Food niche of Exomalopsis (Exomalopsis) fulvofasciata Smith (Hymenoptera: Apidae) in Brazilian savannah: the importance of oil-producing plant species as pollen sources

ABSTRACT Exomalopsis are ground-nesting species, and their food-niche breadth is little known due the difficulty in locating the nests and finding efficient bait plants to attract these bees. Some species of Exomalopsis were recorded as tomato, hot pepper and eggplant pollinators. Information about the food niche could be useful to increase Exomalopsis populations, providing consistent and comparable data for the enrichment of natural and crop areas with adequate plant sources. This study aimed to determine the food niche and the role of pollen size in the diet of E. fulvofasciata. We analysed pollen loads of 28 individuals of E. fulvofasciata collected from bait plants, in two natural areas of the Brazilian savannah. Only five pollen types belonging to the families Malpighiaceae, Solanaceae, Leguminosae, Myrtaceae and Lythraceae were important for this species. This result indicates that E. fulvofasciata is probably a polylectic species. However, we noticed that the Byrsonima used as bait plants contribute significantly for its larval provision, indicating that small pollen grains were more frequently collected.


Introduction
Exomalopsis is the only genus of Exomalopsini in Brazil (Silveira et al. 2002). Some species of this tribe were reclassified as Teratognathini and Tapinotaspidini (one of the groups specializing in oil collection) (Michener and Moure 1957;Silveira et al. 2002;Alves-Dos-Santos et al. 2007). Exomalopsis species build nests in the ground and some species cooperatively provide for each other's cells, and therefore are considered communal (Rozen 2011).
Exomalopsis (Exomalopsis) fulvofasciata Smith is a species of medium-large bee (≥ 1.2 cm in length), according to the classification proposed by Frankie et al. (1983). This species was recorded in Argentina, Brazil and Paraguay (Silveira 2012). The females build their nests in aggregations in open areas with sandy soil (Aranda and Graciolli 2013).
In general, bee females present a high degree of floral constancy when foraging for pollen. This behaviour contributes to the pollination efficiency and represents a mechanism used by bees to save energy in trying to find new pollen sources (Thorp 1979;Chittka et al. 1999).
The identification of the food sources used by bees can be performed directly through observation of their visits to flowers or indirectly by pollen analyses (Cane and Sipes 2006). The pollen analysis has been used for determining both the foodniche breadth presented by bee species and the influence of the flower characteristics (e.g. type of anthers) and pollen grains (e.g. grain size) on the use of pollen sources, such as verified for oil-collecting bees ('Centridini' and Tetrapediini) (Dórea et al. 2010a(Dórea et al. , 2010bMenezes et al. 2012;Rabelo et al. 2012Rabelo et al. , 2014aRabelo et al. , 2014b. Information on the niche breadth allows the classification of the species into specialization categories. Bees are considered monolectic when they collect pollen only from a single source, oligolectic when collecting from sources belonging to one to three families, and polylectic when collecting from sources belonging to four or more families (Cane and Sipes 2006). Although the use of this classification is controversial, it can be associated with other floral characteristics to elucidate the importance of each pollen source.
The study of E. fulvofasciata food niche can contribute data to crop pollination management. Species of Exomalopsis were recorded as pollinators of tomato (Lycopersicon esculentum) (Santos et al. 2014), hot pepper (Capsicum annuum) (Raw 2000) and eggplant (Solanum melongena) (Montemor and Souza 2009). Based on information about the food niche, Exomalopsis populations could be increased by the enrichment of crop areas with other plant sources, since the management of nests in the soil is difficult.
Considering the difficulty in locating the nests of species that build their breeding place on the ground, an alternative method to obtain information about food niche is the use of bait plants. The choice of bait plants can be made based on previous observations. Thus, during the study of the interactions between Byrsonima species and oil-collecting bees in savannah areas, the visits of E. fulvofasciata were frequent (Rabelo, pers. obs.). Among Malpighiaceae, the genus Byrsonima is one of the most frequently visited by oil-collecting bees to take oil and pollen. However, this flowering plant can also be used by other bee species as a pollen source (Teixeira and Machado 2000;Benezar and Pessoni 2006;Costa et al. 2006;Ribeiro et al. 2008;Bezerra et al. 2009;Mendes et al. 2011;Sazan et al. 2014). These records suggest that these plants can be important food sources and an efficient bait plant for E. fulvofasciata.
This study aimed to determine the food-niche breadth and the role of pollen size in the E. fulvofasciata diet using Byrsonima species as bait plants. We investigated both whether E. fulvofasciata is a polylectic species and whether the bait plants contribute significantly to larval provision.

