Introduction
The conservation status of an ecosystem depends mainly on its integrity (Caniani et al., Reference Caniani, Labella, Lioi, Mancini and Masi2016). Environmental integrity and its functional ecological processes are under serious threat from habitat fragmentation and other anthropogenic impacts (Haddad et al., Reference Haddad, Brudvig, Clobert, Davies, Gonzalez, Holt, Lovejoy, Sexton, Austin, Collins, Cook, Damschen, Ewers, Foster, Jenkins, King, Laurence, Levey, Margules, Melbourne, Nicholls, Orrock, Song and Townshend2015; Caniani et al., Reference Caniani, Labella, Lioi, Mancini and Masi2016). The current scenario is a landscape composed of fragments with varying degrees of human intervention, including remnants of the original vegetation and crop fields (Madeira et al. Reference Madeira, Tscharntke, Elek, Kormann, Pons, Rosch, Samu, Scherber and Batáry2016). Human intervention may change communities structure and composition, affect the identity and strength of interactions between species (Tylianakis et al. Reference Tylianakis, Tscharntke and Lewis2006), limit the dispersion of animals that depend on the original habitat and confine them to fragmented remnants of natural vegetation (Giubbina et al., Reference Giubbina, Martensen and Ribeiro2018) potentially resulting in changes in interaction network structure (Tylianakis et al. Reference Tylianakis, Tscharntke and Lewis2006).
Arthropods are the ones most affected by environmental fragmentation, especially those that act as natural enemies in food webs (Grab et al., Reference Grab, Danforth, Poveda and Loeb2018). Solitary bees and wasps are essential components of arthropod communities and are good models for environmental impact studies (Tscharntke et al., Reference Tscharntke, Gathmann and Steffan-Dewenter1998). These insects provide ecological services by acting as pollinators or natural enemies, directly responsible for crop pollination and biological control herbivores (Batra, Reference Batra1984; O’Neill, Reference O’Neill2001; Hoehn et al., Reference Hoehn, Tscharntke, Tylianakis and Dewenter-Steffan2008). The action of natural enemies is essential in maintaining ecosystem balance through an intense regulatory effect on the abundance of herbivorous insects (Sanders et al., Reference Sanders, Kehoe, Cruse, van Veen and Gaston2018; Vidal and Murphy, Reference Vidal and Murphy2018). The study of the effects of natural enemies as regulators allows us to understand the dynamics involved in multitrophic networks (Robinson et al., Reference Robinson, Inouye, Ogilvie and Mooney2017), plant-pollinator and host-natural enemy interactions (Penczykowski et al., Reference Penczykowski, Laine and Koskella2016). The understanding of these dynamics may grant access to the status of an ecosystem, by the assessment of interaction networks structure (Pocock et al., Reference Pocock, Evans, Fontaine, Harvey, Julliard, McLaughlin, Silvertown, Tamaddoni-Nezhad, White and Bohan2016).
Host-natural enemy interactions have their importance for the assessment of ecological networks structure, as organisms with parasitic habits act as connectors of trophic webs (Lafferty et al., Reference Lafferty, Dobson and Kuris2006). Interaction networks allow us to find relevant information such as interconnected species diversity, the way the interactions between species are structured and the stability of these interactions (Pocock et al., Reference Pocock, Evans, Fontaine, Harvey, Julliard, McLaughlin, Silvertown, Tamaddoni-Nezhad, White and Bohan2016). The metrics commonly analysed in interaction network studies are nestedness (sensu Joppa et al. Reference Joppa, Montoya, Sanderson and Pimm2010), modularity (sensu Tylianakis et al., Reference Tylianakis, Laliberte, Nielsen and Bascompte2010) and specialization degree (sensu Blüthgen et al. Reference Blüthgen, Menzel and Blüthgen2006). These metrics relate to network structure and, therefore, to which species are most vulnerable to eventual impacts that may occur (Bascompte et al., Reference Bascompte, Jordano and Olesen2006). The degree of specialization or generalization an interaction network reflects the number of connections that a node has (i.e., host and natural enemies species, in this work) and/or the ability of the connected species to establish the interaction (Pocock et al., Reference Pocock, Evans, Fontaine, Harvey, Julliard, McLaughlin, Silvertown, Tamaddoni-Nezhad, White and Bohan2016).
