Biology ISSN 1435-8603

RESEARCH PAPER Life history trade-offs and evidence for hierarchical resource allocation in two monocarpic perennials G.-X. Cao1,2 & A. C. Worley3 1 Department of Forestry, Agricutural University, Yaan, 2 Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Mengla, China 3 Department of Biological Sciences, University of Manitoba, Winnipeg, Canada

Keywords ABSTRACT ; floral display; flower number; flower size; trade-off. The evolution of floral display is thought to be constrained by trade-offs between the size and number of flowers; however, empirical evidence for the trade-off is Correspondence inconsistent. We examined evidence for trade-offs and hierarchical allocation of G.-X. Cao, Department of Forestry, Sichuan resources within and between two populations each of the monocarpic perennials, Agricutural University, Yaan 625014, China. and C. giganteum. Within all populations, flower size–num- E-mail: [email protected] ber trade-offs were evident after accounting for variation in plant size. In addition, variation in flower size explained much variation in flower-level allocation to attrac- Editor tion, and female and male function, a pattern consistent with hierarchical alloca- J. Arroyo tion. However, between population differences in flower size (C. cordatum) and number (C. giganteum) were not consistent with size–number trade-offs or hierar- Received: 8 November 2011; Accepted: 19 chical allocation. The population-level difference in C. cordatum likely reflects the March 2012 combined influence of a time lag between initiation and maturation of flowers, and higher light levels in one population. Thus, our study highlights one mechanism doi:10.1111/j.1438-8677.2012.00612.x that may account for the apparent independence of flower size and number in many studies. A prediction of sex allocation theory was also supported. In C. giganteum: from one population invested more mass in pistils and less in stamens than did plants from the other population. Detection of floral trade-offs in Cardiocrinum may be facilitated by monocarpic reproduction, production of a single inflorescence and ease of measuring plant size.

ume and number per flower (Vonhof & Harder 1995; Worley INTRODUCTION & Barrett 2000; Sarkissian & Harder 2001; Yang & Guo Natural selection favours the allocation of finite resources to 2004) and seed size and number (Leishman 2001; Roff 2002). different functions in a way that maximises plant fitness By contrast, empirical evidence for trade-offs between flower (Charnov 1982). For a given amount of resources, increased size and number is mixed, with reported relationships rang- allocation to one function should necessitate a decrease to ing from positive to non-significant, to negative (Morgan another, thereby resulting in negative relations or trade-offs 1998; Burd 1999; Worley & Barrett 2000; Worley et al. 2000; between competing plant parts (Lloyd 1987; Roff 2002). Caruso 2004; Delph et al. 2004; Tomimatsu & Ohara 2006; Trade-offs that are widely assumed in theoretical treatments Sargent et al. 2007; Goodwillie et al. 2010). For example, Sar- of plant resource allocation include trade-offs between the gent et al. (2007) found a significant negative correlation size and number of repeated organs such as flowers, seeds between flower size and daily flower number in an indepen- and pollen grains (Cohen & Dukas 1990; Morgan 1993; dent contrasts analysis involving 251 species, but did not Harder & Barrett 1995; Sakai 1995, 2000; Schoen & Ashman detect such trade-offs within the genera Collinsia and Narcis- 1995), between female and male function (Parachinowitsch & sus. In some cases the sign and significance of the correlation Elle 2004; Ashman & Majetic 2006) and between current and between flower size and number varies with population and future reproduction (Charnov 1982; Charlesworth & Morgan plant age (e.g. Worley & Barrett 2001; Caruso et al. 2012). 1991). In animal-pollinated species, resource allocation at Empirical evidence is also mixed for trade-offs between flowering should depend on how flower size and number female and male function, especially at flowering (Fenster & affect pollinator attraction, pollination efficiency and the Carr 1997; Parachinowitsch & Elle 2004; Ashman & Majetic associated marginal fitness gains through female and male 2006). function, but trade-offs are expected to place an important Positive or a lack of a correlation between floral traits in constraint on the trait combinations that evolve. many studies suggests that the importance of trade-offs in The empirical evidence suggests that some trade-offs are theoretical models may be overstated. However, multiple sit- indeed ubiquitous whereas evidence for other trade-offs is uations may obscure trade-offs. (i) Within-season allocation much more inconsistent. Examples of widespread trade-offs to flower size and number may not be made from a common involving size and number include those between pollen vol- resource pool at a single point in time (Bazzaz et al. 1979).

