Ya-P. He 1, Yuan-W. Duan 1,4, Jian-Q. Liu 2,1, and W. K. Smith 3

Transcription

1 Pl. Syst. Evol. 256: (2006) DOI /s Floral closure in response to temperature and pollination in Gentiana straminea Maxim. (Gentianaceae), an alpine perennial in the Qinghai-Tibetan Plateau Ya-P. He 1, Yuan-W. Duan 1,4, Jian-Q. Liu 2,1, and W. K. Smith 3 1 Laboratory of the Qinghai-Tibetan Plateau Biological Evolution and Adaptation, Northwest Institute of Plateau Biology, the Chinese Academy of Sciences, Xining, P. R. China 2 Key Laboratory of Arid and Grassland Ecology, Lanzhou University, Lanzhou, P. R. China 3 Department of Biology, Wake Forest University, Winston-Salem, North Carolina, USA 4 Graduate School of the Chinese Academy of Sciences, Beijing, P. R. China Received February 1, 2004; accepted April 18, 2005 Published online: 23 September 2005 Ó Springer-Verlag 2005 Abstract. Floral closure may be induced by pollination and various other factors, but is rarely studied comprehensively. Different kinds of floral closure should have various effects on reproductive fitness of plants. Two contrasting types of floral closure were observed in the flowers of Gentiana straminea Maxim. in the eastern Qinghai-Tibetan Plateau. The first type occurred prior to pollination during both gender phases, in response mainly to decreasing air temperatures. Flowers closed when decreasing temperatures approached 20 C and subsequently began to reopen the following day during mid-morning when air temperatures warmed to approximately C. This kind of floral closure can protect pollen grains on either stamens or stigmas, increasing fitness of both male and female. Following pollination, permanent floral closure occurred, although there was a delay between the dates of pollination and permanent closure, during which flowers continued to show temporary closure in response to low temperature episodes. The time required for permanent, pollination-induced closure varied according to the age of the gender phase, including a prolonged time before closure if pollination occurred early in the female phase. The retaining of permanent closed flowers increased both approaching (to inflorescences) and visiting (to unpollinated flowers) frequencies of individual plants when with fewer open flowers and the persisting corolla is further beneficial for seed sets of these pollinated flowers. Thus, two separate types of floral closure, one in response to environmental cues and the other in response to the age of each gender stage, appeared to have a strong influence on reproductive fitness in this species. These results revealed a different adaptive strategy of alpine plants in the sexual reproduction assurance in addition to the well-known elevated floral longevity, dominant role of more effective pollinators and increased reproduction allocation in the arid habitats. Key words: Floral closure, temperature, pollination, gender phase, fitness, seed set. In most angiosperms, flowers appear to have a fixed lifespan that is independent of the occurrence of pollination (Primack 1985). In contrast, flower longevity in some species appears to be tightly coupled to the deposition or removal of pollen, although flowers may

2 18 Ya-P. He et al.: Floral closures of Gentiana straminea persist after pollination. Moreover, this persistence may provide an adaptive advantage in attracting pollinators to unpollinated flowers on the same plant (Jones and Cruzan 1999). These species with pollination-induced floral closure usually have flowers that change color, or other morphological traits, indicating the termination of sexual function. However, relatively little is known regarding the duration of sexual function in relation to total flower lifespan, or the possible adaptive advantages of their interaction (Primack 1985). Recent studies on floral morphological changes that appear related to pollination dynamics have shown considerable diversity among different species (e.g. Weiss 1991, 1995; Simons 1992; Oberrath and Bo hning-gaese 1999; Jones and Cruzan 1999). Common floral changes include colour, termination of odor and/or nectar production, changes in flower orientation, and the withering of floral structures and corolla abscission (Gori 1983). Floral closures were frequently observed in gentians (e.g. Mu ller 1881; Weber 1924; Costeloe 1988; Petanidou et al. 1995a, b, c; Kozuharova and Anchev 2001, 2002, 2004), but the adaptive values of such closures appear to be rarely studied in the control experiments (but see Bynum and Smith 2001). Goldsmith and Hafenrichter (1932) reported that flowers of Gentiana algida began to close when temperature decreased under experimental conditions in the laboratory, while Bynum and Smith (2001) found similar results under field conditions and concluded that floral closure was induced by temperature decreases in G. algida associated with afternoon thunderstorms. Moreover, this floral behavior appeared adaptive, preventing pollen removal (rainwash) or damage to paternal and maternal reproductive structures within the upright, tubular corollas of this species. The distribution of gentians is centred in the mountains of Southwest China, particularly the eastern Qinghai-Tibetan Plateau of China. Although more than two thirds of all species of the family are found here, most of which are endemic to this area (Ho and Liu 2001), the reproductive adaptation of these important alpine components on the largest plateau remains poorly understood. Flowers of Gentiana straminea Maxim. from different populations on the Qinghai-Tibetan Plateau were observed to close at night, and only some would reopen on the following day. Open flowers during the day also closed in association with afternoon thunderstorms, reopening with the return of sunlight and warming temperatures. G. straminea is highly self-compatible, but pollinator is indispensable for seed production (He and Liu 2004). It remains unknown how much these closures have contributed to the reproduction success of this outcrossing species. The objectives of this study were to evaluate the apparent occurrence of two distinct types of floral closure, one that is transient and in apparent response to environmental cues and a second that is permanent and induced by pollination. In addition, we also evaluated the adaptive values of both temperature-induced closure and pollination-induced closure, based on seed set of flowers open in all the flowering time and the timing of pollination related to the age of a particular gender phase, and the subsequent success of seed set and contribution to other unpollinated flowers on the same inflorescence. These results will increase our knowledge of the sexual reproduction adaptation of alpine plants to the arid habitats in addition to their previously revealed strategies, such as increased reproduction allocation, the elevated floral longevity and dominant role of more effective pollinators and increased reproduction allocation (e.g. Bingham and Orthner 1998, Ko rner 1999, Bingham and Ranker 2000, Fabbro and Ko rner 2004). Materials and methods Study site and species. The study site was located at the Haibei Alpine Meadow Ecosystem Research Station of the Chinese Academy of Sciences. This station is located on the northeast Qinghai-Tibetan Plateau at N and E,

