Architectural diversity and galling insects on Caryocar brasiliense trees

Galling insects are a highly sophisticated herbivore group on Caryocar brasiliense, a tree that represents the main income source for many communities. The effect of architectural diversity of C. brasiliense trees on galling insect community diversity and abundance was studied. The abundance of adult insects and galled leaves were seven and 1.6 times higher in trees with a greater height/width of canopy (RHW) ratio, respectively. Gall parasitoid richness was 1.8 times greater on trees with higher RHW. Zelus armillatus (Lepeletier & Serville) (Hemiptera: Reduviidae) and ant numbers were 5.8 and 2.7 higher on trees with the largest and smallest RHW, respectively. More complex plant architectures favored species diversity for galling insects and their natural enemies. The competition among four galling insect species for space and feeding and the evidence of “prudence strategy” were, for the first time, observed for galling insects in the Brazilian Cerrado biome.


Results
Architectural diversity and galling insects. In comparison with trees with a smaller RHW, Caryocar brasiliense trees with a greater height/width canopy ratio (RHW) had 7.0 times the number of galling insects, 7.8 times more Eurytoma sp. (globoid galls) adults, 1.6 times the percentage of galled leaves, 7.2 times greater leaflet area with galls, 2.2 times more leaf area (mm 2 ) occupied by Eurytoma globoid galls, and 3.9 times the number of Eurytoma globoid galls. On the other hand, the number of Hymenopteran discoid galls was 7.8 times higher on C. brasiliense trees with lower RHW than on those with greater RHW.

Discussion
Architectural diversity and galling insects. The attacks more frequently of galling insects for taller C.
brasiliense trees or those with wider crowns confirms the hypothesis that larger trees support more species and individuals of galling insects and their communities (i.e. parasitoids and predators) 7,18 and the prediction of the plant vigour hypothesis (PVH). The PVH indicates oviposition preference of females and higher offspring performance of herbivorous insects on fast-growing plants (plant modules). Sawfly galls attacked more frequently longer shoots without galls and those with two, three, four, or five galls were successively longer than with fewer galls 19 . Bigger plants typically support more galling insect species 5,20 .
Galling insect preference for C. brasiliense trees in relation to height and canopy width are of two groups: i) Eurytoma sp. (most abundant) and ii) other three galling insect species. Eurytoma (globoid galls) preferred taller trees (RHW 3) where it showed higher gall numbers and area and consequently a higher percentages of galled leaflets and greater area with galls. The second group, Eulophidae spherical galls, Hymenopteran discoid galls, and Bruchophagus sp. vein galls did not prefer taller trees, but those with wider crowns (i.e. Hymenopteran discoid galls) generally showed higher numbers of its galls. Bruchophagus vein galls were not found on taller trees (RHW 3). The question why Eurytoma sp. attacks more frequently taller C. brasiliense trees and the other three galling insects a wider canopy may be explained by the fact that leaves on the top of taller C. brasiliense trees probably suffer a drying effect from the wind that shorter trees do not especially in regions with high temperatures and sunlight and low relative humidity 10 . Higher fruit production in the basal part than in the apical part of C. brasiliense canopy increased flower and fruit drop by winds 10 . This indicates that leaves more exposed to sunlight and winds can negatively affect their resistance to galling insects 21 . The higher wind and sunlight incidence harden leaves and may also explain the higher population of Eurytoma sp. on taller C. brasiliense trees. These leaves are, probably, a better food source for galling insects and Eurytoma sp. seems to predominate on them. Harsh ecophysiological conditions in the upper canopy of tropical rainforests increase gall-forming populations. Sclerophyll leaves increased with tree height while free-feeding herbivores decreased inversely 22 . Water and nutrient stress in the canopy tree meristem in tall wet tropical rainforests may cause leaf sclerophylly, forming a suitable ecosystem for gall-forming insects 22 . The other three galling insect species also prefer leaves more exposed to wind and sun 12-14 , but they preferred trees with a wider canopy than taller trees, avoiding competition with Eurytoma sp.