Host-natural enemy interaction networks tend to be specialized and modular (Bellay et al., Reference Bellay, Oliveira, Almeida-Neto, Abdallah, Azevedo, Takemoto and Luque2015; Bellay et al., Reference Bellay, Oda, Campião, Yamada, Takemoto, Oliveira, Dáttilo and Rico-Gray2018). Some studies with natural environments show evidence for these patterns. For example, interaction networks between marine fishes and their metazoan parasites present high degrees of specialization of the host-parasite interactions and consequently, higher degrees of modularity (Bellay et al., Reference Bellay, Lima, Takemoto and Luque2011). Studies in altered environments, however, present different patterns, finding networks with more generalized patterns (Matos et al., Reference Matos, Sousa-Souto, Almeida and Teodoro2012; Stangler et al., Reference Stangler, Hanson and Steffan-Dewenter2015). Studies with solitary bees and wasps had also presented evidence for specialized patterns (Lima et al., Reference Lima, Moure-Oliveira and Garofalo2018; Rocha-Filho et al., Reference Rocha-Filho, Moure-Oliveira, Carvalho, Frantine-Silva and Augusto2019). These networks, when occurring in natural environments, are composed by fewer species and modules (Vázquez et al., Reference Vázquez, Poulin, Krasnov and Shenbrot2005).
In addition to the analysis of interaction networks, the diversity indexes are commonly used to verify environmental quality. Areas with few species tend to have lower diversity indexes (i.e., Shannon index) (Lima et al., Reference Lima, Moure-Oliveira and Garofalo2018). On the other hand, areas with generalist species support a higher number of species and consequently higher diversity index values as evidenced in mutualistic networks (Encinas-Viso et al., Reference Encinas-Viso, Revilla and Etienne2012). Meanwhile, simplified areas (i.e., monocultures) have less species and low diversity indexes (Tylianakis et al., Reference Tylianakis, Tscharntke and Lewis2006). Therefore, the structure of the network depends on the interaction type and environmental complexity, among other factors (Bellay et al., Reference Bellay, Oda, Campião, Yamada, Takemoto, Oliveira, Dáttilo and Rico-Gray2018).
Here we analysed host-natural enemies interaction network in a community of solitary bees and wasps in a fragmented landscape composed of a Eucalyptus plantation (considered here as higher anthropic perturbation), a Cerrado area (medium perturbation) and a Riparian forest (lesser perturbation). Once disturbed areas such as monoculture areas usually have a lower environmental complexity (i.e., a measure anthropic impact level), it was expected that networks associated with simplified areas (e.g., Eucalyptus plantation) would reflect the degree of human interference, showing lower diversity and lower degrees of network specialization, as generalist species are generally more resistant to environmental change. On the other hand, natural environments (e.g., Cerrado and Riparian forest areas) tend to maintain the integrity of interactions, in this case, highly specialized host-natural enemy interactions. Hence, we aimed to address whether there is a difference in species composition and diversity of natural enemies’ species between areas with different levels of human impact and whether such impact results in networks with different structures in a solitary hymenopteran community.