3 Ya-P. He et al.: Floral closures of Gentiana straminea 19 and an elevation of 3244 m. The average annual air temperature is )1.7 C with an annual mean maximum of 27.6 C in summer months and mean minimum of )37.1 C during winter (Zhao and Zhou 1999). Mean annual precipitation ranges from 426 to 860 mm, with eighty percent occurring in the short summer growing season from May to September, primarily in July and August. This alpine meadow area is dominated by Kobresia humilis (C. A. Mey.) Serg., while Gentiana straminea Maxim. (Gentianaceae) occurs as a subdominant herb. We carried out the following experiments from 2000 to 2003, and G. straminea initiated flowering in mid-july, 2000, in late July, 2001 and 2002 and mid-august in 2003, and continued flowering until early September, 2000, mid-september, 2001 and 2002 and early October in 2003, respectively. G. straminea, one perennial species of Gentiana sect. Cruciata Gaudin, occurs in alpine environments and high mountains of China and has a welldeveloped basal vegetative rosette with a terminal bud that grows continuously throughout the life of the plant. Annual flowering stems originate from axillary buds on the outer leaves of the basal rosette. Individual plants have 1 6 stems and each stem bears 6 13 flowers in a simple terminal or axillary cymes. Flowers are relatively large (7 cm in length) and have an upright, funnel-shaped yellowish corolla comprising five connate petals. Unlike other arctic gentians, which reproduce both sexually and vegetatively from rhizomes (e.g. G. algida, Bynum and Smith 2001), G. straminea relies solely on sexual reproduction on the Qinghai- Tibetan Plateau (personal observation). Our previous study indicated that G. straminea is highly selfcompatible, but autonomous self-pollination is completely avoided through temporal and spatial separation of male and female reproductive structures (herkogamy and dichogamy) of individual flowers since no seed was produced when flowers were isolated (He and Liu 2004). The field observations revealed that bumblebees (Bombus sushikini Skorikov) were the primary, if not sole, pollinators of G. straminea, matching the relatively large size of the corolla opening in this species (Duan et al. 2005). There are also easily recognized morphological changes that distinguish the termination of male and female phases, i.e. stamens twisting towards the corolla and permanent floral closure, respectively. Temperature-induced floral closure. The temporal dynamics of floral closure was monitored (N=30, five individual plants) in a 2 m 2 m area for 17 days from 16 July to 1 August, During this time period, a wide range of weather conditions occurred, including sunny days, sunny days with thunderstorms in the afternoon and cloudy days. Flower opening and closing was monitored hourly by measuring flower (corolla) opening diameter, along with ambient air temperatures every five minutes while opening and closing were in process in the morning and the closing process in the afternoon during the summers of 2001 (N=18, 18 individual plants at least 5 m apart) and 2002 (N=20, 20 individual plants at least 5 m apart). We also manipulated microenvironmental temperature experimentally to evaluate the influence of temperature on the closing and opening response of flowers in Open flowers (N=5) were given a chilling treatment with ice in water-proofed bag in clear skies (full sunlight) from 13:10 pm to 13:30 pm when the temperature of environment surrounded was above 20 C. Then these flowers were exposed in natural environment but still in shade after they closed completely during this time period to investigate effect of illumination on floral closure. In addition, closed flowers (N=5) were warmed at night by electronic heater. During these two processes, besides measuring their corolla diameter, temperature of environment closed to flowers, relative humidity and illumination intensity (the illumination intensity is zero at night) were also monitored every 5 minutes in daytime and 10 minutes at night. Air temperatures and relative humidity adjacent to individual corollas were measured using a ventilated and shielded thermostat Hydro-Thermometer (Sigma Inc.), while illumination intensity was measured using light meter (Sigma Inc.). Ten individual plants were collected in the field, potted in natural soils, and then carried back to the Haibei Station laboratory in Similar to the procedures of Bynum and Smith (2001), experiments were designed to exclude the possibility that floral closure was in response to changes in incident solar radiation. Different levels of incident irradiance were generated using daylight lamps (measured with a LI-190 sensor for measuring photosynthetically active wavelengths, 0.3 to 0.7 lm; LICOR INC, Lincoln, Nebraska, USA) and air temperature inside a growth cham-

