|Home | About | Journals | Submit | Contact Us | Français|
Variability in embryo development can influence the rate of seed maturation and seed size, which may have an impact on offspring fitness. While it is expected that embryo development will be under maternal control, more controversial hypotheses suggest that the pollen donor and the embryo itself may influence development. These latter possibilities are, however, poorly studied. Characteristics of 10-d-old embryos and seeds of wild radish (Raphanus sativus) were examined to address: (a) the effects of maternal plant and pollen donor on development; (b) the effects of earlier reproductive events (pollen tube growth and fertilization) on embryos and seeds, and the influence of embryo size on mature seed mass; (c) the effect of water stress on embryos and seeds; (d) the effect of stress on correlations of embryo and seed characteristics with earlier and later reproductive events and stages; and (e) changes in maternal and paternal effects on embryo and seed characteristics during development.
Eight maternal plants (two each from four families) and four pollen donors were crossed and developing gynoecia were collected at 10 d post-pollination. Half of the maternal plants experienced water stress. Characteristics of embryos and seeds were summarized and also compared with earlier and later developmental stages.
In addition to the expected effects of the maternal plants, all embryo characters differed among pollen donors. Paternal effects varied over time, suggesting that there are windows of opportunity for pollen donors to influence embryo development. Water-stress treatment altered embryo characteristics; embryos were smaller and less developed. In addition, correlations of embryo characteristics with earlier and later stages changed dramatically with water stress.
The expected maternal effects on embryo development were observed, but there was also evidence for an early paternal role. The relative effects of these controls may change over time. Thus, there may be times in development when selection on the maternal, paternal or embryo contributions to development are more and less likely.
Within developing fruits the number of seeds that mature may vary; moreover, even among seeds that mature, not all develop at the same rate or to the same extent. Seed size is well known to affect seedling performance (Westoby et al., 1992, 1996; reviewed in Halpern, 2005) and seeds that mature more quickly may have less exposure to biotic and abiotic insults and garner more maternal resources if competition among seeds and fruits occurs (Sakai, 2007). While fruit set and final seed mass are often studied, variation in the development of seeds that underlies these patterns is rarely examined.
The classic view of embryo/seed development is that it is largely under maternal control (Raghavan, 2005). Seed size may vary both among and within maternal plants. Within plants, independent assortment of alleles during meiosis contributes to genetic differences among embryos and endosperms that may result in developmental variation among sibs (e.g. Mazer and Gorchov, 1996). In addition, because maternal plant tissues surround and supply the developing embryo and endosperm, maternal plants may be able to selectively provision offspring of differing quality (reviewed in Marshall and Folsom, 1991; Skogsmyr and Lankinen, 2002).
Environmental factors may change maternal influences on seed development. In particular, maternal stress, such as low water or herbivory, can reduce resources available to complete the development of seeds and fruit, and can affect key events both during and after pollination (Lloyd, 1980; Obeso, 2004; Gehring and Delph, 2006). Maternal stress changes the stylar environment in which pollen tubes grow and can alter the outcome of pollen competition (e.g. Haileselassie et al., 2005) and, in turn, alter the paternity of resulting offspring. During seed development, resource limitation may cause maternal plants to make smaller seeds (Vaughton and Ramsey, 1997) or to selectively abort fruits or seeds (Stephenson, 1981; Sun et al., 2004; Wise and Cummins, 2006). If maternal plants do selectively abort embryos, the embryos selected or the strength of selection may differ between stressed and unstressed plants.
Although the mechanisms of control are less obvious, the paternal plant also influences embryo development. The paternal genome might act in two ways: (1) some embryos might develop faster because of correlations between paternal genes that affect mating success and genes that affect offspring growth (Currah, 1981; Sarr et al, 1983; Marshall and Ellstrand, 1986, 1988; Mazer 1987; Marshall and Diggle, 2001; Skogsmyr and Lankinen, 2002; Bernasconi, 2003); (2) paternal alleles contribute to embryo and endosperm vigour and hence variation in mature embryo and seed size (Haig and Westoby, 1989, 1991). The few studies that have investigated maternal and paternal gene expression provide no consensus on the relative contribution of each set of parental genes during embryo development. In arabidopsis, early stages of embryogenesis may be exclusively under maternal control, with expression of paternal alleles beginning at the globular stage onward (Vielle-Calzada et al., 2000; Baroux et al., 2007; Pillot et al., 2010). Recent studies, however, have detected expression of a few paternal loci at earlier stages (e.g. Weijers et al., 2001; Bayer et al., 2009). Moreover, remodelling of male chromatin, which is consistent with active transcription, occurs in the zygote shortly after karyogamy (Baroux et al. 2007).
