Sexual Expression and Reproductive Output in the Ephemeral Geranium Transversale Are Correlated with Environmental Conditions1

RESEARCH ARTICLE

AMERICAN JOURNAL OF BOTANY

Sexual expression and reproductive output in the ephemeral Geranium transversale are correlated with environmental conditions1

Aysajan Abdusalam2,3 , Dunyan Tan 2,5 , and Shu-Mei Chang4,5

PREMISE OF THE STUDY: Abiotic environmental factors are often considered to be important in the distribution and maintenance of variation in sexual systems in fl owering plants. Associations between sexes and abiotic factors are well documented in dioecious systems, but much less is known about this relationship in other sexually polymorphic systems. Species that are highly variable in sexual expression and habitat distribution can provide insights into the role of abiotic factors in maintaining variation in sexual expression.

METHODS: Focusing on a sexually polymorphic species, Geranium transversale, we measured sexual expression at both the fl ower and the plant level and examined vegetative and fl oral traits, pollen deposition, and reproductive success. We also tested for correlations between sexual expression and other traits and examined whether and how these traits covaried with abiotic environmental conditions.

KEY RESULTS: We identifi ed unique variation of sexual expression in G . transversale . There are four sexual morphs that display diff erent combinations of the three fl ower types (pistillate, staminate, and perfect). Sexual morphs that are phenotypically more female (i.e., female and gynomonoecious morphs) are found in wetter and milder environments, and fl ower earlier than morphs that are more male (i.e., hermaphroditic and andromonoecious morphs). Addi- tionally, fl oral organ size and reproductive success are infl uenced not only by the fl ower type but also by the sexual morph of the plant.

CONCLUSIONS: Environmental conditions are likely to cause some of the variation in sexual expression found in G . transversale . Both genetic and ecological factors likely contribute to the maintenance of sexual variation in this species.

KEY WORDS andromonoecy; breeding-system variation; cold desert; fl oral variation; Geraniaceae; gynomonoecy.

S e x u a l e x p r e s s i o n i n fl owering plants is amazingly diverse. Species (e.g., androdioecy and gynodioecy) individuals are found across range from having individuals with perfect fl owers only (hermaph- diff erent lineages ( Barrett and Hough, 2013 ; Elzinga and Varga, roditism) to having distinct individuals that produce either male/ 2017 ). Considering the direct impact that sexual expression has on staminate or female/pistillate fl owers only (dioecy). Many addi- the genetic fi tness of individuals, variation of sexual expression tional polymorphic systems with diff erent combinations of fl owers within and across populations can signifi cantly infl uence evolution- within (monoecy, andromonoecy, and gynomonoecy) or among ary processes such as adaptation and speciation (e.g., Wolfe, 1998; Calviño and Galetto, 2010; Yakimowski and Barrett, 2014). Hence, identifying the conditions and mechanisms through which variation in sexual expression evolves and is maintained is a fundamen- 1 Manuscript received 1 July 2017; revision accepted 10 November 2017. tal question in plant evolution (see reviews by Barrett, 2002 ; 2 Xinjiang Key Laboratory of Grassland Resources and Ecology and Ministry of Education Charlesworth, 2013 ; Zemp et al., 2015 ). Key Laboratory for Western Arid Region Grassland Resources and Ecology, College of Grassland and Environment Sciences, Xinjiang Agricultural University, Ürümqi 830052, Th e evolution of dioecy from hermaphroditism is a well-recognized PR China; evolutionary transition that has been documented in >100 cases 3 Key Laboratory of Ecology and Biological Resources in Yarkand Oasis, College of Biology across distantly related taxa in angiosperms (Bawa, 1980; Renner and Geography Sciences, Kashgar University, Kasghar 844006, PR China; and and Ricklefs, 1995; Webb, 1999; Weiblen et al., 2000; Charlesworth, 4 Department of Plant Biology, 2502 Plant Sciences Building, University of Georgia, Athens, Georgia 30602, USA 2002 ; Renner, 2014 ). Th is repeated evolutionary pattern across a 5 Authors for correspondence (e-mail: [email protected]; [email protected]) substantial number of evolutionary lineages raises the possibility https://doi.org/10.3732/ajb.1700258 that the causes or conditions favoring the selection of separate sexes

1920 • AMERICAN JOURNAL OF BOTANY 104 (12): 1920 – 1929 , 2017; http://www.amjbot.org/ © 2017 Botanical Society of America DECEMBER 2017 , VOLUME 104 • ABDUSALAM ET AL.—VARIATION IN SEXUAL EXPRESSION OF GERANIUM TRANSVERSALE • 1921

