Evidence for Rapid Evolutionary Change in an Invasive Plant in Response to Biological Control

doi: 10.1111/jeb.13078

Evidence for rapid evolutionary change in an invasive plant in response to biological control

M. STASTNY &R.D.SARGENT

Department of Biology, University of Ottawa, Ottawa, ON, Canada

Keywords: Abstract biological control; We present evidence that populations of an invasive plant species that have evolutionary change; become re-associated with a specialist herbivore in the exotic range through herbivory; biological control have rapidly evolved increased antiherbivore defences invasive species; compared to populations not exposed to biocontrol. We grew half-sib fami- plant vigour; lies of the invasive plant Lythrum salicaria sourced from 17 populations near resistance; Ottawa, Canada, that differed in their history of exposure to a biocontrol tolerance. agent, the specialist beetle Neogalerucella calmariensis. In a glasshouse experiment, we manipulated larval and adult herbivory to examine whether a population’s history of biocontrol influenced plant defence and growth. Plants sourced from populations with a history of biocontrol suffered lower defoliation than na€ıve, previously unexposed populations, strongly suggesting they had evolved higher resistance. Plants from biocontrol-exposed populations were also larger and produced more branches in response to herbivory, regrew faster even in the absence of herbivory and were better at compensating for the impacts of herbivory on growth (i.e. they exhibited increased tolerance). Furthermore, resistance and tolerance were positively correlated among genotypes with a history of biocontrol but not among na€ıve genotypes. Our findings suggest that biocontrol can rapidly select for increased defences in an invasive plant and may favour a mixed defence strategy of resistance and tolerance without an obvious cost to plant vigour. Although rarely studied, such evolutionary responses in the target species have important implications for the long-term efficacy of biocontrol programmes.

selection regime may lead to phenotypic divergence of Introduction the new populations relative to their ancestral source The introduction of a species outside of its native range (Vellend et al., 2007; Lawrence et al., 2012). is often accompanied by a series of dramatic changes to Exotic species, and invasive plants in particular, are its ecological interactions. These changes can include often cited as an example of rapid evolutionary change the loss of many of its original associated species, in response to a shift in ecological context (Felker- including competitors, natural enemies and mutualists Quinn et al., 2013; Vandepitte et al., 2014; Wendling (Mitchell et al., 2006; Vellend et al., 2007; Stewart et al., et al., 2015). The loss of natural enemies during intro- 2015). The new ecological context is expected to drasti- duction has been a key explanation for the ecological cally alter selection on the species, both in terms of its advantage and spread of invasive species (i.e. enemy strength and direction (Mooney & Cleland, 2001; Facon release hypothesis (Keane & Crawley, 2002)), and by et al., 2006; Prentis et al., 2008). Over time, the novel extension, a proposed driver of trait divergence between the native and exotic ranges. For instance, release from natural enemies has been hypothesized to Correspondence: Michael Stastny, Atlantic Forestry Centre, Canadian produce a more invasive plant phenotype: genotypes Forest Service, 1350 Regent Street, PO BOX 4000, Fredericton, NB, that invest less in costly defence in favour of increased E3B 5P7, Canada. Tel.: +1-506-452-3026; fax: +1-506-452-3525; e-mail: allocation to growth (vigour) and reproduction are pre- [email protected] dicted to increase in frequency in the new range (i.e.

ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY. J. EVOL. BIOL. 30 (2017) 1042–1052 1042 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Rapid evolution in response to biocontrol 1043

