Mutual complementarity among diverse pollinators as a mechanism underlying open insect pollination in Japanese pear orchards

Cultivated Japanese pear varieties, mostly showing self‐incompatibility, require pollen transfer from other varieties. To elucidate the underlying mechanisms of open Japanese pear insect pollination, we first collected flower‐visiting insects using plastic vials and sticky traps in orchards located at different regions in Japan. Results showed that insects assigned to the families Andrenidae (Hymenoptera), Apidae (Hymenoptera), Halictidae (Hymenoptera), Syrphidae (Diptera) and Empididae (Diptera) are abundant in the orchards. Second, we restricted the flower‐visiting insects to access Japanese pear flowers using bags with mesh sizes of 0, 2 and 3.5 mm. Results indicated that insects which allowed to pass through bags with 3.5‐mm mesh size but not through bags with 2‐mm mesh size contribute primarily to pollination, represented as the fruit‐set ratio and seed number. Third, we measured head and thorax widths of the flower‐visiting insects and counted pollen grains on their body surfaces to estimate their pollination potential. Results indicated that insects assigned to the families Andrenidae, Halictidae, Syrphidae, Bibionidae (Diptera) and Muscidae (Diptera), including species with both widths smaller than 3.5 mm, harbour large quantities of Pyrus pollen grains, in addition to Apis mellifera (Apidae) with both widths greater than 3.5 mm. Consequently, the families Andrenidae, Apidae, Halictidae and Syrphidae might be the most important insect families for Japanese pear pollination. However, species identification of the flower‐visiting insects showed no common key species that contribute remarkably to pollination services other than A. mellifera. Consecutive insect collection using plastic vials and sticky traps demonstrated that compositions of the flower‐visiting insects are fluctuating continuously in the orchards. Nevertheless, year‐to‐year fluctuation of the fruit‐set ratio was less pronounced in open‐pollinated orchards than in hand‐pollinated orchards. These results suggest that mutual complementation among diverse pollinator species might be the mechanism underlying open Japanese pear insect pollination.


| INTRODUC TI ON
The importance of ecosystem services provided by pollinator insects for worldwide agricultural production, including that of fruit, has been well documented (Klein et al., 2003). For fruit production in Japan, the pollination services are estimated to be worth more than 200 billion JPY (ca. 1.8 billion USD) annually (Konuma & Okubo, 2015). This amount corresponds to ca. 30% of the total fruit output per annum. Japanese pear, Pyrus pyrifolia (Burm. f.) Nakai, is an important commercial fruit in Japan. Japanese pears originating from China have been cultivated for more than 1300 years (Saito, 2018). Commercial pear orchards in Japan were established in the late Edo era (1603-1887). Since then, over 1,000 cultivars have been developed (Saito, 2018). Currently, Japanese pears are produced throughout Japan, mostly for raw consumption. They are not cultivated in the Nansei islands (southwestern islands off Kyushu and in the Okinawa archipelago; Saito, 2018). In 2020, their cultivated area and yield in Japan were, respectively, 10,700 ha and 170,500 tons (https://www.maff.go.jp/).
Most cultivated Japanese pear varieties show self-incompatibility (Kikuchi, 1929;Kobayashi, 1971). Therefore, mixed planting with pollenizers has been practiced for pollen transfer among varieties.
In the 1950s, hand pollination became common because of unidentified pollinator decline (The Japanese Society for Horticultural Science, 1973). However, hand pollination is remarkably costly and labour intensive. The short flowering period of Japanese pear and the changeable weather conditions, such as wind and rainfall, in early spring also might adversely affect hand pollination more than open pollination. Additionally, labour shortages exacerbated by Japan's ageing economy has made hand pollination more unsustainable. For those reasons, open pollination by insect pollinators has attracted attention again for Japanese pear production.
The importance of conditions of pollinator diversity, rather than having one or two key species, for pollination services has also been described (Blitzer et al., 2016). Reportedly, a diverse range of insects with functional diversity improved pollination services in apple orchards (Blitzer et al., 2016;Martins et al., 2015). Surveys of pollinator insects were conducted at some Japanese pear orchards during 1971during -1973during (Kumakura et al., 1973Okabe, 1973). However, little information is available about the presence or absence of key pollinator species and pollinator diversity, both of which can affect pollination services and subsequent fruit production of Japanese pears.
For this study, to elucidate the underlying mechanisms of open Japanese pear insect pollination, we first surveyed flower-visiting insects at Japanese pear orchards located in different regions in Japan.
Second, we evaluated the flower-visiting insects for their contribution to pollination, represented as the fruit-set ratio and number of seeds, using bags with different mesh sizes. Third, we measured the head and thorax widths of the flower-visiting insects and counted the pollen grains on their body surfaces. Based on the results and data obtained from consecutive surveys of insects and fruit-set ratios, we discuss mechanisms underlying open insect pollination in Japanese pear orchards.

