| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
0 Department of Ecology and Evolutionary Biology, Corson Hall, Cornell University, Ithaca, New York 14853 USA
Received for publication August 17, 1999. Accepted for publication January 7, 2000.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: Clarkia • development • growth rate • heterchrony • outcrossing • progenesis • selfing • trait covariation
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Jain, 1976
; Lloyd, 1979
; Wyatt, 1983
; Lande and Schemske, 1985
; Wyatt, 1988
; Diggle, 1992
). Flowers of selfing taxa typically have smaller corollas, reduced herkogamy (spatial separation between mature anthers and stigma), and reduced dichogamy (temporal separation between anther dehiscence and stigma receptivity) compared to flowers of related outcrossing taxa (Ornduff, 1969
; Wyatt, 1983
). Reduced herkogamy and dichogamy facilitate self-pollination because mature pollen is brought into close contact with the receptive stigma (Schoen, 1982
; Gottlieb, 1984
; Holtsford and Ellstrand, 1989
). Small corolla size is thought to evolve secondarily in selfing taxa as a means of limiting costly expenditures on pollinator attraction.
Related outcrossing and self-pollinating taxa also often differ in life history and ecology (Ornduff, 1969
; Hill, Lord, and Shaw, 1992
). For example, selfing taxa commonly reach reproductive maturity at an earlier age and at a smaller vegetative size compared to related outcrossers (Stebbins, 1974
; Schoen, 1982
; Wyatt, 1983
; Holtsford and Ellstrand, 1992
), and they occur in habitats at the geographic or ecological limit of the outcrosser's range (Stebbins, 1950
; Baker, 1955
; Lewis and Lewis, 1955
; Vasek, 1964, 1968
; Lloyd, 1965
; Solbrig and Rollins, 1977
; Schoen, 1982
; Wyatt, 1988
). One hypothesis for the evolution of self-pollination in geographically or ecologically marginal environments is the greater reproductive assurance afforded self-pollinating plants in habitats where pollinators may be scarce (Stebbins, 1950
; Baker, 1955
). An alternative hypothesis is that small flower size and self-fertilization arise as by-products of selection for rapid maturation in marginal environments (Arroyo, 1973
; Guerrant, 1989
). In marginal environments, selection may favor individuals that can complete reproduction early, in advance of impending stress (Arroyo, 1973
; Gould, 1977
; Guerrant, 1989
; Eckhart, Geber, and Jonas, 1996
). One way of achieving early reproduction is through a decrease in developmental time, both of individual organs (e.g., flower organs) and of whole organisms (Gould, 1977
). In flowers, shorter organ development time is likely to reduce herkogamy. In addition, in the absence of other developmental changes, a decrease in development time results in smaller floral organs and hence smaller flowers. Under this second hypothesis, the characteristic features of the self-pollinating flower evolve as a by-product of selection for early maturation, and the evolution of self-pollination follows from the changes in floral form (Arroyo, 1973
; Guerrant, 1989
). The rapid maturation hypothesis can also explain the differences in flowering time and whole-plant size that often distinguish related self-pollinating and outcrossing taxa.
Ancestor-descendant changes in mature size and form, such as those that distinguish the flowers of self-pollinating and outcrossing taxa, often arise through changes in the timing or rate of development, i.e., through heterochrony (deBeer, 1958
; Gould, 1977
; Alberch et al., 1979
; Lord and Hill, 1987
; Guerrant, 1988
; Poethig, 1988
; Lord, Eckard, and Crone, 1989
; Kellogg, 1990
; Hill, Lord, and Shaw, 1992
; Gallardo, Dominguez, and Muñoz, 1993
; Sherry, 1994
; Itoh et al., 1998
; Klingenberg, 1998
). A reduction in the duration of organ development is a form of heterochrony that results in progenesis, i.e., in the early maturation of organs at a smaller size and juvenilized form. If self-pollination evolves from outcrossing as a result of selection for rapid maturation, then the selfing flower should be a progenetic form of the outcrossing flower. On the other hand, the reproductive assurance hypothesis for the evolution of self-pollination does not make any specific prediction about the nature of the developmental differences between flowers of related outcrossing and self-pollinating taxa.
In this study, we compare floral organ size and development between the self-pollinating subspecies Clarkia xantiana ssp. parviflora and its outcrossing progenitor C. x. ssp. xantiana to address whether progenesis accounts for subspecific differences in flower size and form. Clarkia xantiana ssp. parviflora inhabits marginal (arid) environments and flowers earlier and at a smaller plant size than its outcrossing progenitor (Eckhart and Geber, 2000
). It also displays floral characters that are typical of the selfing habit, namely small petals and reduced herkogamy and protandry. Lastly, its mature flowers resemble juvenile stages of the flower buds of C. x. ssp. xantiana in several respects. For example, close proximity of anthers and stigma is a characteristic of selfing flowers at anthesis and of outcrossing flower buds pre-anthesis. The smaller petals of selfing flowers also resemble developing petals of outcrossing buds.
