| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Biology, The Pennsylvania State University, University Park, Pennsylvania 16802
Received for publication December 2, 1997. Accepted for publication August 28, 1998.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: Campanula rapunculoides • Campanulaceae • female function • male function • nectar production • phenotypic plasticity • pollen production • pollen viability
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Robertson, Diaz, and MacNair, 1994
; Andersson, 1996
; Campbell, 1996
; Galen, 1996
). We also know that environmental effects can contribute to phenotypic variation in reproductive traits (Bradshaw, 1965
; Solomon, 1985
; Frazee and Marquis, 1994
), which in turn may alter both male and female reproductive success in angiosperms (reviewed in Lloyd 1980
, 1984
; Stephenson, 1981
, 1992
) . But the degree to which environmentally plastic responses in floral and reproductive traits are heritable has only recently attracted attention (Schlichting, 1989
; Mazer, 1992
; Andersson and Shaw, 1994
; Young et al., 1994
). This heritable component of plasticity, or Genotype by Environment interaction (G x E) is a population-wide measure of genotypic differences in the magnitude or direction of response over environments (Stearns, 1992
). Significant G x E interactions correspond to variations in reaction norms (the set of phenotypes produced by a genotype over a range of environments) and provide an indication of the potential for evolution of plasticity (de Jong, 1990
). Despite recent evidence of G x E interactions in some floral and reproductive traits of Phlox (Schlichting, 1989
), Raphanus (Mazer and Schick, 1991a
, b
) and Crepis (Andersson and Shaw, 1994
), it is not known how widespread G x E interactions are among the flowers of various species, or even which aspects of floral morphology and reproductive output (female and male) have greater potential for adaptive plasticity. For the flowers of iteroparous angiosperms, the ability to adapt to temporal heterogeneity may be especially important because environmental and/or developmental conditions experienced by sequential flowers produced on an individual are likely to vary over the course of a season and from one year to the next.
For plants that develop flowers sequentially along a raceme or branch, relative position may have a strong influence on some phenotypic traits, but not others. For example, Diggle (1994)
showed that the degree of hermaphroditism in Solanum hirtum flowers depends on position along a ramet, a condition she calls ontogenetic contingency. Moreover, ontogenic variation in floral traits may mask underlying genetic variation in plants with sequentially developing flowers (Stephenson, Devlin, and Horton, 1988
; Mazer and Delesalle, 1996b
), although this may not be true for all floral traits (e.g., Devlin and Stephenson, 1987
). If variation in the mating environment can select for differences between sequential flowers of an inflorescence (Møller and Eriksson, 1994
; Brunet and Charlesworth, 1995
), we cannot ignore the influence of ontogenetic and/or environmentally induced phenotypic plasticity in floral traits on the evolution of floral morphology.
In the present study, our primary focus was to examine phenotypic variation and plasticity in floral traits in Campanula rapunculoides, a perennial herb that produces multiple flowers on a raceme each year and can be easily cloned to provide replicate genets. Clonal species are ideal for the study of G x E interactions because replicates of all genotypes can be placed in each environment. Consequently, the error term in an Analysis of Variance (ANOVA) is unconfounded with genetic variation among sibs (Stearns, 1992
). Additionally, most measurements can be taken on multiple flowers in C. rapunculoides. Thus, we can use the relative position of a flower along a stalk (the raceme) as a covariate (ANCOVA) to further reduce error variance and we are able to partition phenotypic variation into genetic, environmental, and ontogenetic (i.e., position) sources as well as examine interactions among these factors.
A second focus of our study was to examine the plasticity of trait correlations across environments. Because individual flowers develop over relatively short time spans and are themselves part of a modular body plan, the opportunity exists for traits to compete for resources (Cruden, 1977
; Stephenson, 1981
; Charnov, 1982
; Devlin and Stephenson, 1987
; Stanton et al., 1991
; Ashman and Baker, 1992
; Mazer and Delesalle, 1996a
). Several studies have demonstrated that patterns of allocation to sexual structures can affect patterns of mating and reproductive success (e.g., Stanton et al., 1991
; Stephenson, 1992
; Brunet and Charlesworth, 1995
). Dickerson (1955)
first proposed that negative correlations between traits may constrain evolution. For example, both pollen production and ovule production may have a positive correlation with fitness, but the two traits may be negatively correlated as a result of a trade-off in resource allocation. Thus, to the degree that the environment affects the resources available for allocation to multiple resource sinks, environmental heterogeneity may alter the severity or presence of a resource trade-off.
Of particular interest to our study was the potential trade-off in male and female function. By growing clones of the same genotypes in three environments, we could determine whether the relationships between male and female traits are consistent across environments and thus integrated (Nicotra, Chazdon, and Schlichting, 1997
) or whether the correlation structure varies across environments. Integrated trait complexes may be expected in traits that are formed during a common developmental stage (Berg, 1960
), or in traits that are functionally related and under a common selective force (Scheiner, 1993
). Recently, Armbruster and Schwaegerle (1996)
showed that positive correlations can develop among genetically independent floral traits. However, phenotypic correlations may become uncoupled under certain environmental conditions as has been shown by Schlichting (1986
, 1989
), Schlichting and Pigliucci (1995b)
and Nicotra, Chazdon, and Schlichting (1997)
. Although several studies have examined the structure of correlations in plant traits across environments, few have included more than a couple of reproductive traits (Schlichting, 1989
; Stanton et al., 1991
; Young et al., 1994
; Pigliucci, Whitton, and Schlichting, 1995
) and even fewer have examined both the male and female functions (Mazer and Schick, 1991a
, b
). Yet, as Brunet and Charlesworth (1995)
point out, the factors that influence the pattern of reproductive resource allocation in plants are likely to be of real biological importance and deserve further study.