The pattern changes of the three galling insect species between C. brasiliense trees with low or high Eurytoma sp. density is related to sunlight and wind exposure, gall distribution on the leaf (i.e. border/interior of the plant canopy, border/near the mid vein, distal/proximal-near to petiole), and branch level (i.e. north/south) 21 . Most galling insects preferred the C. brasiliense leaves more exposed to the wind and sunshine, on the branch and at the leaf border and median parts [11][12][13][14] . However, other galling insects attacked other parts of C. brasiliense leaves in the presence of high Eurytoma sp. populations 21 . This suggests competition between the three galling insect species and Eurytoma sp. with the latter showing faster colonization of plants with greater biotic potential 21 . Caryocar brasiliense loses its leaves in Aug/Sep with new leaves appearing at the end of September, a period without rainfall and with strong wind and high sunlight 10 . Eurytoma sp. induces galls on young expanding C. brasiliense leaves when the wasp females were easily found ovipositing on unfolded leaves. Approximately two days later the oviposition site became reddish with a gall visible a few days later 11 . Plant phenology and colonization periods on C. brasiliense leaves explains the decrease of Eurytoma gall abundance as the numbers of other galling insects increased throughout the year as well as the differential temporal distribution of galling insects 23 . Chemical or visual markers may indicate which galling insect dominates a particular ecological niche (i.e., part of a leaf, branch or even a tree) 21 . The genetic differences between C. brasiliense plants and the chemical or morphological composition with different leaf parts, branch positions, sun/wind exposure or relationships with other arthropods (i.e., natural enemies) 21 may also be responsible for this process or these differences.
The reduction of the galling insect communities on C. brasiliense, which acts as a biogeographic islands for these insects, can be explained by the collect of fruits without control affecting plant propagation. Caryocar brasiliense had only about 7.96 and 10.65% of individuals up to 1.0 m tall and 59.58 and 44.73% above 3.0 m (reproductive phase) in the cerrado and pasture areas, respectively 10 . This indicates that collectors remove nearly all fruits from the tree thus reduce C. brasiliense propagation in the cerrado of Brazil. Eurytoma sp. attack more frequently larger trees and the other three species of insect galling the smaller trees, which increases the risk of their extinction due to the inadequate renewal of C. brasiliense trees. This agrees with reports for exotic plants that can extinguish those native to smaller biogeographic islands 16 and to Lepidoptera communities (i.e. generalists x specialists) in urban parks being affected by their size and human disturbance 15 . Architectural diversity and natural enemies. A higher diversity of natural enemies, principally parasitoids, as was found with galling insects in the more complex host individuals, indicates that their populations depend on their prey and host and that they follow of the herbivorous insects 24,25 . The natural enemies found can be divided into parasitoids and predators. The first group (i.e. Sycophila sp. and A. magistretti) showed the same trend as its hosts (galling insects) with higher richness, diversity and abundance on wider crown trees. Sycophila sp., a major Eurytoma sp. 23,26 parasitoid showed a higher survival rate on lower RHW trees and Eurytoma had better fitness on higher RHW trees. The larger RHW trees had the highest number of Eurytoma sp. and Z. armillatus predators 11,26 . These facts showed an evidence of "prudence strategy", whereby predators fed on parasitized prey, preserving the healthy prey as a food reserve for future generations, without endangering prey populations 27 . The predator Z. armillatus may prefer attacking Eurytoma sp. galls parasitized by Sycophila sp. The "prudence strategy" has been observed for Protonectarina sylveirae (Saussure) (Hymenoptera: Vespidae) with the leafminer Liriomyza sp (Diptera: Agromyzidae) and parasitized aphids (Hemiptera: Aphididae) 28 , but this is the first time that it is seen, evidenced, in galling insects. However, a prudence strategy would have to be demonstrated experimentally in future study. Positive relationships between indole butyric acid concentrations and successfully induced globoid galls as well as between the number of adults of the galling Eurytoma sp. and its parasitoid Sycophila sp. were found 26 . Selecting an enemy-free space could be the main reason why sawfly oviposition patterns with higher values on lower quality plants decreases predation and parasitism 8,[29][30][31] .
The similar abundance, richness and diversity of the second group on C. brasiliense with different RHW is due to the fact that they are generalists with little dependence on a single prey, contrary to most parasitoids 32 and these trees support greater diversity of free-feeding herbivore insects [33][34][35][36] . The higher percentages of defoliation and populations of defoliators and ants on C. brasiliense trees may be explained by the impact of predatory ants, bugs and spiders reducing defoliators and leaf miner insects on wider crown plants 35 . Variations in abundance and diversity of sucking insects and their natural enemies on wide crown trees and of the number of species and individuals of natural enemies is similar to those for Dikrella caryocar (Coelho, Leite & Da-Silva) (Hemiptera: Cicadellidae), of ants, predator thrips and lady beetles as well as with higher values on these C. brasiliense trees 36 .
Ants, Epipolops sp., Holopothrips sp., Sycophila sp., spiders, and Z. armillatus can be important to control Eurytoma sp. on C. brasiliense trees 11,23,26,37,38 . The higher numbers of the predator Z. armillatus on C. brasiliense trees at the University Campus might be due to more leaves galled by Eurytoma sp. on these trees than in pastureland and cerrado 35 . Zelus armillatus preyed on Eurytoma galls, which can cover up to 70% of the leaf area 11,26 . These galls can support higher diversity of natural enemies which can kill and cause a top down impact on natural enemies 8,30,31 .
