Discussion
In this study, we investigated how long-term protection from herbivory by overabundant white-tailed deer affected the metabolomic profiles of long-lived woody or semi-woody plant species in a suburban forest understory. Furthermore, by studying two indigenous and two nonindigenous, invasive plant species that are subject to different deer preferences, we could consider how metabolomic responses to deer browse may affect the ecology of invasion in suburban forests. Below, in light of the results, we discuss our hypotheses that 1) the metabolomic profiles of fenced and unfenced plants diverge, and 2) this divergence is more pronounced for species most affected by deer in our site. First, we discuss the overall metabolomic variation among the species.
We studied four diverse species, including the indigenous tree speciesN. sylvatica , the indigenous shrub L. benzoin , the nonindigenous shrub E. alatus , and the nonindigenous semi-woody shrub R. multiflora . All four species are potentially subject to severe negative effects from deer herbivory; they are all long-lived and are therefore exposed to chronic deer browse in from overabundant deer while growing in the deer browse zone. Their shared understory forest environment has very limited light , so losing photosynthetic tissue to deer herbivory can be a serious problem for any of these species. Although the four species share these morphological and ecological similarities, they are rather distant phylogenetically. All belong to different plant Orders (Figure 9 ), which could impose metabolomic variation among them based solely on genetic distance . Additionally, they possess some distinct chemical or morphological features that could affect their metabolomes. Lindera benzoin is notably chemically defended , with very aromatic foliage, R. multiflora is defended with large prickles, and E. alatusproduces tough, corky protrusions from its branches. In contrast,N. sylvatica possesses none of these features. Given this diversity, it was not surprising that the four species’ metabolomes were quite distinct, with only 1.1% of the 2,333 detected metabolite features shared among them, and just 23% of the putative metabolic pathways shared among all four species (Figure 2A, D ).
The dendrogram based on all metabolites produced four distinct clusters for the four species, but the two nonindigenous species, E. alatus and R. multiflora , clustered together (Figure 2C ). In a recent study in eastern US forests, demonstrated metabolomic divergence between a large group of indigenous and nonindigenous species in the same plant community, along with evidence for greater regional invasion frequency in nonindigenous species with greater chemical novelty. In our case, with just two indigenous and two nonindigenous species, it is not possible to confidently attribute the clustering to their indigenous status, although our result is consistent with Sedio et al’s observation. Also, E. alatus and R. multiflora happen to be more closely related phylogenetically than are the two indigenous species we studied (Figure 9 ); so their greater metabolomic similarity could be due simply to genetic relatedness rather than some similar physiology related to being successful invasive species. We also note that the two indigenous species, N. sylvatica and L. benzoin , shared the largest number of putative metabolic pathways that the detected metabolites may be part of, possibly suggesting greater similarity in their physiological ecologies. While variation among the four species based on the metabolite features is quite clear, any prediction that divergence in these metabolites indicates differences in the pathways is rather speculative, since the connection from metabolites to pathways is based on the only knowledge available at this point, which is from the model plant Arabidopsis thaliana . We must be very cautious when applying such evidence to quite different species and contexts , such as woody species in a forest understory.
Even given the differences among the species enumerated above, we had predicted that their shared suburban forest understory environment, with its overabundance of deer, would lead to an across-the-board divergence of the plants’ metabolomes in fenced versus unfenced plots. In support of this hypothesis, we did indeed observe a global divergence (Figure 3A, 3B ). We suspect that the difference was caused mainly by protection from deer herbivory in the fencing treatment, which is widely considered the common effect of deer exclosure . However, preventing deer access also may alter other ecological variables that could affect plant stress and influence a plants’ metabolome, i.e. trampling of plants; soil compaction and therefore more difficult root penetration that negatively affects the soil microbial community and/or access to soil water and nutrients; elimination of deer fecal deposition, thereby altering soil nutrients; and release from herbivory for the entire plant community, creating increased competition with other plants . An important aim of ongoing research in deer-related plant ecometabolomics will be to disentangle all of these possible effects of deer on plant metabolomes in natural communities.
