Pollinator‐mediated interactions between cultivated papaya and co‐flowering plant species

Abstract Many modern crop varieties rely on animal pollination to set fruit and seeds. Intensive crop plantations usually do not provide suitable habitats for pollinators so crop yield may depend on the surrounding vegetation to maintain pollination services. However, little is known about the effect of pollinator‐mediated interactions among co‐flowering plants on crop yield or the underlying mechanisms. Plant reproductive success is complex, involving several pre‐ and post‐pollination events; however, the current literature has mainly focused on pre‐pollination events in natural plant communities. We assessed pollinator sharing and the contribution to pollinator diet in a community of wild and cultivated plants that co‐flower with a focal papaya plantation. In addition, we assessed heterospecific pollen transfer to the stigmatic loads of papaya and its effect on fruit and seed production. We found that papaya shared at least one pollinator species with the majority of the co‐flowering plants. Despite this, heterospecific pollen transfer in cultivated papaya was low in open‐pollinated flowers. Hand‐pollination experiments suggest that heterospecific pollen transfer has no negative effect on fruit production or weight, but does reduce seed production. These results suggest that co‐flowering plants offer valuable floral resources to pollinators that are shared with cultivated papaya with little or no cost in terms of heterospecific pollen transfer. Although HP reduced seed production, a reduced number of seeds per se are not negative, given that from an agronomic perspective the number of seeds does not affect the monetary value of the papaya fruit.

because an increase in arable land usually entails the loss of forest cover and with this, the elimination of valuable floral resources and suitable habitat for pollinators Garibaldi et al., 2014;Winfree, Aguilar, Vázquez, LeBuhn, & Aizen, 2009). Because the expansion of agriculture occurs at the expense of native vegetation, the interface between crops and wild plant species has dramatically increased (Goldewijk, 2001;Klein, Cunningham, Bos, & Stefan-Dewenter, 2008). Therefore, it is crucial to understand the impact of pollinator-mediated interactions between cultivated and wild plants on their reproductive success (Klein et al., 2008;Stanley & Stout, 2014).
Pollination is a complex phenomenon, and its final effect on plant reproductive success is mediated by several pre-(e.g. pollinator visitation rate) and post-pollination (e.g. pollen load quality, pollenstigma interactions) events (Willcox, Aizen, Cunningham, Mayfield, & Rader, 2017). For instance, frequent pollinator visits may produce poor fruit/seed set if the pollen load is dominated by heterospecific pollen (hereafter HP) (Wilcock & Neiland, 2002). Previous studies looking at pollinator-mediated interactions between wild and cultivated plant species have emphasized pre-pollination events with little emphasis on post-pollination events (e.g. Chacoff & Aizen, 2006, Winfree et al., 2008, Bailey et al., 2014. Such studies have also mainly focused on forest proximity (reviewed by Garibaldi et al., 2011, Ricketts et al., 2008, but there has been little attention to other mechanisms that could mediate the effect of co-flowering on crop yield. Studies conducted in natural communities suggest that plant species in co-flowering communities make a variable contribution to the diet of shared pollinators (Bergamo et al., 2017;Carvalheiro et al., 2014). As a result, HP is frequently transferred and deposited on the stigmas of the interacting plants Morales & Traveset, 2008;Tur, Saez, Traveset, & Aizen, 2016). Contrary to the general belief (i.e. that wild plants facilitate crop pollination), the outcome of pollinator-mediated interactions on the reproductive success of co-flowering species in natural communities ranges from negative (competition) to positive (facilitation) (Arceo-Gómez et al., 2016;Muchhala & Thomson, 2012;Tur et al., 2016). If crops co-flower and share pollinators with wild plant species, it is not inconceivable that there may be positive, negative, or even neutral effects, on at least one step of the pollination process. In fact, pollinator sharing and HP transfer may be higher in crops than in wild species because the former have not coevolved with the native plant-pollinator network Morales & Traveset, 2008).
In this study, we looked at pollinator-mediated interactions between papaya (Carica papaya) cultivated on an experimental plantation and the surrounding co-flowering plant community on the Yucatan Peninsula. Wild papaya populations are dioecious and therefore, highly pollinator-dependent (Fuentes & Santamaría, 2014).
Although modern papaya varieties can also produce self-compatible hermaphrodite flowers, previous studies suggest that pollen deposition by pollinators on these flowers significantly increases fruit set and weight (Badillo-Montaño, Aguirre, Santamaría, Martínez-Natarén, & Munguía-Rosas, 2018;Garrett, 1995;Martins & Johnson, 2009). Fruit set in the absence of pollinators is also less attractive to customers owing to the reduced size and round shape of the fruit (Martins & Johnson, 2009;Moo-Aldana et al., 2017). In the study area, cultivated papaya blooms year round and is visited by a wide variety of generalist insects (Moo-Aldana et al., 2017) that also visit many other wild and cultivated plant species around papaya plantations (Badillo-Montaño et al., 2018). Therefore, pollinator-mediated interactions among cultivated papaya and co-flowering plant species are very likely. Using observational and experimental approaches, we dissected the pollination process to see how co-flowering affects both pre-and post-pollination events. Specifically, we assessed pollinator sharing, pollen transfer, and the effects of HP transfer on fruit and seed production in cultivated papaya. Our specific goals in this study were to: (a) identify pollinators shared between cultivated papaya and the surrounding co-flowering plant species, (b) assess the extent to which co-flowering plants contribute to the diet of shared pollinators, (c) determine the degree of HP transfer to cultivated papaya, and (d) determine experimentally whether or not HP pollen load affects the quantity and quality of the yield in this crop species.
The climate is sub-humid, warm with summer rains; the mean annual temperature is 25.3°C (Campos-Navarrete, Abdala-Roberts, Munguía-Rosas, & Parra-Tabla, 2015). The study area is located in a landscape mosaic that encompasses fragments of original forest, Papaya is originally from Mesoamerica but is currently cultivated in several tropical and subtropical regions worldwide (Fuentes & Santamaría, 2014). It is also one of the most economically valuable tropical fruits in the world, and Mexico is its sec-

