Introduction

One factor in the establishment of potentially invasive angiosperm weeds in new regions is the availability of pollen vectors. For potential weeds with specialized pollinators, the co-introduction of suitable pollinators is hypothesized to lead to ‘invasion meltdown’ where the introduced plant and introduced pollinator mutually enhance each other’s spread (Simberloff and Von Holle 1999; Simberloff 2006). As an alternative to the notion of invasion meltdown as a driver of weed proliferation, Olesen et al. (2002) argued that in island ecosystems with very low pollinator diversity, high intraspecific but low interspecific competition would encourage pollinators to evolve into super-generalists, able to exploit a wide variety of plants that might otherwise depend on specialist pollinators.

Homalictus fijiensis (Perkins & Cheesman) (Halictidae) is an endemic bee of the Fijian islands, and previous studies of the pollination systems of Fiji (Crichton et al. 2018; Groutsch et al. 2019; Hayes et al. 2019) showed it has a broad floral host range. The range included an introduced weed Solanum torvum that requires buzz pollination (Staines et al. 2017), and the utilisation of extra-floral nectaries in a native Hibiscus species, suggesting it has evolved to exploit a wide range of food resources (Crichton et al. 2018). Although studies have not explored the efficacy of pollination by H. fijiensis, it has been referred to as a super-generalist pollinator due to visitation of all of the 49 angiosperm species documented in previous Fiji pollination network studies (Crichton et al. 2018; Groutsch et al. 2019; Hayes et al. 2019).

Many angiosperm species with poricidal anthers have distinctive flowers with specialised pollen availability, and are referred to as having a ‘solanoid’ morphology (Faegri 1986). Solanoid plants often do not provide nectar rewards (Vogel 1978), and this is thought to be a strategy that discourages non-capable pollinator visitation (Harder and Barclay 1994; De Luca and Vallejo-Marin 2013). Such anthers have evolved many times in plants as a strategy to prevent excessive pollen loss (Buchmann 1983; Harder and Barclay 1994), restrict access to pollen to capable pollinators (Buchmann and Hurley 1978; De Luca and Vallejo-Marin 2013) and direct pollen delivery by capable pollinators to conspecific plants (Harder and Barclay 1994). To be capable pollinators, bees must be able to sonicate, or ‘buzz’, anthers at a high enough frequency and amplitude to allow pollen grains to be shaken from the cylindrical structure (Buchmann and Hurley 1978; Vallejo-Marín 2019). For some bees, sonication is an innate behaviour (Morgan et al. 2016), and both small and large bees are capable of floral sonication, although reaching the appropriate vibration amplitudes is harder to achieve for smaller bees (Vallejo-Marín 2019; De Luca et al. 2019). However, some bees (including buzzing and non-buzzing bees) can access and potentially transport pollen for solanoid species without sonication by biting through the anthers to access pollen (see Anderson and Symon 1988; Thorp 2000; Solís-Montero et al. 2015). This behaviour, termed pollen larceny (Inouye 1980), is less efficacious than buzz pollination and negatively impacts plant reproduction, but in some cases successful pollination can occur (Hargreaves et al. 2009). Island ecosystems such as Fiji where native pollinators may be unable to buzz pollinate, yet introduced plants requiring specialized pollination are widespread, raises questions as to how these plants overcame pollinator limitations and whether local pollinators have evolved to exploit these alternative floral resources.