In this study, we use Campanula rapunculoides, an iteroparous perennial, to address the following questions: (1) To what extent are reproductive traits plastic? Which traits exhibit significant genetic variation as well as G x E interactions? (2) What effect does flower position (ontogenic effects) have on phenotypic variation? Which traits are most affected? (3) Do traits associated with male fitness (pollen number, pollen vigor) follow the same pattern of phenotypic variation as ovule number, a female trait? (4) Is there evidence for trade-offs between male and female traits? (5) Is phenotypic gender plastic? (6) Does the correlation structure change across environments (or inversely, are reproductive traits integrated)?
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). It overwinters as a rosette, and in July each rhizomatous cluster produces 1–8 flowering stalks (racemes) containing 20–30 blue, bell-shaped flowers, which open acropetally (bottom upwards). Members of the genus Campanula are protandrous: pollen is secondarily presented to stylar hairs (reviewed by Nyman, 1993
) while the stigmatic lobes remain closed (the male phase). Richardson and Stephenson (1989)
demonstrated that pollen removal shortens the duration of the male phase. After several days, during which pollen is removed by bees, particularly Bombus, the stigmatic lobes open and the flower becomes receptive to pollen (the female phase). When the stigmatic lobes first open, the flowers are self-incompatible, however, by the third or fourth day after the lobes open the flowers become partially self-compatible (Richardson et al., 1990
; Stephenson et al., 1992b
).
Experimental design
The plants used in this study were clonally replicated from rootstocks collected in the late 1980s from a natural population established along a shaded roadside near State College, Pennsylvania, USA. In September 1993, replicates in 15-cm pots (6 inch) of each genotype were placed in a 10°C coldroom for 5 wk. This cold treatment stimulates synchronous flowering when plants are returned to the greenhouse. Upon removal from the coldroom, replicate pots from each of seven genotypes were assigned to one of three environmentally controlled rooms in the Buckhout Greenhouse, University Park, Pennsylvania, USA. These three treatments were originally designed to mimic some of the diverse environmental conditions under which C. rapunculoides populations are found in Pennsylvania: full sun (hot), shaded and moist (cool), and crowded (potbound).
Plants in the hot and cool treatments were first subdivided to provide only enough belowground biomass to produce 1–3 flowering stalks per pot. All old soil was removed and replaced with fresh potting soil. The potbound plants were not subdivided, but old soil was replaced before placing the rootstock into a new (same sized) pot. After transplantation, 4–5 pots of each genotype were placed immediately in each of the three greenhouse rooms where they were watered daily and fertilized every other week with a half-strength Peters® (W. R. Grace & Co., Fogelsville, Pennsylvania, USA) solution. Insecticidal sprays were used occasionally as part of normal greenhouse management.
The hot environment was maintained at a near-constant 25°C. Natural light was augmented by sodium lamps in the afternoon to increase light intensity as well as to extend photoperiod to 13–14 h. The cool environment was a room normally devoted to orchid culture. Maintained at 
18°C, the room received only shaded natural light and day length was 10–11 h during the period when the plants were in flower. Because nearby orchids were frequently misted, the humidity was high. Plants in the potbound treatment were kept at 20–25°C, with natural light and day length (10–11 h). As a result of the greater density of flowering racemes in each pot, these potbound plants were frequently water stressed, although they were watered up to twice daily to prevent wilting.
Eleven floral traits were monitored in 96 plants (3 environments x 7 genotypes x 4–5 replicates), which flowered over the 6-wk bloom period (October and November 1993). Flower number per raceme and the floral duration (the number of days a raceme had open flowers) were recorded from two racemes per pot. The remaining traits were evaluated using approximately every fifth flower on one raceme per pot. For these traits, the position of the flower (lowest = 1) was recorded. Corolla size was determined by the length of one corolla lobe from every fifth flower on the raceme. This measurement was highly correlated with total corolla area (r = 0.94, data not shown) and was less destructive than corolla removal. To quantify the energy investment in nectar production per flower, we removed all nectar from a single flower (again, approximately every fifth flower) with a micropipet and washed the nectaries with a known volume (2 µL) of distilled water. This total volume was measured using a calibrated capillary tube. A sample of this diluted nectar was placed on a refractometer. A few nectar samples were too concentrated to be read accurately; these were again diluted with an equal volume of distilled water. Total sucrose equivalents were estimated using a standard curve of sucrose concentrations (mass per volume) read against the refractometer readings, adjusting for all dilutions and total volume of nectar removed for that sample.