158 Plant Biology 15 (2013) 158–165 ª 2012 German Botanical Society and The Royal Botanical Society of the Netherlands Cao & Worley Floral allocation in two moncarpic lilies

(ii) Flower size–number trade-offs may only be evident when top to bottom of the inflorescence (basipetal blooming) in lifetime flower production is considered (Morgan 1998; Wor- C. giganteum, whereas almost all flowers within inflorescences ley & Barrett 2000; but see Caruso et al. 2012). (iii) Flower- open simultaneously in C. cordatum (Cao & Kudo 2008). ing and fruiting may draw resources from the same resource Both species are self compatible, but are visited by a wide pool so that negative relations between size and number may variety of insects (Ohara et al. 2006; Cao, unpublished obser- only be apparent if seed production is included as a compo- vations). Fruits mature at the end of September; each fruit nent of investment per flower (Sato & Yahara 1999). (iv) contains several hundred seeds with thin filmy wings in both Variation in plant size (overall resource availability) or hier- species. archical allocation may mask trade-offs between the alloca- tion of resources to flower size and number, or to female Data collection and male function (Van Noordwijk & de Jong 1986; Houle 1991; de Laguerie et al. 1991; de Jong 1993; Koelewijn & We sampled 30 plants selected to span the full size range of Hunscheid 2000; Worley et al. 2000, 2003; Caruso et al. flowering plants in each of two C. cordatum populations on 2012). For example, higher allocation to flowering may allow the Campus of Hokkaido University (4305¢ N, 14120¢ E) in outcrossing taxa to produce both more and larger flowers, Sapporo, northern Japan in July 2005. One population was and also to invest more floral resources in both female and located in the understorey of a developed deciduous forest male function than do selfing taxa (Ashman & Majetic 2006; (hereafter Cpop-1). The other population (hereafter Cpop-2) Tomimatsu & Ohara 2006; Goodwillie et al. 2010). was located in a relatively open site with sparse canopy cover. In this study, we eliminated or examined three of the four Relative light intensity was 9.7% in Cpop-1 and 44.3% in situations referred to above. We studied two populations Cpop-2. When flowers began to open, three flowers repre- each of the monocarpic perennials, Cardiocrinum cordatum senting each relative position (basal, middle and distal) (Thunb.) Makino and C. giganteum (Wall.) Makino. Mono- within inflorescences were collected. Flowers were separated carpy, production of a single inflorescence and large plant into stamens, pistils and tepals. The number of flowers was size allowed us to accurately estimate lifetime allocation to counted, and plant size was estimated as a product of square flowering in a single season (situation 2). However, differen- stem diameter · stem height (d2 · h, cm3), which was highly 2 tiation of flowers in Cardiocrinum occurs in autumn so that correlated with total vegetative mass (R = 0.94, F1,19 = flower number is determined several months before flowers 298.91, P < 0.000). All reproductive components were oven are fully mature; this raises the possibility that flower size dried to a constant mass and weighed. The data on C. corda- and number depend on different resource pools (situation 1). tum have been used in a previous study (Cao & Kudo 2008), We compared two populations of C. cordatum that differed but allocation to floral display and evidence for hierarchical markedly in light availability and therefore the resources allocation were not examined. available to developing flowers. Finally, measuring plant size We sampled 29 and 28 plants of C. giganteum encompass- and within-flower allocation allowed us to assess whether ing the full size range of flowering plants in BiFengXia allocation patterns were consistent with a two-level allocation (3004¢ N, 10259¢ E) and WangYu (2945¢ N, 10256¢ E) hierarchy, in which resources were allocated first between (hereafter Gpop-1 and Gpop-2, respectively), YaAn District, flower size and number and then among attractive, female Sichuan Province, China, in mid-May 2006. All flowers and male functions within flowers (situation 4). Replicate within inflorescences on each plant were collected as they species and populations allowed us to examine the consis- opened. The inflorescence was divided into basal, middle and tency of allocation patterns within this genus. distal portions. All flowers were separated into stamens, pis- We tested the following specific predictions. (i) Flower size tils and tepals, and the average per flower mass of each com- and number both depend on the resources available for flow- ponent was estimated for each position. At the same time, ering, as indicated by plant size. (ii) For a given plant size, the flower number of each individual plant was recorded. All increases in flower size should be at the expense of flower above- and belowground vegetative parts were collected, oven number because these two components of floral allocation dried to a constant mass and weighed. draw on the same resource pool. (iii) Allocation to attraction, Preliminary analyses indicated that position within the male function and female function within flowers should inflorescence did not influence flower size. Therefore, we depend primarily on flower size rather than on plant size. averaged values from each position to estimate investment per flower (stamens+pistils+tepals) at the plant level.

MATERIAL AND METHODS Statistical analyses Study species Comparison between species and populations within species Cardiocrinum is a genus of three species of bulbous plants of We initially used nested anova to compare floral traits the family . They are native to the , mon- between species and between populations nested within spe- tane China and Japan. Both C. cordatum and C. giganteum cies. These comparisons of floral traits did not take variation are monocarpic perennial herbs. Mature plants are 0.6–1.5-m in plant size into account; our indices of plant size differed tall, have several large cordate cauline leaves, and the inflores- between species and therefore could only be compared cence is a single robust raceme. Flowers are 7–10 cm in between populations. Species and population were both anal- length and open in mid- to late May for C. giganteum and ysed as random effects because we had no a priori expecta- mid- to late July for C. cordatum. Flowers consist of six tions about differences between the two species, or between tepals, six stamens and one pistil; they open sequentially from populations of C. giganteum. Variables were log transformed

