The influence of warming on the biogeographic and phylogenetic dependence of herbivore–plant interactions

Abstract Evolutionary experience and the phylogenetic relationships of plants have both been proposed to influence herbivore–plant interactions and plant invasion success. However, the direction and magnitude of these effects, and how such patterns are altered with increasing temperature, are rarely studied. Through laboratory functional response experiments, we tested whether the per capita feeding efficiency of an invasive generalist herbivore, the golden apple snail, Pomacea canaliculata, is dependent on the biogeographic origin and phylogenetic relatedness of host plants, and how increasing temperature alters these dependencies. The feeding efficiency of the herbivore was highest on plant species with which it had no shared evolutionary history, that is, novel plants. Further, among evolutionarily familiar plants, snail feeding efficiency was higher on those species more closely related to the novel plants. However, these biogeographic dependencies became less pronounced with increasing temperature, whereas the phylogenetic dependence was unaffected. Collectively, our findings indicate that the susceptibility of plants to this invasive herbivore is mediated by both biogeographic origin and phylogenetic relatedness. We hypothesize that warming erodes the influence of evolutionary exposure, thereby altering herbivore–plant interactions and perhaps the invasion success of plants.

biotic resistance hypothesis (Levine, Adler, & Yelenik, 2004). In contrast, if generalist herbivores prefer familiar plants (Agrawal et al., 2005;Liu & Stiling, 2006;Xiong, Dan, Wang, Liu, & Wang, 2008), plant invasion may be promoted by a suppression of native competitors, consistent with the enemy release hypothesis (Colautti, Ricciardi, Grigorovich, & MacIsaac, 2004;Keane & Crawley, 2002). A meta-analysis found that native generalist herbivores preferentially consumed exotic plants (Parker & Hay, 2005), whereas exotic herbivores preferentially consumed native plants (Parker, Burkepile, & Hay, 2006), conferring a competitive advantage to invading plants (Parker et al., 2006) through indirect facilitation (Simberloff & Von Holle, 1999). Such evolutionary mismatches render naïve plants particularly susceptible to herbivory (hereafter termed the "novel interaction hypothesis" (Buckley & Catford, 2016;Carthey & Banks, 2014;Saul & Jeschke, 2015;Verhoeven et al., 2009)). However, studies testing these patterns typically used only a dichotomic analysis In addition to direct evolutionary exposure, herbivory damage may also be affected by plant phylogeny (Craft, Paul, & Sotka, 2013;Hill & Kotanen, 2009;Ness, Rollinson, & Whitney, 2011;Pearse & Hipp, 2009). However, the direction and magnitude of the relationship between plant phylogeny and herbivory pressure are still highly debated. One might hypothesize that exotic plants closely related to natives are more likely to be recognized and attacked by native herbivores and thus encounter resistance to invasion (Levine et al., 2004). Alternatively, exotic plants closely related to natives may be preadapted to existing conditions of herbivory. On the other hand, a generalist herbivore may choose plants based on traits such as nutrition content, regardless of the phylogenetic relatedness between exotic and native plants (Pearse & Hipp, 2009). Testing the feeding efficiencies of exotic herbivores across a phylogenetic gradient of plants may help resolve these opposing hypotheses.
Increasing temperatures could influence consumer-resource interactions with respect to coevolutionary history (Diamond & Kingsolver, 2012). For example, Diamond and Kingsolver (2012) demonstrated that herbivore populations with different evolutionary exposures to host plants varied in their responses to these plants under different temperature. Studies have also linked altered temperature to the movement (Walther et al., 2009), establishment success (Chown et al., 2012), and spread (Stachowicz, Terwin, Whitlatch, & Osman, 2002) of species beyond their natural ranges.
Recently, variation in temperature has been shown to influence competition, insect-plant, and predator-prey interactions involving native and exotic species (Fey & Cottingham, 2012;Fey & Herren, 2014;Lu, Siemann, Shao, Wei, & Ding, 2013). For example, Fey and Herren (2014) found that increasing temperature disproportionately benefited an exotic species compared to a native congener under threat from a shared native predator, resulting in a temperature-dependent enemy release. If increased temperature similarly shapes novel herbivore-plant interactions, it may mediate plant invasion and impact.
In this study, we conducted functional response (FR) experiments to examine how the biogeographic origin and phylogenetic relatedness of plants influence the feeding efficiency of an exotic herbivore under increasing temperatures. The main advantage of the FR method is its derivation of critical parameters of per capita feeding ability, including attack rate, handling time, and maximum feeding rate, rather than less informative "snapshot" measurements of feeding rates by arbitrarily setting one level of resource . Thus, specifically in this study, we use FRs to test (a) whether an invasive generalist herbivore prefers to feed on evolutionarily novel plants; (b) whether its feeding rate is related to the phylogenetic relatedness of the host plants; and (c) how increasing temperatures affect the effects of plant biogeography and phylogeny on herbivore feeding efficiency.