Flowers that were to be used for the determination of the length of the male phase were labeled with a date tag on the first day the corolla opened. At this stage, the stigmatic lobes are tightly appressed and cannot receive pollen. Pollen was not collected, nor were those flowers manipulated, but they were observed daily until the stigmatic lobes separated, marking the end of the male phase. On other flowers, mature pollen was removed with a fresh toothpick and placed in a drop of acetocarmine stain (Marks, 1954
) on a microscope slide. Later, these slides were evaluated for viability of the pollen sample by counting the number of stained (viable) pollen grains in a sample of 300 grains read across the center of the slide. A second sample was taken, usually from the same flower, and spread over a petri dish of 1.5% agar germination media (modified from Brewbaker and Kwack [1963
]) with sucrose adjusted to 11% for optimal germination and growing conditions of pollen tubes. This pollen sample was incubated for 4 h in a lighted growth chamber at 25°C, after which growth was stopped with 70% ethanol. The incubation time was a compromise between maximal germination time (4–6 h) and sustainable in vitro growth of the first-emerging pollen tubes (3–5 h). The percentage of germinated pollen grains at 4 h was calculated from a sample of 300 grains. Both pollen viability and pollen germination percentage data were arcsine transformed (Neter, Wasserman, and Kutner, 1990
) prior to analysis. The lengths of pollen tubes were measured by image analysis using Image® software (Rich, Ranken and George, 1989
). Thirty to 50 tubes per flower were measured in grid sections across the diameter of the plate, but only the averages were used for statistical analysis.
Some flowers (again, approximately every fifth flower) were sacrificed for the determination of pollen grain size, number of pollen grains per flower, and number of ovules per flower. Because pollen is shed prior to anthesis and deposited onto stylar hairs in Campanula (Nyman, 1993
), we obtained the full production of pollen per flower by collecting the whole style at anthesis. These styles were placed in separate glass vials and dried in a 40°C oven for 4 d. Later, this pollen sample was rehydrated with a 0.5% NaCl solution, sonicated to separate pollen grains from the styles, and counted on a particle counter (Elzone 180 with EX 1010 software; Micromeritics Inc., Norcross, Georgia, USA). To determine the number of ovules per flower whole ovaries were placed in individual eppendorf tubes with 70% ethanol. Later, the ovules were removed by teasing apart the ovary under a dissecting microscope and counted. Ovule size was not determined but appeared to be remarkably uniform.
Statistical analysis
Univariate Analyses of Variance (ANOVA) were conducted for each of the 11 traits using a two factor mixed-effects model (PROC GLM; SAS, 1990
). Environment was a fixed effect, while genotype and G x E were random. F tests were constructed under an "unrestricted mixed effects model" (i.e., the Scheffé model of Fry [1992
]). Components of the model and interpretations are given in Table 1. For traits measured on individual flowers, we used analysis of covariance with the flower position as a covariate. Because the development of the flowers on each plant took place within a single environmentally controlled room, significant covariate effects were interpreted as ontogenetic variation and unconfounded with environmental variation.
|
), although whenever an interaction exists (G x E) it means that the expression of a trait is dependant on the environment in which it is measured In other words, when there is a significant G x E interaction, heritability estimates will vary across environments. Similarly, we interpret VE as a population-wide estimate of variance in phenotypic response to changes in environmental conditions (Table 1), but when a G x E interaction is significant in the model, estimates of VE become dependant on the specific genotypes represented. Although the number of genotypes in our study is small (seven), we treat them as a random variable and our test for significant environmental main effects is conservative. Consequently, a large G x E variance will reduce our ability to detect significant VE because the VG x E is the denominator in the mean squares F test in the unrestricted model (Fry, 1992
). In short, a significant G x E interaction means that estimates of broad-sense heritability are environment dependent and/or the shape of the reaction norm is dependent on the specific genotypes tested.
The variance due to interactions of genotype and environment (VG x E) is interpreted as the heritable portion of plasticity (Table 1). Because the three environments are discrete (see Pigliucci, Schlichting, and Whitton, 1995
), and all genotypes (i.e., clones) are represented nearly equally in all environments, this model is highly sensitive to genotypic-specific responses in environmental conditions. In addition, several clonal replicates were available within each genotype x treatment combination, thus the estimation of all variance components are made with fewer confounding assumptions than are normally possible with animal species or annual plants (Stearns, 1992
).
In addition to estimates of variance components from Type III sums of squares (Table 1), we also obtained raw (untransformed) least square means of phenotypic values for each genotype by treatment combination for our depictions of reaction norms. We also used the least square means for pollen number and ovule number and calculated the phenotypic gender of each plant separately for each of the three environments for all 21 combinations of genotypes and environments (7 x 3). To calculate phenotypic gender, we used a modification of Lloyd's (1980)
equation:Gijk = dijk/[dijk + E(aijk)]where Gijk is the phenotypic gender of plant k of genotype i in environment j, dijk is the average number of ovules per flower of plant k of genotype i in environment j, aijk is the average number of pollen grains per flower of plant k of genotype i in environment j, and E is an equivalence factor made up of the ratio of the experiment-wide mean of ovules per flower (
= 162.3) divided by the experiment-wide mean of pollen number per flower ( = 19586). Thus gender can vary between 1 (completely female) and 0 (completely male), e.g., G42. = 0.4 indicates that the average plant of genotype 4 in environment 2 was slightly male, relative to all other plants across all environments. Phenotypic gender was also depicted as a reaction norm using the gender means of all 21 combinations of genotype and environment and variance among replicate plants within that genotype and environment combination.
