Specificity of herbivore‐induced responses in an invasive species, Alternanthera philoxeroides (alligator weed)

Abstract Herbivory‐induced responses in plants can both negatively affect subsequently colonizing herbivores and mitigate the effect of herbivory on the host. However, it is still less known whether plants exhibit specific responses to specialist and generalist herbivores in non‐secondary metabolite traits and how specificity to specialists and generalists differs between invasive and native plant populations. We exposed an invasive plant, Alternanthera philoxeroides, to Agasicles hygrophila (Coleoptera, Chrysomelidae; specialist), Spodoptera litura (Lepidoptera, Noctuidae; generalist), manual clipping, or application of exogenous jasmonic acid and examined both the specificity of elicitation in traits of fitness (e.g., aboveground biomass), morphology (e.g., root:shoot ratio), and chemistry (e.g., C/N ratio and lignin), and specificity of effect on the subsequent performance of A. hygrophila and S. litura. Then, we assessed variation of the specificity between invasive and native populations (USA and Argentina, respectively). The results showed S. litura induced higher branching intensity and specific leaf area but lower C/N ratio than A. hygrophila, whereas A. hygrophila induced higher trichome density than S. litura. The negative effect of induction on subsequent larval growth was greater for S. litura than for A. hygrophila. Invasive populations had a weaker response to S. litura than to A. hygrophila in triterpenoid saponins and C/N ratio, while native populations responded similarly to these two herbivores. The specific effect on the two herbivores feeding on induced plants did not vary between invasive and native populations. Overall, we demonstrate specificity of elicitation to specialist and generalist herbivores in non‐secondary metabolite traits, and that the generalist is more susceptible to induction than the specialist. Furthermore, chemical responses specific to specialist and generalist herbivores only exist in the invasive populations, consistent with an evolutionary change in specificity in the invasive populations.

Previous studies have found consistency on specificity of induced response associated with feeding guilds of herbivores (e.g., chewers versus phloem-feeders, see review in Ali & Agrawal, 2012).
Furthermore, many studies have found that herbivores with distinct diet breadth (specialization) can still influence plant responses differentially (Agrawal, 2000;Van Zandt & Agrawal, 2004;Vogel et al., 2007). However, the systematic review suggests that there is still no consistent pattern of differential elicitation based on the degree of host plant specialization (Ali & Agrawal, 2012). One of reasons for this inconsistent pattern may be that most studies have concentrated more on response of plant secondary metabolites than on other plant traits (e.g., physical defense traits, Barton, 2016). Nevertheless, a meta-analysis of 72 studies suggests that although plant secondary chemicals are seen as fundamental to the defense against insects, some nonsecondary metabolite plant traits (life-history traits, physical defense traits, gross morphological traits, and primary chemistry) can also play an important role in plant defense against herbivores (Carmona, Lajeunesse, & Johnson, 2011). In particular, Carmona et al. (2011) have found that the negative effects of physical defense (e.g., trichome density and leaf toughness) and life-history traits (e.g., the rate of growth and phenology) are stronger than those of secondary metabolic chemistry on specialist herbivores. Thus, it is necessary to assess specificity of elicitation to herbivores with different diet breadths in a wide range of plant traits, especially in traits other than secondary metabolites.
Moreover, despite substantial evidence for specificity of induced response, it is still little known how such specificity could evolve by natural selection (Agrawal, 2011). Some recent studies have indicated that invasive populations of exotic plants may have different specificity of induced response to specialist and generalist herbivores relative to native populations (Huang et al., 2010;Wang et al., 2013). However, the patterns remain highly controversial (Huang et al., 2010;Wang et al., 2012Wang et al., , 2013. Wang et al. (2013) found that specificity of extrafloral nectar (EFN) induction to specialists vs. generalists only exists in native Triadica populations rather than invasive ones, whereas Huang et al. (2010) found the opposite in tolerance responses of Triadica. Therefore, limited traits and plant species investigated restrict broader generalization regarding how invasive and native populations vary in specificity to sets of herbivores with different diet breadth.
In this study, we assessed specificity of induced response in a wide range of plant traits (fitness, morphology, and chemistry) and the potential differences of populations in the alligator weed, A. philoxeroides, which is native to South America, but has colonized large regions of the United States, Australia, and China (Julien, Skarratt, & Maywald, 1995). Here, we compared invasive populations from the United States with native populations from Argentina to understand how plants from these regions respond to the attack by both A. hygrophila (Colepotera: Chrysomelidae), a specialist from South America and S. litura (Lepidoptera: Noctuidae), a generalist from Asia. To better understand the effects of herbivore induction, we followed the recommendations of Ali and Agrawal (2012) and included a manual clipping treatment and an exogenous jasmonic acid (JA) application treatment.
We therefore compared both types of induction treatments relative to undamaged plants and two abiotic stimuli: manual clipping and exogenous jasmonic acid (JA). Specifically, we asked:

1.
Is there any specificity of induced response in plant traits other than secondary metabolites? 2. What is the difference in specificity of induced response between invasive and native populations of A. philoxeroides?

| Study species
Alternanthera philoxeroides (alligator weed) is a perennial herb (Amaranthaceae), which depends on storage roots and rhizomes to overwinter (Jia, Pan, Li, Chen, & Yang, 2009). The plant reproduces primarily through vegetative propagation and emerges from storage roots in spring to form dense monospecific stands (Figure 1a,b;Jia, Pan, Sosa, Li, & Chen, 2010). The native range of this species extends from Argentina (39°S) to southern Brazil (18°S), where it is attacked by as many as 40 insect herbivores, including specialists like the alligator weed thrip, Amynothrips andersoni, alligator flea beetle, A. hygrophila, and a caterpillar, Arcola malloi (Maddox, Andres, Hennessey, Blackburn, & Spencer, 1971), but in its invasive range, which includes both the United States and China, few insects feed on alien A. philoxeroides (Xiaoyun P., personal observation).
Agasicles hygrophila (Coleoptera: Chrysomelidea, Figure 1c) is a strictly specialized leaf-chewing insect that completes its entire life cycle on alligator weed and has a distribution including all of the native alligator weed range. It has been utilized as biological control agent in USA since 1964 and in China since 1986. Despite these efforts, the insect is absent from many invasive populations of alligator weed, especially in terrestrial habitats (Appendix S1; Coulson, 1977;Ma, 2001;Pan, Zhang, Dong, & Li, 2010).
Spodoptera litura (Lepidoptera: Noctuidae, Figure 1d) is a cosmopolitan chewing insect that feeds on approximately 150 plant species from 40 families including plants in the Amaranthaceae (Rao, Wightman, & Rao, 1993).The native distribution of S. litura includes most of Asia, Australia and extends into the south Pacific as far west as Hawaii, but it is not found in North America (Nagoshi, Brambila, & Meagher, 2011;Pogue, 2002).
In this study, we obtained stem fragments from five invasive USA populations and five native Argentine populations in 2005 and 2006, respectively (Appendix S1), and planted them in the glasshouse. We collected 80 adults of A. hygrophila locally and maintained the colony on Chinese A. philoxeroides prior to the experiment. Those collected insects were fed in a growth chamber for 1 month (one generation) before experiment to reduce possible maternal effects (influence of heterogeneity of parental phenotype on their offspring phenotype). We purchased S. litura from the KeYun biocontrol company in Henan prov-

