2
Embryo abortion in natural populations of Cynoglossum officinale (Boraginaceae).
Summary
In many plant species, more flowers and ovules are produced than seem necessary for the production of seeds. Apart from the causal explanation that pollen quantity and quality can limit seed set, considerable attention has been given to the evolutionary hypothesis of selective embryo abortion. Under this hypothesis, the mother plant can increase her fitness by preferentially maturing the offspring of the highest quality. For only very few species it has been established that indeed many embryos are aborted. In this paper, we tested whether pollen was limiting seed production both at the plant and at the individual flower level and we examined embryo abortion in a natural population of Cynoglossum officinale.
Ovules were viewed microscopically. In 46.2% of the ovules embryos were found. The average number of seeds per ovule was smaller and decreased faster in time than the average number of embryos per ovule. As a result, the minimum estimate of the abortion level in- creased with time from 16.4% to 81.8%. Averaged over all flowers, the abortion level is esti- mated at 39%. Embryo or seed production was not limited by either pollen or nutrients. We discuss that both male function and selective embryo abortion can explain the evolutionary advantage of an overproduction of ovules, while bet hedging is an unlikely hypothesis for our study species.
C. Melser, L. Goosen-de Roo, H. Nell and P.G.L. Klinkhamer.
Introduction
Many plant species produce far more ovules than seems to be necessary for the production of their final number of seeds. Selective embryo abortion has been mentioned as one of the evolutionary explanations of this “surplus of flowers”. The overproduction of embryos gives the opportunity for selective maturation of the embryos with the highest fitness later in life.
The selective maturation of offspring with the highest quality could increase the fitness of the maternal parent. This hypothesis has received considerable attention (e.g. Darwin 1883, Janzen 1977, Lloyd 1980, Stephenson 1981, Casper 1984, 1988, Marshall & Ellstrand 1988).
The hypothesis of selective embryo abortion assumes that otherwise viable embryos are aborted. However, little is known about the level of embryo abortion in natural populations.
For numerous species, a low seed production per ovule is not caused by the lack of compatible pollen (e.g. Stephenson 1981, Willson & Burley 1983). Burd (1994) argued that pollen limitation has been underestimated under natural conditions. If so, we would expect a higher seed production when ample additional compatible pollen is applied and levels of em- bryo abortion should be low. Even if the available resources rather than pollination are the limiting factor for the production of mature seeds, pollen availability might still limit the pro- duction of embryos. Individual flowers can then still be pollen limited, while the plant as a whole is not. Under this condition, adding only nutrients will increase the total seed produc- tion, but not equally over all individual flowers. Reduced pollen availability in one flower may increase the seed production in another flower of the same plant, by reallocation of resources. This has been found in e.g. Polemonium foliosissimum (Zimmerman & Pyke 1988), Sorbus aucuparia (Sperens 1996) and Alstroemeria aurea (Aizen & Searcy 1998). Flowers that received extra pollination produced more seeds than control flowers on the same plants, indicating pollen limitation for individual flowers. However, the total seed production of the experimental plants was equal to control plants. Additional pollination might then increase the number of embryos, while the amount of resources available determines the overall level of abortion. It is therefore difficult to infer the level of embryo abortion from pollination experi- ments. However, other methods to estimate levels of embryo abortion present problems as well.
Levels of embryo abortion have been measured indirectly by removing flowers or ovu- les. With this method, an increase in seed set in the remaining flowers or ovules indicates that otherwise aborted embryos can mature into seed (Casper 1984, 1988, Andersson 1990, 1993, Stephenson & Winsor 1986, Lee & Bazzaz 1986, de Jong & Klinkhamer 1989, Chapter 3).
