Sex expression, gender distribution and sex lability

To investigate patterns of sex expression at the plant and flower level, and to quantify the distribution of gender at the population level, we studied seven populations (ABG, BHL, BTF, SG1, SG2, XZ1, XZ2) for the durations indicated in the preceding section. In each population, we randomly sampled 71–401 individuals and counted the total number of flowers, their sexual condition (hermaphroditic or male) and their position on an inflorescence (basal flower 1, next 2, etc.).

To quantify the distribution of gender at the population level we employed Lloyd's measure of phenotypic gender (G_i) using data on the sexual condition of flowers (see Lloyd, 1980a; Lloyd and Bawa, 1984). This index depicts the standardized phenotypic femaleness of a plant in a population as:

where o_i is the maternal investment, p_i is the paternal investment and E is an equivalence factor based on the ratio of investments in maternal and paternal functions in the population as a whole: E = Σo_i/Σp_i. We estimated paternal investment as the number of staminate flowers (the sum of hermaphroditic and male flowers) and maternal investment as the number of ovule-bearing flowers (hermaphroditic flowers). G_i ranges from 0 for plants that produce only pollen to 1 for plants that produce only ovules.

To determine whether L. apertum exhibits labile sex expression over flowering seasons, we monitored the sexual condition of flowers on marked plants for two or three consecutive years (2009–2011) in populations ABG and SG2, respectively, using plastic tags (sample sizes given in Table 1). Each year, we recorded plant height and the number of male and hermaphroditic flowers. We then determined the direction and frequency of annual transitions between the three sex phenotypes (plants with hermaphroditic flowers only, hereafter H; plants with hermaphrodite and male flowers, hereafter H+M; and plants with male flowers only, hereafter M).

Differences between hermaphrodite and male flowers

To investigate potential size differences between hermaphroditic and male flowers we randomly sampled 73 and 66 flowers, respectively, in population ABG, and 23 and 32 flowers, respectively, in population SG2. To control for position effects, only the basal flower within an inflorescence was sampled from each plant. We used digital calipers to measure the radius of flowers. In populations ABG, BTF and SG2 we randomly collected 20–50 flowers per flower type using the same sampling scheme and these were oven dried at 80 °C for 24 h, and then weighed to the nearest 0·01 mg. In population ABG, we estimated pollen production of hermaphroditic and male flowers following methods detailed by Dafni (1992). To test the prediction that hermaphroditic flowers would be larger, weigh more and produce more pollen than male flowers, one-tailed t-tests were used to compare each parameter (flower radius, biomass and pollen production) between flower types within each population. We also used two-way ANOVA to analyse the first two traits with ‘Flower type’ and ‘Population’ as main effects.

Comparisons of hermaphroditic and male flowers

Hermaphroditic flowers were significantly larger than male flowers, based on measurements of flower radius (Fig. 4A: ABG: hermaphroditic 43·5 ± 0·5 mm, n = 73; male 41·6 ± 0·5 mm, n = 66; t ratio = 2·518, P = 0·0065 < 0·05; SG2: hermaphroditic 38·5 ± 0·8 mm, n = 23; male 34·4 ± 0·5 mm, n = 32; t ratio = 3·913, P = 0·0001). The effects of flower type and population on flower size were significant, but the interaction ‘population × flower type’ was not (P = 0·092 > 0·05). Not unexpectedly, the dry biomass of hermaphroditic flowers was significantly greater than that of male flowers (Fig. 4B: ABG: hermaphroditic 86·1 ± 2·8 mg, n = 39; male 64·2 ± 2·5 mg, n = 48; t = 5·801, P < 0·0001; BTF: hermaphroditic 90·7 ± 4·1 mg, n = 24; male 55·3 ± 4·4 mg, n = 21; t = 5·835, P < 0·0001; SG2: hermaphroditic 83·9 ± 4·3 mg; n = 27; male 54·5 ± 3·7 mg; n = 36; t = 5·200, P < 0·0001). Flower type had a significant effect on floral biomass, but the interaction ‘population × flower type’ was not significant (P = 0·163 > 0·05). In population ABG, hermaphroditic flowers produced significantly more pollen (32·5 %) than male flowers (hermaphroditic 329 400 ± 17 398, n = 19 flowers; male 248 522 ± 14 873, n = 26 flowers; t = 3·534, P = 0·0005).

Open in a separate window

Fig. 4.

Differences in floral traits between hermaphroditic and male flowers of Lilium apertum. (A) Flower radius, (B) flower biomass and (C) pollen production per flower. *P < 0·05; **P < 0·001; ***P < 0·0001. Population codes are indicated.