4 20 Ya-P. He et al.: Floral closures of Gentiana straminea ber (model: SPX-250-BG, YUEJIN INC, Shanghai, China) were used to test whether floral closure was induced by incident irradiance level, temperature, and the relative humidity. The male to female stage flowers with different opening duration were tested. Influence of temperature-induced closure on reproduction. To investigate the effects of temperature-induced closure on reproduction, transparent adhesive tape was used to stick around corolla to prevent flowers (N=30) from closing during their flowering time in Seeds from these flowers were compared to those from naturally pollinated flowers. In addition, ten flowers which began to shed pollen just were selected before rain, and the corolla of five flowers were fixed with transparent adhesive tape to keep them open during rain, but other five flowers were kept untouched. Transparent adhesive tape was narrow and stuck to the corolla outside edge to prevent floral closure; these treated flowers showed no morphological appearance with those intact flowers as control. During good weather stage, pollinators visited these treated flowers like the intact flowers. After rain, stamens of all flowers were collected and fixed in FAA in Eppendorf respectively to count number of pollen grains in laboratory under microscope. Rain may decrease pollen germination rate through the damage of rainy dilution (Dafni 1992, Huang et al. 2002). Dafni s (1992) method of pollen germination was followed. Pollen grains from dehisced anthers were placed in sucrose solutions with concentration of 0%, 5%, 10% and 15% in Eppendorf tubes of 1 ml in Both total number of pollen and number of pollen that had germinated in both two experiments were counted under a light microscope after 8 hours. In each germination sample, at least 2000 pollen grains were counted and this experiment was repeated four times. Sexual longevity and pollination-induced closure. In 2002, flower buds of G. straminea in the field were subjected to two different treatments: unbagged controls (N=30, 30 individual plants), and flowers covered with transparent plastic bags (N=29, 29 individual plants). The bag-isolated flowers opened and closed in the same manner as those left open to natural pollination. In order to avoid the greenhouse effects and air block of plastic bags, we perforated the bags. Because bumblebees are the only pollinators of G. straminea in Haibei population, these treatments to bags do not influence our experiment results. The reproductive status of all flowers were monitored continuously throughout a day, along with the degree of floral closure, the timing of permanent closure, and the duration of male and female phases for both control and bagged flowers. Individual flowers (N=128 individual plants) were also isolated before initiation of the female phase, then handpollinated (N=10) on each of the first nine days of the female phase using fresh pollen from different individuals (ensuring outcrossing). One or more anthers were brushed across the unfolded stigma until its surface was entirely covered with pollen. Because stigma receptivity has been reported to approach zero on the 8 th to 10 th day in two other gentian species (Webb and Littleton 1987), the hand-pollination experiments were not prolonged beyond the 9 th day. In these experiments, the time period from pollination to permanent closure was recorded, as well as the occurrence of temporary closure in response to natural temperature changes. To evaluate the relationship between seed set and the length of sexual longevity of the stigma, fruits in the latter group of treatments were harvested before dehiscence. Seed set ratios were calculated using the following formula: seed set = Number of mature seeds/(number of mature seeds + Number of aborted seeds + Number of unfertilized ovule), the same as percent of seed production (Webb and Littleton 1987). Effect of pollination-induced closure on reproductive effort. Firstly, we haphazardly selected 30 pairs of opposite flowers in similar size on different inflorescences to avoid resource allocation bias within the plants. All the flowers were emasculated and bagged before the pistillate stage and then hand-pollinated them with pollen from another plant when the stigma had been bifurcate for about 4 days. Then for each pair, the corolla of one flower was abscised as soon as the flower was pollinated, while another flower was left intact. All flowers were isolated again after hand-pollination, and the fruits were collected in paper bags before they dehisced for seed sets, seed mass and germination. In order to investigate whether flowers closed permanently still contributed to the attraction of other open flowers on the same inflorescence, in full flowering period of the population in 2002, we designed inflorescences in different display by removing flowers, i.e. 2 2, 2 10, 6 2 and

5 Ya-P. He et al.: Floral closures of Gentiana straminea , in which the former denoted the number of open flowers without pollination, while the latter denoted the number of permanent closed flowers after pollination. We arranged these displays into two different comparisons: 2 2 vs and 6 2 vs During the observation, we recorded the number of bumblebees approaching the inflorescence to evaluate approaching frequency, and number of visits to individual flowers to calculate visiting frequency of bumblebees to the unpollinated flowers. Data analysis. The correlation between flower diameters and environmental factors was tested using Pearson Correlation Analysis and Microsoft Excel Seed set variability from pollination treatments (different days from the start of the female phase) was tested by Post hoc-lsd. Oneway ANOVA was used to test for the statistical significance of differences between the sexual organ longevity of the free-pollinated controls and bagisolated treatments. The Independence-samples T Test was used to compare pollen grain number of flowers with stuck corolla and seed sets of fruits from flowers with intact corolla and those from flowers without flowers, approaching frequencies and visiting frequencies to artificially designed Fig. 1. Temperature changes and the corresponding opening (open dots) or closing (solid dots) of unpollinated flowers of Gentiana straminea during 12 hours of a whole day. The numbers (0 to 23) denote the time that starts from 7:30 in the morning to 19:00 in the afternoon. (A) A whole sunny day; (B) A sunny day with thunderstorm in the afternoon, (C) A sunny day in the morning but with clouds in the afternoon; (D) A whole cloudy day