We chose to examine flower size and development in multiple populations of the two subspecies because both subspecies exhibit variation in floral characters, and it is not known which outcrossing lineage might have given rise to self-pollinating forms (Gottlieb, 1984
), or even whether selfing has evolved more than once in C. xantiana.
Another reason for quantifying variation among populations was to compare the within-subspecies pattern of covariation in morphological and developmental traits to the pattern of covariation between subspecies. Such comparisons can give a sense of the degree of developmental or functional integration among flower parts (Armbruster and Schwaegerle, 1996
). For example, similar patterns of covariation within and between subspecies would be indicative either of developmental constraints on the independent evolution of floral parts, or of strong functional integration between floral parts, while dissimilar patterns of covariation would indicate that floral organs can and have changed independently of one another.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). Systematic studies of the genus indicate that self-pollinating taxa are evolutionarily derived from outcrossing ones (Lewis and Lewis, 1955
; Vasek, 1964
; Moore and Lewis, 1965
; Vasek and Harding, 1976
; Vasek and Weng, 1988
) and that self-pollinating species, subspecies, or populations have arisen independently from outcrossing taxa a minimum of ten times (Lewis and Lewis, 1955
; Lewis and Raven, 1958
; Moore and Lewis, 1965
; Vasek and Harding, 1976
; Allen, Ford, and Gottlieb, 1991
).
The species
Clarkia xantiana Gray is native to the southern Sierra Nevada Mountains in Kern and Tulare Counties of California and occurs at a limited number of sites in the Transverse Ranges (Liebre and San Gabriel Mountains) east of Los Angeles (Lewis and Lewis, 1955
; Eckhart and Geber, 2000
) (Fig. 1). Flowers of C. xantiana have an inferior ovary, a small hypanthium, a whorl of four sepals that enclose a whorl of four petals, a whorl of four large stamens, a whorl of four smaller stamens, and a style. Clarkia xantiana ssp. xantiana (hereafter referred to as "outcrosser") has flowers with larger sepals, petals, anthers and styles (Figs. 2–3), and greater herkogamy and protandry than those of C. x. ssp. parviflora (hereafter referred to as "selfer") (Moore and Lewis, 1965
; Gottlieb, 1984
; Sytsma, Smith, and Gottlieb, 1990
; Lewis and Raven, 1992
). Although outcrossing rates have not been measured in the two subspecies, the selfer readily sets seed whereas the outcrosser has much lower fruit set in the absence of pollinators. Electrophoretic studies also reveal that populations of the selfer are either monomorphic or have very low levels of polymorphism at enzyme loci, whereas populations of the outcrosser have much higher levels of polymorphism (Gottlieb, 1984
; Travers and Geber unpublished data). These results are consistent with the notion that C. x. ssp. parviflora is highly selfing.
|
|
). This transect corresponds to a west-to-east moisture gradient. Outcrossers occur alone (allopatrically) in the wetter, western end of the species range; selfers are allopatric in the more arid eastern region. A zone of sympatry is found along the western shore of Isabella Lake and the North Fork of the Kern River where the two subspecies frequently co-occur in mixed populations. At these sites, selfers always flower earlier than outcrossers, with little overlap in flowering time (Eckhart and Geber, 2000
).
Populations
We selected four populations of each subspecies that spanned the range of sepal and ovary lengths measured during an earlier greenhouse study of 45 populations. Each population is referred to by a site number (e.g., 6). Outcrossing populations of subspecies xantiana are designated by an "x" (e.g., 6x), while selfing populations of subspecies parviflora are designated by a "p" (e.g., 6p). Within each subspecies, two out of four populations were selected from the zone of sympatry and two from allopatric regions (Fig. 1). Sympatric populations were chosen from sites where the two subspecies naturally grow side by side or in close proximity (sites 6 and 9). The two allopatric outcrossing populations (13x and 17x) come from the western, wetter end of the species' distribution in the Sierra Nevada, while the two allopatric selfing populations come from the eastern, drier end of the distribution. Subspecies parviflora occurs in two flower colors, pink (as in the outcrosser) and white. This study was limited to an examination of pink-flowered individuals, although white-flowered plants occurred at three of the four parviflora sites (6p, 9p, 23p).