Samplings and records of bee behaviours during foraging activities
We conducted the bee samplings in the warm wet season, due to the higher number of flowering plant species in the Brazilian savannah during this period (Batalha and Matovani 2000). The Exomalopsis bees were collected during their visits to flowers of Byrsonima (Malpighiaceae) used as bait plants. We performed preliminary observations in order to identify plant species that were flowering at the study areas and were attractive to E. fulvofasciata. We chose Byrsonima as bait plant based on frequent observations of floral visits of E. fulvofasciata.
During the sampling period, only B. intermedia was flowering in the ESP, and B. pachyphylla and B. verbascifolia in the SPSCN (Rabelo LS, per. obs.). Consequently, these Byrsonima species were used as bait plants. The flowering of B. intermedia and B. pachyphylla in Brazilian savannah was registered from September to February (Gaglianone 2001) and July to October (Melo et al. 2014), respectively.
All plants used as baits have yellow flowers with very similar morphologies and offer pollen and oil as floral rewards (Anderson 1979). Like other Byrsonima species, they present anthers with rimosa dehiscence, i.e. anther opening occurs through a longitudinal slot. Their pollen grains are small and powdery and can be directly removed either using the fore leg and mid legs or by vibration (Teixeira and Machado 2000;Benezar and Pessoni 2006;Bezerra et al. 2009;Mendes et al. 2011;Sazan et al. 2014).
We performed three samplings, for three days, from 9 am to 1 pm, totalling 36 h of observations in each area (Table 1). We chose the sampling points that presented at least four individuals of Byrsonima with newly opened flowers. These individuals were located no more than 5 m from each other. During the observation period, a researcher walked among the plant individuals and recorded how the females removed the pollen grains from the flowers, since this bee species does not use oil to nest. After the behavioural record the females were captured using entomological nets, individually wrapped in plastic bottles and immediately sacrificed by freezing.
We captured the bees before they incorporated the grains of Byrsonima into the pollen loads. These procedures reduced the potential risk of biases in the evaluation of the importance of bait plant as a food source for E. fulvofasiata. We also record the collections of pollen and oil performed by oil-collecting bees, the main floral visitors of Byrsonima.
Only the pollen loads presented in the legs were used in the pollen analysis, i.e. the pollen grains that were collected in order to be used as larval provision. In the laboratory, the pollen load was removed and packaged in Falcon tubes containing 2 mL of 70% ethanol. All bees were deposited in the Laboratory of Ecology and Behavior of Bees, Federal University of Uberlândia.
The pollen samples were treated by the acetolysis method (Erdtman 1960). Twentyeight pollen samples of E. fulvofasciata were processed (ESP = 14 samples and SPSCN = 14 samples). For each pollen sample, three voucher slides were prepared and deposited in the collection of the Laboratory of Ecology and Behavior of Bees, Federal University of Uberlândia.
The pollen grains were identified based on the characteristics revealed after acetolysis (Salgado-Labouriau 1973; Roubik and Moreno 1991). Voucher specimens were crossreferenced with a database of pollen grain images (Bastos et al. 2008) and by comparison with the specimens catalogued as reference material in the Laboratory of Plant Morphology, Microscopy and Image, Federal University of Uberlândia. In this study, we used the taxonomic classification proposed by the Angiosperm Phylogeny Group (APG II) (Souza and Lorenzi 2005).
After the identification, the morphologically similar pollen grains were grouped into pollen types according to the methods proposed in the literature (Salgado-Labouriau 1973;Vilhena et al. 2012). In some cases, it was possible to determine the species that composed the pollen type.

Pollen analysis: breadth and uniformity of the food niche and classification of pollen types
After the pollen identification, quantitative analysis was accomplished by dividing each slide into four quadrants, in which approximately 100 pollen grains were counted, totalling 400 for each slide and, therefore, 1200 pollen grains per sample. If the quadrant contained less than 100 pollen grains, all of them were counted and expressed as percentages of the pollen sum . Pollen types with proportion lower than 3% in each sample were discarded, as they were considered either contaminants or only sources of nectar. Quantitative analysis was performed using a magnification of 200×. We calculated the niche breadth using the Shannon-Wiener index (H′): H′ = -Σ (pk × ln of pk), where pk represents the proportion of pollen types present in the pollen load as proposed by Camillo and Garófalo (1989). The uniformity of pollen types in the samples was determined according to the Pielou evenness index (J′): J′ = H′/H′max, where H′max represents the logarithm of the total number of pollen types present in the each area. This evenness index ranges between 0 and 1, and values near 1 represent high uniformity in the pollen collection, i.e. the pollen types were collected at similar proportions. Both analyses were performed for each area separately using PAST 2:13 software (Hammer et al. 2001).
According to type of anther, the pollen types were classified into poricidal and non-poricidal based on literature information. The flowers that present pollen inside a tube-like anther with a small apical pore were classified as poricidal sources, while nonporicidal flowers present anthers with longitudinal or valvular opening (Souza and Lorenzi 2005). These categories were determined due to behavioural differences to explore each of these types of resources (Buchmann 1978). They were also classified according to the size of the grains: small (s) = 10-25 µm and medium (m) = 25-50 µm, as proposed in the literature (Barth and Melhem 1988;Bastos et al. 2008).