One such case of the spread of an introduced plant with specific pollinator requirements is that of the invasive neotropical weed Solanum torvum, where it is unclear how pollination services are provided to enable its spread across Fiji. Despite the presence of an introduced buzz pollinator, Amegilla pulchra, it has never been observed on S. torvum (Staines et al. 2017; Groutsch et al. 2019; Crichton et al. 2018; Hayes et al. 2019) and therefore cannot be used to explain the species spread across the Fijian archipeligo. Homalictus fijiensis however has been observed extracting pollen from S. torvum through pollen larceny, where bees were observed extracting pollen from poricidal anthers by biting at their tips with no attempts at locating nectaries (Staines et al. 2017). It is not clear how H. fijiensis is able to identify S. torvum as suitable for pollen larceny, or whether pollen larceny evolved after the original introduction of this species. Pollen larceny of buzz pollinated plants has been reported from regions outside of Fiji, such as small native Trigona bees on Solanum flowers in Australia (Anderson and Symon 1988) and Lasioglossum and Augochlorella bees in Mexico (Solis-Montero et al. 2015), as well as Apis mellifera and Trigona bees on Solanum in Brazil (Vinícius-Silva et al. 2017). Pollen larceny may also be performed by Trigona bees in Costa Rica on Thunbergia grandiflora (Young 1983) and on multiple Melastomataceae species in Brazil (Renner 1983), but notably pollination efficacy was not investigated in these studies. Homalictus fijiensis has been observed on S. torvum and another introduced solanoid plant Dissotus rotundifolia (Crichton et al. 2018; Hayes et al. 2019), but has not yet been observed exploiting any native solanoid plants. Without prior exposure of H. fijiensis to a Fijian native plant with poricidal anthers over evolutionary time, it is difficult to explain the origin of the behaviour of mandibulation to obtain pollen from solanoid plants.

Here, we further explore the breadth of plant species utilized by H. fijiensis with additional emphasis on species with solanoid morphologies. We argue that the presence of native Fijian plants with poricidal anthers provided opportunities for generalist bees, in insular regions such as H. fijiensis, to evolve abilities to recognize and exploit solanoid floral morphologies. Our findings unambiguously demonstrate the super-generalist status of H. fijiensis and provide evidence of the evolution of super-generalists in insular systems, with both positive and negative ecological and economic implications.

Methods

Field observations and species

Staines et al. (2017), Crichton et al. (2018), Groutsch et al. (2019) and Hayes et al. (2019) analysed bee-plant networks covering five bee species and 49 plant species (see Network Analyses below) from Viti Levu in Fiji. Here we add additional records and extend the number of plant species to 60 from those in Crichton et al. (2018). Our observations included four additional solanoid plants, leading to a total of seven solanoid species (see Table 1) of which five are nectarless. All plants were identified using Whistler (1995).

Table 1 Plant species observed with bee interactions used for all analyses
Full size table

Observations were conducted between the 2nd and 21st of April, 2018 from Colo-i-Suva Forest Park, the Laucala Bay campus of the University of the South Pacific (USP), and along roadsides in the Serea and Naqali regions between Colo-i-Suva and Serea, Viti Levu. Field sites were restricted to below 800 m asl as recent studies indicate the existence of multiple Homalictus species that can be morphologically difficult to distinguish from H. fijiensis, but those species are largely restricted to elevations above 800 m asl (Dorey et al. 2019).

Flowering plants were observed from 1 to 2 m (m) distance with recording commencing 2 min after first approach to the plants to allow insects to resettle after any initial disturbance by the observers. The observational time period of 10 min was started after the resettling time, and observers would then watch for interactions between all bee species and flowers. These interactions were recorded quantitatively. Homalictus fijiensis individuals that flew from one flower to another were recorded as a single interaction. Other plants within 3 m of the observation area of the same species were also included in the observation area, and observations were conducted by up to four people at a time. The times of observation were calculated as person-minutes of observation, as detailed in Crichton et al. (2018).

Network analyses

Our 2018 data across all species observed comprised a total of 1460 person-minutes of observation (see Table 1 and below). We combined these data with previous studies conducted by Staines et al. (2017) (670 person-minutes of observation), Crichton et al. (2018) (1260 person-minutes of observation), Groutsch et al. (2019) (825 person-minutes of observation) and Hayes et al. (2019) (540 person-minutes of observation). The combined data set totalled 4755 person-minutes of observations of five bee species and included 60 plant species (Table 1). All observations were transformed into binary data (presence/absence of interactions). Interaction networks were visualized using the visweb and plotweb commands in R ver. 3.3 module Bipartite 2.05 (Dormann et al. 2009).