Pearson product-moment correlations were calculated among the nine variable traits (traits for which the variation was explained by one or more factors) for each environment separately using PROC CORR (SAS, 1990
). Because position was significant for only one trait, we simplified this analysis by using phenotypic values (least square means) for each trait on each raceme averaged over 3–5 flowers on that raceme. These correlations are broad-sense genotypic correlations (Roff, 1995
) because each raceme is of a single genotype grown under a single environmental condition. Because all three correlation matrices use the same genotypes with approximately the same replications (i.e., 3–5 racemes per genotype per environment), differences in the strength or direction of these correlations across environments are considered to be a form of G x E interaction for trait correlations. Significance levels of the 108 correlations were adjusted for multiple inference in two ways: first, the level of significance for all tests was set at P <0.01 to provide a more stringent test of Ho: r = 0. In addition, we used a sequential Bonferroni (Rice, 1989
) and indicate significant correlations at an environment-wide alpha of 0.05 (using k = 36). While the sequential Bonferroni method is conservative for Type I error, it also greatly reduces power relative to the first method. Both methods are presented and we take the view that these two levels of significance represent upper and lower bounds for reasonable significance levels with multiple comparisons in an exploratory analysis. Significant correlations were examined for patterns of sign (positive or negative) across environments using 
2 tests. We also examined differences in sign among traits of the same developmental module (i.e., within the four pollen traits or within the three flower traits of corolla size, nectar content, and duration of male phase) vs. trait pairs representing structures of different development modules (e.g., the correlation between flower number and ovule number).
We further examined the consistancy of correlation patterns across pairs of environments using the integration analysis of Nicotra, Chazdon, and Schlichting (1997)
. Here we used environment pairs (e.g., hot and cool environments) and determined the number of significant correlations (P < 0.01) that did not change sign or vary by more than 0.50 that were common to both environments. That number, divided by the total number of correlations examined (N = 36), yields the level of integration between those two environments expressed as a percentage.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
Highly significant G x E interactions (P < 0.01) were detected in traits representing all four developmental modules: at the level of whole racemes (number of flowers per raceme), individual flowers (corolla size, nectar content, duration of the male phase), pollen traits (pollen viability), and female function (number of ovules per flower) (Table 2). Although these G x E interactions contributed only a small amount to the total phenotypic variance (range 0.05–0.14), this amount was within the range of variance due to environmental main effects (VE) for the same traits. One exception was that of pollen viability, which had a highly significant G x E interaction, but the environmental main effect was not significant (Table 2). This implies no population-wide response among environments for this trait, although one or more genotypes responded in some distinctive manner to environmental change.
The floral position (a covariate) was highly significant for ovule number per flower (F test, P <0.0001; Table 2). Ovule production per flower was inversely related to position with lower flowers yielding 25% more ovules than upper flowers (Fig. 1). The effect of floral position on other traits, including pollen number (Fig. 1) was not statistically significant (Table 2).
|
|
; Kang and Gauch, 1996
). Fan-shaped responses across two environments indicate variation in the magnitude of the response (quantity differences) as exemplified by corolla size and pollen tube growth in the cool vs. potbound environments (Fig. 2). Changes in the rank order of genotypes among environments (quality differences) indicate variation in the direction of the response. The most extreme qualitative differences are exemplified by genotypes 1 and 9 for nectar production (Fig. 2) and duration of the male phase (Fig. 2)—the response of these genotypes is in a direction opposite to the mean response of the other genotypes. The reaction norm for gender (Fig. 3) as well as the reaction norms for pollen number and ovule number (Fig. 2) reveal that genotypes 2, 7, and 9 are female biased, and genotypes 1, 5, 13 and 14 are relatively more male biased or closer to phenotypic hermaphrodites. Gender is phenotypically plastic: all plants became more male (gender shifts towards 0) in the potbound environment (Fig. 3) relative to their phenotypic gender in the other two environments. The range of phenotypic gender in the hot environment was 0.41–0.78 while for the same group of genotypes in the potbound environment, the range was 0.35–0.66. Notably, the reaction norm slopes of the three female biased genotypes between the hot and potbound environments appears to be steeper than the slopes of the males or hermaphrodites (Fig. 3).
|
2 = 7.4, 1 df, P = 0.006). The greater number of positive correlations in the cool environment are also reflected when considering only those correlations which are statistically significant at the P < 0.01 level (i.e., 8:3 for hot, 9:1 for cool, and 6:2 for potbound environments).
|
A third pattern observed in these correlations is that the correlation structure is somewhat dependent on the environment. Of the significant correlations (P < 0.01) only two correlations are significant in all three environments: the positive correlation between pollen number and corolla size and the negative correlation between pollen number and ovule number (Table 3). The remaining 6–9 significant correlations in each environment are not evident in all environments. In other words, the level of overall integration (see Nicotra, Chazdon, and Schlichting, 1997
) is 5% (i.e., 2/36 correlation pairs were significant and did not vary > 0.50 between environments). The level of integration between any two environments ranged from 8 to 16%. Between the hot and cool environments, six significant correlation pairs were shared (16.3% integration), while five correlation pairs were shared between the hot and potbound environments (13.9% integration) and three between the cool and potbound environments (8.3% integration).