With this method, the extreme estimates of viable embryos that are aborted range from 4% in Anchusa officinale (Andersson 1990) and Achillea ptarmica (Andersson 1993), to 56.8% in Cynoglossum officinale (Chapter 3). However, this method provides only a minimum estima- tion of embryos and provides no reliable extrapolation to the numbers of embryos originally present. It remains unknown whether in the undeveloped ovules a seed was present and abor- ted, or there was none. Other estimates of embryo abortion are based on counts of empty seed coats (Levri 1998) or expanded but later aborted ovules (Lee & Bazzaz 1986; Stephenson 1984). With these estimates, only abortions that occur relatively late in embryo development are detected (Shuraki & Sedgley 1996). Moreover, it is not known whether swollen ovules always contain an embryo (Ohad et al. 1996, Chaudhury et al. 1997), nor whether ovules that
remain small, are never successfully fertilized. Direct views of initiated embryos under a microscope have been made for Crypthanta flava and C. flavoculata (Casper 1983), but only swollen ovules were selected for the preparates. Such samples are biased, because abortion may start before ovules start to develop and unswollen ovules with an embryo present are not included. In addition, unpollinated flowers might be present in the population but are not included, as swelling of the ovule might follow after pollination. In an experimental hand- pollinated population of Oxalis magnifica, Guth and Weller (1986) cleared the ovules to detect if embryos were present. They found an average fertilization rate between 48-92%, and an abortion rate of 3-48%. In two natural populations with extra hand pollination of Epilobium angustifolium, also an ovule clearing technique has been used (Wiens et al. 1987).
A fertilization rate of 97% and an abortion rate near 30% were found. The more embryos are aborted under natural conditions, the higher is the potential for selection on embryo quality.
To determine both the level of pollen limitation and abortion rate in the field, we exa- mined ovules microscopically from plants with and without additional pollination and nutrients. We compared the number of seeds with the number of embryos present per ovule in a natural population of Cynoglossum officinale. We compared these two numbers for different treatments: with and without additional pollination and with and without additional nutrients along the flowering season, to answer the following questions: 1) Are plants pollen or resource limited for total seed production? 2) Are individual flowers pollen limited for seed production? and 3) what is the potential for selective embryo abortion? The results give an opportunity to discuss the probability of alternative hypotheses for embryo selection.
Material and Methods
Species
Cynoglossum officinale (L.) is a rosette forming monocarpic, self-compatible perennial. From the main flowering stem, cymes diverge at which flowers develop sequentially. The flowering period starts at the beginning of May, and lasts four to five weeks. Each day, one flower can open at each cyme, and each flower remains open for about two days. The dull red-purple corolla fades to blue before abscission. In our study area, the most common visitors of flowers are bumblebees, while honeybees are less common (de Jong & Klinkhamer 1989). Flowers are hermaphroditic with five anthers and four ovules. The four ovules are symmetrically arranged in a square and may all develop into a seed. However, in the sand dunes, plants usually have, on average, fewer than one seed per flower (de Jong & Klinkhamer 1989). In a garden experi- ment, we showed that this low seed to ovule ratio is at least partly due to abortion of viable embryos, because removal of part of the ovules in each flower increased seed set in the re- maining ovules (Chapter 3). In our study area, additional hand pollination did not increase average seed set per flower (de Jong & Klinkhamer 1989).
Experimental design
Plants were selected in a natural population of C. officinale in the dunes of Meyendel, near The Hague, The Netherlands. Sixty individuals were marked and each individual was randomly assigned to one of the four following treatments:
1. watering with nutrients (n+);
2. additional hand pollination of all flowers on a plant (p+);
3. watering with nutrients and additional hand pollination (n+p+);
4. control, no nutrients or additional hand pollination (c).
The four groups of plants did not differ significantly in size (measured with length of longest leaf) at the moment of selection in the middle of April 1997. Treatments were applied to all flowers of a plant. For watering with nutrients (n+ and n+p+), we used per plant 100 ml standard Hoagland solution in addition to 2 l water three times a week. The plants assigned to this treatment got a shield of corrugated plastic (diameter 20 cm) around the rosette. This corrugated plastic was dug in the ground with a depth of approximately 10 cm and rose above ground for approximately 5 cm. While preventing a run off of the solution added, this low wall did not shade the leaves. The additional hand pollinations (p+ and n+p+) were applied with a small brush, always with a mixture of pollen from neighboring plants and self-pollen.
In another experiment with C. officinale, it was found that, averaged over several maternal plants, self-pollination yields an equal number of seeds per flower compared to cross- pollination (Chapter 6). Treatments were applied during the whole flowering period, from the 7th of May until the 9th of July 1997, three times a week. Flowers were marked with a drop of paint to identify the date of opening. Three times a week one flower was sampled of each maternal plant (if available) that had opened 4 or 5 days before: well after wilting but before visible swelling of the ovules. All sampled flowers were stored in 70% ethanol. After seed ripening and the natural death of the plants at the end of July, all plants were harvested and stored in plastic bags. The vast majority of the seeds (99.6%) adhered to the plant in their original position. After cleaning the roots from sand, plants were weighed (shoot, root and seeds separately) to determine total dry mass.