Finally, the embryo can influence its own development; it is a new organism that is genetically distinct from both parents and all sibs. The embryo genotype, including additive, epistatic, and pleiotropic effects, can influence development independently of maternal, paternal or endosperm effects (Marshall and Ellstrand, 1986; Antonovics and Schmitt, 1986; Marshall, 1988; Dieckmann and Link, 2010). Without any influence of maternal tissues or endosperm, somatic embryos in culture undergo all normal stages of morphogenesis and histogenesis (reviewed in Dodeman et al., 1997).
Most previous studies of maternal, paternal and environmental effects on offspring make inferences based on mature seed size (reviewed in Roach and Wulff, 1987; Mazer and Gorchov, 1996). However, embryo development is a dynamic process during which multiple opportunities exist for maternal, paternal, embryo and environmental influences. The potential for these factors to affect specific developmental stages will be masked by inferences about embryo development based solely on mature seed size. Thus, the study of stages of offspring development other than final seed size is important. Embryo development, however, takes place within the confines of the seed and fruit, making direct observation impossible. Therefore, destructive sampling and histological techniques were used to quantify maternal, paternal, embryo and environmental effects on embryo developmental dynamics.
To report on factors affecting embryo development, features of embryos and surrounding maternal tissues were examined in wild radish, Raphanus sativus. The focus was primarily on 10-day-old (10-d) embryos because they were large enough for accurate measurement and were undergoing critical morphogenetic events such as cotyledon initiation. Nakamura and Stanton (1989) also examined embryo growth in wild radish but in much less detail and at a later developmental stage than reported here. Furthermore, while Nakamura and Stanton (1989) separated maternal environmental effects from genetic effects by moving embryos to tissue culture, in the present study the maternal environment was altered explicitly by imposing water stress.
The following questions were asked. (a) Do maternal plant and pollen donor identity affect characteristics of developing embryos and seeds of wild radish? (b) Are characteristics of 10-d embryos and seeds affected by the timing of earlier reproductive events (pollen tube growth and fertilization order) and does embryo size, in turn, influence mature seed size? (c) What are the effects of maternal plant stress on parental contributions to embryo and seed development? (d) How does stress affect correlations of embryo and seed characteristics with earlier and later reproductive events and stages? (e) Do the relative effects of maternal and paternal plants on embryo and seed characteristics change during development?
Raphanus sativus is a weedy annual found in abandoned fields and along roadsides in California. Plants are visited by an array of pollinators including honeybees, syrphid flies and lepidoptera (Kay, 1976; Stanton, 1987a, b; Stanton and Preston, 1988; D. L. Marshall, pers. obs.). Raphanus sativus is ideal for the present study for several reasons: (a) flowers are numerous (hundreds per plant), allowing many kinds and replications of crosses; (b) developing seeds are large enough to be individually dissected for analysis of embryo characteristics; and (c) embryo stages of development in the Brassicaceae are well studied. Mature seed masses, about 10 mg, are large enough that individual seeds can be weighed. Finally, the wealth of previous work on wild radish places this study in the context of the entire life cycle of this species (e.g. Ellstrand, 1984; Ellstrand and Marshall, 1985a, b, 1986; Marshall and Ellstrand, 1986, 1988; Marshall, 1988, 1991, 1998; Marshall and Fuller, 1994; Marshall et al., 1996, 2007a, b; Marshall and Diggle, 2001).
Eight maternal plants, two each from four full-sibships, and four pollen donors, also from different sibships, were selected. The maternal plants were eight of the 16 plants used by Marshall and Diggle (2001). This report focuses on embryo and seed variables that were not analysed or reported in the earlier work. The plants were chosen primarily to minimize any obvious or cryptic effects of the incompatibility (S locus) genotype and, secondarily, on the basis of suitable genotypes at two electrophoretic loci (PGI and LAP) used to trace paternity in pollinations that used a mixture of different donors. The experimental plants were second generation descendents of seed collected from two wild populations in Riverside, CA, USA. Crosses were initiated by selecting plants haphazardly from among 50 families, and crossing them reciprocally until a group of 16 fully inter-compatible plants was identified (Karron et al., 1990). Because of the sporophytic incompatibility system, all cross-compatible plants have different S alleles. Finding this set of fully crossable plants required nearly 10 000 test crosses. Additional details of the crossing design and paternity analyses are in Marshall and Diggle (2001). Results of mixed-donor pollinations were not informative in the context of the current analysis and are not presented here (details available on request).