Plant Biology 15 (2013) 158–165 ª 2012 German Botanical Society and The Royal Botanical Society of the Netherlands 159 Floral allocation in two moncarpic lilies Cao & Worley to stabilise variances. To facilitate interpretation of the depend primarily on overall flower size rather than on plant results, we present back-transformed EMM (estimated mar- size (prediction 3). Therefore, we analysed each aspect of flo- ginal mean) and values that are one standard error below ral allocation as a function of population, flower mass and (LSE = lower standard error) or above (USE = upper stan- plant size. For each analysis, initial models included popula- dard error) the mean value. tion as a random effect, one or two covariates and all possi- We eliminated two plants a priori that stood out as outliers ble two-way interactions involving covariates. Non-significant in preliminary plots of raw data. Plant 4 from Gpop-2 had interactions were dropped from the model using backwards missing data for within-flower allocation and unusually heavy elimination. We indicate partial regression coefficients with flowers, and plant 28 from Gpop-2 was exceptionally small. the letter b and their standard errors with sb. Partial eta 2 The effects of other outliers are described in the results. squared (gp) values were used to evalute the relative amount of variance accounted for by each factor ⁄ covariate. Flower number and size For all analyses, we examined residuals for homogeneity of At the first level of the hierarchy, we were interested in variances and departures from normality. All analyses were conducted using spss 17.0, except the nested anovas, which whether allocation to flower number and size depended on sas plant size (prediction 1) and whether trade-offs were appar- were analysed using PROC GLM in 9.2. ent after accounting for variation in plant size (prediction 2). We analysed the two species separately because the indices of RESULTS plant size were different, but were able to compare floral traits between populations while accounting for variation in Comparison between species and populations within species plant size. We used analyses of covariance to analyse each of No floral traits differed significantly between species, and flower number and flower size as a function of population, mean plant size and flower number was also similar between plant size and the other component of floral display. How- populations of each species (Table 1). However, all floral ever, our analysis of flower size in C. cordatum included only traits had greater mass in open Cpop-2 than in shaded plant size as a covariate due to the high correlation between Cpop-1. Flower, stamen and tepal mass were all higher in plant size and flower number in this species. Inclusion of Gpop-1 than in Gpop-2 (Table 1). However, pistil mass was both flower number and plant size as covariates resulted in a significantly lower in Gpop-1 (Table 1). These population- variance inflation factor of seven, indicating the presence of level differences in sex allocation were consistent with a strong collinearity. trade-off between female and male function. For each analysis, initial models included population as a random effect, one or two covariates and all possible two-way interactions involving covariates. Non-significant interactions Flower number and size were dropped from the model using backwards elimination. Flower number was positively related to plant size within We indicate partial regression coefficients with the letter ‘b’ populations of both species (Table 2, Figs 1A and 2A). This and their standard errors with sb. These coefficients indicate size-dependence matched our prediction of positive relation- the responses of the dependent variable to one unit change ships between resource status and allocation to flowering. In in a specific independent variable, while all other indepen- C. giganteum, plants from Gpop-1 produced significantly dent variables remain constant. more flowers across all plant sizes (Table 2, Fig. 1A). For plants of a given size, flower number was always nega- Allocation within flowers tively related to flower mass (Table 2, Figs 1C and 2C), as we If resource allocation is hierarchical, allocation to female predicted based on the assumption that these traits draw function (pistil mass), male function (stamen mass) or polli- from the same resource pool. The slope of the flower nator attraction (tepal mass) at the flower level should mass effect was less for C. cordatum (b = )0.32) than for

Table 1. Mean values for plant size and floral traits in two populations each of Cardiocrinum cordatum (Cpop-1, Cpop-2) and C. giganteum (Gpop-1, Gpop-2). All data were natural log transformed prior to analysis of species and population differences. No comparisons between species were significant and P-values refer to population-level comparisons within each species. For ease of interpretation, back-transformed means are given. Values in parenthe- ses correspond to lower (LSE) and upper standard errors (USE).

Cpop-1 (N = 30) Cpop-2 (N = 30) Gpop-1 (N = 29) Gpop-2 (N = 26) trait mean (LSE-USE) mean (LSE-USE) mean (LSE-USE) mean (LSE-USE) plant sizea 1387 (1251–1538) 1553 (1404–1718) 61911 (58817–65168) 69465 (65324–73869) flower number 12 (11.0–13.1) 11 (10.7–12.2) 12 (11.2–12.1) 11 (10.4–11.6) flower mass(mg) 478 (463–493) 663 (649–677)*** 604 (588–620) 531 (520–542)*** stamen mass(mg) 72 (70–74) 86 (84–89)*** 87 (85–89) 79 (77–81)** pistil mass(mg) 97 (94–101) 116 (113–120)*** 50 (48–51) 58 (56–59)*** tepal mass(mg) 307 (296–318) 459 (447–470)*** 454 (442–467) 395 (384–403)***

**P < 0.01, ***P < 0.001 a Plant size is in cm3 for C. cordatum and in mg for C. giganteum, no between species comparison possible.

