Network of Bee-plant Interactions and Recognition of Key Species in Semideciduous Forest

The scanty information about species richness, diversity, taxonomy, distribution and population dynamics, along with the impact of human activities on most bee species represent the major environmental issues that hinder the conservation of native areas in Latin America (Freitas et al., 2009). Determining the diversity of both the fauna and flora of many ecosystems is the starting point for specific investigations, such as assessing the role of pollinators in maintaining the flora (Anacleto & Marchini, 2005). The importance of this matter arises therefrom. Studies about networks of interactions between bees and plants are fundamental for the conservation and management of native pollinators habitat (Biesmeijer et al., 2005). Network approaches to ecological research emphasize the pattern of interactions among species, i.e., how the links (interactions) are systematized within the network, rather than only providing the identity of the species that constitute a community (Bascompte, 2009). Abstract Bees are the most effective pollinators of native plants, contributing to the maintenance of many ecosystems, including forests. Studies about networks of bee-plant interactions are critical for conservation and habitat management of native pollinators. This study aimed to determine the richness of flower-visiting bees in a fragment of the Semideciduous Forest (SF) in Uberlândia-MG, identify the plants visited by them and build the bee-plant interactions network in this fragment. The study was conducted between October 2010 and September 2011 along a transect of 200 meters on the edge of the forest, with monthly collections performed from 8:00 am to 2:30 pm. We yielded the bee-plant interaction networks and calculated the NODF index and betweenness centrality. We collected 70 bee species and 25 plant species. The network of interactions was nested (NODF = 10.97, P = 0.03). Apis mellifera Linnaeus and Merremia macrocalyx (Ruiz & Pav.) O’Donell showed the highest centrality. The nestedness of the bee-plant interaction confers stability to the network, demonstrating its importance for the fragment conservation. It is likely that M. macrocalyx is a key species in this network since this plant attracted the highest number of bee species and showed the highest centrality, contributing to network cohesion. Sociobiology An international journal on social insects


Introduction
The scanty information about species richness, diversity, taxonomy, distribution and population dynamics, along with the impact of human activities on most bee species represent the major environmental issues that hinder the conservation of native areas in Latin America (Freitas et al., 2009).Determining the diversity of both the fauna and flora of many ecosystems is the starting point for specific investigations, such as assessing the role of pollinators in maintaining the flora (Anacleto & Marchini, 2005).The importance of this matter arises therefrom.
Studies about networks of interactions between bees and plants are fundamental for the conservation and management of native pollinators habitat (Biesmeijer et al., 2005).Network approaches to ecological research emphasize the pattern of interactions among species, i.e., how the links (interactions) are systematized within the network, rather than only providing the identity of the species that constitute a community (Bascompte, 2009).
Thus, investigations of bee-plant interaction networks can provide relevant information about the community, such as the identification of key species (Martín Gonzalez et al., 2010) and coexistence and stability of species in the system (Bascompte & Jordano, 2007).Consequently, the identification of key species in a community can be helpful in setting priorities and conservation goals (Jordán, 2009).
Inventories of the bee fauna are crucial for the knowledge of composition, geographic distribution patterns and relationships between fauna and flora.Several factors influence the determination of the richness and abundance of local fauna of bees.For example, the amount of food and diversity of available food sources, which favor the existence of large populations and greater consumer species richness, respectively (Silveira et al., 2002).Several surveys of bee species have been conducted in the Cerrado (Carvalho & Bego, 1996;Andena et al., 2009;Siqueira et al., 2012).The conservation status of this biome is becoming increasingly alarming because its vegetation is being hastily replaced by pasture and crops (Freitas et al., 2009).About 50% of the 2 million km 2 of the Cerrado original biome have been transformed into agricultural areas and pastures, besides other types of use, such as planted forests and urban areas (Silva et al., 2006).As a result of this fragmentation process, plant and animal populations have been lost and extinctions may occur.
The Semideciduous Forests (SF) are among the most threatened and fragmented ecosystems in the world.The history of disturbance in Brazilian SF reduced these forests to small scattered fragments and, consequently, led to biodiversity loss in these environments (Santos et al., 2009).Few studies on bee communities have been conducted in SF and studies on beeplant interactions are also rare in this phytophysiognomy.
Due to anthropic activities, such as habitat destruction for agriculture and excessive use of pesticides, bee populations have been drastically reduced (Michener, 2000).The impoverishment of pollinators in forest fragments can cause changes in gene flow within and among plant populations and directly affect natural regeneration (Engel et al., 1998).Bees are the main pollinators of most species of tropical forest trees (Michener, 2000), playing an important role in the conservation of plants and animals that depend on them.
This study aimed to deepen our knowledge about the richness and composition of flower-visiting bees in a fragment of semideciduous forest, analyzing the patterns of interactions between bees and visited plants.

