Interactions between White-Tailed Deer and Invasive Plants in North American Forests Effects of deer on the photosynthetic performance of invasive and native forest herbs

Overabundant generalist herbivores can facilitate non-native plant invasions, presumably through direct and indirect modifications to the environment that affect plant performance. However, ecophysiological mechanisms behind ungulate-mediated plant invasions have not been well-studied. At a long-term Odocoileus virginianus (white-tailed deer) exclusion site in a temperate deciduous forest, we quantified deer-mediated ecophysiological impacts on an invasive biennial Alliaria petiolata (garlic mustard) and two palatable native herbaceous perennials, Maianthemum racemosum and Trillium grandiflorum. In mid-summer, we found that leaf-level light availability was higher in unfenced areas compared with areas fenced to exclude deer. Alliaria in unfenced areas exhibited 50 % higher mean maximum photosynthetic rates compared with fenced areas. Further, specific leaf area decreased by 48 % on average in unfenced areas, suggesting leaf structural responses to higher light levels. Similarly, Maianthemum had 42 % higher mean photosynthetic rates and 33 % decreased mean specific leaf area in unfenced areas, but these functional advantages were likely countered by high rates of deer herbivory. By contrast, Trillium exhibited significantly lower (26 %) maximum photosynthetic rates in unfenced areas, but SLA did not differ. Deermediated differences in light saturated photosynthetic rates for all three species were only significant during months with overstory tree canopy cover, when light availability in the herb layer was significantly lower in fenced areas. Alliaria’s enhanced photosynthetic rates implicate overabundant deer, a situation that is nearly ubiquitous across its invaded range. Collectively, our results provide empirical evidence that generalist herbivores can alter non-native plant physiology to facilitate invasion.


Introduction
The globalization of human activities has led to widespread movement of plants around the world (van Kleunen et al. 2015) and species invasions are now widely appreciated as a major source of environmental change (Vitousek et al. 1997). In efforts to mechanistically understand, predict, and prevent invasions in an era of ecological change, recent research on plant invasions has focused on functional traits of non-native species that may make them invasive (e.g. van Kleunen et al. 2010) and characteristics of communities that may make them susceptible to invasion (Rejmanek et al. 2013). Habitats characterized by frequent disturbance or altered disturbance regimes have long been considered more susceptible to plant invasions (e.g. Elton 1958;Hobbs and Huenneke 1992). Many invaders come from regions with a long history of human disturbance (e.g. Fridley 2008) and exhibit specific trait adaptations that may allow them to capitalize on high resource availabilities in disturbed environments (Alpert et al. 2000;Davis et al. 2000;Daehler 2003). In addition to altering resource availability, human-mediated disturbances can reduce competition from natives that facilitates invasion by disturbance-adapted species (Daehler 2003).
Untangling the historical, environmental and ecological factors that dictate plant invasion success in disturbed environments remain a significant research challenge, especially for ecosystems stressed by several potentially interacting factors (D avalos et al. 2015a). Temperate forest ecosystems have been subjected to multiple biotic stressors worldwide, including historically unprecedented increases in ungulates (e.g. deer, elk, moose, goats, feral pigs; Singer et al. 1984;Wardle et al. 2001;Côté et al. 2002). In eastern North America, whitetailed deer (Odocoileus virginianus) densities have risen far above what has been considered to be compatible with forest health and regeneration (e.g. Diefenbach et al. 1997), driving further ecological change and associated conservation concerns (McShea 2012). For example, deer can have direct consumptive effects through selective foraging (e.g. Wiegman and Waller 2006; Mason et al. 2010;Averill et al. 2016), indirect effects on biotic and abiotic properties (Rooney and Waller 2003;Heckel et al. 2010;Kardol et al. 2014;Heckel and Kalisz 2017) and can increase exotic seed dispersal (Myers et al. 2004). Together, these effects have resulted in significant and negative community-level impacts (see Habeck and Schultz 2015 for a meta-analysis). Although traditionally studied as separate phenomena, a growing literature has linked deer overabundance to increases in fitness and abundance of invasive plant species (Vavra et al. 2007;Eschtruth and Battles 2009;Knight et al. 2009;Christopher et al. 2014;Kalisz et al. 2014;D avalos et al. 2015a;Shen et al. 2016).