may be ubiquitous in fl owering species. A few important factors size of a population may also cause variation in sex ratios that could select for dioecy include inbreeding depression ( Charles- through genetic drift. Small populations are more likely to be worth and Charlesworth, 1978, 1979 ; Charlesworth, 2006; Eppley strongly affected by random fluctuations in allele frequencies, and Pannell, 2009 ), trade-off s between allocation to male and fe- and this has been proposed to explain why smaller populations male functions when resources are limited ( Charnov et al., 1976 ; of the gynodioecious Lobelia siphilitica often have higher female Charnov, 1979 ; Givnish, 1980 ; Campbell, 2000 ; Parachnowitsch frequency (Caruso and Case, 2007). and Elle, 2004), and sexual selection (Willson, 1979; Th omson and Finally, abiotic factors can interact with endogenous characteris- Barrett, 1981 ; Moore and Pannell, 2011 ; Delph and Herlihy, 2012 ). tics, such as life-stage and plant size, to infl uence sexual expression Unfortunately, it is diffi cult to infer what factors facilitated the evo- in some plants. For example, in sequentially hermaphroditic spe- lution of separate sexes in extant dioecious species because recre- cies, smaller and younger individuals tend to produce male sexual ating the ancestral ecological or genetic conditions of a species is structures but can switch to producing either perfect or female oft en diffi cult, if not impossible. Th is problem, however, can be cir- structures when they grow larger (e.g., Givnish, 1980; Doust and cumvented by examining ecological and genetic conditions associ- Cavers, 1982 ; Matsui, 1995 ; Renner et al., 2007 ). A similar ated with diff erent sexual morphs found in species that exhibit correlation between plant size and sex expression or allocation to within-species variation in sexual expression ( Barrett, 1992 ; Barrett diff erent sexual functions can be found in other sexually polymor- and Hough, 2013 ; Yakimowski and Barrett, 2014 ). phic species such as gynodioecious ( Ashman et al., 2001 ), andro- Abiotic environmental conditions, such as soil moisture, soil fer- monoecious ( Wolfe, 1998 ; Liao and Zhang, 2008 ), monoecious tility, and annual temperature, are oft en correlated with the preva- (Torices and Mendez, 2011), and hermaphroditic (Peruzzi et al., lence of a particular sex in plant populations and are likely to be 2012 ; Lankinen and Green, 2015 ) species. Th e eff ect of abiotic fac- important factors involved in the evolution of separate sexes (e.g., tors on plant size could also infl uence sex expression (Delesalle, Maurice, 1999; Sakai and Weller, 1999; Case and Barrett, 2004). For 1989 ; Zhang et al., 2014). For example, larger plants may have ac- example, in dioecious species, male individuals are more oft en cess to more available resources (both above and below ground) found in lower-quality sites, compared to female individuals, which and tend to have larger fl owers, which could aff ect the number of likely refl ects the outcome of selection acting on the relative re- pollen grains received (e.g., Philipp and Hansen, 2000). Th is can, in source demands of pollen production vs. seed production (Meagher, turn, indirectly shift resource allocation to seed production rather 1984; Th omas and LaFrankie, 1993; Korpelainen, 1998; Queenborough than to pollen function ( Dudash, 1991 ; Wolfe and Shmida, 1997). et al., 2007; Peruzzi et al., 2012). By contrast, in gynodioecious To better understand how environmental conditions may infl u- species that contain both female and hermaphroditic individuals, ence sexual expression in wild populations, we investigated sex ex- populations with females are oft en found in hotter and dryer envi- pression and sex ratios in natural populations of a wild geranium ronments than populations that contain only hermaphrodites species, Geranium transversale, which grows in stony slopes, sandy (Sakai and Weller, 1999; Case and Barrett, 2004; Bishop et al., 2010; deserts, and sandy steppes in central Asia. Geranium transversale ex- Barrett and Hough, 2013 ; Ruff atto et al., 2015 ). One hypothesis to hibits variation in both within-fl ower sexual expression and within- explain this pattern is that, when faced with limited resources, her- plant fl ower-type combinations. To distinguish between variation at maphrodites may emphasize pollen production, allowing seed pro- these two levels, we refer to sex expression of individual fl owers as duction in females to surpass that of hermaphrodites and thus one of three “fl ower types”—perfect, pistillate, and staminate—and enabling the establishment of females in the population (Delph, refer to sex expression of individual plants as one of four “sexual 1990 , 2003 ; Bishop et al., 2010 ). Th ough the associations of sex ra- morphs” that refl ect the combinations of fl owers within a plant: an- tios with environmental conditions diff er in these two instances, dromonoecious (with staminate and perfect fl owers; hereaft er “AN”), they illustrate how abiotic factors can generate selective pressure gynomonoecious (with pistillate and perfect fl owers; hereaft er “GY”), that can aff ect sex expression, an eff ect that has been frequently hermaphroditic (with perfect fl owers only; hereaft er “H”), and female documented in the literature ( Meagher, 1984 ; Grazer et al., 2014 ; (with pistillate fl owers only; hereaft er “F”) (Appendix S1; see Supple- Varga and Kytöviita, 2016). mental Data with this article). Th e habitat of this species varies greatly In addition to being a response to selective pressures, varia- in soil composition and oft en displays steep soil-moisture gradients, tion in sex expression among populations may also be pro- such that closely located individuals can experience dramatically dif- moted by abiotic factors through nonselective mechanisms. ferent environmental conditions that may infl uence sexual expres- For example, an association between environmental conditions sion of these plants. In addition, because G . transversale grows and sex ratios may be observed if populations are phenotypically perennially in a wide array of microsites, we also examined whether plastic and shift resource allocation from one sexual function spatiotemporal variation in ecological factors were associated with to another in response to biotic or abiotic factors ( Lloyd and phenology and sex expression across environmental clinal gradients. Bawa, 1984; Maurice, 1999; Barrett et al., 2000; Sarkissian et al., Using measurements collected from natural populations, we asked 2001; Vitt et al., 2003; Vallejo-Marín and Rausher, 2007; Diggle whether sexual morphs diff er in their (1) fl owering phenology and et al., 2011). As a result, variation in sexual expression among spatial distribution, (2) biomass allocation to diff erent fl oral struc- different populations could be found in different environments tures and reproductive vs. vegetative structures, and (3) seed produc- without possessing heritable genetic differences. For example, tion following open pollination. fl owers of Gossypium hirsutum (Yasuor et al., 2007) and Ipomoea purpurea ( Baucom et al., 2008 ) are feminized after exposure to some herbicides that disrupt anther formation. In addition, her- MATERIALS AND METHODS bivory can also lead to altered sex expression in plants (Hendrix and Trapp, 1981 ; Lehtilä and Strauss, 1999 ; Vallius and Salonen, Study species and study site — Geranium transversale Vedis 2000; Parra-Tabla et al., 2004; Arceo-Gómez et al., 2009). The (Kar. et Kir.) is native to Central Asia, western Siberia, and northern 1922 • AMERICAN JOURNAL OF BOTANY

China, including Kazakhstan, Kyrgyzstan, Russia, Tajikistan, Ürümqi city. In addition to moisture, we also divided the soil type Turkmenistan, Uzbekistan, and the northern Xinjiang Uyghur of our sampled plants into three crude categories based on the tex- Autonomous Region of China ( Xu and Aedo, 2008 ). Its habitat ture of the soil: humus soil (a mixture of plant material and other includes stony slopes, sandy deserts, sandy steppes, and open organic materials and forest residue), stone–humus mixed soil, and fields at elevations of 500–1300 m above sea level. Geranium stony–gravel soil. Among these three soil types, soil moisture levels transversale is a spring ephemeral species that forms clonal ra- diff ered signifi cantly, with the stony–gravel soil containing much mets that are connected by underground rhizomes. Clones can lower moisture than the humus soil, and stone–humus soil in be quite extensive and form a patchy aboveground distribution. between. Each rhizome branch produces a single inflorescence with 8–18 flowers. The sequence of flowering within an inflorescence pro- Biomass allocation to diff erent aboveground structures— T o ceeds in an acropetal and centrifugal manner. Perfect flowers measure how G . transversale plants allocate their resources to are protandrous, with stamens maturing earlier than the pistil diff erent aboveground structures, we randomly sampled two ra- (A. Abdusalam et al., unpublished data), reducing the likelihood mets each from ~15 diff erent colonies for each of the four sexual of self-pollination. Anthers and stigma are also spatially sepa- morphs, recorded the infl orescence height, and recorded the rated when the stigma becomes receptive (Appendix S2), further number of the three fl ower types during the 2011 fl owering sea- reducing the possibility of autonomous selfing. In the study re- son. Colonies were chosen to span the range of sizes observed in gion (Xinjang Uyghur Autonomous Region of China), flowering natural populations. We harvested all the aboveground structures generally starts in mid-April and fruits start to ripen in early and separated the leaves, infructescence stems, and fl owers in June. 2011. To measure the biomass allocation at the single fl ower level, We conducted this study using natural populations located we collected all fl owers from each sexual morph. Th e diameter of around Yamalik Hill (43 °79 ′22 ′ ′–43 °79 ′28 ′ ′N, 87 °56 ′06 ′ ′–87 °56 ′18 ′ ′E; each fl ower was measured to 0.01 mm using a digital Vernier cali- 950–1300 m above sea level), an area of gray desert soil fi lled with per SF2000 (Guanlin, PR China) before we separated the corolla, gravel, dominated by typical cold-desert vegetation. Local weather calyx, stamen, and pistil of every fl ower into separate paper bags. patterns follow a typical continental arid climate ( Abdusalam Each fl oral part was oven dried to constant mass at 80 °C for 48 h and Tan, 2014) with a mean annual temperature of 6.8 °C and in the laboratory, and the biomass of each part was measured mean annual precipitation (including rain and snow) of 234 mm. to 0.01 mg using a Sartorius BS210S electronic balance (Guanlin, Field studies were carried out between April and June in 2011 and PR China). 2014. To estimate pollen and ovule production per fl ower, we sampled fl owers that were fully open but had not yet been visited by pollina- Sexual expression and fl owering phenology of individuals—To in- tors. To separate the eff ects of fl ower type and sexual morphs, we vestigate the sexual expression and phenology of plants in the Ya- sampled 22–48 fl owers of each fl ower type from several individuals malik Hill population, 131 colonies of diff erent sizes (covering for all sexual morphs. Anthers and pistil from each fl ower were col- 1–10 m2 ) were measured before fl owering had started in both lected and placed separately in 1 mL microcentrifuge tubes. Th e 2011 and 2014. Each colony is referring to a cluster of ramets that number of pollen grains in each fl ower was determined by subsam- may or may not be connected belowground but are likely to be pling pollen collected from individual fl owers and counted using an part of the same genetic individual. Clearly distinct colonies (sep- Olympus BH-2 (10× ) light microscope in the laboratory following arated by ≥5 m) were sampled haphazardly across the population the methods described in Han et al. (2008) . Th e number of ovules in the Yamalik Hill area. Plants were labeled in 2011 to enable the per ovary in each fl ower was counted under a stereomicroscope relocation of the same individuals in 2014; labels of 26 plants from (SMZ 1000; Nikon, Japan). 2011 were lost and each was replaced with an individual of similar Pollen size was measured for at least three samples each from size near the location of the original missing plant. We monitored 10–14 individuals of each sexual morph with perfect flowers; the phenology and sexual system of the labeled colonies over the nine individuals were sampled for pistillate flowers (underdevel- two fl owering seasons. Based on these observations, we calculated oped anthers). Pollen samples were collected from single flow- the percentage representation of each sexual morph. In addition, ers, placed on a slide, and observed under the Olympus BH-2 we recorded the fi rst fl owering date for each colony by surveying optical microscope; 10 grains from each were measured using the population every 3 d from mid-April to mid-May of each year. the Motic digital image analysis system (Motic-Advanced 3.2; To document the temporal sequence of fl ower-type appearance Shenzhen, PR China) on a computer connected to the micro- (perfect, pistillate, or staminate) within an infl orescence, we scope. The length and width of each grain was measured, and the closely monitored three infl orescences each in 10 colonies for all volume of the pollen was calculated as length × width × 3.14 / sexual morphs and recorded the sex expression of the fl owers that 4. Similar measures were made for ovules using a stereomicro- opened sequentially. scope (Nikon-SMZ 1000). Though our measurements for pollen and ovule size are likely to be approximations, they provide suf- Soil characteristics associated with the sexual morphs — To deter- ficient information to distinguish between viable and nonviable mine whether the microenvironments differed among sexual (aborted) pollen grains for this study (see below). In addition, morphs, we measured volumetric soil moisture for ~10 colonies of because there were occasionally some ovules in staminate flow- each sex morph during both 2011 and 2014 using a TSC-1 auto- ers, we included staminate flowers in the ovule count data, even matic measuring system with 8 cm probes (Leishen Electron Instru- though these ovules would never be fertilized. ment Company, Shijiazhuang, PR China). Mean average daily temperatures (April–August, 2011–2014) were obtained from the Seed production following open pollination for diff erent sexual Xinjiang Uygur Autonomous Region Meteorological Bureau in morphs— To investigate seed production by the four sexual DECEMBER 2017 , VOLUME 104 • ABDUSALAM ET AL.—VARIATION IN SEXUAL EXPRESSION OF GERANIUM TRANSVERSALE • 1923