| Study sites
This study was conducted at five study sites in Japanese pear orchards of Tochigi (eastern Japan), Tottori (western Japan), central Kumamoto and southern Kumamoto (southern Japan). The survey in Tochigi was administered at the farm of Utsunomiya University (N36°48′E139°98′). The study site (3,600 m 2 ) at the farm had a total of 71 trees of 10 varieties in 2018-2019. In 2020, 55 trees of nine varieties remained after old trees had been cut down. The study site, adjoining crop fields, semi-natural grasslands and orchards mainly cultivating grape, apple and chestnuts was surrounded by secondary Japanese cypress (Chamaecyparis obtusa) forest and rice paddy fields.
The surveys in Tottori were administered at two experimental orchards (Site 02 and Site 05; N35°28′E133°44′). Site 02 (5,000 m 2 ) and Site 05 (1,800 m 2 ) in the Horticultural Research Center, respectively, had 136 trees of seven varieties and 20 rows of seven varieties using the joint-tree training system (Shibata et al., 2011). The Horticultural Research Center, cultivating mainly Japanese pear and additionally cultivating oriental persimmon, citrus and apple, was surrounded by secondary forest consisting of cedar (Cryptomeria japonica) and Japanese cypress and crop fields. Both study sites, each of which was enclosed by a windbreak consisting of Distylium racemosum, sweet

| Collection of flower-visiting insects in 2018 and 2019
Flower-visiting insects were collected directly on the flowers using The collected insects were stored at −20°C until morphological identification to the family level.

| Collection of insects using sticky traps
Results of the survey described above indicated that collection using plastic vials is unsuitable for trapping of some insect families, such as the family Anthomyiidae (data not shown). Consequently, we also surveyed insects inhabiting the study site in Tochigi using sticky traps. The trap comprised a white-coloured plastic board (140 × 340 mm) on which the top side is attached with a transparent sticky plastic sheet (No. 9850, Daikyo Giken-Kogyo Co., Ltd.).
Using a trellis training system that is widely used for Japanese pear production in Japan (NARO, 2006), three traps were set horizontally to the ground from April 12 through April 26 in 2019 and from March 26 through April 9 in 2020. These traps were replaced with new traps on April 19 and April 2 in 2019 and 2020 respectively.
Collected insects were identified directly on the traps based on their morphological characteristics.

| Species identification of insects collected in 2020
In 2020, flower-visiting insects were collected on April 2, April 15, April Morphological characterization of the collected insects was conducted using microscopes (SZX10, SZX7 and SZ43; Olympus Corp.).

| Pollen grain counting and head and thorax width measurements of insects collected in 2020
Estimations  Each tree contains 6-8 replicates to set 24-26 plots for the respective varieties.

| Statistical analysis
The fruit-set ratio and the number of seeds were compared among pollination treatments using a generalized linear mixed model (GLMM), as described by Nikkeshi et al. (2019). Briefly, the fruit set and the numbers of seeds, pollination treatment and tree ID were set, respectively, as response variables, an explanatory variable and a random effect. Tukey's post hoc test was applied to pairs of pollination treatments for a significant difference detection. Statistical analyses described above were performed using software (R ver. 3.5.1; R development Core Team, 2018). The glht function in the multicomp package (Hothorn et al., 2008) and Wald's test with the GLMM function in the glmmADMB package (Skaug et al., 2016) were adopted, respectively, for the Tukey's test and significance assessment of the coefficients.

| Collection of insects using sticky traps
Insects captured using sticky traps are presented in

| Contributions of flower-visiting insects to pollination
Access restriction effects of flower-visiting insects on fruit-set ratios are presented in Figures 1 and 2. The fruit-set ratio in the 3.5M treatment was significantly higher than that of either the 0M or 2M treatment in Tottori and central Kumamoto (p < .05). Significant differences in fruit-set ratios were observed between the 3.5M and 0M treatments in both 2019 and 2020 in Tochigi. The fruit-set ratio in the OP treatment was higher than that in the 3.5M treatment in all surveys, although the difference was not significant (p > .05). The increase in the fruit-set ratio in HPO treatment relative to OP treatment was not consistent between two surveys in Tottori and central Kumamoto. The fruit-set ratio in the 2M treatment was higher than that in the 0M treatment in Tochigi in 2020 (p < .05).
Access restriction effects of flower-visiting insects on seed numbers are presented in Figure 2. Seeds were fewer in the 0M and 2M treatments than in either the 3.5M, OP or HPO treatment in Tottori (p < .05). In central Kumamoto, seed numbers in the 2M treatment were smaller than those in the 3.5M, OP and HPO treatments, although the difference was not significant (p > .05). No significant differences in seed number were found among the 2M, 3.5M and OP treatments in Tochigi. No seed was obtained in the 0M treatment in central Kumamoto and Tochigi (data not shown).
The number of pollen grains on the body surfaces of flowervisiting insects is presented in Table 3. Head and thorax widths of the flower-visiting insects are presented in Table S1. Of the eight Andrenid species, five showed both widths smaller than 3.5 mm, as did 4 of 11 syrphid species. All spe-