Study site
Fieldwork was conducted in an area under restoration located in Sao Carlos, SP, where three fragments could be found: (i) a fragment of Cerrado (−21.972904, −47.881649); (ii) a fragment with Eucalyptus plantation (−21.969998, −47.875637); and (iii) a fragment of Riparian forest along the Espraiado stream (−21.980915, −47.873918). The Cerrado and Riparian forest areas are fragments of secondary vegetation that had been through restoration process for some decades. We selected three sites, each of them inside one of each fragment, with one kilometre apart from each other, where we placed trap nests, assuming a gradient of impact level: (i) a Cerrado fragment (−21.972904, −47.881649), with a record of conservation measures; (ii) a riparian forest (−21.980915, −47.873918), a small fragment with a stream and surrounding dense vegetation; (iii) an abandoned Eucalyptus plantation (−21.969998, −47.875637), where some Cerrado plants begin to grow. The climate of the region varies from tropical wet and dry to humid subtropical according to Köppen’s system, and the vegetation consists predominantly of Cerrado, semi-deciduous and riparian forests, with regeneration areas (Soares et al. Reference Soares, Silva and Lima2003). Data on mean monthly temperature and rainfall were obtained from the São Carlos Station (OMM code: 86845, ‘Instituto Nacional de Metereologia’ – http://www.inmet.gov.br/portal/).
Methods
In each site, we set trap nests for hymenopterans that nest in pre-existing cavities (Westerfelt et al., Reference Westerfelt, Widenfalk, Lindelow, Gustafsson and Weslien2015). Each trap consisted of four trunks of similar size, arranged on a 1.5m wide wooden platform. Each trunk had 34 holes of a given diameter (4, 6, 8 or 10 mm), totalling 136 cavities in each site (Fig 1). The trap nests were made of black cardboard (MacIvor, Reference MacIvor2017). Pieces of about 10 cm long were wrapped into tubes of 4, 6, 8 and 10 mm in diameter and inserted into holes drilled in trunks. The nests were examined biweekly, from October 2017 to October 2018. The occupied nests were removed and replaced with new cardboard tubes of the same diameter.
Figure 1. Representation of interaction network obtained to (a) Cerrado fragment, (b) Eucalyptus plantation and (c) Riparian forest. The modules are represented by different shades in a greyscale. The thickness of edges represents quantity of interaction between the species. Natural enemies are represented by square nodes and hosts by circle nodes. For a list with the codes used in the networks, please see Appendices Table 2.
Material examined
The collected nests were brought to the laboratory, stored in transparent glass tubes plugged with cotton and maintained under controlled temperature (28 ± 2oC), (Gazola and Garofalo, Reference Gazola and Garofalo2009). The nests were observed daily, and emerged adults were collected and euthanized in absolute ethanol. All animals were deposited as vouchers in the Laboratory of Ecological Studies on Ethology and Evolution (LESTES) at the Federal University of São Carlos, Brazil. The material was identified with the aid of a magnifying stereomicroscope Leica MZ95 and consolidated identification keys for each group (Fernández and Sharkey, Reference Fernandéz and Sharkey2006), as well as expert help for some of the species collected.
Data analysis
To address whether there is a difference in species composition and diversity between the sampled sites, we built a matrix with the abundance of each species of natural enemies for each area and used the Past 3.21 software (Hammer et al., Reference Hammer, Harper and Ryan2001) to calculate the Shannon-Winner index (Hʼ) (Poole, Reference Poole1974) in a paired t-test to verify if there is a significant difference between index values among areas (Hutcheson, Reference Hutcheson1970). The host species were not included in these analyses because the diversity of natural enemies is determined by host diversity, which could generate collinearity of data and a false predictive power (Gazola and Garofalo, Reference Gazola and Garofalo2009; Lima et al., Reference Lima, Moure-Oliveira and Garofalo2018). Since species diversity affects coexistence and, consequently, trophic interactions among them (Kéfi et al. Reference Kéfi, Miele, Wieters, Navarrete and Berlow2016; Ohlmann et al. Reference Ohlmann, Miele, Dray, Chalmandrier, O’Connor and Thuiller2019), in this study we consider the relevance of a diversity index to better compare the networks built.