In a temperate deciduous forest in North America, we investigated the effects of overabundant deer on the ecophysiological performance of invasive Alliaria petiolata (Brassicaceae, garlic mustard; hereafter Alliaria) and two commonly co-occurring native species preferred by deer as forage, Trillium grandiflorum and Maianthemum racemosum. Kalisz et al. (2014) recently reported strong population-level demographic declines in Alliaria at this site where deer were excluded. Declines were so strong that in the absence of deer, population growth rates fell below replacement (k <1), which if maintained, projects eventual local extinction of this invader where deer are excluded. Comparable trends in abundance have been reported for additional invaders in other deciduous forests (Eschtruth and Battles 2009;D avalos et al. 2015a;Shen et al. 2016). While other studies have investigated potential physiological impacts of Alliaria on native species (Hale et al. 2011(Hale et al. , 2016 and Alliaria photosynthetic physiology in general (Dhillion and Anderson 1999;Anderson and Cipollini 2013;Stinson and Seidler 2014), the physiological basis for deer-mediated plant invasions remains unexplained.
We hypothesize that persistent browsing by overabundant deer on palatable herbaceous species (Knight et al. 2009) significantly increases light availability, which favours forest edge-or gap-adapted species like Alliaria. To assess this hypothesis, we measured in situ leaf-level ecophysiology of Alliaria and two native perennials, including metrics of leaf photosynthetic performance and specific leaf area. Where deer were allowed access (unfenced areas), we predicted increased photosynthetic rates for Alliaria due to deer-induced disturbances that increase light availability (deer facilitation). In contrast, we predicted the opposite pattern for the palatable natives due to high rates of deer herbivory (which can exceed 75 % of flowering stems; Kalisz et al. 2014) and negative non-consumptive effects (Heckel et al. 2010) that likely constrain their physiological performance. Further, as native understory species are presumably well-adapted to low light levels in the forest interior, we hypothesized that any increases in light availability would be more beneficial to the more light-demanding Alliaria than relatively shade tolerant natives.

Study site and focal species
The study was conducted in a closed canopy deciduous forest in south western Pennsylvania, USA (Trillium Trail Nature Reserve: 40.5201 N;79.9010 W). Trillium Trail is a 16 ha tract of forest within >125 ha of forest and parkland in Pittsburgh, PA. It has a diverse herbaceous understory that has been managed by the Borough of Fox Chapel since 1949. Dominant overstory species include Acer saccharum (sugar maple) and Fagus grandifolia (American beech), as well as Fraxinus spp. (ash), Carya spp. (hickory), Quercus alba and Q. rubra (white and red oak), Prunus serotina (black cherry) and Liriodendron tulipifera (tulip poplar).
Trillium Trail provides an appropriate site for testing ecological hypotheses on deer, plant invasions and their interactions for several reasons. Deer densities at the site rapidly increased in the early 1990s, resulting in annual browse rates on palatable herb species rising as high as 100 %, especially for flowering individuals (Knight et al. 2009). Winter aerial flyovers from 1993 to 2004 confirm high deer densities at an estimated 20-42 deer per km 2 compared with historic estimates of 10-12 deer per km 2 (Kalisz et al. 2014). In addition, the diverse understory is being actively invaded by several woody and herbaceous exotic species. In particular, the non-native biennial herb A. petiolata is now common throughout the forest understory, but this invasion was relatively recent, as Alliaria was reported as rare or nearly absent before 1990 (Knight et al. 2009).
In 1999, the Borough of Fox Chapel erected a 2.5 m tall fence around a 10.06 ha portion of the forest in an effort to protect the understory diversity of Trillium Trail from deer browse (see Knight et al. 2009). Fencing (15 cm Â 15 cm steel mesh) excluded deer, but allowed other animals free access. These fenced and unfenced areas are in the same forest and have similar vegetation to the long-term demographic plots in Kalisz et al. (2014). To avoid damage to the long-term demographic plots, we only studied plants inside and outside of the adjacent larger fenced area. Therefore, our results can be compared with demographic results in Kalisz et al. (2014), but we did not measure the same individuals.
To compare the effects of deer on Alliaria to those of resident species, we also studied two native perennial herbs that commonly co-occur with Alliaria at the site, M. racemosum (Asparagaceae, false Solomon's seal) and T. grandiflorum (Melanthiaceae, white/large-flowered trillium). Other Trillium species are found in adjacent areas of the forest (e.g. T. erectum var. album) and non-flowering individuals are difficult to distinguish. However, T. grandiflorum is the dominant Trillium species in the areas of the forest we sampled. As such, we assume that all individuals we measured were T. grandiflorum.

