Herbivore exclusion drives the evolution of plant competitiveness via increased allelopathy


Author for correspondence:

André Kessler

Tel: +1 607 254 4219

Email: ak357@cornell.edu


  • The ‘Evolution of Increased Competitive Ability (EICA)’ hypothesis predicts the evolution of plant invasiveness in introduced ranges when plants escape from their natural enemies. So far, the EICA hypothesis has been tested by comparing plant vigor from native and invasive populations, but these studies are confounded by among-population differences in additional environmental factors and/or founder effects.
  • We tested the major prediction of EICA by comparing the competitive ability (CA) of Solidago altissima plants originating from artificial selection plots in which we manipulated directly the exposure to above-ground herbivores.
  • In a common garden experiment, we found an increase in inter-specific, but not intra-specific, CA in clones from herbivore exclusion plots relative to control plots. The evolutionary increase in inter-specific CA coincided with the increased production of polyacetylenes, whose major constituent was allelopathic against a heterospecific competitor, Poa pratensis, but not against conspecifics.
  • Our results provide direct evidence that release from herbivory alone can lead to an evolutionary increase in inter-specific CA, which is likely to be mediated by the increased production of allelopathic compounds in S. altissima.


The Evolution of Increased Competitive Ability (EICA) hypothesis has been proposed as a primary mechanism of plant invasiveness (Blossey & Notzold, 1995), when introduced plants escape from their specialist herbivores and pathogens (Keane & Crawley, 2002). Based on a frequently assumed tradeoff between herbivore defense and plant growth (Coley et al., 1985; Herms & Mattson, 1992; Koricheva, 2002), the EICA hypothesis predicts that release from herbivory in introduced ranges selects for plant genotypes with reduced resource allocation to herbivore defense and improved competitive ability (CA) via increased vigor (Blossey & Notzold, 1995).

The EICA hypothesis has been traditionally tested by comparing the biomass production of native and introduced populations of plants in common garden experiments (Blossey & Notzold, 1995). Although several studies have demonstrated the evolution of increased plant growth of invasives (Blair & Wolfe, 2004; Hahn et al., 2012), the support for the hypothesis remains mixed (Bossdorf et al., 2005). An important criticism of this approach is that geographical comparisons alone do not test directly the causal link between herbivore release and the evolution of plant traits, as they can be confounded by additional environmental differences between populations (Colautti et al., 2009) and/or by founder effects (Bossdorf et al., 2005; Dlugosch & Parker, 2008; Keller & Taylor, 2008). For example, plants may rapidly adapt to abiotic environmental variations along the latitudinal gradient (Maron et al., 2007; Etterson et al., 2008). Thus, if native and introduced populations are sampled from different latitudinal ranges, observed differences in CA may arise from adaptation to a variety of environmental factors other than release from herbivory (Colautti et al., 2009). Founder effects may also obscure the effects of selection by herbivory if introduced populations are founded by a limited number of genotypes that differ in mean CA from the native populations (Dlugosch & Parker, 2008). In addition, previous EICA studies have often compared growth differences only in the absence of competition, but the growth under unchallenged conditions does not necessarily reflect plant performance in a naturally competitive environment faced by invasive populations (Bossdorf et al., 2005). Moreover, because the underlying mechanisms of competition may differ between intra- and inter-specific competition (Lankau, 2008), we expect independent evolution of CA for conspecific and heterospecific competitors. However, only a few studies have tested simultaneously both intra- and inter-specific CA (Barney et al., 2009).

To address these issues and to test directly the EICA hypothesis, we asked whether the exclusion of herbivory alone would result in the evolution of intra- and inter-specific CA in Solidago altissima, a perennial plant native to eastern North America. The plant species is particularly relevant because it is an aggressive invasive species in Japan, where specialist herbivores have been absent for > 100 yr (Ando et al., 2010). Moreover, release from herbivory in introduced ranges is likely to have had a strong impact on the evolutionary trajectory of invasive S. altissima populations, because the plants in native ranges encounter a diversity of highly damaging herbivores (Root & Cappuccino, 1992). To circumvent confounding factors associated with previous EICA studies, we used S. altissima plants from an artificial selection experiment in which the above-ground herbivore assemblage has been manipulated by insecticide treatment for 12 yr. We sampled multiple individual plants from insecticide-treated plots (hereafter called ‘H− plots’; = 30 individuals sampled) and untreated control plots (‘H+ plots’; = 29 individuals), and propagated clones from these individuals for a subsequent common garden experiment. By comparing CA of H− clones and H+ clones, we were able to examine evolutionary shifts in competitive phenotype as a result of selection from above-ground herbivores. Because S. altissima primarily reproduces vegetatively (see the 'Materials and Methods' section), evolution in this system is likely to occur predominantly via differential propagation and the mortality of different clones. A previous study examining a subset of clones used here found reduced constitutive resistance to a major herbivore, Trirhabda virgata, in H− clones relative to H+ clones (Bode & Kessler, 2012), suggesting that enemy release resulted in the evolution of plant defense traits in S. altissima.

Here, we tested experimentally the second prediction of the EICA hypothesis that release from herbivory would result in increased CA (Blossey & Notzold, 1995). We assessed the CA in a strict sense (Barney et al., 2009) by growing S. altissima either by itself or with conspecific or heterospecific competitors in a common garden, and assessing the interactions between clone origin (i.e. H− or H+) and competition treatments. Poa pratensis was chosen as a heterospecific competitor, as it is a common competitor of S. altissima in old fields (Carson & Root, 2000). We predicted that H− clones would be less affected than H+ clones by competition. We also expected that the degree to which inter- and intra-specific CA evolved would differ as a result of differential mechanisms through which S. altissima competes with conspecifics and heterospecifics. Solidago altissima produces allelopathic polyacetylene compounds, which inhibit the germination and growth of several heterospecific plants (Kobayashi et al., 1980; Sawabe et al., 1999, 2000; Johnson et al., 2010), but have no previously reported effects on conspecifics. If allelopathic compounds are primarily used for inter-specific competition, intense inter-specific competition would select for increased polyacetylene production, which may provide no advantage in intra-specific competition. Thus, we also explored the differential effect of a major polyacetylene, cis-dehydromatricaria ester (DME), on the germination and seedling growth of S. altissima and P. pratensis by growing them under various concentrations of DME.

Materials and Methods

Study system

The tall goldenrod, Solidago altissima L. (Asteraceae), is a dominant perennial species of the old field community in its native range in eastern North America (Werner et al., 1980). On colonization of an abandoned field, the plants predominantly spread clonally, with recruitment by seed in established fields found to be rare (Cain, 1990). In central New York, S. altissima is attacked by a diverse group of herbivores, many of which are specialists on Solidago species (Root & Cappuccino, 1992). In particular, chrysomelid beetles, Trirhabda virgata and Microrhopala vittata, can have devastating fitness consequences during their outbreak years (Root & Cappuccino, 1992) by increasing stem mortality (Carson & Root, 2000) and reducing rhizome production (Cain et al., 1991). Herbivory by these insects increases understory light availability, and allows understory forbs and grass species to increase in biomass. Poa pratensis is one of the common understory plants found in old fields dominated by S. altissima, and is strongly affected by S. altissima density (Carson & Root, 2000). It is also likely to interact with invasive populations of S. altissima as it is widespread in Japan (Holm et al., 1979). Because of its likelihood of being a major competitor of S. altissima in New York and elsewhere, we chose P. pratensis as an inter-specific competitor for this experiment.

Herbivore exclusion experiment

To examine the evolutionary impact of herbivore removal, we used S. altissima clones obtained from artificial selection plots. The long-term selection plots were established in 1995 in a mid-successional old field at Whipple Farm, Tompkins Co., NY, USA, using the same pool of starting plant genotypes present in the field. Twelve plots (5 m × 5 m) were established, each separated by a 1-m strip of buffer zone. Six were randomly chosen for insecticide treatment (H− plots), which were sprayed with pyrethroid insecticide, fenvalerate (ORTHO® Group, Marysville, OH, USA). The remaining six controls were left unsprayed (H+ plots; plots were established as described in Carson & Root, 2000). The insecticide is effective against external feeders and chewing insects, but not against internal feeders, such as gallers and miners (Root, 1996). Fenvalerate has no known phytotoxic effect on goldenrod, and is not likely to have any fertilizing effects on the plants, but its effect on below-ground herbivores and microbes is unknown (Root, 1996).

In 2008, 16 haphazardly chosen plants from each of the 12 plots (192 plants in total) were used to initiate clone libraries at Cornell University (Ithaca, NY, USA). These plants were collected at least 1 m apart from each other in all directions to capture the maximum genotypic diversity of S. altissima within each plot. Although the plant is able to spread vegetatively, this method should result in the collection of distinct genotypes: Maddox et al. (1989) found that mid-successional fields hold, on average, 2.6 (± 1.3) genotypes of S. altissima within a 1-m-diameter circle. It is possible that we have, in rare cases, collected multiple plants of the same genotype, but this should not constrain our analyses, because we are concerned with the mean shift in competitive phenotypes within a plot as a result of insecticide treatment. To remove potential maternal effects, each clone was grown from rhizomes in the glasshouse for at least two growing cycles.

Common garden experiment

To assess inter- and intra-specific CA, we randomly selected 30 H− and 29 H+ clones (three to eight clones per plot, hereafter called ‘target clones’) from the above collection. Six plants from each target clone were propagated from rhizome cuttings on 1 April 2010. On 13 May 2010, newly sprouted target plants were transplanted into 20-cm azalea pots containing a 1 : 3 ratio of sand (Bonsal American, Charlotte, NC, USA) and potting soil (Sun-Gro, Bellevue, WA, USA), and placed under three levels of competition treatment: control when growing by itself, inter-specific competition when growing with P. pratensis, and intra-specific competition when growing with another S. altissima plant. We used an additive design (instead of a substitutive design) to estimate the importance of competition on target plant growth relative to other factors that could affect its growth (Goldberg & Fleetwood, 1987). In the inter-specific competition treatment, c. 200 seeds of P. pratensis (obtained from Seeds Trust, Inc., Cornville, AZ, USA) were sprinkled evenly on the soil surface around the target plant. As an intra-specific competitor, we selected a standard S. altissima clone expressing intermediate growth rates and polyacetylene production levels (A. Uesugi, unpublished). All 59 target clones were placed into competition against the standard clone to measure the relative strength of intra-specific CA. All plants were placed on the roof of Corson Hall, Cornell University, during the growing season.

To calculate the relative growth rate, we counted the number of newly produced leaves between two survey periods, 16 June and 12–16 July 2010, the period of rapid linear increase in leaf number (Bode & Kessler, 2012). The number of ramets was counted on 12 July 2010 as a measure of lateral vegetative growth. Both measures were chosen as indicators of the plant's ability to efficiently shade out other plants and increase its competitive dominance in the field during the early growing season (McBrien et al., 1983; Bode & Kessler, 2012). Root samples were collected for polyacetylene analyses on 12 July 2010 by removing a few strands of lateral roots from living target plants. Root tissues were frozen with liquid nitrogen, and stored at −80°C for later chemical analysis. All plants were harvested between 3 and 10 September 2010 when the target plants started to bloom. The roots and rhizomes of target plants were carefully separated in soapy water from those of competing plants, and both above- and below-ground tissues were dried at 45°C for 48 h. The dry biomasses of inflorescence and rhizomes were measured as an indication of sexual and asexual reproduction, respectively. Plant samples with accidental infestation by stem gallers (either by Eurosta solidaginis or Rhopalomyia solidaginis) by the end of the season were removed from the biomass analyses.

Root polyacetylene analysis

Approximately 200 mg of fresh root material from each sample were flash frozen with liquid nitrogen and crushed to a powder in a 1.5-ml Eppendorf tube. Samples were extracted in 1 ml of 90% methanol using FastPrep® tissue homogenizer (MP Biomedicals®, Solon, OH, USA) at 6 m s−1 for 60 s with 0.9 g of grinding beads (zirconia/silica, 2.3 mm; Biospec®, BioSpec Products Inc., Bartlesville, OK, USA), sonicated for 6 min and left in the dark at room temperature for 24 h. Fifteen microliters of the supernatant were analyzed for secondary metabolites by high-performance liquid chromatography (HPLC) on an Agilent® 1100 series HPLC equipped with a Gemini C18 reverse-phase column (3 μm, 150 mm × 4.6 mm; Phenomenex, Torrance, CA, USA) using a standard method targeted at phenolic compounds (Keinanen et al., 2001) with a slight modification to quantify polyacetylenes. Our elution system, consisting of solvents 0.25% H3PO4 in water (pH 2.2) (A) and acetonitrile (B), was as follows: 0–5 min, 20% B; 5–35 min, 20–95% B; 35–45 min, 95% B; with a flow rate of 0.7 ml min−1. Four polyacetylene compounds have been isolated previously (Sawabe et al., 2000): methyl ()-2-decane-4,6,8-triynoate (DME), (4)-2,4-decadiene-6,8-diyn-4-olide, (4)-2,4-decadiene-6,8-diyn-4-olide, and methyl 10-[()-2-methyl-2-butenoyloxy]-(2Z,8)-2,8-decadiene-4,6-diyoate. We identified peaks of these compounds by mass spectral comparisons with published spectra (Sawabe et al., 2000), with quantification at 254 nm. The concentration of each compound in each sample was calculated using a DME calibration curve obtained with purified DME (see ‘DME application experiment’). The concentration was expressed as parts per million of DME equivalence.

DME application experiment

We tested allelopathic effects of DME on seed germination and seedling growth, as it is the dominant polyacetylene compound found in S. altissima (see Supporting Information Table S1). We obtained purified DME using the method of Kobayashi et al. (1980). Briefly, the methanol extract of dried root materials was concentrated to 5% of its original volume in vacuo, and the condensate was extracted with hexane. The hexane solution was then chilled to −20°C, which resulted in the formation and precipitation of needle crystals characteristic of DME. The effect of DME on germination and seedling growth of P. pratensis and S. altissima was tested using media containing five concentrations of DME (0, 6, 12, 24 and 48 ppm) to include natural field variation (c. 6 ppm; Kobayashi et al., 1980) using the modification of Johnson et al. (2010). In each of five 500-ml media bottles, 300 ml of distilled water was combined with 0.96 g of Gamborg's B5 basal medium (Sigma-Aldrich) and 1.8 g of Phytagel (Sigma-Aldrich), the pH was adjusted to 6.8 and the bottles were autoclaved for 40 min. Ten milliliters of DME solution in methanol with the above concentrations were added to each bottle at 55°C, and the medium was divided into ten 10-cm-diameter Petri dishes to cool.

Seeds of P. pratensis, obtained from the same source as used in the common garden experiment, were washed with sterile water before sowing. Solidago altissima seeds were obtained from five plants collected from various patches in Tompkins Co. (NY, USA) to ensure some level of the natural variation. To minimize seed loss to molds, the seeds were washed in ethanol for 2 min, undiluted bleach solution for 10 min and rinsed in sterile water. For each plant species, 20 seeds were sown on each Petri dish on 2 August 2012, and placed under artificial light in a laboratory at 23°C. Germination was monitored on days 7, 11, 14 and 24, and surviving seedlings were dried at 40°C for 48 h and weighed to determine the total biomass per dish on day 31.

Statistical analysis

The effects of clone origin and competition on the rate of leaf production, the final mass of inflorescence and rhizome, and root polyacetylene concentrations were analyzed using a Generalized Linear Mixed Model (function ‘lmer’ in the lme4 package of R version 2.8.1; R Foundation for Statistical Computing, Vienna, Austria) fitted by a restricted maximum-likelihood approach after being log transformed to improve normality. Ramet number was analyzed assuming a Poisson error distribution because it is a count measure. In both analyses, the clone origin and plant competitors were modeled as fixed effects, and clones nested within a plot were entered as a random effect. By including plot as a random effect, we were able to consider each plot as a replicate of the herbivore removal experiment. For each measurement, model simplification to obtain appropriate P values for the fixed effects was performed following the method of Crawley (2007). The full model, including clone origin, competitors and origin × competitor provided the most appropriate model in all plant growth measurements, but exclusion of the interaction term was favored for polyacetylene concentrations. F statistics for the overall fixed effects were calculated using the ‘anovaTAB’ function in the MixMod package of R. Because we are interested in assessing inter- and intra-specific CA separately, when the overall interaction is significant, we also report t values (or z values for ramet number) for interactions between clone origin and each of the competition treatments.

The effect of DME application on seed germination was analyzed using repeated-measures ANOVA with DME concentration as fixed effect and monitoring days as random effect. Total seedling biomass and seedling survival rate were calculated per Petri dish, and tested for the DME effect using ANOVA, followed by post hoc tests with Tukey's honestly significant difference (HSD) test contrasting each DME concentration with the 0-ppm control. All statistical analyses were conducted using R.


Common garden experiment

Competition with both P. pratensis and S. altissima competitors reduced significantly the leaf production rate of target plants (F2,263.1 = 94.5, < 0.0001; Fig. 1a). We found no overall effect of clone origin on the leafing rate (F1,8.6 = 0.27, = 0.62) and, in the absence of competitors, the rate did not differ between H− and H+ clones (Fig. 1a). However, the interaction between clone origin and competition was significant (F2,263.1 = 3.74, P = 0.03). In particular, pairwise tests showed a significant interaction with P. pratensis competition (t266 = −2.67, = 0.008), but not with S. altissima competition (t266 = −1.75, = 0.081), suggesting a more pronounced evolutionary shift in inter-specific relative to intra-specific CA. In the presence of P. prantensis, H− clones produced new leaves 28% more rapidly than did H+ clones (Fig. 1a), indicating that H− clones are better able to maintain faster growth under inter-specific competition relative to H+ clones. Similarly, the number of ramets produced decreased with competition (F2,270.9 = 14.6, < 0.0001), but showed no overall effect of clone origin (F1,3.2 = 0.22, = 0.67; Fig. 1b). However, there was a significant interaction between competition and clone origin (F2,270.9 = 3.3, = 0.04). Specifically, H+ clones reduced ramet production in competition with P. pratensis, whereas H− clones maintained an equal level of ramet production in the presence of competition (Fig. 1b), resulting in significant interaction between clone origin and competition with P. pratensis (= 2.10, = 0.036). Such an interaction was not observed for competition with S. altissima (= 0.98, = 0.33), again indicating a stronger evolutionary shift in inter-specific CA.

Figure 1.

The rate of (a) leaf production (new leaves d−1) and (b) ramet production of target plants (± SE) when growing alone (control), with Poa pratensis (Poa) or with Solidago altissima (Solidago). Closed bars, H+ clones; open bars, H− clones.

Competition with P. pratensis and S. altissima also reduced inflorescence mass (F2,244.7 = 230.5, < 0.0001) and rhizome mass (F2,248.8 = 113.7, < 0.0001), but clone origin did not have overall effects (F1,10.9 = 0.04, = 0.8 and F1,10.7 = 0.00, = 1.0 for inflorescence and rhizome mass, respectively). There were no interactions between clone origin and competition for inflorescence and rhizome masses (F2,244.7 = 1.46, = 0.2 and F2,248.8 = 0.61, = 0.5, respectively).

Root polyacetylene production

The production of DME, the major polyacetylene, was significantly higher in H− clones than in H+ clones (F1,8.7 = 9.42, = 0.01; Fig. 2). The presence of competitors did not influence DME production (F2,268.8 = 2.02, = 0.13). Qualitatively similar effects of clone origin were found for the other three polyacetylenes (Table S1), suggesting evolutionary shifts in polyacetylene production.

Figure 2.

Concentration of cis-dehydromatricaria ester (DME, ppm ± SEM) in root tissues in H+ (closed bars) and H− (open bars) clones growing under three competitive treatments, including control, competition with Poa pratensis (Poa) and with Solidago altissima (Solidago).

DME application experiment

DME had an inhibitory effect on P. pratensis seed germination (F4,20 = 3.7, = 0.021) and seedling mass (F4,20 = 9.7, = 0.0002). DME at concentrations of 24 and 48 ppm suppressed significantly seed germination and dry biomass accumulation of plants that germinated (Table 1, Fig. 3b). By contrast, DME did not affect the germination of S. altissima seeds at any DME concentration (F4,20 = 1.73, P = 0.18; Fig. 3a). Instead, DME enhanced seedling survival (F4,20 = 5.2, = 0.0049, mortality was usually associated with the occurrence of seed-borne pathogenic fungi) and increased the dry mass of surviving individuals (F4,20 = 4.78, = 0.0071), particularly at 48 ppm (Table 1, Fig. 3b).

Table 1. Results of the cis-dehydromatricaria ester (DME) application experiment. The effect of DME concentrations of 6, 12, 24 and 48 ppm on seedling germination, growth (seedling dry mass) and survival is shown as mean change in % performance (Δ) relative to performance on media containing 0 ppm DME
Seedling germination Poa pratensis Solidago altissima
ΔdfT-value P ΔdfT-value P
6 ppm−0.7620−0.80.43−0.7520−1.380.18
12 ppm−1.3620−1.430.17−0.9420−1.720.1
24 ppm−2.2520−2.36 0.029 0.13200.240.82
48 ppm−3.3420−3.5 0.002 −0.8620−1.580.13
Seedling massΔLower CIUpper CI P adj ΔLower CIUpper CI P adj
  1. Effect of DME concentration on seed germination was tested with repeated-measures ANOVA, and that of seedling mass and survival was tested with ANOVA with post hoc tests using Tukey's honestly significant difference (HSD) test. Survival rate was not analyzed for P. pratensis because no mortality was observed in germinated seedlings during the study period. P values in bold indicate significance at α = 0.05. CI, confidence interval; df, degree of freedom.

6 ppm0.04−5.595.671.00.1−
12 ppm−0.36−5.995.271.01.26−2.905.420.9
24 ppm−5.7−11.33−0.07 0.046 3.98−
48 ppm−9.02−14.65−3.39 0.001 4.540.388.70 0.028
Seedling survival
6 ppm    −0.08−0.330.170.86
12 ppm    0.03−
24 ppm    0.11−0.140.360.66
48 ppm 0.028
Figure 3.

Effect of cis-dehydromatricaria ester (DME) concentrations on (a) seed germination and (b) seedling growth (mean dry mass (mg) ± SEM) of Poa pratensis (closed circles) and Solidago altissima (open circles). Asterisks indicate significant difference from control treatment (0 ppm) at: *, < 0.05; **, < 0.01.


We used an experimental approach to test one of the main predictions of the EICA hypothesis, namely that release from herbivore pressure will lead to an evolution of increased CA (Blossey & Notzold, 1995). Solidago altissima plants used in our common garden experiment originated from long-term artificial selection plots, in which we manipulated the above-ground herbivore community, whilst controlling for other environmental factors, and using the same pool of starting plant genotypes as present in the old field. Thus, our approach addresses the limitations of traditional EICA studies, whose results are often confounded by additional environmental differences between sample populations and/or founder effects. Moreover, insecticide treatments are known to dramatically reduce herbivore damage on S. altissima and increase its stem density at an ecological time-scale (Carson & Root, 2000), mimicking the enemy-free environment in the introduced ranges. So far, such manipulative experiments to test the evolutionary consequences of herbivore exclusion on plant competitiveness are rare (Agrawal et al., 2012).

Our results partially support the EICA hypothesis: although clone origin did not affect leaf and ramet production in the absence of competitors, a significant interaction between clone origin and competition with P. pratensis suggested that H− clones were able to maintain a higher growth rate than H+ clones under inter-specific competition. The differences between H− and H+ clones indicate the evolutionary shift in inter-specific CA. By contrast, no such interaction was found when competing with conspecifics, indicating that intra-specific CA did not evolve. Although we observed an evolutionary shift in inter-specific CA, as measured by leaf and ramet production during the early growing season, the pattern did not translate into measures of end-of-season asexual (rhizome mass) and sexual (inflorescence mass) reproduction. This is surprising because previous experiments have shown that the inflorescence mass (Carson & Root, 2000) and rhizome production (Cain, 1990) are greater in plants from H− plots relative to those from H+ plots when they are measured in situ. The results may indicate that the removal of herbivores does not select for increased production of reproductive tissues, but rather selects for faster early growth, which could have significant fitness effects in a highly competitive environment. Because light availability is an important factor determining the success of understory plants (Carson & Root, 2000), rapid production of leaves may allow S. altissima plants to efficiently shade out other plants and obtain competitive advantages. In addition, Cain (1990) found that the survival of S. altissima stems in the field increases with plant size in the early growing season. Rapid ramet production also allows plants to grow laterally, thereby occupying more space for growth. Therefore, an increased rate of leaf and ramet production should allow clones to persist and to eventually dominate in highly competitive environments.

That we found the growth advantage of H− clones only under a competitive environment suggests that growth measures in the absence of competition do not necessarily reflect growth in competitively challenged environments, and emphasizes the need for the specific testing of plant CA. Previous studies have shown varying results: For instance, Barney et al. (2009) showed that invasive mugwort (Artemisia vulgaris) populations in North America produced more ramets and total biomass than native European populations, and the difference between continents was exacerbated in inter-specific competition with Solidago canadensis. Similarly, Bossdorf et al. (2004) found growth differences between native and invasive populations of Alliaria petiolata only in a competitive environment, but invasive populations evolved reduced CA relative to native populations, the opposite of the EICA prediction. Leger & Rice (2003), by contrast, found growth differences in invasive and native populations of California poppies (Eschscholzia californica) only in the absence of competition, indicating that CA, in a strict sense, did not evolve in invasive populations. Similarly, Blumenthal & Hufbauer (2007) found differential growth only in the absence of competition across 14 invasive plant species, and suggested that these species may be adapted to disturbed and noncompetitive environments in their introduced ranges. The different outcomes among these studies therefore seem to depend on the competitive environment in which newly established populations evolved. In our study, CA rather than growth in the absence of competitor was selected for, potentially because the initial herbivore exclusion experiment was conducted in an already established old field, where surviving S. altissima plants faced strong competition.

Interestingly, we observed an evolutionary shift in inter-specific CA, but not in intra-specific CA, suggesting that selection for inter-specific competition was stronger than that for intra-specific competition. In support of this hypothesis, Carson & Root (2000) found that herbivory suppresses the growth of S. altissima, which allows understory heterospecific competitors to increase in abundance. When herbivores were removed by insecticide, S. altissima initially competes with these understory plants more intensively than with conspecifics. Thus, the clones from H− plots may represent the winners of such inter-specific competition. Whether inter-specific CA is more likely than intra-specific CA to evolve in invasive populations is an interesting question, but few EICA studies so far have tested both inter- and intra-specific CA simultaneously (Barney et al., 2009).

Differences in the degree to which inter- and intra-specific CA evolve in S. altissima may be explained by mechanisms used by the plant to compete with conspecifics and heterospecifics. We found that H− clones, on average, produced twice as much DME as did H+ clones, a compound previously shown to be allelopathic to Oryza sativa (Kobayashi et al., 1980, 2004; Ito et al., 1998), Miscanthus sinensis, Ambrosia artemisiaefolia (Kobayashi et al., 1980), Lactuca sativa (Sawabe et al., 1999) and Asclepias syriaca (Johnson et al., 2010). Similar patterns were found for the other three polyacetylene compounds, which are also known to exhibit allelopathic properties (Sawabe et al., 2000). We cannot exclude experimentally the possibility that the root chemistry evolution resulted from a direct physiological effect of insecticide, as the effect of fenvalerate on root secondary metabolite production is unknown. However, our preliminary results from a geographic comparison showed congruent evolution of increased polyacetylene production in invasive Japanese genotypes relative to native North American genotypes (A. Uesugi, unpublished), suggesting that the insecticide itself is not the major driver of root chemistry evolution. Thus, herbivore exclusion is the most likely factor causing the evolution of increased polyacetylene production via a change in competitive environment in our experimental populations.

These evolutionary shifts in root chemistry coincided with an evolutionary increase in inter-specific CA, suggesting that allelopathy may play a direct role in inter-specific competition. Consistent with this hypothesis, we found that DME suppresses germination and seedling growth of P. pratensis at 24 and 48 ppm, similar to the active concentrations found in previous studies (Kobayashi et al., 1980, 2004; Ito et al., 1998; Sawabe et al., 1999; Johnson et al., 2010). Although these concentrations are slightly higher than the reported natural levels of DME in soil (c. 6 ppm; Kobayashi et al., 1980), a high concentration of DME within roots (up to 200 ppm) suggests that DME could effectively suppress the growth of competitors that come into contact. Moreover, these high concentrations of DME did not suppress the germination of S. altissima seeds, but enhanced seedling growth by inhibiting seed-borne pathogenic fungal growth, suggesting contrasting direct effects of DME on conspecific and heterospecific competitors. DME may also influence indirectly the growth of heterospecific competitors by changing the microbial community with which each is associated (Mummey & Rillig, 2006; Vogelsang & Bever, 2009; Zhang et al., 2010). In a closely related species, Solidago canadensis, the root exudate, including DME, inhibited colonization of mutualistic mycorrhizal fungi on competing species in invasive ranges (Lu et al., 1993; Zhang et al., 2010). Such allelopathic effects on microbial communities are not likely to influence the growth of conspecifics, which could explain the observed differences between inter- and intra-specific CA of H− clones in our experiment. Although our results are consistent with the hypothesis that selection for increased polyacetylene production mediates the evolution of increased CA against heterospecific competitors, a further study should examine the allelopathic effects of polyacetylenes in bioactive soils under natural conditions.


Using clones originating from an artificial herbivore exclusion experiment, this study provides a direct link between enemy release and the evolution of plant competitiveness implied in the EICA hypothesis (Blossey & Notzold, 1995). A previous study examining a subset of H− and H+ clones used here found that H− clones have evolved a reduced level of constitutive resistance to a major herbivore (Bode & Kessler, 2012). Our study further demonstrates that release from herbivory alone can result in the evolution of CA through increased growth under inter-specific competition mediated by the production of allelopathic compounds. The observed evolutionary shifts occurred within 12 yr of herbivore exclusion, suggesting a rapid evolution of plant competitive traits. A similar, rapid evolution of plant competitive and defense traits in response to experimental herbivore removal has been shown in Oenothera biennis (Agrawal et al., 2012), suggesting a vital role of insect herbivores in driving plant adaptation.

Would we expect a parallel evolutionary shift in comparisons between native and invasive populations of S. altissima? Our results suggest that release from herbivory in introduced ranges could lead to a rapid evolution of CA, which could potentially override the effects of additional environmental differences among populations and/or reduced genetic diversity caused by founder effects (Xu et al., 2010). Moreover, an evolutionary change in allelopathy may provide even greater competitive advantages for introduced populations if polyacetylenes function as novel weapons to which many competitors are not adapted (the Novel Weapon hypothesis; Callaway & Aschehoug, 2000; Qin et al., 2013). Although we observed only the evolution of inter-specific CA in this experiment, we expect intra-specific CA to also evolve if intra-specific competition eventually intensifies as S. altissima becomes dominant (e.g. density in invasive Japanese populations can be more than twice that of native North American populations; Carson & Root, 2000; Nishihiro et al., 2007). These temporal dynamics of selection in invasive populations were observed in Alliaria petiolata, where plants from newly colonized populations were more allelopathic to heterospecific competitors than were older populations (Lankau et al., 2009; Lankau, 2011). A future study will compare native and invasive populations of S. altissima with emphasis on both inter- and intra-specific CAs in a common garden experiment.


We thank Bernd Blossy, Robert Johnson and Tim Connallon for helpful discussions and suggestions on earlier drafts of this study, and Rayko Halitschke for help with chemical analyses. The study was supported with funds from the National Science Foundation (USA, NSF-IOS 0950225) and Cornell University.