Dynamics of gender expression

By investigating marked individuals in populations ABG and SG2 for consecutive years, we found that individuals are able to switch sex from one year to the next (Table 1). Of the six possible annual transitions between sex phenotype classes, all except the transition from H plants to H+M plants were observed. For example, in ABG 35·1 % of H plants in year 1 produced only male flowers (M) the following year, and 42·6 % of H+M plants became M plants in the second year of observation. Similarly, among M plants in year 1, 12·6 and 9·3 % of plants were H+M or H, respectively, the following year. Changes in gender between years were also evident in population SG2. For example, among M plants observed in the first year of observation, 16·2 and 25·8 % of the plants in the following year were either H+M or H, respectively.

In both populations a significant proportion (39·7 % in ABG, 36·4 % in SG2) of flowering plants surveyed in the first year did not flower in the second year. The three sex phenotypes differed with respect to their likelihood of not flowering in the second year of observation. H plants were the least likely to flower, followed by H+M plants, and M plants were the most likely to continue flowering in the second year. Interestingly, with one exception among the sex phenotype (H plants in SG2) the proportion of plants switching gender was lowest among M plants with 42·4 and 29·0 % of plants in ABG and SG2, respectively, remaining M in the second year of observation. Of the 70 and 16 plants that were observed for three consecutive years in ABG and SG2, 17·1 and 18·8 %, respectively, changed sex twice.

The origin, distribution and consequences of dicliny for sexual systems

The origin of unisexual flowers from hermaphroditic flowers broadens the spectrum of sexual systems available to flowering plants, and increases the opportunity for variation in sex allocation and gender strategies. Our study indicates that rather than being of infrequent occurrence, as is the case in most hermaphroditic species, unisexual flowers are a common feature of L. apertum populations. Indeed, the mean proportion of male flowers sampled in our survey was 0·56, with male flowers the most frequent flower type in the majority of populations.

Loss of organ function causing floral unisexuality can occur at all stages of development, ranging from the absence of initiation to the failure of organ maturation, but within a species the stage of organ abortion tends to be similar for both pistillate and staminate flowers (Diggle et al., 2011; Fig. 2). Virtually all male flowers of L. apertum displayed conspicuous non-functional pistils (pistillodes), indicating that the processes responsible for the termination of female function were often late in development, perhaps precluding significant opportunities for reproductive compensation. Indeed, there was no evidence of increased pollen production in male flowers, which were in fact smaller and produced less pollen than hermaphroditic flowers. Significantly, pistillate flowers were never observed in populations of L. apertum, implying that male flower production is an adaptive feature of the size-dependent sex allocation strategy of populations.

The occurrence of male flowers in members of the Liliales is not uncommon and a significant number of taxa are dioecious (e.g. Chamaelirium – Meagher, 1981; Smilax – Sawyer and Anderson, 1998; Wurmbea – Barrett and Case, 1996) or andromonoecious (e.g. Gagea – Wolfe, 1998; Veratrum – Liao and Zhang, 2008; Zigadenus – Emms, 1993). Surveys of male flowers in Fritillaria, Lilium and Tulipa have revealed frequencies ranging from 11 to 38 % (Peruzzi, 2012), indicating that female sterility is also a feature of other normally hermaphroditic taxa of lilies (and see Mancuso and Peruzzi, 2010). A detailed study of flower sex expression in seven populations of F. montana revealed that male flowers represented 14–48 % of the flowers sampled, with the remainder being hermaphroditic (Peruzzi et al., 2012). Similarly, in a survey of 11 populations of hermaphroditic Lloydia serotina, 48·5 % of the flowers were male (Manicacci and Després, 2001). In both these species the vast majority of plants produce a single flower during the flowering season and in both studies plants with male flowers were on average smaller for various traits than those with hermaphroditic flowers. Observations between years in Lloydia serotina provided evidence of switches in gender (Manicacci and Després, 2001; and see Jones and Gliddon, 1999), and these changes were similar to those observed in the present study. Size-dependent gender modification may be especially important in geophytic species with underground storage organs, such as members of the Liliaceae. Indeed, there is evidence for changes in sex allocation with size in Veratrum nigrum (Liao and Zhang, 2008) and Cardiocrinum cordatum (Cao and Kudo, 2008). In these cases gender adjustment is gradual, rather than involving discrete changes in sex phase between seasons, and this may, in part, be because of the larger floral displays that these species possess.