For all flowers on each maternal plant, position, date of flower opening and number of seeds were recorded. Empty seed coats were also registrated. For the analysis of the number of seeds per flower, flowering time was divided in four sequential time periods of equal length over all plants. Number of seeds per flower was averaged per maternal plant for those four periods. The number of seeds per flower were analyzed with an anova for repeated measure- ments for differences among the four treatments in number of seeds per flower (SPSS GLM Repeated Measures Analysis) in time. Two severely damaged plants were excluded from the analysis (one from p+, one from n+). Another four plants (one from treatment n+, three from treatment n+p+) had missing values in the fourth time period and could therefore not be included in the total analysis. These plants did not deviate in overall mean number of seeds per flower from the other plants in the same treatment group. Seed production per flower is positively correlated with plant size (Klinkhamer & de Jong 1987). To correct for differences in seed number per flower due to different plant sizes, we included the total dry mass of the plants as a covariate. Mass of plants is not affected by watering with nutrients (de Jong unpubl. results), so we used dry mass of the plants at the end of the experiment. To detect whether the frequency of flowers without any seed differed between treatments, the
distributions of the seeds per flower of the experimental plants were also compared to the distribution of the control group with a G-test. The total production of flowers and seeds per plant were analyzed with an analysis of variance. To obtain normally distributed residuals, the number of flowers and plant mass were ln transformed. Nutrients and pollination were main factors. For this analysis the four plants with missing values in the fourth time period were included.
Embryo observation
The treatment n+p+ and the control group were analyzed for the presence of embryos in the ovules. Samples for embryo observation were taken over all 15 maternal plants per treatment.
The experiment was divided in two periods of equal length. The time-consuming method to obtain microscopical samples did not allow us to use a more detailed time division. For each time period, one to four ovules were screened for each maternal plant (105 samples). For each plant, data were averaged per period to give an equal representation of each maternal plant.
From the samples of the collected flowers that were stored in 70% ethanol, ovules were isolated and dehydrated in ethanol (96% and 100%, each step for 30 min.) and propylene oxide (Agar Scientific Ltd., Stansted, UK) for 1 h. Thereafter, the ovules were infiltrated with Epon (Serva, Heidelberg, Germany) by successive incubations for at least 1 hour with mixtures of Epon and propylene oxide in the ratios 1:2, 1:1 and 2:1, and incubated in pure Epon, overnight. The specimens were embedded in Epon in BEEM capsules. Epon was allowed to polymerize for 48 h at 60oC. Longitudinal serial sections of 8µ m thickness were cut with glass knives on a LKB pyramitome. The sections were collected onto slides, and after
Figure 1:
Embryo of C. officinale in a pre-globular stage of eight cells. Viewed with a light microscope. Scale bar = P
staining by use of toluidine blue 1% in sodium tetraborate 1% (equal volumes of the two stocks) on a hot plate for at least 5 min. at 60oC, mounted in Epon and polymerized. The presence or absence of an embryo in an ovule was investigated with a light microscope. In this way, the stage of embryo development, from the four-cellular stage and upwards, could be determined (Tokč 1976, see Fig. 1).
Analysis of abortion levels
For the two time periods, we compared the mean number of seeds per ovule with the mean number of embryos per ovule per maternal plant for the treatments n+p+ and control. The difference provided a minimum estimate of the average level of abortion. The average number of seeds per flower were divided by four to achieve the number of seeds per ovule. The data were analyzed with manova (GLM multivariate analysis in SPSS), with dry mass of the plants as covariate, treatments as fixed factor, period as random factor, and average number of seeds per ovule, average number of embryos per ovule and abortion level as variables. In this ana- lysis, the average numbers of seeds and average number of embryos were transformed arcsine square root to obtain normally distributed residuals. The factor period showed a significant de- viation of sphericity, therefore we used a very conservative test for the effect of period, which is the ‘lower-bound’ epsilon for adjusting the degrees of freedom (SPSS 1997). To estimate an average lifetime abortion rate, abortion percentages for the two periods were weighed by the number of flowers produced in the two periods.