Plant culture
Seeds were collected in the wild in 1996 and 1997 and stored dry at 5°C. Prior to sowing in December 1997, we bulked 2–3 seeds per maternal parent for a total of 216 seeds for each population. The seeds from each population were then sown onto the surface of moist Metro Mix 360 (Scotts-Sierra Horticultural Products Co., Maryville, Ohio, USA) in flats, and placed covered in a refrigerator at 5°C for 7 d. Flats were then transferred to growth chambers where germination took place under short daylength (11 h), low temperature (22°C day/15°C night), and high (90%) relative humidity. After 1 wk, 49 plants were selected at random from each population and were transplanted into a planting mixture (1 Metro Mix: 1 fritted clay, Oil Dri, Softco Mead, Elmira, New York, USA) in 156-mL plastic tubes, and growth chamber conditions were changed to longer daylength (15 h), higher temperature (26°C day/20°C night), and lower relative humidity (60%). Transplants were watered from below for the first few weeks. Later on, plants were watered from above 1–2 times per day and were fed a dilute solution of Excel 15–5–15 (Peters) once every 2–3 d.
Experimental design
Transplanted seedlings were split at random into two cohorts: six plants per population were assigned to a macroscopic study of flower development and the rest were reserved for a microscopic study of floral development. Plants for both studies were subdivided between two growth chambers with identical settings. Growth chamber groups were treated as blocks in statistical analyses.
Macroscopic study of flower development
The purpose of this study was to provide estimates of the duration of flower development and of the sizes of mature flower organs. In order to calculate estimates of development time, floral organs were measured daily on the six plants per population. Prior to flower opening, only sepals and ovaries were measured. Measurements began when sepals were first visible, usually at 1–2 mm length, and continued through stigma receptivity, and were made on the first, third, fifth, seventh, and ninth flower of each plant.
Three measures of development time were obtained: (1) the flower plastochron (i.e., the time interval between the initiation of successive flowers on the main stem of each plant), (2) the duration of flower bud development from a size of 
1–2 mm until flower opening, and (3) the duration of protandry (i.e., the length of time between large anther dehiscence and stigma receptivity). The flower plastochron is a measure of development time that pertains to inflorescence ontogeny, while flower bud development time and protandry pertain more directly to the flower itself.
For floral organ size we used (1) the lengths of the ovary, sepal, petal, large anther, large filament, and style at flower opening, and (2) the maximum lengths of these organs. For ovary, petal, and style, maximum lengths were obtained at stigma maturation. Sepals attain their maximum length at flower opening and so we only use this one measurement of size. Large anthers and filaments in outcrossing populations reach maximum length just prior to anther dehiscence, a few days before stigma receptivity. In selfing populations, we were usually unable to measure anther lengths, because they had already begun to dehisce at flower opening.
Microscopic study of flower development
The purpose of this study was to obtain measures of floral organ length from bud initiation to flower opening in order to construct organ growth curves and estimate the relative growth rates of organs. In order to harvest buds of known age from each population, it was necessary that plants from a given population be at approximately equivalent stages of development at the outset of the study. Toward this end, we selected from each population a subset of 12 individuals on a single day when the sepal length of their third flower was between 2.5 and 3.5 mm. This day was designated as day 1 for that population. Day 1 was not the same for all populations because of population differences in the time of initiation of the first flower bud and in the rate and duration of flower bud development. The selection of 12 individuals of similar developmental stage undoubtedly eliminated some of the variation present in a population, but this could not be avoided. Developmental phenology is in fact very uniform among plants within populations, especially of the selfer: under greenhouse or growth chamber conditions, plants from a given population begin flowering within a few days of each other (Geber, unpublished data).
The 12 plants from each population were randomly assigned to one of six harvests, with two plants harvested on day 1, and two plants harvested at successive 5-d intervals until the last harvest on day 26. The 26-d duration was selected so that the oldest (most basal) flowers would have opened on plants assigned to the last harvest. For the early harvests, all flower buds and the apical region of the main inflorescence of each plant were collected. At later harvests, on rare occasions, the youngest buds and apical region were discarded if the inflorescence showed signs of slowed development and only the oldest buds were collected. The mean nodal positions of buds (± SE) at each of the six harvests were: 14.6 ± 0.8 at harvest 1; 19.5 ± 0.7 etc. at harvest 2; 23.4 ± 1.2 at harvest 3; 27.9 ± 1.3 at harvest 4; 31.2 ± 0.9 at harvest 5; and 29.0 ± 1.9 at harvest 6. The range of nodal positions from which buds were collected did not differ between populations.