160 Plant Biology 15 (2013) 158–165 ª 2012 German Botanical Society and The Royal Botanical Society of the Netherlands Cao & Worley Floral allocation in two moncarpic lilies

Table 2. Factors affecting flower size and flower number in Cardiocrinum cordatum and C. giganteum. Separate analyses of covariance were conducted for each species. Initial models included applicable covariates and all possible two-way interactions involving covariates. Non-significant effects were removed using backwards elimination.

Cardiocrinum cordatum effect flower number flower sizea flower number flower size population F1,56 = 0.23 F1,57 = 73.67*** F1,51 = 38.24*** F1,49 = 5.80* plant size F1,56 = 496.94*** n.s. F1,51 = 164.86*** F1,49 = 48.32*** plant size · population n.s. n.s. n.s. F1,49 = 6.98*

flower size F1,56 = 6.65* – F1,51 = 30.57*** –

flower number – n.s. – F1,49 = 35.76***

flower number · population n.s. n.s. n.s. F1,49 = 4.64* model R2 0.90 0.57 0.77 0.61 n.s. = not significant, *P < 0.05, ***P < 0.001, –effect not applicable a Plant size and flower number was highly correlated; therefore, their influence on flower size was considered in two separate analyses.

C. giganteum (b = )0.83). The steep slope for C. giganteum a small proportion of the variation in tepal mass, and was in part due to four plants in Gpop-1 with very large and remained significant only in C. cordatum once variation in relatively few flowers (Fig. 2A–C). Although these plants had plant size and flower mass were taken into account (Table 3). noticeably larger flowers in plots of raw data, their strong As predicted, within-population variation in pistil mass influence on the slope estimates meant that they did not (female allocation) was best explained by flower mass for 2 stand out as outliers in Fig. 2. However, the slope of the both species (gp = 0.36), whereas effects of plant size were 2 flower mass effect still remained quite high (b ± sb = comparatively small (gp = 0.1; Table 3). For a given plant )0.68 ± 0.190) when these four plants were omitted from the size, the relationship between pistil mass and flower mass was analyses. Interestingly, all raw correlations between flower similar across species and populations (C. cordatum: number and mass were non-significant (C. cordatum: both b±sb = 0.72 ± 0.129; C. giganteum: b ± sb = 0.67 ± 0.125). P > 0.35; C. giganteum: both P > 0.85). Thus, the trade-off Population effects on pistil mass (female function) were non- was only apparent when variation in plant size (resource sta- significant in C. cordatum but were significant in C. gigante- 2 tus) was also taken into account. um with gp = 0.45 (Table 3). Thus allocation to female func- In C. giganteum, flower mass within each population was tion remained lower in Gpop-1 (EMM = 48 mg, LSE =46 positively related to plant size and negatively related to flower mg, USE = 50 mg) than in Gpop-2 (EMM = 59 mg, LSE = number, as we predicted (Table 2, Fig. 2B and C). The slopes 57 mg, USE = 62 mg) even after differences in plant and of both relationships were much steeper in Gpop-1 (Table 2, flower mass were taken into account. Fig. 2C). Omission of the four plants with very large and rel- Stamen mass (male allocation) in C. cordatum followed a atively few flowers resulted in similar slopes for the popula- similar pattern to tepal and pistil mass with respect to the ) 2 tions (flower size effect: b ± sb = 0.31 ± 0.088). The relative influence of flower (gp = 0.72) and plant size 2 variance inflation factor for the analysis including both flower (gp = 0.001), despite significant interactions between each number and plant size was below five, indicating the absence variable and population (Table 3). This result was confirmed of strong collinearity in analyses of C. giganteum. when we analysed stamens as a function of flower mass and We did not examine the simultaneous effects of plant size plant size for each population separately. The effect of flower and flower number on flower mass in C. cordatum due to mass within populations was much larger (Cpop-1:F1,27 = 2 concerns about collinearity (see Materials and Methods). 94.84, P < 0.001, gp = 0.78; Cpop-2:F1,27 = 56.74, P < 0.001, 2 When these covariates were considered separately, neither of gp = 0.68) than that of plant size (Cpop-1:F1,27 = 5.07, P = 2 2 them influenced flower mass (Table 2). However, flower mass 0.033, gp = 0.16; Cpop-2:F1,27 = 3.82, P = 0.061, gp = 0.24). was 1.4 times higher in open Cpop-2 than in shaded Cpop1 For a given plant size, stamen mass was positively related to (Table 2, Fig. 1B). flower mass in each population, but the slope was higher for open Cpop-2 (b ± sb = 1.06 ± 0.140) than for shaded Cpop- 1(b±s = 0.70 ± 0.072). Allocation within flowers b In contrast to other analyses, none of flower size, plant size Based on the assumption of hierarchical allocation, we pre- or population explained much variation in stamen mass in 2 dicted that allocation to attraction, female and male function C. giganteum. All three variables had gp < 0.10 (Table 3). For at the level of individual flowers would primarily reflect a given plant size, the relationship between stamen mass and flower mass rather than plant size. Both flower mass and flower mass was similar for both populations (b ± sb = plant size had significant effects on tepal mass (attraction) in 0.30 ± 0.138). Plant size did not affect stamen mass signifi- 2 each species, but flower mass alone accounted for the highest cantly, although its gp came close to that of flower mass; we proportion of the variation within populations (both retained plant size in the model for comparison with C. cord- 2 gp > 0.94; Table 3). The relationship between tepal mass and atum (Table 3). A significant population effect indicated that flower mass was similar for both populations of each species allocation to male function remained higher in Cpop-1 (C. cordatum:b±sb = 1.13 ± 0.035; C. giganteum:b±sb = (EMM = 86 mg, LSE = 83 mg, USE = 90 mg) and lower in 1.18 ± 0.032). Differences between populations accounted for Gpop-2 (EMM = 80 mg, LSE = 76 mg, USE = 84 mg), after

