Flower-visiting insects of genus Melastoma (Myrtales: Melastomataceae) at the Fushan Botanical Garden, Taiwan

Abstract Background We investigated the diversity and behaviour of insects that visit flowers of four native Melastoma (Family Melastomataceae) species of Taiwan and a horticultural hybrid Melastoma species at the Fushan Botanical Garden, Taiwan biweekly from May to August 2020. Visits of flower-visiting insects were classified into seven behavioural categories, based on the insects' behaviour and positions on the flower. The data are further assigned into four insect-flower interactions, namely pollination, herbivory, commensalism and neutralism. Our goal is to provide baseline data of insect-plant interactions of Melastoma, which is a common, but understudied plant genus in the country. New information A total of 1,289 visits to flowers were recorded by at least 63 insect morphospecies belonging to seven orders. The number of insect species recorded per Melastoma species ranged from 9 to 39. Visiting, sonication and passing were the three most frequently recorded types of behaviour, collectively accounting for 90.2% (n = 1,240) of the total observations. Pollination was the most dominant insect-flower interaction, accounting for 70.2% of the total observations, followed by neutralism (20.0%), herbivory (6.3%) and commensalism (3.5%). Sweat bees of the genera Lasioglossum and Maculonomia (Hymenoptera: Halictidae) are considered key pollinators to Melastoma species in Fushan Botanical Garden, based on their high number of visits and sonication behaviour. Our study provides the first list of insects that visit the flowers of all Taiwan's known Melastoma species and description of their interactions with the plants.


Introduction
With over 5,000 species, Melastomataceae represents one of the largest Angiosperm families distributed in the subtropical and tropical regions around the world (POWO 2019). Members of this flowering family have a complicated evolutionary history (Renner 1993, Stein andTobe 1989) and exhibit diverse morphological traits (Dellinger et al. 2018, Renner 1989, Varassin et al. 2008) and reproduction biology (Dellinger et al. 2019, dos Santos et al. 2012, Peng et al. 2014. The diversification of Melastomataceae is partially a result of hybridisation events. Interspecific hybridisation within a genus (Dai et al. 2012, Hawkins et al. 2016) and between genera (Hawkins et al. 2016) have been reported. Empirical studies suggest that hybridisation in some genera of Melastomataceae are likely mediated by specialised insect pollinators. The pollination syndrome in Melastomataceae is mainly, but not exclusively, dependent on bees (superfamily Apoidea) that are able to vibrate pollen from poricidal anthers by sonication (Renner 1989). Although interspecific hybridisation via insect pollinators has been observed in Melastomataceae native to Asia, studies on insect-flower interactions in Melastomataceae are largely focused on New World species (e.g. Brito et al. 2016, Brito et al. 2017, Pereira et al. 2011, Renner 1989).
There are 18 species belonging to 12 genera of Melastomataceae in Taiwan (Huang and Huang 1996). Of these, Melastoma is the most speciose genus with four species. Two of which, namely Melastoma kudoi Sasaki and M. scaberrima (Hayata) (previously known as Otanthera scaberrima, but see Yang and Liu 2002) are endemic to the Island country, whereas the other two species, M. candidum D. Don and M. malabathricum L. are widely distributed in Asia, the Pacific and Australia (GBIF Secretariat 2019). Amongst the four species, M. kudoi is the rarest species, which has only been recorded from the type locality in central Taiwan. The population of M. kudoi is considered highly threatened and included in the national Red List (Editorial Committee of the Red List of Taiwan Plants 2017, listed as M. intermedia Dunn, but see the recent taxonomy revision by Dai et al. 2019) due to habitat disturbance and lack of inclusion in protected areas (Huang and Huang 1996). The other three Melastoma species can be commonly found in the lowlands up to mid-altitude mountainous areas (Huang and Huang 1996). Despite the great richness of Melastoma species in Taiwan, information about pollinators of these species is limited.
To date, only one study on the pollination biology of one Melastoma species, M. candidum, in Taiwan has been published (Liu et al. 2008). Noteworthy, interspecific hybridisation in this genus is often observed in both wild and cultivated plants in China and Southeast Asia (Cai et al. 2019, Zhou et al. 2017. Although genetic introgression has not been reported from Taiwan, co-occurrence of congeners, including the endangered and endemic M. kudoi, is common in Taiwan (C.J. Lin, unpublished data). Moreover, studies show that the primary pollinators for Melastoma species are non-specialised bees (e.g. Amegilla, Nomia ( Maculonomia) and Xylocopa bees for M. affine ( M. malabathricum), Gross 1993;Bombus, Nomia (Maculonomia) and Xylocopa bees for M. candidum, Liu et al. 2008). These generalist bees are also widely distributed in Taiwan (WCY and SSL, unpublished data) and their habitats commonly overlap with Melastoma species in the country. Whether these bees would visit all Melastoma species remains unknown. Therefore, understanding the pollinator fauna of all Melastoma species in Taiwan is essential to protect the Melastoma diversity, particularly the two endemic species, from potential genetic introgression. In the present project, we present the first checklist of flower-visiting insects of all known Melastoma species in Taiwan, based on empirical data.