Study area
This study was conducted in a fragment of semideciduous forest (18º 51' 36'' S and 48º 13' 53'' W) located in the legal reserve of the "Fazenda São José", in Uberlândia, Minas Gerais, Brazil.The fragment has 22 hectares.Its interior shows evidence of selective logging and cattle trails and the surrounding matrix is composed of monoculture eucalyptus plantations, annual crops and pastures (Prado Júnior et al., 2011).
The regional climate presents two distinct seasons, a rainy season that extends from October to March and a dry season, from April to September.The annual rainfall ranges from 1160 to 1460mm/year and the average annual temperature varies between 23 and 25°C, and is uniform throughout the year (Alves & Rosa, 2008).

Data collection
The study was conducted between October 2010 and September 2011, totalizing one year of sampling.Bees were collected with entomological nets during their visits to flowers, and eventually during flight, along a 200 metre transect at the forest edge, the most representative section in number of flowering plants.There were no collections within the fragment due to the difficulty in simultaneously observing and collecting, since the flowers were several feet high, preferably in the canopy.
Bees were collected from 8:00 to 14:30, the time of highest foraging activity.The collection effort involved the work of two collectors during 30 minutes every hour, totalizing 84 hours of sampling.At each time, the temperature and relative humidity were recorded using a digital thermo-hygrometer.
Only the composition of bee species was sampled.The abundance of these insects was not quantified.Bees were sacrificed in a chamber with ethyl acetate.After undergoing pinpricking, they were deposited in the "Coleção do Museu de Biodiversidade do Cerrado (MBC)", located in the "Laboratório de Ecologia e Comportamento de Abelhas (LECA)" of the "Instituto de Biologia (INBIO)" at the "Universidade Federal de Uberlândia (UFU)".
The plants visited by bees were marked, and three flowering branches of each type were collected and deposited in the "Herbarium Uberlandense (HUFU)" of UFU.Plants and bees were subsequently identified using identification keys.We had the help from experts, when necessary.

Data Analysis
To evaluate sampling sufficiency, we elaborated the rarefaction curve using the program EstimateS 8.2.0 (Colwell, 2006).The nonparametric estimator Jack 1 was used to verify if the species richness value found corresponded to the estimated value.This estimator was chosen because it is one of the most accurate richness estimators (Palmer, 1990;Krebs, 1999).
Bee-plant interactions were represented in a bipartite graph (network) generated by the bipartite package (Dormann et al., 2008) in R 3.0.1 software (R Development Core Team -http://www.r-project.org).The centrality of species was calculated by the program Pajek (Program for Large Network Analysis) (Batagelj & Mrvar, 1998).
The nestedness index NODF (Nestedness metric based on Overlap and Decreasing Fill) was calculated using Aninhado 3.0.3software (Guimarães & Guimarães, 2006).We used a null model which assumes that probability of each cell being occupied is the average of probabilities of occupancy of its row and column.Biologically, this means that probability of drawing an interaction is proportional to the level of generalization (degree) of both the animal and the plant species (Bascompte et al., 2003).The higher the value of this index, the greater the nestedness network (Almeida-Neto et al., 2008).Nested networks are characterized by interactions between generalists, specialists interacting with generalists and absent or rare interactions between specialists (Guimarães et al., 2006).
For the calculation of centrality, the Index of Betweenness Centrality was chosen.This index shows the importance of a species as a connector between different parts of the network and evaluates the contribution of each species to its cohesion (Borgatti & Everrett, 2006).

Results
Throughout the study period, 74 bee species distributed in five families were collected (Table 1).Four species were caught only in flight (Anthidioctes megachiloides, Augochlora sp.5, Eulaema nigrita and Lorocanthidium sp.).Apidae was the most represented family with 48 species, followed by Halictidae and Megachilidae.Andrenidae and Colletidae were the less representative families.Meliponina was the group with the highest richness, with 15 species sampled.The rarefaction curve showed no tendency to stabilization.The value found for the richness estimator Jack 1 was 114.51, suggesting that approximately 65% of the species in the area were sampled (Figure 1).Bees visited flowers of 25 plant species belonging to 14 families (Supplementary Material).The most representative families were Sapindaceae, with 4 species of visited plants, followed by Fabaceae and Bignoniaceae, with 3 species each.
The network of interactions consisted of 70 species of bees and 25 species of plants and presented a nested pattern, NODF = 10.97 (P = 0.03) (Figure 2).Merremia macrocalyx  was the plant with the highest betweenness centrality index, followed by Coccoloba mollis and Banisteriopsis argyrophylla (Table 2).Bees that had the highest centrality were Apis mellifera, Paratrigona lineata and Paratetrapedia lugubris, respectively (Table 2).
Bees presented the highest richness in May and September, both with 19 species (Figure 3).The greatest richness of flowering plants visited by bees occurred in April, with nine species (Figure 3).