Data analysis
To compare the richness of pollen sources between the areas and to verify if our sampling effort was enough to record most of the pollen sources expected to be used in each area we constructed extrapolation curves based on the samples using the program EstimateS 9.1.0 (Colwell 2013). This analysis allows the estimation of the total number of expected pollen types present at the samples if the sampling effort reaches the asymptote (Colwell 2013).
The similarity in the utilization of pollen sources by E. fulvofasciata between the two study areas was determined by PS = Σ (lowest percentage for each pollen type) (Brower et al. 1997). This index is also known as the Renkonen index and is based on the relative abundance of the pollen types presented at each area. It ranges from 0 (the females used a completely different group of pollen sources for larval provision) to 100% (the females used exactly the same pollen source and at the same percentages).
The possible difference in the food-niche breadth between the two assessed areas was evaluated by the one-sample t test of Hutcheson, using the PAST 2:13 program (Hammer et al. 2001).
We verified the possible differences in the relative abundance of pollen sources, according to the type of anthers and size of pollen grains, using chi-square tests for each area separately (Zar 2010).
To establish the influence of the use of Byrsonima as bait plant in each evaluated parameter, all the analyses were repeated excluding this pollen type.

Results
We collected 420 bees, 206 at ESP and 214 at SPSCN. Exomalopsis fulvofasciata represented 10.48% of the records, being the second most abundant species. This species occurred in 77.78% and 44.44% of the samplings, at ESP and SPSCN, respectively.
We collected 44 females of E. fulvofasciata during their visits to flowers of Byrsonima, 23 of them to B. intermedia, at the ESP, and 21 to B. pachyphylla, at SPSCN (Figure 1). Visits to flowers of B. verbascifolia were not observed.
In order to collect pollen, E. fulvofasciata approached the flower by the front and fixed itself using the jaws. The female curved its body over the anthers and vibrated the structure. These procedures removed the pollen grains from the anthers and they remained attached to the ventral surface of the female's body.
Sixteen females exhibited pollen collected by vibration from the flowers of Byrsonima only on the surface of the ventral region of their bodies (ESP = 9 and SPSCN = 7). The absence of pollen in their scopa indicates that these bees did not visit other plants before their capture. Twenty-eight females presented pollen in their scopa (ESP = 14 and SPSCN = 14) before their behaviours were recorded. In both cases, the pollen grains from bait plants were not added to their scopa after the visits.
Pollen analyses showed that only five pollen types, belonging to five botanical families, were quantitatively important for E. flulvofasciata (ESP = five types and SPSCN = three types) ( Table 2). Eleven pollen types presented an abundance of less than 3% in each sample and were excluded from the analyses. The extrapolation analysis showed that 87.11% and 100% of the pollen types estimated for ESP (x = 5.69 ± 1.23) and SPSCN (x = 3.00 ± 0.00), respectively, were sampled ( Figure 2). Therefore, these analyses showed that our sample effort was representative considering that we sampled most of the pollen types that should be present in the pollen loads.
The similarity in the use of pollen types between the areas was high (PS = 89.30%), and we observed few pollen types per sample. At the ESP, we recorded one to four pollen types per sample (x = 1.57 ± 0.82), while at SPSCN, one to two pollen types (x = 1.57 ± 0.49) (Figure 3). The Byrsonima type has been the main source of pollen for E. flulvofasciata, 84.55% and 90.72% at the ESP and SPSCN, respectively. This pollen type was present in all samples of both areas and 14 samples presented only Byrsonima type after the exclusion of those with abundance less than 3% (ESP = eight samples and SPSCN = six samples) (Table 2, Figure 3).  The Acosmium and Solanum types were also present in both areas (Table 2), but in low abundance. The Cuphea and Eugenia types occurred only in samples obtained in the ESP. In both areas, there was predominance of the use of plants with non-poricidal anther (χ 2 ESP = 17,585.70; df = 1, p < 0.001 and χ 2 SPSCN = 13,620.97; df = 1; p < 0.001) (Figure 4). Furthermore, most samples presented only pollen types belonging to nonporicidal anthers, 92.86% in the ESP and 92.86% in the SPSCN.   The food-niche breadth of E. flulvofasciata in ESP (H′ = 0.59) was significantly higher than in SPSCN (H′ = 0.35) (t = 25.61, df = 31,767.00; p < 0.001), and, in both areas, the uniformity of pollen types was low (J′ ESP = 0.37 and J′ SPSCN = 0.30).