Plant Biology 15 (2013) 158–165 ª 2012 German Botanical Society and The Royal Botanical Society of the Netherlands 161 Floral allocation in two moncarpic lilies Cao & Worley

A A

B B

C C

Fig. 1. A: Relationships between flower number and plant size

(b ± sb = 0.73 ± 0.033). B: No relationship between flower mass and plant Fig. 2. A: Relationships between plant size and flower number size. C: Relationships between flower mass and flower number (b ± sb = (b ± sb = 0.79 ± 0.062), and B: flower mass (Pop-1: b ± sb = 0.64 ± )0.32 ± 0.125) in Cardiocrinum cordatum. Data presented in the figures 0.095; Pop-2: b ± sb = 0.29 ± 0.093). C: Relationships between flower were adjusted for the effects of other covariates. First, a predicted value mass and flower number (Pop-1: b ± sb = )0.65 ± 0.122; Pop-2: b ± was calculated for each observation based on the intercepts and partial sb = )0.31 ± 0.103) in Cardiocrinum giganteum. Data presented in the regression coefficients from the mixed model, the observed value of the figures were adjusted for the effects of other covariates. First, a predicted covariate of interest and the means of the other covariates. Then the value was calculated for each observation based on the intercepts and residual from the mixed model for each observation was added to its pre- partial regression coefficients from the mixed model, the observed value dicted value. of the covariate of interest and the means of the other covariates. Then the residual from the mixed model for each observation was added to its predicted value. differences in plant and flower mass were taken into account (Table 3), the reverse of the pattern observed for female function. Koelewijn & Hunscheid 2000). Positive phenotypic correla- tions between flower number and plant size are well docu- DISCUSSION mented (Herrera 1991; Morgan 1998; Burd 1999; Sato & Flower size and number Yahara 1999; Worley et al. 2000). For simplicity, we predicted that flower size should also The mutual dependence of flower size and number on plant increase with plant size because stored resources should size (stored resources) in C. giganteum was consistent with influence floral allocation. However, theoretical expectations the expectation that investment in both traits depends on the and empirical patterns for flower size are more complex resources available for flowering at a given point in time than for flower number. Flower size is positively related to (Van Noordwijk & de Jong 1986; de Laguerie et al. 1991; plant size in several species (Morgan 1998; Burd 1999; Wor-

162 Plant Biology 15 (2013) 158–165 ª 2012 German Botanical Society and The Royal Botanical Society of the Netherlands Cao & Worley Floral allocation in two moncarpic lilies

Table 3. Factors affecting per flower allocation to attraction, female function and male function in Cardiocrinum cordatum and C. giganteum. Separate analyses of covariance were conducted for each species. Initial models included applicable covariates and all possible two-way interactions involving covari- ates. Non-significant effects were removed using backwards elimination.

2 2 2 species & effect attraction(tepal mass) gp female function(pistil mass) gp male function(stamen mass) gp

Cardiocrinum cordatum

population F1,56 = 4.45* 0.074 F1,56 = 1.45 0.025 F1,54 = 9.62** 0.151

flower size F1,56 = 1049.08*** 0.949 F1,56 = 31.03*** 0.357 F1,54 = 136.45*** 0.716

plant size F1,56 = 5.43* 0.088 F1,56 = 4.96* 0.081 F1,54 = 0.04 0.001

flower size · population n.s. n.s. F1,54 = 5.50* 0.092

plant size · population n.s. n.s. F1,54 = 8.57** 0.137 model R2 0.98 0.51 0.81 Cardiocrinum giganteum

