Different pollinator assemblages ensure reproductive success of Cleisostoma linearilobatum (Orchidaceae) in fragmented holy hill forest and traditional tea garden

Orchids are generally recognized to have specialist pollination systems and low fruit set is often thought to be characteristic of the family. In this study, we investigated the reproductive ecology of Cleisostoma linearilobatum, an epiphytic tropical orchid, in a holy hill forest fragment and a traditional tea garden in SW China using comparable methods. C. linearilobatum is self-compatible and dependent on insects for pollination. Fruit production in natural conditions was both pollinator- and resource-limited. However, the natural fruit set remained stable over multiple years at both sites. Pollination observations showed that C. linearilobatum has a generalized pollination system and seven insect species were observed as legitimate pollinators. Although the visit frequencies of different pollinators were different in the two sites, the pollinator assemblages ensured reproductive success of C. linearilobatum in both study sites over multiple years. The results partly explain why C. linearilobatum is so successful in the area, and also suggest that holy hill forest fragments and traditional tea gardens in Xishuangbanna are important in preserving orchids, especially those with generalist pollination.

crops are not significant habitats for orchids in Xishuangbanna, but traditional tea gardens, where tea is grown under a canopy of native trees, support a high diversity of orchids 19,20 .
In our systematic field surveys of orchid species diversity in Xishuangbanna, we found that Cleisostoma linearilobatum is the most abundant species in both 'holy hills' and traditional tea gardens at elevations between 1000-1600 m. We were curious about why this orchid was so successful in these different habitats and therefore decided to compare the reproductive ecology of C. linearilobatum in holy hills and traditional tea gardens. Here we present the results of our investigations, which addressed three principal questions: (1) Is the natural fruit set of C. linearilobatum different between holy hills and traditional tea gardens? Does its fruit production remain stable between years? (2) What are the pollinators of C. linearilobatum, and is the pollinator assemblage different between the two different habitats? (3) Why is C. linearilobatum so successful in this area?

Results
Floral phenology and morphology. Cleisostoma linearilobatum was the most abundance orchid species in our study sites, with 66 mature individuals found in in HHF and 777 in TTG. It flowered from May to September (mainly early June to late July) in the study sites. It produces 1-3 paniculate pendulous inflorescences from the base of the stem. The panicle, with mean length of 17.72 ± 6.46 cm (N = 30), contains 57.00 ± 33.10 flowers (N = 30; Fig. 1c). Flowers were small, purplish-red in colour (Fig. 1d), and opened gradually from the bottom to the top of each panicle.

Hand-pollination experiments and natural fruit sets.
In the four hand-pollination treatments, no fruit was found in the bagging and emasculating treatments in either site ( Table 1), suggesting that spontaneous autogamy and apomixis did not occur in C. linearilobatum. The fruit sets of the selfing and crossing treatments were not significantly different within or between HHF and TTG, but both were significantly higher than natural fruit sets in both study sites (P selfing & natural < 0.001, P crossing & natural < 0.001; Table 1). These results indicate that C. linearilobatum is completely self-compatible and dependent on insects for fertilization. The natural fruit sets of C. linearilobatum in the two study sites were stable over the 3 years from 2011 to 2013, with no significant differences between sites (F = 0.140, P > 0.05; Table 1) or among years (F = 0.138, P > 0.05; Table 1).

Floral visitor observations.
Overall, 140 h of visitor observations were made at HHF over three years (2011 to 2013), and 84 h were made at TTG over two years (2012 and 2013). A total of seven insect species were identified as pollinators of C. linearilobatum in the two study sites. Among them, 6 species were observed in HHF and 5 species in TTG, with 4 species common to both sites. All 7 insects were observed visiting flowers for nectar and carrying pollinia away when leaving flowers. Eumenes sp.1, with the most visits observed, was the main pollinator of C. linearilobatum in HHF (Fig. 1e). It visited flowers more frequently and for longer periods than the other 5 species in three observation years, but was not observed in TTG (Fig. 3). In TTG, four insect species, Prionyx sp., Amegilla yunnanensis, Megachile dimidiate and Vespa sp. (Fig. 1f-i) visited flowers of C. linearilobatum frequently and for longer periods, and there was no significant difference in their visiting frequencies between the two years (t Prionyx = 0.577, P > 0.05, t Megachile = − 0.310, P > 0.05, t Amegilla = 0.091, P > 0.05, t Vespa = − 0.378, P > 0.05; Fig. 3). Although these four insects were also observed in HHF, their visiting frequencies were significantly lower than in TTG, and the visiting frequency of Amegilla sp. was significantly different among years (F = 8.034, P < 0.05). A species of Apis (Fig. 2j) was occasionally observed visiting flowers of C. linearilobatum in HHF, but not in TTG; another species, Eumenes sp. 2 (Fig. 2k), was only observed in TTG. In 2013, diurnally bagged flowers showed no evidence of insect visitation at night, as their pollinia remained intact in the next morning.

