Pollen sources used by Frieseomelitta Ihering 1912 (Hymenoptera: Apidae: Meliponini) bees along the course of the Rio Negro, Amazonas, Brazil

Insect pollination has influenced the evolution and diversification of angiosperms. Consequently, knowing plants used as food sources by bees, the most important pollinator group, is the first step toward understanding how their ecological relationships works. Pollen source information is also highly relevant for bee management and associated fruit and seed production. Accordingly, to improve understanding of the trophic ecology of these bees and their relationships with native Amazonian plants the current study identified, quantified and compared pollen stored in nests of Frieseomelitta stingless bees along the Rio Negro, Amazonas, Brazil. A total of 31 pollen pots were sampled and found to contain 65 pollen types distributed across 52 genera and 29 botanical families (predominantly Arecaceae, Araliaceae, Fabaceae and Urticaceae). Euterpe was the commonest pollen type, being present in 32.2 % of the analyzed samples. Although the studied bees were generalists, pollen analysis suggested that different Frieseomelitta species may have distinct food preferences. The pollen profile of the studied bees was influenced more by nest location than by species phylogenetic proximity. The current study also provides a list of important plants for native bee management, which could improve beekeeping when grown near managed meliponarine colonies.


Introduction
Insect pollination influenced the evolution and diversification of Angiosperms (Lima 2000;Soltis et al. 2019), especially in the tropics where the dependence of plants on biotic pollinators appears to be higher than in more temperate environments (Maués et al. 2012;Rech et al. 2016).Within the tropics, where pollinators are central to biodiversity maintenance, bees appear as especially important, being identified as key floral visitors in many plant families (Ollerton 2017).For bees, pollination is driven by their dependence on plant-produced trophic resources since, from the larval phase until senescence, nectar and pollen provide bees with their sole sources of glucose and protein, respectively (Corbet et al. 1991;Nogueira-Neto 1997;Absy et al. 2018).
Therefore, understanding the network of interactions between bees and plants is essential, as knowledge of the diet of each bee species provides the first step towards understanding the multiple levels of plant-insect interdependence, and the dynamics of these relationships (Absy et al. 2018).In addition, such studies can generate data that informs guidelines for defining conservation strategies for both plants and pollinators, as well as assisting in the evaluation of pollination as an ecosystem service (Kearns et Acta Botanica Brasilica -34(2): 371-383.April-June 2020 al. 1998; Kremen et al. 2004;Wolowski et al. 2018).In the Neotropics stingless bees of the Tribe Meliponini (Family Apidae) are especially important in this regard (Imperatriz-Fonseca & Nunes-Silva 2010).
One of the ways of accessing species involvement in the bee/plant interaction network is through direct observation (Freitas et al. 2014).However, due to the number of species involved, this method becomes very complex in tropical forest environments (Absy et al. 2018).Instead, palynology offers an indirect tool for the identification of the pollen collected by bees, one which allows access to the preferred food sources of various bee species, as well as others used as alternative and occasional resources (Rech & Absy 2011a;b).This method can be very effective because it provides access to the full spectrum of resources used by the bees, such as tall flowering trees, small flowering herbs and/or individual plants far-removed from study colonies, and which cannot be accessed easily by direct observation (Imperatriz-Fonseca et al. 1989).Therefore, once the dynamics and resources of meliponines are known, it becomes possible to test their effectiveness as pollinators, and so manage them to increase fruit production and so reduce extinction risk for endangered plant species, among other benefits.
Frieseomelitta is an ecologically important stingless bee genus in Amazonia, where its members are commonly called breu, moça branca or marmelada (Nogueira-Neto 1997).The genus is poorly known, with scattered papers on biogeography, phylogeny and autoecology of member species (Teixeira 2003).Moreover, it comprises a relatively large, diversified group, with a wide geographical distribution, ranging from the Mexican southwest to the Brazilian southeast (Marques- Souza et al. 1995;Silveira et al. 2002).Thus, many species still lack key baseline information.Species belonging to this genus are generally small (about 6 mm) and usually build their nests in rotting holes in tree trunks.The literature indicates that these bees are generalist foragers, visiting many plant species, but concentrating sequentially on a few species (Marques- Souza et al. 1995;Teixeira 2003).