population F1,51 = 1.84 0.035 F1,51 = 41.21*** 0.447 F1,51 = 5.18* 0.092

flower size F1,51 = 1363.51*** 0.964 F1,51 = 28.54*** 0.359 F1,51 = 4.72* 0.085

plant size F1,51 = 8.12** 0.137 F1,51 = 5.65* 0.100 F1,51 = 3.34 0.058 model R2 0.97 0.61 0.31 n.s = not significant, *P < 0.05, **P < 0.01,***P < 0.001 ley et al. 2000) but not in others (Sato & Yahara 1999; Although patterns of resource allocation within popula- Iwaizumi & Sakai 2004; C. cordatum, this study). Early tions were consistent with expectations based on size–number models of size–number trade-offs predicted that optimal off- trade-offs and the assumption that floral resources were allo- spring size should be independent of resource status because cated from a common pool at a single point in time, differ- the increase in fitness associated with producing an addi- ences between populations were not. C. cordatum plants from tional offspring was assumed to remain the same, regardless the open habitat produced flowers 1.4 times larger than those of how many offspring were produced (Smith & Fretwell from the understorey without reducing flower number. This 1974; Lloyd 1987). However, negative interactions among was presumably because the photosynthates available to flowers, such as geitonogamy, pollen discounting or compe- developing flowers were increased in the open habitat (Cao & tition for resources, may reduce the advantage of producing Kudo 2008). In C. cordatum, 13C tracer experiments have additional flowers relative to the advantage of producing shown that photosynthetic products produced by leaves are larger flowers and therefore introduce a positive relationship mostly transferred to the inflorescence (G Kudo, unpublished between plant size and flower size (Venable 1992). Although observations). In contrast to the effect of light levels in flower size did depend on plant size in C. giganteum, within C. cordatum, additional resources (nutrients) increased flower population variation in flower size in both species number rather than flower size in Plantago coronopus (Koe- (CV = 1.6–2.8%) was much lower than variation in flower lewijn & Hunscheid 2000) and Fragaria virginiana (Bishop number (CV = 8.8–19.2%), indicating that flower size may et al. 2010), both perennial species with indeterminate flow- be less sensitive to changes in resource status than flower ering. number (Smith & Fretwell 1974; Lloyd 1987). In C. giganteum, large plants from population one pro- As predicted, trade-offs between flower size and number duced larger flowers than plants of a similar size from popu- were apparent within populations of Cardiocrinum when we lation two. This increase in flower size appeared to occur at controlled for variation in plant size. Although accounting the expense of flower number. However, plants from popula- for plant size often makes the relationship between flower tion one also produced an average of three more flowers than size and number less positive, it has not always revealed phe- plants with the same vegetative mass (plant size) and flower notypic trade-offs (Morgan 1998; Burd 1999; Worley & Bar- mass from population two. Differences between habitats were rett 2000; Worley et al. 2000; Caruso et al. 2012). Identifying not readily apparent at the time of sampling, but defoliation an appropriate measure of plant size may be problematic in experiments on C. giganteum indicate that current photosyn- polycarpic species due to trade-offs between reproduction thates contribute significantly to flower size but do not affect and vegetative growth (Zhang & Wang 1994). For example, a flower number (G. Cao, unpublished observations), presum- phenotypic trade-off between flower number and investment ably because all flowers differentiate prior to the growing sea- per flower occurred within selfing and outcrossing varieties son (preformation). Thus, the proximate cause of the of the annual Impatiens hypophylla (Sato & Yahara 1999). population-level difference in flower number remains unclear, However, annual life histories did not increase the incidence but seems unlikely to involve differences in current photosyn- of negative correlations between flower size and number thate production. within 32 Mimulus guttatus populations, or within lineages native to California, USA (Caruso et al. 2012). The fact that Allocation within flowers trade-offs between flower size and number were readily apparent in Cardiocrinum species likely reflects several aspects Variation in within flower allocation in both species was con- of their biology, including monocarpic reproduction, produc- sistent with expectations based on hierarchical allocation; tion of a single inflorescence and ease in measuring plant and most variation in attraction and primary sexual allocation flower size. was directly related to flower size. This result must be inter-

Plant Biology 15 (2013) 158–165 ª 2012 German Botanical Society and The Royal Botanical Society of the Netherlands 163 Floral allocation in two moncarpic lilies Cao & Worley preted with some caution given that the separate components Our study of two Cardiocrinum species revealed consistent of floral allocation were summed to estimate flower size. phenotypic evidence for a trade-off that is expected but infre- Nonetheless, in both species, flower size explained variation quently demonstrated. Our ability to detect trade-offs of in per flower allocation to primary sexual allocation that was flower size and number within populations of each Cardiocri- not explained by plant size alone. Positive correlations among num species was likely enhanced by monocarpic reproduc- primary sex functions and attraction within flowers are com- tion, the ease of obtaining accurate estimates of plant size, mon (Parachinowitsch & Elle 2004; Ashman & Majetic 2006) and the fact that light levels were relatively uniform within and may often reflect the effects of resource availability on populations. Resource allocation appeared to be hierarchical flower size. In this study, the influences of flower size on at the time of floral differentiation. However, photosynthesis allocation to tepals and pistils in both species, and stamens during flowering allowed increases in flower size without a in C. cordatum, were largely similar between populations and reduction in flower number, and thus masked trade-offs of species. This occurred even though the contribution of stored flower size and number between populations of C. cordatum. versus current photosynthates to flower mass differed between Our study gives insight into the flexibility of resource alloca- populations of C. cordatum, and some C. giganteum plants tion in species with a time lag between differentiation and differed markedly from the rest of the population in their maturation of flowers, and demonstrates one mechanism that allocation to flower size and number. The responsiveness of may account for the apparent independence of flower size floral components to resource status in Cardiocrinum is not and number in many studies. The capacity for dynamic surprising given that flowers are preformed and reproduction adjustment of resource allocation is likely to be widespread is monocarpic. By contrast, nutrient additions increased (Diggle 1997; Worley & Harder 1999). Further work is flower number but did not affect per flower gynoecium or needed to determine how developmental limits on the alloca- androecium mass in Plantago coronopus, a perennial species tion of newly acquired resources affect genetic variation in with indeterminate flowering (Koelewijn & Hunscheid 2000). resource allocation patterns. Interestingly, we found lower allocation to female function and higher allocation to male function in one population of ACKNOWLEDGEMENTS C. giganteum compared to the other. The differences between male and female allocation were evident whether or not vari- We thank Chris Caruso for comments on an earlier version ation in plant size and flower size were taken into account. of the manuscript. This project was supported by the Evidence for trade-offs between male and female functions at National Natural Science Foundation of China (30870388) flowering in simultaneous hermaphrodites has also been and Creative Group Programme for National ‘211 Project’ in inconsistent (Parachinowitsch & Elle 2004; Ashman & Majetic Sichuan Agricultural University, and a Discovery Grant to 2006; but see Mazer et al. 1999), so this pattern warrants A. Worley from the Natural Sciences and Engineering further investigation with more populations. Research Council of Canada.

