Metabolomics provides an unprecedented window into diverse plant secondary metabolites that represent a potentially critical niche dimension in tropical forests underlying species co-existence. Here, we used untargeted metabolomics to evaluate chemical composition of 358 tree species and its relationship with phylogeny and variation in light environment, soil nutrients, and insect-herbivore leaf damage in a tropical rain forest plot. We report no phylogenetic signal in most compound classes, indicating rapid diversification in tree metabolomes. We found that locally co-occurring species were more chemically dissimilar than random, and that local chemical dispersion and metabolite diversity was associated with lower herbivory, especially that of specialist insect herbivores. Our results highlight the role of secondary metabolites in mediating plant-herbivore interactions and their potential to facilitate niche differentiation in a manner that contributes to species coexistence. Furthermore, our findings suggest that specialist herbivore pressure is an important mechanism promoting phytochemical diversity in tropical forests.

Plant secondary metabolites were extracted and analyzed from leaves using ultra-high performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/MS), following Sedio et al. (2021) with slight modifications. Three branches were taken from different directions, and 10 leaves per branch were sequentially selected from top to bottom on each branch to avoid overestimating the percent herbivory and ignoring certain types of herbivores (Woodman & Fernandes, 1991). All collected leaves were scanned, and the leaf area was calculated using ImageJ software (Abramoff et al., 2004). The percent herbivory for each leaf was calculated as the ratio of damaged area to estimated undamaged area, with higher percentages indicating greater herbivore damage (Kurokawa & Nakashizuka, 2008). We classified herbivore damage into broad categories (e.g., hole feeding, margin feeding) according to Labandeira et al. (2007) and further divided them into three diet-breadth categories: generalized, intermediate, or specialized (Labandeira et al., 2007; Wang et al., 2022). We used published data from our laboratory for 9 soil nutrient variables in the plot (Hu et al., 2012, Yang et al., 2014), including pH, total N, total P, total K, available N, extractable P, extractable K, total C and bulk density. Light availability was measured using hemispherical photographs taken with a digital camera and analyzed using the Gap Light Analyzer Version 2.0 software to calculate the canopy gap fraction, representing the non-vegetated portion of the image (Frazer et al., 2000). To quantify variation in percent herbivory and chemical diversity in different habitats, the plot was divided into slope, ridge and valley habitats from Yang et al.