| Shared pollinators
A circular area ca. 177 ha with the experimental plantation at its center was delimited to record flower visitors. Floral resources (wild and cultivated plants) were scattered within this area; however, vegetation closer than 3 m or within the plantation was cleared. Thus, cultivated papaya was the only source of floral resources within the plantation, no cultivated papaya was observed outside of the plantation in the sampled area. A portion of two forest patches also fell within the sampled area. The radius of this area (751 m) was slightly smaller than the mean foraging distance (815 ± 10 m) of bees, with body size similar to species found in the study area (Araujo, Costa, Chaud-Netto, & Fowler, 2004;Zurbuchen et al., 2010). From April through June 2016, floral visitors were recorded for three consecutive days, twice a month. Flower visitors were surveyed from 0700 to 1300 hr, the observed peak of activity for flower-visiting insects  Flower stigmas were collected after 48 hr and immediately fixed in ethanol 70% (Arceo-Gómez et al., 2016). Once in the laboratory, the stigmas were rehydrated with water, decolorized with NaOH (5 N) at 37°C for 12 hr, and stained with aniline blue 0.3% for 18 hr (Alonso et al., 2013). Then, the stigmas were placed on microscope slides and observed with a fluorescence microscope (Leica DM1000, Germany) under a 515-560 nm excitation filter at magnifications of 10×, 20×, and 40× (Kearns & Inouye, 1993). Based on pollen morphology, conspecific pollen (hereafter CP) and HP in stigmatic loads were identified and counted. Although the pollen of papaya can be identified easily, HP was not identified to the species level because pollen morphology is not species-specific in some plant groups or differences are not observable with the technique we used.

| Fruit production and seeds
To assess the effect of HP transfer on fruit and seed production, a hand-pollination experiment was conducted. Thirty-three hermaph- To do so, we first harvested the pollen of as many anthers as possible per species, and then we took approximately the same amount of pollen in volume of each species and mixed it to obtain a homogenous mixture. Therefore, the proportion of pollen per species in the mixture used for hand pollinations was approximately 1:1:1. A second subgroup of 33 flowers (treatment 2) was hand-pollinated with a mix of CP from the plantation, pollen from wild papaya, and pollen from M. dissecta (1:1:1), following the procedure described for treatment 1. Pollen from wild papaya was used in treatment 2 because it is known that cultivated varieties share some pollinator species with wild papaya when they co-occur (Moo-Aldana et al., 2017). M. dissecta and M. oleifera were selected because they co-flower with cultivated papaya, share some pollinator species (3) with cultivated papaya, and produce abundant and accessible pollen. The same number of flowers (11) per variety and per treatment was selected. Before pollen was placed on the stigma, flowers were carefully emasculated. In all cases, pollen was placed on the stigma until it was saturated. After hand-pollinating the flowers, all flowers were bagged with a mosquito net. Fruit set was recorded weekly and, once ripe, fruit were weighed and seeds counted.