Leaf gas exchange
We measured individuals of the three species from late July 2014 to June 2016, from both fenced and unfenced areas. Individuals were selected haphazardly each day, alternating the daily order that species were measured to prevent any bias by time of day. This approach maximized sample size while permitting inferences on population trait means through the year. For each species, all measured individuals were separated by at least 2 m. For Trillium and Maianthemum, only adults were measured. For Alliaria, 1st year rosettes (i.e. not bolting) were measured in July-March, and 2nd year adults (i.e. bolting/ flowering) were measured in April-June (senescence).
Gas exchange measurements were performed in situ on intact mature leaves using two LI-6400XT portable photosynthesis systems equipped with a CO 2 control module, 2 cm Â 3 cm leaf cuvette and a red-blue LED light source (LI-COR, Lincoln, NE, USA). All leaves filled the leaf cuvette entirely. Measurements were made under ambient temperature and humidity, with reference chamber CO 2 concentration at 400 lmol mol À1 . Leaves were photo-induced at a moderate irradiance level (200-400 lmol photons m À2 s À1 ) until they equilibrated. Light levels were then progressively increased until leaves were light saturated (800-1500 lmol photons m À2 s À1 ) prior to employing the measurement routines described below. All individuals were light saturated at the highest light levels, with no apparent signs of photoinhibition.
Seasonal measurements. For each species, lightsaturated photosynthetic rate (A sat ) was recorded after equilibration to the measurement light level for at least 2 min at high photosynthetic photon flux density (PPFD; 800-1500 lmol photons m À2 s À1 ) and reaching defined stability parameters based on photosynthetic rate and stomatal conductance. Measurements were taken at least once per month, under ambient leaf temperature and humidity. As a biennial with green leaves throughout the winter, Alliaria was measured in all months except February and March (due to snow cover and low temperatures). The native perennials (both deciduous) were measured from leaf expansion in spring until leaf senescence (Maianthemum: May-August; Trillium: April-June). Cumulatively, these gas exchange measurements resulted in >2800 logged observations over 37 days: 240 total Alliaria individuals (mean number of individuals measured per month 6 1 SE: 22 6 4), 125 Maianthemum (29 6 8) and 79 Trillium (11 6 6).
Midseason light and CO 2 response curves. We measured leaf photosynthetic responses to irradiance (PPFD) in 10-12 steps from 0 to 1500 lmol photons m À2 s À1 . Net photosynthetic rate was recorded after allowing the leaf area inside the cuvette to equilibrate for at least 2 min at each PPFD and achieving defined stability parameters described above. Leaf temperature was maintained at 25 C under ambient humidity.
Light response curve parameters were estimated through non-linear least squares regression of a non-rectangular hyperbola (Marshall and Biscoe 1980) [see Supporting Information-File S1]: where A net and A max are the area-based net and maximum gross photosynthetic rates (lmol CO 2 m À2 s À1 ), respectively, / is the apparent quantum yield (mol CO 2 mol photons À1 ), R d is daytime dark respiration rate (jA net j at no light; lmol CO 2 m À2 s À1 ) and h is curve convexity (dimensionless). Light compensation point (LCP) was calculated as the x-axis intercept, and light saturation point (LSP) as the PPFD when 90 % of A max (model asymptote) was achieved.
After light response curve measurement, leaves were re-stabilized at the saturated light level to measure photosynthetic responses to changing intercellular CO 2 pressure (photosynthetic rate vs. intercellular CO 2 partial pressure; A/C i curves). A/C i measurements were taken from the same leaves (or from adjacent nodes) as those for light response curves. While maintaining saturating light levels (800 lmol photons m À2 s À1 ), reference chamber CO 2 concentration was varied in 11 steps (400, 300, 200, 100, 50, 400, 400, 600, 800, 1000, 1500 lmol CO 2 mol À1 ). Net photosynthetic rates were recorded at each step following equilibration, as described above.