All buds up to 15 mm in length and all apices were fixed by vacuum infiltration of 4% glutaraldehyde in 0.1 mol/L cacodylate buffer in 0.15 mol/L sucrose. For all but the smallest buds, one sepal was removed to allow better penetration of the fixative to internal organs. Fixed buds were rinsed, dehydrated through an ethanol series, and embedded in JB-4 resin (Electron Microscopy Sciences, Ft. Washington, Pennsylvania, USA). Flower buds were sectioned at 2–3 µm using an Olympus CUT 4060 rotary microtome equipped with a glass knife holder. Sections were stained with 0.1% toluidine blue in 0.005% sodium carbonate (pH 5.8). We measured the lengths of the sepal, ovary, petal, style, and the combined length of the large anther and filament (anther and filament are indistinguishable during their earliest developmental stages) to the nearest 0.05 mm using a Zeiss Stemi 2000C dissecting microscope equipped with an eyepiece micrometer (Figs. 4 and 5). Flower buds >15 mm were too large to embed and were fixed in formalin–acetic acid–ethanol, and floral organs were measured after dissection, as described above.
For each organ measured, length data from successive flower buds within plants and from different plants from the six harvests were pooled in order to construct a plot of organ length vs. time throughout the period of bud development. In pooling data within and between plants, we made the assumption that flower buds at successive nodes and from different plants were representative of different ages of a flower at the same node position. This assumption is justified only if floral bud development is the same for all buds, independent of node position (see Analysis section).
Analysis
All statistical analyses were performed with SAS statistical software (JMP Statistics and Graphics Guide, 1996
; or SAS Statistical Software, 1996
).
Development time and floral organ sizes
Flower plastochron was calculated after the method of Hill and Lord (1990)
, using data from the macroscopic study of flower development. Plots of loge(sepal length + 1) vs. time were linear during the period that included sepals of 3 mm length, indicating that growth was exponential during this phase. The 3 mm length was chosen as the reference length. The time interval (tref) from bud initiation to attainment of the reference length was calculated as follows:
The plastochron is estimated as the difference in tref between flowers at successive nodes. In this case, where flowers were measured at every other node, the plastochron was calculated as the difference in tref divided by 2. For each population, an estimate of the plastochron was obtained based on the intervals between the first and third, third and fifth, fifth and seventh, and seventh and ninth flowering nodes.
Each of the three developmental time parameters, and the size of each organ, was analyzed by a repeated-measures analysis of variance, where the repetition consisted of measurements at successive flower nodes. The statistical model included the following effects: block, subspecies, population nested within subspecies, node position of the flower, and interactions between node position and subspecies, and between position and population. Block was treated as a random effect, and the other factors were treated as fixed effects. Population was treated as fixed because populations were selected explicitly to span the known range of floral variation rather than at random. The covariance structure of the data was assumed to be compound symmetric, because no improvement in fit was obtained with an autoregressive structure (Crowder and Hand, 1993
). For anther length, we only analyzed data from outcrossing populations, and the subspecies and subspecies by position interaction were therefore dropped from the statistical model.
In assessing the statistical significance of a particular effect on a trait, we accounted for the fact that we were performing multiple analyses on a set of nonindependent traits by adjusting the significance levels of our tests by the sequential Bonferonni procedure (Rice, 1989
).
Test of flower bud equivalency
To test the assumption of equivalency of floral organ development at successive node positions, we compared the allometry of sepal to ovary length of buds from the different harvests of the microscopic study of flower development. We reasoned that if the allometry of sepal to ovary did not differ between harvests, then flower development was likely to be similar across node positions, because, for a given bud size, buds from earlier harvests originated from more basal node positions compared to buds from later harvests. Allometric relationships were determined by least-squares linear regression of loge(sepal length) vs. loge(ovary length).
Because the relationship between log-transformed sepal and ovary lengths was curvilinear when flower buds in all size classes were included in a single regression, we divided the data into three size classes of flower buds: sepal 
6 mm, 6 mm < sepal 12 mm, sepal > 12 mm. Only outcrosser buds were included in the largest sepal size class.
To compare organ allometry among harvests in each of the three bud size classes, we compared the intercepts and slopes of the sepal-to-ovary allometric relationship using multivariate analysis of covariance, with block, subspecies, and population nested within subspecies as main effects, and the mean flower node position of buds from each harvest as the covariate, since mean node position increases with later harvests. A statistically significant effect of the covariate would indicate that floral allometry, and by inference, flower development, varies across node positions.
In pooling data from several buds within plants in the regression analyses, the assumption of statistical independence of observations was violated since flowers within plants are likely to be correlated. We do not consider this to be a serious problem because of the strong intrinsic relationship between sepal and ovary sizes (Niklas, 1994
).
Analysis of organ relative growth rate
To construct curves of organ growth it was necessary to establish the age of each fixed flower bud. The age of a bud was determined from our estimates of flower plastochron obtained in the macroscopic study of flower development and the bud's nodal position relative to the youngest bud primordium on the inflorescence axis from which the bud was collected. The youngest bud was arbitrarily assigned an age of 1. The bud and its floral organs at j node positions basal to the youngest bud was assigned a plastochron age equal to the product of the plastochron and j (1 j 36). We used the population mean value of the plastochron to determine the plastochron ages of the buds collected from plants from that population.