Figure 2:
Relation between the number of seeds per flower and the total plant mass (at the end of the experiment). Control: no nutrients or additional handpollination. p+: additional handpollination of all flowers on a plant. n+p+: watering with nutrients and additional handpollination. n+: watering with nutrients.
Results
Number of seeds per flower
The average number of seeds per flower over all plants was 1.13 (n=58; se=0.0424). Plants that have a higher mass at the end of the season, also produced more seeds per flower (Fig. 2;
F-value=8.58; df=1,49; p=0.0051). The number of seeds per flower decreased strongly along the season (Fig. 3; F-value=21.591; df=1,49; p<0.0001). Contrary to our expectations, the four treatments did not differ significantly from each other (F-value=0.893; df=3,49; p=0.4515).
When plant mass was not included as a covariate, this result did not change.
To examine whether the lack of sufficient pollination would increase the number of seedless flowers, the frequency distributions of seeds per flower per plant within the experi- mental treatments, were compared to the control group (Fig. 4). Although each treatment differed significantly from the control, the actual differences were very small. The percentage of e.g. empty flowers in the control group was 46.9%, and the application of extra hand polli- nation decreased this percentage only to 43.1%.
Total number of flowers
The heaviest plants at the end of the season had a greater total number of flowers per plant (F- value 110.0, df=1,53 p<0.001). The treatments appeared to have no significant additional effect on the total number of flowers per maternal plant (F-value =2.28, df=3,53 p=0.090).
Figure 3:
Average number of seeds per flower for the different treatments in time - estimated marginal means at mean plant mass (se). control: no nutrients or additional handpollination. p+: additional handpollination of all flowers on a plant. n+p+: watering with nutrients and additional handpollination. n+: watering with nutrients. Time periods are of equal length in days over the flowering period of all plants.
Without including plant mass as a covariate, the treatment effects remained insignificant, indicating that the effect of plant mass did not mask an eventual effect of treatment.
Total number of seeds
The plants with more mass at the end of the season had a greater total number of seeds per plant (F-value=200.9; df=1,53; p<0.0001). The experimental treatments had no effect on the total number of seeds per maternal plant (F-value=2.19, df=3,53 p=0.101). The average mass per seed was not correlated with plant mass at the end of the season (F-value=0.77, df=1,53;
p=0.385). Neither did the treatments affect average mass of the seeds (F-value=0.49, df=3,53;
p=0.693). Without including plant mass as a covariate, the treatment effects remained insignificant, indicating that also without accounting for plant mass, there were no significant differences between the treatments.
Figure 4:
Frequency distributions of seeds per flowers for all plants of the different treatments. A control plants (no nutrients or additional handpollination); B p+ (additional handpollination of all flowers on a plant; compared with control with the G-test: X2=50,7; df=4; p<0.001); C n+ (watering with nutrients; X2=17.3; df=4; p=0.002); D n+p+ (watering with nutrients and additional handpollination;
X2=56.9; df=4; p<0.001).
Abortion levels
Number of embryos per ovule and corresponding abortion levels were determined in the control and n+p+ treatment. The analysis with average number of seeds per ovule, number of embryos per ovule and abortion level per plant as dependent variables and with plant mass as covariate, and treatment and period as factors, yielded significant results (see Table 1). With analyzing only two treatment groups (n+p+ and control) instead of the full four when analyzing the number of seeds per flower, plant mass was not significant in the analysis.
Neither differed the two treatments significantly. Time period was highly significant, indicating overall changes in the variables (embryos, seeds per ovule and abortion level) along the season. The interaction between time and treatment was not significant, indicating that the two treatments reacted similarly in time.
Embryos were recognized from a four-cellular stage until a globular stage. The majo- rity of embryos found were in a pre-globular stage (Fig. 1). Pollen tubes that pass the ovule without fertilizing it were seen inside the hole of the ovule in 11.8% of the preparates in
Table 1:
Manova table for the embryos, seeds and abortion per ovule. Values for embryos and seeds are square root arc sinus transformed.
