Pollen types collected by Tetragonisca angustula (Hymenoptera: Apidae) in dry vegetation in Northeastern Brazil

Knowing the floral origin of the pollen collected by native bees in Neotropical ecosystems enables us to understand the dynamics of the interdependent relationships between the insects and the native flora. The objectives of this study were to: (i) investigate the spectrum of plant species from which pollen is collected by Tetragonisca angustula in caatinga (dry vegetation) areas in the semiarid region of Northeastern Brazil; (ii) identify pollen types that could be used as regional geographical markers and (iii) determine the pollen niche breadth and the pattern of use of floral resources by this stingless bee. In total, 23 samples of the pollen stored by T. angustula were collected monthly and subjected to acetolysis. Of the 45 pollen types identified that of Prosopis juliflora (Fabaceae) and Solanum (Solanaceae) were the most frequently represented in the samples and 19 were collected by the bees at both study sites. Species characteristic of caatinga, such as Poincianella pyramidalis (Tulasne) L.P. Queiroz and Senna macranthera (de Candolle ex Colladon) H.S. Irwin & Barneby, were among the pollen identified and they were used as indicators of the geographical origin of the samples. The average values of the pollen niche breadth (H’) and equitability (J’) indicate that T. angustula is a generalist and homogeneous forager of the floral resources in caatinga.


INTRODUCTION
Over the past several decades, deforestation, habitat loss and the introduction of exotic species have resulted in a decline in populations of native pollinators. This phenomenon has led the international scientific community to attempt to understand the implications of this decline for global ecosystems and propose mechanisms for reducing this decline (Allen-Wardell et al., 1998;International Pollinators Initiative, 1999;Biesmeijer et al., 2006;Freitas et al., 2009;Burkle & Alarcón, 2011;Giannini et al., 2012). As the environmental services provided by native bees in particular are important it is essential to understand those aspects of their biology, behaviour and ecology that can be used to develop conservation initiatives designed to maintain pollination (Allen-Wardell et al., 1998).
The pollinators are primarily female bees that collect pollen as the main protein source for themselves and their larvae (Michener, 2007). In the Neotropical region, however, it is difficult to carry out continuous and systematic studies of the flora that provide this protein diet for native bees. In Brazil, such studies have been ongoing for decades; however, they have not been done in all the biomes in Brazil (Barth, 2004;Borges et al., 2006).
Among the Brazilian biomes, the phytogeographic domain of caatinga is the largest gap in our botanical knowledge, largely because of the erroneous belief that this type of vegetation originated from the modification of another plant formation (Giulietti et al., 2004). According to Queiroz (2006), historically, the species diversity of the flora of the caatinga is considered to be low and include few endemics. It is suggested that the caatinga does not have an autochthonous flora and that most of its elements are derived from chaco and Atlantic rain forest (Rizzini, 1963;Andrade-Lima, 1981). These hypotheses were subsequently rejected (Prado & Gibbs, 1993). Recent studies indicate that the flora of the caatinga is surprisingly diverse (Giulietti et al., 2002(Giulietti et al., , 2004(Giulietti et al., , 2009Albuquerque et al., 2012) and includes 4,320 species of angiosperms (Forzza et al., 2010), which have relationships with a range of pollinators including 187 species of native bees . There is a mosaic of different ecosystems in the dry vegetation in the caatinga, which covers an area of approximately 800,000 km² and practically the entire semiarid climate region in Brazil, from the northeast through a region north of the State of Minas Gerais (Prado, 2003;Giulietti et al., 2004).
Among the Meliponini reported in the Brazilian caatinga, Tetragonisca sp. group angustula (Latreille, 1811) is one of the most well-known and studied groups. The geographical distribution of this group of stingless bees goes beyond the Brazilian territory, extending from Mexico to Argentina (Camargo & Pedro, 2012). Although the pollen collected by T. angustula has been studied in many different types of Brazilian vegetation (Iwama & Melhem, 1979;Imperatriz-Fonseca et al., 1984;Carvalho et al., 1999;Morgado et al., 2011;Novais & Absy, 2013), few studies have been done in caatinga. Novais et al. (2006) performed a palynological study of the honey of this species collected in a hyperxerophytic caatinga region at the Canudos Biological Station in the northeastern micro-region of Bahia State. Recently, the pollen contents of honey from T. angustula colonies at Itaberaba and Ruy Barbosa in Bahia State were investigated (Novais et al., in press). However, there are no reports of palynological studies of the pollen stored by T. angustula foraging in caatinga vegetation.
Because there is no information on the floral spectrum foraged by T. angustula for collecting pollen in dry vegetation in the caatinga region of the Brazilian semiarid region we asked the following questions: (i) what pollen types reflect the flora visited by T. angustula in the caatinga of Bahia?; (ii) which of the various pollen types collected by this species can be used as regional geographical markers?, and (iii) does the pattern of use of pollen resources in the areas of caatinga investigated indicate a homogeneous or heterogeneous use of the local flora?