Each of the four pollen donors (A, B, C and D) was crossed with each of the eight maternal plants. Replicate crosses were made so that ovaries and developing fruits could be collected at various time intervals (described under tissue collection, below). Each pollen donor by maternal plant cross was repeated four times per collection interval.
Pollinations were performed by collecting pollen in a small Petri dish and applying the pollen with tissue-wrapped forceps. Stigmas were thoroughly coated with pollen. Pollinations placed several hundred pollen grains on the stigma and produced numerous pollen tubes in the style (D. L. Marshall, pers. obs.). Both the number of pollen grains and pollen tubes were greater than the number of ovules (mean ovule number per flower was 6·8).
To test for effects of maternal condition on the reproductive process, the eight maternal plants were divided into two treatments. One maternal plant from each family was a control and the other was given reduced water. Control and water-stressed plants were selected at random within families. Stressed plants were given half of the control amount of water daily. This resulted in nearly daily wilting. Treatments were initiated before pollinations began.
To examine early reproductive events, replicate collections were made at the following post-pollination intervals: 30 min to score pollen germination; 6 h to score pollen tube length; 12, 16, 20 and 24 h to score pollen tube location within ovaries and ovule fertilization, and after 3 d to score fertilization and seed size. These data were reported in Marshall and Diggle (2001) and are used here only to describe correlations with later embryo and seed characteristics. Gynoecia were collected at 10 d for detailed analyses of embryos and seeds. Fruits were collected at maturity to score seed mass.
Measurement of pollen germination, pollen tube growth, fertilization, size of 3-d seeds, seed mass and seed location within the ovary are described in Marshall and Diggle (2001). Gynoecia collected at 10 d were fixed in 3 : 1 70 % ethanol : glacial acetic acid for 24 h and then stored in 70 % ethanol until scoring. Stored tissue was prepared for examination by rehydration and clearing in 0·8 m NaOH for 30–50 min at 60 °C, and staining overnight in 1 mg mL−1 aniline blue with 0·1 mg mL−1 ethidium bromide in 0·33 m K3PO4.
Developing seeds were dissected from the cleared ovaries and flattened slightly under a cover slip so that the embryo was visible through the intervening tissue. Embryo and seed characteristics were measured at ×50 with a Zeiss Axioskop equipped for Normarski optics. A video camera (Panasonic CCTV camera #WV-BD400) was used to capture the image of the seed. Embryo developmental stage (aborted, proembryo, globular or heart-shaped) was noted and MorphoSys (Meacham and Duncan, 1990) was used to measure total embryo length (including the suspensor), length and width of the embryo proper, seed area and width of the funicular vascular trace. An index of seed area was created by tracing the circumference of the slightly squashed seed on the captured image and calculating the resulting area.
To examine the relationship between 10-d embryo and seed measures (described in the previous paragraph) and final seed mass, individual seeds were weighed to the nearest 0·01 mg.
Some variables were combined to reduce the number of variables considered: (a) embryo length and width (these were measures of the embryo proper and did not include the suspensor) were multiplied to produce an estimate of embryo area; (b) a principal components analysis was used to summarize data on the time course of ovule fertilization (Hotelling, 1933). The mean proportion of ovules fertilized at 12, 16, 20 and 24 h post-pollination for each maternal plant by pollen donor combination was used in the principle component analysis. The first axis accounted for 80 % of the total variation and the scores on this first axis were used as a variable, ‘fertilization’, to examine correlations between the time course of fertilization and 10-d embryo characters.
Proportion of heart-shaped embryos and proportion of pollen grains germinated at 30 min were arcsine square-root transformed prior to analysis to meet assumptions of normality. None of the other variables required transformation.
Measures of embryo and seed characteristics at 10 d were compared among maternal families, pollen donors and stress treatments using a series of ANOVAs in which embryo and seed characteristics were the dependent variable and maternal family, pollen donor, stress treatment and their two-way interactions were the independent variables. The pollen donors and maternal families were treated as fixed effects because they were not randomly chosen, rather they were chosen based on careful screening for S-alleles and electrophoretic patterns. The screening was done to minimize effects of incompatibility in the mating system and to trace paternity.