| Data analyses
Sampling completeness of flower-visitor interactions was assessed by comparing the number of observed and expected interactions based on the Chao 2 estimator and pooling the data of all plant species (Chacoff et al., 2012). To visually analyze the structure of the co-flowering plant-pollinator network, we built a quantitative plant-pollinator network using the bipartite package for R (R Core Team, 2017). Then, we ran a hierarchical cluster analysis based on among-plant dissimilarity (Bray-Curtis) in terms of the pollinator species using the average agglomerative method (Everitt & Hothorn, 2011;Goslee & Urban, 2007). Cluster uncertainty was assessed with bootstrap resampling methods (1,000 replicates) implemented in the pvclust package of R 3.3.3. (Suzuki & Shimodaira, 2006). While the cluster analysis allowed us to identify similarity between plant species in terms of pollinator identity and frequency, for a given pair of plant species, the influence (i.e. pollen transfer) of one species (acting plant) on the other (target plant) may be asymmetrical (Bergamo et al., 2017). To address potentially asymmetrical pollinator-mediated interactions among pairs of co-flowering plant species, we calculated Müller's index (Müller, Adriaanse, Belshaw, & Godfray, 1999)  respectively. The variety of cultivated papaya was included as a random factor to account for any among-variety variation in all models.

| Shared pollinators
During the study, 18 plant species co-flowered with cultivated papaya, and all of them shared at least one pollinator species with cultivated papaya (Figure 1), except for two Citrus species (lemon and orange) that were not visited. Thirty-four pollinator species and 5, 201 flower visits were recorded. The most frequent pollinators were Apis mellifera (39.4%), and the native bee species Trigona fulviventris (21.5%) and Nannotrigona perilampoides (5.9%). The remaining 33.2% of visits were by social, eusocial, and solitary bee species (21 species), some species of Lepidoptera (5), Diptera (2), and Coleoptera (1)

| Fruit production and seeds
Flowers that were hand-pollinated with a mix of CP and HP (treatments 1 and 2) and CP alone (control) did not statistically differ in fruit set ( 2 2 = 0.99, p = 0.95) or fruit weight (F 2,50 = 1.12, p = 0.33). However, a difference among treatments was found in seed number F I G U R E 2 Hierarchical cluster of dissimilarity for pollinator assemblages in a community of co-flowering plant species in a landscape mosaic on the Yucatan Peninsula. Values at the nodes are the times (in percentage) that a focal cluster appeared in 1,000 bootstrap iterations F I G U R E 3 Müller's index for a community of co-flowering plant species in a landscape mosaic on the Yucatan Peninsula. Black bars represent the index when cultivated papaya is the target species and white bars when cultivated papaya is the acting species per fruit ( 2 2 = 42.01, p < 0.001). Flowers pollinated with a mix of HP and CP produced significantly fewer seeds than flowers pollinated only with CP (Table 1).