6 22 Ya-P. He et al.: Floral closures of Gentiana straminea Fig. 2. Correlations between corolla diameters (y-axis) and ambient temperature (x-axis) of unpollinated 449 flowers of Gentiana straminea. The correlation equation was y = , x ) , r =0.606,p <0.001 display between 2 2 and 2 10, and 6 2 and 6 10 (SPSS Inc. 2002). All statistical tests were checked for validity and accuracy as described in Zar (1999). Results Temperature-induced floral closure. The closing dynamics of unpollinated flowers on individual days selected as representative of the four distinct weather patterns that commonly occur in this area are shown in Fig. 1. From observations over a 17 day period of continuous monitoring, flowers began to open each day when temperatures rose to 11 C and became fully open at 18 C on sunny or cloudy days (Fig. 1A, C). However, flowers started to close when temperatures decreased to near 20 C during late afternoon. On several rainy days, temperatures in the studied area never rose to 11 C, and all flowers remained closed. On sunny days with afternoon thunderstorms, which occur frequently in the Qinghai-Tibetan Plateau during July and August, flowers were typically open before the storm approached, but closed during it s passing when temperatures fell to below 20 C, Fig. 3. Changes of corolla diameter vs. temperature adjacent to corolla, illumination intensity and relative humidity every ten minutes when flowers of Gentiana straminea were cooled with ice in water-proofed bags and then exposed under natural condition but still in shade where temperature of surrounding environment was above 20 C in Corolla diameter is denoted by open squares, temperature by open diamonds, relative humidity by solid squares and illumination intensity by solid diamonds. The correlation coefficients between corolla diameter and temperature, relative humidity, illumination were respectively (p < 0.01),)0.756 (p < 0.01) and )0.178 (p =0.526)

7 Ya-P. He et al.: Floral closures of Gentiana straminea 23 Table 1. Temperature dependence of the floral opening and closing processes of Gentiana straminea in the morning and evening Means ± standard deviation (sample size) Independence sample T-test Opening process (18) Closing process (20) T-value P-value Initial temperature ( C) ± ± 2.02 ) Final temperature ( C) ± ± Range of temperature 8.75 ± ± 3.11 ) change ( C) Duration time (min) ± ± ) Temperature change ratio ( C /min) ± ± and then re-opened again with warming temperatures (Fig. 1B). Cloudy days with maximum temperatures of approximately 15 C resulted in only a few hours, or less, with open flowers (Fig. 1D). A relatively strong, linear relationship occurred between ambient temperature and flower diameter in the opening and closing processes (y = x ) , r = 0.606, p < 0.001, N=449; Fig. 2). Also, although air temperature increased slightly faster in the morning than it decreases in the afternoon, final temperatures were nearly identical for the opening and closing cycles (Table 1). For the gradually changed temperature on sunny days, flowers took over two hours from the time of initial closing to the fully open state, but slightly less time (96 minutes) from the start of the closure to full closure (t = )1.959, df = 36, P = 0.058). However, according to our detailed records during observations on closing dynamics of unpollinated flowers during the thunderstorm, these flowers took less than 15 minutes to close before thunderstorm and less than 20 minutes to reopen when the temperature increased above 15 C (Fig. 1B). In the laboratory, such a process became faster than that during the thunderstorm when the temperature manually decreased and increased in a fast way within several minutes. Fig. 4. Changes of corolla diameter vs. temperature and relative humidity every five minutes when flowers of Gentiana straminea were heated at night when the illumination intensity was zero in The symbols are the same as those in Fig. 3. The correlation coefficients between corolla diameter and temperature, relative humidity were respectively (p < 0.01) and )0.934 (p <0.01)

8 24 Ya-P. He et al.: Floral closures of Gentiana straminea When the flowers were cooled with ice and then exposed in natural condition in full skies, the closing and opening response could be observed easily. In the two processes, changes in corolla diameter showed highly correlated with both temperature and relative humidity (Fig. 3). The correlation coefficients between corolla diameter and temperature, relative humidity were (P < 0.01), )0.756 (P < 0.01) respectively. But changes in corolla diameter appeared to correlate illumination poorly (Fig. 3), for the correlation coefficient was only )0.178 (P = 0.526). The same results could also be observed when flowers were heated at night without illumination (Fig. 4). The correlation coefficients between corolla diameter, temperature and relative humidity were (P < 0.01), )0.934 (P < 0.01), respectively. Table 2 shows the floral opening and closure under the controlled experiments through a combination of temperature and illumination in the laboratory under the relative humidity between 5% and 40%. When the temperature was less than 15 C, flowers remained closed even under the high irradiance level (226 lmol m )2 s )1 PAR) and remained open when the temperature increased above 20 C under fully dark conditions (0.0 lmol m )2 s )1 ). The flowers with different opening duration from the male to the female phase with more than 3 days showed no corresponding difference in the opening and closing process. When the temperature was controlled at 20 C or10 C, the flowers kept open or closed even the relative humidity varied from 5% to 40%. Influence of temperature-induced closure on reproduction. After raining, pollen grains of stamens of the stuck-corolla flowers without transient closures at the male stage decreased more than 30% compared with those of the control (N=5). The effects of rain damage on the pollens were evaluated by the germination rate of simulated dilution in sucrose solution. On average, over 60% pollen grains germinated in 10% and 15% sucrose solution, and there was no significant difference between them. However, germination rate of pollen grains in distilled water was low, and there was significantly difference between 0 and 10% sucrose solution (P < 0.01) (Table 3). These results indicated that pollen grains after rain suffer from decreased germination rate Fig. 5. The time period from pollination to permanent closure (Mean ± 1SE) after hand-pollination at the different stigmatic duration ages (days) of Gentiana straminea