Factor F value df p value
Intercept 31,80 3 0,000
Plant mass 2,30 3 0,088
Treatment 0,19 3 0,906
Time period 71,80 3 0,000
Treatment x time 0,86 3 0,469
Factor Variable F value df p value
Corrected Model embryo 1,50 4 0,214
Intercept 16,93 1 0,000
Plant mass 0,00 1 0,969
Time period 5,45 1 0,023
Corrected Model seed 49,84 4 0,000
Intercept 86,00 1 0,000
Plant mass 3,11 1 0,083
Time period 193,57 1 0,000
Corrected Model abortus 0,79 4 0,540
Intercept 4,52 1 0,038
Plant mass 0,53 1 0,471
Time period 2,36 1 0,130
which there was no embryo visible. If there was an embryo present, pollen tubes that did not fertilize the ovule were seen inside the hole of the ovule in 1.9% of the preparates. The number of embryos per ovule per period and per treatment is depicted in Fig. 5. The corrected total model of the manova with plant mass, treatment, time period and interaction between treatment and time period as factors and embryo presence, seed presence and abortion level as variables was not significant for the number of embryos per ovule (see Table 1; average 0.461, se=0.052, n=59). However, although the number of embryos per ovule was not significant for the general model, it had a significant effect for time period. This indicates a trend that the number of embryos per ovule declined from 0.59 to 0.34 along the season, which is a decline of 43.1% ((0.591-0.336)/0.591).
The average number of seeds per ovule was significantly affected by period (Table 1, Fig. 5), and declined with 87.7% ((0.494-0.061)/0.494). While the number of embryos de- creased significantly along the season, the decrease in absolute number of aborted embryos per ovule (the absolute difference between number of seeds per ovule and the number of embryos per ovule) was not significant (Table 1, Fig. 5; average 0.185, se=0.055, n=59). The model intercept for abortion (Table 1) differed significantly from zero (p=0.038), indicating that abortion took place. The fraction of aborted embryos increased from an initial 16.4%
((0.591-0.494)/0.591) to 81.8% ((0.336-0.061)/0.336). The difference between the two periods was only on the margin of significance (p=0.0660, Fisher exact test), probably due to our limited sample size.
Figure 5:
Average number of seeds and embryos per ovule per treatment and per time (se). The level of abortion can be found by substracting the average number of embryos per ovule by the average number of seeds per ovule. To maximize the clarity of the graph, standard errors of the mean are drawed only one-sided.
Time periods are of equal length in days over the flowering period of all plants.
To calculate the overall percentage of abortion during the reproductive life of a plant, the average percentages of the two time periods were weighted by the number of flowers produced in both periods. The number of flowers in the first period was on average 216.8 (se=19.21) and was significantly smaller in the second time period (average 115.0, se=13.9;
F=18.2, df=1,57 p<0.0001). Thus, the percentage of abortion over all flowers of a plant is estimated to be
( ) ( ) 39.0%
0 . 115 8 . 216
0 . 115
% 8 . 81 8 . 216
% 4 .
16 =
+
× +
×
Discussion
Embryo abortion
Our experiment has shown that a considerable fraction of the embryos is aborted in C. offici- nale and a low seed set is not due to limitation of pollination. In the start of the flowering sea- son, embryo abortion occurs with a minimum of 16.4%. During time, this percentage increases to minimally 81.8%. If we weigh those percentages with the open flowers in the two time periods, the weighted average of aborted embryos over all flowers is estimated as 39.0%.
This is within the range of estimations for several other species, for which abortion of seed embryos has been estimated under not-natural conditions: 7.6% for Cassia fasciculata (Lee &
Bazzaz 1986), 20.4% and 47.7% for Cryptantha flava (Casper 1984, 1988) and 46.2% for Lotus corniculatus (Stephenson & Winsor 1986). In another experiment with C. officinale we removed three of the four ovules in all flowers of experimental plants. The seed set in the remaining ovule was more than doubled, indicating that otherwise aborted embryos were viable and capable of maturing into seeds. In this indirect way, abortion of viable embryos over all flowers of a plant was estimated at 56.8% (Chapter 3). Both experiments provide a minimum estimation and indicate that there is a high potential for selective embryo abortion.
Why could we not detect an embryo in all ovules?
Embryos were found only in 46.2% of all preparates, including a treatment with supplemental pollen. This percentage is a minimum estimate of embryo presence, because we might have missed embryos that were aborted at a very early stage. Under laboratory conditions where flowers or ovules were removed (Table 2B), the seed set per flower can approach three seeds per flower under laboratory conditions. Compared to an artificial environment, conditions in the field might be suboptimal for embryo formation. However, there are also other reasons for the absence of embryos that we can not exclude. The number of embryos declines with time.