Stage of embryo development (aborted, proembryo, globular or heart-shaped) was compared among maternal families, stress treatments and pollen donors using a logistic regression.
To determine whether variation in 10-d embryos and seeds was related to processes of earlier and later development, correlations were examined between the seed and embryo characteristics at 10 d and measures of pollen tube growth, fertilization and mature seed mass. For these correlations, the data points were the average values of each measurement for each maternal plant by pollen donor combination, making the sample size 32. First patterns of correlations were examined for the entire data set. Then, because differences in patterns of development when plants were under stress were expected, separate correlations for control and water-stressed plants were performed.
Because previous data (Marshall and Ellstrand, 1988) showed that the order of fertilization can depend on a seed's position within the fruit, position was scored within the fruit for each embryo and seed measured. However, preliminary analyses revealed that seed position within the ovary had virtually no effect on the features of interest here. Therefore, to avoid over-specification of the ANOVA models, and to concentrate on more interesting patterns, position within the ovary was ignored in the analyses reported here.
All statistical analyses were performed in SAS version 9.1.3 (SAS Institute, Cary, NC, USA).
Ten days after pollination the frequency of embryos in the aborted, proembryo, globular or heart-shaped developmental stage varied among pollen donors (pollen donor effect, Table 1). In particular, embryos sired by donor B were more likely to have reached the heart-shaped stage than those sired by other donors, while those sired by donor A were more likely to be in the globular stage (Fig. 1). In contrast, the frequencies of embryos at a particular stage did not differ significantly among maternal families (maternal family effect, Table 1). Additional effects of individual maternal plants may be strong, however, as the interaction between maternal family and stress treatment is significant (Table 1). Because there was only one plant per family allocated to each treatment, the family × stress interaction includes both differences among individual maternal plants within families and effects of water stress within individual families.
Means of all 10-d embryo and seed measures varied significantly among pollen donors (Tables 2 and and3).3). Pollen donor identity even had a significant effect on the funicular vascular trace, a maternal tissue. Embryo and seed measurements also varied among maternal families (maternal family effect, Table 2) and among maternal plants within families when they received different treatments (maternal family × stress treatment effect, Table 2). Maternal family by pollen donor effects were not particularly strong (Table 2), suggesting that effects of pollen donor identity were relatively consistent across maternal families.
As expected for characteristics determined during maternal provisioning, either maternal family or the interaction between maternal family and stress explained the most variance in all embryo measurements (Table 2). Pollen donor identity did not explain as much of the variance in embryo and seed measurements (Table 2); however, for embryo area at 10 d, pollen donor identity explained 4·15 % of the variance, more than that explained by maternal family (3·45 %; Table 2).
All of the measurements of 10-d embryo and seed size were positively correlated, although not all correlations were statistically significant (Table 4). While total embryo length, which includes the suspensor, was positively correlated with the area of the embryo proper (r = 0·48), it was more strongly correlated with seed area (r = 0·59). Embryo area also was significantly correlated with seed area (r = 0·49) and with the width of the funicular vascular trace (r = 0·58).
Total embryo length, which includes the suspensor, and seed area at 10 d were not significantly correlated with any of the characteristics measured earlier or later in development (Fig. 2). Particularly surprising is that the 10-d seed area was uncorrelated with the 3-d seed area. In contrast, both the area of the embryo proper at 10 d and the width of the funicular vascular trace were positively correlated with measures of fertilization (fertilization is a composite measure of 12–24 h fertilization patterns, see Materials and methods) and of seed area at 3 d. Both the embryo and the funicular vascular trace were larger when fertilization occurred faster. None of the 10-d embryo and seed measures was correlated with final seed mass. Note, however, that this analysis combines control and water-stressed plants. When only control plants are considered, embryo area at 10 d is correlated with both fertilization and seed area at 3 d, and the percentage of ovules in the heart-shaped stage is positively correlated with both seed area at 3 d and final seed weight (Fig. 3). These relationships suggest that early fertilization and rapid early growth are associated with more rapid passage through early developmental stages and larger mature seeds.
