Floral traits of mammal‐pollinated Mucuna macrocarpa (Fabaceae): Implications for generalist‐like pollination systems

Abstract Floral traits are adapted by plants to attract pollinators. Some of those plants that have different pollinators in different regions adapt to each pollinator in each region to maximize their pollination success. Mucuna macrocarpa (Fabaceae) limits the pollinators using its floral structure and is pollinated by different mammals in different regions. Here, we examine the relationships between floral traits of M. macrocarpa and the external morphology of mammalian pollinators in different regions of its distribution. Field surveys were conducted on Kyushu and Okinawajima Island in Japan, and in Taiwan, where the main pollinators are the Japanese macaque Macaca fuscata, Ryukyu flying fox Pteropus dasymallus, and red‐bellied squirrel Callosciurus erythraeus, respectively. We measured the floral shapes, nectar secretion patterns, sugar components, and external morphology of the pollinators. Results showed that floral shape was slightly different among regions and that flower sizes were not correlated with the external morphology of the pollinators. Volume and sugar rate of nectar were not significantly different among the three regions and did not change throughout the day in any of the regions. However, nectar concentration was higher in Kyushu than in the other two regions. These results suggest that the floral traits of M. macrocarpa are not adapted to each pollinator in each region. Although this plant limits the number of pollinators using its flower structure, it has not adapted to specific mammals and may attract several species of mammals. Such generalist‐like pollination system might have evolved in the Old World.

Previous studies showed that pollinators induce selective pressures on floral traits. Plant adaptations to specific pollinators can occur when their distribution ranges closely overlap with those of their pollinators. However, when the distributional ranges of plants do not overlap with those of their specific pollinators, some plants change their flowering phenology, shape, color, nectar secretion pattern, and/or volatile components to attract another pollinator (Boberg et al., 2014;Nagano et al., 2014;Sun, Gross, & Schiestl, 2014). Finally, plants may speciate in each site (Fleming et al., 2001;Forest et al., 2014;Gowda & Kress, 2013). Otherwise, a plant might adapt its floral traits to several pollinator species (i.e., become generalized).
Insect-or bird-pollinated plants have been targeted in previous studies, and no studies have investigated the adaptations of mammal-pollinated plants to different mammalian pollinators in different regions. Although the diversity of mammal-pollinated plants is relatively low compared with insect-pollinated plants, there are many mammal-pollinated species throughout the world, especially in the tropics (Carthew & Goldingay, 1997;Fleming & Kress, 2013).
However, pollination ecology of mammal-pollinated plants has been studied mainly in Australia and Africa, with only a few studies in Asia (Willmer, 2011).
Mucuna macrocarpa (Fabaceae) is a woody vine plant which is distributed from Southeast Asia to Kyushu, Japan (Tateishi & Ohashi, 1981). This flower is papilionaceous and utilizes a special pollination stage called "explosive opening" (Toyama, Kobayashi, Denda, Nakamoto, & Izawa, 2012). Stamens and pistils are tightly enclosed by the keel petals and are exposed when the flower opens explosively. The explosive opening is facilitated by different mammals (explosive openers) in different regions: Japanese macaques (Macaca fuscata) and Japanese martens (Martes melampus) in Kyushu; Ryukyu flying foxes (Pteropus dasymallus) on Okinawajima Island; and redbellied squirrels (Callosciurus erythraeus), Formosan striped squirrels (Tamiops maritimus), and masked palm civets (Paguma larvata) in Taiwan Kobayashi et al. 2017;Toyama et al., 2012). Among them, Japanese macaques and redbellied squirrels open the most flowers in Kyushu and Taiwan, respectively, compared with other explosive openers. Thus, these two species are the main pollinators in these regions Kobayashi et al. 2017). Except for Japanese macaques, all explosive openers hold the wing petals with their forelegs and insert their snouts into the gap between the wings and banner petals and then push up the banner using their snout to feed on the nec-  (Kobayashi, 2017). Although bees collect pollen and stigma attaches to their body in some cases, most pollen grains are removed by explosive openers Kobayashi et al. 2017;Toyama et al., 2012). Thus, explosive openers are considered as the main pollinators Kobayashi et al. 2017;Toyama et al., 2012).
In this study, we aimed to reveal the relationships between floral traits (shape and nectar) and characteristics of explosive openers (external morphology and daily activity patterns) in three regions.
We examined the following hypotheses: (1) Floral shapes differ among the three regions; (2) face size, which is an important trait for opening flowers explosively, is different among explosive openers; (3) floral shape correlates with face size of explosive openers; (4) nectar secretion patterns differ among the three regions; and (5) nectar secretion patterns correlate with the activity patterns of the main explosive openers.  (Figure 1). Kyushu is a large island, but distribution of M. macrocarpa is limited to the study site. This species grows in the southward steep slope .

