Size-Dependent Sex Allocation within Flowers of the Annual Herb Clarkia unguiculata (Onagraceae): Ontogenetic and among-Plant Variation

by Susan J. Mazer, Kelly Ann Dawson
Citation
Title:
Size-Dependent Sex Allocation within Flowers of the Annual Herb Clarkia unguiculata (Onagraceae): Ontogenetic and among-Plant Variation
Author:
Susan J. Mazer, Kelly Ann Dawson
Year: 
2001
Publication: 
American Journal of Botany
Volume: 
88
Issue: 
5
Start Page: 
819
End Page: 
831
Publisher: 
Language: 
English
URL: 
Select license: 
Select License
DOI: 
PMID: 
ISSN: 
Abstract:

.4mericaii Journd of Botany 88(5): 819-83 1. 1001.

OF THE ANNUAL HERB CURKIAUNGUICUWTA

AMONG-PLANT VARIATION'

SUSANJ. MAZER'AND KELLYANNDAWSON

Department of Ecology. Evolution and Marine Biology. University of California, Santa Barbara, Califor~iia 93106 USA: and Departmelit of Ecolog} and F,volutionary Biology, University of Calilbmia, Irvine, California 92697 USA

The relative allocation of resources to male and female functions may vary among flowers within and among iildividual plants for many reasons. Sever;~l theoretical models of sex allocatio~~ a

in plants predict a positive correlatiol~ betwecn the resource status of tlo.ruer or individual and the proportion of rcproductiue resources allocated to female function. These models assume that, independent of resource status, a negative correlation exists between male and feniale investment. Focusing on the allocation of resources within flowers, we tested tllese theoretical predictions and this assumption usii?g the annual Clndiin ritzgrricr~lntir (Onagraceae). \%'e also sought prelii~~inarqevidence for a genetic component to these relationships. From 116 greenhouse-culti~ated plants representing 30 field-collected maternal fan~ilies, multiple flowers and fruits per plant were sampled for gamete production, pollen :ovule ratio, seed ~~un~ber, ovule abortion, seed biomasslfruit, mean individual seed mass, and petal area. IS sex allocation changes as predicted, then (1) assuming that flowers produced early have access to more resources than those produced lates basal flowers should exhibit a higher absolute a~icl proportional investn~ent in feniale function than distal flowers and (2) plants of high resource status (large plants) should produce tlowers with a higher proportional investment in female function than those of low resource status. Within plants. variation in floral traits conformed to the first prediction. Anrong plants and families, no significant effects of plant size (dry sreni biotnass) on intraforal proportional sex allocation were observed. W-e detected no evidence for a negative genetic correlatioil hetweeii male and female in\.estment per flower, even when controlling for plant size.

Key words: Cliirkin unguiczllatii; gender; Onagraceae; pollell : ovule ratio; seed mass; sex allocation; size dependence

In hermaphroditic plants, the proportional allocation of re- sources to male vs. female f~mction (or sex allocation) varies at all ecological levels. Within individuals, sex allocation often changes with floral position or as plants age. For example, ovule number (but not pollen production) declines significant- ly oxper time among sequentially produced flowers of Carrtpnnula rapuncz~loides (Vogleb Peretz, and Stephenson, 1999), and the pollen: ovule ratio declines over time in the highly selfing ~russica napus (Damgaard and Loeschcke, 1994). Similarly, the proportion of male relative to bisexual flowers in Anthrisc~~~

sylvestri~, Leptospermum myv.rinoides, and L. continerttale (Myrtaceae) increases among upper and outer branch positions (O'Brien, 1994: Spalik and Woodell, 1994). In Arisaenza rriphyllum, younger plants are more likely to be male than female (Policansky, 1981; Bierzychudek, 1982, 1984; Clay, 1993). These results suggest that gcnder allocation responds to changes in the resource status of individual flowers or whole plants. By contrast, strong ontogenetic changes in gender allocation exhibit no clear temporal pattern in Spevg~tlaria rrzurina (Mazcr and Delesalle, 1996a, b), and shifts in gender allocation are seen in some plants independently of

u

' hlanuscript received 23 March 2000; revision accepted 1 August 2000.

The authors thank Ann Sakai, Steve Weller. Justin Epting. Camille Barr, ll~ine Rankin. Syndallas Baughman, Keith Vogelsang, and partic~ilarly Tia- Lynn Ashmati and an anonymous reviewer for comments on an earlier version of the manuscript; hlichelle Buran, Kiln Derbyshire. Shannon Fearnley. and Mari Kube for greenhouse and lab assistance; and the Coiumittee on Research at the University of California, Santa Barbara. Further support came from NSF DEB-9815300 to SJM. SJM thanks Mark Camara, Johanna Schmitt, Spencer C. H. Barrett, Stephen Wight, and John Damuth for valuable dis- cussions and criticism.

Author for correspondence (e-mail: mazer@lifesci.ucsb.edu).

variation in resource status (Wolfe, 1992; Diggle, 1995, 1997; Ashman and Hitchens, 2000).

Among conspecific individuals, sex allocation may vary for genetic or environmental reasons. The relative production of ovules and pollen per flower differ among maternal families in the bee-pollinated Canzpariulu rc~p~punculoides

(Vogler, Per- etz, and Stephenson, 1999) and among predominantly selfing lines of Senecio vulgaris (Damgaard and Abbott, 1995), Brassica ~~apus

(Damgaard and L,ocschcke, 1994), and Spergularia ~izarinn(Mazer and Delesalle, 1996a, b). In several studies, variation in sex allocation appcars to be environmentally in- duced. For example, allocation to male flowers is positively corselated with an individual's height relative to its conspecific neighbors in Anzbrosia artemi.siiJ'olia (Lundholm and .4arssen, 1994; see also Ackerly and Jasienski, 1990). and the flowers of small plants have higher pollen: ovule ratios than flowers on large plants in Triphyllztni ~vecturiz and T. gnrizd(jlllorw7~ (Wright and Barrett, 1999). In Raphanus sntivus, increases in local population density result in the production of relatively male-biased flowers (Mazer, 1992).

Among populations, variation in sex allocation may result from natural selection, phenotypic plasticity, or genetic drift (Benseler, 1975; Lovett Doust and Cavers, 1982a, b; Schoen, 1982; Delesalle and Mazer, 1995). Finally, among species, var- iation in sex allocation is often associated with breeding sys- tem and inferred to be the result of adaptive evolution (Queller, 1984; McKone, 1987; Ganeshaiah and Uma Shaanker, 1991).

The recognition that natural selection may discriminate among bisexual genotypes on the basis of their combined male and female fitness motivated the development of theoretical models that predict the conditions that favor particular patterns of sex allocation (Charlesworth and Charlesworth, 198 1, 1987;

s+.3

Large plants, early flowers, or high-resource flowers

a,

Small plants, late flowers, or low-resourceflowers

" opt "opt

Proportion allocated to male function (x)

Fig. 1. Hypothetical relationships between a whole plant's or an individual flower's female and male fitness (estimated as the number of seeds produced or sired) and the proportion of reproductive resources allocated to male func- tion. Total individual or pepflower fitness is maximized at the proportion of male investment where the absolute values of the slopes of the male and female fitness curves are equal. These fitness gain curves illustrate size- or resource-dependent changes in optimal gender allocation when the absolute fitness return for increases in the proportional allocation of resources to either male or female allocation increases with plant or floral resource status. When the male fitness gain curve decelerates faster than the female fitness gain curve for both small and large plants (and low- and high-resource flowers), the optimal allocation of resources to male function decreases as plant size or floral resource status increases (modified from Klinkhamer, de Jong, and Metz, 1997; Brunet and Charlesworth. 1995).

Charnov, 1982; Lloyd, 1984; Lloyd and Bawa, 1984; Geber and Charnov, 1986; Charlesworth and Morgan, 1991; Spalik, 1991). These models use an evolutionarily stable strategy (ESS) approach to determine the optimal allocation of repro- ductive resources to male relative to female function; their principles are described in detail elsewhere (Charnov, 1982; Lloyd, 1984; Brunet and Charlesworth, 1995; Klinkhamer, deJong, and Metz, 1997; Campbell, 1998). These models share the assumption that, among genotypes with similar amounts of resources, there is an intrinsic negative correlation between the allocation of reproductive resources to male vs. female functions.

One aim of these models is to determine the optimum al- location to male function given specific relationships between allocation and fitness. The optimum is estimated from the shapes of the male and female fitness functions (Fig. 1) and is located where the absolute values of the slopes (or selection gradients; Morgan and Schoen, 1997) of the two curves are equal. When the shapes of the fitness gain curves depend on plant size or resource status, the optimum male allocation sim- ilarly depends on these parameters (de Jong and Klinkhamer, 1989; Klinkhamer, 1997). For example, where male reproduc- tive success is a decelerating function of male investment and female reproductive success is a linear (or less rapidly decel- erating) function of female investment (the complement of male investment), then as plant size increases the optimum proportional allocation to female function also increases (Fig. 1). This prediction has been supported by the phenotypic cor- relations observed between plant size and allocation to female function in many monocarpic species (de Jong and Klinkhamer, 1989; Kudo, 1993; Klinkhamer et al., 1997).