Data from all flowers of each population and block were pooled, and for each organ, organ size was plotted against plastochron age. To reduce the systematic increase in variance associated with larger organs, organ size was log-transformed [loge(organ size + 1)]. Gompertz curves were fit to the data for each population and block, as follows:
). The maximum relative growth rate (natural log transformed of data taken in millimetres per day) of an organ occurs at the inflection point of the Gompertz curve and is estimated as (ra/e) (Tipton, 1984
). For each organ (ovary, sepal, petal, combined long anther and filament, and style), two estimates of ra/e were obtained per population, with one estimate from each block. To compare maximum relative growth rates of all organs among populations and subspecies, we performed a multivariate analysis of variance of relative growth rate estimates for the five organs. The statistical model included effects of block (random) and of subspecies and population nested within subspecies (fixed). We also performed univariate analyses of variance on each organ because the multivariate analysis indicated significant effects of subspecies and population on organ growth rates.
Morphological and developmental covariation
We used principal components analysis to ask whether subspecific differences in developmental time and organ growth rate paralleled differences in flower size. For a set of p characters (e.g., organ sizes), principal component analysis defines a new set of p axes (principal component axes or PCA), which are linear combinations of the original traits, i.e., PCAi = 
aijXj, where Xj is the jth original trait (i, j = 1 to p). The absolute value of the coefficient (or loading), aij, measures the importance of the jth character in defining the direction of the ith axis. The first PCA axis defines the direction of greatest trait variation among individuals in the data set; the second PCA axis is orthogonal to the first in the direction of the second highest level of variation, and so on, up to the pth PCA (Harris, 1985
). Principal components analysis is especially useful if the first few axes account for a major portion of variation in traits, since this variation can then be summarized with respect to a smaller set of composite variables defined by the axes. Plotting of individuals (in our case, populations) with respect to these first few axes allows one to visualize their location in the space defined by the axes (e.g., "morphological" or "developmental space"). In analyses of organ size or growth rate variation, the first PCA often has positive loadings for all traits and can therefore be thought of as a composite measure of overall size or magnitude. Our purpose in doing principal components analysis was (1) to determine how well subspecies were differentiated in "morphological" and "developmental" space and (2) to ask whether the differentiation with respect to flower development (growth rate and duration) paralleled the differentiation in flower size.
We conducted separate principal components analyses on organ size traits (PCASZ), on organ growth rates (PCAGR), and on development time (duration of bud development and protandry, PCADT). For analyses of size and growth rates we only included data on the lengths of ovaries, sepals, petals, and styles, since these were the organs for which we had both final size and growth rate data in both subspecies. In the analysis of development time, we excluded the plastochron, since this is a developmental character of the whole plant, rather than of individual flower buds. The principal component axes were derived from the covariances of mean trait values for each population and block. We report results on the first one or two axes from each analysis, because, in every case, these accounted for > 90% of the trait variation.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
|
|
|
|
|
|
|
|
Trait covariation within vs. between subspecies
Patterns of covariation in floral morphology and development across the two subspecies are not always matched by within-subspecies patterns. For example, while the correlations between overall flower size (PCA1SZ) and organ growth rates (PCA1GR) and between organ growth rates (PCA1GR) and development time (PCA1DT) are strongly negative (Fig. 8c, r SZ,GR =-0.83. P < 0.0001; Fig. 8d, r GR,DT =-0.85, P < 0.0001), these same correlations are more weakly negative in ssp. xantiana (r SZ,GR =-0.47, P = 0.24; r GR,DT =-0.38, P = 0.36) and are weakly positive in ssp. parviflora (r SZ,GR = 0.43, P = 0.29; r GR,DT = 0.49, P = 0.22). Differences in the sign of these correlations between the two subspecies are not statistically significant, perhaps due to the small sample size of populations.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
).