| D ISCUSS I ON
In this study, we have shown that cultivated papaya shares at least one pollinator species with the majority of co-flowering plants in the vicinity of an intensive experimental plantation. Despite this extensive pollinator sharing, observed HP transfer in cultivated papaya tends to be low. HP loads had no negative effect on fruit production or weight; even though experimental HP loads were greater than the loads on open-pollinated flowers. We suggest that the effect of co-flowering plants on crop yield of papaya is positive because these plants contribute to the diet of shared pollinators with little cost in terms of HP transfer.
Cultivated papaya shared at least one pollinator species with 88% of the co-flowering plant species in the study area. Therefore, we think that cultivated papaya has been successfully integrated into the plant-pollinator network of co-flowering plants, and this may have occurred because this crop has a generalist pollination system (Moo-Aldana et al., 2017). For instance, cultivated papaya was visited by A. mellifiera and T. fulviventris, which were the most common pollinator species in the whole network, and thus, the visit of these two bee species alone would result in its successful incorporation into the network. Although nonnative plants have not coevolved with native flora and pollinators, previous studies suggest that a large floral display as well as a generalist pollination system may facilitate their integration into the plant-pollination network (Jakobsson, Padron, & Traveset, 2008;Memmot & Waser, 2002). In the specific case of our study system, we also think that the presence of wild papaya in the study area may have played a role in the incorporation of cultivated papaya because the local pollinator fauna were already familiar with floral resources offered by cultivated papaya owing to the similarity in the floral traits that attract pollinators (Moo-Aldana et al., 2017).
The results of the hierarchical clustering and Müller's index coincided in that cultivated papaya has a relatively strong pollinatormediated interaction with wild papaya, C. nucifera and L. anagyroides (Figures 2 and 3). This was also probably due to floral similarity between wild and cultivated papaya. For C. nucifera and L. anagyroides, this probably occurred because these plant species have massive floral displays (Meléndez-Ramírez et al., 2004;Stawiarz & Wróblewska, 2013) that attract a widely diverse and abundant assemblage of generalist insect visitors, many shared with cultivated papaya. Müller's index also indicates a slightly stronger influence of papaya as an acting than as a target species, a finding that needs further attention. If co-flowering plants actually receive pollen from cultivated papaya, this may lead to CP loss (Morales & Traveset, 2008) as well as the transfer of papaya pollen to wild plant species with potentially negative effects on their reproductive success (Stanley & Stout, 2014).  (Ekroos et al., 2015;de Waal, Anderson, & Ellis, 2015).
That is, to reduce foraging cost, pollinators may move preferentially within the plantation. The transfer of CP not assisted by pollinators (i.e. autonomous self-pollination) is likely to be negligible in the plantation because CP load on stigmas was similar between female and hermaphrodite flowers.
Positive covariation between CP and HP on the stigmas of plants, as we observed in cultivated papaya, has been interpreted as evidence of facilitative interactions between co-flowering plants (Tur et al., 2016). The rationale behind this is that when a community of co-flowering plants contributes to the diet of shared pollinators, it increases pollen transfer in general (Tur et al., 2016).
However, we cannot consider this to be evidence of facilitation if HP reduces plant reproductive success; an aspect rarely evaluated in co-flowering plant communities (e.g. Carvalheiro et al., 2014, Tur et al., 2016, Bergamo et al., 2017. Using a proportion of HP of about 66%, we did not detect any effect of HP on fruit set or weight in cultivated papaya. In our sample, 87% of stigmas had a proportion of HP ≤66%. Therefore, the degree of HP transfer typically seen in open-pollinated flowers has no negative effect on fruit production or quality. In one sense, the results of our experiment can be seen as an exacerbated effect of HP loads. Although seed production does not affect the economic value of papaya, the observed negative effect of high proportions of HP on seed production indicates that HP affects a post-pollination process related to ovule fertilization and/or seed development (Aizen & Harder, 2007;Wilcock & Neiland, 2002). This may be relevant for other crops where seed yield is of primary interest to farmers, such as sunflower (Nderitu, Nyamasyo, Kasina, & Oroje, 2008) and almond (Dag, Zipori, & Pleser, 2006).
Although we have assessed the consequences of co-flowering on the different pre-and post-pollination events in cultivated papaya, we recognize that more effort is needed to bridge these pollination events. The identification of HP to the species level for all the coflowering species may help to link pollinator sharing and HP transfer; however, this may require a comprehensive reference collection of pollen and/or the use of more sophisticated techniques of pollen identification (e.g. scanning electron microscopy, and spectroscopy).
Experiments emulating the actual quantity and quality (i.e. species of origin) of HP on stigmas would be the optimal approach for demonstrating a link between HP transfer and reproductive success;

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

DATA ACCE SS I B I LIT Y
Data used in this study is available in Dryad https://doi.org/10.5061/ dryad.4rs855v. TA B L E 1 Mean values (±1 SE) for fruit set, fruit weight and number of seeds per fruit of cultivated papaya under three hand-pollination treatments (Conspecific, Heterospecific-1 and Heterospecific-2). Treatment Heterospecific-1 consisted of a mix of conspecific pollen and pollen from Merremia dissecta and Moringa oleifera (1:1:1). Treatment Heterospecific-2 consisted of a mix of conspecific pollen from the plantation, pollen from wild papaya and pollen from M. dissecta (1:1:1). Different superscript letters indicate statistically significant differences between treatments