Study areas
This study was carried out in the municipalities of Itaberaba (12°26´18.8˝S, 40°13´12.7˝W) and Ruy Barbosa (12°17´27.6˝S, 40°30´03.0˝W), which are located in the semiarid climate zone in the state of Bahia, Northeastern Brazil (Fig. 1). The annual precipitation in this area does not exceed 800 mm and the average temperature is approximately 25°C (EMBRAPA, 2013).
The most diverse plant families in the region include Fabaceae, Euphorbiaceae, Apocynaceae, Orchidaceae, Rubiaceae and Asteraceae, of which there are more than 120 recognized species (Cardoso & Queiroz, 2008). Depending on altitude, various plant formations can be distinguished locally, from caatinga with palm trees at lower altitudes (up to 450 m), to vegetation similar to savanna and areas of rupestrian fields (at approximately 900 m) (Cardoso & Queiroz, 2008).

Sampling and laboratory processing
Samples of pollen (23 in total) collected from stores in T. angustula colonies in the study areas were examined. At Itaberaba, samples were collected (n = 10) monthly during 2011, with the exception of July and November when there was a scarcity of pollen in the colonies. At Ruy Barbosa, the samples were collected monthly (n = 13) between October 2010 and December 2011, with the exception of June and November. Each month, the samples were collected from the same colony, mainly from the five pollen pots that had been sealed most recently by the bees, which were usually of a lighter colour than older pots. The material was collected from the food pots using pieces of disposable plastic straws (approximately 3 cm), which were packed in lidded pots, labelled and kept in a refrigerator until used in the laboratory procedures.
The protocol adopted was that of Novais & Absy (2013), which included drying the material in an oven at 40°C to stabilize the weight, dissolving in warm water and 95% ethanol, dehydrating in glacial acetic acid and acetolysis (Erdtman, 1960).
For each sample, three slides were prepared using glycerin gelatin and at least 500 pollen grains were quantified and identified using optical microscopy. The botanical identification of the pollen types was determined using catalogues (Roubik & Moreno, 1991;Silva, 2007) and the pollen library at the Laboratory of Plant Micromorphology of Bahia State University at Feira de Santana. Because each pollen type represents a morphological entity that potentially includes various species or genera (Joosten & de Klerk, 2002), the author was cited for each species when, for the first time, the taxonomic entities are referred to in the text.
After the microscopic analysis, the classes of occurrence (C.O.) were established based on the presence or absence of each pollen type in the samples as follows: r -rare (present in < 10% of samples); i -infrequent (10-20%); f -frequent (21-50%); and vf -very frequent (> 50%) (Jones & Bryant, 1996). These classes do not consider the number of pollen grains in each sample.

Ecological analysis
The pollen niche breadth of T. angustula indicated by each sample was estimated using Shannon-Weaver's diversity index (H') (Shannon & Weaver, 1949), which was calculated using the following formula: In the above equation, pi is the proportion of each pollen type i in the sample. This proportion is given by , where ni is the ni N number of grains of each pollen type i and N the total number of pollen grains in the sample.
Pielou's equitability index (J') (Pielou, 1977) was used to indicate the dynamics of the use of pollen resources by the bees. This was calculated using the following equation: All analyses were performed using PAST (Palaeontological Statistics) software version 2. 16 (Hammer et al., 2001).

RESULTS
Our data indicate that T. angustula used a large number of plant species as 45 pollen types are listed in the spectra (Tables 1 and 2). Of these 19 were common to both study areas, 10 were recorded only in the samples from Itaberaba and 16 only in the samples from Ruy Barbosa. In total, 29 pollen types belonging to 18 plant families were recorded in the samples from Itaberaba (Table 1) and 35 from 25 families in the samples from Ruy Barbosa (Table  2). Of the plant families the Fabaceae was the most frequently represented family, with eight morphologically different pollen types (Tables 1 and 2).      The main pollen types present in the samples collected at Itaberaba include the following: Brosimum, Prosopis juliflora, Senna macranthera and Solanum (Table 1). At Ruy Barbosa, the main types were Brosimum, Heteropterys, Prosopis juliflora, Schinus, Senna macranthera and Solanum (Table 2). These pollen types were present in more than 50% of the samples analyzed ("very frequent" class of occurrence) and greater than 10% in at least one of the samples.