evolution of increased competitive ability, or EICA, We designed a study to examine whether defensive hypothesis (Blossey & Notzold, 1995)). However, exper- and vegetative phenotypes of exotic populations of an imental support for increased vigour and reduced anti- invasive plant reflect recent selection imposed by re- herbivore defences in comparison with native association with its specialist natural enemy through populations has been mixed (Stastny et al., 2005; Abhi- biocontrol. Using a glasshouse experiment, we lasha & Joshi, 2009), and these two predicted outcomes addressed this largely overlooked question in the are not always examined jointly (Felker-Quinn et al., prominent wetland invader, purple loosestrife (Lythrum 2013). salicaria, hereafter Lythrum), that had, until its recent Traits involved in antiherbivore defence may function biocontrol programme, enjoyed over 150 years of pro- not only to reduce the amount of herbivory (i.e. resis- liferation in North America without its specialist herbi- tance), but also to compensate for its negative effects vores. By manipulating the presence and absence of on fitness (i.e. tolerance); plant tolerance to herbivory feeding damage by its specialist herbivore, the leaf bee- may be especially useful in mitigating the impacts of tle Neogalerucella, we examined the evidence for rapid generalist enemies acquired in the new range (Schaff- adaptation to biocontrol using replicate Lythrum popula- ner et al., 2011). Furthermore, the cost of resistance tions from a small, homogenous portion of its exotic traits (e.g. secondary metabolites) has proven difficult range that differed in their history of exposure to bio- to demonstrate (Zust€ et al., 2015); tolerance may be control. We predicted that plants from populations that linked to general plant vigour (Simms, 2000), and her- have been under selection by biocontrol for the last bivores may select for mixed defence strategies that two decades would suffer lower herbivory (i.e. show employ both resistance and tolerance (Carmona & For- increased resistance) and/or exhibit a greater ability to noni, 2013). Consequently, the evolutionary responses compensate for its effects (i.e. show increased toler- of invasive plants to the novel ecological context of ance), compared to plants from ‘na€ıve’ populations reduced enemy pressure are likely to be complex, par- (without prior exposure to biocontrol) whose antiherbi- ticularly given the suite of additional biotic and abiotic vore defences should reflect putatively relaxed levels. differences complicating comparisons between the Plants from populations with a more recent history of native and exotic ranges. exposure (< 20 years) were expected to exhibit inter- More direct evidence for rapid evolutionary change mediate levels of defence. Finally, we explored the rela- may be detected in situations where an invasive spe- tionship between resistance and tolerance among plant cies becomes re-associated with its key natural ene- genotypes from each of the three categories of biocon- mies in the new range. Specialized natural enemies, trol history to evaluate the role of joint defence strate- introduced incidentally or intentionally, can thus act gies and their long-term evolutionary implications for as agents of renewed natural selection on defence biological control in the exotic range of Lythrum. traits that, over many generations, may have relaxed or even reversed (Zangerl & Berenbaum, 2005; Rapo Materials and methods et al., 2010; Jogesh et al., 2014). Biological control agents of plants, in particular, provide a convenient Study species system to test this prediction: these strictly specialized herbivores tend to have a strong impact on the fit- Lythrum salicaria is a self-incompatible, herbaceous ness of the target invasive species and can be rela- perennial native to Eurasia. Since its introduction in tively unimpeded by ecological factors such as the early 19th century, it has become a prominent predation and competition in the exotic range invasive species in wetlands and disturbed, seasonally (Muller-Sch€ arer€ & Schaffner, 2008). As the biocontrol wet habitats across many regions of temperate North agent becomes successfully established, individuals of America. As Lythrum can form dense stands with pro- the invasive species least impacted by its herbivory, lific seed production, the species began to attract con- that is those that possess the greatest resistance and/ siderable public attention in the late 20th century. or tolerance, should contribute more genes to dwin- Government-backed management efforts resulted in dling future generations. Over time, populations the introduction of several species of insect biological under biocontrol may thus begin to diverge from control agents (Blossey et al., 2001). Lythrum has also ‘na€ıve’ populations that remain free of biocontrol. been the subject of basic research on evolutionary Long-term evolutionary change towards elevated changes in invasive species (Colautti & Barrett, 2013), defence may eventually compromise the efficacy of including the original test of the ‘Evolution of Increased biological control (Muller-Sch€ arer€ et al., 2004; Seast- Competitive Ability (EICA)’ hypothesis (Blossey & Not- edt, 2015). Yet, in spite of the numerous biocontrol zold, 1995). programmes introduced over the past half-century, Over 150 years after Lythrum’s initial introduction, experimental studies of the evolutionary responses of several species of its European specialist insect herbi- invasive plants to their biocontrol are nearly nonexis- vores were released in North America in the early tent (Rapo et al., 2010). 1990s as biological control agents. The Chrysomelid leaf

ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY. J. EVOL. BIOL. 30 (2017) 1042–1052 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 1044 M. STASTNY AND R. D. SARGENT