Network parameters for our binary data were explored using the command networklevel in the Vegan module in R ver. 3.3. The parameters of interest were (i) C-scores, which quantify levels of aggregation or disaggregation, typically used to explore competition between bees for plants or plants for bees; (ii) connectance, which measures the observed number of interactions compared to the possible number; and (iii) niche overlap for both plants and bees separately. We also examined nestedness of the interaction network using the function nestednodf which calculates the NODF parameter which estimates nestedness for both trophic levels as well as the combined levels (Almeida-Neto et al. 2008). We estimated the significance levels for all parameters using the oecosimu command in the Vegan module, with 2000 Monte Carlo simulations for null models.

Interactions with Melastoma denticulatum

Particular focus was placed on locating and observing plant species with poricidal anthers, in particular the native M. denticulatum. A further six solanoid plants were identified with poricidal anthers and all of these were introduced species (Table 1). When H. fijiensis was observed interacting with M. denticulatum flowers, video evidence was taken using Samsung S4 phone camera to identify potential pollen larceny behaviours.

Specimens of H. fijiensis were captured from flowers by sweep nets during observational periods for pollen count analyses. Bees were caught once they had interacted with M. denticulatum flowers, and were immediately placed into separate 3 mL pre-filled 98% ethanol vials with as little handling and abrasive contact as possible to prevent loss of pollen or contamination. Samples of M. denticulatum pollen were taken by preserving anthers in 3 mL pre-filled 98% ethanol vials.

A standard slide mount of M. denticulatum pollen was made by crushing anthers with watchmaker forceps and preparing a microscope slide viewed using a Nikon Eclipse 50i compound microscope at ×40 magnification (following methods detailed in Staines et al. 2017). Photographs were taken using a Nikon 5000 SLR camera as references for mature and immature M. denticulatum pollen grains. Pollen samples from vials of H. fijiensis were obtained by vigorously shaking the vials to suspend pollen and then taking 0.5 mL aliquots to place on a slide and covered with a 2.5 cm × 5.0 cm coverslip. Each pollen grain was then recorded as either M. denticulatum or as ‘other pollen’. The slide was observed in sequential zig–zag transects until 100 total pollen grains were observed or until the entire space under the coverslip was observed before a maximum of 100 pollen grains were counted.

Results

Bee-plant network analyses

The binary plant-bee interaction network from our combined 2016–2018 data are shown in Fig. 1. This figure indicates a much wider range of host plants for H. fijiensis than for the other bee species, with Apis mellifera having the next broadest host range, followed by Braunsapis puangensis, and then Amegilla pulchra and Megachile species. Of the seven plant species with poricidal anthers, only one, Senna surattensis, was visited by a bee species (Apis mellifera) in addition to H. fijiensis.

Fig. 1
figure 1

Network diagram of the bee-plant interactions observed in Fiji. Native species are indicated by green boxes, introduced species by black boxes, and plants with poricidal anthers are in red text. The network diagram contains bee-plant interactions as indicated in Table 1. Homalictus fijiensis was seen to be the sole visitor of all but two solanoid plant species

Full size image

Of the plants observed to be receiving visits from only a single bee species, 14 of these were visited by H. fijiensis, one was visited by Amegilla pulchra, and three were visited by B. puangensis (Fig. 1). Four of these ‘single-pollinator’ plants were native to Fiji and only one of these was not visited by H. fijiensis. Of the four plant species receiving visits from all bee species, only one, Vitex trifolia, was a native Fijian species.