| Common garden experiment
We used a completely randomized design (CRD) in the experiment.
On 28 August 2015, 60 seedlings from each population with similar leaf numbers (two pair of leaves) were chosen and transferred to 400 ml plastic pots with fertile soil (Beilei Organic Fertilizer Co., Ltd., Zhenjiang, China) with the content of N, P, K ≈ 2 ± 0.3% (untransformed mean ± SD and the same below, dry weight basis), organic matter ≈35 ± 5% (dry weight basis), water ≈45 ± 3%, and pH = 6 ± 0.5 (total n = 600). Those 60 seedlings of each populations were averagely divided into four replicated plots (150 plants per plot). We randomly arranged the pots in one plot and caged each individual plant over the course of the experiment to prevent disturbance from other insects in the glasshouse.
On September 21, we randomly assigned one-fifth of the plants (n = 3) of each population in each plot to one of five treatments (unmanipulated control, A. hygrophila herbivory induction, S. litura herbivory induction, manual clipping, and exogenous JA) and each population/ treatment combination was replicated four times (four plots). For herbivory induction treatments, we applied two second-instar A. hygrophila or two third-instar S. litura larvae on each individual plant. A preliminary observation suggested that these instars would provide the equivalent amount of damage (50 ± 10% leaf area damage after 12 hr). Those two larvae were, respectively, confined on two fully expanded new leaves, which represented 25% leaf area (eight leaves totally in each plant, Appendices S2 and S3). We removed larvae when the two caged leaves were consumed completely (for 2d). For the manual clipping treatment, we clipped two leaves over 2d (one leaf per day) by cutting the whole leaf at the base of the stem (Appendix To avoid cross-interference by herbivore-induced volatiles (HI-VOCs), we grouped 120 plants of the same treatment at the beginning of induction and we: (1) kept about 1.5 m of distance between F I G U R E 1 (a) Alternanthera philoxeroides reproduces from storage roots, (b) a dense monospecific stand of A. philoxeroides, (c) a second-instar larva and two adults of Agasicles hygrophila feeding on A. philoxeroides, (d) three third-instar larva of Spodoptera litura feeding on Drimiopsis kirkii each treatment group with its neighbors (the effective radius of volatiles can be range from 60 cm to 1 m [Heil, 2014]), (2) assigned the control treatment to be upwind of other treatment groups, and (3) made the JA induction at an equivalent, adjacent glasshouse module (these two adjacent glasshouse modules were the same size

| Larval bioassay
Seven days after induction treatments, 12 plants of each population/ treatment combination were divided into three groups (10 populations × 5 treatments × 4 replicates = 200 plants per group). Two groups were challenged by either three second-instar A. hygrophila or third-instar S. litura larvae, and the remaining group was kept for phenotypic measurements. We retained about 1.5 m of distance between each bioassay group and its neighbors to prevent crosstalk.
The bioassay larvae were weighed to 0.0001 g initially and allowed to feed freely in nylon mesh bags for 4 days, after which the larvae were removed and reweighed by electronic balance (FA2104, Shanghai, China). We used the difference between post-and prelarval biomass as our measurement of larval growth gain in the challenge bioassay.

| Plant measurements
On 9 October 2015, the plants in the group not challenged by herbivores were harvested to measure plant fitness, morphological, and chemical traits. We measured leaf areas with a leaf area meter (Li-3100, Li-Cor Inc., Lincoln, NB, USA) and determined leaf trichome numbers by removing a 0.5-cm-diameter leaf disk from the first fully expanded leaves and counting the total number of trichomes on the top and bottom surfaces. All plant material was separated into below and aboveground (leaves, branches, and stems) parts, and dried at 60°C for 72 hr prior to determination of dry biomass. We used plant total, aboveground, and belowground biomass as our estimates of plant fitness (fitness traits). Our morphological traits included trichome density (sum of top and bottom trichome number on the disk), root:shoot ratio (RSR; ratio of belowground biomass and aboveground biomass), branch intensity (BI; ratio of branch dry weight and stem dry weight), specific stem length (SSL; ratio of length of stem in cm and stem dry weight in g), and specific leaf area (SLA; ratio of leaf area in cm 2 and leaf dry weight in g).
To obtain sufficient material for chemical analyses, we pooled dry leaves of four replicate individuals within each population/treatment combination and ground them to a fine powder. To measure total triterpenoid saponins, we used UV spectrophotometry (Wang, Xu, & Wang, 2011). Leaf powder was weighed, soaked in diethyl ether, heated in a water bath at 30°C for 4 hr, and then centrifuged. The sediment was extracted by methyl alcohol and chloroform, successively. The chloroform was desiccated, and the extract was dissolved in 200 μl methyl alcohol. The solution was measured at 215 nm, and the concentrations was calculated by regression, using oleanolic acid as a standard. To measure total flavonoids, we also used UV spectrophotometry (Li, Zhang, Xu, Wang, & Zhang, 2017;Zhang, Li, He, Chen, & Liu, 2012).
The leaf powder was weighed, soaked in 70% ethyl alcohol, and then centrifuged. The supernatant was collected, desiccated to remove the ethyl alcohol, and further extracted in distilled water and 1 ml of 99% ethyl acetate. The ethyl acetate was removed by desiccation, and the extract was dissolved in 1 ml of 70% ethyl alcohol with 1 ml KOH.
After 5 min, the solution was diluted to 5 ml using 70% ethyl alcohol.
The absorbance of the extract was measured at 395 nm, and concentrations were calculated by regression, using luteolin as a standard.
To determine the lignin concentrations, we used the Klason method (Effland, 1977), mixing equivalent leaves of each population of parent material together and using the Klason lignin and absorbance by UV spectrophotometry at 280 nm for each combination of population and treatment (Johnson, Moore, & Zank, 1961). To determine total carbon, total nitrogen, and the C/N ratio, we used an elemental autoanalyzer (FlashEA 1112 Series, Thermo Inc., Milan, Italy).