morphs, we marked ~30 inflorescences for F, H, GY, and AN fl ower types sequentially within the fl owering season, both starting plants in 2011. For each infl orescence, we labeled two fl owers for with single-sex fl owers (pistillate fl owers for GY and staminate fl ow- each sex-expression type as our target fl owers and allowed open ers for AN morphs) and later switching to producing only perfect pollination to occur. About 4 wk aft er fl owering, we collected fl owers. Th e number of single-sex fl owers produced were generally mature fruits from the labeled fl owers to determine their seed low: <5 fl owers for GY plants and <3 staminate fl owers for AN. Pistil- set. Seed set was measured as the proportion of available ovules late fl owers have nondehiscent anthers with small and aborted pollen (identifi ed as the sum of developed and aborted seeds and unfer- grains inside, and staminate fl owers have shriveled stigma and pistil tilized ovules) within a fl ower that successfully turned into seeds. but still contain intact ovules inside the ovaries. Th e proportion of Aborted seeds were much smaller and shriveled and so were easily each sexual morph in the population did not diff er signifi cantly be- distinguishable from the healthy, mature seeds. Unfertilized tween 2011 and 2014 (χ 2 = 7.36, df = 3, P > 0.05). Specifi cally, GY ovules remained small but were identifi able as residual structures plants constituted the most common plant type within the study in the fruits. Seed mass was measured for six to eight individuals population (39% and 44% for 2011 and 2014, respectively), followed in each sex morph using mature seeds collected from unmanipu- by H (35% and 30%) and AN (16% and 22%) morphs, and the F lated infl orescences. We measured the mass of the bulked seeds morph remained rare in both years (10% and 4%). (100 seeds each) to obtain the mean seed mass for each plant. To Temporal surveys showed that sexual morphs diff ered in when measure pollen deposition on the stigmas of the perfect and pistil- they started fl owering: plants that started fl owering early (mid- late fl owers, we collected ~26 stigmas from both types of fl owers April) tended to be GY or F morphs (38 of 39 plants), whereas AN in each sexual morph and counted the number of deposited con- and H morphs usually did not start fl owering until early May, with specifi c pollen grains under a microscope. many individuals not fl owering until the last survey in mid-May. Th e trend was similar in both 2011 and 2014 (Fig. 1; χ 2 = 12.5, df = Statistical analysis — We used analysis of variance (ANOVA) to 9, P > 0.05). test for the diff erences among sexual morphs for plant height, total fl ower number, aboveground biomass, and total biomass of Microsite characters of diff erent sexual morphs— Of the fi ve en- fl owers. When multiple measurements were taken from a single vironmental measurements analyzed by PCA, PC1 (soil moisture, individual, we included individuals as a random eff ect in our elevation, and soil type) explained 53% and PC2 (average tempera- ANOVA. We tested for diff erences on three types of traits among ture and rainfall) explained 24% of the overall variance (Appendix S3). the three fl ower types (perfect, pistillate, and staminate): fi rst, the Th e sexual morphs diff ered signifi cantly with regard to PC1 (F = dry mass of the entire fl ower as well as individual fl oral compo- 3, 40 11.51, P < 0.001). Patterns in association between the four sexual nents (corolla, calyx, stamen, and pistil); second, number and size morphs and individual microsite factors contributing to PC1 of the gametophytes (pollen and ovules); and fi nally, number of emerged when we examined the three factors separately (F = seeds produced following natural pollination. If ANOVA indi- 3, 43 12.99 for soil moisture, F = 8.52 for elevation, all P < 0.001; cated signifi cant eff ects of the main variable, Tukey’s HSD test was 3, 99 Fig. 2 ). Because AN and F morphs were missing in one of the soil performed. Data that were not normally distributed (ratio of each types, this data set is not suitable for chi-square tests of homoge- morph, biomass of each part, fruit and seed set) were transformed neity in soil distribution. However, the diff erent category missing before analysis to ensure homogeneity of variance, with untrans- in each of these morphs illustrated the diff erences in how sexual formed means reported. Because the environmental factors we morphs are associated with soil types recorded in this study. More measured were intercorrelated, we used a principal component specifi cally, AN were only found in rocky or rocky–humus mixed analysis (PCA) to generate two orthogonal principal components soil types located in higher elevation where the soil moisture tends (PCs) that best described site variation. Th is PCA included only to be low, whereas F tend to be found in soil with high humus 44 sites where soil moisture was measured individually. Th ese sites content ( Fig. 2A ). H and GY plants were found in intermediate were evenly distributed among the four sexual morphs. Using the two main PCs with highest eigenvalues (both >1), we conducted a one-way ANOVA to determine whether the four sexual morphs differed in their environmental conditions. Categorical data (fl owering time, soil type distribution, and sex ratios) were analyzed using chi-square tests to test for differences in the distribution of sexual morphs in different categories or between different years. We also carried out correlation analyses between flowering time and the two PCs. All statistical analyses were performed using SPSS version 13.0.