Possibly later in the season abortion occurs in an earlier stage and embryos are more difficult to detect. Another explanation for the decline of embryos along the season may be that there is a maternal cause of fertilization failures or that the quality of ovules declines along the season.
Pollen limitation
Two lines of evidence in this paper suggest that embryo production is not limited by pollen in C. officinale, neither on plant level, nor on the level of individual flowers. In this
experiment, extra pollination did not increase the total seed set of the plant, neither increased the average number of embryos per ovule. Also, if flowers without any seed (which represent 46.9% of the total number of ovules) would indicate that they were not pollinated at all, then we would have expected that the fraction of seedless flowers would diminish substantially if we apply extra pollination, regardless of eventual shifts in the other seed categories. However, the percentages seedless flowers in the p+ treatments are rather similar to the percentages of the control group.
In addition, pruning of flowers in plants of C. officinale increased seed set by 74%
under natural conditions (de Jong & Klinkhamer 1989). This indicates that ovules that other- wise would not have produced seeds were pollinated.
Pollen quality
One might suggest that low seed set is caused by a difference in siring success between self- and cross-pollen. However, the observed low presence of embryos is not likely due to in- breeding depression. Plants of C. officinale have moderate levels of selfing in the field (on average 32% estimated by Vrieling et al. 1999) and have no differences in seed set between seeds from self- or cross-pollinations (de Jong & Klinkhamer 1989, Chapter 6). Moreover, effects of deleterious alleles would neither explain the decrease of number of embryos in time, because the genetic composition of pollen arriving on the stigma is not likely to change syste- matically in time.
Sometimes pollen tubes bypass zygotes as they grow into the ovary (Sayers & Murphy 1966, Hossaert & Valéro 1988) and this leads to non-fertilization of the ovule. It is unknown whether this failure is due to the pollen tubes, or to maternal effects. We found this for 11.8%
of the preparates where no embryo was present. However, this percentage is too low to count for all cases where we did not detect an embryo. Ovarian fertilization failures may not explain the absence of embryos in all samples.
Nutrient limitation
In contrast to the present experiment, watering caused a significant increase in seed set per flower, while additional pollination did not in an earlier experiment with this species (de Jong
& Klinkhamer 1989). The absence of an effect of watering in our experiment might well be due to meteorological conditions. At the time of possible water limitation in this experiment, the months May, June and July in 1997 appeared to be relatively wet with a total precipitation of 205 mm. (KNMI 1997). The normal total amount for this time of the year is 179 mm (KNMI 1997), whereas the year 1986 in which the experiments of de Jong and Klinkhamer (1989) were performed had a relatively dry period from May to July with 168 mm of total precipitation (KNMI 1986).
Although water might not have been a limiting factor in 1997, nutrients could still have been limiting seed production. Moreover, as resources become depleted along the season, the rate of embryo abortion could increase with time (Casper & Niesenbaum 1993, Levri 1998, Parra Tabla et al. 1998). However, adding nutrients and pollination from the onset of flowering onwards, did not increase the average seed set per flower, neither increased the average number of total seeds produced. The average seed set per flower seems thus not to be limited by nutrient availability in the environment during the period of seed set. In the n+p+
Chapter 2 Table 2:
Average number of seeds per flower of C. officinale for different field populations of Meyendel, the Netherlands, and different experi- mental conditions of plants, collected from populations of C. officinale in Meyendel.
A Natural field populations in different years.
