Herbivore pressure on goldenrod (Solidago altissima L., Asteraceae): its effects on herbivore resistance and vegetative reproduction

Journal of Ecology 1, 1, 795 81 doi: 1.1111/j.1365-745.1.1958.x Herbivore pressure on goldenrod (Solidago altissima L., Asteraceae): its effects on herbivore resistance and vegetative reproduction Robert F. Bode and Andre Kessler* Department of Ecology & Evolutionary Biology, Cornell University, E441 Corson Hall, Ithaca, NY 14853, USA Summary 1. Ecological and physiological costs of resistance to herbivory are core concepts in the study of plant insect interactions, but identifying them remains challenging. These costs are most obvious when trade-offs in resource allocation occur between different growth and resistance traits.. We used plant genotypes collected from long-term herbivore exclusion plots and from plots with natural herbivory. We evaluated putative trade-offs between resistance to two different herbivore species (the larvae of the beetle Trirhabda virgata and the moth Spodoptera exigua), and between resistance and growth as a potential outcome of differential natural selection. 3. We hypothesized that long-term exclusion of herbivores would result in differential natural selection on plant resistance traits that are expressed by marked differences in mean resistance between plants from populations with and without herbivores. The results supported this hypothesis. Genotypes from herbivore exclusion plots were less resistant to the more common herbivore in the system, T. virgata, than genotypes from control plots. In contrast, the genotypes from the two herbivory regimes were equally resistant to the rarer S. exigua caterpillars. As a consequence, resistance to one herbivore species did not correlate with resistance to another, suggesting independent evolution of resistance to multiple herbivores. 4. Plant growth correlated positively with resistance to Trirhabda, but not to Spodoptera, and only in plants from herbivore exclusion plots, leading us to reject the hypothesis of a growth resistance trade-off. 5. Synthesis. Our results suggest that correlations between growth and resistance are context dependent and may only be apparent in populations relieved from certain natural pressures, such as in natural populations relieved from natural selection by herbivores. Key-words: cost of defences, herbivory, natural selection, plant herbivore interactions, plant growth, plant resistance, Solidago altissima Introduction Plant defence theory assumes the existence of metabolic costs for the production of plant chemical defences. Within this theory, plants are thought to have a finite pool of resources (resource allocation principle), available to be allocated to different functions including growth, reproduction and defence (Fagerstrom 1989; Karban & Baldwin 1997). It would follow that an increased allocation of resources into defence compound production should be as obvious as physiological trade-offs (such as reduced allocation to growth) as well as *Correspondence author. E-mail: ak357@cornell.edu reproductive outputs as an integral consequence. Although costs should be measured in terms of fitness to be considered in an evolutionary sense (Koricheva ; Karban 11), tradeoffs between different life functions (growth and reproduction) have been used as a valuable proxy for the existence of costs of defences. This is largely because estimating costs of plant resistance to herbivores directly is experimentally challenging for various reasons: defence metabolites may be highly effective for a low metabolic cost (Kakes 1989), may be degraded and reused (Mihaliak, Gershenzon & Croteau 1991) or may have roles in both primary and secondary metabolism (Seigler & Price 1976; Arnold & Targett 3). Trade-offs may also reflect feedback loops between the evolutionary and ecological Ó 1 The Authors. Journal of Ecology Ó 1 British Ecological Society

796 R. F. Bode & A. Kessler levels characteristic of the evolution of herbivore resistance. Thus, the identification of potential trade-offs as well as the conditions determining them is crucial for our understanding of the evolution of plant insect interactions. Plant resource allocation has been extensively studied both within and among species, but with mixed results. A number of studies have found that increased resistance to herbivory is costly in terms of growth and plant fitness (Mooney & Chu 1974; Bentley & Whittaker 1979; Herms & Mattson 199; Baldwin & Hamilton ; Zavala et al. 4; Zavala & Baldwin 6). In contrast, other studies have found no evidence of trade-offs between resistance and growth or reproduction, (Bowers & Stamp 199; Adler, Schmitt & Bowers 1995; Strauss & Agrawal 1999; Siemens et al. ; Arnold & Targett 3; Jones et al. 6; Lankau & Kliebenstein 9) or trade-offs between resistance to different herbivore species (Koricheva, Nykanen & Gianoli 4; Rudgers, Strauss & Wendel 4). Finally, the plant vigour hypothesis would suggest a trade-off between the number of branches on a plant and its resistance to herbivore attack (especially galling herbivores) (Price, 1991). Possible reasons for the inconsistent observations of trade-offs associated with resistance are the relative differences in the time plant populations have been exposed to a particular combination of agents of natural