The significant number of plants producing only male flowers in members of the Liliaceae has resulted in some confusion concerning the sexual systems of populations. For example, based on studies of Fritillaria, Lilium and Tulipa, Peruzzi (2012) and Peruzzi et al. (2012) concluded that species in these genera were androdioecious (and see Mancuso and Peruzzi, 2010). In contrast, based on very similar patterns of sexual variation, Manicacci and Després (2001) concluded that Lloydia serotina was not androdioecious, as had been previously suggested, but was andromonoecious. This inference was based on the assumption that all plants within the populations if large enough over their lifetimes would produce hermaphroditic flowers. Given the overall rarity of androdioecy in angiosperms, and the severe constraints on the spread and maintenance of female sterile plants in predominantly hermaphroditic populations (reviewed by Pannell, 2002a), this interpretation seems more probable. The occurrence of plants with only male function in populations of Liliaceaous genera (including L. apertum), although superficially resembling androdioecious gender dimorphism, is more likely to reflect the influence of resource status on hermaphroditic species with size-dependent gender modification (and see Pannell, 2002b). However, long-term demographic studies would be required to completely rule out the presence of female sterile plants. Notwithstanding the difficulties in establishing the sexual system of these populations, in any given season the strong bias towards ‘maleness’ in populations will have important implications for mating biology, particularly the intensity of intrasexual (male–male) competition for mates.

Does L. apertum exhibit gender diphasy?

Gender diphasy may also be difficult to distinguish from more gradual forms of size-dependent gender adjustment in hermaphroditic species with gender monomorphism. The essential feature of gender diphasy is that individuals in a population belong to a single genetic class in which their sex expression in any particular season is controlled by circumstances, particularly plant size. Thus, within a flowering season populations should be composed of two primary gender classes: male-functioning and female-functioning plants. Although female-functioning plants can also produce pollen, the majority of their fitness is likely to be obtained through their ovules because of the strongly male-biased sex allocation at the population level, as we observed in L. apertum. Thus, the sex phenotypes H and H+M in our study are often referred to as ‘female’ (see Lloyd and Bawa, 1984; Schlessman, 1988), even though they are not strictly unisexual. Despite the occurrence of two primary gender classes in diphasic populations, from the perspective of between-season, lifetime fitness all plants on average should contribute genes to the next generation as both male and female parents. As a result, the lifetime functional gender of populations with gender diphasy will be unimodal, rather than bimodal as occurs in species with gender dimorphisms such as androdioecy.

As discussed above, although we cannot completely rule out the possibility that some L. apertum plants are genetically determined males, we believe this is unlikely for two reasons. First, given the various constraints on the evolution and maintenance of androdioecy it would be unexpected that male plants in androdioecious populations would produce fewer and smaller flowers with less pollen than coexisting female-functioning plants, as occurs in L. apertum. Both theoretical studies and empirical evidence indicate that for males to be maintained in androdioecious populations it is necessary for them to compensate for their lack of female function (Lloyd, 1975; Charlesworth, 1984; Pannell, 2002a, b). This is usually achieved by a significant increase in pollen production compared with hermaphrodites, the reverse pattern to what we observed in L. apertum. Second, we directly observed between-year switches in gender from male to female function and from female to male function associated with increases and decreases in plant size, respectively. These patterns have not been reported in androdioecious species and are more consistent with the seasonal switches in gender that are the hallmark of gender diphasy.

The lifetime functional gender of populations of gender-diphasic species should be monomorphic rather than dimorphic. However, in any given season the distribution of gender variation in a population will be discontinuous, as we have shown (Fig. 3). This pattern is the direct result of the ‘choice’ that plants make between two contrasting modes of pre-fertilization investment in maleness versus femaleness. This discontinuity differs from what would be observed in most hermaphroditic species that exhibit gradual gender adjustment. In these species phenotypic gender is regulated by modifications to the sex allocation of flowers, and by post-fertilization phenomena associated with fruit and seed abortion (Lloyd and Bawa, 1984). Gender diphasy thus represents an extreme case of size-dependent sex allocation in which populations in any given season are made up of two distinct categories or phases corresponding to male- versus female-functioning plants, despite populations being fundamentally hermaphroditic.

We thank J.-G. Chen and D.-L. Peng for assistance with data collection in the field and glasshouse. This work was supported by the National Science Foundation of China (31100179), China Postdoctoral Science Foundation (2012T50787) and Western Light Talent Culture Project to Z.-Q.Z. and by a Discovery Grant from the Natural Sciences and Engineering Research Council of Canada to S.C.H.B.