The current study is part of a series conducted in the Rio Negro-Amazonas region aiming to gain key baseline in a variety of aspects of meliponin ecology.Previous results of this initiative were presented by Rech & Absy (2011a;b) who worked with several genera of stingless bees from the region, including: Aparatrigona; Cephalotrigona; Nogueirapis; Oxytrigona; Partamona; Plebeia; Ptilotrigona; Scaptotrigona; Scaura; Schwarzula; Tetragonisca and Trigona.In these studies, the authors identified important meliponine diet plants and noted temporary specialization events (collection concentrated on specific pollen sources).The results of these previous studies also indicated that collection site was more important than phylogeny in the amplitude and identity of trophic resources used as food by the stingless bees along the Rio Negro.
The current study continues the investigation of pollen resources used by Meliponini species from the Rio Negro region so as to further improve understanding of the trophic ecology of these bees and their relationships with Amazon region native plants.Based on the previous findings, we expect that Frieseomelitta pollen diversity will be more strongly influenced by the sampling site than by the species phylogenetic relationship.Also, given the non-aggressive behavior exhibited by Frieseomelitta (Marques- Souza et al. 2002), we expect they will have a generalist pollen profile similar to the other non-aggressive bees already studied in Amazon, and have foraging behavior characterized by a series of brief resource-specialization events (sensu Rech & Absy 2011a;b).

Materials and methods
The studied pollen material was collected along a stretch of approximately 1,600 km, between the municipalities of Manaus and São Gabriel da Cachoeira, western Brazilian Amazonia (Fig. 1).The collection area was covered by seasonally-flooded igapó vegetation, with some points under the influence of never-flooded Terra-Firme rainforest.According to Wittmann et al. (2006), the seasonally-flooded Amazon forest environment has many species tolerant to flooding, in a system considered the world's most speciesrich rainforest.
Collections of material used for this study were made in the natural habitat of the bees by Dr. João Maria Franco de Camargo (in memorian) and team (M.Mazucato and SRM Pedro) in five communities (Fig. 1) along the Rio Negro main river and its tributaries, during a collection expedition carried out from 15/ July/ 1999 to 15/ August/ 1999.Analyzed pollen samples were collected from nests of Frieseomelitta Ihering bees.Nests found in nature were opened and pollen pots already closed by bees removed and stored in labeled plastic bags and then placed in boxes to avoid pollen pot breakage.
Collection occurred at each nest found.Each pot of closed pollen was considered a separate sample and, for this reason, the number of samples per species varied according to the number of pots found per nest.In total, 31 samples from four bee species (Frieseomelitta flavicornis Fabricius, 1798; Frieseomelitta sp.Ihering, 1912;Frieseomelitta portoi Friese, 1900;Frieseomelitta trichocerata Moure, 1990), encounted across a total of five different locations (Tab.1), were analyzed.For analysis samples were withdrawn from pots using sterile plastic straws, placed in Petri dishes, weighed (to the nearest 0.5 g), then placed in falcon tubes with 3 ml of glacial acetic acid.After 24 hours, the material was acetolysed, following the protocol described by Erdtman (1960).Slide assembly was performed with glycerinated gelatin and sealed with parafinn (Salgado-Labouriau 1973).Three slides were prepared per sample.