selection, and the environmental conditions under which trade-offs are measured. Assuming that there is a mix of plant genotypes in a newly established population, natural selection is expected to alter the frequency of genotypes and thereby the relative abundance of particular phenotypes. Thus, trade-offs can be best measured in systems that control for (or maximize) genetic and phenotypic variability (Strauss et al. ) or minimize the number of interacting species for less diffuse selection (Stinchcombe ; Hull-Sanders et al. 7). More specifically, long-term selection by dominant herbivore species could make trade-offs more difficult to detect because of reduced phenotypic diversity in the population, while trade-offs with resistance to a weak (low dominance) herbivore selection agent may be obscured by other agents of selection (Conner 1; Stinchcombe ). In long-term experiments, selection by forces such as competition may minimize the trade-off between growth and resistance. The long-term removal of insect herbivory as an agent of natural selection from a system could be a useful way to determine whether there are trade-offs between resistance and other traits. In the absence of herbivory (Stinchcombe ), natural selection should favour plants that invest fewer resources into resistance against one attacker and more into other functions, such as growth, reproduction, competition and or resistance to other attackers. A seminal long-term study by Richard B. Root et al. examined the effects of herbivore exclusion by insecticide application on the growth of a community-dominating plant species, Solidago altissima (L.), and on plant community composition (Cain, Carson & Root 1991; Root 1996; Carson & Root 1999, ). This work revealed how longterm (7 1 year) herbivore exclusion can have strong positive effects on components of plant sexual reproduction, including larger inflorescences, and a greater proportion of blooming stems (Root 1996), as well as on components of asexual reproduction, such as production and growth of rhizomes (Cain, Carson & Root 1991), compared to plants under normal herbivore pressure. However, the extent to which these phenotypic changes have a genetic basis remains unclear. In other words, do plants in populations not under selection by herbivores have on average lower resistance and more vigorous growth than plants in populations under selection by herbivores? Using a long-term herbivore exclusion experiment like the one described above to address this question has a number of additional benefits. Comparing growth rates in plants from herbivore exclusion plots to those from plots with natural herbivory in a common garden could more clearly reveal potential tradeoffs with resistance, since, for example, the absence of selection by herbivores and thus the relatively more important competition with other plants could place slow-growing, well-defended plants at a disadvantage. Here, we measure trade-offs between resistance and growth, and between resistance to two herbivores in S. altissima plants from populations under natural herbivory conditions and plants from populations where herbivores were excluded over a 1-year period. More specifically, we evaluate mean plant resistance to one of the major herbivores, the chrysomelid beetle Trirhabda virgata (J.L. LeConte), and the rarer larvae of the noctuid moth, Spodoptera exigua (Hu bner), in plants from natural herbivory and herbivore exclusion regimes. Growth and defence against multiple herbivores are of particular importance in the S. altissima system, where dominance of canopy space ensures maximization of fitness, and a large community of insect herbivores is present (Root & Cappuccino 199). Plant systems with such diverse arthropod communities and demonstrated reductions in plant fitness due to herbivory naturally evoke questions about the specificity of plant resistance. Is there diffuse selection by multiple herbivore species on a generalized resistance to herbivory? Or, is resistance specific to individual herbivore species or small groups of species? If resistance could evolve specifically to particular herbivore species, we would expect the strongest effects of general herbivore exclusion on the resistance to herbivore species that have the strongest influence on plant fitness. Comparing plant genotypes from different selection regimes (with and without herbivory) for their resistance to a common, plant fitness-affecting herbivore (T. virgata) to a relatively rare, less-damaging herbivore (S. exigua) allowed us to address the hypothesis of whether specific resistance to herbivores has evolved and to measure potential costs of that specificity. Specificity of resistance will be evident if plants from the two selective environments differ in their resistance to the two different herbivores. Here, we address four major hypotheses. (i) Herbivory is a strong selective agent on herbivore resistance, so that plants from natural herbivory plots should have, on average, higher resistance than plants from herbivore exclusion plots. (ii) There is a trade-off between growth and resistance. (iii) There is a trade-off between resistance to one herbivore species and resistance to another. (iv) The ability to detect a relationship between resistance and growth varies with the selective environment.