The overall divergence in the metabolomes of fenced and unfenced plants is by no means the entire story; rather, there were notable distinctions among the species/fencing treatment combinations in metabolites that were the main drivers of the fenced/unfenced divergence (Figure 3C ). For the top five of these metabolites, a significant difference between fenced and unfenced samples within a species appeared in just some: N. sylvatica for metabolites 1, 2, 4, and 5; L. benzoin for metabolites 2 and 3; R. multiflora and E. alatus for metabolite 5. It should be noted that these were conservative pairwise tests, and the means in Figure 3C suggest additional marked differences, i.e. metabolite 3 in R. multifloraand N. sylvatica and metabolites 2 and 5 in E. alatus . In any case, the species’ responses to protection from deer were variable for these five metabolites, and there was variation in whether the five increased or decreased when exposed to deer. For metabolites 1, 2, and 4, the significant differences were due to greater production in plants exposed to deer, but for metabolites 3 and 5, plants protected from deer by fencing had greater amounts. These findings on the top five metabolites responsible for the global fenced/unfenced divergence therefore partially support the idea that metabolites involved in plant defense responses would be downregulated in fenced plots. More detailed knowledge of these chemicals’ roles in the defense physiology of woody, understory plants is needed to fully understand why some were upregulated and some downregulated.
The separate analyses on each species provides further insight on the responses of defense-related metabolites to deer fencing and also enables discussion of our second hypothesis, that the effect of fencing on a plant’s metabolomic profile is stronger for plant species that are more preferred by deer. Support for this hypothesis would add credence to the idea that it is herbivory by deer, rather than other fencing-related impacts, that affects the metabolome. Out of the four species, N. sylvatica stood out as the one with a metabolomic profile most affected by protection from deer; only for this species did the fenced and unfenced samples clearly cluster into separate groups on both the PCA based on all of its detected metabolites and in the HCA based on its top 30 significantly different metabolites (Figure 4 vs. Figures 5, 6, and 7 ). Some of these top metabolites are putatively involved in pathways that produce intermediate compounds that can be used to produce defense secondary metabolites, e.g. pentose phosphate pathway and cyanoamino acid metabolism , or are involved in indirect defenses, e.g. monoterpenoid biosynthesis . Although fencing did not separate the other three species into distinct treatment-based clusters on the PCA plot, a significant numbers of metabolite features exhibited significant changes in their accumulation based on fencing treatment (Figures 5D, 6D, and 7D ).
Deer did, indeed, browse N. sylvatica frequently, but the other species also were commonly browsed (Figure 1 ). In the summer of 2018, the same season when the samples were taken, N. sylvaticaand E. alatus were browsed the most and at very similar rates, but the differences among the four species was not statistically significant. The browse rates from all observations pooled across all observed individuals from 2012-2019 provides more comprehensive knowledge of deer preference for these four species in the site, and revealed that the two indigenous species, N. sylvatica andL. benzoin , were preferred over the two non-indigenous species,E. alatus and R. multiflora . Full support for our second hypothesis would therefore require that both N. sylvatica andL. benzoin exhibit more divergent metabolomic profiles between the fenced and unfenced plants than did the less-preferred nonindigenous species, yet such divergence was only clear in N. sylvatica .
There are several aspects of herbivory that our data did not address and which may be key to understanding the variation among species in their metabolomic responses to protection from deer; these are all worth further study. First, as in other studies, e.g. , we used the proportion of observed plants with presence of browse signs as our metric for deer browse. A more fine-grained metric that captures the intensity, rather than presence, of browsing on a species could be a better predictor of how metabolomes respond to overabundant deer.