beetle Neogalerucella (Galerucella) calmariensis (and clo- several vegetative traits (relative growth rate, leaf size, sely related N. pusilla; hereafter, Neogalerucella) became specific leaf area) were measured to address separate the most widespread and effective in suppressing plant research objectives (M. Stastny & R. Sargent, unpub- growth and reproduction (Blossey et al., 2001). The lar- lished data). Then, due to space constraints, we selected vae and adults feed on foliage and meristematic tissue, eight families from each population such that the selec- including developing inflorescences. In most temperate tion roughly captured the mean and spread in the regions, the insect completes two generations per year; observed traits as calculated from the larger sample. A larval feeding in the first generation (mid-spring to late bootstrap simulation ascertained that this haphazard spring) can strongly impact plant growth and reproduc- selection method did not result in a distribution of traits tive ability, whereas the second generation (mid-sum- that was measurably different from a random sample of mer) additionally damages inflorescences (Dech & families. In February 2014, plants from the selected Nosko, 2002). families were bulk-germinated from the same seed material under conditions as described above, to produce eight replicates (half-sibs) for each of the eight Study sites and plant material families per population (i.e. 17 populations 9 8 fami- Seed material was collected from 17 populations of lies 9 8 replicates = 1088 plants). Lythrum spanning a relatively small and climatically uniform region of eastern Ontario, Canada, near Experimental treatment and data collection Ottawa (sampled area approximate dimensions: 140 km east–west and 120 km north–south; Table A1). Six After 9 weeks, we measured plant height (hereafter, were among the original (1990s) release sites of ‘initial size’), which is strongly correlated with biomass Neogalerucella in eastern Ontario (hereafter, ‘release’), as in young plants (M. Stastny and R. D. Sargent, unpub- determined by records from the Ontario Ministry of lished data), to assess differences in early plant growth Natural Resources (St. Louis, 2014). Six populations before the experimental treatments were imposed; cru- were identified as having been colonized by Neogaleru- cially, initial size was nearly identical among the three cella more recently, during the subsequent spread of the categories of biocontrol history (see Results and beetle in the region (hereafter, ‘recent’). The final five Fig. S1). In a split-plot design, we randomly assigned populations had not experienced any beetle herbivory four of the eight replicates in each half-sib family to (hereafter, ‘na€ıve’), as ascertained through repeated the herbivory treatment, which consisted of placing field surveys of the areas (see Table S1). We were four Neogalerucella larvae onto each plant. Second- and unable to quantify the incidence of other types of her- third-instar larvae were collected from a large bivory, which primarily included sporadic browsing by Neogalerucella population in a wetland near Ashton, deer, and floral damage by another introduced, com- Ontario (45.16N, 76.03W), located within the sampled monly co-occurring insect, Nanophyes marmoratus; how- region but not from any of the original biocontrol ever, Neogalerucella is the dominant herbivore (where release sites. To discourage the larvae from leaving their present) in this region (St. Louis, 2014). The popula- assigned plant, a transparent tube enclosure made of tions were selected haphazardly while avoiding any sys- cellulose acetate was placed around each plant. After tematic bias in site or population characteristics with 4 days of feeding, we removed the enclosures and any respect to the history of biocontrol, as well as excluding remaining larvae (others had pupated in the soil), and isolated small pockets of plants that appeared to be the estimated defoliation as the percentage of the leaf area result of a recent disturbance and colonization. At each consumed (hereafter, ‘larval defoliation’); such visual site, seeds from approximately 40 plants, spaced at least assessment provides a straightforward and repeatable 5 m apart, were collected in 2012; these represent half- measurement of damage (Johnson et al., 2016). In the sib maternal lines (hereafter, families). control treatment, the other four replicates in each In January 2013, 16 randomly selected families per family were grown in the absence of herbivory inside population were bulk-germinated under glasshouse identical tube enclosures. conditions on moist soil (Metro-Mix, Sun Gro Horticul- Approximately 1 week later, over 400 adult beetles ture) in narrow plastic containers (66 mL) placed in emerged from the soil over several days. The beetles racks in bottom-watering trays (Stuewe & Sons Inc., were allowed to move freely and feed on the plants in Tangent, OR, USA). Seedlings were thinned randomly the herbivory treatment, temporarily segregated with a until only a single individual remained in each con- mesh screen from the control treatment in a split-plot tainer, producing three replicates per family (i.e. design. All beetles were removed after 6 days, prior to 17 9 16 9 3 = 816 plants). Racks were rotated among oviposition, and again we visually estimated the total the trays weekly to minimize positional effects. Plants per cent leaf area consumed, now representing cumula- were grown at 26:18 °C and at 16-h: 8-h light: dark tive larval and adult herbivory (hereafter, ‘final defolia- photoperiod until flowering; due to the absence of pol- tion’). Our experimental treatment thus constituted a linators, the plants did not set seed. During this period, short-term bout of herbivory by the two life stages of

ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY. J. EVOL. BIOL. 30 (2017) 1042–1052 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Rapid evolution in response to biocontrol 1045

