Ecological and metabolomic responses of plants to deer exclosure in a suburban forest

Abstract

Trees and shrubs in suburban forests can be subject to chronic herbivory from abundant white-tailed deer, influencing survival, growth, secondary metabolites, and ecological success in the community. We investigated how deer affect the size, cover, and metabolomes of four species in the understory of a suburban forest in central New Jersey, USA: the woody shrubs Euonymus alatus and Lindera benzoin, the tree Nyssa sylvatica, and the semi-woody shrub Rosa multiflora. For each species, we compared plants in 38 16 m² plots with or without deer exclosure, measuring proportion cover and mean height after 6.5 years of fencing. We scored each species in all plots for deer browsing over 8 years and assessed selection by deer among the species. We did untargeted metabolomics by sampling leaves from three plants of each species in an equal number of fenced and unfenced plots, conducting chloroform-methanol extractions followed by LC-MS/MS, and conducting statistical analysis on MetaboAnalyst. The proportion of a species browsed ranged from 0.24 to 0.35. Nyssa sylvatica appeared most selected by and susceptible to deer; in unfenced plots, both its cover and mean height were significantly lower. Only cover or height was lower for E. alatus and L. benzoin in unfenced plots, while R. multiflora height was greater. The metabolomic analysis identified 2333 metabolites, which clustered by species but not fencing treatment. However, targeted analysis of the top metabolites grouped by fencing for all samples and for each species alone and was especially clear in N. sylvatica, which also grouped by fencing using all metabolites. The most significant metabolites that were upregulated in fenced plants include some involved in defense-related metabolic pathways, e.g., monoterpenoid biosynthesis. In overbrowsed suburban forests, variation of deer impact on species' ecological success, potentially mediated by metabolome-wide chemical responses to deer, may contribute to changes in community structure.

Keywords: browsing pressure; ecometabolomics; suburban forests; white‐tailed deer.

FIGURE 1

Boxplots of proportion cover in the herb layer of four species in fenced and unfenced plots in Herrontown Woods Preserve in suburban central New Jersey, USA after 6.5 years of deer fencing exclosure or no fencing (N = 20 fenced plots and 18 unfenced plots).

FIGURE 2

Boxplots of heights of four species in the understory of fenced and unfenced plots in Herrontown Woods Preserve in suburban central New Jersey, USA after 6.5 years of deer fencing exclosure or no fencing (Euonymus alatus, fenced N = 65, unfenced N = 18; Lindera benzoin, fenced N = 23, unfenced N = 41; Nyssa sylvatica, fenced N = 19, unfenced N = 11; Rosa multiflora fenced N = 11, unfenced N = 17).

FIGURE 3

Comparison of the metabolite profiles of the species in the fenced and unfenced plots. (a) Number of common and unique metabolite features identified for the four woody tree species is depicted in the Venn diagram. (b) Principal component analysis (PCA) of all metabolite features grouped the samples based on species. (c) The dendrogram displays the relationship of the samples based on all metabolites and shows the presence or absence of fences. (d) Number of common and unique metabolic pathways predicted based on the differentially accumulating metabolites.

FIGURE 4

Identification of metabolites that accumulate differentially in the four species following the treatment gradient. (a) Orthogonal partial least squares discriminant analysis (OPLS‐DA) separated the samples into eight distinct groups corresponding to species and containment in fence or unfenced plots. (b) Important features that contributed to the PLS‐DA‐based separation of the samples are depicted with their pattern of accumulation, shown by the color code. (c) Normalized concentrations (mean ± SE) of the top five metabolite features identified by PLS‐DA. Within each metabolite, concentration was dependent on the combination of species and fencing treatment (ANOVA species × treatment interactions: Metabolite 1: F _(3,42) = 3.17; p = .03; metabolite 2: F _(3,42) = 3.87, p = .01; metabolite 3: F _(3,42) = 3.05, p = .03; metabolite 4: F _(3,42) = 3.96, p = .01; metabolite 5: F _(3,42) = 2.99, p = .04). Different letters indicate statistically significant differences by Tukey HSD among all means within a metabolite.

FIGURE 5

Untargeted metabolomic analysis of fenced and unfenced Nyssa sylvatica samples. (a) Principal component analysis (PCA) of all metabolite features grouped the samples into two clusters that correspond with the fencing treatment. (b) Partial least squares discriminant analysis (PLS‐DA) identified the top 15 metabolites that accumulated significantly differently among the fencing treatments and grouped the samples based on those metabolites, showing separation of the samples into distinct groups based on treatment. (c) Hierarchical cluster analysis (HCA) computed based on the top 25 statistically significantly different metabolite features grouped the samples into two fencing treatment‐based groups. The relative concentration of each metabolite is indicated by the color scale, and the relationship among the samples is indicated by the color‐coded dendrogram on the top of the HCA plot.

FIGURE 6

Untargeted metabolomic analysis of fenced and unfenced Lindera benzoin samples. (a) Principal component analysis (PCA) of all metabolite features did not group the samples into two clusters that correspond with the fencing treatment. (b) Partial least squares discriminant analysis (PLS‐DA) identified the top 15 metabolites that accumulated significantly differently among the fencing treatments and grouped the samples based on those metabolites, showing separation of the samples into distinct groups based on treatment. (c) Hierarchical cluster analysis (HCA) computed based on the top 25 statistically significantly different metabolite features grouped the samples into two fencing treatment‐based groups. The relative concentration of each metabolite is indicated by the color scale, and the relationship among the samples is indicated by the color‐coded dendrogram on the top of the HCA plot.

FIGURE 7

Untargeted metabolomic analysis of fenced and unfenced Rosa multiflora samples. (a) Principal component analysis (PCA) of all metabolite features did not group the samples into two clusters that correspond with the fencing treatment. (b) Partial least squares discriminant analysis (PLS‐DA) identified the top 15 metabolites that accumulated significantly differently among the fencing treatments and grouped the samples based on those metabolites, showing separation of the samples into distinct groups based on treatment. (c) Hierarchical cluster analysis (HCA) computed based on the top 25 statistically significantly different metabolite features grouped the samples into two fencing treatment‐based groups. The relative concentration of each metabolite is indicated by the color scale, and the relationship among the samples is indicated by the color‐coded dendrogram on top of the HCA plot.

FIGURE 8

Untargeted metabolomic analysis of fenced and unfenced Euonymus alatus samples. (a) Principal component analysis (PCA) of all metabolite features did not group the samples into two clusters that correspond with the fencing treatment. (b) Partial least squares discriminant analysis (PLS‐DA) identified the top 15 metabolites that accumulated significantly differently among the fencing treatments and grouped the samples based on those metabolites, showing some separation of the samples into groups based on treatment. (c) Hierarchical cluster analysis (HCA) computed based on the top 25 statistically significantly different metabolite features did not clearly group the samples into two fencing treatment‐based groups. The relative concentration of each metabolite is indicated by the color scale, and the relationship among the samples is indicated by the color‐coded dendrogram on top of the HCA plot.