Reduced flower size, herkogamy, and duration of protandry are the norm in selfing Clarkia taxa compared to their outcrossing relatives (Lewis and Lewis, 1955
; Vasek, 1958, 1964
; Moore and Lewis, 1965
; Holtsford and Ellstrand, 1992
; Sherry, 1994
; Sherry and Lord, 1996
). Moore and Lewis (1965)
studied two of the C. xantiana populations included in this study (6x and 6p) and reported that the selfers had smaller organs (they did not measure ovary length) and reduced protandry and herkogamy compared to outcrossers. Sherry (1994)
studied the development of flowers from selfing and outcrossing populations of Clarkia tembloriensis and reported that, except for ovaries, flower organs were smaller in selfing flowers. The morphology and phenology of selfing flowers appear to be effective in promoting selfing. Outcrossing rates are lower in populations of C. tembloriensis with the self-pollinating flower form than in populations with larger flowers (Holtsford and Ellstrand, 1989
) and are strongly negatively correlated with the degree of protandry and herkogamy (Holtsford and Ellstrand, 1992
). In C. xantiana, populations of the white- and pink-flowered ssp. parviflora are much less genetically variable at allozyme loci than populations of spp. xantiana at the same and other locations, as would be expected if selfing rates are higher in spp. parviflora (Gottlieb, 1984
; Travers and Geber, unpublished data).
Flower development
In order to determine the developmental basis of differences in mature flower form between the two subspecies of C. xantiana, we compared flower development time and organ relative growth rates between selfing and outcrossing flowers from four populations of each subspecies. Measures of developmental time were determined from daily measurements of the same flower buds, from bud sizes of 1–2 mm to stigma presentation in open flowers. Organ relative growth rates, on the other hand, were derived from measurements of sectioned flower buds at different stages of development and collected from successive node positions on inflorescences. The accuracy of the relative growth rate estimates depends on the validity of several assumptions. First, flower buds at successive nodes and from different plants must be representative of different ages of a flower at the same node position. In other words, flower bud development must not differ with node position. We tested this assumption by comparing the allometry of sepal to ovaries in buds from six successive harvests that included buds drawn from progressively later nodes. Because there was no evidence of an effect of node position on allometry (Table 4), we concluded that bud development was independent of node position, at least within the range of nodes included in this study. The second assumption underlying the estimates of organ relative growth rate is that flower bud age for a population of plants can be accurately estimated from the population's mean flower plastochron. Plastochron interval was determined for flowers from node positions 1–9, while the buds used in the estimation of organ relative growth rates came from node positions 1–36. We cannot be certain therefore that plastochron values (and hence flower age estimates) from lower nodes are accurate for buds at higher nodes. On the other hand, in our analysis of variation in flower plastochron, we found no consistent effect of node position on plastochron interval (Table 1, Fig. 6). We therefore felt that the mean flower plastochron over nodes 1–9 was our best estimate of a population's plastochron for all nodes.
Developmental basis of subspecific differences in flower form
The developmental comparisons between selfing and outcrossing flowers of C. xantiana were motivated by contrasting hypotheses about the evolution of self-pollination. The first hypothesis states that the selfing flower form evolved as a consequence of direct selection for the ability to self-pollinate in the arid environments occupied by ssp. parviflora. According to the second hypothesis, selfing arose indirectly as a by-product of selection for rapid maturation. The first hypothesis does not make any specific predictions concerning the developmental basis of subspecific differences in flower form, whereas the second hypothesis predicts that a reduction in flower development time (i.e., truncation and progenesis) should account, at least in part, for the smaller size and altered phenology of parviflora flowers. An alternative heterochronic mechanism for producing a smaller flower is through a reduction in the rate of organ development (e.g., slower relative growth rate), with no change in development time. This latter mechanism, which is termed deceleration and which results in neotony, would not be consistent with selection for rapid maturation.
There is clear evidence of progenesis in ssp. parviflora relative to ssp. xantiana. Selfing flowers required 26% less time to develop from a bud size of 1–2 mm to flower opening and the period of protandry was 71% shorter (Table 2A). The flower plastochron interval is also shorter in the selfers compared to outcrossers (Table 2A). These findings are consistent therefore with the hypothesis that selection favored rapid development in ssp. parviflora.
There is no evidence of a deceleration in development rate and of neotony in ssp. parviflora relative to ssp. xantiana. To the contrary, the relative growth rates of all flower organs are accelerated in the selfer compared to the outcrosser (Table 5). Acceleration ranged from a 23% increase in anther/filament relative growth rate to a 57% increase in ovary relative growth rate.. Acceleration of development rate is consistent with selection for rapid maturation to the extent that it allows flower to achieve a mature size in a minimum amount of time. Selection for small flower size in selfing flowers would not necessarily be expected to generate selection for accelerated organ growth rates. Indeed, accelerated organ growth rates mean that flowers of ssp. parviflora are larger than they would otherwise have been had growth rates not differed between the subspecies. Final organ sizes are smaller in ssp. parviflora than in ssp. xantiana, because of shorter development time (Fig. 8b) and in spite of higher organ relative growth rates (Fig. 8d).