The fourth, and perhaps most interesting pattern is that of significant negative correlations between ovule number and male traits, including pollen traits and duration of the male phase. Negative correlations between ovule and pollen numbers were among the most consistent relationships between any pair of traits in the correlation matrices (hot: r = -0.448, cool: r = -0.499, potbound: r = -0.480, Table 3). Most of the pollen traits were positively correlated with each other and negatively correlated with ovule number (Table 3). Negative correlations between ovule number and duration of the male phase were larger in the hot environment (r = -0.779), than in the cool (r = -0.377) or potbound (r = -0.414).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Stearns, 1992
). Yet our data agree with the findings of several other studies (Mazer and Schick, 1991a
; Andersson and Widen, 1993
; Conner and Via, 1993
; Campbell, 1996
; Robertson, Diaz, and MacNair, 1994
; Young et al., 1994
; Galen, 1996
) that have detected moderate levels of heritability in floral traits. In Campanula rapunculoides, the majority of traits under study (nine out of 11) had significant (P < 0.01) genotypic variation, and in these traits heritability estimates ranged from 0.21 to 0.38 (Table 2). Moreover, significant genotypic variation was detected in traits representing all levels of functional organization within floral structures, male function (pollen traits) and female function (ovules), and at the level of whole flowers (corolla size, nectar) and whole plants (flowers per inflorescence).
What maintains this genotypic variation? Clearly, it is unlikely that floral traits are neutral with respect to selection nor is it likely that selection is weak. In fact, Galen (1996)
demonstrated that in Polemonium the size of the corolla can respond to directional selection in as few as two generations under natural conditions. Similarly, selection can operate on phenotypic variation in pollen vigor in a single generation (see Quesada, Winsor, and Stephenson, 1993
, 1996
; Jóhannsson and Stephenson, 1997
). Perhaps the simplest explanation for the maintenance of genetic diversity of traits under selection is that selection fluctuates across environments: phenotypes favored in one environment are disfavored in another, and environments vary periodically (Via and Lande, 1985
; Falconer, 1989
; Stearns, 1992
; Whitlock, 1996
; but see Sasaki and Ellner, 1997
). Several workers have enlisted this argument to explain the high heritability so often observed in floral morphology (Young et al., 1994
; Galloway, 1995
; Wilson and Thomson, 1996
). At the extreme, the environment may be so unpredictable that selection leads to a stable "bet-hedging strategy," as has been documented for male function in plants of Aralia hispida (Thomson and Barrett, 1981
). If, however, environments vary predictably (as in the changing resource conditions as a plant ages) or stochastically within some range of temperature or light, plants with appropriate plasticity may have a fitness advantage (Bradshaw, 1965
; Schlichting and Pigliucci, 1995a
). This view has been recently supported by two studies that have demonstrated adaptive plasticity in plant height in Nicotiana and Brassica (Schmitt, McCormac, and Smith, 1995
) and Impatiens (Dudley and Schmitt, 1996
). In both studies, plants that produced a "normal" phenotype in response to sun vs. shade environments had greater fitness than those that had been induced to produce the phenotype "normally" associated with the alternate environment.
In our study, pollen traits exhibited no environmental or developmental plasticity and only small amounts of G x E interactions (Table 2), thus we conclude that reaction norms for pollen size, number, and vigor have little potential to evolve and that pollen traits are generally buffered against environmental or ontogenetic influences in this species. Stable allocations to male function traits have been found by Willis (1993)
for pollen size, Wolfe (1992)
for corolla size, and Stanton et al. (1991)
for pollen number and size. Recently, Conner et al. (1996)
found that selection acted on female reproductive success in Raphanus, but not male reproductive success. In other species, however, pollen size and/or pollen vigor are known to vary with environmental conditions, such as soil fertility, leaf herbivory, temperature and mycorrhizal infection, and the physiological status of the plant (see reviews by Stephenson et al., 1992a
; Delph, Jóhannsson, and Stephenson, 1997
). Campanula rapunculoides appears to respond to different environmental conditions by altering flower number, flower size, ovule number, etc., but maintaining consistent pollen number and pollen size.
An interesting exception to the relative stability of pollen traits was pollen viability. Crossed reaction norms (Fig. 2) show that some genotypes have highest pollen viability in the hot environment and different genotypes have highest viability in the potbound environment. Other studies using both cultivated and wild species (e.g., Zamir and Gadish, 1987
; Jóhannsson and Stephenson, 1998
) have also found pronounced effects of developmental temperature on pollen viability. If selection acts on pollen viability (see Delph, Jóhannsson, and Stephenson, 1997
), then some amount of phenotypic variation could be maintained in this trait by environmental conditions that vary within or between years. Yet this cannot entirely explain the substantial genotypic variation in pollen viability (35% of total phenotypic variance; Table 2) given that only a few genotypes had rank order changes across environments (Fig. 2). Other factors, perhaps acting at later developmental stages such as specific male-female interactions (Schlichting and Devlin, 1989
; Snow and Spira, 1991
), may influence male success and thus maintain genetic variation in pollen viability measured at anthesis.