| Statistical analyses
We used mixed-

| Plant fitness traits
The type of treatment had a significant effect on plant aboveground biomass (Table 1). Clipping treatment plants had 25% greater aboveground biomass than controls and 43% greater aboveground biomass than feeding damage plants (Figure 2b), but similar with JA.
Treatments had no significant effects on plant total biomass or belowground biomass (Table 1, Figure 2a,c).

| Plant morphological traits
The treatments had significant effects on all morphological traits (Table 2)

| Plant chemical traits
The treatments had significant effects on total triterpenoid saponin concentrations and C/N ratio (  Figure 4d). with the S. litura had lower C/N ratios relative to the two abiotic treatments (12.3% and 9.3% on average for clipping and exogenous JA, respectively, Figure 4a). For total flavonoids and lignin, there were no significant differences among treatments (Table 3, Figure 4b,c).
In addition, although the two biotic stimuli did not differ in their effects on total triterpenoid saponins concentrations, the S. litura treatment decreased C/N ratios significantly relative to the A. hygrophila treatment (16.0% on average, Figure 4d). In contrast, the clipped plants had higher total triterpenoid saponins concentrations than those induced with JA (23.3% on average, Figure 4a), but did not differ in C/N ratio.

| Effect on insect performance
The specificity of effect can be detected from significant interactive effect between treatment and bioassay (Table 4). Compared to growth gain of S. litura in control, the S. litura experienced a significantly decreased growth gain in the two herbivory and clipping treatments (−69%, −56%, and −36% on average, respectively, relative to control, Figure 5), while not affected by JA. Additionally, the A. hygrophila was also negatively affected by the two herbivory damage and clipping induction (−31% and −32% and −7%, respectively, on average relative to the control,) but positive affected by JA (22% on average, Figure 5).

| Variation in specificity between native and invasive populations
For most plant traits, the treatments did not differ between the native and invasive populations (no significant interactive effect of treatment and origin in Tables 1-3). Nevertheless, there was a significant interaction between origin and induction treatment on total triterpenoid saponins concentration and C/N ratio (Tables 2 and 3).
T A B L E 1 The mixed-model ANOVA tests the effect of induction treatment, continental origin (native vs. invasive), their interaction on three square-root transformed fitness traits (total biomass, aboveground biomass, and belowground biomass). Population (Origin) of Alternanthera philoxeroides was treated as a random factor. Statistical significance is marked in bold and indicated as: *p < .05, **p < .01, ***p < .001 Compared to the control, the invasive populations had lower total triterpenoid saponins concentrations in the S. litura treatment (−19% on average) and two abiotic treatments (−20% for clipping and −30% for JA on average) but no significant induced changes in the A. hygrophila treatment (Figure 6a). Nevertheless, native populations significantly increased triterpenoid saponins concentrations to all the induction treatments compared to controls (70%, 88%, 91%, and 52% on average, respectively) ( Figure 6a). The invasive populations had lower total triterpenoid saponin contents in the S. litura treatment relative to the A. hygrophila treatment (−26%; Figure 6a), while native populations responded with similar triterpenoid saponin contents to the two biotic treatments (Figure 6a). Invasive populations damaged by A. hygrophila had higher total triterpenoid saponins than either of the two abiotic treatments (38% for clipping and 62% for JA on average, respectively; Figure 6a), while there were no significant differences between S. litura and the two abiotic treatments. In native populations, the two biotic treatments had similar total triterpenoid saponins with clipping treatment but significantly higher total triterpenoid saponins than the JA treatment (12% for A. hygrophila and 24% for S. litura on average, respectively; Figure 6a). for JA on average, respectively), but there was no significant difference between the A. hygrophila treatment and the two abiotic treatments.
In native populations, there were no significant differences between biotic and abiotic treatments.
Additionally, the population origin did not influence the magnitude of growth gain differences between the two herbivores in response to the treatments (no significant interactive effect among treatment, bioassay, and origin in Table 4).