Sampling description: Melastoma flower visiting insect survey
Data on the diversity of insects that visited the flowers of all Melastoma species were obtained biweekly at FBG from 7 May 2020 to 19 August 2020. Melastoma species included M. malabathricum L., M. candidum D. Don, M. kudoi Sasaki and M. scaberrima (Hayata). We primarily follow the taxonomy of Yang and Liu (2002), but treat M. septemnervium as a synonym of M. candidum as suggested by the backbone of most catalogues (GBIF Secretariat 2019). Melastoma malabathricum is the only species of the four that is native to this region of Taiwan (CJL, unpublished data). For M. candidum, we included both the typical purple-flowered form and the white-flowered form. Ten wild M. malabathricum individuals were selected along the trails adjacent to the nursery. For the remaining species, including the white flower variant of M. candidum, 10-15 planted individuals for each type/species were used from the nursery. Ten planted individuals of a horticultural hybrid of M. scaberrima and M. kudoi (tentatively named as Melastoma kudoi x Melastoma scaberrima) were also included. For each survey session, observations of insects were made by 2-4 people at the same time for two consecutive days. The observation began roughly 45 mins after sunrise, usually between 6:30 am and 6:45 am and ended at around 11:30 am when flowers began closing or were out of pollen (JCCH, unpublished data). In the early stage of the study by mid-June 2020, continuous observations were made for M. malabathricum in the trails and the rest of the species in the nursery alternatively at 20-min intervals. After the end of M. malabathricum flowering season in mid-June, the observations were made continuously for all samples in the nursery site. Additional data, made by random observations outside of the scheduled survey sessions during the weekly phenology suvey (usually one hour in the morning) in another project during the same period, were also included to maximise our understanding of the diversity of flower-visiting insects. Taxa and behaviour (see the next section for details) of insects with body length > 3 mm present on the adaxial surface of flowers were recorded. Insects were identified visually in the field to the finest taxonomy level, whenever possible. For pollinators that could not be identified in the field, 1-3 individuals of each morphospecies were collected using a butterfly net or a plastic bag. All insect species were identified morphologically, following existing keys (Dubitzky et al. 2008, Hsu et al. 2018, Johnson and Triplehorn 2005, Starr 1992).

Behaviour and insect-flower interaction classification
Types of behaviour of insects visiting flowers of Melastoma species were recorded by direct observations in the field. Further confirmations were made, based on pictures and videos taken using phone cameras. Seven behaviour categories were defined, depending on how insects interact with the flower and the location on the flower where the behaviour occurred, namely sonication, visiting, stamen herbivory, petal herbivory, recycling, drinking and passing (Table 1, Fig. 2). We did not include pollen theft, another important insect behaviour related to interactions with flowers reported in other studies (e.g. Hargreaves et al. 2009). Despite bees often being observed placing their mouth parts at the porous dehiscence of the anther during our observation, there was no evidence that they removed pollen grains from the anthers. In many cases, bees stepped on anthers before they inserted their tongues and then sonicate the anthers afterwards. In other cases, especially near the end of the flowering season or at the last two hours before flowers closed, bees often left the flowers without sonicating the anthers after they performed such behaviour. Therefore, instead of pollen theft, we assume that bees assess pollen capacity of the anthers using both mouth parts and legs before they decide to buzz flowers. Under this context, both types of behaviour were included into the category of visiting. The observations of insect behaviour were further assigned into four types of insect-plant interactions, namely pollination, herbivory, commensalism and neutralism, based upon expected direct effects of each behaviour category for both insects and flowers (Table 1).