Effects of herbivore pressure on goldenrod 797 Materials and methods STUDY SYSTEM Solidago altissima is a common perennial forb found in abandoned agricultural fields in eastern North America. The clonal, rhizomatous growth of Solidago allows for perpetual maintenance of established lines and rapid propagation of genetically identical plants. In the field, above-ground biomass dies back every season and re-grows from below-ground rhizomes, so individual plants can theoretically persist in the population indefinitely unless they are replaced by later successional plant species or are exposed to lethal disturbance, such as intraspecific competition and herbivory (Carson & Root 1999, ). Moreover, once an old field community is established, there is very little recruitment of new plants from seeds (McBrien, Harmsen & Crowder 1983), meaning that most new growth comes from established plants. Two species of beetles (T. virgata and Microrhopala vittata (Fabricius), both Coleoptera: Chrysomelidae) among others feed on Solidago and are hypothesized to act as keystone herbivores, dramatically impacting plant fitness and plant community composition during outbreak years (McBrien, Harmsen & Crowder 1983; Carson & Root ). To compare the effects of long-term herbivore exclusion on plant resistance to different herbivore species, we used performance of the common T. virgata and a low abundance generalist moth larva (S. exigua, Lepidoptera: Noctuidae) as two independent measures of herbivore resistance. Trirhabda virgata is a specialist commonly found on Solidago (Root & Cappuccino 199), while S. exigua is a generalist less frequently found on Solidago (personal observation). Spodoptera exigua is capable of completing its life cycle on Solidago and can remove substantial amounts of leaf tissue, similar to amounts removed by Trirhabda (personal observation). We used these species to compare resistance of two herbivores of the same feeding guild (Maddox & Root 1987; Hull-Sanders et al. 7) but with different relative ecological impacts on Solidago. PLANTS AND RESISTANCE MEASUREMENTS We transplanted 14 propagated plants from each genotype into 15-cm-diameter azalea pots with Metro Mix Ò (Sun-Gro, Bellevue, WA, USA) soil. All plants had a single ramet at the beginning of the growing season. These were grown outside under ambient light and temperatures from late April to October, which is the natural growing season in upstate New York. There was no herbivory on these plants prior to our experiment. For each genotype, we used eight plants for growth measurements and six plants for herbivore resistance measurements. To measure herbivore resistance, plants were grown for 8 weeks. Thereafter, three plants of each genotype were infested with two second-instar S. exigua larvae (purchased from BioServ, Frenchtown, NJ, USA), grown on diet from BioServ for 6 days and acclimated on S. altissima leaf tissue for 4 h. Three plants from each genotype were infested with two T. virgata larvae (second or third instar, collected locally). We used two individual insects per plant to ensure the full availability of all plants as replicates in the study. Thus, our measurements (mean performance per plant and per plant genotype) integrate over the insects potential genotypic variation in tolerance of plant resistance. Trirhabda virgata larvae of similar size and instar were used for the experiment. We used second-instar S. exigua larvae since that instar is large enough to be weighed accurately, but not so large that the individuals show high variation in size. To determine initial larvae mass per plant, we averaged the initial mass of both larvae. All plants were enclosed in mesh bags with the larvae for 7 days. After the 7 days, all larvae were removed, and a final mass measurement was taken in the same way as the initial mass measurement. Growth rate of herbivores was calculated as (final mass ) initial mass) initial mass. Growth rates for herbivores were averaged on all plants of each genotype to provide genotype means. Resistance was calculated as 1 growth rate for both insects. Resistance between regimes was compared using Student s t-test of genotype means with jmp 8. (SAS Institute Inc., Cary, NC, USA). HERBIVORE EXCLUSION PLOTS AND PLANT MATERIAL Herbivore exclusion (herbivory )) plots were established in an old field containing S. altissima at Whipple Farm, in Ithaca, New York, USA (4 5 W, 76 31 W) (as in Cain, Carson & Root 1991; Carson &Root).Twelve5 5 m plots with a -m gap between each plot were marked and assigned to one of the two treatment regimes. Alternate plots (six in total) were sprayed with Fenvalerate following the manufacturer s instructions [ORTHO Ò Group, Marysville, OH, USA,.7 ll of the active compound (Esfenvalerate) in water per m ] every other week during the growing season (May to September) for 1 years (1996 8). Six plots were left unsprayed (herbivory +). Fenvalerate has been shown to have no significant direct effect on plant mass or flowering (Carson & Root ). During the 1-year treatment period, there were at least two local outbreaks of T. virgata and M. vittata that left more than 9% of all S. altissima plants in the herbivory (+) plots severely damaged. We randomly collected 16 individual plants (separate genotypes) from these plots, eight from herbivory ()) plots and eight from herbivory (+) plots. All plots were represented by at least one genotype, with multiple genotypes from some plots ( each from plots, 3 from 1 plot). All plants were derived from rhizome samples collected at least 4 m apart within a