To verify the structure of the interaction network host-natural enemies, we built three matrices of weighted adjacency containing the amount of host cells parasitized by each natural enemy, one matrix for each study site. Hence, we calculated the degree of network specialization, the degree of nestedness and the modularity. The degree of network specialization (H2’ index) is a two-dimensional measurement ranging from 0 (extreme generalization) to 1 (extreme specialization) (sensu Blüthgen et al., Reference Blüthgen, Menzel and Blüthgen2006). We estimated the significance of this index using the Monte Carlo procedure with null model for 1,000 randomizations. These analyses were performed using the bipartite package (Dormann et al., Reference Dormann, Gruber and Fründ2008) of the R 3.5.1 software (R Core Team, 2018). To verify the presence of nestedness in the networks, we used the NODF index (Nestedness metric based on Overlap and Decreasing Fill – Almeida-Neto et al., Reference Almeida-Neto, Guimaraes, Loyota and Ulrich2008) using incidence matrices (presence and absence). This index ranges from 0 (no nestedness) to 100 (perfect nestedness). We estimated the significance of NODF index using the Monte Carlo procedure with 1,000 randomizations to the null model Ce, which keeps the total value of fixed lines during randomizations (see Guimarães and Guimarães, Reference Guimarães and Guimarães2006). We used the software ANINHADO 3.0 (Guimarães and Guimarães, Reference Guimarães and Guimarães2006) to address the NODF values and their significance. We calculated modularity of the networks using the ComputeModules from the bipartite package of the R 3.5.1 software, which uses a QuanBioMo (Q) algorithm for quantitative data matrices (sensu Dormann and Strauss, Reference Dormann and Strauss2014). We tested the modularity through null models with 1,000 randomizations using R2d method, generating ZQ values equivalent to the z score of a normal distribution. ZQ values above 2.0 represent significant modularity (sensu Dormann and Strauss Reference Dormann and Strauss2014). The modules formed in this analysis were represented in the network by using different shades in a greyscale.
We use additional metrics to verify species specialization inside interaction networks. From the three matrices, we calculated the degree of specialization (d’ index) of the parasitoid species using the package bipartite of the R 3.5.1 program. The index d’ is a measure of the normalized Kullback-Leibler distance measuring the specialization of a sort based on the frequency of the total number of network interactions. According to Blüthgen et al. (Reference Blüthgen, Menzel and Blüthgen2006), this index ranges from 0 to 1, indicating extremely generalization and specialization, respectively. The strength of an interaction (i.e., the frequency that natural enemies parasitize hosts; prevalence of natural enemies) is indicated by graphic representations, where the force is indicated by the thickness of lines between taxa (Berlow et al., Reference Berlow, Neutel, Cohen, Ruiter, Ebenman, Emmerson, Fox, Jansen, Iwan, Kokkoris and Logofet2004).
Results
In total, 1,586 individuals emerged from 234 trap nests collected. There was emergence of natural enemies in 80 of these nests, which were used in the present study. We sampled 15 species of natural enemies distributed among 12 families and five different insect orders, which were Coleoptera, Diptera, Neuroptera, Lepidoptera and Hymenoptera. We sampled natural enemies of Apidae, Chrysididae, Eulophidae, Gasteruptiidae, Ichneumonidae, Leucospidae, Megachilidae and Mutillidae. For each sampled area, four different host species have been identified. In all cases where there was parasitism, it was possible to determine the host and the number of affected cells. In the Eucalyptus site, we found 11 species of natural enemies in 44 trap nests, while we sampled nine species of natural enemies in 33 nests and five species of natural enemies in ten trap nests, in Cerrado and the Riparian forest, respectively (see Appendices Table 1).
The diversity indexes found were 2.07 for Eucalyptus plantation, 1.75 for the Cerrado fragment and 1.44 for the riparian forest, and there is difference only between the index values for Eucalyptus plantation and riparian forest sites (t-test, t = 2.83, df = 27.168; p = 0.01). The values of diversity indexes between Cerrado and Eucalyptus, and Cerrado and Riparian forest were not different (t-test, t = −1.82, df = 99.209; p = 0.07 and t-test, t = 1.38, df = 26.513; p = 0.18, respectively).
The three networks of interactions evaluated were modular and non-nested, showing high degrees of specialization (H2ʼ) to Cerrado and Riparian forest (Table 1). We construct a graphic representation showing the strength of the interactions between the host and its natural enemy (Figure 1).
