Plant phenological asynchrony and community structure of gall‐inducing insects associated with a tropical tree species

Abstract The dynamics of occurrence of target organs in plant populations produces windows of opportunity that directly and indirectly affect the structure of herbivore communities. However, mechanisms that drive herbivore specialization between resource patches are still poorly known. In this study, we tested three hypotheses related to variation in host plant phenology and community structure (i.e., composition, richness, and abundance) of gall‐forming species: (a) plants with early leaf‐flushing in the season will have greater vegetative growth and high contents of secondary chemical compounds; (b) gall‐inducing insect community structure changes among temporary resource patches of the host; and (c) interspecific competition is a probable mechanism that drives gall‐inducing insect community structure on Copaifera langsdorffii. We monitored daily a total of 102 individuals of the super‐host C. langsdorffii from August 2012 to May 2013, to characterize the leaf flushing time of each host plant. The leaf flushing time had a positive relationship with the number of folioles per branch and a negative relationship with branch growth. We sampled a total of 4,906 galls belonging to 24 gall‐inducing insect species from 102 individuals of C. langsdorffii. In spite of some gall‐inducing species presented high abundance on early leaf‐flushing plants, direct and indirect effects of plant phenology on galling insect abundance was species dependent. At the community level, our study revealed that the quality and quantity of plant resources did not affect the richness and abundance of gall‐inducing insects associated with C. langsdorffii. However, the richness and composition of gall‐inducing species varied according to the variation in leaf flushing time of the host plant. The results of null model analysis showed that galls co‐occurrence on C. langsdorffii trees differ more than expected by chance and that interspecific competition can be one potential mechanism structuring this gall‐inducing insect community.


