Species‐specific plant‐mediated effects between herbivores converge at high damage intensity

Abstract Plants are often exposed to multiple herbivores and densities of these attackers (or corresponding damage intensities) often fluctuate greatly in the field. Plant‐mediated interactions vary among herbivore species and with changing feeding intensity, but little is known about how herbivore identity and density interact to determine plant responses and herbivore fitness. Here, we investigated this question using Triadica sebifera (tallow) and two common and abundant specialist insect herbivores, Bikasha collaris (flea beetle) and Heterapoderopsis bicallosicollis (weevil). By manipulating densities of leaf‐feeding adults of these two herbivore species, we tested how variations in the intensity of leaf damage caused by flea beetle or weevil adults affected the performance of root‐feeding flea beetle larvae and evaluated the potential of induced tallow root traits to predict flea beetle larval performance. We found that weevil adults consistently decreased the survival of flea beetle larvae with increasing leaf damage intensities. In contrast, conspecific flea beetle adults increased their larval survival at low damage then decreased larval survival at high damage, resulting in a unimodal pattern. Chemical analyses showed that increasing leaf damage from weevil adults linearly decreased root carbohydrates and increased root tannin, whereas flea beetle adults had opposite effects as weevil adults at low damage and similar effects as them at high damage. Furthermore, across all feeding treatments, flea beetle larval survival correlated positively with concentrations of carbohydrates and negatively with concentration of tannin, suggesting that root primary and secondary metabolism might underlie the observed effects on flea beetle larvae. Our study demonstrates that herbivore identity and density interact to determine systemic plant responses and plant‐mediated effects on herbivores. In particular, effects are species‐specific at low densities, but converge at high densities. These findings emphasize the importance of considering herbivore identity and density simultaneously when investigating factors driving plant‐mediated interactions between herbivores, which advances our understanding of the structure and composition of herbivore communities and terrestrial food webs.


Method
To examine whether there are wide variations in composition and abundance of herbivores across tallows (Triadica sebifera) in real conditions, we conducted a field survey in July 2018 in Dawu,Hubei,China (31.58°N ,114.18°E). We randomly selected 100 tallow plants (80-120 cm in height) along farmland roadsides. The plants were separated by at least 50 m. We carefully checked each leaf and recorded the identity and abundance of herbivores on each plant. We collected herbivores that we could not identify in the field, stored them in glass tubes with 95% alcohol and took them back to laboratory for identification. To quantify natural damage levels, we randomly selected 4 branches that were located diagonally on each plant and visually assessed the percentage of leaf damaged area (to the nearest 5%) for the terminal 10 leaves (40 leaves in total for each plant). A leaf was classified as 100% damaged when it was rolled by weevil adult or when caterpillars had consumed all but the petiole. We determined leaf damaged area of each plant by averaging the visual estimates for all 40 leaves. Leaf damage areas were used for parameterization in the following experiments.

Statistical analyses
To investigate the association between herbivore abundance and tallow leaf damage in the field survey, we performed a simple linear regression analysis across 100 surveyed plants for each herbivore. To account for the interactive impacts between herbivores, we incorporated all herbivores in a multiple regression analysis.
Model selection was performed using a stepwise regression with backwards removal based on Akaike's Information Criterion (Burnham and Anderson 2002). Ricania speculum was excluded from analysis because it is not a leaf chewing herbivore and does not cause leaf damage that can be measured with our methods.
All analyses were performed in R, version 3.6.3 (R Development Core Team 2020). The package 'STATS' was used to conduct the simple and multiple regression analyses. The package 'MASS' was used to conduct the stepwise regression analysis.

Results
In the field, tallow plants experienced a wide range of feeding intensity and the percentage of leaf damaged area ranged from 0 to 71.5% (mean ± se, 19.11 ± 1.64 %; Fig. S1a). There were 16 species found on the leaves of tallow, including eight Lepidoptera species, seven Coleoptera species and one Hemiptera species (Table S1). Flea beetle adult (Bikasha collaris) and weevil adult (Heterapoderopsis bicallosicollis) were two of the most abundant herbivores, in the proportion of 69.8% and 12.0% in total and with 16.4 and 2.8 per plant, respectively (Fig. S1b). In both simple and multiple regression analyses across 100 surveyed plants, the percentage of leaf damaged area was strongly positively correlated with the abundance of these two herbivores (Table S1- (Table S1-S2), suggesting that they are major drivers of leaf damage on tallow.

Table S1
Results of simple linear regressions between the percentage of leaf damaged area and herbivore abundance on 100 surveyed tallow (Triadica sebifera) using linear model. Each herbivore was analyzed separately. Ricania speculum was excluded from analysis because it is not a leaf chewing herbivore and does not cause leaf damage that can be measured with our methods. Coefficient estimates in each model are presented. The statistical significance was estimated by t tests. Significant effects (P < 0.05) are shown in bold. SE = standard error.