| Specificity of elicitation in different plant traits and specificity of effect
In this study, we found that these plants can express specific elicitation to S. litura and A. hygrophila in some traits. In detail, the S. litura induced significantly decreased C/N ratio relative to the A. hygrophila and other treatments in our study. This C/N ratio response is consistent with a previous finding that generalist swift moth Endoclita excrescence decreases the foliage C/N ratio of willow species (Utsumi & Ohgushi, 2009) and indicates that plants can express specific elicitation to herbivory in C/N ratio. However, few studies have concentrated on the specificity of elicitation in C/N ratio to feeding by two herbivores that differ in diet breadth (Mooney, Tiedeken, Muth, & Niesenbaum, 2009;Wang et al., 2012), and the responses of C/N ratio in our experiment are in contrast to those studies. Mooney et al. (2009)  Triadica sebifera tends to similarly increase C/N ratio in response to specialist, Gadirtha inexacta, and two generalists, Grammodes geometrica and Cnidocampa flavescens. The reason underlying the differential response of C/N ratio in different research systems may be that the defenses of the herbaceous plants which evolved in fertile soils on disturbed sites like A. philoxeroides are generally nitrogen rather than carbon based and have higher tolerance to T A B L E 2 The mixed-model ANOVA tests the effect of induction treatment, continental origin (native vs. invasive), their interaction on five morphological traits (root:shoot ratio, RSR, Box-Cox transformed branch index, BI, reciprocal-specific leaf area, SLA, specific stem length, SSL, square-root total trichome density). Population (Origin) of Alternanthera philoxeroides was treated as a random factor. Statistical significance is marked in bold and indicated as: *p < .05, **p < .01, ***p < .001 F I G U R E 3 Effect of the induction treatments on morphological traits of Alternanthera philoxeroides, including (a) root:shoot ratio (RSR), (b) branching intensity (BI), (c) specific leaf area (SLA), (d) specific stem length (SSL), and (e) trichome density. AH, Agasicles hygrophila damage, SL, Spodoptera litura damage, Clip, clipped leaves, JA, exogenous jasmonic acid. Data are means ± 1 SE, and different letters indicate significant differences among means following LSD-adjusted post hoc contrasts herbivory relative to more slowly growing woody plants (Bryant, Chapin, & Klein, 1983;Niesenbaum, 1992;Zhang, Pan, Zhang, He, & Li, 2015). Therefore, the C/N ratio response may vary in different types of plant growth form. However, it is still unclear whether the C/N ratio response to different herbivory depends on the strategies of resource allocation between plant defense and growth, and it is necessary to test the specificity of elicitation in C/N ratio to generalist and specialist in diverse research systems.
Our finding also showed that A. philoxeroides exhibited significantly higher trichome density following damaged by the specialist A. hygrophila than when damaged by generalist S. litura. It is consistent with an emerged finding that plants can express specific elicitation in trichome to different herbivores especially with distinct diet breadth (Traw & Dawson, 2002).
Additionally, our results show that the induced plant responses to biotic stimuli were different from the responses to abiotic stimuli in fitness (e.g., aboveground biomass), chemical (e.g., concentrations of triterpenoid saponin and C/N ratio), and some morphological traits (e.g., SLA and trichome density). These results indicate that most plant traits have some specificity of elicitation to biotic and abiotic stimuli although with some overlap, consistent with previous observations (Mooney et al., 2009;Travers-Martin & Mueller, 2008).
The specific responses to biotic and abiotic stimuli may be based on two facts: (1) plants have recognition of herbivore-specific cues such as herbivore saliva (Heil, 2009;Hilker & Meiners, 2010) and (2) clipping is usually an one-off wounding event, while real herbivores normally cause a longer time of damage to plant tissue, which could increase the intensity of induction (Moreira, Zas, & Sampedro, 2012).
In this study, we also found that no matter what the induction treatment is, the generalist S. litura experienced more decrease of growth gain on induced plants relative to control than did specialist A. hygrophila. (Figure 4). Considering that S. litura is a typical generalist herbivore and A. hygrophila is a strictly specialist (see Section 2), that result may support the emerged view that generalists are more susceptible to induced plant defense than are specialist herbivores (Coley, Bateman, & Kursar, 2006;Cornell & Hawkins, 2003;Van Zandt & Agrawal, 2004). However, more recent studies indicate that the relative susceptibility between generalist and specialist herbivores is correlated with specific plant traits. For example, a recent meta-analysis found that physical and life-history plant traits are both negatively correlated with susceptability of specialist herbivores, while there was no consistent relationship between those traits and susceptibility of generalist herbivores (Carmona et al., 2011).
To rule out the chance that these differences in response are due to a difference in the amount of damage per se, we carefully standardized the amount of damage that the two herbivores applied to plants (cages were used to limit each herbivore on a specific leaf [Appendix S3] and just two leaves of each plant were eaten completely. These two leaves were 25% of total leaf area). Also, the observed differences were not caused by different feeding guild of herbivores, because the two kinds of larvae are both chewing feeders and have similar feeding behaviors, in that both herbivores cut the leaf from margin of leaf to midrib (Appendix S4).