Discussion
The network was essential for the identification of keyspecies in the studied area.The nestedness pattern found in this study corroborates other researches in other types of vegetation.
One of the most important factors associated with the richness of bees is environmental heterogeneity.The more varied the floristic composition of the site, the greater the possibility of available niches providing a larger number of species living in the same area (Andena et al., 2009).Regarding bee fauna in any particular location, the greater the diversity of plants, the greater the variety of bees, since they are their potential pollinators (Michener, 2000).Thus, a high diversity of plants contributes to a high diversity of pollinators, which are one of the key components responsible for the maintenance of plant diversity (Michener, 2000).
For the maintenance of potential pollinators in an area, the preservation of nesting sites for bees, such as rotten tree trunks and bounds in the case of solitary bees are essential.Trees with hollows and abandoned ant and termite nests should be preserved to be used for social bee nesting sites (Cortopassi-Laurino et al., 2009).Thus, the conservation of the studied area subject of this research is extremely relevant.Its significance lies in the fact that it is a fragment located near an urban area that maintains important and diverse plant species, offering a possible refuge, and providing food resources for many bee species.
Regarding the bee richness, it was expected that the rarefaction curve was not stabilized, since samplings including all the species present in an area is not common in insect inventories due to the high diversity of this group (Gotelli & Colwell, 2001;Brosi et al., 2007).The extreme incidence of the Apidae family seems to be a characteristic pattern in the Brazilian Cerrado biome, since similar results have been found in other surveys of different types of vegetation in the Cerrado biome (Carvalho & Bego, 1996;Antonini & Martins, 2003;Anacleto & Marchini, 2005;Andena et al., 2012).
Apidae is one of the most diverse families widely distributed in Brazil and worldwide.It occurs in different biomes and under different environmental characteristics (Michener, 2000), which may explain the major representativeness of this family in this study and in others.
The Halictidae family, the second most representative in this study, has a worldwide distribution.However, it is more diverse in temperate regions, despite possessing some unique Neotropical genera (Michener, 1979).In ecosystems with disturbances, there is a trend towards increased species of Halictidae (Roubik, 1989), which may indicate that the studied semideciduous forest is a disturbed area.
The Andrenidae and Colletidae families had low occurrence in this study, and are poorly represented in the Neotropical region (Silveira & Campos, 1995), which may have contributed to the results.Colletidae is an Australian family, although some genera occur in South America (Silveira et al., 2002).Silveira & Campos (1995) found that bees of the Cerrado biome compared to the fauna from other Brazilian biomes are characterized by the high representativeness of the tribes that collect oil (Centridini, Tapinotaspidini and Tetrapediini).For example, the species Centris tarsata, Epicharis flava, Paratetrapedia flaveola, Tetrapedia diversipes, and bee species of the subtribe Meliponina, have high representativeness, as we verified in this study.
The high centrality of Apis mellifera shows that this species is essential for the maintenance of the interactions network studied.Apis mellifera was the species that had the broadest niche, as it interacted with the largest number of plant species in the present study.Similar results were found in an area of Cerrado biome in the state of São Paulo (Andena et al., 2012) and in a review of inventories carried out in several regions of Brazil (Kleinert & Giannini, 2012).This fact may be related to its long daily and annual periods of foraging, high population density and sophisticated communication system (Roubik, 1989), allowing for a large number of plant species to be visited by these bees, resulting in a broader niche (Andena et al., 2012).A. mellifera is an exotic species scattered throughout various biomes, well adapted to different climatic conditions and presents generalist behaviour (Kleinert & Giannini, 2012).It is among the most important pollinators of natural environments and crops (Potts et al., 2010).
Amidst the most generalist bees sampled in this study, there were several species of Meliponina, such as Paratrigona lineata, Tetragona clavipes, Scaptotrigona depilis and Trigona spinipes.Other surveys carried out in the Cerrado biome have also reported large amplitude niches for some of these species (Antonini & Martins, 2003;Nogueira-Ferreira & Augusto, 2007).The niche size may be related to their eusocial group behavior, perenniality of their colonies and generalized foraging and recruitment habits (Roubik, 1989).In general, social species are more generalists and, therefore, have broader niches than solitary bees (Biesmeijer et al., 2005).