Embryos developing on plants given reduced water were less likely to reach the heart-shaped stage by day 10 than those on control plants (stress treatment, Table 1 and Fig. 4). All 10-d embryo and seed measurements were smaller on plants that received less water (Table 5) and most of these differences were statistically significant (stress treatment, Tables 2 and and5).5). Effects of the water-stress treatment on embryo and seed measures varied across both maternal families and pollen donors (maternal family × stress treatment and pollen donor × stress treatment, Table 2). In the case of maternal families this may be in part due to effects of individual maternal plants within families.
Correlations of 10-d embryo and seed measurements with other variables were strikingly different between control and water-stressed plants (Fig. 3). Half of the correlations changed in direction. For example, all characteristics are positively correlated with pollen germination in low water plants and negatively correlated in control plants. This is statistically significant for embryo area, where faster pollen germination is correlated with larger embryos on low water plants and smaller embryos on control plants. The correlations of 10-d embryo size and stage with 3-d seed area and of 10-d embryo stage with final seed weight observed in control plants are absent in low-water plants, suggesting that early seed growth only is a good predictor of sustained embryo growth under benign conditions.
Examination of the amount of variance in embryo and other characteristics that is explained by maternal family and pollen donor reveals that the influence of the maternal family and pollen donor may change over time (Fig. 5). Maternal family explained more variance than pollen donor for most characteristics and most time points, e.g. the rate of pollen germination on stigmas, the number of ovules fertilized after 12 h, seed area at 3 and 10 d, and final seed mass. Pollen donor identity explained only a small proportion of the variance at most time points, with the exception of pollen tube length and 10-d embryo area. Pollen donor contribution to the variance of these characters was large, and exceeded that of maternal family.
Variation in mature seed size among and within plants is common (Mazer et al., 1986; Leishman et al., 2000; Holland et al., 2009). Underlying this variation is a complex array of interactions involving genes, genomes, and environment that may change during development. In this paper, the dynamics of embryo and seed development are addressed in Raphanus sativus. As expected, a strong maternal influence was found, but also significant paternal effects were found on embryo and seed characteristics at different times during development. Analysis of correlations among the measurements at 10 d with those taken at earlier and later stages showed that the timing of fertilization affected features of 10-d embryos and seeds, and that stage of embryo development at 10 d affected final seed size.
Early embryo and seed development is expected to be primarily under maternal control because these processes occur within maternally derived tissues. The strong effect of maternal family on 10-d embryo and seed characteristics was clear in all analyses. The maternal family influenced embryo size but not the developmental stage. In other words, although embryos and seeds of different maternal parents may grow at different rates (achieving different sizes after 10 d), the embryos pass through the developmental stages of proembryo, globular and heart-shaped at the same rate. Maternal control of rate of growth and rate of development must be separate to some extent. Perhaps the mechanisms controlling embryo phase transition are not completely coupled to those controlling components of embryo growth such as cell division and enlargement. Nakamura and Stanton (1989) also found that maternal plant identity of R. sativus affected developmental stage of older (14 d) embryos. However, the stages recognized in their analysis (torpedo to inverted-U stage) are more reflective of size than morphogenesis. While the transition from globular to heart-shaped stages (as in the current study) involves the initiation of cotyledons, the transition from torpedo to U-shaped is primarily the result of enlargement of the embryo within the confines of the anatropous seed. Thus, while maternal effects on size are likely to persist into later stages of embryo development in R. sativus (to day 14), it is unclear whether effects on stage persist. Although there are many studies of maternal effects on features of mature embryos and seeds, very little work has examined maternal effects during early seed development (Ma, 2000; Russinova and de Vries, 2000; Autran et al., 2005; Mosher et al., 2009; Adamski et al., 2009).
In contrast to the widely held expectation of strong maternal effects, paternal effects on early plant development are generally thought to play a smaller role. Surprisingly, pollen donor identity affected many embryo and seed characteristics in R. sativus. The strongest paternal effects were on embryo size measures, where paternal effects were stronger than maternal family for embryo area (Table 2, proportion of variance explained). Embryos sired by different donors grew at different rates and, in contrast to the findings for maternal family, they also passed through developmental stages at different rates. Paternity of embryos also influenced 10-d seed area and the maternally derived funicular vascular trace, although the effects were much smaller than maternal effects (Table 2). The paternal effect on funicular vascular trace, the source of nutrients for the developing seed and embryo, is particularly interesting, as it provides evidence that this provisioning is not entirely under maternal control. The pollen donor effect may be indirect through embryo development; wider funicular traces are associated with larger embryos (Table 4). Perhaps the greater sink strength of large embryos elicits development of greater conducting capacity in the funiculus. It was not possible to determine the cell types involved or the extent to which cell number and cell size contributed to the increased size. The absence of a significant maternal plant × paternal plant interaction on funiculus size indicates that maternal plants are not supporting embryos differentially based on their genotype (sire identity).