Pollen sources used by Frieseomelitta Ihering 1912 (Hymenoptera: Apidae: Meliponini) bees along the course of the Rio Negro, Amazonas, Brazil Pollen type identification was performed by comparison with the reference collection for the current study areas maintained at the Laboratório de Palinologia of the Instituto Nacional de Pesquisas da Amazonia (INPA), as well as consultation with specialized literature (Roubik & Moreno 1991;Carreira et al. 1996;Lorente et al. 2017).For pollen grain taxonomic characterization, the concept of "pollen type" proposed by Joosten & Klerk (2002) and Klerk & Joosten (2007) was adopted.Measurements and photomicrographs were obtained with a Zeiss PrimoStar microscope combined with the AxionCam ICc image capture program.
For each sample, 600 pollen grains were counted for richness/abundance quantification and statistical analysis.To ensure that 600 pollen counts was enough to reach the saturation of the curve required for our analyses, we produced species accumulation curves using rarefaction for each sample (run using R software -R Development Core Team 2019).Following Ramalho et al. (1985), a minimum representation of 10 % in the sample was used to define when a plant species was considered attractive to the bee species in question.
To visualize the operational limits underwhich the beeplant interactions were operating, information relating to  floral biology (flowering strategy, compatibility system, flower morphology, nectar production) was obtained via bibliographical research for species operationally defined as attractive.When investigating temporary specialization events (collection concentrated on a specific pollen source), a minimum cut-off of 90 % representation of a single pollen type in a sample was used (following Rech & Absy 2011a; b).We used collected pollen type abundances to perform a pair-wise cluster analysis.Similarity between each pair of samples was determined using the Bray-Curtis Index, with this index chosen due to its robustness and because it ignores the multiple zeros that are not necessarily true absences in the data matrix (Michin 1987).As defined by Bray & Curtis (1957), the index of dissimilarity is: Where C ij is the sum of the lowest values for only those species in common between both sites, and S i and S j are the total number of specimens counted at both sites.The Index value varies from 0 to 1, being 1 when both communities have identical compositions.We used a dominance index to evaluate the frequency distribution of pollen types in each sample.For this we used the Simpson Diversity Index ((ni/N) 2 ), where ni is the amount of pollen types in the sample i and N is the total pollen grains counted in the sample.Dominance varies from zero to one, being 0 when all pollen types have identical frequencies, and 1 when all pollen-grains come from a single species.Analyzes were performed using PAST software (Hammer et al. 2001).
Sample specimens of the studied bee species were deposited in the "Camargo" -RPSP collection in the Departamento de Biologia, Faculdade de

Results
We analysed a total of 31 pollen pots and identified 65 pollen types assigned to 29 botanical families.Of these, 24 were identified to species, 37 to genus category and three to botanical family (Tab.2).On average, per pot pollen samples contained seven pollen types (range three to 12 types per sample).For most samples, mean species accumulation curve showed saturation below 600 pollen grains, so indicating sampling completeness (Fig. 2).The species Frieseomelitta flavicornis (nest two) and Frieseomelitta portoi (nest five) had the broadest pollen spectra (21 and 18 pollen types, respectively).The most restricted pollen spectrum was recorded for Frieseomelitta trichocerata (nest seven) with just three types (Tab.3).For F. flavicornis, the pollen type with the highest relative frequency was Cecropia type, which was dominant at the two locations at which the bee was sampled, followed by Swartzia type at Curicuriari and Schefflera type at Santa Isabel.For F. portoi nests, the pollen spectrum was distinct at the two sampling localities (Igarapé Açu and Santa Isabel).At Igarapé Açu, the most commonly-collected pollen types were Cecropia type, followed by Schizolobium amazonicum (both with 34 % frequency) and Pourouma type (~12 %).At Santa Isabel, the most frequent pollen types were Ryania speciosa (~38 %), Iriartella setigera (~33 %) and Mabea nitida (~13 %) (Tab.2).