| Study sites
The annual mean temperature (2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017) in Kamae, the nearest meteorological observation point, is 17.4°C, and annual mean precipitation is 2,401 mm (Japan Meteorological Agency, 2018). On Okinawajima Island, this species is distributed throughout the island and mainly grows along the valley (Kusumoto & Enoki, 2008). The annual mean temperature (2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017) in Oku, the nearest meteorological observation point of OY, is 20.8°C, annual mean precipitation is 2,611 mm, and those in Naha, the nearest meteorological observation point of OU, are 23.4°C and 2,159 mm (Japan Meteorological Agency, 2018). In Taiwan, this species is also distributed from north to south, but elevation of distribution area is ranging from 100 to Bureau, 2018). All study sites are evergreen forests, but the flora is quite different among these sites. In addition, the individual planted in TK originated from near the study site. were Japanese martens, Ryukyu flying foxes, red-bellied squirrels, Formosan striped squirrels, and masked palm civets, which use their snouts for explosive opening Kobayashi et al. 2017;Toyama et al., 2012). Japanese macaques were the main openers in Kyushu; however, as they opened flowers using their hands, we did not measure any facial parts. Samples of Japanese martens were those killed by cars, which were stored in the freezer (n = 16), and live individuals captured for ecological research (n = 10) on the Tsushima Islands, Japan. Captive Ryukyu flying foxes at the University of the Ryukyus were measured (n = 16).

| Measurements of flowers and explosive openers
Eleven were rescued from the wild, and five were born in captivity.
Specimens of red-bellied squirrels (n = 31) and masked palm civets (n = 17) stored in the National Museum of Natural Science of Taiwan and Endemic Species Research Institute were measured. We could not find available specimens of striped squirrels, so we used the mean value of skull data (n = 4) measured by Chang (2011). The external morphology of opener snouts was compared using PCA.
Finally, the relationship between the floral gap size and snout size of the explosive openers was investigated in each region. When the explosive openers opened flowers using their snouts, the direction of snout insertion into the flower gap was almost fixed in the rightside upward direction (see Kobayashi et al. 2017;Toyama et al., 2012). Thus, the gaps of flower banners were compared using the width of rhinarium or width of upper jaw of openers, and the gaps between the tops of banners and the tops of flower wings were compared using the height of rhinarium or height of upper jaw of openers (see Figure 2).

| Measurement of nectar
To investigate the daily change in nectar secretion, the volume, weight, and sugar concentration of nectar were measured every 3 hrs. Mature flowers were picked from four individuals in Kyushu, five individuals (OY: 1, OU: 4) on Okinawajima and six individuals (TB: 5, TK: 1) in Taiwan. Nectar was collected from inside the calyx, and volume was measured using a microsyringe (MS-N100; Ito Corporation, Tokyo, Japan). Nectar weight was measured using a digital portable balance (TR-SC30; Pepaless, Hyogo, Japan) and sugar concentration (Brix index) was measured using a handheld

| Comparison between floral and opener shapes
All flower parts were significantly different among regions, with the longest length observed in Kyushu and the shortest on Okinawajima (ANOVA for each flower trait; p < 0.05) ( Table 1). The maximum and minimum values of each part were observed in the different regions.
The PCA of floral parts showed that floral shape had the least difference among the three regions (Figure 3). In addition, the external morphology of the openers' snouts differed among the three regions. Snouts of Japanese martens and masked palm civets were larger than those of squirrels (Figure 4). Snouts of Ryukyu flying foxes were medium-sized; however, the rhinarium was larger than that of any other opener (Figure 4).