For the five bee and 60 plant species in our study we observed 132 species-pair interactions out of a theoretically maximum of 300. Our Monte Carlo simulations show that this value is significantly higher (P < 0.001) than expected from a null model. Our value is also higher than any of the 29 plant-pollinator network studies reviewed by Olesen and Jordano (2002), and which ranged across very diverse continental and island habitats. Our connectance value reflects that three of the four introduced bee species in Fiji are generalists and the endemic H. fijiensis is a super-generalist, as is evident from Fig. 1. The estimated number of compartments in our networklevel analyses was 1.0, suggesting that we do not have well-defined clusters of interacting species, which is also evident in the interaction network in Fig. 1. Niche overlap for plants was 0.59 and for bees was 0.37 for plants; both were statistically different from random interaction models (P < 0.001, for both cases) and indicate moderate niche overlaps.

The C-score for plants was 0.17 (P < 0.001) and for bees was 0.30 (P < 0.001). The relatively lower C-score value for plants suggests they may be competing for bee pollinators less than bees compete for floral resources, though the C-score value for bees is still low, suggesting little competition among the species. However, it is likely that the C-score value for bees partially reflects the situation where all but two plant species with poricidal anthers only received visits from H. fijiensis; this situation may not reflect competition between bees for plant species and instead reflect differential abilities of bees to access solanoid flowers, with H. fijiensis being the only species to commonly exploit such flowers in Fiji. However, exclusive visits by H. fijiensis to some plant species extends beyond just the solanoid plant species (Fig. 1).

Lastly, we examined nestedness in the combined data set, and this is visualized in Fig. 2 which indicates differences in the breadth of host plants for each bee species. Host breadth for each species was 16 plants for Amegilla (1 unique), 25 for Braunsapis (3 unique), three plants for Megachile, 35 plants for Apis (1 unique) and 53 for Homalictus (14 unique). The NODF values for plants, bees and the combined matrix were similar (7.10, 8.61 and 7.10, respectively, each with P < 0.001) substantiating the apparent nestedness in Fig. 2.

Fig. 2
figure 2

Nestedness of the bee-plant network of the listed observed species in Table 1. Observation of an interaction between the bee species and plant is indicated by black squares, green squares and green text indicate interactions between Fijian native Homalictus fijiensis and native plant species, and solanoid plants with poricidal anthers are in red text

Full size image

Bee interactions with Melastoma denticulatum

The observations were conducted at several roadside locations along the Suva-Monasavu road in the Serea district at elevations of ~ 20–30 m asl. A total of 240 person-minutes of observations were conducted in 2018 across nine observational events of M. denticulatum. A total of 13 interactions between H. fjiiensis and M. denticulatum were observed and 19 bee specimens (with 6 specimens caught outside the observation time) were acquired for pollen analysis (see below). Observations of these long interactions between H. fijiensis and M. denticulatum not only showed the bees concentrated their physical contact to the anthers, but also crawling over the petals and stigma while mandibulating the anthers before leaving (see video in electronic supplementary material). The behaviour of H. fijiensis also showed no signs of attempts to locate nectaries, nor behaviours consistent with sonication.

Homalictus fijiensis was the only bee observed to approach or make contact with M. denticulatum flowers, despite a large number of Apis mellifera and B. puangensis present in the immediate areas (< 2 m distance from M. denticulatum plants) where they primarily foraged on Breynia disticha, Pentas lanceolata and Clerodendrum thomsoniae in the vicinity of the observation areas. Amegilla pulchra was never observed on or near M. denticulatum flowers.

The 19 H. fijiensis specimens collected after interacting with M. denticulatum flowers were used to examine the prevalence of M. denticulatum pollen grains on bees. This prevalence ranged up to 81%, and M. denticulatum pollen was present on all bees except for one (Fig. 3), but for eleven specimens fewer than 100 pollen grains could be counted on the pollen slide preparations. This may not be surprising because bees were collected after their first observed interaction with M. denticulatum flowers, and those bees may have only just started foraging shortly before collection. Both mature and immature M. denticulatum pollen grains were found during microscope examination (see Fig. 4), suggesting that some pollen was retrieved before maturation, consistent with pollen larceny (Solís-Montero et al. 2015; Staines et al. 2017; De Luca et al. 2019).