Neogalerucella that was realistic in the extent and nature The effects of the experimental treatment and bio- of feeding, which typically includes apical meristem control history on two aspects of plant vigour, final damage. above-ground biomass and branching (secondary The plants were subsequently grown under glass- shoots), were tested in a linear mixed effects model house conditions for another 9 weeks. Harvested (lmer) that included the fixed effects of herbivory treat- above-ground biomass was dried at 65 °C for 48 h and ment, biocontrol history (specified with a priori orthog- weighed, and secondary shoots (branching propensity) onal contrasts), the respective interactions and the were counted. Three weeks after harvest, we surveyed random effects as described above. The R package the shoots regrowing from the unharvested below- lmerTest was used to obtain P-values for the five con- ground parts. We measured the total length of all the trasts based on Kenward–Roger’s approximations (Hale- regrowth (i.e. the sum of all shoots from each root koh & Højsgaard 2014). To test potential differences in crown) as an indirect measure of below-ground alloca- initial plant vigour prior to experimental treatments, tion and subsequent regrowth capacity. we used the same approach to analyse initial plant height. Plant regrowth capacity (total length of regrowing Statistical analyses shoots) was analysed in a generalized linear mixed All of the analyses involved fitting linear mixed effects effects model (glmer with a gamma link function) that models (function lmer), generalized linear mixed included the fixed effects of herbivory treatment, bio- effects models (function glmer), or linear models control history (specified with a priori orthogonal con- (function lm) using R software (version 3.3.1; R Devel- trasts), the respective interactions and the random opment Core Team 2016). In the mixed effects models, effects as described above. biocontrol history was a three-level fixed effect (na€ıve, We further examined the patterns of variation in recent, release) that was crossed with the experimental growth and defence traits in an analytical framework treatment (two-level fixed effect denoting herbivory that focused on family means pooled within each cate- vs. control) to analyse herbivory effects on the gory of biocontrol history (means calculated from four response variables of biomass and regrowth, as well as replicates per family per treatment). First, we tested to assess the variation in initial plant size immediately whether plant vigour in the absence of damage was prior to the treatments. In all mixed effects models, correlated with vigour under herbivory, using Pearson population (nested within biocontrol history, family correlation of the above-ground biomass in the two (nested within population) and rack (a physical unit treatments. In a similar approach using family means, holding replicates; not a true block) were included as we then explored a putative trade-off between resis- random effects. The inclusion of all fixed (overall) and tance and tolerance to herbivory. For each of the 136 random effects was determined using log likelihood half-sib families, we first calculated resistance, defined tests (Bolker 2008). The models specified a priori as the family mean of % final defoliation subtracted orthogonal contrasts (Ruxton & Beauchamp, 2008) from 100 (i.e. [100 – damage]). Second, we calculated that tested specific biological hypotheses concerning an absolute metric of tolerance (i.e. impact of damage the history of exposure to biocontrol. The first contrast on biomass) as the difference between the mean family tested whether na€ıve populations differed from those biomass in the control and the herbivory treatments. with biocontrol history (i.e. the average of recent and Accordingly, increasingly negative values indicated fam- release populations). The second contrast compared ilies with lower tolerance (i.e. those exhibiting the the recent and longest history of biocontrol, that is greatest absolute reductions in growth following dam- tested the effect of the length of the plant’s recent age) and positive values indicated overcompensation association with Neogalerucella. In two-factor analyses, under glasshouse conditions. Using separate linear respective interactions involving each of these two regressions for the three categories of biocontrol his- contrasts and the herbivory treatment were tested in tory, we then examined the relationship between resis- addition to the overall effect of the treatment; there- tance and absolute tolerance across families. To account fore, these models involved five orthogonal contrasts for differences in defoliation and plant size, respec- (Table S2). tively, we additionally examined two alternative met- The effect of biocontrol history on larval and final rics of tolerance: (i) an impact on biomass per unit (cumulative) defoliation was tested in generalized linear damage, in which the absolute difference between the mixed effects models (glmer) with biocontrol history as mean family biomass in the control and the herbivory a fixed effect, and the random effects as described treatments was divided by the mean of its defoliation above. The analysis employed a gamma link function (Lehndal & Agren, 2015), and (ii) a proportional (i.e. that best approximated the distribution of each of the relative rather than absolute) impact of herbivory on two response variables, as checked by standard diagnos- biomass, in which the difference was divided by the tics. Degrees of freedom were estimated through penal- biomass in the control treatment (Wise & Carr, 2008). ized quasi-likelihood (glmmPQL function in R). For all three approaches, we compared the regression

ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY. J. EVOL. BIOL. 30 (2017) 1042–1052 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 1046 M. STASTNY AND R. D. SARGENT

lines for the three categories in an analysis of covari- (a) 30 ance (ANCOVA), fitting a linear model (function lm). We specified two a priori orthogonal contrasts for the main effect of biocontrol history, as described above. For each 25 contrast, a significant interaction term involving resistance (covariate) and the main effect would indicate a difference between the respective regression slopes. 20

Results 15

Resistance Plants derived from na€ıve populations suffered signiﬁ- 10 cantly higher larval and cumulative herbivory than % larval defoliation those from populations with a history of exposure to biocontrol. Speciﬁcally, larval defoliation was 35% 5 greater (Fig. 1a), and cumulative (combined larval and adult) defoliation was 26% greater (Fig. 1b) in the populations that had not come into contact with Neogalerucella 0 compared to those exposed to biocontrol (a priori con- NAÏVE RECENT RELEASE trasts: t = 3.54 and Pr(>|z|) = 0.0033; t = 2.306; Pr 14 14 40 (>|z|) = 0.021, respectively), in spite of similar initial (b) plant size immediately prior to the application of the larvae (Fig. S1; aprioricontrasts within the herbivory treatment: t133 = 1.18 and Pr(>|z|) = 0.238; t133 = 0.822; Pr(>|z|) = 0.412, respectively). Larval and cumulative 30 defoliation was similar between populations from the original release sites and those more recently exposed to biocontrol (Fig. 1; a priori contrasts: t14 = 0.043 and Pr(>|z|) = 0.816; t = 0.483; Pr(>|z|) = 0.629, 14 20 respectively).

Plant vigour % final defoliation The patterns of herbivory described above shaped the 10 differences in the final above-ground biomass at the end of the experiment, which was reduced overall by 21% by Neogalerucella feeding (t21.4 = 7.40, Pr(>| | < z ) 0.0001). Plants from populations associated with 0 biocontrol (i.e. recent and release) were, on average, NAÏVE RECENT RELEASE 40% larger after the herbivory treatment, when com- € pared to plants from naıve populations (Fig. 2a; a priori Fig. 1 Per cent leaf area consumed through larval herbivory (a) = contrast within the herbivory treatment: t14 2.94, Pr and through cumulative (larval and adult) herbivory (b) by the (>|z|) = 0.011). However, because plants from biocon- biocontrol beetle Neogalerucella calmariensis, comparing populations trol-associated populations were also 21% larger in the of Lythrum salicaria in eastern Ontario, Canada, with no exposure control group compared to plants from na€ıve popula- to biocontrol (na€ıve; n = 5), with more recent (< 20-year) 0 tions, the impact of herbivory on biomass differed exposure (2 ; n = 6), and from original biocontrol release sites with ~ = depending on the history of exposure to biocontrol (in- 20 years of association (release; n 6). Error bars indicate standard errors. teraction effect involving a priori contrast: t942.4 = 1.95, Pr(>|z|) = 0.046). For instance, whereas herbivory reduced biomass in all three categories, its impact was original release populations reached similar final bio- € most pronounced in the naıve populations (29% mass (Fig. 2a; a priori contrast: t15.4 = 0.778, Pr(>| decrease between control versus herbivory; Fig. 2a) and z|) = 0.448). Although plants from the populations the weakest in populations from the original release associated with Neogalerucella also showed increased sites (15% decrease; Fig. 2a). The length of biocontrol production of secondary shoots by 17% (control) and association was not a significant predictor of vigour, as 36% (damaged) compared to those from na€ıve popula- plants in both the more recently exposed and the tions, these differences were only significant under