Discussion
We observed few pollen types that were quantitatively important for E. fulvofasciata, among which the Byrsonima type was the main pollen source. Considering the spatial and temporal differences in the samples analysed, we can conclude that the use of Byrsonima species as floral resources is a frequent choice of E. fulvofasciata, at least during the wet season and flowering period of these Malpighiaceae species.
Although Byrsonima flowers are non-poricidal, E. fulvofasciata used vibration to collect pollen. The use of vibration on flowers with non-poricidal anthers allows the collection of a larger amount of pollen grains per unit time of handling compared to other behaviours (Buchmann 1985). Thus, it may represent a more economical mechanism of collecting pollen. This strategy has been observed in bees of the genera Centris and Epicharis, considered efficient pollinators of Byrsonima species (Gaglianone 2003;Teixeira and Machado 2000;Bezerra et al. 2009). Taking into account the frequency of visits to flowers and behaviour of by this species of bees, we concluded that E. fulvofasciata probably contributes to the reproduction of B. intermedia and B. pachyphylla.
The high abundance and frequency of the Byrsonima pollen type in the samples, as well as its features, may also influence the predominant use of small pollen grains. Each bee species requires a specific pollen volume for larval provision according to its body size (Müller et al. 2006). Hence, the variation in pollen size (Barth and Melhem 1988) can favour the establishment of a trade-off between the number and the size of pollen grains. Thus, the bees can invest in gathering either many small grains or a smaller number of larger grains. In this context, the small size of pollen can facilitate the collection of a greater number of grains in a few visits to Byrsonima, minimizing the energy expenditure of female bees. This pattern of predominant use of pollen grains with specific size has also been observed for other solitary bee species, such as Centris (Heterocentris) analis (Fabricius) and Centris (Hemisiella) tarsata Smith, which use medium grain sizes (Rabelo et al. 2014a), and Tetrapedia diversipes Klug and Diadasia spp., which use large grain sizes (Sipes and Tepedino 2005;Menezes et al. 2012).
Other pollen sources were also identified. Plants offering only pollen as floral resource, such as Solanum and Eugenia (Souza and Lorenzi 2005), were used as pollen sources and their importance for E. fulvofasciata can be determined based on pollen abundance. On the other hand, flowers that offer pollen and nectar as floral rewards, for example, Cuphea and Acosmium, can be exploited for the collection of both. There are records of nectar exploitation of Cuphea's flowers by Xylocopa and Centris (Ribeiro et al. 2008;Ramalho and Rosa 2010). There are reports on other species that forage on Acosmium's flowers for nectar (Viana and Kleiner 2006), and the high abundance in some samples also suggests their use as a pollen source.
These native plant species used as provision sources for E. fulvofasciata could be used in order to maintain this bee in crop areas, since it is still not possible to manage the nests of this ground-nesting species. The presence of natural areas around the crop could ensure important pollen sources, e.g. Byrsonima, and suitable nesting sites, since this species nests on the ground and some agricultural practices can destroy its nest. Exomalopsis fulvofasciata is considered a pollinator of hot peppers (Raw 2000) and has the potential to pollinate plants that require buzz pollination, since it is able to vibrate in order to collect pollen, as observed in this study.
We also found few pollen types per sample. This pattern may be ascribable to floral constancy. Bees may forage on several flowers of the same species to collect resources (Cane and Sipes 2006). This behaviour results in energy savings by reducing their energy expenditure to locate and handle various floral types (Thorp 1979). As regards to the plants, their floral constancy may favour the process of pollen transport between individuals of the same species promoting pollination and act as a mechanism of speciation (Chittka et al. 1999).
The floristics of the areas (Lopes et al. 2011) and the predominant use of the Byrsonima type may have promoted the high similarity in the use of pollen sources between the study areas. The greatest contribution of the Byrsonima type consisted in providing a low uniformity index in the collection of resources. These results, associated with the great proportion of pollen grains from Byrsonima in all the samples, reinforce the choice of this species during the period of increased availability of floral sources.
Based on the number of plant families used as pollen sources, E. fulvofasciata could be considered a polylectic species (Cane and Sipes 2006). However, considering that our records are restricted to the warm and wet season, additional studies on the dry and cold season are useful to verify if E. fulvofasciata is active in this period and to obtain new information about its food niche.
Our results indicate that E. fulvofasciata is probably a polylectic species. However, we noticed that the Byrsonima type contributes significantly to its larval provision, indicating that small pollen grains were more frequently collected. Taking into account the frequency of visits to flowers and the behaviour during the visits, we also considered that E. fulvofasciata is probably a pollinator of B. intermedia and B. pachyphylla.

Disclosure statement
No potential conflict of interest was reported by the authors.