Mean seeds/flower s.e. n Year Additional Remarks Reference
1.36 0.01 24 1984 Klinkhamer & de Jong 1987
0.76 0.01 24 1984 Klinkhamer & de Jong 1987
0.67 0.02 21 1984 Isolated plants Klinkhamer & de Jong 1987
0.62 0.05 54 1984 Klinkhamer & de Jong 1987
0.76 0.07 24 1984 de Jong & Klinkhamer 1989
1.18 0.06 56 1985 de Jong & Klinkhamer 1989
0.88 0.05 20 1986 de Jong & Klinkhamer 1989
1.0 10 1987 de Jong & Klinkhamer 1991
0.76 0.04 35 1990 Klinkhamer & de Jong 1993
0.93 0.07 40 1991 de Jong unpubl data
1.49 0.09 15 1997 this experiment (control)
Embryo abortion in natural populations37 B Experimental treatments
Mean
seeds/flower s.e. n Pollination Environ- Additional Additional Remarks Reference
ment nutrients
2.77 0.13 5 H G + Removal of ovules Melser (unpubl data)
2.29-2.86 3-4 H G +/- Removal of flowers at 150C,
with or without Ga or Aux addition Luckerhoff (unpubl data)
2.07 0.12 5 H G + Melser (unpubl data)
1.97 0.13 10 O+H A - Removal of ovules Melser & Klinkhamer (in press)
1.82-2.20 3-11 H G +/- With or without Ga
or Aux addition Luckerhoff (unpubl data)
1.62 0.11 15 O+H F - this experiment (p+)
1.57 0.07 15 O F + this experiment (n+)
1.50 0.09 15 O+H F + this experiment (n+p+)
1.10 88 O F + Transplanted in field de Jong & Klinkhamer 1991
1.09 0.13 10 O+H A + Melser & Klinkhamer (in press)
1.01 0.11 7 H G + Melser & Klinkhamer (in prep)
0.85 0.09 10 O+H A - Melser & Klinkhamer (in press)
0.84 0.05 31 O F - Water addition Klinkhamer & de Jong 1993
0.79 0.06 13 H A + Melser & Klinkhamer (in prep)
0.72 0.1 10 H G + Self pollination de Jong & Klinkhamer 1989
0.72 0.03 10 H G + Klinkhamer & de Jong 1987
0.66 0.21 10 H G - Cross pollination de Jong & Klinkhamer 1989
O = Open pollination H = Hand pollination F = Field population
G = Growth chambers of 200C. unless stated otherwise.
A = Artificial population in garden Ga = Gibberelline
Aux = Auxine
treatment, the seed to flower ratio was still far below its maximum and embryos were aborted at any time period. For our semelparous species, the ability to take up nutrients later in the flowering season might be limited and the internal nutrient status of the plant might have been the limiting factor. In another experiment (Chapter 3), the addition of nutrients during flowering did not increase the number of seeds per flower, but rather increased the number of flowers of the plants. This indicates that nutrients added were used for both extra flower and extra seed production, and although we know that seed number per flower can be adjusted (Table 2), plants maintained seed to flower ratios close to one under a large range of conditions in the field (Table 2A).
Functional explanations of embryo abortion Selective seed abortion hypothesis
The level of embryo abortion we found, shows that in this natural population of C. officinale there is ample room for the selective abortion of embryos of relatively low quality, and this may explain the low seed to flower ratio. In this experiment, we did not test for offspring qua- lity due to selection among embryos. However, in another study with C. officinale with an im- posed lower abortion level (a reduction in abortion level over 50%) by removing three of the four ovules in each flower, the resulting offspring had a 14% lower survival (after 10 days) compared to offspring from control plants (Chapter 3). In that paper, a model was used to quantify the required increase in offspring quality to compensate for a lower seed to flower ratio if an equal total number of offspring would be produced. The model showed that a very small increase in survival may already lead to considerable differences in seed to flower ratio of the maternal plants (Chapter 3). These results suggest that selective abortion of embryos may also explain a considerable part of the low seed set per flower.
Sex-allocation hypothesis
The hypothesis that assumes that fitness might increase by producing seedless flowers that act mainly as pollen donors is called the sex-allocation hypothesis (e.g. Willson 1979, Queller 1983, Sutherland & Delph 1984). The model in Rademaker (1998) shows that the fitness gain by the male function of flowers explains a considerable reduction of seed number per flower for C. officinale. With a model, based on an evolutionary stable strategy (ESS: a population where each mutant will have a lower fitness compared to the resident population), the optimal seed number per flower due to sex-allocation theory for C. officinale is estimated to be 0.46.
Obviously, the hypothesis of sex-allocation plays an important part in reducing the seed numbers per flower.
Bet hedging strategy hypothesis
Plants might produce a surplus of flowers as a strategy to exploit unpredictable favorable con- ditions during seed set or as a compensation for ovule losses through e.g. herbivory. These factors have been lumped together under the concept of “bet hedging” (e.g. Stephenson 1980, Udovic 1981, Kozlowski & Stearns 1989, Ehrlén 1991). Under this hypothesis, seed produc- tion is expected to be close to its maximum in case of favorable conditions during seed set or in case of extreme losses of ovules.