9 Ya-P. He et al.: Floral closures of Gentiana straminea 25 Table 2. Opening and closing states of Gentiana straminea under different temperature and relative illumination intensity in laboratory (3/3, maximum illumination, 226 lmol m )2 s )1 PAR). ) indicated that flowers remained closed, and + opening Relative illumination intensity Temperature ( C ) 0 1/3 2/3 3/3 5 ) ) ) ) 10 ) ) ) ) 15 ) ) ) ) through rainy dilution. Fruit set, number of seeds and seed set from flowers with stuck corolla were far low compared to control, but individual seed mass and seed germination rate from flowers with stuck corolla were higher than control (Table 4). Sexual longevity and pollination-induced floral closure. The distinct morphological indicator marking the transition from male to female phases (the stamens curving towards the corolla lobes), allowed us to record the duration of the two phases separately. Initially, Fig. 6. Seed set (Mean ± 1 SE) of outcrossing pollination at different stigmatic duration of the female phase flowers of Gentiana straminea. A correlation test is shown in Table 4 the delay between pollination and permanent closure was unknown, so we defined the female phase as lasting from the initial open of stigma to the occurrence of permanent closure. The male phase of flowers left to free pollination lasted 3.03 ± 1.19 days, and the female phase 5.11 ± 2.18 days before permanent closure (Table 5). During both male and female phases, all flowers displayed temporary closures that corresponded to cooler temperatures. The bagged flowers had a longer male phase (4.75 ± 1.07 for bagged flowers) than the free-pollination flowers (F = , df = 1, P < 0.001). All bagged flowers never showed permanent closure before turning brown and wilting during the 24-day observation period. However, all laboratory flowers in the female phase showed temporary closures, opening and closing with rises and falls in temperature, respectively. Flowers isolated in the laboratory never closed permanently till wilting. Hand-pollinated flowers did not close immediately on the day of pollination, but closed late in the afternoon similar to the unpollinated flowers (temporary closure). Also, all pollinated flowers reopened and closed on the day following the day of pollination. The time period from pollination to permanent closure was approximately two days if pollination occurred during the first to fourth day of the female phase, and longer in flowers pollinated at a later stage (Fig. 5). The older the female phase at the time of

10 26 Ya-P. He et al.: Floral closures of Gentiana straminea Table 3. No. of pollen grains from one flower in staminate stage of Gentiana straminea and their germination rate in different concentration of sucrose. Values with the different subscripted letters are statistically significant different at 0.01 level Concentration of sucrose No. of pollen grains germinated Total No. of pollen grains Germination rate (%) Mean S. E. Mean S. E. Mean S. E. 0% a % b % c % c 1.80 pollination, the longer was the delay to permanent closure (Pearson Correlation, r=0.287, n=106, P=0.005). Furthermore, permanently closed flowers had no visible change in color compared to unpollinated flowers until the seeds had matured (about a month later). Because most flowers continued to open for 5.1 ± 2.18 days from the end of the male phase to permanent closure (Table 5), the pollination of these flowers occurred from around the second to fifth day of the female phase. The seed set ratio for hand-pollinated plants on different days of the female phase peaked on the 2 nd, 3 rd, 4 th, but decreased sharply on the 5 th day (Fig. 6, Table 6). This decrease continued to be negatively associated with stigmatic duration of the female phase (r=)0.468, n=128, p<0.001). Effect of pollination-induced closure on reproductive effort. Compared to fruits from flowers with intact corolla, both number of seed per fruit (p<0.01) and seed set (p<0.01) decreased significantly after corolla were abscised, indicating protection by the corolla during process of seed development, but seed germination rate (p=0.919) did not change greatly (Table 7). According to observations to inflorescences with artificially designed displays, both approaching (to inflorescences) and visiting (to unpollinated flowers) frequencies increased significantly when there were 2 open flowers on one inflorescence (Table 8). Nevertheless, when there were more open flowers on one inflorescence, bumblebees showed no preference to designed inflorescences either with fewer or more closed-flowers, for no difference was found in both approaching frequency (t = 1.45, p=0.15) and visiting frequency (t = 0.40, p = 0.69) to 6 2 display and 6 10 display (Table 8). Although the bumblebees Table 4. Effect of corolla stuck on reproductive variables over the entire flowering period of Gentiana straminea in (Mean ± 1SE, with sample size in parenthesis). The mean difference is significant at the 0.01 level in one row Structure Treatment Percentage reduction Control Stuck Fruit set (%) 80 (N=20) 40.7 (N=27) 50.4 No. of seed/fruit 122 ± 13* (N=20) 43 ± 12* (N=27) 64.8 Seed set (%) 63.6 ± 5.5* (N=20) 20.6 ± 5.8* (N=27) 67.6 Individual seed mass (mg) ± (N=10) ± (N=10) )18.2 Seed germination rate (%) 82.0 ± 0.00 (N=5) 86.4 ± 0.00 (N=5) )5.4