These models have also been applied to individual flowers (Brunet and Charlesworth, 1995) in which allocation to male function is measured as pollen production per flower (Fig. 1). Here, the male and female fitness gain curves measure the number of seeds that are, respectively, sired or produced by a flower with a given allocation to male gametes. As in the whole-plant model, a trade-off between investment in male vs. female function per flower is assumed: controlling for resource status, flowers producing relatively large amounts of pollen must produce relatively few ovules or seeds. This model also assumes that for a given level of pollen production, flowers of high resource status will produce more or higher quality seeds than flowers of low resource status. Given male fitness func- tions that decelerate more rapidly than female ones, this model predicts that the optimum proportional allocation to female function will be greater in flowers of high than low resource status.

In many species of annual plants, floral resource status may be strongly coiselated with floral position on an inflorescence. Annuals often produce flowers over a protracted period along stems that decrease in diameter distally and that bear leaves at basal but not distal nodes. In such species, flowers (and fruits) produced relatively late (and distally) may depend on resources not used first by the fruits and flowers produced earlier. If there is such competition among flowers for resourc- es, then flowers produced later will have access to fewer re- sources than those produced earlier. This conclusion is consis- tent with the results of experiments by Solomon (1988) and Guitiin and Navaiso (1996). If sex allocation among flowers responds to resource status as Brunet and Charleworth's (1995) model predicts, then successively produced flowers should be- come relatively male biased. An alternative adaptive expla- nation may also account for the production of increasingly male-biased flowers over time; protandry in sequentially flow- ering plants may favor the production of increasingly male- biased flowers over time (Brunet and Charlesworth, 1995), as observed in Aquilegia caerillea (Brunet, 1996).

In sum, these models predict that flowers on plants of high resource status should have higher female investment than those on ~lants of low resource status and that ontogenetic

u

change in floral gender should progress from female- to male- biased flowers. To date. few em~irical studies ~rovide data that can be used to evalujte these'models. 0nl; two studies of perfect-flowered species have compared gender expression of flowers in small vs. large individuals; both found that small individuals tend to produce male-biased flowers (Damgaard and Loeschke, 1994; Wright and Barrett, 1999). Second, all studies to date have examined phenotypic correlations between plant size and sex allocation rather than estimating genetically based correlations. Phenotv~ic correlations are ~roblematic be

, L

cause they cannot distinguish between environmental and ge- netic causes of covariation between size and sex allocation (Rose and Charlesworth, 1981). To detect evidence for the joint evolution (or "coevolution") of resource status and sex allocation, it is necessary to seek evidence for a genetic cor- relation between the two traits.

The present study examines both ontogenetic variation in sex allocation among flowers within plants and correlations between size and sex allocation among whole plants repre- senting distinct maternal sibships. This allows us: (1) to eval- uate whether within a given population sex allocation responds to changes in resource status in a parallel manner within and among plants as predicted by theory, (2) to seek preliminary evidence for a genetic relationship between plant vigor and floral gender expression, and (3) to seek preliminary evidence for genetically based negative relationships between male and female investment per flower and between other floral traits.

METHODS

Study organism and seed source-Clarkin ~ttlgrric~rlrirn

Dougl. (Onagra- ceae) is a hermaphroditic annual herb that occupies roadside embankments,

May 200 1] MAZERAND DAWSON-SIZE-DEPENDENT. SEX ALLOC,\TI~NIN CL-~RKIA

foot paths, and slopes of oak and oak-pine woodlands in the Coastal Ranges and Sierra Nevada of California, USA (Vasek. 1964, 1971). The species is highly variable morphologically and with respect to the existence oi' supernumerary chrorilosomes and chromosomal reassnngements (Lewis, 1951; Lewis ancl Lewis, 1955: Mooring, 1958, 1960; Holsinger, 1985). C. unglliczllatci is protandrous, herkogatnous, and self-compatible but predorriillalltly outcrossing (r = 0.96 in one wild population; Vasek, 1965). The eight anthers begin to dehisce as soon as the bud opens, exposing pollen that may remain ~iable for 4 d (Smith-Huerta and Vasek, 19811. Several days after the bud opens, the stigma becomes exserted and receptive, and petals reach their max- imum si~e. Slopes and canyons of the Santa Ynez mountains (Santa Barbara County, California) support small populations of C. ur~guiculatii.In Rattle- snake Canyon there are several small (N = 10-300 flowering plants) popu- lations of C ili~gllicirlam.The seeds used here were collected from one such population in July 1996 (USGS Santa Barbara Quadrangle, 7.5 lnin series; Township 4 N;section 3, northeast quadrant). Seeds were collected from 30 rnater~ial plants and stored in envelopes at room temperature.

Experirizenfal design-On 22 January 1997. seeds from each maternal plant were sown in four 18 cm long X 1cm diameter cylindrical plastic tubes with drainage holes at their base (Stuewe and Sons, Corvallis. Oregon, USA). Tubes were tilled to a depth of 10 cm with crushed dry clay; a 7-cm layer of UC xeric mix soil was added and topped with a 0.5-cm layer of gravel. In each tube, 10-20 seeds were sown and watered daily in a UCSB greenhouse. Tubes were placed vertically in plastic racks that rested in a plastic tray filled with -3 cm of water. As seedlings emerged. they were thinned to one robust seedling per tube. This nonrandom sampling within maternal families should have reduced the potential for maternal environmental effects to influence progeny performance. A total of 116 progeny flowered. Throughout the ex- periment, plants remailled under ambient conditions with 16 h day18 11 night supplen~ental lighting.

Plants began to flower on 15 April 1997. As each plant flowered, we sam- pled flowers in sets of three along the plant's primary meristem. The first (most basal) flower in each set was collected tvhen in bud to determilie its pollell production before anther dehiscence. The bud was placed in an open 35-mL glass vial in a dust-fl-ee cabinet for 48 11. The anthers were then re- moved, slit longitudinally with a dissecting needle to facilitate dehiscence. and returned to the vial for I wk. The vial was then capped until the pollell was counted (see below). Tlie second flower was collected when S~~lly

open and frozen in a 2-mL Eppendorf tube. In June-July 1997. the fro~en flowers were dissected to determine ovule number per flower, and the area of one petal was estimated as follows. The length of the petal (in millimeters) was measured from its tip to the base of the blade where its width had narrowed to 3 mni; petal width was measured as the maximum width pelpe~ldicular to the length: and petal area was estimated in square millimeters as length X width. Tlie third (most distal) Bower in each set was hand-pollinated with a mixture of pollell sampled from two plants representing other maternal fam- ilies. These flowers were allowed to mature to determine the number of' viable seeds, rnean individual seed mass, and the total seed mass per fruit.

Flowers were collected and pollinated from I5 April to 2 June 1997 and fruits were collected when dry (May-July 1997). For each set of three flowers, the proportion of o\ule abortion was estimated as:

{[l~umberofovules produced by the unpollinated flower in set r I

[number of seed\ produced by the pollinated flower In set xl ]

+ [number of ovules produced by the unpollinated

flower In set xJ

The pollen: ovule (P:O) ratio was estimated for each set of flowers as the ratio of the number of pollen grains produced by the anthers in the bud to the number of ovules in the ilissected flower. When available, a total of five sets were sampled along the primary branch of each plant. A sixth bud was sarnpled if present. As each plant began to senesce, water was withheld until the plant was completely dry. The mean dry nmss of steni tissue was deter- mined after removing all roots, fruits, and leaves.

Pollen grains were counted ~ising an Elzone 180PC Particle Counter (for- merly Particle Data, Inc., currently Mlcromeritics Instrument Cosporation, Norcross, Georgia. USA; Devlin, 1988). Twenty-four hours before couliting pollen, the vial containing the anthers was filled with a known quantity of a 2% NaCl solution (-30 mL) and sonicated for 2 mill to facilitate the release of pollen from the anthers. Immediately before using the particle counter, samples were sonicated again for 4 tnin. If the pollen had not I'ully dehisced from the anthers. they were permitted to soak for another 24 h, by which time all pollen was relea3ed from the softened anthers. Using the particle counter, we determined the number of pollen 91-ains in five 0.5-mL samples of this solution, shaking the sample between each count to prevent pollen settling. Based on examination of the solution under a microscope, particles between 75 and 160 p.111 in diameter were generally considered to represent pollen grains (the exact size range included in a given sample depended on the bo~~nds

of the frequency distributio~l of particles detected by machine). The 0.5-mL salnples with the highest and the lowest pollen counts were eliminated from analysis. We estimated the number of grains per flower by n~ultiplying the mean number of grains in the three remaining 0.5-mL samples by twice the number of milliliters of solution in the vial.