In contrast to our finding of low phenotypic plasticity in pollen traits, most of the remaining floral traits with significant genetic variation also had phenotypic plasticity, which included a significant G x E interaction (P < 0.05 for flower number, P < 0.001 for corolla size, nectar quality, duration of the male phase, and ovule number per flower for both G x E and environmental variances; Table 2). Our finding of widespread and highly significant G x E interactions in floral traits of C. rapunculoides differs from other studies of G x E in floral traits that have shown only few (Mazer and Schick, 1991a
; Andersson and Shaw, 1994
; Schlichting and Pigliucci, 1995b
) or no G x E interactions involving floral traits (Andersson and Widen, 1993
; Young et al., 1994
; Armbruster and Schwaegerle, 1996
). Because the very definition of a G x E interaction underscores its dependence on particular environments and genotypes, it is difficult to determine how universal, or important, G x E interactions are in the maintenance of genetic variation or in the evolution of floral morphology. For example, our study with C. rapunculoides reveals that male traits are less plastic and have fewer G x E interactions than other floral traits, but many more studies with other species are needed before we can suggest that this is a general characteristic of male function in perennials. Our results are consistent with the findings of Mazer (1992)
with Raphanus, which indicate that traits associated with male reproductive success are less sensitive to environmental conditions than traits associated with female reproductive success.
Some comments on the detection of G x E interaction
Three factors may have increased our likelihood of being able to detect G x E interactions: First, we used three, discrete environmental conditions. Studies that rely on the genetic correlations of traits over two environments or that use only a single environmental variable (e.g., density or nutrients) may not provide sufficient environmental variation for plasticity to be revealed (Gillespie and Turelli, 1989
). Unfortunately, experiments in discrete environments cannot elucidate the mechanisms behind phenotypic plasticity (Pigliucci, 1996
), but they may be useful in addressing the questions regarding which traits are most affected by G x E interactions.
Second, we used an iteroparous perennial with a relatively long reproductive cycle (6–8 wk from bolting to final flower senescence) in which flower production and fruit maturation overlap on the same raceme. By using a species that frequently encounters variations in temperature, light, nutrients, and crowding, both within and between reproductive cycles, we may have been more likely to find traits that were environmentally plastic to these conditions (de Jong, 1990
). It remains to be seen whether iteroparous perennials have more G x E interactions in floral traits than annuals or semelparous perennials. Finally, our use of clonal replicates within environments and multiple measures of floral traits per plant provided an estimate of experimental error that was unconfounded with genetic variance among replicates (Stearns, 1992
) or ontogenic factors (Mazer and Delesalle, 1996b
). This enabled us to detect significant G x E interactions even when they were only a small fraction of the overall phenotypic variation (e.g., pollen viability G x E accounted for only 5% of total phenotypic variation). At this point we do not know whether this amount of variation is ecologically significant (see Via et al., 1995
). Future experimental efforts to quantify G x E interactions in natural plants may benefit from statistical methods available in the agricultural literature (reviewed by Kang and Gauch, 1996
) that partition G x E interactions to allow statistical tests for the magnitude of rank order changes (e.g., Wu and Stettler, 1997
) or to identify which genotypes show the highest plasticity across environments (i.e., stability analysis; Pritts and Luby, 1990
). While the statistical approaches of the horticulturalists often require balanced designs and many (>10) environments (see Kang and Gauch, 1996
), their methods can be readily applied to any plant species, cultivated or wild, that can be clonally replicated.
Integration of floral traits
Selection acts not on isolated traits, but on phenotypes comprising suites of characteristics (Berg, 1960
). On the one hand, selection on functionally related traits such as floral traits can create positive correlations among traits that are under independent genetic control (Armbruster and Schwaegerle, 1996
). On the other hand, negative correlations may constrain evolution and thus maintain considerable genetic variation in fitness traits (Via and Lande, 1985
). But are these responses integrated such that the correlation pattern between traits in one environment is repeated in other environments? While some have argued that correlation patterns are important to preserve functional integration of trait complexes (Cheverud, 1996
; Conner and Via, 1993
; Scheiner, 1993
), Schlichting and Pigliucci (1995b)
suggest instead that plant responses to environmental change are unlikely to be adequate if they are only proportional. Their study (Schlichting and Pigliucci, 1995b
) as well as another by Waitt and Levin (1993)
demonstrated that both the magnitude and the direction of correlations among floral traits of Phlox were, in fact, environment dependent.
Most of the significant phenotypic correlations between traits in our study were positive and moderately consistent across environments, particularly those in the same functional unit (e.g., the five traits associated with male function; Table 3). Negative correlations were less frequent and were largely restricted to pairs of traits associated with different functional units (e.g., any of the male function traits vs. ovule number, or between flower number and ovule number). The degree of integration in the correlation architecture across all three environments was modest (5%), suggesting that correlation structure depends on the environment. The level of integration for any two environments was higher, particularly between the hot and cool environments where 18% of the correlations were shared. This level of integration is comparable to the 15–20% level of integration found in growth and photosynthetic traits of Piper spp. by Nicotra, Chazdon, and Schlichting (1997)
.