Complexity of plant architecture (i.e. biogeographic island) favored the diversity of galling insect species and their parasitoids and the predator Z. armillatus. The patterns of natural enemies indicate that their populations depend on prey/hosts and that they follow those of the herbivorous insects. The competition among the four galling insect species for shelter and feeding space and the evidence of "prudence strategy" were observed for the first time for galling insects. The pasture area has 84.2% of the soil covered by grass, 5.8%, 4.8% and 2.8% by herbs, small and tall trees, respectively with an average of 42 C. brasiliense trees per ha 34 .

Host Plants Studied and Galling Species.
Caryocar brasiliense trees can reach over 10 m tall with a 6 m canopy width 10 . Its fruits have an internal mesocarp rich in oil, vitamins, and proteins with many compounds of medicinal importance. Moreover, it is used by humans for food, production of cosmetics, lubricants, and in the pharmaceutical industry [40][41][42] .
Adult (reproductive stage) C. brasiliense trees in the pasture area were two to nine meters tall with a two to 11 meters width canopy 34 . The hymenopteran galls on C. brasiliense leaves are Bruchophagus sp. vein galls (Hymenoptera: Eurytomidae), Eulophidae spherical galls (Hymenoptera), Eurytoma sp. glodoid galls (Hymenoptera: Eurytomidae) and Hymenopteran discoid galls, the morphology, natural history and tree distribution of which have been described [11][12][13][14]21 . Descriptions of the natural enemies and other herbivores in this C. brasiliense system were also provided 21 .
Study Design. The study design was completely randomized with 36 replications (36 trees). Data was collected from C. brasiliense adult trees (producing fruits) every 50 m along a 600-m transect. Insect species (i.e., rare species) data was collected over three consecutive years. No fertilizers or pesticides were used.
The distribution of galling insects and their galls, predators, and leaf percentages infested with galls (three leaflets/leaf) were recorded in 12 fully expanded leaves of 36 C. brasiliense trees (one leaf per canopy vertical and horizontal stratifications). Sampling was performed monthly in the morning (7-11 AM) by visual observation 43 . Insects were collected with tweezers, brushes, or aspirators and preserved in vials with 70% alcohol for identification by taxonomists. Twelve leaves per tree were collected and transported to the laboratory. Gall size was measured with a digital caliper (accurate to the nearest 0.1 mm). Leaves were scanned and their area and those of each leaf with galling species calculated using a computer program. Subsequently, leaves of each sample were placed in white plastic pots (temperature of 25 °C), and the emergence of galling insects, hyperparasitoids, inquilines and parasitoids evaluated on alternate days for 30 days. The insects emerged were collected and preserved as described for identification by taxonomists. The voucher number for spiders is IBSP 36921-36924 (Instituto Butantan, São Paulo state, Brazil) and that for insects is 1595/02 and 1597/02 (CDZOO, Universidade Federal do Paraná, Paraná state, Brazil).
SCiEntiFiC REPORTS | 7: 16677 | DOI:10.1038/s41598-017-16954-6 Statistical Analyses. Averages were made by reducing the data to leaflet/tree. The richness and species diversity of galling insects and their parasitoids and predators were calculated per tree. The diversity was calculated using Hill´s formula 44 and the species richness with Simpson indices 45,46 . The height and width (RHW) ratio on C. brasiliense trees and percentage survival of galling insect adults [(number of adults/number of galls)*100] and of the parasitoid adults [(number of adults/number of host adults)*100] were calculated. The ratio of height/ width categories for C. brasiliense trees are: RHW 1: 1.00 to 1.64, RHW 2: 1.64 to 2.10 and RHW 3: 2.11 to 2.53 m with 12 trees per group.
The effect of RHW on ecological indexes and numbers of individuals per herbivore species and natural enemies (transformed to √x + 0.5 or arcsine for percentage data whenever necessary) was tested with ANOVA (P < 0.05) and Scott-Knott's test (P < 0.05) and regression analysis were performed with the System of Statistical and Genetics Analysis of the Universidade Federal de Viçosa (UFV) 47 .
The Spearman correlation matrix, among the most significant characteristics, was calculated. The matrices were submitted to correlation networks 48 . The edge thickness was controlled by applying a cut off value of 0.33 (value from which the Spearman correlation became significant, meaning that only edges with | | ≥ . r 033 ij were highlighted). All the analyses were performed using the R software version 3.4.1 49 . The correlation network procedure was performed using the package qgraph 48 .