Second, plants are affected by herbivory not just by whether they are attacked, but also by their tolerance of any herbivory that does occur , and the four species we studied may differ in tolerance. For example,N. sylvatica is the only tree of the four species we studied, and this plant architecture, with one central stem and terminal bud, could make its seedlings less tolerant of a deer browse event that removes that bud, compared to more branched shrub species with more meristems . We may then expect a lack of tolerance to be correlated with stronger chemical defenses against herbivory , although some recent evidence for this idea is equivocal or negative ). In any case, variation in tolerance to deer herbivory may correlate to variation in a species’ metabolomic profile.
Finally, our framework for investigating the effect of deer on the plants’ metabolomes relied on comparing plants that had been protected from chronic deer herbivory for years (in fenced plots) versus those exposed to deer (in unfenced plots). Given the overall high browse rates on these species, it was reasonable to assume that at least some of the exposed plants would have experienced herbivory in the recent past, leading to metabolomic profiles different from protected plants. However, little is known about the timing of herbivory-induced metabolite production in long-lived woody plants. While there is evidence that woody plants maintain increased levels of some induced defense chemicals for months following herbivory , other studies in herbaceous plants showed that the metabolomic response to herbivory can be very rapid and then quickly wane . Similarly, the priming of the metabolome caused by an herbivory event, which readies the plant or neighboring plants to rapidly defend against subsequent herbivory, may persist for months in woody plants, and even throughout the plant life cycle of herbaceous plants, or may last only days .
The four species included in this study exhibited variation in their metabolomes, but also in their ecological success in response to protection from deer. The connection between these responses is worth consideration. Nyssa sylvatica was the only species of the four that, in both measures of ecological success, percent cover and height, showed statistically significantly greater values in fenced vs. unfenced plots after 6.5 years (Figure 8 ). This suggests that N. sylvatica is particularly vulnerable to deer pressure in this forest. It was indeed browsed at one of the highest rates, suffering tissue loss, but it also was the one species that showed a clear metabolomic difference in fenced vs. unfenced plots, with increased production in deer-exposed plants of potentially costly secondary metabolites that are involved in defense-related pathways. The increased metabolite production did not appear to protect it from browse relative to the other three species, since it had among the higher browse rates in unfenced plots. Thus, we may expect this indigenous tree species to decline under severe, chronic deer herbivory, as in this suburban forest.
In contrast, one of the invasive species, R. multiflora , actually did better in the unfenced plots exposed to deer pressure, with a significantly greater mean height and a trend of increased cover. Although R. multiflora was browsed, it had among the lower browse rates and much less difference between metabolites in fenced vs. unfenced plants. This all suggests that R. multiflora is resistant and/or tolerant of deer browsing and, under the severe deer pressure in suburban forests, even may gain a competitive advantage over other, more susceptible species. Thus, we may expect this nonindigenous, invasive species to increase in the plant community.
It is important to note that species’ responses to deer in a community are specific and likely cannot be generalized into, for example, indigenous vs. nonindigenous invasive species. Of the other two species, the indigenous L. benzoin did not show a strong difference among the metabolites in fenced vs. unfenced plants and exhibited only a weak positive effect on growth when protected from deer in the fences, but it had among the highest browse rates. The nonindigenous, invasive E. alatus was as highly browsed as the indigenous N. sylvatica in the summer of 2018, yet like the indigenous L. benzoin , had only a weak positive growth response to protection from deer, and did not have clear difference in its metabolites in fenced vs unfenced plants.
This study has set the stage for further research on how severe, chronic deer pressure affects the metabolomes of long-lived species in forests, with possible consequences for the ecology of communities that are now composed of a mix of indigenous and nonindigenous species, as in many suburban forests. While our research clearly indicates that protection from overabundant deer in suburban forests can influence a plants’ metabolome, including affecting metabolites putatively involved in plant defense and stress pathways, it is only a first step. Needed next are ecometabolomic studies that include quantification of deer preference and herbivory intensity in various ways, measurements of tolerance to deer herbivory, documentation of the timing of metabolomic responses to deer herbivory in long-lived plants, and determination of the chemical identities of significant metabolite features.