11 Ya-P. He et al.: Floral closures of Gentiana straminea 27 approached the inflorescences with permanently closed flowers, they never visited these closed flowers. Discussion Adaptive benefits of two floral closures. Two different types of floral closure appeared to exist in G. straminea on the Qinghai-Tibetan Plateau of Northern China, a temporary closure elicited by cooling air temperatures, plus a permanent closure in response to pollination. Unpollinated flowers of G. straminea that were experimentally isolated (bagged) from pollinators in the field (or by placing them in the laboratory) never closed permanently like pollinated flowers before they wilted, although they continued to undergo temporary closure in response to cooler temperatures. Each of these responses has been reported to occur separately in different gentian species (Webb and Littleton 1987, Bynum and Smith 2001), but not together. Furthermore, the data reported here confirm previous results reported for G. algida where sunlight level did not influence closure, and temperature decreases associated with afternoon thunderstorms showed a strong correlation with temporary floral closure (Bynum and Smith 2001). However, both temperature and relative humidity seemed to show strong correlations with temporary closure of G. straminea in the chilling and warming experiments, while illumination level also did not influence closure (Fig. 3 and Fig. 4). But flowers did not respond to the change of the relative humidity between 5% and 40% in the laboratory when the temperature remained constant. Likewise, the change of relative humidity between 25% and 75% in the controlled experiments of G. algida did not trigger the floral closures and reopening of this species (Bynum and Smith 2001). The opening and closing of the unpollinated flowers of G. straminea were mainly triggered by temperature and the irradiance and the relative humidity have fewer effects on such a process. The correlation of the relative humidity to this process in the field is mainly caused by the fluctuation of the temperature. These observations are consistent with the findings of Goldsmith and Hafenrichter (1932) and Bynum and Smith (2001) on the floral movements of G. algida. The flowers of this species did not demonstrate diurnal rhythms in opening times and would remain open under the constant temperature and close under the decreased temperature. This sensitivity was dependent upon the magnitude of the temperature change regardless of the time of day. Short-term floral closure in G. algida in response to frequent afternoon rain showers was suggested to be adaptive, providing protection for reproductive structures within the upright, tubular flowers. Damage to female parts and removal of pollen by rainfall was associated with decreased reproductive success in G. algida (Bynum and Smith 2001). Pollen grains of G. straminea decreased greatly in number after rain and this direct damage obviously reduces its paternal fitness. In addition, wetting of pollen is known to affect viability and germinability, in some species, by either increasing mortality or inducing germination prior to pollination (see reviews by Bynum and Smith 2001). G. straminea pollen grains were seriously damaged in water with a low proportion of germination (Table 3). This was also observed in Pulsatilla cernua of Ranunculaceae (Huang et al. 2002). Because rain and thunderstorms in the Qinghai-Tibetan Plateau occur frequently in July and August when G. straminea is at its peak flowering period, the temporary closure of this species obviously protects its pollen grains from rain damage. Moreover, rain could potentially affect pollen dispersal by altering flower-pollinator interactions. The stuck-corolla flowers were observed to gather and retain rainwater for several hours, which presumably diluted the nectar of these flowers. Because the visits of bumblebees are affected by differences in nectar quality (Marden 1984), it is likely that rainfall could reduce nectar quality and results in fewer pollinator visits to G. straminea flowers. Compared with those of the control plants, the fruit set and number of seeds and

12 28 Ya-P. He et al.: Floral closures of Gentiana straminea seed set of the permanently stuck flowers of G. straminea decreased more than 50% (Table 4). These maternal decreases might result from such an affect with fewer pollinators, removal of pollens on the stigmas by rain and possible low pollen germination when without closure at the low temperature. The failure to close after pollination in these stuck-corolla flowers might also affect female reproductive efforts (see the following discussions). Both paternal and maternal fitness of G. straminea could be negatively impacted via several mechanisms if flowers were not able to close in response to the falling temperature, which occurs during the thunderstorms and night. Floral changes following pollination vary greatly among angiosperms and include changes in color, scent or nectar production, orientation, withering, and abscission (Gori 1983). Pollination-induced closure might reduce the costs of open flowers, which include increased transpirational water loss (Nobel 1977), floral respiration (Ashman and Schoen 1994), nectar production (Schemske 1978, Harder and Barrett 1992, Ashman and Schoen 1997), maintaining flowers in stressful environments (Gori 1983) and retention of the corolla (and its color). In addition, floral closure induced by pollination may enhance floral display on a whole-plant basis, attracting pollinators from even greater distances (Jones and Cruzan 1999, Oberrath and Bo hning- Gaese 1999), which could act beneficially to direct pollinators to unpollinated flowers (e.g. Kerner 1902, Arditti and Flick 1976, Stead and Moore 1979, Casper and La Pine 1984, Oberrath and Bo hning-gaese 1999), leading to an increase in foraging and pollination efficiency (Gori 1983, Weiss 1991). Our results demonstrated that if there were fewer open flowers on an inflorescence of G. straminea, increased closed number of flowers might attract more pollinators from a long distance and increase both approaching frequency of the inflorescences and visiting frequency of unpollinated flowers. However, this did not occur if there were enough open flowers on an inflorescence (Table 8). Webb and Littleton (1987) suggested that permanent closure following pollination might protect ovaries from predators, pollen tube growth and accelerate development of the fertilized ovules. Our experiments demonstrated that fruit and seed set had been greatly reduced if corolla was abscised, which confirmed that closed corolla could protect growth of pollen tube from arid environments and accelerate development of the fertilized ovules. Therefore, permanent closure and retained-corolla followed pollination should play an important role in enhancing male and female fitness especially in later flowering period of the whole population. Floral longevity, pollination, and sexual function. Recent studies on floral longevity have indicated that floral longevity in male and female phases shows a plastic response to fitness, and is affected by variables such as pollen removal and deposition (Devlin and Stephenson 1984, 1985; Richardson and Stephenson 1989; Preston 1991; Proctor and Harder 1995; Sargent and Roitberg 2000; Evanhoe and Galloway 2002; Sato 2002). However, these studies did not evaluate interactions between the length of the female phase and the timing of pollination, i.e. the relation between the time of pollen deposition relative to stigma lifespan. For example, permanent floral closure induced by pollination might be prolonged if the stigma is not optimally receptive to fertilization at the time of pollination, enhancing the probability and efficacy of fertilization. In the present study, the corolla of G. straminea was retained until seeds developed to maturity. The floral lifespan measured here for G. straminea was greater than its sexual longevity, and the male phase lasted longer in bagged than in unbagged control flowers. Most likely, this was the result of pollen removal in control plants by pollinators who accelerated completion of the male phase, for there was inverse relationship between the frequency of insect visits and pollen retention (Mallick 2001), indicating no pollinator limitation in G. straminea. Similarly, for the shortage of pollinators resulting in no pollen