The biochemical photosynthetic model developed by Farquhar et al. (1980) was fit using non-linear least squares regression and the simultaneous estimation method described by Dubois et al (2007): where A net is the net photosynthetic rate, A v and A j are the CO 2 assimilation rates limited by RuBisCO (Ribulose-1,5-bisphosphate carboxylase/oxygenase; carboxylation) and RuBP (Ribulose-1,5-bisphosphate; substrate regeneration), respectively, and R day is the daytime mitochondrial respiration rate. RuBP-saturated (CO 2 limited) photosynthesis was modelled as: where C i is intercellular CO 2 partial pressure (Pa), K c and K 0 are Michaelis-Menten constants for carboxylation and oxygenation (40.4 Pa and 24.8 kPa, respectively), O is the O 2 concentration (21 kPa), V c,max is maximum carboxylation rate (lmol CO 2 m À2 s À1 ), R day is the daytime mitochondrial respiration rate (lmol CO 2 m À2 s À1 ) and C* is the CO 2 compensation point in the absence of mitochondrial respiration (3.7 Pa). RuBisCO kinetic constants were obtained from von Caemmerer (2000). RuBP-limited photosynthesis was modelled as: where J max is the light saturated maximum rate of photosynthetic electron transport.

Specific leaf area
On 29 June 2015 (Alliaria and Maianthemum) and 22 June 2016 (Trillium), mature leaves of 19-100 individuals of each species (Alliaria: n ¼ 50 1st year rosettes, 50 adults; Maianthemum, n ¼ 50; Trillium, n ¼ 19) were sampled by walking three parallel transects and haphazardly collecting 2-5 leaves per individual and ensuring individuals of each species were separated by at least 2 m. Entire leaves were sampled except for Trillium, where leaf samples of fixed area (5.32 cm 2 ) were taken. Fresh leaves were scanned and leaf area determined using ImageJ (Schneider et al. 2012). Leaves were oven dried at 60 C for at least 48 h. Specific leaf area (cm 2 g À1 ) was calculated as the leaf surface area per g dry mass.

Understory light availability
To determine if light levels were different inside versus outside of the fenced area, we used light measurements (PPFD) taken at the height of the individual plants we studied (5-60 cm) that were recorded concurrently with gas exchange measurements with an external LI-COR quantum sensor on the chamber head.

Data analysis
All statistical analyses were performed in R (R Core Team 2015) [see Supporting Information-Tables S1-S3 for trait data]. When necessary, data were log-transformed to meet normality and homoscedasticity assumptions for model residuals.
Comparing effects of fencing on leaf traits. For univariate traits, we fit linear mixed-effect models separately for each species (Bates et al. 2011). In models for midseason photosynthetic and SLA data, individual and measurement date were random effects. For light response and CO 2 response curve parameter fits, values were weighted by the precision of the estimates (standard error À1 ). We analysed A sat data using separate ANOVAs for each species with fixed factors of fencing (fenced or unfenced) and month. We tested for differences within each month with t-tests implemented through the multcomp package in R (Hothorn et al. 2008).
Determining canopy closure. To quantify canopy closure, we measured understory PPFD with a quantum sensor (Onset Computer Corporation, MA, USA) at a fixed location within the fenced deer exclusion area [see Supporting Information- Fig. S1]. We then modelled mean daily PPFD (10:00-14:00 h) from April-July using a generalized additive model (GAM), implemented through the mgcv package in R (Wood 2016). After complete canopy closure, light levels are likely to vary little between days. We calculated the derivative of the GAM regression line and its 95 % confidence interval to estimate the first day when the slope of the regression was not significantly different from zero, indicating complete overstory canopy closure [see Supporting Information- Fig. S2].
Modelling impact of deer exclusion on understory light availability. For each plant, we calculated mean PPFD conditions experienced while the gas exchange data were collected. On the basis of our analysis of the process of canopy closure (described above), we used data collected between mid-May and late June to determine how deer exclusion affected light conditions before and after complete canopy closure. We modelled log(mean PPFD) as a function of day of year (continuous), fencing (unfenced vs. fenced) and a (day of year) Â (fencing) interaction using the lme function in the nlme package in R (Pinheiro et al. 2016). Day was included as a random effect. Estimated variances were allowed to change as the growing season progressed and also between fenced and unfenced treatments using the varPower function in lme. We tested the significance of model terms using likelihood ratio tests of nested models. The best model was re-fit without the log transformation to report estimates on the original scale.