One trait, pollen viability, showed significant correlations with several traits that either changed sign or became nonsignificant across environments. Notably, pollen viability was also the only trait that had a highly significant G x E interaction in the ANOVA (P < 0.001) without a significant environmental variance (P = 0.44, NS). This situation is depicted by reaction norms that cross, such that the average environmental response is flat, but some genotypes exhibit significant phenotypic plasticity across environments (Fig. 2). We suggest that the G x E interaction for pollen viability may be an indication of environment-dependent correlations with other traits that resulted in differences in provisioning of pollen grains in the different genotypes. Given some variation in resources available to these genotypes over their life history, genotypic variation in pollen viability could be maintained (1) by selection favoring different genotypes in different years or different times in the growing season (e.g., Galloway, 1995
) and/or (2) as a consequence of a correlation of pollen viability and other traits that are phenotypically plastic such as flower number or corolla size (e.g., Campbell, Waser, and Price, 1994
). There is no reason to expect that only one mechanism (differential selection vs. correlation with other traits) serves to maintain genetic diversity in floral traits. Yet the ability to sort out the effects of interacting factors may be important to understanding the forces driving floral evolution, however technically difficult. The confounding effects of multiple factors will especially obscure relationships among trait pairs with negative correlations in some (or all) environments (Via and Lande, 1985
).
Consistency vs. plasticity of gender
The phenotypic gender of individual hermaphrodite plants in a population is frequently a quantitative trait that varies with the genotype, the environment, and the physiological status of the plant (Lloyd, 1980
; Devlin and Stephenson, 1987
; Thomson and Barrett, 1981
; Winsor and Stephenson, 1987
). On the surface, floral gender in C. rapunculoides appears to be fixed. The diagrams of reaction norms (Fig. 2) reveal that those genotypes that have the lowest pollen number per flower (genotypes 9, 7, and 2, respectively) also have among the highest ovule numbers per flower, and this trend is rather consistent across the three environments. The differences in pollen and ovule numbers per flower among individuals result in consistent negative correlations between pollen number and ovule numbers in each of the three environments (Pearson correlations r = -0.45, -0.50, and -0.48; Table 3). Both ovule and pollen number are significantly influenced by genotype, and ovule number varies across environments and position on the raceme (Table 2; Fig. 1).
As a consequence of the variation in ovule production, individual flowers of C. rapunculoides become phenotypically more male towards the tip of the raceme (Fig. 1) and the gender of plants varies both within and across environments (Fig. 3). The trend of increasing maleness in later flowers observed in other species has usually been attributed to diminished resources available to terminal flowers (Lobelia cardinalis, Devlin and Stephenson, 1987
; Aquilegia caerulea, Brunet and Charlesworth, 1995
; and Solanum spp., Solomon, 1985
; Diggle 1994
). A common, but increasingly questioned assumption of sex allocation theory (reviewed in Charnov, 1982
) is that increased investment in one gametic type necessarily incurs a cost with respect to the allocation of resources to the alternate gametic type (see Goldman and Willson, 1986
; Devlin and Stephenson, 1987
; Mazer, 1992
). The variability in ovule numbers and the constancy of pollen production for all floral positions found in this study suggest that, at the level of individual flowers, investments to female function vary with the resource base, while investments to male function are held constant. These results, however, are limited to ovule production and not the full investment to female function (i.e., seeds produced), most of which occurs postzygotically (Schlichting and Devlin, 1989
; Stephenson, 1992
).
The results of our study do, however, give support to another commonly held prediction of sex allocation theory, viz. that allocation to the least costly gamete (generally considered to be pollen) will be favored under stressful conditions (Charnov, 1982
). The reaction norms for gender based on pollen and ovule numbers averaged for individual racemes (Fig. 3) reveal a shift towards greater relative expression of maleness (i.e., decreased phenotypic femaleness) in the resource-constrained potbound environment relative to the expression of gender in the cool and hot environments. If gender modification is to respond appropriately to changing environmental conditions, a population must have heritable genetic variation for the traits that comprise gender expression as well as heritable variation in the plasticity of those traits (Goldman and Willson, 1986
; Mazer, 1992
). By this standard, our study demonstrates that C. rapunculoides has the potential for the evolution of plasticity of reproductive traits, and consequently of gender itself.
In conclusion, C. rapunculoides alters floral traits and traits related to the female function in response to environmental conditions, while maintaining consistency in traits related to the male function (pollen number per flower, pollen size, pollen germinability, and pollen tube growth). Likewise, ovule production decreases as a function of flower position on the raceme, while other floral traits, including traits related to male function, remain constant. This study reveals significant genetic determination of both pollen and ovule production (and seven other floral traits) and evidence of a genetic trade-off between male and female function among genotypes. Most important, our results show a significant genotype x environment interaction for ovule production as well as for four other floral traits and pollen viability. Thus, even within this small sample of individuals from one population, we have found variation within the norms of reaction for several traits associated with reproductive success. Additional studies are necessary to determine whether this plasticity is adaptive and whether selection can act upon it under natural conditions of variable pollinator activity and changing resources.
|
FOOTNOTES |
|---|
2 Current address: Department of Biological Sciences, University of Pittsburgh, Pittsburgh, Pennsylvania.
3 Author for correspondence (FAX: 814-863-1553, Email: as4{at}psu.edu
).
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
———, and R. G. Shaw. 1994 Phenotypic plasticity in Crepis tectorum (Asteraceae): genetic correlations across light regimens. Heredity 72: 113–125.[ISI]
———, and B. Widen. 1993 Reaction norm variation in a rare plant, Senecio integrifolius (Asteraceae). Heredity 73: 598–607.[ISI]
Armbruster, W. S., and K. E. Schwaegerle. 1996 Causes of covariation of phenotypic traits among populations. Journal of Evolutionary Biology 9: 261–276.[CrossRef][ISI]
Ashman, T-L., and I. Baker. 1992 Variation in floral sex allocation with time of season and currency. Ecology 73: 1237–1243.[CrossRef][ISI]
Berg, R. L. 1960 The ecological significance of correlation pleiades. Evolution 14: 171–180.[CrossRef][ISI]
Bradshaw, A. D. 1965 Evolutionary significance of phenotypic plasticity in plants. Advances in Genetics 13: 115–155.