Accuracy of the Elzor~e particle couizter-In the course of using the par- ticle counter, we became concerned that it might not tlistinguish between viable and aborted pollen grains (which we occasionally observed under a microscope). In addition, we discovered that the shaking of the saline solution that was necessary to preveut Clarkia pollen from sinking introduced air bub- bles that were counted by the particle counter. We tested the accuracy of the particle counter by cornparing the number of pollen grains per Rower that it estirnated with the ~iumber counted by hand. The three-cornered pollen grains of C. ~lr~gliic~ilntnare large enough (80-160 ~nlin diameter) that they can be easily co~nlted using a dissecting niicroscope, although this ~iiethod takes much more time than the particle countel:

For this comparison, we sarnpled 38 fresh flowers of C. rri~glticr~iafa

from the same field population used in this experiment. From each flower, we di- vided the four large and four small anthers into two 2.0-11iL Eppendorf tubes, each containing two large and two small anthers. The tubes were left un- capped for 1 wk to facilitate anther dehiscence.

The tubes with the pollen to be hand-counted were treated as follows. One nlilliliter of 50:50 ethanol : glycerin solution was added and the contents vor- texed thoroughly. From 19 of these san~ples, 20 p.L were added to each of five glass slides along with one drop of Alexander's stain (Alexander, 1980). On the slide surface, the solution was mix-pipetted with a clean micropipette tip and then covered with a cover slip. The slides were etched with a 16 X 16 mm grid to facilitate counting. On each ~lide. all "norn~al" and "aborted" grains were counted. Grains assumed to be normal were those with turgid walls, of full size. and that had absorbetl the red stain. Grains aqsumed to be aborted were deflated, a maxirnunl of 80 pm in diameter, and remained green and t~.anslucent. The number of grains produced by an entire flower was es- timated by calculating the mean number of grains per slide and multiplying this value by 100. For these 19 samples, we estimated pollen production per flower using sample sires of both three and five slides per sarnple. The cor- relation between the mean number of pollen grains per 20 fiL solution using three vs. five slides was very high (r = 0.96, A' = 19, P < 0.0001). From the remaining 19 pollen samples we limited the salnple sire to three slides. The pollen in each of the Eppendorf tubes containing the other four anthers was transferred to a 35-nlL glass vial and then counted using the particle counter as described above.

Among the 38 pollen samples, there was no significant dilTere~lce among methods in the estimated number of pollen grains per flower (F?,,,= 2.66. P > 0.07). The meall number of normal grains per flower estirnated by hand was 12005 (SD = 3672, range: 4533-1 8 733); the mean number of all grains (normal and aborted) per flower estimated by hand was 12996 (SD = 3674, range: 6000-19993); and the mean ~iurnber estimated by the pollen counter was 14 189 (SD = 4935, range: 5101-27229). The mean proportion of abort- ed grains per sample was 0.08 (SD = 0.09). Assuming that the hand-counted estimates of pollen production are more accurate (they contained no bubbles), the particle counter overestimated the total number of grains per flower bq

9% axid the number of normal grains by 18%.

To determine whether the estimates derived from the particle counter ac-

curately predicted the values estiniated by hand, a regression was concluctetl.

The regression of the number of normal grains per flower estimated by hand

on the number estimated by the particle counter was highly significant (Y =

0.32.r + 7719, dl' = 37, P < 0.005), but the co~~elation coefficierit (I-) was

only 0.45 (though significant; P < 0.005). The relatively low value of r

i~idicates that the estimates of pollen production per Rower aud the pollen:

ovule ratios reported below are not highly accurate. Although the particle

counter introduced error, it is unlikely to have introduced bias into our data

set. The comparisons of the mean pollen production per flower and mean

po1leii:ovule ratios among the sets of flowers arid maternal families sampled

in this experiment shoold not be compromised even though the values of these

variables were overestimated by the particle counter.

Statistical ai~alysis and irzterpretatioiz-Vnricitior~ ciilioizg 177citeriznlJinzilics iil porn1 t~.aits-Pheilotypic differences aniong the progeny of different field- collected materrla1 plants may be due to additive or nonadditive genetic var- iation and/or due to envirorimentally induced maternal effects, Konadclitive sources of genetic variation among families include dominance, epistasis. or differences a~nong families in inbreeding depression. Plants in at least one wild population of C. zlng~iicuiotti exhibit outcrossing rates close to 100% (Vasek. 1965), however, suggesting that interfamily variation in inbreeding depression is likely to be very low.

Raising the progeny urider uniiorm greenhouse conditions ancl free of com- petition increases the likelihood that differences observed among the progeny of distinct maternal families have a genetic basis, hut additive and nonadditive sources of variation cannot be distinguished. Also, maternal environmental effects on progeuy phenotype can create the false appearance of heritable variation aniong families. In all studies based on inaternal families, inferences concerning the genetic basis of differences among maternal sibships s nu st be viewed with caution. By using only robust seedlings from each family, how- ever, we ai~necl to minimize environmentally induced maternal effects. The sixfold variation observed among families in final vegetative size (see below) suggests that ~naternal families exhibited significant genetic variation in re- source acquisition or in resource use efficiency, at least under greenho~ise conditions.

To evaluate whether maternal families differed in the mean values of their floral traits, we conducted one-way ANOVAs (Statview. SAS Institute, 1999) using data from each floral position separately. Pollen per flower, mean in- dividual seed mass, and petal area were lognormally transformed to improve normality.

Fanlily-hosed co~c~rintiolz aiijoi~g /lorn/ tmits-Genetic covariation among maternal families was estimated using maternal family means of floral traits, where each family's mean was calculated from the phenotypic means of the individuals representing it. Although we aimed to collect five sets of flowers fro111 the primary flowering sten1 of all individuals, not all stems produced five buds, flowers, ancl fruits. Many individuals coritributed no more than four sets. In the case of pollen production, many of the first buds sampled were eliminated frorn the analysis because the pollen did not fully dehisce frorn their anthers (from the first buds, the anthers had not been slit prior to drying). Therefore, to control for phenotypic variation among individuals and families that might be due to the combination of unequal flower production (across all positions) and ontogenetic effects oil floral traits, we first created a balanced data set with which to calculate maternal family means. Each record in this data set represented an individual plant, ancl for all traits except for pollen production, the data set included only those individuals that contributed the first four sets of flowers and fruits (no other sets were included). In the case of pollen production. many buds in the first set were eliminated from the analysis because the pollen did not fully dehisce fro~ii their anthers. Conse- quently. to estimate family rnearis for pollen production per flower and the pollen: ovule ratio, we analyzed data from all pla~lts contributing buds froni the second, third, and fourth sets (no other sets were included).

When estimating covariances, environmentally induced differences among maternal faillilies in plant vigor could bias among-family correlations towards positive values. We therefore controlled for differences among families in plant size by estimating partial correlation coefticients between floral traits, controlling for stem biomass. The balanced data set and the partial correlations allowed the detection of among-family correlations between floral traits in- dependent of stem biomass (S.4S. 1989) and floral position. All traits except for the P:O ratio ancl the proportion of ovule abortion were lognorn~aily trans- formed to reduce heteroscetlasticity

These analyses were used to seek evideuce for genetic covariation between components of reproduction, and, in particulal; to seek evideuce for trade-offs between male and female investment. In accortlance with the view that corolla size contributes more to inale than to female fitness in animal-pollinated out- crossiilg taxa (Charnov. 1979; Willson, 1979: Bell. 1985). we considered pol- len production and petal area to represeiit male in~estment. The fact that C. irr~g~~iciilnta

petals do not reach maximum size until the anthers have dehisced (i.e., when the flower is just beginning its female phase) may challenge the view that petal size contributes more to male than to female fitness in this species. At the whole plant level, however. if large-flowered genotypes attract more pollen vectors than small-flowered plants, the11 large-petalled flowers may contribute disproportionately to male fitness.

These family means were also used to calculate residuals of ovules per flowet pollen per flower, seeds per flower, and petal area on stem biomass. Regressions between pairs of these residuals were conducted as a second means of detecting trade-offs between male and female reproductive conporients independent of plant size.

Ontogoletic cilnrrge in floriil trc~its-Ontogenetic change in sex allocation was observed by examining how traits related to female and to male function changed as one progresses fl-om early to late flowers along the primary flow- ering stern. Temporal change in floral traits was detected statistically usi~lg a repeated measures ANOVA (SAS Institute, 1989). Maternal family and in- dividual were independent variables, with floral traits measoretl repeatedly within individuals.

Size-depri~deilt cizol?ge in Jior-01 traitc. ai?ioizg i~icifenzol fc~ilziiies cind tziiro~zg iiidividurzis-To detect size-dependent changes in per-flower sex allocation. we examined the regressions among maternal family means between the mean floral phenotype of each trait and mean vegetative stem biomass (Statview, SAS Institute, 1999). To control for phenotypic differences among maternal families that might be due to the combination of unequal flower production per family and ontogenetic change in floral traits. we used constructed two balanced data sets.