RESULTS

Phenological diff erences among sexual morphs— Our observations clearly identifi ed four sexual morphs in G . transversale in the Yamalik Hill population: H, GY, andromonoecious AN, and F individuals FIGURE 1 Breakdown of the four sexual morphs for the colonies that be- (Appendix S1). H and F morphs produce only one fl ower type— came actively fl owering during the three survey periods in 2011 and perfect and pistillate fl owers, respectively, throughout the entire 2014 (H = hermaphroditic, AN = androdioecious, F = female, GY = gyno- fl owering season. By contrast, GY and AN morphs produce two dioecious; total N = 131 for both years). 1924 • AMERICAN JOURNAL OF BOTANY

FIGURE 2 Soil type distribution (A), soil water content (B), and elevation (C) of the four sexual morphs of Geranium transverseale (AN = andromonoecious, H = hermaphroditic, GY = gynomonoecious, and F = female). Error bars = SE. In A and B, sample sizes for AN, H, GY, and F are 30, 32, 27, FIGURE 3 Mean infl orescence height, total number of fl owers produced, and 15, respectively; in C, sample sizes for AN, H, GY, and F are 13, 16, 13, dry aboveground biomass, and dry vegetative biomass at harvest for the and 10, respectively. Letters above the bars indicate statistical results; four sexual morphs (H = hermaphroditic, GY = gynomonoecious, AN = an- means that share the same letter do not diff er signifi cantly (at P = 0.05). dromonoecious, and F = female). Error bars = SE. Sample sizes are in parentheses. Letters above the bars indicate statistical results; means that soil water content and elevation and in all soil types (Fig. 2A–C). share the same letter do not diff er signifi cantly (at P = 0.05).

Th e sexual morphs did not diff er with regard to PC2 (F 3, 40 = 0.72, P = 0.54). However, fl owering time was signifi cantly positively corre- ( F1, 52 = 2.02, P > 0.05; Fig. 5 ), and both were smaller than perfect lated with PC2 (Pearson r = 0.78, P < 0.001) but much less so with flowers. PC1 (r = 0.33, P = 0.03, but not significant after Bonferroni All perfect fl owers allocated biomass to the four fl oral structures correction). similarly, regardless of the morph from which they were collected (Appendix S4A). Overall, the corolla had the highest biomass in all Biomass allocation to vegetative and reproductive structures — fl owers (~40% of the total biomass), with roughly equal propor- Sexual morphs diff ered in many of the traits measured (both veg- tions of biomass allocated to the remaining three structures—pistil, etative and reproductive). H plants had signifi cantly taller stamen, and calyx. Biomass allocation in pistillate fl owers diff ered infl orescences than the other three sexual morphs, and the AN slightly between the two sexual morphs that produce them: F plants plants were the shortest (F 3, 119 = 8.93, P < 0.001; Fig. 3A) . GY and AN morphs produced slightly, but signifi cantly, more fl owers than

F morph plants, with H in between ( F 3, 119 = 5.32, P < 0.05; Fig. 3B ). Finally, the overall aboveground biomass (including vegetative and reproductive structures) were lowest in GY plants (F 3, 119 = 3.82, P < 0.05; Fig. 3C). Combined, the GY morph had smaller stature than, but produced a similar number of fl owers as, the other morphs ( Fig. 3A–D ). When we examined the sexual morphs separately, there was a positive correlation between the total number of fl owers and the aboveground biomass (Fig. 4A–C) , except for AN plants for which no relationship was found (Fig. 4D). A similar pattern was found when we separated unisexual and perfect fl owers (data not shown).

Floral resource and sex allocation of each sexual morph — When all flower types were examined together, perfect flowers were significantly larger than the other two flower types (F 2, 178 = 33.36, P < 0.001; Fig. 5 ) . Though perfect flowers in the AN morph were significantly smaller than the perfect flowers found FIGURE 4 in the other morphs (F 2, 95 = 27.83, P < 0.001; Fig. 5), they were Relationship between the number of fl owers produced and the same size as the staminate flowers on AN plants. Pistillate plant size (vegetative biomass) for the four sexual morphs: (A) hermaph- flowers in the F and GY morphs were similar in their diameters roditic, (B) female, (C) gynomonoecious, and (D) andromonoecious. DECEMBER 2017 , VOLUME 104 • ABDUSALAM ET AL.—VARIATION IN SEXUAL EXPRESSION OF GERANIUM TRANSVERSALE • 1925

FIGURE 6 Mean (A) number of pollen grains per fl ower, (B) pollen size, (C) FIGURE 5 Flower diameter of the three flower types (pistillate, perfect, number of ovules per fl ower, and (D) ovule size for the three fl ower types and staminate) produced by the four sexual morphs (H = hermaphro- (pistillate, perfect, and staminate) produced by the four sexual morphs ditic, F = female, AN = andromonoecious, and GY = gynomonoe- (H = hermaphroditic, GY = gynomonoecious, AN = andromonoecious, and cious). Error bars = SE. Sample sizes are in parentheses. Letters above F = female). Error bars = SE. Sample sizes are in parentheses. Letters above the bars indicate statistical results; means that share the same letter the bars indicate statistical results; means that share the same letter do not do not differ significantly (at P = 0.05). Capital letters are for within- diff er signifi cantly (at P = 0.05). Capital letters are for within-sexual-morph sexual-morph comparisons between different flower types, and low- comparisons between diff erent fl ower types and, and lowercase letters are ercase letters are for comparisons of the same flower types among for comparisons of the same fl ower types among sexual morphs. sexual morphs. tended to allocate a higher proportion of biomass to the calyx than significantly between perfect and pistillate flowers and, to a to the pistil, whereas GY plants tend to allocate similar amounts to lesser extent, among different sexual morphs (F 1, 129 = 282.12 for the corolla and calyx (Appendix S4B). When examining all fl ower flower type and F 3, 129 = 25.17 for sexual morphs, both with P < types together, pistillate fl owers allocated a signifi cantly lower per- 0.001; Fig. 7A) . More specifically, when examining all morphs × centage of biomass to the stamen than the other morphs (F 3, 47 = together, perfect flowers received almost 10 the number of pol- 115.37, P < 0.001; Appendix S4C). However, staminate fl owers al- len grains as pistillate flowers ( Fig. 7A ). Among the perfect located similar amounts to the pistil and stamen, a pattern similar flowers borne on the three sex morphs, the numbers of pollen to perfect fl owers. grains received by GY and AN were significantly higher than