In sum, selection for rapid maturation may have caused or contributed to the evolution of selfing in C. xantiana. However, because direct selection for self-pollination in ssp. parviflora could also have caused floral development to change by the same heterochronic means, definitive answers about the selective basis of the selfing flower form await measurement of natural selection in the wild on development time vs. on the ability to self-pollinate. There are reasons to think that the ability to self-fertilize may be directly advantageous, and hence directly selected, in arid regions of C. xantiana's geographic distribution. Members of the genus Clarkia are pollinated by an array of specialist bees that use Clarkia pollen to provision their young (McSwain, Raven, and Thorpe, 1973
). The specialist pollinators are absent from arid areas where ssp. parviflora occurs alone (Fausto, Eckhart, and Geber, 1999
). Field studies of natural selection on development time vs. the ability to self-pollinate are now ongoing (Geber and Eckhart, unpublished data).
A review of the literature on comparisons of outcrossing and selfing flower forms indicates that a variety of heterochronic mechanisms have been involved in the evolution of self-pollination. Shorter development time is found in buds of selfing Mimulus guttatus (Fenster et al., 1995
) and in selfing Limnanthes alba flowers (Guerrant, 1988
). By contrast, flowers from selfing populations of Arenaria uniflora develop over a longer period of time than their outcrossing counterparts (Wyatt, 1984a, b
; Hill, Lord, and Shaw, 1992
). In comparisons of cleistogamous and chasmogamous flowers, developmental differences appear to be more consistent, at least with respect to the flower bud. Cleistogamous flower buds are reported to mature in less time than chasmogamous buds in Viola odorata (Mayers and Lord, 1983
), Lamium amplexicaule (Lord, 1982
), Collomia grandiflora (Minter and Lord, 1983
), and Astralagus cymbicarpos (Gallardo, Dominguez, and Muñoz, 1993
).
There is also no universal pattern with respect to differences in flower plastochron between selfing and outcrossing taxa. Cleistogamous flowers of V. odorata have a shorter plastochron than chasmogamous flowers (Mayers and Lord, 1983
), but flower plastochron is longer in selfing populations of Eichhornia paniculata (Morgan and Barrett, 1989
) and A. uniflora (Hill, Lord, and Shaw, 1992
). Lastly, both acceleration and deceleration of development rate are observed in selfing flowers compared to outcrossing ones. As in C. xantiana, faster growth rates are reported in buds of selfing M. guttatus (Fenster et al., 1995
) and in selfing L. alba flowers (Guerrant, 1988
). Faster growth rates also characterize cleistogamous relative to chasmogamous flower buds (Lord, 1982
; Mayers and Lord, 1983
; Minter and Lord, 1983
; Gallardo, Dominguez, and Muñoz, 1993
). However, selfing flowers of A. uniflora develop at a slower rate than outcrossing flowers (Wyatt, 1984a, b
; Hill, Lord, and Shaw, 1992
).
Taxonomic variation in the developmental mechanisms underlying the transition from outcrossing to selfing may reflect in part the diversity of ecological causes that have favored this transition (Diggle, 1992
). For example, selfing flowers of Arenaria uniflora are not progenetic forms of outcrossing ones since the duration of development is longer and the rate of development is slower in selfing compared to outcrossing flowers. Selfing is thought to have evolved in marginal populations of A. uniflora, not as a by-product of selection for rapid maturation, but because of pollinator competition from A. glabra, or as a means of avoiding hybridization with A. glabra, a congener that is found at the margins of A. uniflora's range (Wyatt, 1986
; Fishman, 1996
).
Developmental constraint and selection in the evolution of the selfing flower
In spite of taxonomic variation in the nature of outcrosser-selfer differences in flower development, one pattern does emerge from this collection of studies. Outcrosser-selfer shifts in flower bud development time, development rate, and plastochron are correlated among taxa. In other words, where the shift to a selfing flower form has involved a reduction in flower bud development time, it is more likely than not to have also involved an increase in flower bud development rate and a decrease in flower plastochron.
Developmental correlations among flower parts and across taxa can be a sign either of development and genetic constraints or of past selection (Berg, 1960
; Armbruster, 1991
; Armbruster and Schwaegerle, 1996
; Hufford, 1997
; Cresswell, 1998
). Developmental correlations can be caused by pleiotropic effects of genes such that change in one aspect of development (e.g., developmental duration) forces change in other aspects of development (e.g., developmental rate). Alternatively selection to improve the functional integration of flower parts can also lead to the evolution of strong genetic and developmental correlations.
In order to address the cause of correlated shifts in development processes (or size of flower parts) between ssp. xantiana and parviflora, we examined trait covariation among populations within each subspecies. Similar patterns of trait covariation within and between subspecies would bolster an argument of developmental constraint. Dissimilar patterns of trait covariation, on the other hand, would suggest that correlated changes in development processes (or size of parts) are the result of selection for a flower that is developmentally and/or functionally integrated.