Brewbaker, J. L., And B. H. Kwack. 1963 The essential role of calcium ion in pollen germination and pollen tube growth. American Journal of Botany 50:747–758.
Brunet, J., and D. H. Charlesworth. 1995 Floral sex allocations in sequentially blooming plants. Evolution 49: 70–79.[CrossRef][ISI]
Campbell, D. R. 1996 Evolution of floral traits in a hermpahorditic plant: Field measurements of heritabilities and genetic correlations. Evolution 50: 1442–1453.[CrossRef][ISI]
———, N. M. Waser, And M. V. Price. 1994 Indirect selection of stigma position in Ipomopsis aggregata via a genetically correlated trait. Evolution 48: 55–68.[CrossRef][ISI]
Carr, D. E., and C. B. Fenster. 1994 Levels of genetic variation and covariation for Mimulus (Scrophulariaceae) floral traits. Heredity 72:606–618.
Charnov, E. L. 1982 The theory of sex allocation. Princeton University Press, Princeton, NJ.
Cheverud, J. M. 1996 Developmental integration and the evolution of pleiotrophy. American Zoologist 36: 44–56.
Conner, J., S. Rush, S. Kercher, and P. Jennetten. 1996 Measurements of natural selection on floral traits in wild radish (Raphanus raphanistrum). II. Selection through lifetime male and total fitness. Evolution 50: 1137–1146.[CrossRef][ISI]
———, and S. Via. 1993 Patterns of phenotypic and genetic correlations among morphological and life-history traits in wild radish, Raphanus raphanistrum. Evolution 47: 704–711.
Cruden, R. W. 1977 Pollen-ovule ratios: a conservative indicator of breeding system in flowering plants. Evolution 31: 32–46.
De Jong, G. 1990 Quantitative genetics of reaction norms. Journal of Evolutionary Biology 3: 447–468.
Delph, L. F., M. H. Jóhannsson, and A. G. Stephenson. 1997 How environmental factors affect pollen performance: ecological and evolutionary perspectives. Ecology 78:1632–1639.
Devlin, B., and A. G. Stephenson. 1987 Sexual variations among plants of a perfect-flowered species. American Naturalist 130:199–218.
Dickerson, G. E. 1955 Genetic slippage in response to selection for multiple objectives. Cold Spring Harbor Symposium on Quantitative Biology 20: 213–224.[ISI][Medline]
Diggle, P. K. 1994 The expression of andromonoecy in Solanum hirtum (Solanaceae): phenotypic plasticity and ontogenetic contingency. American Journal of Botany 81: 1354–1365.[CrossRef][ISI]
Dudley, S. A., and J. Schmitt. 1996 Testing the adaptive plasticity hypothesis: Density-dependent selection on manipulated stem length in Impatiens capensis. American Naturalist 147: 445–465.
Falconer, D. S. 1989 Introduction to quantitative genetics, 3rd ed. Longman Scientific & Technical, Essex.
Frazee, J. E., And R. J. Marquis. 1994 Environmental contribution to floral trait variation in Chamaecrista fasciculata (Fabaceae: Caesalpinoideae). American Journal of Botany 81: 206–215.[CrossRef][ISI]
Fry, J. D. 1992 The mixed-model analysis of variance applied to quantitative genetics: biological meaning of the parameters. Evolution 46: 540–550.[CrossRef][ISI]
Galen, C. 1996 Rates of floral evolution: adaptation to bumblebee pollination in an alpine wildflower, Polemonium viscosum. Evolution 50: 120–125.
Galloway, L. F. 1995 Response to natural environmental heterogeneity: maternal effects and selection on life-history characters and plasticities in Mimulus guttatus. Evolution 49: 1095–1107.
Gillespie, J. H., and M. Turelli. 1989 Genotype-environment interactions and the maintenance of polygenic variation. Genetics 121: 129–138.
Goldman, D. A., and M. F. Willson. 1986 Sex allocation in functionally hermaphroditic plants: a review and critique. Botanical Review 2: 157–194.
Jóhannsson, M. H., and A. G. Stephenson. 1997 Effects of pollination intensity on the vigor of the sporophytic and gametophytic generations of Cucurbita texana. Sexual Plant Reproduction 10:236–140.
———, and ———. 1998 Effects of temperature stress during microsporogenesis on pollen performance in a wild and cultivated Cucurbita pepo L. (Cucurbitaceae). International Journal of Plant Science 159: 616–626.[CrossRef]
Kang, M. S., and H. G. Gauch, Jr. [eds.]. 1996 Genotype-by-environment interaction. CRC Press, Boca Raton, FL.
Lloyd, D. G. 1980 Sexual strategies in plants. III. A quantitative method for describing the gender of plants. New Zealand Journal of Botany 18: 103–108.
———. 1984 Gender allocations in outcrossing cosexual plants. In R. Dirzo and J. Sarukhan [eds.], Perspectives on plant population ecology, 277–300, Sinauer, Sunderland, MA.