Th e number and size of pollen grains diff ered signifi cantly that received by H stigmas (F 2, 77 = 22.57, P < 0.001; Fig. 7A). among perfect fl owers produced by diff erent sexual morphs (F 2, 116 = Between the two morphs bearing pistillate flowers, GY-pistillate 28.68, P < 0.05 and F 2, 113 = 9.48, P < 0.05, respectively) ( Fig. 6A, B ) . flowers received significantly more pollen grains than F-pistillate Specifi cally, pollen production (pollen number) in perfect fl owers fl owers (F 1, 51 = 9.91, P < 0.001). follows the order of H > GY > AN plants. By contrast, Seed set and seed mass were also infl uenced by both the fl ower pollen size in perfect fl owers is signifi cantly larger in AN and H type and the plant sexual morph. Overall, perfect fl owers had sig- than in GY perfect fl owers (Fig. 6B). Interestingly, pistillate fl ow- nifi cantly higher seed set per fl ower (F 1, 320 = 106.34, P < 0.01) but ers in this species still produce pollen grains inside the nondehisc- lower seed weight (F 1, 34 = 7.65, P < 0.01) than the pistillate fl owers ing anthers, but the grains are much smaller and most likely (Fig. 7B). If we only consider perfect fl owers, seed set in AN was nonviable. signifi cantly higher than seed set in GY and H plants (F 2, 199 = Th ere were no overall eff ects of sexual morphs in regard to the 10.16, P < 0.01). For pistillate fl owers, GY has signifi cantly lower number of ovules in their perfect fl owers ( F 2, 58 = 2.50, P > 0.05; seed set (F 1, 119 = 4.48, P < 0.001) and seed weight (F 1, 14 = 325.63, Fig. 6C) or in ovule size (F 2, 94 =1.10, P > 0.05; Fig. 6D). However, P < 0.01) than the F morph (Fig. 7C). pistillate fl owers of GY plants produced more and larger ovules than F plants (F 1, 45 = 10.46, P < 0.01 for ovule number and F 1, 56 = 7.01, P < 0.001 for size; Fig. 6C, D ). Interestingly, there are nor- DISCUSSION mal-sized ovules inside the ovary of staminate fl owers, but far fewer than in perfect and pistillate fl owers ( F = 9.49, P < 0.001; 2, 184 Unique sex-expression system in Geranium transversale— Our 2 yr Fig. 6D). of fi eld observation revealed unique sex-expression patterns in Gera- nium transversale that have not been reported before. Specifi cally, in Female reproductive success of pistillate and perfect flowers — addition to the perfect and pistillate fl owers commonly found in The number of pollen grains deposited on the stigma differed other Geranium species ( Ågren and Willson, 1991 ; Williams et al., 1926 • AMERICAN JOURNAL OF BOTANY

(only pistillate fl owers) individuals and, occasionally, intermediate individuals with both pistillate and perfect fl owers (same as the GY individuals in this study). Geranium transversale contains all of these previously reported morphs and exhibits similar morphological patterns, such as similar vegetative structures among F and H, smaller petal size and fewer pollen received in pistillate than in perfect fl owers, and heavier seeds produced by F than by H ( Ågren and Willson, 1991; Chang, 2006). A fourth, AN, morph that produces staminate fl owers before switching to producing perfect fl owers has, thus far, been reported only in G . transversale . In contrast to the diff erence in fl ower diameter between the pistillate and perfect fl owers on the GY morph, staminate fl owers are similar in size to the perfect fl owers on the same plant and contain normal-sized ovules inside the ovary. Th e only apparent diff erence between staminate and perfect fl owers is that the styles are nonfunctional or aborted in the staminate fl owers. We suspect that the production of staminate fl owers (more specifi - cally, aborted styles) is a response to poor environmental conditions, rather than a genetically determined modifi cation, because genetic determination is oft en associated with a reduction in both reproductive structures and gametes. For example, in other Geranium species in which male sterility is genetically controlled, pistillate fl owers found in F and GY morphs contain reduced anthers with only unde- veloped pollen grains (Ågren and Willson, 1991; Williams et al., 2000; Asikainen and Mutikainen, 2005; Chang, 2006). We hypothe- size that both environmental and genetic factors contribute to the spectrum of sex expression in G . transversale , and we plan to test these hypotheses in future studies.

Environmental association with sex expression— Regardless of whether the production of single-sex flowers is under genetic or environmental control, our results show a clear association between sexual morph and environmental/microsite conditions. Specifically, if we consider the four sexual morphs by their relative level of “femaleness” from low to high as (1) AN, (2) H, (3) GY, and (4) F, we find that as the environmental conditions become less harsh (less rocky soil with more organic materials and moisture) the level of femaleness increases. The perceived harsh conditions around the AN morph does not appear to be re- flected in the vegetative size of these plants, a trait commonly used as a surrogate for environmental quality in field studies ( Delph, 1990 ; Wolfe and Shmida, 1997 ; Calviño and Galetto, 2010 ). This pattern suggests that water availability may not be the driving force for the spatial distribution observed here because soil type is also correlated with other environmental factors, including elevation and temperature. Collectively, our results indicate that abiotic conditions contribute to the spatial, and likely also the temporal, variation of sex expression in wild FIGURE 7 Mean (A) pollen deposition, (B) seed set, and (C) seed mass for populations of G . transversale . the pistillate and perfect fl owers produced by the four sexual morphs (H = As discussed above, there are at least two non–mutually exclu- hermaphroditic, GY = gynomonoecious, AN = andromonoecious, and F = sive hypotheses that can explain an association between sexual female). Error bars = SE. Sample sizes are in parentheses. Letters above the expression and the environment as revealed by our results. First, bars indicate statistical results; means that share the same letter do not sex expression might be a plastic trait that can change in response diff er signifi cantly (at P = 0.05). Capital letters are for within-sexual-morph to environmental quality ( Vitt et al., 2003 ; Case and Barrett, 2004 ; comparisons between diff erent fl ower types, and lowercase letters are for Lu et al., 2012). If this is true, we would expect to see sex expression comparisons of the same fl ower types among sexual morphs. “switch” when we move plants to a diff erent environment. Some 2000 ; Asikainen and Mutikainen, 2005 ; Chang, 2006 ), we also found support for this idea comes from a recent study on the gynodioe- staminate fl owers consistently produced by AN individuals. To our cious Geranium sylvaticum by Varga and Kytöviita (2016) , in which knowledge, only two types of breeding systems have been found in nearly half of the transplanted individuals showed some level of sex Geranium species thus far—hermaphroditism and gynodioecy, the lability in response to light condition in the transplant site. Addi- latter consisting of hermaphroditic (only perfect fl owers) and female tionally, nine of the 21 species reviewed in the same study found DECEMBER 2017 , VOLUME 104 • ABDUSALAM ET AL.—VARIATION IN SEXUAL EXPRESSION OF GERANIUM TRANSVERSALE • 1927