| Experimental organisms
The golden apple snail (Pomacea canaliculata; Gastropoda, Ampullariidae) is native to freshwater wetlands of South America and has been introduced widely into Asia since the 1980s (Hayes et al., 2015;. In its introduced range, it has caused significant damage to agricultural production (Cowie, 2002), wetland plants (Fang, Wong, Lin, Lan, & Qiu, 2010), and ecosystem functioning (Carlsson, Brönmark, & Hansson, 2004) and is listed among the "Top 100" invasive species by the International Union for Conservation of Nature (Lowe, Browne, Boudjelas, & Poorter, 2000). It is an omnivorous species that feeds predominantly on aquatic and semiaquatic macrophytes (Qiu & Kwong, 2009 In the FR experiment, five plant species that share an evolutionary history with P. canaliculata were chosen as follows: Alternanthera philoxeroides (alligator weed), Eichhornia crassipes (water hyacinth), Ipomoea batatas (sweet potato), Myriophyllum aquaticum (parrot feather), and Pistia stratiotes (water lettuce). In addition, we chose the following five plant species with which P. canaliculata has had no evolutionary exposure: Apium graveolens We collected all host plant species on the day of the experiment to ensure their freshness. Each of these species is common in wetlands and agricultural areas of South China, and the snail was observed to consume all these plants to some degree in the laboratory. Cultivated plants may have higher nutrient content and thus higher palatability, which could potentially confound the effect of nutritional quality with that of the evolutionary novelty; therefore, we tested differences in nutritional quality across the host plant species. We found no evidence that major nutritional (total N) and physical (dry matter content) properties of evolutionarily novel plants differed from those of evolutionarily familiar plants (see Supporting Information).

| Functional response experiment
The FR experiment was conducted in the summer of 2015 in three consecutive experimental blocks. For each block, the experimental units were allocated to five tanks (200 cm × 70 cm × 50 cm) using a split-plot design. The first block started at 08:00 on May 29 and ended at 20:00 on May 31 (60 hr); similarly, the second and third blocks started at 08:00 on June 4 and June 10, respectively, and ended at 20:00 on June 6 and June 12, respectively. Prior to the experiment, the snails were held without food for 24 hr to allow for standardization of hunger levels (Alexander, Dick, Weyl, Robinson, & Richardson, 2014;Xu, Mu et al., 2016). Snails with similar body size were used to minimize variation in the FR data due to body mass effects (mean body mass with shell: 13.04 ± 0.05 g). Fresh leaves of all 10 plants were picked, weighed, and allocated to experiment units. The natural air temperature range over the experimental duration was 22-26°C at nighttime and 26-32°C during daytime.
In the split-plot design, the tank with tap water was the wholeplot experimental unit and temperature was the whole-plot factor.
We had five whole-plot units with water temperature controlled at 26, 28, 30, 32, and 34°C. To control the water temperature, we heated the tank using water tank heaters (mean temperatures during the experiment were 26.39 ± 0.21, 28.45 ± 0.19, 30.53 ± 0.18, 32.34 ± 0.20, and 34.38 ± 0.28°C, respectively). The tank water was maintained at 40 cm height. Within each tank, 70 boxes (12 cm × 10 cm × 6 cm) were used as subplot experimental units with plant species and plant biomass as subplot factors. These boxes consisted of 10 plant species, each having seven biomass gradients (wet weight 1, 2, 4, 6, 8, 10, and 12 g) and randomized with respect to position in tank. In each box, a golden apple snail was introduced.
The experiment thus consisted of 1,050 (3 × 5 × 10 × 7) experimental units. At the end of the experiment, the leaves were removed from the boxes and allowed to dry in air for 3 hr to evaporate surface moisture before they were measured. Gross consumption by the snails was determined by subtracting final weight from initial weight.
In order to characterize the potential variation of plant weights due to imbibition or air drying, we conducted a control experiment under
In a previous study, we successfully characterized the FRs of this herbivorous snail to derive its impacts on plant resources (Xu, Mu et al., 2016). In the context of herbivores, they may rarely suffer searching limitations, owing to concentrated and ubiquitous plant resources (Spalinger & Hobbs, 1992). Thus, herbivores are unlikely to experience the classic hypothesis of competition for time between searching and handling prey (Holling, 1959), rather, cropping and chewing have been viewed as the competing processes for herbivores (Spalinger & Hobbs, 1992). Based on this assumption, and accounting for the "digestive pause" (Holling, 1966) where Ne is the consumed biomass of plant, N is the initial biomass, a is the "attack (cropping) rate", h is the "handling (chewing and digestion) time" for per capita biomass, and T is the total experiment time. This recursive function can be resolved using the Lambert W function (Corless, Gonnet, Hare, Jeffrey, & Knuth, 1996): In the laboratory, we firstly derived the FR parameters and curves for the snail toward each of the 10 plant species at each experimental temperature. The parameters a and h for the FRs were estimated using nonlinear least squares regression and the "lambert W" function of the package "emdbook" (Bolker, 2010 Then, Rogers' random predator equation was used to each data set to construct 95% confidence intervals around the mean FR curves.