The first data set consisted of maternal faniily means calculated using only those individuals for which data from the first four sets of flowers and fr~~its were available; family lneans for pollen production and for the P:O ratio were based on intlividuals contributing the second through fourth sets of buds and flowers. This data set includeti information froni -85 individuals for all traits except pollen. for which 63 i~ldividuals were included.

The second data set consisted of maternal family lneans calculated using data froin all individuals, hut only from the first two sets of flowers atid fruits; family means for pollen production were based on buds from the second and third sets. This data set included information fi.om -114 individuals for each trait. The first data set has the advantage of including more flowers per in- dividual, while the second data set includes more individuals per faniily. These data sets allowed us to determine whether, controlling for flower num- ber and position, there is a relationship between sex allocation and vegetative size among maternal families. If variation among maternal families in size is due to genetic variation, a significant regression coefficient indicates a genetic basis to the relationship between size and phenotype.

Regressions were also condircted using individual plant means (including individuals contributing the first four sets of flowers only) to determine ahether phenotypic regressions, in which size differed alnong individuals by a fac- tor of 23, differed from the among-family regressions, where size differed aniong family means by a factor of six.

May 200 11 MAZERAND DAWSON-SIZE-DEPENDENTSEX ALLOCATION IN CLARKIA

TABLE1. Suinmary of one-way ANOVAs to detect significant differences among maternal families with respect to floral traits. For all flora positions, fernale reproductive components exhibit higher variation among families than male reproductive components. df = model, error degrees of freedom. Boldfaced values are statistically significant at the P < 0.05 level.

Ln(pollen1flower)

Ovuleslflower

Seedslfruit

Total seed mass

Ln(mean individual seed mass)

Proportion ovule abortion

Ln(peta1 area)

Pollen : ovule ratio

RESULTS

variatiorz floral traits-within and amollg individuals of C. unguiculatn, phe~lotypic variation i11 primary and secondary sexual traits is high. In this study, individual flowers produced 42-126 ovules (mean = 84.0, SD = 16.47, N = 49i flowers) and -4800 to -29000 pollen grains (mean = 12406, SD = 4075, N = 354 buds). The pollen : ovule ratio ranged from 56 to 427 (mean = 146, SD = 55.08, N = 351 buds and flowers). Fruits produced 5-95 viable seeds per fruit (mean = 47.4, SD = 21.96, N = 451 fruits), with mean in- dividual seed mass per fruit ranging from 0.18 to 1.10 mg (mean = 3.97, SD = 1.33, N = 439 fruits). The area of single petals ranged from 63 to 195 mm2 (mean = 121.0, SD = 21.03, N = 491 flowers).

Amo~lg all individuals monitored in this experiment that contributed data from both the first and second sets of flowers sampled (or the second and third bud, for pollen), mean ovule productio~l per flower ranged from 57 to 118 (N = 116 indi- viduals), and mean pollen production per flower ranged from 6585 to 25 237 grains (N = 70). The pollen : ovule ratio among individuals contributing data from the second and third sets of flowers and buds varied from 80 to 346 (N = 58). Individuals produced a mean of 5.5-86.5 viable seeds per fruit (N = 114) with mean individual seed mass ranging from 0.19 to 0.85 mg (N = 113). The mean area of a single petal varied among i~ldividuals from 85.3 to 181.2 mm2 (N = 116). Stem biomass among individuals ranged from 0.10 to 2.31 g (N = 116).

Variatiorz amorzg materrzal families irz floral traits-For most traits and floral positions, only female traits differed sig- nificantly among maternal families (Table I). For at least three of the five floral positions, families differed significantly with respect to the mean number of ovules per flowel; the number of seeds per fruit, total seed mass per fruit, and the estimated ~ronortion of ovule abortion. Significant differences amone

L " L

maternal families were detected for mean individual seed mass only for the first and second fruits sampled, Maternal families differed with respect to the pollen : ovule ratio oIlly for the last flower sampled,

Fumily-based covariation amorzg floral traits-Male VJ. fe- innle investment-The partial correlations revealed no signifi- cant correlations between any pair of male and female repro- ductive components (Table 2). Similarly, In-ln regressio~ls among family means based on the residuals of ovules per flow- er, seeds per fruit, pollen per flowel; and petal area 011 stem biomass detected no negative relationships between male and female reproductive components (N = 30 families; residual of 111[pollen/flower] vs. residual of ln[ovules/flower]: y = 0.05~ -0.0004, P > 0.87; residual of ln[pollen/flower] vs. residual of ln[seeds/fruit]: y = -0.03~-0.0001, P > 0.99; residual of ln[petal area] vs. residual of ln[seeds/fruit]: j= -0.003x, P > 0.96; residual of ln[petal area] vs. residual of ln[ovulesl flower]: y = 0.07x, P > 0.9999).

Covnriation ninong female reproductive components-Sig- nifica~lt partial correlations between several pairs of female traits were detected (Table 2). Independently of stem biomass, two negative correlatio~ls appeared: families producing many ovules per flower or many seeds per fruit produced seeds of low mass. Consistent with the latter correlatio~~,

families with high rates of ovule abortion produced relatively few seeds per fruit, and those seeds were of relatively high individual mass. In snite of the trade-off between individual seed mass and seed number per fruit, families produci~lg many seeds per fruit also exhibited higher total seed mass per fruit.

Orztogerzetic change in floral traits-Repeated-measures

824 AMERICANJOURNALOF BOTANY [Vol. 88

TABLE2. Partial correlation coefficients among maternal family means, controlling for variation in stem biomass. All variables except for the proportion of ovule abortion and pollen : ovule ratio were In-transformed. Significant coefficients appear in boldface type. P values (in paren-

theses) are not reported for nonsignificant coefficients.
S~ngle petal area    Ov~iles per Rower    Pollen: ovule ratlo    No. vlable seeds per frolt    Pl.oportlon ovule abortlo11    Total seed inasslfr~iit    Mean ind~v~dual seed nia\s per fiult
Pollen grainslflowei Single petal area

Ovules per flower Pollen :ovule ratio

No. viable seedslfruit

Proportion ovule abortion Total seed masslfruit

ANOVA detected significant effects of floral position on ovule number per flower, the P:O ratio, viable seeds per fruit, total seed mass per fruit, and mean individual seed mass (Table 3). Flowers produced relatively late produced significantly fewer ovules than those produced earlier (Fig. 2). Similarly, seed number and total seed mass per fruit declined significantly from basal to distal positions. The reduction in seed number in distal positions was so great that, even though total seed mass per fruit declined distally, mean individual seed mass increased towards the distal end of the flowering branch.

The proportion of ovule abortion increased distally, but not significantly. The reductions in ovule and seed number from basal to distal positions were similar enough that seed set (seeds per ovule) remained constant. In contrast to the steady declines exhibited by these components of female reproduc- tion, positional changes in pollen production were erratic. Howevel; the distal decline in ovule number was so strong that the P:O ratio increased significantly from basal to distal po- sitions. Petal area remained constant with position.

Size-dependent change in %oral traits among maternal families and among individuals-Stem mass provided a re- liable measure of growth, reproduction, and apparent plant vieor. Stem mass was ~ositivelv correlated with lifetime flower

L

production, branch number, and total branch length, and neg- atively correlated with branch length per flower (Fig. 3). Fam- ilies of relatively large individuals produced more flowers over their lifetime and more flowers per unit branch length than families of small individuals. Among families, lifetime flower production increased proportionately with plant size (the slope of the In-ln regression of lifetime flower production 011stem biomass is not significantly different from 1.00).

Among maternal family means, the only traits that covaried significantly with vegetative size were ovule and pollen pro- duction per flower (Fig. 2). As mean plant size increased, both mean ovule and pollen production per flower increased. The slopes of the In-ln regressions of ovule and pollen production per flower 011vegetative stem biomass were statistically indis- tinguishable (Fig. 2), indicating that male and female primary investment per flower increased at the same rate with increas- ing stem biomass. The P:O ratio remained constant as plant size increased, indicating no significant change in sex alloca- tion with plant size. These results are qualitatively identical to those based on maternal family means calculated from the first two sets of flowers (or second and third sets, for pollen traits; analyses not shown). These results also mirror the phenotypic regressions based on individual means (individuals contribut- ing four sets of flowers are included and all variables were ln- transformed: ovules/flower vs. stem biomass (SB): j= 0.15~

+
4.53, P < 0.0001, N = 85; petal area vs. SB: y = -0.04~
+
4.77, P > 0.07, N = 85; pollenlflower vs. SB: y = 0.12~
+
9.50, P < 0.01 18, N = 63; pollen :ovule ratio vs. SB: y = -0.05~+ 4.97, P > 0.37, N = 57; seedslfruit vs. SB: j= 0.10~+ 3.94, P > 0.30, N = 69; mean seed mass vs. SB; y = -0.06~+ 1.28, P > 0.29, N = 69; proportion ovule abor- tion vs. SB: y = 0.021~-1.01, P > 0.83, N = 69).
DISCUSSION

We examined patterns of floral sex allocation both within individuals among successively produced flowers and across individuals and maternal families differing in total plant size. Our data are partly consistent with the widely cited theoretical predictions that: (1) flowers should become more male-biased from basal to distal positions (i.e., as plants grow) and (2) flowers produced by large individuals (i.e., plants of high re- source status) should be more female-biased than those pro- duced by small individuals. Ontogenetic changes in floral sex allocation agreed with theoretical predictions, but the observed changes in mean floral phenotype in response to total plant size did not. Flowers sampled from large individuals and from vigorous maternal families produced more pollen and more ovules than those from less vigorous plants, but the relative allocation to male vs, female primary sexual traits did not change with total plant size.