Pollen donor identity also significantly affected the developmental stage of 14-d R. sativus embryos, explaining about 2 % of the variance (Nakamura and Stanton, 1989). After 9 additional days in tissue culture (23 d of development in total), where the maternal environment could no longer affect development, paternal effects on embryo size were slightly larger than maternal effects (Nakamura and Stanton, 1989). Recent work on embryo development in the model organism Arabidopsis thaliana has shown that paternal alleles can be expressed as early as 3–4 d after fertilization (Vielle-Calzada et al., 2000; Weijers et al., 2001; Bayer et al., 2009); and it is generally accepted that the zygotic genome (including paternally contributed alleles) assumes greater control by the globular phase of embryo development (Baroux et al., 2007). Thus, paternal influence on early embryo growth dynamics may be a general phenomenon and warrants further study in other organisms.
Larger 10-d embryos (and their associated larger seed area and vascular trace) are associated with delayed pollen germination, more rapid fertilization, and larger seed size at 3 d post-fertilization (Figs 2 and and3,3, control plants). Delayed pollen germination may be due to interference competition between pollen grains that, in turn, may affect embryo quality (Marshall et al., 1996). The association of both embryo size and vascular trace with rapid fertilization suggests that once pollen has germinated, rapid pollen tube growth and earlier fertilization result in larger embryos and greater vascular supply, probably because the embryos have a longer time to develop. The relationship of earlier fertilization to larger 10-d embryos and seeds does not appear to explain the differences between maternal families and pollen donors in embryo and seed characteristics since Marshall and Diggle (2001) (Table 3) found no significant differences among the four donors in the proportion of ovules fertilized at any of the time intervals examined. Maternal families do differ in timing of fertilization (Marshall and Diggle, 2001); however, rank order among families at time of fertilization (3 > 1 > 2 = 4) does not correspond to their rank order of embryo area at 10 d [for embryo area (1 = 2 = 3 > 4)].
The size advantage at 10 d conferred by earlier fertilization did not persist to the end of embryo development (Figs 2 and and3).3). Marshall and Diggle (2001) found that pollen-tube growth rate and timing of fertilization also did not predict mature seed mass. Nakamura and Stanton (1989) also found that embryo size at 14 d was not a good predictor of mature seed size in R. sativus. In contrast to the lack of a persistent effect of embryo size, embryos that had achieved the most advanced stage of development (heart shaped) by 10 d, were associated with larger seed at 3 d and greater mature seed size. This association between the rate of morphogenesis and final seed size has not been reported previously.
All of the 10-d embryo and seed measures were smaller in the water-stressed plants (Table 5), indicating that maternal stress affects embryo development. The difference in embryo size at 10 d did not translate into smaller seeds at maturity; final seed mass was indistinguishable in the two treatments (table 9 in Marshall and Diggle, 2001). Water-stressed plants had fewer seeds per fruit, however, suggesting that stress may have resulted in embryo abortion. Embryo development also may have been slower for low-water plants as they had fewer heart-shaped embryos at 10 d (Table 1 and Fig. 4). Taken together, these data suggest that water stress resulted in slower embryo and seed development to a smaller size at day 10, differential abortion of the smallest seeds prior to maturation, and reallocation of resources such that fewer but not smaller seeds mature. A similar pattern of developmental delay/suppression has been observed in the bushtype bean (Phaseolus vulgaris; Yanez-Jimenez and Kohashi-Shibata, 1998), lentils (Lens culinaris; Shrestha et al., 2006), and soybeans (Glycine max; DeSouza et al., 1997). Developing seeds in water-stressed plants were likely to be smaller or to stop growing altogether.
The effects of pollen donor on developmental stage do not vary with stress treatment (Table 1). However, paternal effects on embryo area, seed area and width of the funicular vascular trace do change with stress (Table 2). This raises the possibility that maternal condition may affect the expression of paternally inherited alleles.