Dominance analysis showed a clear pattern of high values in most analised samples.Two locations, Ponta do Camucurí (PC) and Ilha de Pinto (IP), showed high dominance values and a reduced number of pollen types per sample (Fig. 3).Per pot profile for F. flavicornis revealed a consistent pattern, with high dominance being found in all analysed pots from the two study areas, but this species also had a broad pollen spectrum.Frieseomelitta sp.showed high dominance (above 0.9) in three of the six pots analyzed, with Croton cajucara as the dominant species in all samples.For F. portoi from Igarapé Açú (IA) and Santa Isabel, dominance was low (averages of 0.2 and 0.3, respectively), with the species having pots containing up to ten pollen types.For F. trichocerata, four of the six samples analyzed from Santa Isabel showed high dominance, with Euterpe being the commonest pollen in these samples.The same pattern was observed in samples at two other communities (Ponta do Camucurí and Ilha de Pinto), where dominance was high.In both of these sites, only one sample from each bee species was analyzed, and both showed greatly reduced pollen spectra, with three and four types, respectively.
Cluster analysis was performed, and partitioned the pollen profile in two ways: (i) separating all pots for each species (Fig. 3) and (ii) by grouping pots from the same bee nest at the species/site level (Fig. 4).The analysis showed that, for the same species, diet similarity between bee colonies was low.The only species where the diet was similar between different sample locations was F. flavicornis, which concentrated its collection on Cecropia type pollen at both sample localities.Both F. trichocerata and Frieseomelitta sp. from different localities formed distinct clusters based on their collections, and between-collection similarity was low.Although the species F. trichocerata and F. portoi grouped together, similarity between them was almost zero, indicating their diets were strongly dissimilar (Fig. 4).
In general, collection pot cluster analysis found an internal consistency for pots from the same nests.This led to the creation of four groups: Group 1 was formed by F. flavicornis, even though the species was sampled at different locations (Curicuriari and Santa Isabel), and characterized by high similarity (above 0.6).This group also containing a single F. portoi pot, which like those of F. flavicornis, was dominated by Cecropia type pollen.However, the similarity of F. portoi within the group was low; Group 2 was composed almost entirely of F. trichocerata, which showed high between-pot similarity.Sampled pots all came from the same locality; pots from the same species, but from different localities, showed distance-related variation in diet content (Fig. 4); Group 3 was formed by Frieseomelitta sp., collected at Ponta do Camucurí, plus a single pot of F. trichocerata from Ilha do Pinto.Although the latter placed in this group analytically, it showed a low similarity with the other pots.Finally, Group 4 was composed only of F. portoi pots collected at Santa Isabel, which showed a consistent pattern of having Ryania speciosa and Iriartella setigera as the main pollen types.However, the single pot collected for F. portoi at another community (Igarapé Açú), pollen prevalence may also be a consequence of its being in flower when other species were not.In this scenario, this pollen source becomes highy attractive to bees as more profitable sources are absent.
Similarly, several other pollen types recorded here as important food items for the studied bees (Astrocaryum type, Bactris brongniartii, Bactris gasipaes, Cocos nucifera, Euterpe precatoria, Euterpe type, Iriartella setigera and Iriartella type) were all palms.A scientometric study by Souza et al. (2018) for the period between 2005 and 2017 found clear evidence that the Arecaceae family is one of the botanical families most often visited by bees.Such data agree with other studies (Aguiar 2003;Marques-Souza 2010), which have recorded palm pollen in Frieseomelitta collections.
Bee visitation to palm trees may occur principally because the large numbers of flowers grouped together in clusters allow extensive foraging (Barfod et al. 2003).According to Oliveira et al. (2003), although some palms have specific characteristics that favor visitation by beetles, their morphology also allows visitation by other insects, including bees, revealing a pattern of generalist interactions.In addition, in the Amazon, palm trees have both wide natural distribiutions and are commonly cultivated for their economic potential (which ranges from human and animal food to biodiesel production: Oliveira & Rios 2014).The resulting abundance of palms is good for bee-keeping and especially for Amazonian Meliponiculture.