some evidence of sex lability (Varga and Kytöviita, 2016, and refer- morphs do not). A high selfi ng rate could also explain the low ences therein), further supporting phenotypic plasticity in sex ex- percentage of ovules that matured into seeds if inbreeding depres- pression. Similar behavior is also seen in sequentially hermaphroditic sion is high during the seed development stage, as found in G . species in which individuals generally express only one sexual func- maculatum ( Chang, 2007 ). However, because we measured each tion at any given time but can switch back and forth across years or fi tness component (fl ower number, seed set, and seed weight) us- seasons (Case and Barrett, 2004). In those species, female function ing separate subsamples, our results cannot be used to compare is more likely to be expressed when plants experience “better” envi- the total fi tness of diff erent sexual morphs. Future studies that ronmental conditions ( Freeman et al., 1980 ; Dawson and Geber, collect longitudinal data for all these fi tness components on the 1999 ), similar to what we found in the present study. same infl orescences as well as total fl ower number of the entire Alternatively, though not mutually exclusive with the above idea, clonal patch will allow us to compare the relative fi tness of diff er- sex expression could be genetically determined and the observed ent morphs. sex-expression patterns may have arisen through natural selection Future studies that transplant individuals of diff erent sexual by environmental factors. Th is would imply that each sexual morph morphs into a range of environments, as seen in the fi eld, should is most adapted, in relation to the other morphs, to the local envi- be done to help determine whether sex expression is a plastic trait ronments where they are found and that moving plants would not in G . transversale that can “switch” in response to the environmen- be expected to cause the sexual expression of these plants to change. tal conditions, similar to the fi ndings of Varga and Kytöviita Also, the plants moved to their nonnative sites would be expected to (2016) . Additionally, genetic studies involving crossbreeding be- display lower relative fi tness than the morph found native to that tween diff erent sex morphs and the sex-expression patterns seen environment. Th e idea that environments may select for a particu- in their progeny can help determine the relative contributions of lar sex is not new in the literature on the evolution of gynodioecy or genetic and ecological factors to the sex expression of this species. dioecy. Th e sex-diff erential-plasticity hypothesis ( Delph, 1990 , In the three congeneric species studied thus far, evidence exists 2003 ) proposes that because hermaphroditic plants have alternative that the expression of femaleness (i.e. the production of female ways to gain fi tness, through either seed or pollen, whereas females morph or pistillate fl owers) is, to a large degree, a heritable trait can gain fi tness only through seeds, it is plausible that hermaphro- attributable to genotypes at a sex-determining locus (or loci) (S.-M. dites are more plastic in resource allocation to their seed fi tness Chang, unpublished data). Th e results of the present study there- than females. As such, in harsher environments, hermaphrodites’ fore add to existing evidence that the genus Geranium is a good relative seed fi tness may become lower than females’, and this allows model system for studying the ecology and evolution of variation the establishment of female individuals. Th is hypothesis, hence, in sex-expression systems. provides an explanation for why we oft en observe an association between harsh environment and the presence of females in gynodioecious species. However, our results are inconsistent with this ACKNOWLEDGEMENTS hypothesis because G . transversale females are found in the least Th e authors thank P. Sabit and Y. Alim for help in the fi eld. harsh environment, opposite from what the sex-diff erential-plasticity Laboratory work was carried out in Xinjiang Key Laboratory of hypothesis predicts. Instead, our data are consistent with observa- Grassland Resources and Ecology and at Ministry of Education Key tions of many dioecious species that females are found in spatially Laboratory for Western Arid Region Grassland Resources and “better” environments than males. For example, female plants are Ecology, College of Grassland and Environment Sciences, Xinjiang oft en found in wetter and low-altitude environments in Podocarpus Agricultural University, PR China. Comments from N. Armstrong nagi (Nanami et al., 2005). Th is pattern is usually explained by the and two anonymous reviewers have signifi cantly improved this generally higher resource requirements by the female than the male paper. Th is study was partially supported by the National Natural function in fl owering plants ( Charnov,1982 ; Sun et al., 2006 ). Science Foundation of China (NSFC, U1603231, 31400279). Th e deposition of pollen onto fl owers was highly dependent on the fl ower type in G . transversale. Th e number of pollen grains on perfect fl owers was, on average, 10× higher than deposition onto LITERATURE CITED pistillate fl owers. Higher pollen deposition on perfect fl owers is commonly observed in other gynodioecious (e.g., Van Etten and Abdusalam , A. , and D.-Y. Tan . 2014 . Contribution of temporal fl oral closure to reproductive success of the spring-fl owering Tulipa iliensis. Journal of Chang, 2014) or gynomonoecious species (e.g., Mamut et al., Systematics and Evolution 52 : 186 – 194 . 2014) and has been attributed to the fact that the larger diameter Ågren , J. , and M. F. Willson . 1991 . Gender variation and sexual diff erences in of perfect fl owers attracts more pollinators than smaller pistillate reproductive characters and seed production in gynodioecious Geranium fl owers. Th is may also be the case for G . transversale, though we maculatum. American Journal of Botany 78 : 470 – 480 . do not have pollinator visitation data. Higher pollen deposition in Arceo-Gómez , G. , V. Parra-Tabla , and J. Navarro . 2009 . Changes in sexual ex- perfect fl owers may also be due to self-pollination. Although au- pression as result of defoliation and environment in a monoecious shrub in togamous self-pollination in G . transversale is likely to be infre- Mexico: Implications for pollination. Biotropica 41 : 435 – 441 . quent because of the spatial and temporal separation (herkogamy Ashman , T. L. , J. Pacyna , C. Diefenderfer , and T. Left wich . 2001 . Size- and protandry; see Appendix S2) between male and female func- dependent sex allocation in a gynodioecious wild strawberry: Th e eff ects tions, geitonogamous self-pollination between fl owers of the of sex morph and infl orescence architecture. International Journal of Plant Sciences 162 : 327 – 334 . same plant remains a possibility. In fact, our fi nding that pistillate Asikainen , E. , and P. Mutikainen . 2005 . Preferences of pollinators and her- fl owers in GY received more pollen than pistillate fl owers in F bivores in gynodioecious Geranium sylvaticum. Annals of Botany 9 5 : supports the possibility that pollen transfer between fl owers of the 879 – 886 . same plant is at least partially responsible for pollen deposition Barrett , S. C. H. 1992 . Gender variation and the evolution of dioecy in (since GY morphs contain pollen-producing perfect fl owers but F Wurmbea dioica (Liliaceae). Journal of Evolutionary Biology 5 : 423 – 444 . 1928 • AMERICAN JOURNAL OF BOTANY