Correlations between the earlier events/stages (pollen germination through 3-d embryos) and mature seed mass with measurements of 10-d embryos and seeds are significantly affected by stress. Not only did the magnitude of correlation differ between the treatments but the direction changed from control to low-water plants in 15 out of 24 associations (Fig. 3). For example, embryo area was positively correlated with pollen germination in the low-water plants and negatively correlated in the control plants; i.e. in control plants delayed pollen germination is associated with larger embryos whereas under water stress rapid germination is associated with larger embryos (Table 4 and Fig. 3). This may be due to greater ability of non-stressed maternal plants to set up a selective environment for pollen-tube germination, leading to higher quality of embryos fertilized by those pollen tubes that eventually begin to grow. In water-stressed plants, in the absence of a selective filter at the stigma, rapid germination will be associated with early fertilization and more successful embryos.
Consideration of the features of 10-d embryos and seeds reported here, in combination with the fertilization dynamics and mature seed sizes reported in Marshall and Diggle (2001), shows that both maternal and paternal parents contribute significantly to variation in the processes of fertilization and development of embryo and seed (Fig. 5). The magnitude of the contributions of maternal family and pollen donor, however, varies among processes and changes over time. Maternal family explained more of the variance in most characteristics at each stage, yet pollen donor effects were stronger than maternal family effects on pollen tube length at 6 h and on area of 10-d embryos (Fig. 5). Differences among maternal plants in prezygotic processes such as pollen germination, pollen tube growth, and fertilization have been found in numerous species and have been taken as evidence for maternal sorting among mates (reviewed in Marshall and Folsom, 1991). Detection of both maternal and paternal contributions to variance in pollen tube length is consistent with the expression of phenotypic differences in growth rate among male gametophytes produced by the different donors, and differences among maternal plants in sorting among the donor phenotypes (Marshall and Folsom, 1991).
Maternal and paternal lines contribute more-or-less equally to variance in 10-d embryo area while the maternal family effect on 10-d seed area was much larger. Although the areas of 10-d embryos and seeds are correlated (Table 5), the contrast in the components of variance contributed by maternal and paternal parent to embryo versus seed size (Fig. 5) suggest that the two traits can be influenced independently and that expansion of seed size does not just passively keep pace with embryo growth. The greater maternal effects on seed area compared with embryo area at 10 d might be due to direct controls on cell cycle (cell division) in the developing seed coat (Adamski et al., 2009) or may be indirect due to physical constraints of ovary size (seed packing within the ovary) and/or initial ovule size.
In R. sativus, maternal plants clearly exert considerable control over embryo development; however, maternal contribution changes over the course of embryo ontogeny and there are windows in time when pollen donor effects emerge. For example, pollen donors differ in the proportion of embryos that reach the heart-shaped stage by 10 d and these more advanced embryos tend to be associated with larger (and potentially more successful) mature seeds. Natural selection can act at any stage of the life of an individual (e.g. Schlichting and Pigliucci, 1998), including during embryogenesis, and depending on the timing, selection will act on maternal and paternal effects differently. In natural populations of R. sativus, seeds within fruits often have different sires and may be genetically diverse (e.g. Marshall and Ellstrand, 1986). Furthermore, differential survival of developing embryos of R. sativus does occur: mean ovule number per fruit is 6·8, but only an average of 4·7 seeds per fruit reach maturity, and stress reduces embryo and seed survival still further (Marshall and Diggle, 2001). Although the maternal and paternal effects described here are statistical measures, they may reflect some underlying effects of alleles contributed by the maternal and paternal plants. In that case, maternal and paternal alleles will experience different selection depending upon the stage of development at which embryo death occurs. Differential levels of additive (and other components of) genetic variance during embryogenesis have been studied extensively in animals (e.g. Atchley and Rutledge, 1981; Cheverud et al., 1983, Atchley, 1984; Badyaev and Martin, 2000; Zelditch et al., 2004); and the potential for selection at these early stages of development to influence phenotypic evolution is generally acknowledged (e.g. Wolf et al., 2001; Poe, 2006). The present data show that examination of genetic variation and selection during embryogenesis might be equally interesting for plants.
In summary, the expected maternal effects on embryo development were found, but there is also evidence for an early paternal role. The relative effects of these controls may change over time. Thus, there may be times in development when selection on the maternal, paternal or embryo contributions to development are more and less likely. Additional studies of sequential effects on embryo development are needed.
The authors thank M. Folsom for developing methods for tissue preparation and microscopy, and collection of data. Initial financial support was provided by National Science Foundation Grant BSR 88-18552 to D.L.M. Support for manuscript preparation was provided by NSF DEB 03-44560 to P.K.D.