Proceedings of the Royal Society of London B, 332, Houle D. (1991) Genetic covariance of fitness corre- REFERENCES 91–102. lates: what genetic correlations are made of and Charnov E.L. (1982) The theory of sex allocation. why it matters. Evolution, 45, 630–648. Ashman T.-L., Majetic C.J. (2006) Genetic constraints Princeton University Press, Princeton, NJ, USA. Iwaizumi G.M., Sakai S. (2004) Variation in flower on floral evolution: a review and evaluation of pat- Cohen D., Dukas R. (1990) The optimal number of biomass among nearby populations of Impatiens terns. Heredity, 96, 343–352. female flowers and the fruits-to-flowers ratio in textori (Balsaminaceae): effects of population plant Bazzaz F.A., Carlson R.W., Harper J.L. (1979) Contri- plants under pollination and resource limitation. densities. Canadian Journal of Botany, 82, 563–572. bution to the reproductive effort by photosythesis American Naturalist, 135, 218–241. de Jong G. (1993) Covariances between traits deriving of flowers and fruits. Nature, 279, 554–555. Delph L.F., Gehring J.L., Frey F.M., Arntz A.M., Levri from successive allocations of a resource. Functional Bishop E., Spigler R.B., Ashman T.-L. (2010) Sex-allo- M. (2004) Genetic constraints on floral evolution Ecology, 7, 75–83. cation plasticity in hermaphrodites of sexually in a sexually dimorphic plant revealed by artificial Koelewijn H.P., Hunscheid M.P.H. (2000) Intraspe- dimorphic Fragaria virginiana (Rosaceae). Botany, selection. Evolution, 58, 1936–1946. cific variation in sex allocation in hermaphroditic 88, 231–240. Diggle P.K. (1997) Extreme preformation in alpine Polyg- Plantago coronopus (L.). Journal of Evolutionary Burd M. (1999) Flower number and floral compo- onum viviparum: an architectural and developmental Biology, 13, 302–315. nents in ten angiosperm species: an examination of analysis. American Journal of Botany, 84, 154–169. de Laguerie P., Olivieri I., Atlan A., Gouyon P.H. assumptions about trade-offs in reproductive evolu- Fenster C.B., Carr D.E. (1997) Genetics of sex alloca- (1991) Analytic and simulation models predicting tion. Biological Journal of the Linnean Society, 68, tion in Mimulus (Scrophulariaceae). Journal of Evo- positive genetic correlations between traits linked 579–592. lutionary Biology, 10, 641–661. by trade-offs. Evolutionary Ecology, 5, 361–369. Cao G.-X., Kudo G. (2008) Size-dependent sex allo- Goodwillie C., Sargent R.D., Eckert C.G., Elle E., Leishman M.R. (2001) Does the seed size ⁄ number cation in a monocarpic perennial herb, Cardiocri- Geber M., Johnston M.O., Kalisz S., Moeller D.A., trade-off model determine plant community struc- num cordatum (Liliaceae). Plant Ecology, 194, Ree R., Vallejo-Marin M., Winn A. (2010) Corre- ture? An assessment of the model mechanisms and 99–107. lated evolution of mating system and floral display their generality. Oikos, 93, 294–302. Caruso C.M. (2004) The quantitative genetics of floral traits in flowering plants and its implications for Lloyd D.G. (1987) Selection of offspring size at inde- traits variation in Lobelia: potential constraints on the distribution of mating system variation. New pendence and other size-versus-number strategies. adaptive evolution. Evolution, 58, 732–740. Phytologist, 185, 311–321. American Naturalist, 129, 800–817. Caruso C.M., Maherali H., Benscoter A.M. (2012) Harder L.D., Barrett S.C.H. (1995) Mating cost of Mazer S.J., Delesalle V.A., Neal P.R. (1999) Responses Why are trade-offs between flower size and number large floral displays in hermaphrodite plants. Nat- of floral traits to selection on primary sexual infrequently demonstrated? A test of three hypothe- ure, 373, 12–515. investment in Spergularia marina: the battle ses. International Journal of Plant Sciences, 173, Herrera C.M. (1991) Dissecting factors responsible between the sexes. Evolution, 53, 717–731. 26–35. for individual variation in plant fecundity. Ecology, Morgan M.T. (1993) Fruit to flower ratios and trade-offs Charlesworth D., Morgan M.T. (1991) Allocation of 72, 1436–1448. in size and number. Evolutionary Ecology, 7, 219–232. resources to sex allocation in flowering plants.