Barrett , S. C. H. 2002 . Th e evolution of plant sexual diversity. Nature Reviews. Dudash , M. R. 1991 . Plant size eff ects on female and male function in her- Genetics 3 : 274 – 284 . maphroditic Sabatia angularis (Gentianaceae). Ecology 72 : 1004 – 1012 . Barrett , S. C. H. , A. M. Baker , and L. K. Jesson . 2000 . Mating strategies in Elzinga , J. A. , and S. Varga . 2017 . Prolonged stigma and flower lifespan monocotyledons. In K. L. Wilson and D. A. Morrison [eds.], Systematics in females of the gynodioecious plant Geranium sylvaticum. Flora 226 : and evolution of monocots, 256–267. CSIRO, Sydney, Australia. 72 – 81 . Barrett , S. C. H. , and J. Hough . 2013 . Sexual dimorphism in fl owering plants. Eppley , S. M. , and J. R. Pannell . 2009 . Inbreeding depression in dioecious pop- Journal of Experimental Botany 64 : 67 – 82 . ulations of the plant Mercurialis annua : Comparisons between outcrossed Baucom , R. S. , R. Mauricio , and S. M. Chang . 2008 . Glyphosate induces tran- progeny and the progeny of self-fertilized feminized males. Heredity 102 : sient male sterility in Ipomoea purpurea. Botany 86 : 587 – 594 . 600 – 608 . Bawa , K. S. 1980 . Evolution of dioecy in fl owering plants. Annual Review of Freeman , D. C. , K. T. Harper , and E. L. Charnov . 1980 . Sex change in plants: Ecology and Systematics 11 : 15 – 39 . Old and new observations and new hypotheses. Oecologia 47 : 222 – 232 . Bishop , E. J. , R. B. Spigler , and T. L. Ashman . 2010 . Sex-allocation plasticity Givnish , T. J. 1980 . Ecological constraints on the evolution of breeding sys- in hermaphrodites of sexually dimorphic Fragaria virginiana (Rosaceae). tem in seed plants-dioecy and dispersal in Gymnosperms. Evolution 3 4 : Botany 88 : 231 – 240 . 959 – 972 . Calviño , A. , and L. Galetto . 2010 . Variation in sexual expression in relation to Grazer , V. M. , M. Demont , Ł. Michalczyk , M. J. G. Gage , and O. Y. Martin . plant height and local density in the andromonoecious shrub Caesalpinia 2014 . Environmental quality alters female costs and benefi ts of evolving gilliesii (Fabaceae). Plant Ecology 209 : 37 – 45 . under enforced monogamy. BMC Evolutionary Biology 14 : 21 . Campbell , D. R. 2000 . Experimental tests of sex-allocation theory in plants. Han , Y. , C. Dai , C. F. Yang , Q. P. Wang , and T. J. Motley . 2008 . Anther ap- Trends in Ecology & Evolution 15 : 227 – 232 . pendages of Incarvillea trigger a pollen-dispensing mechanism. Annals of Caruso , C. M. , and A. L. Case . 2007 . Sex ratio variation in gynodioecious Botany 102 : 473 – 479 . Lobelia siphilitica : Eff ects of population size and geographic location. Hendrix , S. D. , and E. J. Trapp . 1981 . Plant-herbivore interactions: Insect Journal of Evolutionary Biology 20 : 1396 – 1405 . induced changes in host plant sex expression and fecundity. Oecologia 49 : Case , A. L. , and S. C. H. Barrett . 2004 . Environmental stress and the evolu- 119 – 122 . tion of dioecy: Wurmbea dioica (Colchicaceae) in Western Australia. Korpelainen , H. 1998 . Labile sex expression in plants. Biological Reviews of the Evolutionary Ecology 18 : 145 – 164 . Cambridge Philosophical Society 73 : 157 – 180 . Chang , S. M. 2006 . Female compensation through the quantity and quality Lankinen , Å. , and K. K. Green . 2015 . Using theories of sexual selection and of progeny in a gynodioecious plant, Geranium maculatum (Geraniaceae). sexual confl ict to improve our understanding of plant ecology and evolu- American Journal of Botany 93 : 263 – 270 . tion. AoB Plants 7 : plv008 . Chang , S. M. 2007 . Gender specifi c inbreeding depression in a gynodioecious Lehtilä , K. , and S. Y. Strauss . 1999 . Eff ects of foliar herbivory on male and fe- plant, Geranium maculatum (Geraniaceae). American Journal of Botany 94 : male reproductive traits of wild radish, Raphanus raphanistrum. Ecology 80 : 1193 – 1204 . 116 – 124 . Charlesworth , B. , and D. Charlesworth . 1978 . A model for the evolution of Liao , W. , and D. Y. Zhang . 2008 . Increased maleness at fl owering stage and dioecy and gynodioecy. American Naturalist 112 : 975 – 997 . femaleness at fruiting stage with size in an andromonoecious perennial, Charlesworth , D. 2002 . Plant sex determination and sex chromosomes. Veratrum nigrum. Journal of Integrative Plant Biology 50 : 1024 – 1030 . Heredity 88 : 94 – 101 . Lloyd , D. G. , and K. S. Bawa . 1984 . Modifi cation of the gender of seed plants in Charlesworth , D. 2006 . Evolution of plant breeding systems. Current Biology varying conditions. Evolutionary Biology 6 : 255 – 338 . 16 : R726 – R735 . Lu , Y. , Y. B. Luo , and S. Q. Huang . 2012 . Eff ects of soil moisture and fl o- Charlesworth , D. 2013 . Plant sex chromosome evolution. Journal of ral herbivory on sexual expression in a gynodioecious orchid. Journal of Experimental Botany 64 : 405 – 420 . Systematics and Evolution 50 : 454 – 459 . Charlesworth , D. , and B. Charlesworth . 1979 . A model for evolution of hetero- Mamut , J. , Y. Z. Xiong , D. Y. Tan , and S. Q. Huang . 2014 . Pistillate fl owers styly. American Naturalist 114 : 467 – 498 . experience more pollen limitation and less geitonogamy than perfect fl owers Charnov , E. L. 1979 . Simultaneous hermaphroditism and sexual selection. in a gynomonoecious herb. Th e New Phytologist 201 : 670 – 677 . Proceedings of the National Academy of Sciences, USA 76 : 2480 – 2484 . Matsui , K. 1995 . Sex expression, sex change and fruiting habit in an Acer rufi - Charnov , E. L. 1982 . Th e theory of sex allocation. Monographs in Population nerve population. Ecological Research 10 : 65 – 74 . Biology 18. Princeton University Press, Princeton, New Jersey, USA. Maurice , S. 1999 . Gynomonoecy in Silene italica (Caryophyllaceae): Sexual Charnov , E. L. , J. J. Bull , and J. M. Smith . 1976 . Why be an hermaphrodite? phenotypes in natural populations. Plant Biology 1 : 346 – 350 . Nature 263 : 125 – 126 . Meagher , T. R. 1984 . Sexual dimorphism and ecological diff erentiation of male Dawson , T. , and M. Geber . 1999 . Sexual dimorphism in physiology and mor- and female plants. Annals of the Missouri Botanical Garden 71 : 254 – 264 . phology. In M. L. Geber., T. E. Dawson, and L. F. Delph [eds.], Gender and Moore , J. C. , and J. R. Pannell . 2011 . Sexual selection in plants. Current Biology sexual dimorphism in fl owering plants, 175–215. Springer, Berlin, Germany. 21 : R176 – R182 . Delesalle , V. A. 1989 . Year to year changes in phenotypic gender in a mon- Nanami , S. , H. Kawaguchi , and T. Yamakura . 2005 . Sex ratio and gender- oecious cucurbit Apodanthera undulata. American Journal of Botany 76 : dependent neighboring eff ects in Podocarpus nagi , a dioecious tree. Plant 30 – 39 . Ecology 177 : 209 – 222 . Delph , L. F. 1990 . Sex-diff erential resource allocation patterns in the subdioe- Parachnowitsch , A. L. , and E. Elle . 2004 . Variation in sex allocation and male– cious shrub Hebe subalpina. Ecology 71 : 1342 – 1351 . female trade-off s in six populations of Collinsia parvifl ora (Scrophulariaceae Delph , L. F. 2003 . Sexual dimorphism in gender plasticity and its consequences s.l.). American Journal of Botany 91 : 1200 – 1207 . for breeding system evolution. Evolution & Development 5 : 34 – 39 . Parra-Tabla , V. , V. Rico-Gray , and M. Carbajal . 2004 . Eff ect of defoliation Delph , L. F. , and C. R. Herlihy . 2012 . Sexual, fecundity, and viability selec- on leaf growth, sexual expression and reproductive success of Cnidoscolus tion on fl ower size and number in a sexually dimorphic plant. Evolution 66 : aconitifolius (Euphorbiaceae). Plant Ecology 173 : 153 – 160 . 1154 – 1166 . Peruzzi , L. , E. Mancuso , and D. Gargano . 2012 . Males are cheaper, or the ex- Diggle , P. , V. S. Di Stilio , A. R. Gschwend , E. M. Golenberg , R. C. Moore , treme consequence of size/age dependent sex allocation: Sexist gender di- J. R. W. Russell , and J. P. Sinclair . 2011 . Multiple developmental pro- phasy in Fritillaria montana (Liliaceae). Botanical Journal of the Linnean cesses underlie sex differentiation in angiosperms. Trends in Genetics 2 7 : Society 168 : 323 – 333 . 368 – 376 . Philipp , M. , and T. Hansen . 2000 . Th e infl uence of plant and corolla size on Doust , J. L. , and P. B. Cavers . 1982 . Sex and gender dynamics in Jack-in-the- pollen deposition and seed set in Geranium sanguineum (Geraniaceae). pulpit, Arisaema triphyllum (Araceae). Ecology 63 : 797 – 808 . Nordic Journal of Botany 20 : 129 – 140 . DECEMBER 2017 , VOLUME 104 • ABDUSALAM ET AL.—VARIATION IN SEXUAL EXPRESSION OF GERANIUM TRANSVERSALE • 1929