Control Herbivory recent biocontrol association and those from original = = (a) 14 release sites (a priori contrast: t14 0.772, P 0.453).

12 Tolerance Across family means, plant vigour (above-ground bio- 10 mass) in the presence versus absence of herbivory was positively correlated in each ofthethreecategoriesofhis- 8 tory of exposure to biocontrol (Pearson’s r = 0.536, 0.523, 0.310; P < 0.001, P < 0.001, P = 0.032 for na€ıve, recent 6 and release families, respectively). In other words, genotypes that showed vigorous growth when undamaged 4 also tended to perform well under herbivory, irrespective of biocontrol history. However, when the absolute difference in mean family biomass between the two treatments Above-ground biomass (g) 2 (i.e. metric of tolerance) was ﬁtted against mean family resistance (= 100 – % defoliation), outcomes of these lin- 0 ear regressions became contingent on biocontrol history. NAÏVE RECENT RELEASE Families from na€ıve populations showed no relationship between resistance and tolerance (Fig. 3a; adj. 2 (b) 25 R = 0.011, P = 0.458). Yet, there was a signiﬁcant positive association between these two axes of plant defence among families from populations exposed to biocontrol 2 20 more recently (Fig. 3b; adj. R = 0.095, P = 0.019), and an even stronger association among those from the release sites (Fig. 3c; adj. R2 = 0.198, P < 0.001); overall, 15 the least resistant families tended to be the least tolerant of herbivory. Correspondingly, the slope of the regression line involving the families from the na€ıve populations differed from those in the two groups with biocontrol his- 10 tory (ANCOVA, interaction involving resistance and apriori contrast: t135 = 2.185, P = 0.031), whereas the slopes for the two groups with previous exposure to biocontrol did 5 not differ (interaction term: t95 = 1.184, P = 0.238). We Number of secondary shoots found similar patterns using two alternative metrics of tolerance: one that accounted for differential resistance 0 among plants, and the other that focused on proportional NAÏVE RECENT RELEASE (relative rather than absolute) impacts on biomass (see Statistical Analyses). Respectively, resistance was most Fig. 2 Final above-ground biomass (a) and number of secondary strongly positively associated with both per unit damage shoots (b) in Lythrum salicaria in the absence (control; light bars) tolerance (Fig. S3) and with proportional tolerance and presence (dark bars) of manipulated herbivory by specialist (Fig. S4) among families from the release populations, leaf beetle Neogalerucella calmariensis, comparing populations from eastern Ontario, Canada, with no exposure to biocontrol (na€ıve; whereas this association was consistently absent among € n = 5), with more recent (< 20-year) exposure (20; n = 6), and the naıve families with no history of biocontrol (slope not from original biocontrol release sites with ~ 20 years of association different from zero: P = 0.254 and P = 0.127, respec- (release; n = 6). Error bars indicate standard errors. tively). In both supplementary approaches, regression slopes for the families from the na€ıve populations consis- herbivory (Fig. 2b; aprioricontrasts by treatment: tently differed from those with a history of biocontrol (in- t14 = 1.42, Pr(>|z|) = 0.177, t14 = 2.68, Pr(>|z|) = 0.026, teraction term involving aprioricontrast: t135 = 2.116, respectively). P = 0.036 and t135 = 2.208, P = 0.029, respectively), but Post-harvest regrowth mirrored the general patterns not between the two categories with the previous expo- observed for the above-ground biomass. Collectively, sure to biocontrol. plants from populations associated with biocontrol tended to regrow 25% more vigorously overall, com- Discussion pared to the na€ıve populations (Fig. S2; a priori contrast: t14 = 2.59, Pr(>|z|) = 0.021). Regrowth capacity The ability of invasive species to undergo rapid evolu- was similar among plants from populations with a tionary change has been invoked as one of the key

ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY. J. EVOL. BIOL. 30 (2017) 1042–1052 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 1048 M. STASTNY AND R. D. SARGENT

(a) 50 Neogalerucella leaf beetles, received signiﬁcantly lower NAÏVE defoliation than na€ıve populations with no exposure to biocontrol, strongly suggesting that they have evolved 0 increased resistance to herbivory (Fig. 1). Moreover, biocontrol-associated populations were more vigorous € –50 adj. R2 = –0.011 than the naıve populations in terms of their above- P = 0.458 ground biomass, branching and regrowth ability, both Tolerance in the presence and absence of experimental herbivory, –100 and were better able to compensate for the effects of 40 50 60 70 80 90 Neogalerucella damage on their growth, suggesting that they have evolved increased tolerance to herbivory. (a) 50 RECENT Finally, by examining plant genotypes pooled across the categories of biocontrol history, we present evi- 0 dence for an increasingly robust, positive association between resistance and tolerance in response to exposure to biocontrol, suggestive of selection for a mixed –50 adj. R2 = 0.094 P = 0.019 defence strategy under increased plant vigour.

