown. Part of the selection between selfed and out- crossed seeds seems to be genetically determined by both additive and recessive deleterious
alleles. The variation in genetic load could be due either to differential purging of alleles in the different parental lineages, or to newly generation through random mutations. With reces- sive deleterious alleles present in the population, an initial surplus of embryos is especially advantageous in self-compatible species because selective abortion enables to enjoy the advantages of selfing at relatively low costs. Further experiments will focus on the heritability of the relative abortion rates of selfed and outcrossed offspring.
Acknowledgements
We thank C. van der Veen-van Wijk, H. Nell, H. de Heiden, J. Mols and A. Bijleveld for field and greenhouse assistance and seed weighing. J.M.M. Peters-van Rijn made the RAPD- method suitable for E. vulgare. H. Jansen (CPRO Wageningen), advised helpful with the statistics. T.J. de Jong, E. van der Meijden and D.L. Mulcahy made useful comments on an earlier version.