| INTRODUC TI ON
The resources obtained by plants are allocated to different functions such as growth, reproduction, and production of chemical compounds associated with defence against herbivorous insects (Costa, Reis-Júnior, Queiroz, Maia, & Fagundes, 2016;Herms & Mattson, 1992;Stamp, 2003). Variation in leaf flushing time within a plant population can play a relevant role in plant-herbivore interactions (Fagundes, 2014;Forister, 2005). Directly, plant phenology can affect herbivore actions using the host satiation mechanism or producing resource patches that vary over time (Aide, 1993;Fei, Gols, & Harvey, 2014;Souza & Fagundes, 2017;Van Schaik, Terborgh, & Wright, 1993). Indirectly, variation in leaf flushing time produces resource patches that change in quality for herbivores. For example, in seasonal environments, plants with early leaf-flushing can accumulate more resources during an extended vegetative stage, resulting in both a greater vegetative growth and accumulation of secondary compounds, which in turn, affect herbivores performance (Costa et al., 2016;Fagundes et al., 2016;Forister, 2005;Yukawa, 2000).
Temporal dynamics of occurrence of target organs in an asynchronous plant population produce windows of vulnerability in which different herbivore species can specialize (Fei et al., 2014;Fox, Waddell, Groeters, & Mousseau, 1997;Yukawa, 2000). In fact, plants with early-leaf flushing are usually more attacked by sessile herbivores such as gall-inducing insects, because those plants provide more resources and longer time for herbivore development (Forister, 2005;Yukawa & Akimoto, 2006). In contrast, herbivores with a short larval phase can better adapt to plants with late leaf flushing Singer & Parmesan, 2010). Different races of the same herbivore species can use temporary resource patches in different ways (Mopper, 2005). These findings suggest that specialization of gall-inducing insects on temporary resource patches can shape the structure of herbivore communities (i.e., composition, richness, and abundance of species) associated with a host plant species (Edmunds & Alstad, 1978;Egan & Ott, 2007;Mopper, 2005;Oliveira, Mendonça, Moreira, Lemos-Filho, & Isaias, 2013;Yukawa & Akimoto, 2006). Competition (Kaplan & Denno, 2007;Mopper, 2005), escape from natural enemies (Egan & Ott, 2007;Karban & Agrawal, 2002), and resource depletion (Revilla, Encinas-Viso, & Loreau, 2014) are possible biological factors used to explain the variation in the herbivore community structure among temporary resource patches of the host plant. However, the role of those biological factors in shaping the structure of natural communities has been little tested experimentally (see Gurevitch, Morrison, & Hedges, 2000;Oliveira et al., 2013;Pereira et al., 2014).
Gall-inducing insects exhibit at least three characteristics that make them excellent models to access the importance of interactions within and between different trophic levels in the organization of herbivore communities. First, the adult life time of specialized gall-inducing insects is relatively short and gall-inducing insects are extremely specialized on the host plant taxon and tissue exploited (Carneiro et al., 2009;Fagundes, Neves, & Fernandes, 2005). This characteristic leads to an intimate temporal relationship between the life cycle of gall-inducing insects and the occurrence of the target organ in the plant. Second, the galls are sessile and their formation is associated with an alteration in plant resource allocation, which occurs a few hours after oviposition (Höglund, 2014), with possible effects on the oviposition site selection by females of other galling insects (Cornelissen, Guimarães, Viana, & Silva, 2013). Third, the success of the gall-inducing insect also depends on the selection of the most vigorous organ of the plant (resource quality) and the number of oviposition sites (resource abundance) available for its development (Höglund, 2014;Price, 1991). Moreover, the performance of gall-inducing insects is also affected by the interaction between plant phenology and top-down forces (Fagundes et al., 2005;Hood & Ott, 2010;Johansson et al., 2015) as well as by biological interactions that occur within the same trophic levels (Cornelissen et al., 2013;Johansson et al., 2015).
Previous studies have shown that competition between endophagous insects as leaf miners and gall-inducing insects can arise when sites for oviposition and insect development are a limiting resource (Cornelissen et al., 2013). When experimental manipulations are impossible due to the nature of the study system, the comparison of the observed patterns of species occurrence with patterns that might be expected to occur by chance alone is a conventional approach to study competition between species (Cornelissen & Stiling, 2008;Morin, 2011). Many authors have used analysis of co-occurrences to show the role of competition on community organization of sedentary insects on host plant (e.g., Cornelissen et al., 2013;Kaplan & Denno, 2007;Tack, Ovaskainen, Harrison, & Roslin, 2009). Copaifera langsdorffii (Fabaceae) is a super-host species for gall-inducing insects (Sensu Veldtman & McGeoch, 2003) and exhibits substantial interplant variation in the timing of the leaf flushing (Fagundes, 2014). We used the system Copaifera langsdorffii-galling insects to test the following hypotheses: (a) plants with early leafflushing in the season will have greater vegetative growth and high concentrations of secondary chemical compounds; (b) gall-inducing community structure changes among temporary resource patches of the host and (c) the actual pattern of gall co-occurrence on host plants differ from that expected by chance.

| Study area
This study was carried out in a private reserve with an area of approximately 25 ha, located in Montes Claros, northern Minas Gerais, Brazil (16 o 40′26″S and 43 o 48′44″W). The soil of the study area is dystrophic and covered by trees and shrubs with distorted branches and thick bark, with height between 2 and 10 m (Souza, Solar, & Fagundes, 2015). The region is located in the transition between the biomes Cerrado and Caatinga of Brazil. Its climate is semiarid with a dry season from March to September and a rainy season from October to February. The average annual temperature is 23°C, and the rainfall is approximately 1,000 mm/year, mainly concentrated from November to January (Costa et al., 2016).

| The system
Copaifera langsdorffii Desf. (Fabaceae) is a tropical tree species that reaches up to 20 m in height in the Brazilian Cerrado. In the study area, these trees show complete deciduousness in the dry season (mainly from July to September) and leaf flushing occurs right after the fall of old leaves produced in the previous year. Flowering occurs from November to December and fruits ripe from August to September (Fagundes et al., 2013). Fruiting set is supra-annual, that is, there are years of intense fruiting set followed by years of low or no fruit production (Souza & Fagundes, 2017). C. langsdorffii has the most diverse fauna of gall-inducing insects in the Neotropics (24 species) and a broad diversity of free-living herbivore insects (Costa et al., 2016).