In Clarkia xantiana, there is a suggestion that the pattern of trait covariation differs within subspecies as compared to between subspecies. Thus, the correlations between organ size and relative growth rate and between developmental duration and rate are negative between subspecies and within ssp. xantiana, but these correlation are positive within ssp. parviflora (Fig. 8d). The shift in the sign of these correlations between ssp. xantiana and ssp. parviflora suggests that developmental duration can be uncoupled from developmental rate, and these two aspects of development are not inextricably linked by pleiotropic effects of genes. The uncoupling of developmental timing and organ growth is also supported by findings of character covariation among populations of C. unguiculata (Jonas and Geber, 1999
), a species that is in the sister clade to C. xantiana. In C. unguiculata, the positive correlation between herkogamy and protandry that is characteristic of interspecific differences in the genus (Lewis and Lewis, 1955
; Vasek, 1964, 1968
; Moore and Lewis, 1965
; Holtsford and Ellstrand, 1992
) appears to have been disrupted in some populations; in these populations, plants have long-duration protandry but virtually no herkogamy. Since herkogamy in Clarkia is primarily the result of continued growth of the style after flower opening (Table 1), it appears that the association between organ growth and developmental duration can also be disrupted. Fenster et al. (1995)
in studies of outcrossing and selfing species of Mimulus have also found evidence that the duration of floral development is genetically independent of the rate of development.
There are now quite a few studies of genetic variation and covariation in floral form (Shore and Barrett, 1990
; Meagher, 1992
; Conner and Via, 1993
; Mitchell and Shaw, 1993
; O'Neil and Schmitt, 1993
; Carr and Fenster, 1994
; Robertson, Diaz, and MacNair, 1994
; Stanton and Young, 1994
; Young et al., 1994
; Bradshaw et al., 1995
; Andersson, 1996, 1997
; Campbell, 1996
; Galen, 1996; Mazer and Delesalle, 1996
). While the verdict is still out on the relative strengths of constraint vs. selection in shaping the genetic and developmental architecture of floral form, it is clear that developmental constraints are not so large as to prevent the independent evolution of floral characters under the appropriate selective regime (Stanton and Young, 1994
; Delph, Galloway, and Stanton, 1996
; Eckhart, 1999
).
|
FOOTNOTES |
|---|
2 Current address: Department of Plant Sciences, University of Cambridge, Downing Street, Cambridge, CB2 3EA, UK.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Allen, G. A., V. S. Ford, and L. D. Gottlieb. 1990 A new subspecies of Clarkia concinna (Onagraceae) from Marin County, California. Madroño 37: 305–310
Andersson, S. 1996 Floral variation in Saxifraga granulata: phenotypic selection, quantitative genetics and predicted response to selection. Heredity 77:2 17–223
———. 1997 Genetic constraint on phenotypic evolution in Nigella (Ranunculaceae). Biological Journal of the Linnean Society 62: 519–532[CrossRef]
Armbruster, W. S. 1991 Multilevel analysis of morphometric data from natural populations: insights into ontogenetic, genetic, and selective correlations in Dalechampia scandens. Evolution 45: 1229–1244
———, and K. E. Schwaegerle. 1996 Causes of covariation of phenotypic traits among populations. Journal of Evolutionary Biology 9: 261–276[CrossRef][ISI]
Arroyo, M. T. K. 1973 Chiasma frequency evidence on the evolution of autogamy in Limnanthes floccosa. Evolution 27: 679–688
Baker, H. G. 1955 Self-compatibility and establishment after ‘long-distance’ dispersal. Evolution 9: 347–348[CrossRef][ISI]
Berg, R. L. 1960 The ecological significance of correlation pleiades. Evolution 14: 171–180[CrossRef][ISI]
Bradshaw, H. D. Jr, S. M. Wilbert, K. G. Otto, and D. W. Schemske. 1995 Genetic mapping of floral traits associated with reproductive isolation in monkeyflowers (Mimulus). Nature 376: 762–765[CrossRef]
Campbell, D. R. 1996 Evolution of floral traits in a hermaphroditic plant: field measurements of heritabilities and genetic correlations. Evolution 50: 1442–1453[CrossRef][ISI]
Carr, D. E., and C. B. Fenster. 1993 Levels of genetic variation and covariation for Mimulus (Scrophulariaceae) floral traits. Heredity 72: 606–618[ISI]
Conner, J. K., 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
Crowder, M. J., and D. J. Hand. 1993 Analysis of repeated measures. Chapman and Hall, New York, New York, USA
Cresswell, J. E. 1998 Stabilizing selection and the structural variability of flowers within species. Annals of Botany 81: 463–473<
} = 0.05)