27
T. angustula very frequently collected pollen from Prosopis juliflora (Swartz) de Candolle. In addition to being present in all of the pollen samples analyzed from both study areas (n = 23), P. juliflora occurred with a frequency of greater than 10% in 20 of the samples (86.96%) (Tables 1 and 2). The second most represented pollen type was Solanum, which occurred in 90% of the samples from Itaberaba and 92.31% of those from Ruy Barbosa. The third most used pollen type at Itaberaba was Senna macranthera, which occurred in 80% of the sam-       ples from this municipality, followed by Poincianella pyramidalis (70%), Mikania (60%), Brosimum (60%), Myrcia (50%) and Heteropterys (50%). At Ruy Barbosa, Brosimum and Schinus each occurred in 84.62% of the samples, followed by Senna (76.92%) and Heteropterys (61.54%).

28
At Itaberaba, only the Mikania and Poincianella pyramidalis pollen types occurred at a frequency lower than 10% in more than 50% of the samples (Table 1). At Ruy Barbosa, the pollen types included in the "very frequent" class of occurrence demonstrated a greater than 10% frequency in at least one sample (Table 2). Although a large number of pollen types were recorded, less than 25% of these were found consistently at a high percentage (> 10%) (Ramalho et al., 1985) in the T. angustula collections (Tables 1 and 2).
In 2011, eight pollen types were found in at least three consecutive months at Itaberaba, Poincianella pyramidalis (Jan-May), Prosopis juliflora (Jan-Dec), Solanum (Feb-Dec), Senna macranthera (Feb-Aug), Zornia echinocarpa (Apr-May), Myrcia (May-Aug), Brosimum (Jun-Dec) and Mikania (Sep-Dec). The occurrence of these pollen types in consecutive months was more evident in the second half of 2011 (Table 1) (2011) of pollen types that were recorded simultaneously in the pollen spectra obtained for 2 semiarid regions in Bahia in Northeastern Brazil. The data are derived from the palynological analysis of the pollen stored by Tetragonisca angustula.  Table 2).
In the months when the collections were performed at both study sites, the number of pollen types identified in the samples from Itaberaba each month was slightly lower than at Ruy Barbosa (Fig. 2, Tables 1 and 2). Furthermore, in these months, the number of pollen types recorded simultaneously in the spectra for the two areas varied from three (in Jan and Apr 2011) to eight (in Dec 2011) (an average of 4 ± 1.5 s.d., n = 9) (Fig. 2, Table 3). In total, eleven pollen types were recorded in the samples of pollen collected by Tetragonisca angustula in the same months at Itaberaba and Ruy Barbosa, and include the families Anacardiaceae, Arecaceae, Asteraceae, Faba-ceae, Malpighiaceae, Malvaceae, Moraceae, Myrtaceae and Solanaceae (Table 3, Fig. 3).
The average equitability at Itaberaba was 0.50 ± 0.15 s.d. (n = 13), with the minimum and maximum recorded in Jan 2011 (J' = 0.24) and Jun 2011 (J' = 0.69), respectively (Fig. 4, Tables 1 and 2). At Ruy Barbosa, the The monthly H' and J' values for Itaberaba decreased or increased in parallel (Table 1). At Ruy Barbosa, however, this parallelism in behaviour was not consistent. In 2011, in Feb-Mar and Mar-Apr, opposite trends were recorded in the H' and J' values: the H' index decreased when the J' value increased and vice versa (Table 2).