Tolerance In spite of the extensive body of research on invasive –100 species, our study is one of the first investigations of 40 50 60 70 80 90 evolutionary changes in plant resistance within the exotic range that may have arisen in response to bio- 50 control, although we do not identify specific resistance (c) RELEASE traits. We are aware of only three other studies that tested a similar prediction and only one in the context 0 of biological control. In a common-garden study of adj. R2 = 0.198 tansy ragwort, Jacobaea vulgaris, Rapo et al. (2010) –50 P = 0.0009 found that biocontrol-exposed exotic populations from New Zealand and North America suffered higher her- Tolerance –100 bivory by specialist flea beetles than those with no prior exposure to biocontrol, although their divergent pat- 40 50 60 70 80 90 terns in defensive chemistry were more equivocal and Resistance did not match the predictions. A second study, contrast- ing New Zealand populations of the wild parsnip, Pasti- Fig. 3 Linear regressions of resistance (100 minus per cent nacea sativa, recently re-associated with its specialist defoliation by leaf beetle Neogalerucella calmariensis larvae and (nonbiocontrol) herbivore (Jogesh et al., 2014), adults) and tolerance (absolute difference in above-ground detected similar secondary chemistry in populations biomass between undamaged and damaged plants), comparing with and without the parsnip webworm, but also half-sib families of Lythrum salicaria from eastern Ontario, Canada, showed that the 3- or 6-year exposure to its herbivory with a) no exposure to biocontrol (n = 40 families); b) more recent (< 20-year) exposure (n = 48 families); and c) from original coincided with reduced negative effects of damage, that biocontrol release sites with ~ 20 years of association (n = 48 is increased tolerance. Finally, rapid evolution of resis- families). Each point represents a mean of four family replicates. tance was demonstrated in the introduced goldenrod Solidago gigantea in Japan in response to the spread of an exotic herbivorous lace bug in the past decade factors behind their success in the new range (Keller & (Sakata et al., 2014). All other evidence for evolution- Taylor, 2008; Renaud et al., 2015; Wendling et al., ary change in antiherbivore defence in invasive species 2015). Biological control presents a unique opportunity comes from comparisons between exotic and native to examine the potential of introduced organisms to populations (Bossdorf et al., 2005; Stastny et al., 2005; rapidly respond to selection. Following a period of Colautti et al., 2009; Felker-Quinn et al., 2013), which refuge from specialist natural enemies (Keane & Craw- involve longer time scales for natural selection to act, ley, 2002), an invasive species is hypothesized to face as well as a wider plethora of biological, genetic and strong selection when re-associated with a key antago- abiotic factors that could generate phenotypic differ- nist that is itself likely to have few natural enemies or ences between the ranges (Keller & Taylor, 2008). competitors (Muller-Sch€ arer€ et al., 2004; Rapo et al., Evolutionary change towards larger, more vigorous 2010; Seastedt, 2015; Paynter et al., 2016). In a rare test phenotypes has been reported in the exotic ranges of a of this hypothesis, we found that Lythrum from popula- number of invasive plant species (Felker-Quinn et al., tions that had the longest (approx. 20-year) history of 2013; Turner et al., 2013) and is often attributed to a re-association with their specialist herbivore, reduced investment in costly antiherbivore defences

