Speciose tropical plant genera harbor a large diversity of insect herbivores and such systems are important in understanding the generation of tropical diversity. Here we extend previous food web studies across Ficus to include not only insect herbivore data but also trait and plant chemistry data in a phylogenetic context. We generated a multi-gene phylogeny of 21 Ficus species representing a sympatric lowland rainforest community in Papua New Guinea. We then investigated the trade-offs between several key defensive traits (triterpene diversity and content, cysteine protease activity, trichome density) and palatability (specific leaf area, Nitrogen content and Carbon content) as well as their effects on caterpillar abundance. Given the dominance of Ficus also as a highland genus in PNG we extended our study to include sampling sites along a 2,500m elevational transect. We studied the same suite of chemical and physical defences in five species of Ficus with wide elevational ranges. Our hypotheses were generated under existing models of insect-plant co-evolution, predicting an escalation of traits across the phylogeny and a trade off in defensive investment along the elevational gradient.
We found several significant negative correlations between traits, notably between triterpene diversity and total triterpene content. The most effective trait in reducing caterpillar abundance was the activity of cysteine proteases found in Ficus latex; we suggest that other traits may have a more important role in influencing caterpillar community structure. Furthermore, we show the major influence of phylogeny occurs at mid-level nodes and that sister species tend to be more divergent in trait space than expected by chance. In combination with the fact that we found limited evidence for trait escalation across our phylogeny these results suggest that the evolutionary dynamics of herbivore pressure acting in local communities may force divergence in defensive traits between closely related species. For highland Ficus we found species specific trends in defensive investment that i) showed correlations with herbivore inflicted damage, ii) demonstrated within trait trade-offs consistent with our lowland results and iii) generally showed strong elevational trends. This situation, with labile suites of defensive traits being adopted at a community level and showing variability along environmental gradients suggests a system in flux, which may be a more realistic hypothesis for species rich plant communities growing in sympatry than a constant escalation of defensive traits.