The analysis of sugar component ratio showed that sucrose
was dominant in all regions. The ratio of sucrose was over 65%, and the range of mean sugar ratio was 1.93-2.07 in the three regions ( Figure 7). In addition, the sugar component ratio between morning and night was the same in all regions (chi-square test; Kyushu: χ 2 2 = 0.00, p > 0.05, Okinawajima: χ 2 2 = 0.00, p > 0.05, Taiwan: χ 2 2 = 0.05, p > 0.05) (Figure 7). Many previous studies clarified that some plants adapt their floral shapes to the main pollinators in each region when the main pollinator differs among regions (Anderson & Johnson, 2008;Boberg et al., 2014;Johnson & Steiner, 1997). In this study, the floral shape and size of M. macrocarpa slightly differed among regions. In addition, there were specific mammalian openers in each region Kobayashi et al. 2017;Toyama et al., 2012), and snout sizes of these openers were also different among regions.

| D ISCUSS I ON
However, contrary to our third hypothesis, floral dimensions did not correlate with snout sizes of main openers. Thus, we concluded that this plant did not adapt its floral shapes to individual pollinators. One possible reason might be that mammals have a higher intraspecific size variation (including differences between sexes or among ages) than insects or birds and cannot exert sufficient selection pressure on the plants they pollinate.
Nectar secretion patterns were not adapted to mammalian openers in each region in this study. Bat-pollinated plants generally secrete nectar at night (Faegri & van der Pijl, 1979;Willmer, 2011).
The concentration and sugar ratio of nectar in Mucuna are determined by the species of pollinator (Agostini et al., 2011;Liu et al., 2013), except for M. macrocarpa (Table 3). Although the squirrelpollinated M. sempervirens has some similarities in nectar secretion patterns (i.e., nectar concentration, sugar ratio, and daily variation) (Table 3) Although the floral traits of M. macrocarpa, such as shape and nectar secretion patterns, did not match the main pollinators in each region, there were small differences among regions. Floral traits only differed in biotic factors, but also in abiotic factors (e.g., Campbell, 1996;Petanidou, Goethals, & Smets, 2000). Therefore, abiotic factors, such as temperature, precipitation, and soil conditions, should be considered as well as genetic drift caused by bottlenecks and/or founder effects on floral traits in future.
In the eastern Caribbean, bird-pollinated plants have different pollinators in different regions and have evolved their floral traits depending on the traits of the pollinator in each region (Gowda & Kress, 2013;Temeles & Kress, 2003). In the Sonoran Desert, batpollinated cactus species are known to adapt to both bird and bat pollinators (i.e., become more generalized) (Fleming et al., 2001 (Schrire, 2005), and many species are pollinated specifically by birds or bats (Agostini, Sazima, & Sazima, 2006;Cotton, 2001;Grünmeier, 1993;Hopkins & Hopkins, 1993;Sazima, Buzato, & Sazima, 1999). However, plants pollinated by multiple mammals are found only in Asia Kobayashi et al. 2017;Toyama et al., 2012). One reason why bat-pollinated plants have become specialized or adapted to local vertebrate pollinators in the New World is the high diversity of nectarfeeding specialist birds and bats (Fleming & Kress, 2013). Conversely, the diversity of nectar-feeding mammals is low in the Old World, especially in Asia (see Fleming & Kress, 2013), while the diversity of omnivorous mammals, such as squirrels, macaques, and civets, is high in TA B L E 3 Comparison of components of nectar secretion patterns among species Southeast Asia (Corlett, 2007). Therefore, in case the nectar-feeding specialist bats and birds are lacking, plants have evolved nonflying mammal-dependent and/or generalist-like pollination systems.
However, this hypothesis is based on the information of one specific genus. Mammalian pollinators are not well clarified in Asian regions. Further researches clarifying pollinators of mammalpollinated plants are needed to establish this hypothesis.

ACK N OWLED G M ENTS
We thank Atsushi Nakamoto (Okayama University of Science), Yoko Okawara, Hiroaki Ui, Keiichiro Abe (University of the Ryukyus), National Museum of Natural Science (Taiwan), and Taiwan

CO N FLI C T O F I NTE R E S T
None declared.