under reduced herbivore pressure. Renewed selection The diversity of herbivore community, another on defence traits imposed by re-associated natural ene- proposed mechanism for the maintenance of higher mies, such as biocontrol agents, may then occur at the levels of both strategies (Turley et al., 2013), is too expense of vegetative growth. Instead, we found that depauperate in Lythrum’s exotic range to drive the biocontrol-associated Lythrum populations were more mixed defence strategy. Alternatively, it could arise vigorous than the na€ıve populations both in the pres- through other phenotypic traits that influence damage ence and in the absence of Neogalerucella herbivory and/or compensatory growth, such as allocation or phe- (Figs 2a,b and S2). These findings suggest that nology, which may itself be under selection by special- increased defences may not incur an obvious cost in ist herbivores (Fukano et al., 2013). terms of growth, at least under glasshouse conditions The observed phenotypic divergence among popula- that unavoidably ignore ecological costs, natural phe- tions differing in the history of exposure to biocontrol nology or the perennial life history of Lythrum. Alter- may have been influenced by transgenerational (mater- natively, given that we observed no differences in the nal) effects, or it may only partly reflect the evolution- initial plant size (immediately prior to herbivory ary change in the populations under biocontrol. manipulation), final differences in above- Maternal effects were not accounted for by our study; ground biomass may have been driven by phenological however, these tend to be most prominent in the early differences rather than growth rate; Neogalerucella has stages of growth (Roach & Wulff, 1987; Weiner et al., been shown to impose selection on earlier flowering in 1997). We found no differences in plant size after the Lythrum (Thomsen & Sargent, 2017). Regardless of the first 9 weeks among the three categories of biocontrol mechanism, the greater vigour of the more resistant history (Fig. S1), strongly suggesting that variation in phenotypes also points to the role of other aspects of the maternal environment was unlikely to generate the the phenotype beyond the resistance traits that have observed differences in vigour. Whereas we cannot been the focus of prior studies (Rapo et al., 2010; entirely exclude the possibility of maternal effects influ- Jogesh et al., 2014). Specifically, our findings under- encing plant defence traits (Agrawal, 2002), it is unli- score the possibility that rapid evolutionary responses kely that the divergence in the patterns of covariance of Lythrum to biocontrol additionally involve tolerance in resistance and tolerance (Fig. 3) could be attributed – a compensatory strategy under selection by the bio- solely to maternal effects. Finally, the source material control agent that likely involves components of plant in our experiment may have affected our ability to growth, such as photosynthetic ability, allocation and evaluate long-term consequences of the inferred evolu- phenology (Fornoni, 2011). tionary change. By collecting seed, we may have The ability of plants to tolerate herbivory is an undersampled the existing heritable variation in the underexplored area of invasion biology, in spite of its biocontrol-associated populations by omitting the non- implications for long-term outcomes of biocontrol pro- reproducing (and potentially less resistant and/or less grammes (Wang et al., 2011). In populations with the vigorous) plants that may not have yet been fully erad- more recent and especially with the longest (~ 20-year) icated through genotype sorting. Under certain ecologi- history of biocontrol exposure, the most resistant geno- cal conditions (e.g. decline in Neogalerucella densities), types were also best able to compensate for specialist these long-lived genotypes may once again contribute damage (Fig. 3), counter to theoretical predictions (but to future generations, thereby influencing the strength see Leimu et al., 2006). In contrast, genotypes from and direction of natural selection on defence and vig- na€ıve populations exhibited no covariance between our through evolutionary feedbacks (Lankau & Strauss, these two axes of plant defence. Alternative metrics of 2011). tolerance which accounted for genotypic differences in One of the challenges of comparing populations of defoliation (Fig. S3) or plant size (Fig. S4), respectively invasive plants with different history of exposure to (see Wise & Carr, 2008; Lehndal & Agren, 2015), biocontrol is the issue of nonrandom sampling (Carson yielded a similar pattern: genotypes with the longest et al., 2008). Specifically, the observed phenotypic history of biocontrol consistently exhibited the stron- divergence may be due to pre-existing differences gest positive association between resistance and toler- among populations, or population types, rather than ance. Our results suggest that tolerance traits may be adaptation in response to biocontrol. Crucially, across an important component of a plant’s defensive strategy all measured variables the populations with the more against negative fitness impacts of specialist herbivores recent (< 20-year) history of biocontrol, which con- (Carmona & Fornoni, 2013), possibly because under sisted of a random sample from hundreds of Lythrum high densities of Neogalerucella following the initial populations secondarily colonized by Neogalerucella, build-up of biocontrol, few Lythrum individuals may be tended to align with those with the longest history of able to minimize damage through resistance traits. The biocontrol, suggesting an outcome of genotype sorting observed positive covariance between the two axes of that favoured the more resistant or tolerant individuals. defence may be a result of a genetic correlation These patterns were robust and consistent with an between resistance and tolerance (Stinchcombe, 2002). adaptive response to selection, as previously

ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY. J. EVOL. BIOL. 30 (2017) 1042–1052 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 1050 M. STASTNY AND R. D. SARGENT