DISCUSSION AND CONCLUSIONS
This study indicates that T. angustula collected pollen from a wide spectrum of flowers but especially from Prosopis juliflora (Fabaceae) and Solanum Linnaeus spp. (Solanaceae), which makes up a major part of the pollen spectrum throughout most of the period studied. The consistent use of certain floral sources by the bees and other pollinators may reflect the ability of these insects to "learn" or "memorize" their floral characteristics, such as colour, odour, size, and patterns, which function as indicators of rewards, such as nectar, pollen or resin (Gonzalez et al., 1995;Chittka et al., 1999;Glover, 2007;Nicolson, 2007). Although the search for particular flower sources can often be inferred from pollen spectra, this fact does not necessarily imply that these species provide a greater amount of the resource than other species. Chittka et al. (1999) suggest that in the search for a preferred species, bees could neglect flowers of other species of plants that offer as much or more of the resource than the preferred plants. In addition, pollen is not always a reliable indicator of a nectar source because there are no in depth studies of how the availability of resources varies even within the same genus or of the floral and reproductive biology of a large number of species of plants in areas such as the caatinga (Iwama & Melhem, 1979;Machado & Lopes, 2004;Santos et al., 2010;Roubik & Moreno, 2013). Therefore, it is essential to carry out more field studies to ratify or rectify the inferences made based on pollen spectra (Roubik & Moreno, 2013).
The pollen grains in bee products provide clues to their geographical origin because the key plant species that characterize the region where they were produced can be identified (Maurizio, 1975;Louveaux et al., 1978). Giulietti et al. (2002) lists 318 species endemic to areas of caatinga in Northeastern Brazil. Of the genera identified in this study, 13 were represented in the pollen spectra recorded for Itaberaba and Ruy Barbosa, however, only a few of these can be considered to be geographical indicators of this type of vegetation, such as Poincianella pyramidalis, Senna macranthera and Zornia echinocarpa (Giulietti et al., 2004). Queiroz (2008) reports that Poincianella pyramidalis (Tulasne) L.P. Queiroz var. pyramidalis is characteristic of areas of sensu stricto caatinga, where small-sized trees are predominant but do not form a continuous canopy, the trees and bushes exhibit xerophytic characteristics and the herbaceous plant layer is only present during the rainy season. P. pyramidalis var. pyramidalis is typical of the state of Bahia and adjacent regions of caatinga in Pernambuco and Alagoas States, mainly occurring in arboreal caatinga. The majority of the insects visiting the flowers of P. pyramidalis seek nectar, although the volume of this resource per flower is limited, which may cause the visitor to forage from a larger number of flowers, thereby increasing the chance of effectively pollinating the species (Leite & Machado, 2009).
There are three different varieties of Senna macranthera (de Candolle ex Colladon) H.S. Irwin & Barneby occurring in areas of caatinga (Queiroz, 2008), of which only S. macranthera var. micans (Nees) H.S. Irwin & Barneby is reported by Cardoso and Queiroz (2008) in the region of the Serra do Orobó [Orobó Mountains] in the municipalities of Itaberaba and Ruy Barbosa. Pollen from the genus Senna Miller is recorded in various palynological studies carried out in areas of caatinga (Novais et al., 2009;Oliveira et al., 2010; and S. macranthera is an important source of pollen used by the bees in this type of vegetation (Maia-Silva et al., 2012). It is likely that T. angustula can collect the pollen remaining in the flowers of this species after their anthers have been shaken by the vibrations generated by larger bees, such as those of the genera Xylocopa Latreille and Bombus Latreille (Machado & Lopes, 2004;Maia-Silva et al., 2012). Queiroz (2008) reports that Zornia echinocarpa (Moricand ex Meissner) Bentham occurs in restinga (coastal vegetation) and caatinga on sandy soil only in the state of Bahia. Cardoso & Queiroz (2008) confirm the presence of this species in the Serra do Orobó. Pollen from Zornia echinocarpa is among the dominant types in the pollen loads of Apis mellifera Linnaeus foraging in the caatinga region of Canudos, Bahia (Novais et al., 2010). In our study, Zornia echinocarpa pollen was present at a low level in the spectra.
In contrast, pollen types indicating genera and species important for bees in the caatinga region are recorded, both in nectar [e.g., Prosopis juliflora (Fabaceae) and Waltheria Linnaeus (Malvaceae)] and pollen [e.g., Solanum (Solanaceae)] (Machado & Lopes, 2004;Maia-Silva et al., 2012). The Prosopis juliflora pollen type occurred most frequently in the samples studied. This species is not native to caatinga, however, it occurs spontaneously in various regions where agricultural activities have ceased (Queiroz, 2008). Prosopis juliflora may be pre-adapted to surviving in caatinga because of its rapid growth and resistance to drought (Sajjad et al., 2012). In the study areas, this species flowered practically throughout the entire year, constituting a continuous source of resources for local bees.