The nestedness of the network observed in this study means that plants with few interactions are associated with generalist animals.However, specialist animals are associated with plants with many interactions, and generalists of one group interact with generalists of the other group, forming a dense core of interactions (Lewinsohn et al., 2006).Nested networks of bee-plant interactions were also found in studies in the Caatinga in the state of Bahia (Pigozzo & Viana, 2010) and in another area of cerrado sensu stricto in the State of São Paulo (Andena et al., 2012).
Nested networks are thoroughly cohesive, present a dense mass of interactions that extend throughout the community.They also have heterogeneity in their distribution of connections and possess possible alternative routes in response to environmental perturbations (Bascompte et al. 2003;Bascompte & Jordano, 2007).Thus, nested networks are asymmetric, with generalist species interacting with specialists, providing pathways for rare species to withstand environmental adversities and alternative routes for system responses to perturbations, such as deletion of a mutualist (Jordano, 1987).These characteristics exert fundamental influence on the network stability (Bascompte et al., 2006;Santos et al., 2010) and maintenance of biodiversity (Bascompte et al., 2006).Thus, the network of bee-plant interactions found in this study has important properties for the maintenance of the community of bees and plants.
In pollination systems, the most generalist species are usually key species in the network (Martín González et al., 2010).These species are vital to the network structure, functioning and resilience, playing a pivotal role in community cohesion (Martín González et al., 2010;Kleinert & Giannini, 2012).Not all nodes (species) are equally important for the dynamics and stability of the system (Jordán, 2009;Martín González et al., 2010).The importance of a node can be quantified by centrality indices (Freeman, 1979).In the present study, the plants that had the highest centrality indices compared to other species were Merremia macrocalyx, Coccoloba mollis and Banisteriopsis argyrophylla, which were also the plant species visited by more bee species.These plants are essential to the structure and stability of the network (Martín González et al., 2010), in other words, they are indispensable to the maintenance of the community, and suppression of these species can quickly affect other species (Jordán, 2006).
M. macrocalyx presents the highest centrality index and can be considered a key species in the obtained network of interactions.It is extremely relevant for conservation in the studied fragment since it contributes to the maintenance of the existing bee community.As a climbing plant, M. macrocalyx is a substantial element in the structure of the forest, helping maintain optimal microclimate conditions for germination and contributes to increased resistance to wind on the edges of the fragment (Engel et al., 1998).The considerable number of bee species foraging on M. macrocalyx observed in this study may be associated with the supply of available resources offered by its flowers, as corroborated by Neves et al. (2006).They reported that pollen remains available from the time of anthesis, early in the morning, and decreases after 15h.Nectar secretion is intensive in the morning with a decreasing tendency during the day.
Coccoloba mollis, the second most important species in our study, has inflorescences in terminal panicles, with many small and fragrant flowers with nectaries (Lorenzi, 1992).According to Lorenzi (1992), the flowers of this species are considered apicultural, which explains the high number of bee species collected visiting its flowers.The greatest richness of bees occurred in May and September coinciding with the flowering periods of M. macrocalyx and C. mollis, respectively, which attracted a large proportion of species sampled in these months.
The network of bee-plant interactions presented is nested.This aspect can be positive in case of disturbances since nested networks offer alternative routes for the maintenance of rare species.M. macrocalyx can be considered a key species in the studied area since it attracted a large number of bee species and presented the highest centrality in the network of interactions.This plant species is essential to the cohesion of the network as a whole, and greatly relevant to the conservation of the Cerrado biome.

Fig 1 .
Fig 1. Rarefaction curve representing the cumulative number of species sampled in function of the quantity of samples in Semideciduous Forest of "Fazenda São José", Uberlândia-MG, in 2010 and 2011.

Fig 2 .
Fig 2. Bipartite graph of bee-plant interaction's network sampled in the Semideciduous Forest of "Fazenda São José", Uberlândia-MG, in 2010 and 2011.Bees are represented on the left side of the graph and plants on the right side.

Fig 3 .
Fig 3. Richness of bees and plants and averages of abiotic data collected throughout the year of study at "Fazenda São José", Uberlândia-MG.(A) Average temperature.(B) Average relative humidity.

Table 2 .
Values of the Betwenness Centrality Index for species of the network of bee-plant interaction sampled in the Semidecidous Forest of "Fazenda São José", Uberlândia-MG, in 2010 and 2011.
Note: Bee species that were not presented in the table obtained index equal to zero.