Fig. 3
figure 3

Graph of the percentage of Melastoma denticulatum pollen grains found in ethanol washings of Homalictus fijiensis captured after interacting with M. denticulatum flowers. Unless otherwise indicated by numbers within bars, sample size was 100 pollen grains

Full size image
Fig. 4
figure 4

a Homalictus fijiensis performing pollen larceny by chewing on the anthers of the Melastoma denticulatum flower; b microscope image of M. denticulatum pollen taken under ×40 magnification; (i) a mature pollen grain of lighter colour; (ii) an immature pollen grain of darker colour. See video in electronic supplementary material

Full size image

Discussion

Our data indicate five clear results. Regarding super-generalism: (i) H. fijiensis has a substantially broader host range than the introduced bee species, including Apis mellifera, a species that is widely regarded as a super-generalist; (ii) interaction networks display significant nestedness for both bees and plants; (iii) the parameter estimating disaggregation, C-score, was significantly different from the null model for plants and bees, suggesting either competitive exclusion or host/bee specialization; and regarding pollen larceny: (iv) H. fijiensis also visits six other solanoid plant species (five of which are nectarless) and only one of those species was observed to be visited by another bee species (Apis mellifera); and (v) H. fijiensis extracts pollen from the poricidal anthers of the native Fijian solanoid plant M. denticulatum by biting at anthers. The implications for these findings are discussed below.

The super-generalism of H. fijiensis suggested by Crichton et al. (2018), and extended by our new data and covering multiple solanoid plants, is interesting. Apis mellifera is regarded as a super-generalist, but its host range in Fiji is narrower than H. fijiensis. Importantly, most plants in our interaction network are not native to Fiji. Apis mellifera evolved in Eurasian continental regions (Ruttner 1988) that have much greater plant diversities than Fiji, so the greater level of super-generalism in H. fijiensis, compared to Apis mellifera, cannot be explained by exposure to a larger number of host plant species over evolutionary time. Instead, it seems likely that H. fijiensis has evolved its host breadth in a situation where intra-specific competition was high and floral morphologies were diverse.

We observed H. fijiensis visiting a total of seven plant species with poricidal anthers (five of which are nectarless), but only M. denticulatum is native to Fiji. Of these seven plants, only one was visited by a bee (Apis mellifera) other than H. fijiensis, suggesting a strong ability for bees introduced to Fiji to recognize and avoid floral morphologies that do not provide nectar or high pollen rewards. This was particularly evident from our observations of M. denticulatum plants where non-native bee species were highly abundant in the immediate vicinity but never approached M. denticulatum flowers, despite their showy appearance. It is therefore likely that bees unable to exploit solanoid plants have evolved or efficiently adapted behaviour to avoid those plants.

Staines et al. (2017) showed that H. fijiensis exhibits pollen larceny for the introduced solanoid weed S. torvum, and when doing so may effect pollination as pollen was present on the bodies of H. fijiensis and contact is made with both the anthers and stigma. The same was observed in our study of M. denticulatum. Visits to this plant never involved attempts to locate nectaries suggesting that these bees have evolved the ability to recognize plants with poricidal anthers, which was a puzzle because S. torvum was only introduced to Fiji in the 1800s. There is only one potentially indigenous Solanum species, Solanum viride (Smith 1991), however, it is unclear if this species existed there prior to the arrival of the Lapita or the Polynesians. All other species of Solanum are considered to be of known introduced origin to the islands. Pollen larceny has not been recorded for any halictine bees that we are aware of, but was also recorded from a further five introduced plants in the present study. Melastoma denticulatum is native to the Indo-Papuan, Fijian and some other South West Pacific (SWP) regions (Meyer 2001) and has a broadly similar floral appearance to S. torvum (see Fig. 2 in Staines et al. 2017), so it is possible that Homalictus evolved an ability to recognize this kind of floral morphology before these introduced solanoid plants arrived in Fiji. In particular, a lack of floral specialisation and broad host range of H. fijiensis would have likely assisted in the exploitation of the familiar morphology of the invasive S. torvum (Padyšáková et al. 2013).

