DISCUSSION
The results of this study highlight the coupling of functional traits associated with plant growth forms specific of different environmental conditions with the differential production of glucosinolates (GSLs) across Cardamine species. Specifically, we found thatCardamine species cluster into four main groups. Each group, being anchored within a major climatic zone of the Alpine elevation gradient, expressed different levels of phytochemical diversity, and exhibited an overexpression of unique GSLs; indoles being the signature of the alpine and low elevation groups, and aliphatics the signature of two mid-elevation zones. Such habitat-driven phytochemical convergence had variable consequences on herbivores belonging to different diet breadths and feeding guilds. We thus suggest that the identity and diversity of secondary metabolites within a given species is determined by convergent adaptation to the local abiotic and biotic conditions, ultimately affecting different herbivores, variously.
One major prediction for explaining variation in phytochemical diversity across species is that phylogenetic conservatisms for phytochemical production should result in closely-related species being more phytochemically-similar that distantly-related species (Futuyma & Agrawal 2009a). On the contrary, we found that the diversity of GSLs was not explained by phylogeny. This is in contrast to the phylogenetic conservatism reported across different families of plants (Wink 2003; Wink & Mohamed 2003; WInkler & Mitter 2008), or within genera; such as the production of aliphatic and branched-chain GSLs in the genusStrepthantus (Cacho et al. 2015), or the production of cardenolides in the genus Asclepais (Agrawal et al. 2009; Rasmann & Agrawal 2011). However, we interpret the lack of phylogenetic signal in GSL production in our system with caution, as the reduced number of investigated species impairs the ability to fully tease apart potential patterns that might emerge when assessing more species-rich clades (Swenson 2019). Nevertheless, our results are indicative of other factors, other than shared evolutionary history, in driving the variable production of GSLs across species having colonized different habitats. Accordingly, previous studies also found ecological convergence in chemical defensive profiles across species, independently of phylogenetic relationship (Kursar & Coley 2003; Salazar et al.2016).
Here, we expanded on this previous work by integrating large-scale ecological gradients, and we observed a significant correlation between plant functional traits, which are associated with the specific niche of the species within each elevation zone, and the GSLs matrix. These results build on previous work showing, across 15 differentCardamine species, a strong correlation between climatic variables and 10 functional traits related to abiotic tolerance, growth and defence (Defossez et al. 2018). Taken together, these results suggest that climatic factors force species into specific growth forms (Wright et al. 2004; Díaz et al. 2016), and likewise dictate the shape and structure of the phytochemicals to be produced. However, our result are less in line with predictions of the screening hypothesis (Berenbaum et al. 1991; Duffey & Stout 1996), but more with the resource availability hypothesis (Coley et al.1985); alpine species, for which herbivore pressure is the lowest (Pellissier et al. 2016), but growing in resource-poor environments, expressed the highest number for practically all indices of phytochemical GSL diversity. In other words, we observed a less direct effect of herbivory pressure than that of the habitat on phytochemistry (Richards et al. 2015). We observed that alpine species expressed the highest phytochemical diversity, particularly when compared to mid-elevation plant species. We argue that the higher costs associated with replacement of biomass loss in the harsher environment, characteristics of high elevation zones (Korner et al. 1989; Chapin & Korner 1995), could be an explanation for the increased GSL diversity as observed in our study. At high elevation, the cost to recover tissue lost is strongly limited by the paucity of resources and the cold temperatures. Therefore, for these alpine species, the fitness costs of herbivory cannot be outweighed by the energy saved in reduced levels of defences (Bryant et al. 1983). The production of defence strategies is therefore more linked to the impact of herbivory based on resources available, than solely on herbivore pressure (Coleyet al. 1985). Therefore, while alpine species (group 1) are characterized by a combination of traits conferring high abiotic resistance (e.g. lower SLA values, tougher leaves, and slow growth), they also integrate higher levels of phytochemical diversity for likely withstanding the scattered, but potentially lethal, attack of herbivores (Rasmann et al. 2014a). Low-elevation species, on the other hand, experience a constantly high pressure by herbivores. Thus, while expressing traits relating to fast growth and lower abiotic resistance (higher SLA values and softer leaves), they also express higher GSL diversity, particularly compared to the species within the two mid-elevation groups. Species occupying mid-elevation zones of forest habitats are typically comprised of species with high biomass production (especially species in group 2) and high carbon to nitrogen ratio (CN) (Defossez et al. 2018), which suggest a preference toward investing in tolerance instead of defenses for those species (Núñez-Farfán et al. 2007). In sum, our results suggest that where plant species, independently of their phylogenetic relatedness, share a common compendium of ecological variables, such as common herbivore pressure, similar resource levels, or similar climates, plants are also likely to defend themselves with a similar set of chemical molecules.