Other sources of pollen for bees in the study areas were the genera Brosimum Swartz, which is common in pollen spectra although predominantly an anemophilous species (Martins & Batalha, 2006), and Virola Aublet (Lena & Oliveira, 2006). Cesário & Gaglianone (2008) report that the simple flowers of Schinus Linnaeus are a source of pollen and nectar for various species of insect. Heteropterys Kunth, although considered to be a source of oil for Centridini bees in areas of caatinga (Aguiar et al., 2003), was also regularly visited during several months at Itaberaba and Ruy Barbosa and, therefore, is also a pollen source (Rocha-Filho et al., 2012).
Compared with that of other Meliponini, the trophic niche of Tetragonisca angustula is relatively small (H' < 1.5) in the two areas studied (Carvalho et al., 1999;Oliveira et al., 2009;Ferreira et al., 2010). However, values close to those recorded in our study were obtained for T. angustula at other Brazilian locations. Novais & Absy (2013)  In general, the trends in H' and J' run parallel with one another. However, in some cases, this is not the case as was recorded from November to December 2010, at Ruy Barbosa. Although the value of H' was the same (1.18) in both months, that of J' increased (from 0.46 to 0.54), due to the decrease in the dominance of the pollen of Prosopis juliflora, which was the predominant pollen (71.01%) in November 2010. While the most dominant pollen type in December 2010 was Solanum pollen type (57.12%), but at a frequency lower than that recorded for Prosopis juliflora in the previous month. In contrast, in December 2010, the frequency of P. juliflora pollen was 17.97%. Thus, the co-dominance of the Solanum and Prosopis juliflora pollen types in December 2010 may explain the increase in uniformity recorded in the spectrum (J').
Another example occurred from April to May 2011 when there was an increase in H' and a decrease in J'. This decline was due to the dominance of Solanum pollen type (54.88%). In April 2010, Prosopis juliflora dominated the spectrum (54.91%), followed by Solanum (24.91%) and Senna macranthera (14.15%). In May 2010, the Solanum pollen type was followed by that of Prosopis juliflora (12 .37%) and Heteropterys (11.67%).
The lowest values of these last two types, compared to the frequencies of Solanum and Senna macranthera in April 2010, indicated a lower uniformity in the use of pollen resources in May 2010.
The uniform use of pollen sources was slightly lower at Itaberaba than at Ruy Barbosa. Carvalho et al. (1999) recorded that compared with Apis mellifera and three species of Meliponini of the genera Nannotrigona Cockerell, Partamona Schwarz and Plebeia Schwarz, T. angustula foraged less uniformly at Piracicaba (São Paulo). The diversity of floral sources used by Tetragonisca angustula, together with the consistency exhibited by this species in the use of these resources, confirmed the generalist habit of this species. The small size of this insect may result in reduced energy expenditure (Imperatriz-Fonseca et al., 1984) per flower while foraging and probably per minute of flight, making various plants equally attractive for supplying food rewards. We may hypothesize that flowers offering a low quantity of resource (pollen or nectar) should be almost as attractive to small bees as those which offer high pollen/nectar rewards. This can be advantageous if they are competing with other foragers for the same resources.
Therefore, we conclude that T. angustula exploits a significant number of species of plants in the study areas, which confirms that this species has a polylectic foraging strategy. A number of species of plants can be used as indicators of caatinga vegetation, such as Poincianella pyramidalis and Senna macranthera. The pollen types Prosopis juliflora and Solanum were the most frequently recorded in the samples, demonstrating their prioritization as sources of pollen by Tetragonisca angustula at Itaberaba and Ruy Barbosa. The ecological analyses indicated a homogeneous use of floral resources by T. angustula in the areas of caatinga studied. High homogeneity implies low dominance of a single pollen type in the pollen spectrum. That the exotic Prosopis juliflora is such an important source of pollen for this native bee may have implications for the interactions between native and non-native species in this and other eco-regions that contain generalist species with broad niches.
ACKNOWLEDGEMENTS. The authors thank C. de Souza Dias for contacting the bee keepers who provided samples from their private apiary; the regional office of the Bahia Company of Agricultural Development (EBDA, Empresa Baiana de Desenvolvimento Agrícola S.A.) in Itaberaba for their logistic support; the board of directors of the Agricultural Family School Mãe Jovina for access to the apiary at the institution in Ruy Barbosa; the beekeepers Alex and Manoel for providing samples for the study; B.H. Andrade-Silva for his help in the field and labora-tory; L. Santos Damascena for generating the map used in this study; and the National Council of Scientific and Technological Development (CNPq, Conselho Nacional de Desenvolvimento Científico e Tecnológico) for partially financing this study and for the following grants: 575747/2008-0, 143084/2009-7, 303557/2010-9 and 477127/2011-8.