| Data collection
In March 2012, we selected and marked 102 individual adults of C. langsdorffii with a height that ranged between 8 and 12 m, abundant crown, and absence of lianas and parasitic plants. C. langsdorffii trees showed irregular distribution in the study area and the distance between trees varied from 7 to 45 m. We daily monitored all trees selected for the study, from August 2012 to May 2013 to determine the period of new leaf flushing of each individual plant.
In June 2013 (prior to leaf fall), we collected ten terminal branches (approximately 30 cm long) from each tree and in this way, we were able determine the relative temporal variation (days) in leaf flushing among individual trees in the population. We collected those branches from different parts of the tree crown, to obtain samples from the whole individual and avoid a bias caused by the effect of microhabitat on branch development (Costa, Fagundes, & Neves, 2010). We used those samples to quantify branch growth (mean internodes length), average number of folioles per branch, leaf phenol concentration, and intensity of attack by gall-inducing insects. We determined the growth of each branch by dividing the length of the main branch by its number of nodes. We identified the galls present in those branches following Costa et al. (2010).

| Phenol quantification
The leaf phenol content was determined with spectrophotometry using the Folin-Ciocalteu methodology (Swain & Hillis, 1959), and which included gallic acid as standard. Initially, leaf extract was prepared for each individual plant by diluting 0.5 g grinded leaves in one mL of methanol. The 0.5 g of leaves was obtained from 50 mature leaves collected from those ten branches in each plant according to the method mentioned above. Next, 500 μl of the extract was transferred to tubes containing 250 μl of Folin-Ciocalteu's reagent.
After waiting for ten minutes, 500 μl of a sodium carbonate solution (10% w/v) was added to the samples. The tubes were then allowed to stand at room temperature for 30 min before absorbance was measured at 743 nm. The concentration of polyphenol in the samples was derived from a standard curve of gallic acid ranging from 20, 30, 40, 60, and 80 μg/ml. Total phenolic content was expressed as gallic acid equivalents (GAE) in mg/g of dry leaf extracts.