In accordance with alpine species bearing the highest chemical diversity values, caterpillars, especially the specialist P. brassicae grew less on those plants. Particularly, these plants produced the highest H values. However, our results do not fully concord with the general view that GSL are more efficient against generalist than specialists (Schlaeppi et al. 2008; Schweiger et al. 2014; Rasmannet al. 2015). P. brassicae feeds exclusively on plants producing GSLs (Chew 1988), also utilizing these compounds for host recognition and as feeding stimulants (Moyes et al. 2000). Interestingly, it has been shown that that ovipositing P. rapaefemales respond more strongly to indole GSLs, such as glucobrassicin, (Rodman & Chew 1980; Renwick et al. 1992; Huang et al.1994), which is also a GSL characterizing the alpine species. Therefore, the slow-growing and comparatively very small alpine Cardaminespecies needed to evolve specific GSL combinations, through high H values, that are toxic to the specialist herbivores, but this hypothesis needs to be tested thoroughly using mixtures of compounds.
Concerning aphids, we found that the generalist aphids M. persicae grew more on plants with lower FDiv values (i.e. species in group 2). Therefore, for generalist aphids, our results support the prediction of a negative correlation between the functional chemical diversity/divergence of GSLs and herbivore performance (Dyer et al. 2018). That said, it has been argued that GSLs in general are less toxic to aphids than to caterpillars, because aphids avoid the activation of GSLs by the enzyme myrosinase (de Vos et al. 2007). Nevertheless, indole GSLs are thought to be less stable, and activate spontaneously in the absence of myrosinase. Consequently, indole GSLs alone have been shown to impair the growth of the generalist aphid M. persicae when added to an artificial diet or overexpressed in host plants (Kim & Jander 2007; Kim et al.2008). On the contrary, specialist aphids, such as B. brassicae , are able to accumulate aliphatic GSLs (Franciset al. 2001). In line with these findings, we suggest that aphids are impaired by the indole GSLs, which are more produced by plants in group 1 and 4, and less produced by the plants from group 2, as well by a GSL chemical mixture that favour functional divergence.
In summary, this study, by combining metabolomics analyses with insect bioassays on plants growing along steep ecological gradients, provides a novel approach for explaining the cause and consequences of variations in phytochemical diversity across plant species. By including several indices of phytochemical diversity, we took a step further in mechanistically disentangling the effects of different metrics of phytochemical diversity on insect herbivore resistance. For instance, we observed that groups of plants bearing practically identical chemical richness values (S) can have completely different GSLs compositions. This indicates that focusing on arbitrarily-selected indices of phytochemical diversity can be misleading in interpreting the metabolomics data and their effects (Wetzel & Whitehead 2020). Taking into account different factors determining such diversity, such as compound class, metabolites’ molecular metrics, or biological activity, we were able to add a functional dimension to phytochemical diversity, as was for instance done for cardenolides in milkweeds using polarity values (Rasmann & Agrawal 2011). We thus argue that the classical indices of phytochemical diversity used so far (total amount, number of compounds, Shannon diversity), should be expanded to include functional axes of chemical diversity, in order to be able to interpret the biological activity of secondary metabolites in a more precise and ecologically relevant manner , and to integrate these novel axes related to plant defenses into the functional syndrome of plant growth forms.