Interestingly, several native Homalictus species have been recorded on flowers of M. denticulatum in New Caledonia (Pauly et al. 2015) and Capsicum annuum from Papua New Guinea (Ibalim et al. 2020), while two other halictines, Lasioglossum sp. and Nomia sp. (with six other bee species) have been recorded from M. affine from tropical Australia but pollen larceny was not observed for these halictines (Gross 1993). The Fijian Homalictus species comprise a separate clade to both New Caledonian and Papua New Guinean species (Groom et al. 2014; Ibalim et al. 2020), suggesting either that the ability to utilize plants with poricidal anthers has evolved multiple times or that it evolved prior to the crown age of the Fijian clade, which is at least 400,000 ya (Groom et al. 2014).

It is puzzling that H. fijiensis is able to exploit so many solanoid plants in Fiji when reports of similar behaviour for non-buzz pollinators are very rare for other regions (but see Gross 1993 for other bee species). We suggest two possible factors: (i) Fiji does not have any native buzz-pollinating insects (Groom et al. 2014; Staines et al. 2017; Dorey et al. 2019), so plants with poricidal anthers that have colonized this archipelago would have comprised a pollen resource that was unavailable to other invertebrate pollinators; and (ii) prior to human settlement, approximately 3000 ya (Clark and Anderson 1999) Fiji largely comprised dense tropical forests that are unsuitable for ground-nesting bees (Ash 1987, 1992), so competition to exploit local floral resources in non-forested areas could have been intense. Because Fijian native bee diversity is extremely low in lowland habitats, with only one species prevalent below 800 m asl (Michener 1979; Dorey et al. 2019), intra-specific competition to exploit all available resources could have been high, selecting for super-generalism rather than species-specific niches.

Future research

Our finding that H. fijiensis commonly visits solanoid plants raises the issue if this bee may be able to pollinate solanaceous species, as has been suggested with S. torvum (Staines et al. 2017). To confirm pollination and determine the efficacy of pollen larceny, however, further research is required (see Gross 2005; Solís-Montero et al. 2015). If H. fijiensis is found to successfully effect pollination, the combination of pollen larceny behaviour and wide host range will pose a greater invasion threat from introduced solanaceous weeds because pollination services may be provided without invader complexes, previously thought to be required (Simberloff and Von Holle 1999; Simberloff 2006). However, this behaviour of H. fijiensis also has the potential to improve solanaceous crop pollination services in tropical countries. The only solanaceous crop plant in our study was C. frutescens, but visits to S. torvum suggests that this bee may also exploit other cultivated solanceous crops, such as tomatoes, eggplants, capsicums and chillies. The use of native bees to pollinate crop plants is being investigated world-wide to secure the agricultural industry in the face of declining insect populations (Sánchez-Bayo and Wyckhuys 2019) and native carpenter bees have been investigated as an alternative to bumble bees in Australia (Hogendoorn et al. 2000). In Brazil, native bee alternatives for pollinating crop plants have been successful, where 10 of 13 native bee species studied could pollinate Lycopersicon esculentum (Bartelli and Nogueira-Ferreira 2014). It was also suggested that even in the presence of typical crop pollinating species there are greater benefits in utilising many pollinating species to avoid reliance and overexploitation of a single species but a crucial factor in the successful use of native species for agricultural crop pollination was to understand commercial management of each species (Bartelli and Nogueira-Ferreira 2014).

The ability of H. fijiensis to exploit a very wide variety of introduced weeds with varying floral morphologies may extend to introduced crop species that benefit from pollination services. However, in this sense, super-generalism in insular bees may be a two-edged sword, encouraging the spread of weedy species but could possibly provide additional or alternative pollinators for agricultural crop pollination. More research is needed regarding population management and sustainability for a super-generalist such as H. fijiensis to ensure effectiveness in supporting crop pollination.