Queenborough , S. A., D. F. R. P. Burslem , N. C. Garwood , and R. Valencia . Van Etten , M. L. , and S. M. Chang . 2014 . Frequency-dependent pollinator 2007 . Determinants of biased sex ratios and inter-sex costs of reproduc- discrimination acts against female plants in the gynodioecious Geranium tion in dioecious tropical forest trees. American Journal of Botany 9 4 : maculatum. Annals of Botany 114 : 1769 – 1778 . 67 – 78 . Varga , S. , and M. Kytöviita . 2016 . Light availability aff ects sex lability in a gy- Renner , S. S. 2014 . Th e relative and absolute frequencies of angiosperm sexual nodioecious plant. American Journal of Botany 103 : 1928 – 1936 . systems: Dioecy, monoecy, gynodioecy, and an updated online database. Vitt , P. , K. E. Holsinger , and C. S. Jones . 2003 . Local diff erentiation and plas- American Journal of Botany 101 : 1588 – 1596 . ticity in size and sex expression in jack-in-the-pulpit Arisaema triphyllum Renner , S. S. , L. Beenken , G. W. Grimm , A. Kocyan , and R. E. Ricklefs . 2007 . (Araceae). American Journal of Botany 90 : 1729 – 1735 . Th e evolution of dioecy, heterodichogamy, and sex expression in Acer. Webb , C. J. 1999 . Empirical studies: Evolution and maintenance of dimorphic Evolution 61 : 2701 – 2719 . breeding systems. In M. A. Geber, T. E. Dawson, and L. F. Delph [eds.], Renner , S. S. , and R. E. Ricklefs . 1995 . Dioecy and its correlates in the fl owering Gender and sexual dimorphism in fl owering plants, 61–95. Springer, New plants. American Journal of Botany 82 : 596 – 606 . York, New York, USA. R u ff atto , D. M. , D. N. Zaya , and B. Molano-Flores . 2015 . Reproductive suc- Weiblen , G. D. , R. K. Oyama , and M. J. Donoghue . 2000 . Phylogenetic analysis cess of the gynodioecious Lobelia spicata LAM.(Campanulaceae): Female of dioecy in monocotyledons. American Naturalist 155 : 46 – 58 . frequency, population demographics, and latitudinal patterns. International Williams , C. F. , M. A. Kuchenreuther , and A. Drew . 2000 . Floral dimorphism, Journal of Plant Sciences 176 : 120 – 130 . pollination, and self-fertilization in gynodioecious Geranium richardsonii Sakai , A. K. , and S. G. Weller . 1999 . Gender and sexual dimorphism in fl ower- (Geraniaceae). American Journal of Botany 87 : 661 – 669 . ing plants: A review of terminology, biogeographic patterns, ecological cor- Willson , K. F. 1979 . Sexual selection in plants. American Naturalist 113 : relates, and phylogenetic approaches. In M. A. Geber, T. E. Dawson, and L. 777 – 790 . F. Delph [eds.], Gender and sexual dimorphism in fl owering plants, 1–31. Wolfe , L. M. 1998 . Regulation of sex expression in desert and Mediterranean Springer, New York, New York, USA. populations of an andromonoecious plant (Gagea chlorantha , Liliaceae). Sarkissian , T. S. , S. C. H. Barrett , and L. D. Harder . 2001 . Gender variation in Israel Journal of Plant Sciences 46 : 17 – 25 . Sagittaria latifolia (Alismataceae): Is size all that matters? Ecology 82 : 360 – 373 . Wolfe , L. M. , and A. Shmida . 1997 . Th e ecology of sex expression in a gyno- Sun , S. G. , Y. Lu , and S. Q. Huang . 2006 . Floral phenology and sex expres- dioecious Israeli desert shrub (Ochradenus baccatus ). Ecology 78 : 101 – 110 . sion in functionally monoecious Rhoiptelea chiliantha (Rhoipteleaceae). Xu , L. R. , and C. Aedo . 2008 . Geraniaceae . In Z. Y. Wu and P. H. Raven [eds.], Botanical Journal of the Linnean Society 152 : 145 – 151 . Flora of China, vol. 24, 73–263. Science Press, Beijing, China, and Missouri Th omas , S. C. , and J. V. LaFrankie . 1993 . Sex, size, and inter-year variation Botanical Garden Press, St. Louis, Missouri, USA. in fl owering among dioecious trees of the Malayan rainforest understory. Yakimowski , S. B. , and S. C. H. Barrett . 2014 . Variation and evolution of sex Ecology 74 : 1529 – 1537 . ratios at the northern range limit of a sexually polymorphic plant. Journal of Th omson , J. D. , and S. C. H. Barrett . 1981 . Selection for outcrossing, sexual selec- Evolutionary Biology 27 : 1454 – 1466 . tion, and the evolution of dioecy in plants. American Naturalist 118 : 443 – 449 . Yasuor , H. , J. Riov , and B. Rubin . 2007 . Glyphosate-induced male sterility in Torices , R. , and M. Mendez . 2011 . Infl uence of infl orescence size on sexual glyphosate-resistant cotton (Gossypium hirsutum L.) is associated with in- expression and female reproductive success in a monoecious species. Plant hibition of anther dehiscence and reduced pollen viability. Crop Protection Biology 13 : 78 – 85 . (Guildford, Surrey) 26 : 363 – 369 . Vallejo-Marín , M. , and M. D. Rausher . 2007 . Th e role of male fl owers in an- Zemp , N. , R. Tavares , and A. Widmer . 2015 . Correction: Fungal infection in- dromonoecious: Energetic costs and siring success in Solanum carolinense duces sex-specifi c transcriptional changes and alters sexual dimorphism in L. Evolution 61 : 404 – 412 . the dioecious plant Silene latifolia. PLOS Genetics 11 : e1005536 . Vallius , E. , and V. Salonen . 2000 . Eff ects of defoliation on male and female Zhang , Z. Q. , X. F. Zhu , H. Sun , Y. P. Yang , and S. C. H. Barrett . 2014 . Size- reproductive traits of a perennial orchid, Dactylorhiza maculata. Functional dependent gender modifi cation in Lilium apertum (Liliaceae): Does this Ecology 14 : 668 – 674 . species exhibit gender diphasy? Annals of Botany 114 : 441 – 453 .