13 Ya-P. He et al.: Floral closures of Gentiana straminea 29 deposition on the stigma in bagged flowers, the female phase increased from about 5 days to almost 24 days, or until flowers were fully withered. Seed set obtained after hand-pollination of flowers demonstrated that stigma receptivity was highest on day 2 5 of the female phase (Fig. 6). Subsequently, seed sets decreased with stigma ageing. However, most flowers in this area were pollinated during the four days of when stigmas were at their peak seed-set. Webb and Littleton (1987) also reported that seed set in two gentian species decreased to zero on the 8 th and 10 th days of the female phase, compared to the 9 th day reported here, and seed set of G. straminea still kept 0.40, obviously higher. When hand-pollinated in the first five days of the female phase, flowers did not initiate permanent floral closure immediately, but maintained temporary closure patterns for approximately 2 days (Fig. 5). This delay in permanent closure also increased with the duration of the female phase beyond day 5 (Fig. 5). There could be several possible explanations for the observed delay between pollen deposition and the morphological change indicating pollination and permanent floral closure in G. straminea. First, ovule fertilization, and not pollen deposition, may be a possible cue leading to permanent floral closure. Although large numbers of pollen grains were manually deposited on the stigma, some grains may not have germinated, or there may have been an inadequate amount of time for grown pollen tubes to achieve full growth. Such an explanation seems particularly plausible in the light of previous studies on other species in which ovule fertilization has been found not to occur for 24 or more hours after pollination (Mazer et al. 1986, Cruzan 1989, Oberrath and Bo hning-gaese 1999). Second, continued opening of pollinated flowers enhances and extends the attractiveness of the plant s floral display, especially if insufficient pollen deposition delays permanent closure, ensuring that pollinators will visit and pollinate again. This latter mechanism may continue to function even if the stigmas receive adequate pollen grains under hand-pollination conditions. Since the duration of the female phase may be prolonged under suboptimal conditions for pollen deposition (Evanhoe and Galloway 2002), the greater delay between the start of the morphological and functional female phase may be adaptive. Ultimately, such coordination between pollen deposition and stigma receptivity may lead to higher fertilization efficiency. The reductions in the male and female phases induced, respectively, by pollen removal and deposition are in agreement with previous studies (Devlin and Stephenson 1984, 1985; Richardson and Stephenson 1989; Preston 1991; Proctor and Harder 1995; Sargent and Roitberg 2000; Evanhoe and Galloway 2002; Sato 2002). These studies also demonstrated that the male and female phase gradually shortens with the accrual of better Table 5. Male and female duration of flowers from opening to permanent closure of Gentiana straminea in different treatments manipulated artificially in the field. The male phase was calculated from opening to stamen twisted towards corolla and the female duration from stigma opening to corolla closure or wilting, and the total floral duration from flower opening to closure or wilting Duration (sample size, mean duration ± Std. Dev.) One-way ANOVA Free pollination Bag-isolation F value P value Male phase (days) ± ± <0.001 Female phase (days) ± ± <0.001 Total flower duration (days) ± ± <0.001