Type of behaviour
Sign of expected effect Definition Pollination Sonication +/+ Emit buzz sounds when contacting stamens or pistil, producing vibrations that attempt to expel pollen out from anthers Table 1.
The interactions, definitions and expected effects of the seven types of behaviour of insects that were observed to visit the flowers of Melastoma species. "+", "-" and "0" signs denote positive, negative and neutral effects, respectively, of each type of behaviour on the insect (before the left slash) and the plant (after the left slash).

Insect-flower interaction
Type of behaviour    . 2020) is a summary of the flower-visiting insect occurrence records, based on the observations of this project. The information of flower visiting and flower are addressed in "occurrenceRemarks" and "associatedTaxa", respectively. The dataset is in Darwin Core and published on GBIF.

Column label Column description
occurrenceID An identifier for the Occurrence (as opposed to a particular digital record of the occurrence). In the absence of a persistent global unique identifier, construct one from a combination of identifiers in the record that will most closely make the occurrenceID globally unique. basisOfRecord The specific nature of the data record. kingdom The full scientific name of the kingdom in which the taxon is classified. phylum The full scientific name of the phylum or division in which the taxon is classified. class The full scientific name of the class in which the taxon is classified. order The full scientific name of the order in which the taxon is classified. family The full scientific name of the family in which the taxon is classified.

genus
The full scientific name of the genus in which the taxon is classified.
specificEpithet The name of the first or species epithet of the scientificName. infraspecificEpithet The name of the lowest or terminal infraspecific epithet of the scientificName, excluding any rank designation. taxonRank The taxonomic rank of the most specific name in the scientificName.
identificationRemarks Comments or notes about the Identification.

Results
A total of 1,298 insect visits were observed, which generated 911 occurrence records of flower-visiting insects, of which more than one-third of the visits were made to the horticultural hybrid species, Melastoma kudoi x Melastoma scaberrima (n = 437). Of the remaining observations, 12-19% were recorded for each of the remaining species/forms and only 3.8% of the observations were recorded for M. scaberrima. Around 15.6% and 56.3% of the insects sampled could be identified to species and genus, respectively and the rest are identified to family or higher levels ( Table 2). The number of insect taxa recorded from each Melastoma species ranged from 9 to 39 morphospecies, for a total across all Melastoma species of at least 63 insect morphospecies of seven orders (  Visiting, sonication and passing were the three most commonly-recorded types of behaviour, comprising 37.3%, 32.8% and 20.0%, respectively, of the total observations of behaviour (n = 1,240). The other four behaviour categories only accounted for less than 10.0% of the total observations. With 870 observations, pollination was the most dominant insect-flower interaction recorded on Melastoma species, followed by neutralism (n = 248), herbivory (n = 78) and commensalism (n = 44).
Pollinating insects that demonstrated sonication behaviour were exclusively bees in families Apidae and Halictidae (Hymenoptera: Superfamily Apoidea). Amongst all sonicating bees, sweat bees of genera Lasioglossum and Maculonomia were the two most common taxa, accounting for 89% of all flower visits (Fig. 3). There was a higher diversity of pollinator taxa showing visiting behaviour than other types of behaviour on Melastoma flowers, including insects of 22 families of all seven orders. Lasioglossum and Maculonomia bees, adult insects of Coleoptera (mainly families Chrysomelidae and Elateroidea) and Formicidae (Hymenoptera) were the four most frequently encountered taxa in our samples (Table 1Fig. 4).