plot to minimize the probability of re-sampling the same genotypes. The plants were grown in a common garden greenhouse for three cycles (one cycle = rhizome grown until flowering, rhizomes cut and re-planted) to eliminate potential maternal effects before being used in our experiment. PLANT GROWTH MEASUREMENTS To measure growth, eight plants (not used in the bioassays) from each of the 16 genotypes were surveyed for above-ground shoots (ramets) in October, when all above-ground shoots for the season have been produced. Because all rhizomes do not necessarily produce plants in the next season (Cain, Carson & Root 1991), and plants are able to re-grow above-ground tissue after heavy leaf loss with no measurable fitness reduction (Meyer 1998), we used ramet number as our measure of asexual reproduction rather than rhizome mass, rhizome number, above-ground biomass or leaf number. However, it is important to note that we have found that ramet number correlates well with leaf number (linear fit, r =.336, P =.184, n = 16) and aboveground biomass (linear fit, r =.49,P =.76, n =16),signifying that high ramet production results in increased canopy space dominance and thus higher competitive ability. The number of ramets at the end of the season also correlates well with the number of rhizomes produced (linear fit, r =.463,P =.36, n = 16) and total length of rhizomes (linear fit, r =.45, P =.8, n =16). Moreover, in an old field community, asexual reproduction may be a more relevant measure of fitness, since new Solidago are unlikely to come from seeds (McBrien, Harmsen & Crowder 1983). Additional ramets, however, translate into a larger lifetime sum of inflorescences and thus more seeds, placing ramet number as the most reliable and reasonable fitness proxy for this species. We compared ramet numbers between the different genotypes and two treatment regimes using an anova (jmp 8.; SAS Institute). Relationships between ramet

798 R. F. Bode & A. Kessler number and resistance were analysed using linear regression with jmp 8. (SAS Institute). Results RESISTANCES TO TWO HERBIVORES ARE NOT CORRELATED Plantsfromtheherbivory(+)regimewereonaveragemore resistant to T. virgata larvae than plants from the herbivory ()) regime. The larvae had a higher mass gain on plants from herbivory ()) plots than on plants from herbivory (+) plots (Student s t-test, t = ).41, P =.33, n = 16, Fig 1a). Several larvae moulted into their pupal stage or died and were excluded from measurements, since larvae begin losing mass once they moult into their pupal stage. Survivorship did not differ between the two regimes (Student s t-test, t =.388, P =.74, n = 16). The growth rate of S. exigua larvae was equal on plants from herbivory ()) plots or herbivory (+) plots (Student s t-test, t = ).11, P =.9166, n = 16, Fig 1b). There was a wide variety of larval sizes, even between individuals feeding on the same genotype. Survivorship was lower in larvae feeding on plants from herbivory (+) plots, although this difference was not significant (Student s t-test, t = )1.8, P =.11, n = 16). The growth rate of T. virgata larvae did not correlate with the growth rate of S. exigua larvae, regardless of the herbivory regime their food plants derived from (linear fits; herbivory ()) r =.7, P =.6953, n =8,herbivory(+)r =.3, P =.678, n =8). Trirhabda Growth Rate Spodoptera Growth Rate.5 1.5 1.5 14 1 1 8 6 4 (a) (b) Herbivory ( ) Herbivory (+) Herbivory ( ) Herbivory (+) Fig. 1. Herbivore performance on plants from long-term herbivore exclusion (herbivory ())) and natural herbivory (herbivory (+)) plots. (a) Mean mass gain (±SEM) of Trirhabda virgata larvae. (b) Mean mass gain (±SEM) of Spodoptera exigua larvae. The asterisk (*) designates significantly different means as informed by Student s t-test of genotype means (P <.5). * HERBIVORE RESISTANCE DOES NOT SHOW A TRADE- OFF WITH PLANT GROWTH We found a positive correlation between mean resistance to T. virgata (1 growth of larva) and single-season asexual reproduction of S. altissima plants in herbivory ()) plots (linear fit, r =.51, P =.436, n = 8, Fig. a). The resistance to T. virgata did not correlate with ramets per plant in genotypes from herbivory (+) plots (linear fit, r =.53,P =.5815, n = 8). Mean ramet production was not correlated with mean resistance to S. exigua in genotypes from herbivory ()) plots (linear fit, r =.86, P =.488, n =8)orherbivory(+) plots(linearfitr =.7,P =.5195, n = 8). Genotypes varied significantly in growth rate (anova F 14,18 = 4.153, P <.1 Fig. b). However, the growth rates (genotype means) did not vary with treatment (Student s t-test, t = )1.78, P =.963). Discussion Plants from herbivory (+) plots had higher resistance to T. virgata than plants from herbivory ()) plots. This suggests that herbivory is a major agent of natural selection in the S. altissima system. It is reasonable to assume that chrysomelid Average ramets per plant Average ramets per plant 4 3.5 3.5 1.5 1.5..4.6.8 1 4 3.5 3.5 1.5 1.5 (a) (b) 1/Trirhabda growth rate 1-A4 1-C 1-C 3-D4 6-C3 8-A 8-B1 8-C3 11-B 11-C1 -B4 4-C3 5-A3 7-B 7-D4 9-D Herbivory ( ) Herbivory (+) Fig.. (a) Relationship between plant growth (average number of ramets per plant) and resistance to Trirhabda virgata for genotypes from the herbivore exclusion regime (m, black line) and the natural herbivory plots (h, grey line). Only the relationship in the herbivore exclusion regime is statistically significant (P <.5). (b) Variation in mean (±SEM) ramet production between genotypes propagated from herbivory ()) regime and herbivory (+) regime.

Effects of herbivore pressure on goldenrod 799 beetles like T. virgata and M. vittata areamongthemajor agents of selection on plant resistance in the herbivore community of S. altissima. Previous studies in the same system have shown that outbreaks of chrysomelid beetle species can affect long-term S. altissima fitness and thus its competitive ability. This damage is to such an extent that Solidago s dominant status in the community is weakened and overall plant species diversity is increased compared to populations without herbivory (Carson & Root ). If the removal of herbivores results in reduced mean resistance of the plants in the population, it is likely a result of the missing positive selection on resistance, mainly by the chrysomelid beetle species as well as a potential negative selection on resistance traits that are costly in the absence of herbivores. Such potential costs of resistance have long been discussed and are major concepts in our understanding of plant insect interactions (Koricheva ; Koricheva, Nykanen & Gianoli 4). The resource allocation principle assumes that plants have a finite pool of resources that can be allocated into growth, reproduction and resistance (Karban & Baldwin 1997). Here, we indirectly tested for two potential costs of resistance: (i) that the resistance to one herbivore compromises the resistance to another and (ii) that the investment into resistance compromises growth of new ramets. Our results clearly indicate that resistance to one herbivore species does not compromise the resistance to another herbivore in this system. Such ecological costs, where resistance to one herbivore may increase susceptibility to another, have been documented in a variety of systems (Bergelson & Purrington, 1996; Strauss et al. ; but see also Koricheva, Nykanen & Gianoli 4) and are thought to be important drivers of plant-mediated interactions among the members of the arthropod community (Viswanathan, Narwani & Thaler 5). However, we also did not observe the opposite effect, cross-resistance, which has been found in other systems (e.g. Kessler & Baldwin 4; Viswanathan, Narwani & Thaler 5). The performance of the generalist lepidopteran S. exigua was not different on plants from herbivory (+) and herbivory ()) plots, and we detected no correlation between the resistance to T. virgata and the resistance to S. exigua. This result is remarkable because both herbivore species belong to the same feeding guild and cause relatively similar amounts of damage. Although this result does not entirely exclude the possibility that the resistance to one herbivore may increase the susceptibility to another herbivore or pathogen, it suggests that different resistance traits and mechanisms may be under selection by different herbivore species (even in the same feeding guild) and may thus evolve independently. Support for this hypothesis comes from an experiment with a close relative of S. altissima, S. gigantea (Aiton). In invasive S. gigantea populations growing in Europe, plants have lower resistance to S. exigua (Hull-Sanders et al. 7) but not to T. virgata compared to plants in America. In this study, diterpenes were hypothesized to mediate resistance to S. exigua. We hypothesize that the reduced herbivore pressure in Solidago populations invasive to Europe resembles a natural experiment, similar to our herbivore exclusion experiment in the native habitat. The mechanisms of resistance to a coleopteran herbivore may be different from those that function for a lepidopteran herbivore (Hull-Sanders et al. 7; Huang et al. 1), for example, because the digestive physiologies of these two orders of herbivores are not the same (Jongsma & Bolter 1997). Insect physiology may explain why different herbivore species may not synergistically select for the same plant defences. We did not investigate the specific effects of compounds on each herbivore in this study, but from the data, we expect little overlap between resistance traits that show a high correlation with a particular herbivore species. Future studies would have to test this hypothesis in this system and expand it to comparative studies with multiple plant species to test underlying mechanistic and functional principles. In addition to the conclusion that different resistance traits potentially mediate resistance to different herbivore species, our data also suggest differential selection by S. exigua and T. virgata. The fact that we only see clear resistance differences between the long-term herbivory treatments when resistance is measured with T. virgata larvae (which have been hypothesized to negatively affect plant fitness), but not with S. exigua larvae, indicates that T. virgata, or a species complex responding to similar resistance traits as T. virgata, is likely to be the agent of selection on resistance traits (Carson & Root ). Future studies should explicitly test individual herbivore species as agents of selection on resistance traits to establish causal links between, for example, plant secondary metabolite production and natural selection by particular herbivore species. Our data did not support the existence of a trade-off between herbivore resistance and plant growth in the S. altissima system. The resource allocation principle predicts that decreased resistance would correlate with increased growth if plants reallocate resources from growth to herbivore resistance (Herms & Mattson 199; Karban & Baldwin 1997) and that, as a consequence, a trade-off of resistance with growth would be evident in plants from herbivore exclusion plots. Whereas in our study a correlation between resistance and growth appeared to be dependent on the selection regime (herbivory (+) vs. herbivory ())) (Fig. a) and was only apparent among plant genotypes from herbivore exclusion plots, growth and resistance were instead positively correlated among plants from the long-term herbivore exclusion plots. It is important to note that all experiments were carried out in the absence of competition, which, in combination with herbivory, has been hypothesized to have significant effects on plant performance (Carson & Root ). Thus, the full costs of resistance may only be apparent when measured in a competitive environment. Alternatively, there may not be trade-offs between resistance and above-ground ramet production, but with other plant functions mediating resistance, or physiological pleiotropic effects may link growth and parts of the secondary metabolite production positively with a net positive effect on resistance. For example, under conditions of high competition, but no herbivory (e.g. herbivory ()) plots), there may be selection for genotypes that can both grow and produce certain secondary metabolites with multiple physiological functions, including resistance, as would be expected if mechanisms of

8 R. F. Bode & A. Kessler chemical resistance are linked to protection from abiotic physiological stresses (Siemens et al. ). Thus, the observed positive correlation of growth and resistance in plants from the herbivory ()) plots could be a side effect of the production of multifunctional compounds that mediate generalized resistance to multiple herbivores. The higher resistance against T. virgata larvae in herbivory (+) plots, but no loss of resistance between plants from different plots to S. exigua supports this hypothesis. These data suggest that two mechanisms may be involved: first, the existence of agroup of general secondary metabolites with potentially multiple functions, which may not primarily be under selection by herbivores, but can mediate some generalized resistance. Second, another group (or groups) of compounds could be under selection by specific herbivores and thus specifically mediate resistance to particular herbivore species. Although such a mechanism could explain the positive correlation between vegetative reproduction and resistance in plants from herbivore ()) plots in the absence of competition, it still does not explain why specific resistance is reduced in those plots. Under conditions of low competition and high herbivory, one might predict a negative correlation between growth and resistance, as genotypes with extreme resistance phenotypes are favoured at the cost of growth (Kato, Ishida & Sato 8). Our data suggest a cost of specific resistance to T. virgata beetle larvae, although not in terms of above-ground ramet production and general growth. Following the above model (in which plants divide resources into a specific resistancemediating proportion and a general proportion of the total secondary metabolism of the plant), specific resistance, independently from the production of compounds with alternative functions, could trade-off with above- and belowground biomass production and with competitive ability of the plant that could be a function of below-ground growth and allelopathy production. Direct evidence for a potential tradeoff between the production of allelopathic compounds and the production of other secondary metabolites comes from a chemical analytical study on S. altissima. In this study, the production of the potent allelopathic polyacetylene dehydromatricaria ester was negatively correlated with the production of most major terpenoid classes (Johnson, Halitschke & Kessler 1). A more detailed study of the secondary metabolism of the plant genotypes used here will reveal whether traits that mediate specific resistance are traded off with traits that enhance competitive ability. In conclusion, our study found a positive correlation between growth and resistance, but only when herbivores were excluded from the system. This finding contradicts expectations of the growth differentiation balance hypothesis, which would predict that plants will either grow or defend (Herms & Mattson 199). Our findings also do not support the plant vigour hypothesis, as herbivores on faster-growing plants did not perform better than those on slow-growing plants (Price, 1991). However, this correlation may depend on the context in which the plants are growing (Stinchcombe ). It is possible that certain resistance traits have functions other than herbivore resistance (Seigler & Price 1976; Arnold & Targett 3), that the production of some general secondary metabolites is positively linked with plant growth (Siemens et al. 3; Jones et al. 6) and that the production of specific resistance-mediating secondary metabolites trades off with factors other than growth (Johnson, Halitschke & Kessler 1). One of the most important findings in this study is that relationships between different fitness-mediating traits, such as growth and resistance, may only be seen under certain environmental contexts, such as the removal of herbivore pressure (Stinchcombe ). Acknowledgements We thank Rayko Halitschke, Ellen Fagan, Kimberly Morrell, Justin Porter and Adam Damon for help with plant and insect growth measurements and sample preparation and Richard B. Root and Peter L. Marks for the early maintenance of the herbivore exclusion plots. A special thanks to Akane Uesugi, Stuart Campbell and Marta del Campo for constructive comments on experimental procedures. Thank you also to Jennifer Thaler, Anurag Agrawal, Su-Sheng Gan and Carrie Bode for commenting on an earlier version of the manuscript, and Kimberly Morrell for reviewing a late version of the manuscript. This project was funded by an NSF DDIG (DEB-1176), NSF (IOS-955) and FFF HATCH grants of Cornell University. References Adler, L.S., Schmitt, J. & Bowers, M.D. (1995) Genetic variation in defensive chemistry in Plantago lanceolata (Plantaginaceae) and its effect on the specialist herbivore Junonia coenia (Nymphalidae). Oecologia, 11, 75 85. Arnold, T.M. & Targett, N.M. (3) To grow and defend: lack of tradeoffs for brown algal phlorotannins. Oikos, 1, 46 48. Baldwin, I.T. & Hamilton, W. () Jasmonate-induced responses of Nicotiana sylvestris results in fitness costs due to impaired competitive ability for nitrogen. Journal of Chemical Ecology, 6, 915 95. Bentley, S. & Whittaker, J.B. (1979) Effects of grazing by a Chrysomelid beetle, Gastrophysa viridula, on competition between Rumex obtusifolius and Rumex crispus. Journal of Ecology, 67, 79 9. Bergelson, J. & Purrington, C.B. (1996) Surveying patterns in the cost of resistance in plants. American Naturalist, 148, 536 558. Bowers, M.D. & Stamp, N.E. (199) Chemical variation within and between individuals of Plantago lanceolata (Plantaginaceae). Journal of Chemical Ecology, 18, 985 995. Cain, M.L., Carson, W.P. & Root, R.B. (1991) Long-term suppression of insect herbivores increases the production and growth of Solidago altissima rhizomes. Oecologia, 88, 51 57. Carson, W.P. & Root, R.B. (1999) Top-down effects of insect herbivores during early succession: influence on biomass and plant dominance. Oecologia, 11, 6 7. Carson, W.P. & Root, R.B. () Herbivory and plant species coexistence: community regulation by an outbreaking phytophagous insect. Ecological Monographs, 7, 73 99. Conner, J.K. (1) How strong is natural selection? Trends in Ecology & Evolution, 16, 15 17. Fagerstrom, T. (1989) Anti-herbivory chemical defense in plants a note on the concept of cost. The American Naturalist, 133, 81 87. Herms, D.A. & Mattson, W.J. (199) The dilemma of plants to grow or defend. Quarterly Review of Biology, 67, 83 335. Huang, W., Siemann, E., Wheeler, G.S., Zou, J.W., Carrillo, J. & Ding, J.Q. (1) Resource allocation to defence and growth are driven by different responses to generalist and specialist herbivory in an invasive plant. Journal of Ecology, 98, 1157 1167. Hull-Sanders, H.M., Clare, R., Johnson, R.H. & Meyer, G.A. (7) Evaluation of the evolution of increased competitive ability (EICA) hypothesis: loss of defense against generalist but not specialist herbivores. Journal of Chemical Ecology, 33, 781 799. Johnson, R.H., Halitschke, R. & Kessler, A. (1) Simultaneous analysis of tissue- and genotype-specific variation in Solidago altissima (Asteraceae) rhizome terpenoids, and the polyacetylene dehydromatricaria ester. Chemoecology,, 55 64. Jones, T., Kulseth, S., Mechtenberg, K., Jorgenson, C., Zehfus, M., Brown, P. & Siemens, D.H. (6) Simultaneous evolution of competitiveness and

Effects of herbivore pressure on goldenrod 81 defense: induced switching in Arabis drummondii. Plant Ecology, 184, 45 57. Jongsma, M.A. & Bolter, C. (1997) The adaptation of insects to plant protease inhibitors. Journal of Insect Physiology, 43, 885 895. Kakes, P. (1989) An analysis of the costs and benefits of the cyanogenic system in Trifolium repens L. Theoretical and Applied Genetics, 77, 111 118. Karban, R. (11) The ecology and evolution of induced resistance against herbivores. Functional Ecology, 5, 339 347. Karban, R. & Baldwin, I.T. (1997) Induced Responses to Herbivory. The University of Chicago Press, Chicago, IL, USA. Kato, T., Ishida, K. & Sato, H. (8) The evolution of nettle resistance to heavy deer browsing. Ecological Research, 3, 339 345. Kessler, A. & Baldwin, I.T. (4) Herbivore-induced plant vaccination. Part I. The orchestration of plant defenses in nature and their fitness consequences in the wild tobacco Nicotiana attenuata. The Plant Journal, 38, 639 649. Koricheva, J. () Meta-analysis of sources of variation in fitness costs of plant antiherbivore defenses. Ecology, 83, 176 19. Koricheva, J., Nykanen, H. & Gianoli, E. (4) Meta-analysis of trade-offs among plant antiherbivore defenses: are plants jacks-of-all-trades, masters of all? The American Naturalist, 163, E64 E75. Lankau, R.A. & Kliebenstein, D.J. (9) Competition, herbivory and genetics interact to determine the accumulation and fitness consequences of a defence metabolite. Journal of Ecology, 97, 78 88. Maddox, G.D. & Root, R.B. (1987) Resistance to 16 diverse species of herbivorous insects within a population of goldenrod, Solidago altissima genetic variation and heritability. Oecologia, 7, 8 14. McBrien, H., Harmsen, R. & Crowder, A. (1983) A case of insect grazing affecting plant succession. Ecology, 64, 135 139. Meyer, G.A. (1998) Pattern of defoliation and its effect on photosynthesis and growth of Goldenrod. Functional Ecology, 1, 7 79. Mihaliak, C.A., Gershenzon, J. & Croteau, R. (1991) Lack of rapid monoterpene turnover in rooted plants implications for theories of plant-chemical defense. Oecologia, 87, 373 376. Mooney, H.A. & Chu, C. (1974) Seasonal carbon allocation in Heteromeles arbutifolia, a California evergreen shrub. Oecologia, 14, 95 36. Price, P.W. (1991) The plant vigor hypothesis and herbivore attack. Oikos, 6, 44 51. Root, R.B. (1996) Herbivore pressure on goldenrods (Solidago altissima): its variation and cumulative effects. Ecology, 77, 174 187. Root, R.B. & Cappuccino, N. (199) Patterns in population-change and the organization of the insect community associated with goldenrod. Ecological Monographs, 6, 393 4. Rudgers, J.A., Strauss, S.Y. & Wendel, J.E. (4) Trade-offs among anti-herbivore resistance traits: insights from Gossypieae (Malvaceae). American Journal of Botany, 91, 871 88. Seigler, D. & Price, P.W. (1976) Secondary compounds in plants primary functions. The American Naturalist, 11, 11 15. Siemens, D.H., Garner, S.H., Mitchell-Olds, T. & Callaway, R.M. () Cost of defense in the context of plant competition: Brassica rapa may grow and defend. Ecology, 83, 55 517. Siemens, D.H., Lischke, H., Maggiulli, N., Schurch, S. & Roy, B.A. (3) Cost of resistance and tolerance under competition: the defense-stress benefit hypothesis. Evolutionary Ecology, 17, 47 63. Stinchcombe, J.R. () Environmental dependency in the expression of costs of tolerance to deer herbivory. Evolution, 56, 163 167. Strauss, S.Y. & Agrawal, A.A. (1999) The ecology and evolution of plant tolerance to herbivory. Trends in Ecology & Evolution, 14, 179 185. Strauss, S.Y., Rudgers, J.A., Lau, J.A. & Irwin, R.E. () Direct and ecological costs of resistance to herbivory. Trends in Ecology & Evolution, 17, 78 85. Viswanathan, D.V., Narwani, A.J.T. & Thaler, J.S. (5) Specificity in induced plant responses shapes patterns of herbivore occurrence on Solanum dulcamara. Ecology, 86, 886 896. Zavala, J.A. & Baldwin, I.T. (6) Jasmonic acid signalling and herbivore resistance traits constrain regrowth after herbivore attack in Nicotiana attenuata. Plant Cell and Environment, 9, 1751 176. Zavala, J.A., Patankar, A.G., Gase, K. & Baldwin, I.T. (4) Constitutive and inducible trypsin proteinase inhibitor production incurs large fitness costs in Nicotiana attenuata. Proceedings of the National Academy of Sciences of the United States of America, 11, 167 161. Received 7 July 11; accepted 16 January 1 Handling Editor: Judith Bronstein