Table 1. Nestedness (NODF), modularity with interaction strengths (Q) with z-score values (zQ), number of modules (Mod), degree of network specialization (H2’) calculated to natural enemies in a fragment of Cerrado in restoration, a Eucalyptus plantation and a Riparian forest, sampled between October 2017 and October 2018, in São Carlos, SP, Brazil.
Meaning values to *p<0,05 and NS to non-significant values.
The degree of species specialization (d’ index) calculated for each natural enemy in each of the areas is shown in Table 2. The riparian forest had two species with maximum specialization degree (d’ = 1.00). The Cerrado fragment had four species with specialization degree higher than (d’ = 0.5), one of them with maximum degree. The Eucalyptus plantation had lower values of specialization degree, with the higher value found being (d’ = 0.54). The natural enemy Chrysis sp. (Hymenoptera: Chrysididae) has a higher degree of specialization in Trypoxylon sp2 (Hymenoptera: Crabronidae) in the area of Cerrado (dʼ = 0.75). In Eucalyptus plantation area, the higher degree of specialization occurs between Anthrax sp1 (Diptera: Bombyliidae) and Anthidiini bees (Hymenoptera: Megachilidae) (dʼ = 1.00). The interactions observed in the riparian forest, the maximum value of dependence (dʼ = 1.00), was observed in the interactions between Coelioxoides sp. (Hymenoptera: Apidae) and Tetrapedia diversipes Klug (Hymenoptera: Apidae), Chrysis sp. (Hymenoptera: Chrysididae) and Pseudodynerus sp (Hymenoptera: Vespidae) and between Ichneumonidae and Zethus sp. (Hymenoptera: Vespidae).
Table 2. Specialization degrees (d’) of natural enemies sampled at three spots in São Carlos, SP, Brazil, between October 2017 to October 2018. Values higher than (d’= 0.5) are bolded.
Discussion
Our results partially corroborate our initial hypothesis. We show the area with higher human interference, a Eucalyptus plantation, had higher values of diversity of natural enemies, greater number of species in the interaction network, lesser degree of specialization in the interaction and lesser modularity, than Cerrado and Riparian forest environments, respectively. Thus, the gradient from highest to lowest degree of anthropic perturbation was evidenced in the species diversity index, the size of the interaction network and the specialization indexes of the host-natural enemy network. The low degree of nestedness and lack of significance of this index to all the three sampled areas is an indicative of a specialized pattern of these networks.
Recent research has shown that networks involving hosts and parasites have high specialization indexes (Pereira-Peixoto et al. Reference Pereira-Peixoto, Pufal, Staab, Martins Feitosa and Klein2016; Araújo et al., Reference Araujo, Fagundes and Antonini2018; Lima et al., Reference Lima, Moure-Oliveira and Garofalo2018). This pattern is due to the nature of these interactions, which involves specific mutual adaptations because of coevolution between these organisms (Gómez et al., Reference Gómez, Ashby and Buckling2015). Here, the specialist pattern was evidenced in all sampled areas. In this case, there was either a lack of nestedness or values of specialization that range from medium values (0.47 in the Eucalyptus plantation) to 1.0 in the riparian forest. The medium values of specialization of the network found for the Eucalyptus plantation can be related to the specialization degree of the species involved. Most of the diversity (10 out of 11 species) presented a tendency of generalization, with d’ values under 0.5. Generalization favours interactions between species and contributes to a higher diversity (Vázquez et al. Reference Vázquez, Poulin, Krasnov and Shenbrot2005), represented here by Shannon index, which was higher in the Eucalyptus plantation. Generalist species are less sensitive to land use (Holzschuh et al., Reference Holzschuh, Steffan-Dewenter and Tscharntke2010) and thus are more likely to survive in less structured habitats (e.g., monocultures) (Pereira-Peixoto et al., Reference Pereira-Peixoto, Pufal, Staab, Martins Feitosa and Klein2016). Therefore, higher values of species diversity of natural enemies found in the Eucalyptus plantation may reflect a generalization of the interactions established between the hosts and parasitoids and not necessarily anthropogenic impact.