Discussion
This study provides the first checklist of flower-visiting insects to all Melastoma species in Taiwan with an emphasis on insect-plant interactions, based on our field observations. Our data show a diverse flower-visiting insect fauna of at least 63 morphospecies which is higher than observations in similar studies on Melastoma (Gross 1993, Liu et al. 2008, Peng et al. 2014, Peng et al. 2012. The majority of the insects exhibited sonicating and visiting behaviour, which presumably can be linked to pollination interaction. Buzzpollinating bees of the families Apidae and Halictidae and particularly members of the genera Lasioglossum and Maculonomia, were the most common pollinators of Melastoma plants in our study site. These findings support the previous conclusion that this genus is primarily buzz-pollinated and highly dependent on bees for pollination.
Despite the commonality in the dependence of buzz-pollinating bees, our results reveal a different bee pollinator composition to other studies on Melastoma plants, even for the same plant species. Liu et al. (2008) studied pollination biology of Melastoma candidum and other three confamiliar species in Melastomataceae in central Taiwan and found that bees of genera Bambus and Xylocopa (both Apidae) are the primary pollinators. Studies on M. malabathricum (affine) in Australia (Gross 1993) reported Xylocopa, Amegilla (family Apidae) and Nomia (Maculonomia) as the main pollinators. Studies on several Melastoma species in southern China suggested that Bambus and Xylocopa, as well as Amegilla bees, are the most important pollinators (Liu et al. 2008, Luo et al. 2008, Peng et al. 2012, Peng et al. 2014. Except Maculonomia bees, these common bee pollinators of Melastoma , particularly the genus Amegilla, represent the minority in our observations. The discrepancy between studies could be explained by the variations in local bee fauna. Landscape features (Ferreira et al. 2013, Sritongchuay andBumrungsri 2016), elevation effect (Hoiss et al. 2015) and biogeography (Traveset et al. 2016) could greatly shape bee assemblages and associated pollination networks via trait-filtering resource partitioning and phenological mismatches between pollinators and plants. This might not be the case in this study, because Amegilla and Bambus bees are both considered common and abundant at the Fushan Botanical Garden (WCY and SSL, unpublished data). For example, Amegilla were abundant at the nursery, but rarely visited Melastoma flowers throughout the study period (JCCH, YCH, WCY and SSL, unpublished data). A possible cause of the shifted pollination niches is that local bees might not recognise the experimental Melastoma plants as an available food resource (Williams 2002) since three of the four plant species are not native to Fushan. Nevertheless, lack of experience cannot completely explain why these bees did not visit the native M. malabathricum often. Other studies show that inter-specific competition of pollinators and pollens mediated by floral neighbourhoods (Bruckman and Campbell 2014) and the presence of a super pollinator (Gross andMackay 1998, Thomson 2004), respectively, could also significantly change the pollinator-plant partnership. Further studies are necessary to clarify the causes of the shifted pollination network in Fushan.
The occurrence of herbivores and their damage to flower structures could supress the pollination process in several ways. First, complete loss of stamens and pollens inside certainly terminate the further chances of pollen transferring. Konzmann et al. (2020) demonstrated that physical modification of anthers could greatly affect the efficiency of pollen-spreading to bees in a neotropical Melastomataceae species. In this case, bees may fail to load pollen to their body if they sonicate damaged stamens, regardless of the amount of pollen remaining inside. Loss and modification of stamen(s) and petal(s) could also reduce the chance of flowers being visited by bees as these floral traits are often found as a resource guide in Melastomataceae plants (Larson andBarrett 1999, Luo et al. 2008). Moreover, in our observations, many herbivorous insects, particularly those with large body size, hindered other flower visitors by active-guarding behaviour or simply covering the reproductive organs with their body (as shown in Fig. 2c). Such trait-mediated processes, mediated by flower herbivores, could also diminish the pollination process at an early stage (Gonçalves-Souza et al. 2008). While reproduction biology is recognised as an essential part of plant conservation, identifying key pollinators and pollination mechanisms becomes fundamental (Havens et al. 2006, Moza andBhatnagar 2007). Without such information, cultivation of closelyrelated species with high hybridisation potential, as observed in Melastoma (Dai et al. 2012, Liu et al. 2014 in ex situ collection sites, may increase chances of genetic introgression (Lozada-Gobilard et al. 2020). The hybridisation risk in ex situ collections might be more severe for sanctuaries in the tropics as most countries in the tropical regions usually have mega-diverse flora, but often grow high numbers of species in a confined area due to lack of sufficient infrastructure. Noteworthy, Target 8 of the Global Strategy for Plant Conservation aims to preserve at least 75% of threatened species of global flora by 2020. Following the Target, many national and regional botanical gardens, for example, Taiwan Forestry Research Institute, have been expanding their ex situ collections since 2012 (Botanic Gardens Conservation International 2012). Further studies on how environmental and ecological factors may drive pollination networks are helpful in preventing ex situ plant conservation from accidental hybridisation events.