Differences between previous studies and the current study-This study differs from previous ones in several ways. First, we examined size-dependent sex allocation at two levels:

(I) as plants grew and (2) among plants and maternal families differing greatly in whole-plant size. Other studies to date of floral sex allocation have been conducted either within or among plants. This approach allowed us simultaneously to test the two predictions just mentioned and to ask whether within a given population floral traits are equally sensitive to onto- genetic factors and to whole-plant attributes. We found that ontogenetic variation in sex allocation was much greater than variation among plants of different size. The lack of variation detected among individuals and families in sex allocation is consistent with the growing view that natural plant populations

May 20011 MAZERAND D~U'SON-SIZE-DEPENDENTSEX ALLOCATION IN CLARKIA 825

TABLE3. Surn1n:rry of repeated-measures ANOVA to detect significant effects of flower position on gender-related traits. Type IT1 sulns of squares were used. All traits except for the number of viable seeds per fruit and total seed mass per fruit were In-transformed to improve normality.

Significant results are indicated in boldface type.

Ovoles per lloi*-er

Source of variation df MS F

Maternal family 27 0.1369 1.98

Error 35 0.0691 Position 4 0.3009 22.15 Maternal family X Position 108 0.0198 1.46

Error 140 0.0135

Pollen: ovule ratio (~ndlv~duals 5eth 2-4

contr~but~nn

Source of \ariation df MS F

Maternal farnily 23 0.1727 1.19

Error 35 0.1155 Position 2 0.6837 7.59 Maternal family X Position 46 0.1 I90 1.32

Error 70 0.0901

'hul 5eed ma\s per fruit

Source of vnr~ation df MS F

Maternal family 23 0.0003 4.65

Emor 25 0,0001 Position 4 0.0002 9.04 Maternal family X Position 92 0.00002 1.12

Error 100 0.00002

Tndiv~dusl petal area

Source of \ar~at~on df MS F

hfaternal family 27 0.091 0 1.06

Enor 35 0.0854 Position 4 0.0 173 1.00 Maternal farnily X Position 108 0.01 62 0.93

Error 140 0.01 73

harbor greater genetically based variation in resource-garner- ing ability among individuals than in the proportional alloca- tion of resources to male vs. fernale investment (Campbell, 1997; Fenster and Carr, 1997).

Second, this study was unusual in its focus on gender al- location within bisexual flowers. Most previous studies have been conducted either on monoecious (or andromonoecious) species, where the ratio of male to female or male to her- inaphroditic flowers is used to estimate sex allocation, or 011 perfect-flowered species, where the ratio of lifetime flower production (a proxy for male investment) to lifetime fruit or seed production (female investment) is used to estimate sex allocation. Within-flower changes in gender allocation are dis- tinct from those meas~tred at the whole-~lant level in that the former require developmental "decisions" to occur as each flower is produced. By contrast, the change from female to male flower production in monoecious species occurs once during the development of an inflorescence or individual, and changes in the fruit to flower ratio depend on post-fertilization abortion occurring throughout fruit development. While nil- merous studies of ontogenetic changes in reproductive components of hermaphroditic flowers have been conducted (Ash- man, 1992; Stoecklin and Favre, 1994; Vogler, Peretz, and Stephenson, 1999; Medrano, Guitian, and Guitian, 2000), de- tailed studies of both male and female prilnary sexual traits are relatively rare (Mazer and Delesalle, 1996a, b: Lehtila and Strauss, 1999; Ash~nan and Hitchens, 2000).

Third, by raising offspring derived from maternal families, we sought preliminary evidence of genetic variation in both

Pollen pcr Rowel (indlvidunis contrib~iting huds 7-4 only)

P    df    \.IS    F    P
0.0288 0.0001 0.0178    25 3 8 2 50 76    0.1492 0.1440 0.2245 0.0775 0.0925    1.04 2.43 0.84    0.45 15 0.0950 0.7455
only) Seeds per fruit

P df MS F P

0.3170 23 1879.8 2.93 0.0031 30 642.1 0.0010 4 233 1.7 1 1.44 0.0001

0.1451 92 165.11 0.81 0.8545 120 203.74

Mean ~nd~v~dunl seed nlaq

P dC MS F P

0.0002 23 0.3178 1.13 0.3823

25 0.28 14 0.0001 4 0.0721 3.46 0.0109 0.2834 92 0.0240 1.15 0.2454

100 0.0209

Prilportion of ovule abortloii

P df MS F P

0.4257 29 0.1055 3.29 0.0014

29 0.0321 0.4103 4 0.0 182 1.90 0.1 145 0.6419 92 0.0082 0.86 0.7686

floral gender and plant vigor. One advantage of examining size-dependent sex allocation among maternal families raised under uniform co~ldit~ons

is that an observed correlation be- tween plant si~e and gender is likely to have a genetic basis. By contrast, phenotypic acsociations between plant size and sex allocatioli observed in a heterogeneous field environment (e.g., Wright and Bassett, 1999) are more likely to be environ- mentally induced. Establishing a genetic basis for the associ- ation between size and gender is important because only such genetic associatio~ls can be the outcome of natural selection (and not simply environmentally induced). Here we detected no evidence for a genetic basic to size-dependent sex expres- sion.

The lack of variation in floral sex allocation among plants that differ greatly in biomass suggests that resources alone rnight not be the most important cause of position effects 011 floral sex allocation within plants. Architect~~ral or develop- mental changes associated with floral position rnay have dif- ferential mechanical effects on stamen vs. carpel development, causing position effects on sex allocation independent of re- source availability. In addition, differential movement or pro- duction of horinones throughout an illflorescence or flowering branch could affect the relative development of inale vs. fe- male floral parts. Moreover, differences among flowers in gene expression might be position dependent, resulting in pheno- typic changes along the axis of an inflorescence. While some experimental evidence indicates that resource availability does vary with floral position (Solomon, 1988; Guitikn and Navar- so, 1996), developmental responses to position effects on vas-

Flower Order

Ln(Vegetative Stem Biomass)

I

35 54j
0 1 2 3 4 5 6

Flower Order

Ln(Vegetative Stem Biomass)

E .2 .4 .6 .8 1 1.2

.-.

Flower Order w Vegetative Stem Biomass (g)

Flower Order Ln(Vegetative Stem Biomass)

Fig. 2. Size-dependent ourogenetic and among-family variation in floral traits. Panels on left illustrate ontogelietic changes progressing frotn basal to distal flowers: panels on right illustrate changes in mean phenotype among families of different vegetative size. Repeated ANOVAS (Table 3) conducted to detect significant ontogenetic changes are based on a balanced data set (including individuals contributing four sets of flowers or fruits or three sets of buds). The panels on the left show the phe~~otypic

means and SE for all flowzrs obcerved at each position (sample sizes appear helotv each mean). F value refers to the effect of position (or set) on phenotype detected by the repeated ANOVA (df indicated in Table 3). Superscripts signify the significance of the F value: * P <

0.05: "9 p < 0~01;:*:$q: P < 0.001; "+**: P < 0.0001; ns = nonsignificant. Bivariate plots on the right show the regressions of the mean phenotype of each trait on mean dry stem biomass: each point represents a materual family mean based on the lirst four sets of flowers produced by fanlily members.

May 20011

Flower Order Ln(Vegetative Stem Biomass)

y = 0.15x+ 9.50; R2 = 0.28; P = 0.0059

F=2.43'lS

Flower Order Ln(Vegetative Stem Biomass)

0123456 Flower Order

Vegetative Stem Biomass (g)

Flower Order Ln(Vegetative Stem Biomass)

Fig. 2. Continued.

Ln(Vegetative Stem Biomass) Ln(Vegetative Stem Biomass)

Ln(Vegetative Stem Biomass) Ln(Vegetative Stem Biomass)

Fig. 3. Relationship among maternal familier between lifetime mearures of reproduction or size and aboveground vegetative stem biomass. Vigorous genotypes appear to be more efficient at fower production (they produce less rtem length per flower) than lerr vigorour genotyper.

culature, hormone availability, or gene expression cannot be ruled out as proximal causes for position-dependent sex allo- cation. Indeed, the expression of some genes have been found to be associated with flower or inflorescence position (Maz- zucato et al., 1999; Yu et al., 1999), potentially accounting for observed position effects on sex allocation.