Ovule losses due to herbivory were neglectable in our experiment as, as far as we know, is true for most populations of C. officinale. But even in experiments that could be interpreted as mimicking extreme seed predation by the removal of flowers or destruction of three of the four ovules per flower (Table 2B, Chapter 3) while ensuring sufficient pollination, seed production is never close to four seeds per flower, although seed set in the remaining flowers or ovules is increased. Moreover, if unpredictable damage would be a key factor under field conditions, then seed production in the field should approach 100% at least in some years in some populations. In the dune populations of Meyendel, this was never found, despite the fact that eleven populations were studied in seven different years (Table 2A).
In this experiment, we also created unforeseen favorable conditions by adding water and nutrients from the onset of flowering onwards. With this treatment, seed set per flower was not significantly increased and remained under the maximum possible, also with extra hand pollination. One might argue that our experiment did not provide optimal conditions.
However, until now values anywhere near the maximum seed set of four seeds per flower has never been reached in the field in the Netherlands (Table 2A), while the wide distribution of C. officinale in Europe (de Jong, Klinkhamer & Boorman 1990) does not suggest the Netherlands to be a marginal environment. The number of embryos per flower exceeds all values of seeds per flower ever found in our study area. Reports for seven years showed that average seed set never increased above 1.49, while our microscopical views revealed an average presence of at least 1.85 (0.462 x 4 ovules per flower) embryos per flower in the field.
Based on the arguments above, we believe that the bet hedging strategy is not the sole explanation for the average low number of seeds per flower in C. officinale.
The three evolutionary hypotheses that determine the optimal seed number per flower are not mutually exclusive and all three may act at the same time. The comparison of the importance of the three hypotheses is beyond the scope of this chapter. A mathematical model compares the relative importance of the hypotheses of sex-allocation and selective embryo abortion (Chapter 8).
Increase of abortion level in time
During the flowering season, the percentage of aborted embryos increased considerably. The internal status of nutrients of the plants could be the limiting factor and the uptake of nutrients late in the flowering season might be limited. Also, the filling of newly produced seeds cost time, and at the end of the flowering season, the chance increased that time will run out. It has been suggested that the production of empty flowers as a relatively "quick" investment in pollen instead of seeds might then be more advantageous late in the season. This pollen may still be the father of seeds that are produced on other plants. This might especially be the case for plant populations that have a wide time range in flowering periods. Pollen from a plant at the end of the flowering stage, can then still arrive on plants in which seed production is taking place. However, the flowering period of C. officinale is rather synchronized over all plants in a population and if late flowers only produce pollen as a quick investment, none of this pollen will probably be able to sire seeds on another plant. Summarized, there are several evolutionary hypotheses that could explain the increase in embryo abortion along time. To analyze the relative importance of those explanations is beyond the scope of this chapter.
The increase in level of embryo abortion in time, gives an elegant opportunity to test for the hypothesis of selective embryo abortion. Under this hypothesis, differences in compe- titive power of the embryos should reflect quality differences of the offspring later in life.
Selection of the best offspring could then increase the fitness of the maternal plant. The em- bryos that are initiated late in the flowering season are competing with many other maturing seeds on the same plant. Resources may be harder to retrieve than early in the season and only few embryos with the highest competitive power may mature into seeds in the latest flowers.
A study to differences in offspring quality from relatively early initiated embryos compared to embryos that are initiated later in the season is presented in Chapter 6. Results suggest that higher levels of embryo abortion, induced by the ending of the flowering season, result in seeds that have less mass, but nevertheless, the offspring survives significantly longer.
Selective seed abortion might thus provide an adequate, although not exclusive, explanation for the overproduction of ovules for C. officinale.
Acknowledgements
Mieke Piekaar, Joep Bovenlander and Kees Koops were of essential help with collecting all the seed data. Gerda Lamers prepared the epoxy resin mixture, Peter van Mulken provided the printing of the photo of the preparates and Martin Brittijn drawed the figures. We also are grateful for all comments on earlier drafts by Tom de Jong, Eddy van der Meijden and Peter van Tienderen.