| Fruit-set ratio in the OP treatment at openpollinated and hand-pollinated orchards
Fruit-set ratio in the OP treatment is shown in Table 4, including the data presented in Figure 1. Results showed that fluctuation of year- to-year fruit-set ratios was less pronounced in open-pollinated orchards (Tochigi and southern Kumamoto) than in orchards in Tottori and central Kumamoto.
Reportedly, insects assigned to the families Andrenidae, Apidae and Halictidae, and Syrphidae were abundant at Japanese pear orchards in Ishikawa (N36°E136°) and Fukushima (N37°E140°), Japan (Kumakura et al., 1973;Okabe, 1973 and L. proximatum (Halictidae), were commonly observed in the three regions (Table 3). However, they did not constitute large proportions in any region. These results suggest that no key species exists for pollination services in Japanese pear orchards.
Our access restriction experiment in Japanese pear orchards showed a primary contribution to pollination services of insects that had passed through bags with 3.5 mm mesh. The families Andrenidae, Halictidae, Syrphidae, Anthomyiidae and Empididae included species, possibly accessible to the flowers in the 3.5M treatment, with head and thorax widths smaller than 3.5 mm (Table S1). In addition, A. mellifera (Apidae) and some species assigned to the families Andrenidae and Syrphidae with both widths larger than 3.5 mm (Table S1) might also contribute to pollination services because fruitset ratios in the OP treatment were higher than that in the 3.5M treatment (Figure 1), but not significantly (p > .05). The six families included species harbouring large quantities of Pyrus pollen grains on their body surfaces (Table 3). Consequently, we concluded that the six families are the main components (hereinafter designated as small insects) responsible for pollination services.
Seed set has been recognized as an important indicator of fruit development: it influences fruit size, shape and eventually marketability (Brookfield et al., 1996;Buccheri & Di Vaio, 2005;Fountain et al., 2019;Garratt et al., 2014Garratt et al., , 2016Matsumoto et al., 2012;Monzón et al., 2004;Volz et al., 1996). Increased fruit-set ratios and seed set in the 2M treatment relative to the 0M treatment were observed in Tochigi (Figures 1 and 2). This finding suggests the presence of minute insects with head and thorax widths smaller than 2 mm, with involvement in pollination services at the study site.
Some species assigned to the families Halictidae, Anthomyiidae and Empididae might have contributed to the increase. Reportedly, some gall midge species (Cecidomyiidae) are involved in pollination of Artocarpus plants (Moraceae) (Gardner et al., 2018;Sakai et al., 2000). In Tochigi, insects assigned to Cecidomyiidae were collected using sticky traps in large numbers in 2019 and 2020 ( Table 2). In addition, Zygothrica flies (Drosophilidae) are known to be involved in pollination of Dracula orchids in Andean cloud forests (Endara et al., 2010). The functional significance of these very small insects in pollination demands future investigation.
One index representing the pollination potential of flowervisiting insects is the number of pollen grains on the insect body surfaces (Devoto et al., 2011;Forup et al., 2008;Lopezaraiza-Mikel et al., 2007). Our surveys showed that insects assigned to the families Andrenidae, Apidae, Halictidae and Syrphidae harboured large quantities of Pyrus pollen (Table 3). Reportedly, stigma contact and visitation rates (flowers/ min) of pollinators in combination with the number of pollen grains affected fruit-set and seed-set of pear (Alarcón, 2010;Monzón et al., 2004;Popic et al., 2013). To evaluate the pollination potential of the small insects observed at the study sites in the three regions, their stigma contact rate and pollinating effectiveness must be examined in future studies. Additionally, the pollen grains deposited on a stigma must be quantified, as suggested by King et al. (2013).  work was supported financially by the Ministry of Agriculture, Forestry and Fisheries, Japan, through a research project entitled 'Monitoring and enhancement of pollinators for crop production.'

CO N FLI C T O F I NTE R E S T S
The authors declare that they have no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available in (Sonoda, 2021a(Sonoda, , 2021b