Midseason light and CO 2 response curves. Detailed light response curves for Alliaria rosettes measured during a single point mid-summer (July) show clear photosynthetic differences between fenced and unfenced areas, with greater gross maximum photosynthetic rates (A max ) in unfenced areas, as well as a significantly greater mean LSP (64 % greater; Fig. 3A, Table 2A). Likewise, both maximum carboxylation capacities (V c,max ) and Figure 2. Light saturated photosynthetic rates (A sat ) for (a) Alliaria petiolata, (b) Maianthemum racemosum and (c) Trillium grandiflorum in unfenced (red triangles, solid line) and fenced (blue circles, dashed line) areas. Points denote monthly mean 6 1 SE. Grey region illustrates approximate period of overstory canopy shading. Asterisks indicate statistical significance for differences between fenced and unfenced by month: *P < 0.05; **P < 0.01; ***P < 0.001.  Table 2. Mean trait values (61 SE) in fenced and unfenced areas for (a) Alliaria, (b) Maianthemum and (c) Trillium. Statistical differences were determined using likelihood ratio tests (v 2 with 1 df) that compared models for each trait with and without deer fencing fixed effect (unfenced or fenced) and random effects of individual and/or measurement date. rates of electron transport (J max ) were significantly higher in unfenced versus fenced areas (Table 2A). Mid-season response curves for Maianthemum showed qualitatively similar differences between fenced and unfenced areas, as found for Alliaria (Fig. 3B, Table 2B). In particular, A max LSP, LCPs, and apparent quantum yield were greater in unfenced compared with fenced plants. However, V c,max and J max did not vary significantly.
Unlike Maianthemum and Alliaria, Trillium exhibited significantly lower A max in unfenced compared with fenced areas (Fig. 3C) but no other measured parameters differed by treatment (Table 2C).
Specific leaf area. Impacts on leaf morphology varied between adult and rosette leaves for Alliaria, so we report them separately. Mean specific leaf area (SLA) for rosette leaves was significantly higher in fenced compared with unfenced areas, and these leaves were also significantly larger (Table 2A). However, Alliaria adults did not differ in SLA or leaf size between fenced and unfenced areas. Like Alliaria rosettes, Maianthemum had greater SLA and larger leaves on average in fenced areas (Table 2B). By contrast, SLA did not differ for Trillium (Table 2C).

Discussion
Despite the growing body of evidence that invasion can be strongly modulated by overabundant ungulate herbivores (Vavra et al. 2007;Eschtruth and Battles 2009;Knight et al. 2009;Kalisz et al. 2014;Christopher et al. 2014;D avalos et al. 2015a;Shen et al. 2016), many studies on the functional ecology of plant invasions do not consider these effects. Here, we investigated the potential physiological consequences of deer exclusion on Alliaria. We found pronounced differences in the photosynthetic performance of this invader between fenced and unfenced areas. Notably, during summer months, Alliaria had a $40 % reduction in mean light-saturated photosynthetic rates (A sat ) and 93% increase in SLA when deer were excluded. Reduced A sat and increased SLA are both known responses to shade (Valladares and Niinemets 2008). The palatable native perennials differed in response to deer exclusion: Maianthemum showed deer-mediated photosynthetic increases while Trillium exhibited nearly the opposite pattern. While our data suggest the importance of deer on invasive plant performance, we did not measure additional factors that might also contribute to these responses, including earthworm invasion (Nuzzo et al. 2009) and the interaction between earthworms, deer and plant invasions (D avalos et al. 2015a, b).
Why did Alliaria exhibit strong physiological advantages due to overabundant deer? In line with our predictions, Alliaria photosynthetic rates were congruent with previous demographic results from the study site, where deer not only facilitate, but enable, positive population growth rate (Kalisz et al. 2014). Knight et al. (2009) reported relatively greater area unoccupied by herbaceous vegetation in unfenced (deer access) plots, which suggested declines in the native understory canopy and presumably higher understory light levels. The current study quantitatively confirmed this idea, with a nearly threefold mean increase in leaf-level light availability where deer were present. We propose that these deerassociated differences in understory light availability are due to high browse pressure on palatable natives (Kalisz et al. 2014), as well indirect effects on other less preferred plant species in the community (Rooney and Waller 2003;Heckel et al. 2010).