164 Plant Biology 15 (2013) 158–165 ª 2012 German Botanical Society and The Royal Botanical Society of the Netherlands Cao & Worley Floral allocation in two moncarpic lilies

Morgan M.T. (1998) Female fertility per flower and Sato H., Yahara T. (1999) Trade-offs between flower direct and correlated responses to selection on trade-offs between size and number in Claytonia number and investment to a flower in selfing and flower size and number. Evolution, 54, 1533–1545. virginica (Portulacaceae). American Journal of Bot- outcrossing varieties of Impatiens hypophylla (Bal- Worley A.C., Barrett S.C.H. (2001) Evolution of floral any, 85, 1231–1236. salminaceae). American Journal of Botany, 86, display in Eichhornia paniculata (Pontederiaceae): Ohara M., Narumi T., Yoshizane T., Okaysu T., 1699–1707. genetic correlations between flower size and num- Masuda J., Kawano S. (2006) Life-history mono- Schoen D.J., Ashman T.-L. (1995) The evolution of ber. Journal of Evolutionary Biology, 14, 469–581. graphs of Japanese plants: Cardiocrinum cordatum floral longevity: resource allocation to maintenance Worley A.C., Harder L.D. (1999) Consequences of (Thunb.) Makino (Liliaceae). Plant Species Biology, versus construction of repeated parts in modular preformation for dynamic resource allocation by a 21, 201–207. organisms. Evolution, 49, 131–139. carnivorous herb, Pinguicula vulgaris (Lentibularia- Parachinowitsch A.L., Elle E. (2004) Variation in sex Smith C.C., Fretwell S.D. (1974) The optimal balance ceae). American Journal of Botany, 86, 1136–1145. allocation and male–female trade-offs in six popu- between size and number of offspring. American Worley A.C., Baker A.M., Thompson J.D., Barrett lations of Collinsia parviflora (Scrophulariaceae). Naturalist, 108, 499–506. S.C.H. (2000) Floral display in Narcissus: variation American Journal of Botany, 91, 1200–1207. Tomimatsu H., Ohara M. (2006) Evolution of hierar- in flower size and number at the species, popula- Roff D.A. (2002) Life history evolution. Sinauer, Sun- chical floral resource allocation associated with tion, and individual levels. International Journal of derland, MA, USA. mating system in an animal-pollinated hermaphro- Plant Sciences, 161, 69–79. Sakai S. (1995) Evolutionarily stable selfing rates of ditic herb, Trillium camschatcense (Trilliaceae). Worley A.C., Houle D., Barrett S.C.H. (2003) Conse- hermaphroditic plants in competing and delayed American Journal of Botany, 93, 134–141. quences of hierarchical allocation for the evolution selfing modes with allocation to attractive struc- Van Noordwijk A.J., de Jong G. (1986) Acquisition of life history traits. American Naturalist, 161, 153– tures. Evolution, 49, 557–564. and allocation of resources: their influence on vari- 167. Sakai S. (2000) Size–number trade-off and allocation ation in life history tactics. American Naturalist, Yang C.-F., Guo Y.-H. (2004) Pollen size–number to flower production in animal-pollinated plants. 128, 137–142. trade-off and pollen–pistil relationships in Pedicu- Plant Species Biology, 15, 261–268. Venable D.L. (1992) Size–number trade-offs and the laris (Orobanchaceae). Plant Systematics and Evolu- Sargent D.R., Goodwillie C., Kalisz S., Ree R.H. variation of seed size with plant resource status. tion, 247, 177–185. (2007) Phylogenetic evidence for a flower size and American Naturalist, 140, 287–304. Zhang D.-Y., Wang G. (1994) Evolutionarily stable number trade-off. American Journal of Botany, 94, Vonhof M.J., Harder L.D. (1995) Size–number trade- reproductive strategies in sexual organisms: an inte- 2059–2062. offs and pollen production by papilionaceous grated approach to life history evolution and sex Sarkissian T.S., Harder L.D. (2001) Direct and indirect legumes. American Journal of Botany, 82, 230–238. allocation. American Naturalist, 144, 65–75. responses to selection on pollen size in Brassica rapa Worley A.C., Barrett S.C.H. (2000) Evolution of floral L. Journal of Evolutionary Biology, 14, 456–468. display in Eichhornia paniculata (Pontederiaceae):

Plant Biology 15 (2013) 158–165 ª 2012 German Botanical Society and The Royal Botanical Society of the Netherlands 165