| Phylogeny reconstruction
To verify the phylogenetic structure of these species, for each of the 10 species, we searched GenBank for matK gene sequences, which are commonly used in published plant phylogenies (Cadotte, 2013).

| Statistical analyses
Using a restricted maximum likelihood method in the "lme" function of the "nlme" package, for the laboratory experiment, we fitted linear mixed models (LMMs) to the parameters of the FRs by log transformation (Pinheiro & Bates, 2006). In these models, temperature and the origin of species (evolutionarily familiar vs. novel) were used as fixed effects. We tested their main and interaction effects on three parameters derived from the Type II FRs as follows: attack rate, handling time, and estimated maximum feeding rate. The tank was used as a random effect to describe the error structure of a split-plot design (Crawley, 2012). The notation of the model was as follows: where y is the parameter derived from Type II FR model (a, h, or maximum feeding rate), β 0 is the intercept, β temp and β origin are the coefficients associated with the fixed effect variable temperature and origin. term origin characterizes the interaction between these two factors. b is random effect (tank) depicting the error structure of the split-plot design and ε is the remaining variance.
TA B L E 2 Results of linear mixed models predicting the log response of attack rate (a), handling time (h), and maximum feeding rate (max) derived from the Type II functional responses (FR) to temperature × origin and temperature × phylogenetic distance. "Origin" denotes whether plant species have the same biogeographic origin (five familiar species) or different biogeographic origin (five novel species) with respect to the herbivore. "Phylo" denotes the mean phylogenetic distance of each familiar species to five novel species. For FR~temp × origin model, there were 50 samples (FR parameters of 10 species in each tank × 5 tanks). For FR~temp × phylo, there were 25 samples (FR parameters of 5 familiar species in each tank × 5 tanks). In both models, there are five groups (5 tanks and temperature treatment was assigned at each tank), and thus, the tank was used as random effect to account for the autocorrelation within a tank for our split-plot design Then, in order to account for the possible effects of phylogenetic nonindependence among species on our analyses, we also conducted a PGLS (phylogenetic generalized least square regression) to examine the effect of origin and its interactions with temperature. As the PGLS required the data point must be equal to the number of tips in the phylogenetic tree (i.e., only one data point can be included for each species), we first performed the PGLS at each of the five temperature levels. Then, through averaging the data for a given species derived from different temperatures, we conducted another PGLS to test the mean effect of origin. These analyses were complished using the "gls" function in "nlme" package by hypothesizing a Brownian Motion model (Swenson, 2014).
Finally, based on the reconstructed phylogenetic tree, we calculated the mean phylogenetic distance of each familiar species to five novel species using the "cophenetic.phylo" function of the "ape" package.
For example, the phylogenetic distance between the familiar species

| RE SULTS
Pomacea canaliculata had markedly different FRs when consuming the evolutionarily familiar plant species than when consuming the evolutionarily novel plants (Figure 2, Supporting Information Figure S1; lack of 95% CI overlap at all temperatures), with significantly lower attack rates (Table 2, Figure 3a), higher handling times, and lower maximum feeding rates (Table 2, Figure 3b,c).
There was a significant interaction effect of temperature and plant origin for the handling time and maximum feeding rate (Table 2), with higher temperature enhancing the maximum feeding rate of the herbivore when consuming the evolutionarily familiar plants, but not when consuming the evolutionarily novel plants (Table 3, Figure 3b,c). This interaction effect and the main effect of temperature did not exist for the attack rate (Table 2, Figure 3a), indicating that temperature alters the handling efficiency, but not the attack ability, of the herbivore on evolutionarily familiar plants.
After accounting for the phylogeny nonindependence, the results of PGLS were consistent with that of LMMs except for the attack rate. Across all temperature, the evolutionarily familiar plants yielded higher handling time (t = 4.804, p = 0.001) and lower maximum feeding rates (t = 5.535, p = 0.001) than the evolutionarily novel plants, but attack rate was not significantly different between the two origins (t = 0.001, p = 0.999) (Figure 4).
The mean phylogenetic distances of each familiar plant species to the five novel species significantly correlated with the handling time and maximum feeding rate of P. canaliculata toward these familiar F I G U R E 3 The relationships between temperature and attack rate (a), handling time (b), and maximum feeding rate (c) of the invasive herbivore Pomacea canaliculata toward evolutionarily familiar and novel plant species. Points represent the mean values of evolutionarily familiar or novel plant species after log transformation (n = 5). The linear fits come from two-variable (temperature and origin) mixed models. Error bars represent standard errors species (Table 2). Plant species more closely related to novel species yielded less handling time and more herbivory pressure (Table 3, Figure 5b,c). Increasing temperature did not significantly shift this relationship (Table 2, Figure 5b,c). The phylogenetic distance, temperature, and their interaction effect did not affect the attack rate of P. canaliculata (Table 2, Figure 5a).