With 16 collected species recorded, Fabaceae was another botanical family strongly represented in the studied Frieseomelitta bee collections.This corroborates several previous studies that found this family to be important in Frieseomelitta bee diets (Aguiar 2003;Teixeira 2003;Teixeira et al. 2007;Marques-Souza 2010;Aleixo et al. 2013).A review of 28 studies published between 1977 and 2013 (Freitas & Novais 2014) confirms this, reporting 610 types of pollen, from 94 botanical families, of which 129 types belong to the Fabaceae, the greatest contribution by a single family.Nationally, the dominance of Fabaceae was confirmed in the diets of bees through the work of Souza et al. (2019), who found the it to be the family most strongly represented in palynological surveys published between 2005 and 2017.
The similarity analysis performed for bee nests examined in the current study showed a weak clusting between diets of the same bee species at different sites.In contrast, different pots of the same nest were frequently grouped together.Samples of F. portoi from two localities, for example, were located in different groups, while all pots from F. trichocerata clustered in Group 3. It is likely that local plant availability drove this pattern.For example, collections from Igarapé Açú concentrated mainly on Cecropia type pollen, while bees from Santa Isabel, had a preference for Ryania type pollen.Therefore, for Frieseomelitta, patterns of pollen resource composition supported the idea that locality has a stronger influence than species phylogenetic proximity when explaining pollen resource profile of stingless bees, reinforcing the idea that this genus has a generalist, flexible and opportunist foraging behavioral strategy (Rech & Absy 2011a).
The present study showed that some plant species (Cecropia type, Bactris gasipaes, Schefflera type, Schizolobium amazonicum, Cynometra marginata, Croton cajucara, Euterpe type, Ryania speciosa, Iriartella setigera and Scleria type, in order of importance, based on relative frequency) are important for feeding colonies of the bees studied here, as well as showing such bees to be potential pollinators for these species (Fig. 5).These data also indicate that plants of these genera hold promise for use in areas where bee-keeping includes meliponary, and so deserve future studies of their interactions with pollinating stingless bees.The importance of local supply and behavior in the food profile of native bees combined with the importance of both a large number of cultivated palms and Cecropia type as pollen resource to Frieseomelitta, raise important questions about Meliponiculture and forest management in the Amazon region.These, however, will be a matter for future studies.

Figure 1 .
Figure 1.Map showing sample sites along the Rio Negro channel, Amazonas, Brazil.
Filosofia, Ciências e Letras de Ribeirão Preto, Universidade de São Paulo.Pollen slides were placed in the Palynological Library of the Laboratório de Palinologia of the Instituto Nacional de Pesquisas da Amazônia-INPA.Bee specimens were identified by João M. F. Camargo, using the classification of Camargo & Pedro (2008).Plant nomenclature followed Tropicos.org (GARDEN-MOBOT MB 2016) and the classification proposed by APG III (2009) and IV (2016).

Figure 2 .
Figure 2. Mean accumulation curve (rarefaction) calculated for each counted sample of pollen found inside the nests of Frieseomelitta spp.along Rio Negro riverside, Amazonas, Brazil.Upper and lower lines correspond to the standard errors.

Figure 3 .Figure 4 .
Figure 3. Cluster analysis of the pots of each nest of Frieseomelitta flavicornis, Frieseomelitta sp, F. portoi and F. trichocerata along the Rio Negro, Amazonas, Brazil, using pollen spectra identity per colony and a cluster method based on Bray-Curtis Index similarity.Diferents symbols relate to different species, and colours are associated to the sampling site.

Table 3 .
Relative frequency (%) of pollen types (per pot) found inside the nests of Frieseomelitta flavicornis, Frieseomelitta sp, F. portoi and F. trichocerata found along Rio Negro riverside, Amazonas, Brazil.Bee species in the columns are numbered according to Table1, and the plant species in the rows are numbered according to Table2.Sites are coded as follows: SI -Santa Isabel do Rio Negro; CUR -Curicuriari; IA -Igarapé Açu; PC -Ponta do Cumucuri and IP -Ilha do Pinto.