Zmplicatiorzs for models of size-dependent sex allocation-

If the observed changes in sex allocation per flower within plants and among families had been quantitatively similar arzrl were the result of adaptive evolution, this would suggest that the shapes of the fitness gain curves for high- vs, low-resource flowers mirror those for high- vs. low-resource individuals in

C. unguiculata. In this study, however, ontogenetic change in floral sex allocation was much greater than size-related change among families.

If the changes in floral sex allocation observed here are adaptive, this would suggest that the male and female fitness gain curves for high- vs, low-resource status flowers within C. urzguiculata plants differ qualitatively from those for flowers produced by high- vs. low-resource status individuals. Our re- sults suggest that as one progresses from basal to distal flowers within plants, the optimum proportional investment in male function increases, as predicted by Fig. 1. In contrast, the ab- sence of size-dependent P:O ratios observed among families suggests that the per-flower male and female fitness functions for small vs, large plants do not generate size-dependent op- tima for primary sex allocation. In this study we did not es- timate sex allocation at the level of whole plants, so it is un- known whether the fruit: flower ratio increases (becomes more female-biased) with increasing size (see Aker, 1982; Watkin- son, 1982; de Jong and Klinkhamer, 1989; Klinkhamer and de Jong, 1993).

Previous evidence for resource-dependent sex allocatiorz-

Previous studies of size-dependent patterns of sex allocation in monocarpic, perfect-flowered plants have suggested that the shapes of male and female fitness functions may be resource dependent. For example, Klinkhamer and de Jong (1987, 1993) found strong correlations between plant size and com- ponents of male and female reproduction in C~~rzoglossi~i~z

ofjcinale (Boraginaceae). In the field, small plants have higher rates of ovule abortion and fewer seeds per fruit per gram of dry mass than large plants. In addition, small plants produce more flowers per gram of dry mass than large plants. Both patterns indicate increased proportional male investment among small plants.

Similar size-dependent changes in whole-plant sex alloca- tion have been found in the perennial Yucca whipplei (Agavaceae) (Aker, 1982). Relative to the size of the basal rosette, lifetime flower production was higher in small plants than in large ones. In addition, mean seed mass increased with plant size in one of two populations studied. Vi~lpia fascicillata (Watkinson, 1982) exhibited a similar pattern: lifetime seed production increased more rapidly with plant size than did total flower production. In a review of 34 studies representing

May 200 11 MAZERAND DAWSON-SIZL-DEPENDENTSEX ALLOCATION IN CUKKIII 829

31 hermaphroditic, monocarpic species, de Jong and Klink- hamer (1989) report that in 28 cases there was a positive phe- notypic correlatio~l between size and some standardized mea- sure of feinale investment (sensu Lloyd and Bawa, 1984). These results are consistellt with the view that the shapes of male and female fitness curves are similar to those shown in Fig. 1.

Male vs. female investment: no negative genetic correlations-We detected no evidence for a negative corre- lation among family means between male and female com- ponents of reproduction, even when controlling for total plant size (Table 2). The absence of significant negative coi-relations between male and female i~lvestinent suggests that genetic var- iance in the resource-garnering ability (or resource use effi- ciency [RUE]) of whole plants may be high enough to obscure a negative genetic covariance between resources invested in male and female function (van Noordwijk and de Jong, 1986; Houle, 1991). Indeed, when the genetic coefficients of varia- tion (CV,) based on maternal family means of aboveground stem biomass and the pollen :ovule ratio are estimated (Houle, 1992). it appears that there is higher genetic variation in re- source acquisition (or RUE) than in gender expression. The CV, for In-transformed stem biomass was 36.6% while that of the In-transformed pollen :ovule ratio was 1.5%. This result supports the view that the relatively high variation among fam- ilies in their levels of resource acquisition or RUE makes it difficult to detect negative relationships between components of reproduction. However, the inability to detect a trade-off between inale and female investment even when stem biomass is controlled statistically suggests either that high variation among maternal families in resource status cannot alone ex- plain the lack of a negative correlation or that stem bionlass is not a good assay for the resources used for gamete produc- tion.

The one trade-off between male and female gender expres- sion observed here is the negative correlation between the P:O ratio and seeds per fruit (r = -0.32, P > 0.14; Table 2), which merits further exploration. Significant negative relationships were detected between some components of female reproduc- tion (e.g., seed number per fruit vs. mean individual seed mass), suggesting that resources not allocated to one trait may be diverted to another. In the current study, however, this kind of diversion did not generally occur between male and female reproductive components.

Several other studies have similarly detected 110 negative genetic correlations between male and female reproductive compoilents in wild species with bisexual flowers. Mossop, Macnair, and Robertson (1994) cultivated ten clones of Minzuli~s guttatu,~ from a natural population, providing measures of anther mass (which is correlated with pollen production), pollen viability, corolla mass, ovary size, and fruit mass. These traits were measured under two treatments: one in which all flowers were pollinated (high-resource demand) and another in which flowers were removed after anthesis (low-resource demand). In spite of high between-clone variation, there was no evidence for a trade-off between anther mass and ovary size, even when corolla size was controlled for statistically. Similarly, plants producillg larger quantities of viable pollen grains per flower exhibited no deficits in flower production, ovary or fruit size, nor did they show an increased rate of decline in the phenotypic value of these traits as plants age. In another study of M. guttatus (Robertson, Diaz, and Macnair, 1994), quantitative genetic parameters were estimated for 20 floral traits. Although both pollen quality and ovule production exhibited significant levels of additive genetic variation, there was no detectable genetic correlation between them. A positive correlation between pollen and seed production, however, was detected, suggesting that variation among clones in resource- garnering ability may exceed variation in the proportional al- location of resources to male vs. female investment. In a third study of M. guttat~~s,

Fenster and Carr (1997) raised full-sib families representing a highly selfing and an outcrossing pop- ulation and measured ovule and pollen production per flower, pollen size, and corolla size. No negative genetic correlations were detected among these traits. even when corolla size was controlled statistically. The authors suggested that the positive correlations observed between co~nponents of male and female reproduction may be due to variation among genotypes in re- source-garnering ability.

To test the assurnption that increased allocation to male function results in reduced female function in Iponzopsis ag- gregata, Campbell (1997) examined floral traits among 32 pa- ternal half-sib families and 229 maternal half-sib families. She detected no significant negative genetic correlation between stamen biomass and any measure of male or feinale invest- ment: corolla mass, calyx mass, pistil mass, and total seed mass per fruit. Indeed, among paternal family means, stamen biomass was positively correlated with corolla, pistil, and seed mass. These results again suggest that genetic variation in re- source-garnering ability may obscure trade-offs between the allocation of resources to male vs. female function. A similar conclusion was drawn in a study of tristylous Lythrunz sali- curia (O'Neil and Schmitt. 1993).

By contrast, several studies have detected evidence for ge- netically based trade-offs between male and female investment at the level of individual flowers. Artificial selection over three generations to reduce anther number per flower resulted in a compensatory increase in ovule production per flower in Spergularia marina, and selection to increase ovule number per flower resulted in a decline in anther number (Mazer. Dele- salle, and Neal, 1999). A study of Camparzula ra~>u~zculoides detected a significant negative correlation among 11 clones between ovule and pollen production per flower (Vogler, Per- etz, and Stephenson, 1999). Among hermaphroditic genotypes in the gynodioecious Fragaria virginiana, pollen production per flower was negatij-ely coirelated with the probability that a flo~ler would develop into a fruit (Ashman, 1999). Rameau and Gouyon (1991) reported negative genetic correlations be- tween seed mass and the number of viable pollen grains pro- duced in horticultural clones and hybrids of Gladio1u.r.Finally, Atlan et al. (1992) reported a negative coi-relation among ma- ternal family means between the number of germinating seeds per fruit and the number of full pollen grains per flower in

Tf~ytnus vulgaris.

Ideiztifyiizg the adaptive signiJicance of size-dependent sex allocation-It is not possible to assert whether observed changes in sex allocation are adaptive without knowing the shapes of male and female fitness functions depicted in Fig.

1. Predicting the optimal allocation to male (or female) func- tion requires precise estimates of the curvature of these func- tions, which may be impossible to obtain without huge sample sizes and a large number of molecular markers. Moreover, the shapes of male and female fitness gain curves may depend on the quality of the environment in which plants grow: a pos- sibility that has not yet been explored. It may therefore be unrealistic to expect to estimate these parameters for a wide range of species under field conditions.