| D ISCUSS I ON
Pomacea canaliculata consumed evolutionarily novel plants to a greater degree than evolutionarily familiar plants, consistent with the novel interaction hypothesis (Morrison & Hay, 2011;Parker et al., 2006). Warming, however, increased the consumption rate on evolutionarily familiar plants and thus reduced the herbivory difference between evolutionarily novel and familiar plants. The feeding rate of P. canaliculata was higher for plant species more closely related to the palatable novel plants, and this trend was unaffected by warmer temperatures.
Our findings do not support the argument that the biogeographic origin of species has no bearing on its potential ecological impact (Davis et al., 2011;Valéry, Fritz, & Lefeuvre, 2013) and suggest that evolutionary novelty may benefit the herbivorous snail (Verhoeven et al., 2009). Five of the experimental host plants originate from South America, the same biogeographic region as P. canaliculata. A shared evolutionary history may have caused these plants to evolve some resistance to herbivory, whereas the plants that originated from other continents suffered higher consumption perhaps owing to lack of adaptation. Conversely, evolutionary novelty has paradoxically been cited to explain the release from natural enemies, by arguing that enemies have not been selected to counter the novel plant's defenses (Callaway & Aschehoug, 2000;Cappuccino & Arnason, 2006). Distinguishing these two contrasting processes requires consideration of underlying mechanisms such as recognition-based and toxin-based defenses of novel plants (Verhoeven et al., 2009). Among the plants used in our study, perhaps those familiar with P. canaliculata have produced phytochemicals or developed physical traits to reduce the intensity of herbivory by this snail, whereas novel plants may lack such adaptations. We focused on how increased temperature mediates this novel interaction rather than the specific mechanisms underlying it, although the latter facet deserves further exploration.
Temperature reduced the handling time and increased the maximum feeding rate of P. canaliculata when consuming evolutionarily familiar plants but not novel plants. This finding, on the one hand, has implications for the management of exotic species TA B L E 3 Estimated effects of models predicting the log response of attack rate (a), handling time (h), and maximum feeding rate (max) derived from the Type II functional responses to plant origins × temperature and phylogenetic distance × temperature. "Origin" denotes whether plant species have same biogeographic origin (five familiar species) or different biogeographic origin (five novel species) with the herbivore. "Phylo" denotes the mean phylogenetic distance of each familiar species to five novel species An impediment to assessing and comparing the ecological impacts of invasions is the lack of standardized methods for F I G U R E 5 The relationships between phylogenetic distance and attack rate (a), handling time (b), and maximum feeding rate (c) of five familiar species at 26, 28, 30, 32, and 34°C. The linear fits come from two-variable (phylogenetic distance and temperature) mixed models. Phylogenetic distance denotes the mean phylogenetic distance of each familiar species to five novel species measuring and interpreting per capita effects Ricciardi, Hoopes, Marchetti, & Lockwood, 2013). By using FR experiments, we can better depict resource use and avoid the many pitfalls of "snapshot" assessments where resource levels are arbitrarily fixed Xu, Mu et al., 2016). In our study, both handling time and maximum feeding rate varied with plant origin and phylogeny, and the effect of origin was mediated by increasing temperature. However, attack rate was only affected by origin, but not by plant phylogeny and temperature. To explain this result, it would be necessary to explore the feeding behavior of the herbivore and bioenergetic tradeoffs (Jeschke et al., 2002;Spalinger & Hobbs, 1992). These distinct responses may indicate that the invertebrate herbivore has not yet evolved the ability to identify and distinguish temperature-altered phytochemicals and therefore attacks plants without much discrimination while still exhibiting differential handling efficiencies. Alternatively, warming may enhance enzyme activity related to handling efficiency, but not attack behavior. In any case, the observed changes of handling time and maximum feeding rate indicate that plant origin and phylogenetic relationship can affect invasive herbivore-plant interactions, and such relationships are likely to be altered with increasing temperature.

ACK N OWLED G M ENTS
We thank Shaopeng Li, Xubing Liu, Nancai Pei, and Hongmei Song for

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
XDM and MX designed the experiments and wrote the first draft of the manuscript. XDM, MX, DL, and HW conducted the experiments and assembled the data. XDM, MX, AR, JTAD, YCH, and QWW performed data analysis and revised the manuscript.

DATA ACCE SS I B I LIT Y
Data files that support this study can be obtained from: https://doi.