The relative importance of direct and indirect deer effects versus other factors remains unknown. Additionally, we only measured instantaneous PPFD at the leaf-level of each target plant, so the relative contributions by the overstory, mid-story, shrub or herb layer to these differences in light availability remain unclear. Given deer have been excluded for only 15 years at this site and the browse line is typically <2 m, we assume the overstory canopy is similar in fenced and unfenced areas where we took our measurements. Further research is needed to confirm this assumption at this site. In a study of 44 exclosures in the eastern US, Rohleder (2013) found no difference in overstory canopy cover between fenced and unfenced plots, but the density of vegetation <2 m was significantly greater in fenced plots.
Alliaria's trait responses to deer exclusion closely matched expectations from the ecophysiological literature on trait responses and adaptations to shade (Valladares and Niinemets 2008). First, decreased areabased photosynthetic rates (A max , A net ), carboxylation rates (V c,max ) and electron transport rates (J max ), as we found for Alliairia in fenced areas during summer, are characteristic of shade tolerant species (Valladares and Niinemets 2008) and within species, individuals acclimated to low light conditions (Evans and Poorter 2001). Further, both SLA and leaf area were greater in fenced areas, as expected for forest herbs under low light levels (Rothstein and Zak 2001). In both shade tolerant and intolerant species, increasing SLA is a common response to low light, which is thought to maximize carbon gain per unit mass by increasing light capture area (Evans and Poorter 2001). We did not measure soil nutrients or leaf chemistry (e.g. leaf N) and these factors could potentially differ between fenced and unfenced areas. Collectively, Alliaria's traits in fenced areas are indicative of a shift towards a more shade tolerant strategy, although its growth and reproduction are substantially reduced in low light environments (Stinson and Seidler 2014). Trait differences between fenced and unfenced areas are likely due to acclimation to growth conditions (phenotypic plasticity) rather than genetic changes. However, if deer-mediated selection pressures remain strong, there is potential for evolutionary change (Heckel and Kalisz 2017).
Light availability is often the dominant limiting factor for plant growth in the forest understory (Finzi and Canham 2000) and the main resource axis that differentiates ecophysiologicial strategies of forest herbs (Neufeld and Young 2014). The abundance of invasive species in forested habitats has been empirically linked to higher light environments (e.g. Dreiss and Volin 2013). While shade tolerant strategies can permit invasion and persistence in forest understories (Martin et al. 2009), both natives and non-native invaders in the understory can benefit from increased light (Heberling and Fridley 2016). Alliaria has highest fitness in forest edge habitats but can survive and reproduce in a variety of light environments (Meekins and Mccarthy 2001). Its invasion is particularly puzzling as it can invade seemingly intact forest understories (Nuzzo 1999;Rodgers et al. 2008) where light availability is relatively low. We suggest that overabundant deer create light conditions in the forest interior similar to edge habitats. Indeed, previous studies report positive correlations between Alliaria photosynthetic rates and irradiance (Dhillion and Anderson 1999;Myers and Anderson 2003;Myers et al. 2005). One hypothesis for Alliaria's success focuses on its 'extended' phenology as a biennial, compared with the native species in communities it invades (Engelhardt and Anderson 2011;Smith and Reynolds 2015). This 'extended phenology' hypothesis emphasizes the importance of Alliaria's ability to capitalize early-and late-season light levels for invasion success. Further, Stinson and Seidler (2014) found Alliaria growing in forest edge habitat conditions achieve higher photosynthetic rates, lower SLA, and higher fecundity than those in the forest interior (low light). In a common garden, they found that these physiological differences between forest edge and interior populations are due to plasticity and conclude that genetic adaptation to shade in this species might be constrained by trade offs with plastic trait responses that optimize fitness at higher light availabilities. Our results broaden these prior findings to connect light-associated advantages for Alliaria's physiology to deer overabundance, a phenomenon that is nearly ubiquitous across its invaded range (McShea 2012) yet only briefly mentioned in a major review of factors driving Alliaria's invasive success (Rodgers et al. 2008). Similar to responses in Alliaria, the native perennial Maianthemum exhibited higher photosynthetic performance and lower SLA compared with fenced plants. These functional shifts toward a high light-adapted plant strategy are expected when considering light-level differences, as this result is not due to deer per se but rather deer-mediated indirect effects via differences in light availability. The contrasting result seen in Trillium, which displayed lower photosynthetic rates in June in unfenced compared with fenced areas, is surprising. As both Trillium and Maianthemum are presumably light-limited under closed canopies (Neufeld and Young 2014), both species