| Statistical analysis
Generalized linear models (GLMs) were used to assess the effects of the variation in plant flushing time on branch growth, average number of folioles, and leaf phenol contents. The branch growth, the average number of folioles, or the leaf phenol content were considered as the response variables, and the relative variation in leaf flushing time among plants was considered the explanatory variable.
Considering that the response variables are continuous, we tested the models using a F test based on the Gaussian distribution. After those tests, the models were submitted to analysis of residuals to test their adequacy to statistical assumptions (Crawley, 2007).
To determine the effects of the variation in leaf flushing time, branch growth, leaf phenol contents, and average number of folioles per branch on the average abundance of each gall species, total richness, and mean of total abundance of galls (i.e., sum of all galls divided by number of branches sampled per plant), we performed generalized linear models (GLMs). In model constructions, the abundance of each gall-inducing insects as well as total richness and average abundance of gall-inducing were used as response variables. The explanatory variables were leaf phenol contents, branch growth, average number of folioles, leaf-flushing time, and the interaction among all explanatory variables. However, these tests were done only for gall-inducing insect species that were present at least in 25% of all sampled plants to avoid that high number of zero values mask the result of statistical analysis. We tested the models related to gall-inducing insect richness with a chi-squared test based on the Poisson distribution as they contained count data. To test the models of average abundance, we used a F test based on the Gaussian distribution, as they contained continuous data (Crawley, 2007). To control for the effect of the entry sequence of explanatory variables used in the models, we use a stepwise model selection analysis in the package MuMIn for R (Bartón, 2015). In this analysis, the selection of the most parsimonious model is based on the values of the Akaike Information Criterion (AIC). Finally, we submitted the most parsimonious model to an analysis of residues to test for the adequacy of the model to statistical assumptions (Crawley, 2007). In all analyses, we tested for the presence of influent points that could interfere with the results. We removed those influent points from significant models and ran the analyses again. All analyses were run in the program R (R Core Team, 2015).
To test whether the community composition of gall-inducing in- We used null models to compare the observed and random patterns of gall-inducing insect occurrence in all plants of the population. A C-score index (Stone & Roberts, 1990), calculated based on the matrices of presence/absence of galls per plant, was used to quantify the co-occurrence of galls. The null hypothesis was that the presence of one gall species in the plant does not influence the occurrence of other gall species in the same plant. Hence, when the value of the co-occurrence index calculated for the original matrix (observed C-score) was outside 95% of the values found in the distributions of randomized C-scores, the null hypothesis must be rejected. In that case, we infer that the distribution of gall-inducing species on plants is determined by interspecific biological interactions (Ribas & Schoereder, 2002). To test for variation in the distribution of observed and expected matrices, we used fixed-fixed models with 5,000 randomizations. All analyses were conducted in the R software (R Core Team, 2015).  (Table 1). In fact, the abundance of six species TA B L E 1 Minimum adequate models showing the effects of the explanatory variables (variation in leaf flushing time, branch growth, and leaf phenol contents) on the abundance of gall-inducing species associated with Copaifera langsdorffii F I G U R E 2 Relationship between gall-inducing insect abundance and variation in leaf flushing time (a,b), branch growth (c), and leaf phenol contents (d) of Copaifera langsdorffii plants was negatively related with leaf flushing time (Figure 2a,b), two gall-inducing species were positively related with branch growth (Figure 2c), two gall-inducing insects were negatively related with leaf phenol content (Figure 2d), and seven gall-inducing insects were not statistically affected by explanatory variables (Figure 3a,b). A total of nine gall-inducing species presented low frequency on sampled plants (<25%) and was not statistically analyzed (Figure 3c-e). The observed values of the C-score index did not fall within the 95% limits of frequency distribution of the 5,000 randomized matrices (p = 0.019, Figure 6). Thus, the null hypothesis was rejected and the hypothesis of biological mechanisms conditioning the gallinducing insect occurrence on C. langsdorffii must be accepted. Phenological asynchrony in plant populations can affect plantherbivore interactions directly by producing patches of temporally unpredictable resources (Fei et al., 2014;Souza & Fagundes, 2017) and indirectly by producing resources that vary in quality and quantity over time (Forister, 2005;Nord et al., 2011;Parachnowitsch et al., 2012). In a population perspective, early leaf-flushing plants usually produce more vigorous growth modules (Fox et al., 1997;Vitou, Skuhravá, Skuhravý, Scott, & Sheppard, 2008;Yukawa, 2000), which are preferentially attacked by herbivores (Price, 1991). In our study, of the 15 more frequent gall-inducing insect species that were statistically analyzed, eight gall species were directly affected by plant phenology or indirectly by plant quality (see Figure 2a (Faria & Fernandes, 2001;Price, 1991). Thus, the dispute for more adequate sites for oviposition is a possible factor that drives the organization of gall-inducing insect communities especially in super-host plants (Reitz & Trumble, 2002). This idea is in agreement with our results, because the richness and composition of galling insects were not affected by quality of the host plant, but these two aspects of the community changed as a function of leaf flushing time of the host plants.

| D ISCUSS I ON
In spite of many studies showing that the galling community Hence, our null models analysis suggests that co-occurrence of galls on C. langsdorffii trees differ more than expected by chance and interspecific competition can be one potential mechanism structuring this gall community.
In summary, our study showed that C. langsdorffii is a deciduous tree species with broad intrapopulation phenological asynchrony in leaf flushing time. In spite of some gall species presenting high abundance on early-flushing plants, direct and indirect effects of plant phenology on galling insect abundance was species dependent. Finally, the structure of galling insect community associated with C. langsdorffii change in function of leaf flushing time of host plant and interspecific competition seem be responsible by these gall community organization.

ACK N OWLED G M ENTS
The study was carried out with financial support from Fundação LGOL, RCFX, and MF formatted and submitted the manuscript.

CO N FLI C T O F I NTE R E S T
None declared.

DATA ACCE SS I B I LIT Y
The data supporting this study and all analyses are available at doi: 10.5061/dryad.fk0v0fq.