Discussion
Orchids are primarily pollination limited, and many factors may affect male and female reproductive success including floral traits 5 . Cleisostoma linearilobatum produces many flowers per inflorescence and these flowers have nectar as a reward for pollinators. The lack of fruit production in the bagging treatments indicated that C. linearilobatum was dependent on insects for pollination, and that spontaneous self-pollination did not occur. The fruit sets of selfing and crossing treatments were not significantly different, indicating that C. linearilobatum is completely self-compatible. Natural fruit sets were significantly lower than the fruit sets of any selfing and crossing treatments, which were also relatively low (54.08-67.00%, Table 1 ), suggesting that fruit production of C. linearilobatum in natural conditions was both pollinator-and resource-limited.
Most orchids have specialist pollination systems and about 60% have only one recorded pollinator 7,8,21 . The advantage of specialization is that pollen transfer between flowers is very efficient 22,23 , but these plants are often pollinator limited, which is the most frequent reported cause of reproductive failure in orchids 5,24,25 . Generalist pollination may alleviate this problem, partucularly in habitats that are marginal for insect activity 13,14 . Our observations suggested that Cleisostoma linearilobatum possess a generalized pollination system with 7 morphologically similar and effective pollinators in the two study sites. All 7 insects are very common in the study area and similar in body size. Amegilla yunnanensis, Megachile dimidiate, and many species of Prionyx, Eumenes and Apis are knew as common pollinators of other flowering plants. Although the pollinator assemblages and the visiting frequencies of different pollinators were different, the natural fruit sets of C. linearilobatum were stable over the years in both study sites (Fig. 3). This may partly explain why C. linearilobatum is so successful in this area.
Xishuangbanna is one of the orchid hotspots in China 16,17 . However, most species are threatened due to the habitat loss and over-collection. Although about two-thirds of orchid species occur in nature reserves, there are still 142 species recorded only outside the nature reserves, and most of these are found in forest fragments 18 . A recent investigation suggested that fragmentation had no significant impact on the species diversity and abundance of epiphytic orchids at elevations both above and below 1 000 m, but this may reflect the short history of forest fragmentation in this area 26 . Some studies have shown that fragmentation may directly affect pollinators of orchids and lead to reproductive failure in many orchid species 27,28 . Nevertheless, for C. linearilobatum, even in the seriously fragmented habitat of HHF, the natural fruit set remained stable over the years of the study and did not differ from TTG. Six insect species were observed as pollinators of C. linearilobatum in HHF. The visiting frequencies of different insects may differ between years, but all pollinators contributed to a stable fruit set.
In Xishuangbanna, the major commercial crops are rubber and tea. Rubber plantations are not suitable habitats for orchids in Xishuangbanna, but traditional tea gardens, where tea trees are grown under a canopy of native trees, support a high diversity of orchids 19,20 . About 5494 ha of these tea gardens still persist in Xishuangbanna at elevations of 1200-1800 m 29 and they are also important for other native species 19 . This situation is very similar to the role of shade-coffee and cacao plantations in South America as refuges for tropical wild orchids 30,31 . Our study confirmed that in addition to the nature reserves, the 'holy hills' and traditional tea gardens in Xishuangbanna also act as orchid refuges, especially for those species with generalist pollination. As suggested 18 , continued protection with incentives for the continuation of the traditional conservation practice for holy hills and with traditional management for tea gardens are certainly needed and important.    Methods Study sites. This study was conducted in a holy hill forest of the Dai minority and a traditional tea garden of the Jinuo minority in Xishuangbanna, Yunnan province, SW China. All holy hill forests and traditional tea gardens in Xishuangbanna are out of the nature reserves. Holy hills are sacred forested hills believed to be the dwelling place of the spirits of ancestors by Dai people, and are traditionally protected by customary rules. Nowadays, there are 328 holy hill forests remaining in the area, but they have been severely fragmented and vary greatly in size 20 . The traditional tea gardens, where tea is grown under a canopy of native trees, support a high diversity of orchids and other plants, and the premium paid for the tea they produce in a traditional manner has ensured that 5494 ha persist in Xishuangbanna at altitudes of 1200-1800 m 20 .

Study species. Cleisostoma
Blume is a genus with around 100 species distributed in the Asian tropics. Little is known about the pollination biology of Cleisostoma except that C. parishii was said to be autogamous 32 . Cleisostoma linearilobatum is a small epiphytic orchid widely distributed in Southeast Asia including Xishaungbanna in Southwest China 33 . In our study areas, it usually grows on tree trunks in tropical montane evergreen broad-leaved forests at altitudes of 1000-1600 m.
In our study sites, a total of 13 orchid species in 10 genera were found in HHF, while 38 orchids of 17 genera were found in TTG, growing both on tea trees and shade trees 20  Hand-pollination experiments and natural fruit sets. The breeding system of C. linearilobatum was evaluated by different hand-pollination treatments in the two study sites in 2011. The treatments were: (i) bagging, inflorescences were bagged throughout without pollination; (ii) emasculating, pollinia were removed before flowers opened and then inflorescences were bagged throughout; (iii) selfing, inflorescences were bagged before flower opening and flowers were hand-pollinated with pollinia from the same flower, and then inflorescences were bagged again; (iv) crossing, inflorescences were bagged before flower opening and flowers were hand-pollinated with pollinia of another individual, and then inflorescences were bagged again.
All treatments were conducted in HHF and TTG during 7-13 and 9-15 June 2011, respectively. The fruit set of each treatment was counted about 4 weeks later (28-30 July). The natural fruit sets of C. linearilobatum in the two study sites were investigated by randomly marking inflorescences from different individuals for three consecutive years (2011-2013). The numbers of flowers, inflorescences and individuals that were used in hand-pollination treatments and investigations of natural fruit set are summarized in Table 1. Statistical data analysis. The difference in floral longevity among treatments and in visiting frequency in HHF among three consecutive years was analyzed using one-way ANOVA, and the visiting frequency in TTG between 2012 and 2013 was compared using a Paired-Samples T-test. A General Linear Model was used to explore the relationships between natural fruit set among different years and between study sites, and the fruit set of different hand-pollination treatments, and between study sites. All statistical analyses were performed by SPSS ver. 13.0 for Windows.