should benefit from increased understory light availability. However, June leaf-level PPFD data taken concurrently with gas-exchange data for Trillium indicated the opposite pattern, with PPFD slightly lower in unfenced areas (mean June PPFD 6 SE: 34 6 4 lmol photons m À2 s À1 vs. 23 6 8 lmol photons m À2 s À1 , for fenced vs. unfenced areas, respectively). (Note: Trillium was the only species measured in June 2016, so these PPFD data should be interpreted with caution). One explanation for the decrease in photosynthetic rates of Trillium in the presence of deer could be that the species is less phenotypically plastic with respect to its leaf physiology than Maianthemum, as suggested by its slightly earlier spring emergence, mid-summer senescence and lack of SLA signal. As a result, C gain in late spring, after canopy closure when understory light levels substantially decrease, might be constrained such that light availability is not the factor explaining the difference in photosynthetic rates between fenced and unfenced areas. Additionally, Trillium is a highly preferred forage of deer and may be more likely to be browsed when growing in microsites with higher light availability than Maianthemum. Therefore, our sampling of Trillium in unfenced areas may have overrepresented relatively shadier microsites despite deer access. Another possibility is that Trillium recruitment was biased toward high light microsites where deer have access. More data are required to fully address these hypotheses, including measurements of plant size, age, and herbivory and flowering history. Together, our data suggest that while deer exclusion influence plant ecophysiology, the response can differ by species.
Previous research has strongly linked deer to Alliaria invasions (e.g. Kalisz et al. 2014), but does the species' widespread invasive success require deer? There are many additional factors that are increasingly appreciated as causing ecosystem modification that can affect invasion success, including invasive earthworms (e.g. D avalos et al. 2015a). Invasive earthworms are ubiquitous at our study site and have been found to play an important role in facilitating plant invasion (Nuzzo et al. 2009). Further, it is important to note that our study does Heberling et al not provide direct, causal evidence that deer are solely responsible for increases in light availability and the ecophysiological differences we report, or that light availability alone fully explains differences in photosynthetic rates between fenced and unfenced areas. Additionally, the intensity and ecological importance of species' responses to deer exclusion will likely differ based on covarying, idiosyncratic factors such as community context, competition, forest successional stage, local deer abundance, landscape conditions and disturbance history. However, in combination with demographic models and population-level studies (Eschtruth and Battles 2009;Kalisz et al. 2014), our new results highlight the potential importance of deer for plant invasion. At our study site, deer were only recently overabundant (since $1992), and Alliaria was rare prior to this deer overabundance (Knight et al. 2009

Conclusions
Ecosystems subjected to frequent disturbance have long been thought to be more susceptible to invasion (Elton 1958;Hobbs and Huenneke 1992). However, seemingly undisturbed forests are not immune to invasion, especially by shade-tolerant non-native species (Martin et al. 2009), and all forests are subject to anthropogenic disturbance to some degree. Modern forests are commonly subject to multiple ecological stressors (D avalos et al. 2015a), including modifications to the abiotic and biotic environment modulated by deer overabundance. In the present study, we found the presence of deer can modify seasonal light availability in the understory herb layer. As a result, invasive Alliaria expressed a sun-adapted physiological strategy when deer were present. Although deer-mediated facilitation in Alliaria supports the 'passengers' model of species invasions (MacDougall and Turkington 2005), previous studies at the same site show that, even when deer are present, Alliaria removal can significantly benefit native plant vital rates (Brouwer et al. 2015).
Overabundant ungulates act as a strong biotic force that result in significant interspecific shifts in phylogenetic diversity (Begley-Miller et al. 2014) and communityweighted functional trait composition (Mason et al. 2010). Little research has been published on intraspecific ecophysiological differences mediated by generalist mammalian herbivores. In line with previous studies (e.g. Knight et al. 2009;Kalisz et al. 2014), this link between deer presence and individual plant performance highlights the need for management of deer to control invasions, particularly by plant species requiring higher light, openings in the understory canopy or disturbance.

Conflicts of Interest Statement
No conflicts of interest.

Supporting Information
The following additional information is available in the online version of this article-File S1. Annotated R code used for fitting photosynthetic models to response curve data. Table S1. Light response curve summary data. Table S2. A/C i curve summary data. Table S3. SLA data. Figure S1. Understory light level data used to estimate date of overstory canopy closure. Figure S2. Derivative approach to estimate date of spring overstory canopy closure.