14 30 Ya-P. He et al.: Floral closures of Gentiana straminea Table 6. A test of the seed sets through hand pollination and the stigmatic durations of the female phase of Gentiana straminea by Post Hoc Tests (LSD). The mean difference is significant at the 0.01 level * * * )0.8 ) * )6.6 )8.0 ) )16.4* )17.8* )15.6* )9.8 7 )4.7 )25.2* )26.5* )24.3* )18.6* )8.7 8 )16.4* )36.9* )38.3* )36.1* )30.3* )20.5* ) )24.6* )45.1* )46.5* )44.3* )38.5* )28.7* )19.9* )8.2 fitness levels. Schoen and Ashman (1995) predicted similar results, based on their floral longevity model, suggesting that if costs of floral maintenance are constant, floral longevity should be shortened in response to increased fitness. Although a few previous studies have found differences in the durations of the morphological and functional female phases, this difference did not vary greatly with the duration of the female phase (e.g. Evanhoe and Galloway 2002). This variable delay could also be considered a part of the floral longevity response to variations in fitness (such as changes in stigma seed-set). Based on a survey of floral longevity in angiosperms, Ashman and Schoen (1996) predicted that floral longevity measured under field conditions was very close to the maximum possible for the species being considered. The total floral longevity in G. straminea (male and female longevity) for flowers experiencing free pollination (6 8 days) was much shorter than floral longevity under isolated conditions (25 28 days). However, these results may also indicate that most flowers undergoing free pollination were pollinated around the second to fifth day of the female phase, when the stigmas were most receptive. The floral longevity exhibited by G. straminea in the field likely represents a genetically determined minimum, similar to those observed in G. pneumonanthe (Petanidou et al. 2001) and Campanula americana (Evanhoe and Galloway 2002). Summary and conclusions The current study identifies two types of floral closure and their causes in one gentian, G. straminea from the Qinghai-Tibetan Plateau. Both types of closure appear to contribute to the sexual reproductive success of this species. In addition, the coupling of floral longevity between the male and female phases is affected by variables such as pollen removal and deposition, possibly enhancing reproductive fitness. Gentianaceae species are dominant in alpine ecosystems in the Asian highlands (Ho and Liu 2001) and frequently occurs in other alpine areas (Mu ller 1881, Bynum and Smith 2001). The adaptive values of these two contrasting floral closures revealed here must partly account for the reproductive fitness of other Gentianaceae species in arid habitats, because this phenomenon has been observed in diverse species of this family with outcrossing breeding system although without careful evaluation (e.g. Mu ller 1881; Weber 1924; Costeloe 1988; Petanidou et al. 1995a, b, c; Kozuharova and Anchev 2001, 2002, 2004). Because of the reduced insect diversity, abundance, and activity in alpine ecosystems, classical views predicted increasing self-compatibility, apogamy, and vegetative reproduction with increasing altitude (e.g. Mosquin 1966), but recent researches revealed that insect pollination is far more important than wind pollination and outcrossing appears to be the dominant breeding system in the

15 Ya-P. He et al.: Floral closures of Gentiana straminea 31 Table 7. Seed number, seed set, and germination rate of mature seeds in the abscised corolla (T) and control (C) (Mean ± 1SE, sample size in parenthesis). All flowers were bagged again after hand-pollination. The mean difference is significant at the 0.01 level in one group No. of seed Seed set Germination rate of mature seeds T ± 11.5** (N=20) 0.63 ± 0.05** (N=20) 0.45 ± 0.10 (N=6) C ± 12.6** (N=21) 0.80 ± 0.03** (N=21) 0.44 ± 0.17 (N=6) Table 8. Approaching (to the inflorescence) and visiting (to unpollinated flowers) frequency (times/fower/ 30 min) of the artificial different combinations of opening and permanently closed flowers of Gentiana straminea. In each combination, the former denotes the number of opening flowers, and the latter is the number of the permanently closed flowers. Sample sizes of 2 2 vs and 6 2 vs combinations are respectively 108 and 94 repeats. The mean difference is significant at the 0.05 level in one group Approaching frequency 0.45 ± 0.09* 0.68 ± 0.10* 0.46 ± ± 0.10 Visiting frequency 0.57 ± 0.11* 0.77 ± 0.13* 0.78 ± ± 0.17 alpine ecosystem (Ko rner 1999, Bingham and Ranker 2000). The presumed reduction in pollinator diversity and visitation rate at high altitude was demonstrated to be compensated by the increased floral longevity and the more effective pollinators at both individual species and community level (Primack 1985, Bingham and Orthner 1998). Furthermore, alpine plants allocate more above-ground biomass to flowers, which therefore maintain the flowers longer and guarantee them to have more chances for pollination and adequate seed production (Fabbro and Ko rner 2004). In addition to these mechanisms, two identified floral closures of G. straminea are similarly alternative strategies in guaranteeing sexual reproduction of alpine plants because the successful seed formation within the short growing season is especially crucial for all outcrossing species occurring in the alpine habitats (Ko rner 1999). We thank Ms Zhang Qian and Ms Yang Huiling for their valuable help in field observations and experiments, and Dr. J. Blackwell for editing and improving the English of this manuscript. We are also indebted to Prof. Ekaterina Kozuharova and an anonymous reviewer, who gave constructive suggestions for the manuscript. Dr. He Yaping started the field experiments and investigations in 2000 and 2001, and Dr. Duan Yuanwen continued all observations in 2002 and Both authors contributed equally to this work. Support for this research was provided by National Natural Science Foundation of China (Project ), and the Chinese Academy of Sciences (Key Innovation Plan KSCX-SW-106, Special Fund of Outstanding PhD Dissertation to Liu Jian-quan, and Innovation Research Fund to He Ya-ping). References Arditti J., Flick H. (1976) Post-pollination phenomena in orchid flowers, I. Excised floral segments of Cymbidium. Amer. J. Bot. 63: Ashman T. L., Schoen D. J. (1996) Floral longevity: fitness consequences and resource costs. In: Lloyd D. G., Barrett S. C. H. (eds.) Floral biology. Chapman and Hall, New York. Ashman T. L., Schoen D. J. (1994) How long should flowers live? Nature 371: Ashman T. L., Schoen D. J. (1997) The cost of floral longevity in Clarkia tembloriensis: an experimental investigation. Evol. Ecol. 11: Bingham R. A., Orthner A. R. (1998) Efficient pollination of alpine plants. Nature 391: Bingham R. A., Ranker T. A. (2000) Genetic diversity in alpine and foothill populations of Campanula rotundifolia (Campanulaceae). Int. J. Plant Sci. 161: