TRICOMM: Structure, assembly and evolution of natural tritrophic communities

植物群落、昆虫食草动物和它们的昆虫寄生蜂天敌提供了地球上大多数已知物种。这些社区包括导致经济损害的相互作用，如农作物病虫害，以及其他有益于人类社会的社区，如生防剂。尽管它们很重要，但我们仍然对决定哪些物种吃其他物种或被其他物种吃的东西知之甚少。我们知道的最多的是植物和食草动物之间的联系，对食草动物和寄生虫的了解较少，对这三个层次的组合模式的了解也较少。一个关键的问题是，这种三个水平的(三营养的)物种关联性在多大程度上是由植物特性从“自下而上”、由寄生蜂从“自上而下”或它们的某种组合构成的。“自下而上”的观点认为，通过植物对食草动物的影响，食草动物和寄生虫的相互作用是由营养水平较低的过程构成的。相反，自上而下的观点认为寄生虫和食草动物的相互作用推动了草食动物防御的进化，这些特征对于构建寄生虫群落比发现它们的寄主植物更重要。这个项目使用最先进的统计方法评估了这些替代模型及其组合的证据，这些方法需要三种类型的数据：(I)相互作用矩阵，总结一个营养级别的物种和另一个营养级别的物种之间的联系；(Ii)草食动物防御特征数据和(Iii)植物、草食动物及其寄生虫的完整物种水平系统发育。发现植物系统发育是植物-食草动物和食草动物-寄生蜂相互作用的有力预测因素，这将支持自下而上的观点。相比之下，发现食草动物和寄生虫的相互作用强烈地被食草动物的防御特征所预测，这将支持自上而下的观点。首先，我们将评估物种特性和特征对植物-食草动物和食草动物-寄生蜂相互作用的影响，首次测试自下而上和自上而下过程的相对重要性。我们将使用来自北半球三个区域数据集的超过50,000条特定的植物-草食动物-寄生虫相互作用的自然群落记录，其中包括树木、赤眼蜂草食动物和脉管寄生虫。这些群体已经独立进化了足够长的时间，为我们的假设提供了很大程度上独立的测试。其次，我们询问我们三个区域社区的食草动物是否独立进化出了类似的防御系统。如果自上而下的影响很强，而食草动物的防御针对寄生蜂攻击行为的基本方面，那么选择应该有利于类似的防御特征的重复进化。巨蜂食草动物生活在虫胆内，这是一种复杂的新型植物组织，幼蜂诱导其发育。寄生蜂都是通过钻穿胆汁组织来攻击胆小蜂的，之前的研究表明，一些胆汁特征(如脊椎或粘性树脂的外壳)的进化使这一点变得更加困难。最后，我们将评估我们的统计模型预测寄生蜂攻击新的或未采样的食草动物的能力如何，当我们所知道的是它在哪种植物上，它具有哪些虫害特征，以及它与其他五倍子的关系。我们的方法涉及到基于模型的预测，我们拥有真实的数据，因此通过交叉验证，我们可以评估我们模型的准确性(即预测是否没有偏见)和精确度(即预测是否具有高度置信度)。这种方法在预测新出现害虫的天敌和天敌的非靶标受害者方面具有特别的价值，我们将把它应用于预测攻击全球害虫板栗小蜂的天敌。

Communities of plants, insect herbivores, and their insect parasitoid enemies provide most of the known species on Earth. These communities include interactions that lead to economic damage, such as pests of crops, and others that benefit human societies, such as biocontrol agents. Despite their importance, we still know little about what determines which species eat, or are eaten by, other species. We know most about links between plants and herbivores, less about herbivores and parasitoids, and less again about patterns over all three levels combined. A key question is the extent to which such three level (tritrophic) species associations are structured from the 'bottom-up' by plant traits, from the 'top-down' by parasitoids, or some combination of these. The 'bottom-up' view regards herbivore-parasitoid interactions as structured by processes happening a trophic level lower, via the effects of plants on herbivores. In contrast, the 'top-down' view sees parasitoid-herbivore interactions as driving the evolution of herbivore defences, and these traits as more important for structuring parasitoid communities than the host plants on which they are found.This project assesses the evidence for these alternative models, and their combinations, using state of the art statistical methods that require three types of data: (i) an interaction matrix, summarising links between species in one trophic level and those in another; (ii) herbivore defence trait data and (iii) complete species-level phylogenies for plants, herbivores and their parasitoids. Finding that plant phylogeny is a strong predictor of both plant-herbivore and herbivore-parasitoid interactions would support the bottom-up view. In contrast, finding that herbivore-parasitoid interactions are strongly predicted by herbivore defensive traits would support a top-down view. First, we will estimate the effects of species identity and traits on plant-herbivore and herbivore-parasitoid interactions, providing the first test of the relative importance of bottom-up versus top-down processes. We will use over 50,000 records of specific plant-herbivore-parasitoid interactions for natural communities comprising trees, gallwasp herbivores, and chalcid parasitoids, sampled from three regional datasets that span the Northern Hemisphere. These communities have evolved independently for long enough to provide largely independent tests of our hypotheses. Second, we ask whether herbivores in our three regional communities have independently evolved similar sets of defences. If top-down effects are strong, and herbivore defences target fundamental aspects of parasitoid attack behaviour, then selection should favour the repeated evolution of similar sets of defensive traits. Gallwasp herbivores live inside galls, complex novel plant tissues whose development the larval wasps induce. Parasitoids all attack gallwasps by drilling through gall tissues, and previous work suggests that some gall traits (such as coatings of spines or sticky resins) have evolved to make this more difficult. Our hypothesis is that such gall traits will both structure parasitoid communities and have evolved repeatedly.Finally, we will assess how well our statistical models predict which parasitoids attack a novel or unsampled gallwasp herbivore when all we know about it are which plant it is on, which gall traits it has, and how it is related to other gallwasps. Our approach involves making model-based predictions for gallwasp-parasitoid interactions for which we have real data, so that via cross-validation we can assess the accuracy (i.e. whether predictions are unbiased) and precision (i.e. whether predictions are made with high confidence) of our model. This approach could be of particular value in predicting the natural enemies of emerging pests and the non-target victims of natural enemies, and we will apply it to predicting the enemies attacking oriental chestnut gallwasp, a global pest species.