An alternative approach to detecting the adaptive signifi- cance of size-dependent sex allocation would be to compare patterns of gender change in species with different mating sys- tems or pollen vectors. For example, wind-pollinated species may be more likely to exhibit increases in proportional allo- cation to male function as plant size increases than are ento- inophilous species (Bickel and Freeman, 1993; Dajoz and Sandmeier, 1997). This has been proposed to be the result of the increased dispersal ability of pollen with increased plant height in wind-pollinated taxa. One would also predict that completely autogamous taxa should not exhibit size-dependent sex allocation, as all flowers contribute equally to both male and female function. By such comparisons. it may be easier to find support for the argument that size-dependent sex al- location is the adaptive outcome of natural selection than to rely on the estimation of fitness f~~nctions.

LITERATURE CITED

ACKERLY,D. D., AND M. JASIENSKI.1990. Size-dependent variation of gen- der in high density stands of the monoecious annual, Al?lbro~osia rrrteini- sifolirr (Asteraceae). Oecologia 82: 474-477.

AKER, C. L. 1982. Regulation of flower. fruit and seed production by a monocarpic perennial. Yircca rvi~i/~p/ei.

Jorrrnal ofEcology 70: 357-372. ALEXANDER, 1980. A versatile stain for pollen, fungi. yeast and bac-

M. I? teria. Staiiz Tec,hi~ology 55: 13-1 8. ASHMAN. T.-L. 1992. Indirect costs of seed production within and between seasons in a gynodioecious species. Oecologia 92: 266-272.

. 1999. Determinants of sex allocation in a gynodioecious wild straw- berry: implications for the evolution of dioecy and sexual dimorphism. Joi~rrznlof' Evohrtioriary Biology 12: 648-66 I.

----, AND M. S. HITCHENS. 2000. Dissecting the causes of variation in intra-inflorescence allocation in a sexually polymorphic species, Frogaria vitginiarlci (Rosaceae). Anzericar? Jo~lrnril of Bottr~i? 87: 197-204.

ATLAN.A,, P-H. GOUYOY. T. FO~JRUIAL. D. POZ.IENTE, AND D. COI:VET. 1992. Sex allocation in an hermaphroditic plant: the case of gynodioecy in Thyri~us 1~t1gnri.s L. Joitrrinl (gEvol~rtiona,:v Biology 5: 189-203.

BELL, G. 1985. On the function or flowers. Proceeditig~ of the Royal Socien of London, Srries B 224: 223-265. BENSELER,R. W. 1975. Floral biology of California buckeye. Mnrlroiio 23: 41-53.

BICKEL,A. M., AND D. C. FREEMAN. 1993. Effects of pollen vector and plant geometry on floral sex ratio in monoecious plants. Ainericarl Mid- land Natrrm1i.st 130: 239-247.

BIERZYCHUDEK,1982. The demography of Jack-in-the-pulpit, a forert pe-

P rennial that changes sex. Ecnlogiccrl A4onograph.v 52: 335-351. . 1984. Determinants of gender in Jack-in-the-pulpit: the influence of plant size and reproductive history. Oeculogia 65: 14-18. BRGNET,J. 1996. Male reproductive success and variation in fruit and seed set in Aquilegia caerirlen (Ranunculaceae). Ecolo,qy 77: 2458-2471. -, AND D. CHARLESWORTH.

1995. Floral sex allocation in sequentially bloo~ning plants. Evolirtioii 49: 70-79. C.AMPBELL,D. K. 1997. Genetic co~~elation

between biomass allocation to Inale and fe~nale functions in rt natural plant population. Heredity 79: 606-614.

1998. Variation in lifetime male fitness in Ii~oiiioi~.sis

rrggregufa:tests of sex allocation theory. Ainericaii iVatrrrali.st 152: 338-353. CHARLESWORTII. 1981. Allocation of resources

D., AND B. CHARI.ESWORTH. to male and female functions in hermaphrodites. Biological Jo~ri.iiril of the Liilizr~iri Society 15: 57-74.

. AXD ---. 1987. The effect of investment in attractive str~~ctilres on allocation to male and female f~lnctions in plants. Evolutioil 41 : 948

968.

, AND M. T. MORGAN. 1991. Allocation of resources to sex functions in flowering plants. Philosi,]~hiccil Trcrri.srrctioils (fthe Royal Society of Loildoii, Series B 332: 91-102.

CHARWOV. 1979. Simultaneous Ilermaphroditism and sexual selection.

E. L. Proceerfings of the Nrltioiznl Academy of Sciences, USA 76: 24110-2484. . 1982. The theory of sex allocation. Princeton University Presr, Princeton. New Jersey. USA. CLAY,K. 1993. Size-dependent gender change in green dragon. Arisrreina

dracoiztiirm (Araceae). Ainei.icnrl Jorrrnal of' Botaizy 80: 769-777. DAJOZ,I., A~D 1997. Plant size effects on allocation to male

M. SAVDMEIER. and female functions in pearl millet, a hermaphroditic wind-pollinated species. Caiinrliuiz Jourrliil of Botnny 75: 228-235.

DA~~GAARU, 1995. Positive correlatiolls between self-

C.. AND R. J. A~BOTT. ing rate and pollen-ovule ratio within plant populations. Evol~rtioii49: 214-217.

, AND V. LOESCHCKE. 1994. Genotypic variation for reproductive characters, and the intluence of pollen-ovule ratio on selfing rate in rape seed (Bias.sicn iial~rts). Joirrncil c>j'Evol~rtio~icrry

Biology 7: 599-607. DE SONG,T. J.. AND P G. L.. KLINKHAMER.

1989. Size-dependency of rex- allocation in hern~aphroditic, monocarpic plants. F~rii~tio~iril

Ecologj, 3: 20 1-206. DELESAL.~.~. S. J. MAZER. 1995. The structure of phenotypic

V. A,, AND variation in gender and iloral traits within and among pop~~lations of Spergirlcrriii niririn~~ (Cavyophyllaceaej. Ai?re~.icirn Jouri?al of Aotnily 82: 798-8 10.

DEVLIN,B. 1988. The effects of strers on reproductive characterr ol'I,obelia ctrrcliiinlis. Ecology 69: 1716-1720.

DIGGLE,P K. 1995. Architectural effects and the interpretation of patterns of fruit and seed development. Aniii~nl Review oj'Ecology and Systeiiz- cities 26: 531-552.

1997. Ontogenetic contingency and floral moiphology: the effects of architecture and resource limitation. Ii~tei.national J~~rrrial of Plniit Scieric.es 158 (supplement): s99-s107.

FENSTFR, C. B., AND D. E. CAKR. 1997. Genetics of rex allocation in Mimulio (Scrophulariaceae). Jour7tril ufEvolrrrioiing; Biology 10: 641661.

GANESHAIAI-I.K. N., AND R. UMA SHAANKER. 1991. Floral sex ratios in monoecious species-why are trees more male-biased than herbs? Cirrrent Science 60: 3 19-32 1.

GEBER, M. A.. AND E. L. CHAKNOV. 1986. Sex allocation in hermaphrodites with partial overlap in maleifemale resource inputs. Joi~rizalof Theoreticiil Biology 1 18: 33-43.

GUITIAN,J., AND L. NAVARRO. 1996. Allocation of reproductive resources within inflorercences of Petrocopti.~ graridijoro (Caryophyllaceae). C(iiiatliarz Jor~rnal of Botrlii); 74: 1482-1486.

Hor.sracr~, K. E. 1985. 4 phenetic study of Clnrkia ~rrig~licirlnti~

Lindley (Onagraceae) and its relatives. Systeiilatic Biology 10: 155-165. HOUL.E. D. 1991. Genetic covariance of fitness correlates: what genetic cor- relations are made of and why it matters. Evol~rtion45: 630-648. . 1992. Comparing evolvability and variability of quantitative traits. Genetics 130: 195-204. KLINKHAMFR, 1997. Size-dependent allocation to male and feiriale

P G. L.

reproduction. In E A. Bazzaz and J. Grace reds.], Plant resource allo-
cation, 21 1-229. Acadenlic Presr, London, UK.
-,    AND T. J. DE JON(;. 1987. Plant sire and seed production in the
monocarpic perennial Cynoglossirin officiizr~le L. Nerv Plzytvlogist    106:
773-784.
-.    AND ----.    1993. Phenotypic gender in plants: effects of plant hize
and environment on allocation to seeds and flowers in Cyrioglossraii of
ficiricile. Oiko.5 67: 81-86.
--. AND H. METZ. 1997. Sex and size in cosexual plants. Ti.~iid.s iii Eculogy anil Evolrrtion 12: 260-265. KUDO, G. 1993. Size-dependent rerource allocation pattern and gender var- iation of Alzei~zone debelis Fisch. Plant Sl)ecie.\ Biology 8: 29-34.

LEHTILA,K., AND S. Y. STRA~ISS.1999. Effects of foliar herbivory on male and female reproductive traits of wild radish. Rapiiaiitrs i.ol~I?cirii.trr~cin. Ecology 80: 116-114.

LEWIS. H. 1951. The origin of supernumerary chromosomes in natural pop- ulations of Clarkia elegarz.\. Evolirtion 5: 142-157. , AND M. E.LEWIS. 1955. The genus C/nr./;icr. Urziversity of Cfl/iJbi.i!ia P~rh1icntioii.s iii Botai~y 20: 241-392.

LLOYD. D. G. 1984. Gendel- allocations in outcrossing corexual plants. Iii R. Dirzo and J. Sarukhan leds.], Perspectives in plant population ecology. 277-300. Sinauel; Sunderland, Massachusetts. US4.

-. AND K. S. BAWA. 1984. IvIodification of gender in seed plants in varying conditions. Evoli~rioizar-y Biology 17: 255-338. LOVETT DOIJST, J., AND F! N.CAVERS. 1982a. Resource allocation and gen-

May 20011 MAZERAND DAWSON-SIZE-DEPENDENT SEX ALLOCATION IN CLARKIA

der in the green dragon Arisnemn dmcontiuri~ (Araceae). Ainericniz Mid- lcind Aintilriilisf 108: 144-148.

-, AND -. 1982b. Sex and gender dynamics in jack-in-the-pulpit, Arisiienln triphyllurti (Araceae). Ecolog)) 63: 797-808.

LCNDHOLM,J. T., AND L.W. AARSSEN. 1994. Neighbour effects on gender variation in Arizbrrjsin iirtemisi$o/ia. Ciinndicir~ Jo~tnlnl of Bototzy 72: 794-800.

MALEK.S. J. 1992. En~ironmental modification of gender allocation in wild radish: consequences for sexual and natural selection. It2 Robert Wyatt [ed.], Ecology and evolution of plant reproduction: new approaches, 18 1

225. Chapman and Hall, New York, New York, USA.

, AND V. A. DEI.PSALLE. 1996a. Ternporal instability of genetic com- ponents of floral trait variation trait: maternal family and population ef- fects in Spergitlarin iiznrit~ti (Caryophyllaceae). Eiol~~tioiz50: 2509- 2515.

-. AND ---, 1996b. Floral trait variation in S~~erg~~lnrinti1nri170 (Caryophyllaceae): ontogenetic, maternal family, and population effects. Heredity 77: 269-281.

--, AND P. R. NEAL. 1999. Responses of floral traits to selection on primary sexual investment in Spergctlnrin marina: the battle between the sexes. Evoli~tio~~

53: 717-73 1. ~~~AZLI!CATO, AND G. P.SORESSI. 1999. Effect of

A,, G. TFSA, T. BIANCARI, gibberellic acid treatments, environ~nental conditions, and genetic back- ground on the expression of the pasthenocarpic fruit mutation in tomato. Protol~lasma208: 18-25.

MCKONE, M. J. 1987. Sex allocation and outcrossing rate: a test of theoret- ical predictions using bromegrass (Bromus). Evolittion 41: 591-598. MEDI<ANO,M.,l? GUITIAN.

AND J. GUITI;\N. 2000. Patterns of fruit and seed set within inflorescences of Pnncrcrtiur?l ninrititi~ctri-i (Amaryllidaceae): nonuniform pollination, resource limitation, or architectural effects? Aiiierican Journ(i1 ~/Bottitzj 87: 493-501.

MOORING,

J. S. 1958. A cytogenetic study of Clarkin utzg~iiculntn. I. Trans- locations. Airzerirnrl J~ILU~CII

of Botii~zy 45: 233-242. . 1960. A cytogenetic study of Clarkii~ unguiculata. 11. Supernumer- ary cbromosonles. Aiizericaiz Journal of Botany 47: 847-854. M~II~AN,

M. T., AND D. J. SCHOEN. 1997. The role of theory in an emerging new plant reproductive biology. Treiltl,~ in Ecology nnd Evol~~tiorz 12: 231-234.

MOSs(>P, R., hf. R. MACNAIR. AND A. W. ROBERTSON.1994. \%-ithill-popu- lation variation in sexual resource allocation in ~~Iiiricllir Filnc

gutt~~tus. tioilal Ecology 8: 410-418. O'BIIIEN, S. F! 1994. Andromonoecy aud fruit set in Lc/~to.spertiiunz mny1,si- noiiles and L. continetzfale (Myrtaceae). Austinlian .lourtzn/ of Botntiy

42: 751-762.

O'Nrlr., P.,AND J. SCHILIITT.1993. Genetic constraints 011 the independent evol~~tion

of ~iiale and female reproducti\,e characters in the tristylous plant Lythnini siilictrrin. E~~oliitioiz 47: 1457-147 1. POLICANSKY, 1981. Sex choice and the size advantage model in jack-in-

D. the-pulpit (Ariscretiz~~ Proc~editzgs of thr Ncitioniil Acadenly

tri~~lzyllunz). of Scier~ces, USA 78: 1306-1308. QUELL~R, 1984. Pollen-ov~~le ratios and hermaphrodite sexual allo-

D. C.

cation strategies. Evol~ttioiz38: 1148-1 151. RAMPAIJ,C., AND l?-H. GOUYON. 1991. Resource allocation to growth, re- production, and survival in Glnc1iolu.s:the cost of male function. Joirrr~al

of E~~olufionii~

Biology 4: 291-307. ROBERTSON,A. W., A. DIAZ, AND M. R. MACNAIR. 1994. The quantitative genetics of floral characters in Mir?~~~lus

gclttatus. Heredit,y 72: 300-3 11. ROSE, M. R., AND B. CHARLESWORTH.

1981. Genetics of life-history in Drosol~hila melatiognster. I. Sib analysis of adult females. Getletics 97: 173-

186. SAS. 1989. SASISTAT user's guide, version 6.04 edition. SAS Institute, Cary, North Carolina, USA. SCHOEN,D. J. 1982. Male reproductive effort and breeding system in an

hermaphroditic plant. Orcologia 53: 255-257. SIL~ITH-HUFRTA,

N. L., AND E C. VASEK. 1984. Pollen longevity and stigma pre-emption in Clarkia. Aiilerictirl Jourt~crl qf Botaily 71: 1183-1 191.

SOLOIL~ON.B. P. 1988. Patterns of prefertilir.atio11 and postfertilization re- source allocation within an inflorescence: evidence for interovary corn- petition. Aiizet,icnt7 Joiiriznl of Botnnj 75: 1074-1079.

SPALIK, K. 1991. On evolution of andromouoecy and 'overproduction' of flowers: a resource allocation model. Biologicti1 Jo~ournnl (I/the Litzizearz Societv 42: 325-336.

, AND S. R. J. WOODELL. 1994. Regulation of pollen production in Aiirizrisc~rs .sjlvestris, an andromonoecio~~s

species. It~terricztiorml Jo~tri~al of Plnrit Sciences 155: 750-754. STOECKLIN, 1994. Effects of plant size and morphological

J., AND P. FAVR~. constraints on variation in reproductive components in two related spe- cies of E~~ilobiuriz.

Jo~lrtzalof Ecology 82: 735-746. VAN KOOKDWIJK,

.4. J., AND G. DE JONG. 1986. Acquisition and allocation of resources: their influence on variation in life history tactics. America11 ~Vntiiriilist128: 137-142.

VASEK, E C. 1964. The evolution of Clarkin ~ti~giriculnta

derivatives adapted to relatively xeric environments. Evoliitioii 18: 26-42. , 1965. Outcrossing in natural populations. 11. Clarkicl c~i~giticularri. Evol~lrioil 19: 152-156. . 1971. Variation in marginal populations of Cln~kicl. Ecology 52: 1046-1051.

VOGLER, D. W., S. PERETZ, AND A. G. STEPHENSON. 1999. Floral plasticity in an iteroparous plant: the interactive effects of genotype, environment, and ontogeny in Cnrizpc~rrilla ri1~1uricu1oidt.s (Campanulaceae). American Jour~znl qf'Rotiiny 86: 482-494.

WATKINSON, 1982. Factors affecting the density response of l'ulpin fns-

A. ciculiitu. .loir~~rznl of Ecology 70: 149-1 61. WILLSON, M, E 1979. Sexual selection in plants. Amei,ican N~t~tralist 113: 777-790.

WOLFE, L. M. 1992. Why does the size of reproductive structures decline through time in Hjdi.o~~hj//un~ (Hydrophyllaceae)? De-

c~p~~erzn'iculc~titi?z veloprnental constraints 1,s. resource limitation. Aii~ericaii Joitnz~l (j'Bot- any 79: 1286-1290.

WRIGHT, S. I,, AND S. C. H. BARRETT. 1999. Size-dependent gender modi- fication in a hermaphroditic perennial herb. Proceediizgs of the Roynl Society of Loizdoii, Series R 266: 225-232.

Yu, D., M. KOTILAINEN, E. POLLANEN, M. MEHTO, P. ELOMAA,Y. HELAR- IUTTA,V. A. ALBERT. AND T H. TEERI. 1999. Organ identity genes and modified patterns of flower development in Gei6er.n hybriiln (Asteraceae). Plntzt Jo~tn~cil 17: 51-62.

Comments
  • Recommend Us