| Variation of specificity between native and invasive plant populations
For the comparison of elicitation to the two herbivores, the invasive populations expressed lower total triterpenoid saponins and C/N ratio to generalist S. litura relative to specialist A. hygrophila, whereas native populations had no such specificity to the two herbivores ( Figure 6). These results support the view that invasive and native populations differ in their specificity of elicitation to herbivores with distinct diet breadth (Huang et al., 2010;Wang et al., 2013).
Furthermore, these results indicate that specificity of elicitation only existed in invasive A. philoxeroides populations rather than in native populations. This finding is consistent with previous research that invasive Triadica sebifera populations can express specific tolerance to generalist and specialist herbivores, while native populations did not have this specificity (Huang et al., 2010). However, in contrast, other important research on indirect defense (extrafloral nectar, EFN) indicates loss of specificity to generalist and specialist herbivores in invasive populations (Wang et al., 2013). Those conflicting results suggest that reduced herbivory may not lead to a consistent variation in specificity of induced response between invasive and native populations; rather variation in specificity between two populations may be correlated with the function of specific plant traits (Wang et al., 2013). Nevertheless, further research is still needed to assess the variation in specificity to herbivores with different diet breadth between invasive and native populations in diverse research systems.
For comparison of specific response to biotic and abiotic treatments, there is different specificity between invasive and native populations. This result is consistent with previous findings in Boechera stricta that the specificity of elicitation to biotic and abiotic stimuli exists in some genotypes but not in others (Manzaneda, Prasad, & Mitchell-Olds, 2010). Besides, in our study, the specificity of elicitation to biotic and abiotic stimuli in invasive populations also varied between types of biotic stimuli. In other words, specificity only existed between either of A. hygrophila or S. litura treatment and abiotic treatments in invasive populations. However, in native populations, the specific elicitation to biotic and abiotic treatments did not vary with different biotic stimuli ( Figure 6). Finally, it is also worth noting that if ignoring variation of invasive and native populations, the induced response of triterpenoid saponins has no specificity to the two herbivores ( Figure 4a). This result suggests that a lack of distinguishing plant population variation will cover up the specificity of plant induced response (Uesugi et al., 2013).
Overall, our study shows strong evidence for both specificity of elicitation in a wide range of traits, especially in traits other than secondary metabolites and specificity of effect. Furthermore, our data demonstrate that invasive populations can express specificity of elicitation to different herbivores, but native populations cannot.
We also suggest that invasive and native populations had different specificity of elicitation to biotic and abiotic stimuli, especially, only either S. litura or A. hygrophila damage treatment expressed this specificity in invasive populations rather than both in native populations. However, the difference in herbivore performances did not vary between invasive and native populations. Documenting variation in specific plant induced response among populations under different herbivore pressure, like that which exists between the native and invasive populations, is one way to improve our understanding of the evolution of specificity of induced response (Bingham & Agrawal, 2010;Carrillo, McDermott, & Siemann, 2014). Future work may consider variation of specificity of induced response between invasive and native populations to herbivores with close phylogenetic relationship but different diet breadth.

CONFLICT OF INTEREST
None declared.