On the other hand, the areas of Cerrado and riparian forest (i.e., lower anthropic impact) presented lower species diversity of natural enemies, but higher values of specialization of networks and species. Moreover, networks also exhibited more interaction modules. These areas may have higher structural complexity of the habitat – places with higher amount and diversity of niches. The complexity of an environment is positively correlated with the niche diversity (Mougi and Kondoh Reference Mougi and Kondoh2016), contributing to species diversity (Tylianakis et al., Reference Tylianakis, Tscharntke and Lewis2006). Nevertheless, one must consider that environmental complexity must primarily favour the host when assessing host-parasite interactions, because parasitism depends on host diversity (Lagrue e Poulin, Reference Lagrue and Poulin2015).
In the Cerrado area, Chrysis (Hymenoptera: Chrysididae) and Coelioxys (Hymenoptera: Megachilidae) species presented higher specialization index (d’) among hosts. Several authors showed that cleptoparasitoid Coelioxoides (Hymenoptera: Apidae) usually steal nests from related lineages, such as Tetrapedia diversipes Klug, 1810 (Hymenoptera: Apidae) (Araújo et al., Reference Araujo, Lourenço and Raw2016; Rocha-Filho et al., Reference Rocha-Filho, Rabelo, Augusto and Garófalo2017; Lima et al., Reference Lima, Moure-Oliveira and Garofalo2018). In this study, flies of the genus Anthrax (Diptera: Bombyliidae) presented lower values of specialization, which agrees with the literature concerning this natural enemy (Krombein, Reference Krombein1967; Gazola and Garófalo, Reference Gazola and Garofalo2009; Mesquita and Augusto, Reference Mesquita and Augusto2011). Tetrapedia diversipes and Centris analis (Fabricius, 1804) (Hymenoptera: Apidae) bees registered the higher rates of parasitism. This was probably consequence of the higher abundance of nests built by these host species. These organisms are commonly found in higher abundance in studies with trap nests (Alves-dos-Santos, Reference Alves-dos-Santos2003; Buschini and Wolff, Reference Buschini and Wolff2006; Gazola and Garófalo, Reference Gazola and Garofalo2009; Araújo et al., Reference Araujo, Lourenço and Raw2016; Araújo et al., Reference Araujo, Fagundes and Antonini2018).
All these aspects demonstrate the complexity of host-parasite interactions and ecological interactions. Land use and anthropogenic impact also affect diversity and abundance of trap nesting insects (Albrecht et al., Reference Albrecht, Duelli, Schmid and Müller2007). These impacts may negatively influence trophic interactions (e.g., parasitism), even more intensely in fragmented and isolated fragments (Klein et al., Reference Klein, Steffan-Dewenter and Tscharntke2006). To assess these interactions in the context of conservation demands a macrovision of ecological communities and a microvision of natural history of the species involved in the connections. In other words, the task of interpreting the results of network analyses with metrics is even more meticulous.
Conclusion
In conclusion, this study showed the complexity of the ecological interactions in a fragmented landscape, through building and analysis of host-natural enemy interaction networks. The sampled area constitutes a region with fragments in recent process of restoration. This history can be determinant to the diversity of solitary wasps and bees sampled and the ways they explore resources, being generalists or specialists. Therefore, the history of land use and the fragmentation process must be accounted for in areas with distinct gradients of anthropogenic impacts to more precisely understand its effects.
Acknowledgments
All authors were responsible for study design, analyses and writing. We thank Felipe Varussa for helping with species identification.
Financial Support
RGF thanks the National Council for Scientific and Technological Development (CNPq) for a productivity grant (process 307836/2019-3). GGS thank CNPq for financial support.
Appendices
Table 1. Natural enemies and their hosts, sampled in trap nests collected between October 2017 and October 2018 in three areas of São Carlos, SP. NN is Number of Nests and NC is Number of Cells.
Table 2. List of codes to identify species in the network representations.