demonstrated in Lythrum’s rapid local adaptation to cli- Natural Sciences and Engineering Research Council of mate (Colautti & Barrett, 2013), rather than prior dif- Canada (NSERC) Discovery Grant and an Early ferences or stochastic phenotypic selection (Keller & Researcher Award (Ontario Ministry of Research and Taylor, 2008; Colautti & Barrett, 2013). Unfortunately, Innovation) to RDS. we do not have records or accurate estimates of the length of time of biocontrol exposure in these popula- References tions. In animal systems, a similar approach of using replicate populations varying in the length of associa- Abhilasha, D. & Joshi, J. 2009. Enhanced fitness due to higher tion with a novel predator has demonstrated rapid evo- fecundity, increased defence against a specialist and toler- lution of defences, for instance in frogs adapting to an ance towards a generalist herbivore in an invasive annual – invasive crayfish (Nunes et al., 2014). Experimental plant. J. Plant Ecol. 2:77 86. Agrawal, A.A. 2002. Herbivory and maternal effects: mecha- manipulations of such chronosequences of (re-)associa- nisms and consequences of transgenerational induced plant tions between species can help elucidate changes in the resistance. Ecology 83: 3408–3415. underlying traits and explore their long-term coevolu- Blossey, B. & Notzold, R. 1995. Evolution of increased compet- tionary dynamics. itive ability in invasive nonindigenous plants - a hypothesis. The potential for an invasive species to evolve in J. Ecol. 83: 887–889. response to pressure exerted by biological control has Blossey, B., Skinner, L.C. & Taylor, J. 2001. Impact and man- been largely overlooked in evolutionary and applied agement of purple loosestrife (Lythrum salicaria) in North ecology, with a very limited number of empirical stud- America. Biodivers. Conserv. 10: 1787–1807. – ies (Rapo et al., 2010; Paynter et al., 2016) in spite of Bolker, B. 2008. Ecological Models and Data in R. pp. 254 258. the existence of multidecadal biocontrol programmes Princeton University Press, Princeton, NJ. Bossdorf, O., Auge, H., Lafuma, L., Rogers, W.E., Siemann, E. for many exotics. Our findings suggest that rapid evolu- & Prati, D. 2005. Phenotypic and genetic differentiation tionary change among populations exposed to biocon- between native and introduced plant populations. Oecologia trol can occur over very short timescales and involve 144:1–11. diverse aspects of the plant’s phenotype. The re-associa- Carmona, D. & Fornoni, J. 2013. Herbivores can select for mixed tion between Lythrum and Neogalerucella leaf beetles in defensive strategies in plants. New Phytol. 197: 576–585. the past two decades appears to have selected for Carson, W.P., Hovick, S.M., Baumert, A.J., Bunker, D.E. & increased resistance, plant vigour and ability to tolerate Pendergast, T.H. 2008. Evaluating the post-release efficacy of damage, compared to populations still unaffected by invasive plant biocontrol by insects: a comprehensive – biocontrol. In addition, the trend towards a positive approach. Arthropod-Plant Interact. 2:77 86. relationship between resistance and tolerance among Colautti, R.I. & Barrett, S.C.H. 2013. Rapid adaptation to cli- mate facilitates range expansion of an invasive plant. Science the genotypes from populations under biocontrol, con- 342: 364–366. trary to the expected trade-off (Lehndal & Agren, Colautti, R.I., Maron, J.L. & Barrett, S.C.H. 2009. Common 2015), suggests selection for a mixed defence strategy garden comparisons of native and introduced plant popula- without an obvious cost in terms of plant vigour. Evi- tions: latitudinal clines can obscure evolutionary inferences. dence of such rapid phenotypic differentiation has Evol. Appl. 2: 187–199. important implications for the evolutionary dynamics Dech, J.P. & Nosko, P. 2002. Population establishment, disper- of invasive species in the exotic range, as well as for sal, and impact of Galerucella pusilla and G. calmariensis, intro- the long-term efficacy of biological control. On the duced to control purple loosestrife in Central Ontario. Biol. – other hand, these phenotypic changes may in turn Control 23: 228 236. select for evolutionary responses in the biocontrol agent Facon, B., Genton, B.J., Shykoff, J., Jarne, P., Estoup, A. & David, P. 2006. A general eco-evolutionary framework for (McEvoy et al., 2012) if sufficient genetic variation understanding bioinvasions. Trends Ecol. Evol. 21: 130–135. exists, particularly given the much shorter generation Felker-Quinn, E., Schweitzer, J.A. & Bailey, J.K. 2013. Meta- times of insects. Further research is needed to examine analysis reveals evolution in invasive plant species but little the potential for such coevolutionary dynamics support for evolution of increased competitive ability (EICA). between invasive plants and their insect biocontrol Ecol. Evol. 3: 739–751. agents. Fornoni, J. 2011. Ecological and evolutionary implications of plant tolerance to herbivory. Funct. Ecol. 25: 399–407. Fukano,Y.,Tanaka,K.&Yahara,T.2013.Directionalselectionfor Acknowledgments early flowering is imposed by a re-associated herbivore - but no – We thank Jake Russell-Mercier, Max Schmidt, Braydon evidence of directional evolution. Basic Appl. Ecol. 14: 387 395. Halekoh, U. & Højsgaard, S. 2014. A Kenward-Roger Approxi- Hall and Nada Bochor along with past and present mation and Parametric Bootstrap Methods for Tests in Linear members of the Sargent Lab for technical support; Jes- Mixed Models The R Package pbkrtest. J. Stat. Software. 59: sica Forrest, Stuart Campbell and Nicolas Rode for dis- 1–32. cussions; and Howard Rundle, Rees Kassen and Jogesh, T., Stanley, M.C. & Berenbaum, M.R. 2014. Evolution anonymous reviewers for comments on earlier versions of tolerance in an invasive weed after reassociation with its of the manuscript. The study was supported by a specialist herbivore. J. Evol. Biol. 27: 2334–2346.

Johnson, M.T.J., Bertrand, J.A. & Turcotte, M.M. 2016. Preci- Roach, D. & Wulff, R.D. 1987. Maternal effects in plants. Annu. sion and accuracy in quantifying herbivory. Ecol. Entomol. Rev. Ecol. Syst. 18: 209–235. 41: 112–121. Ruxton, G.D. & Beauchamp, G. 2008. Time for some a priori Keane, R.M. & Crawley, M.J. 2002. Exotic plant invasions and thinking about post hoc testing. Behav. Ecol. 19: 690–693. the enemy release hypothesis. Trends Ecol. Evol. 17: 164–170. Sakata, Y., Yamasaki, M., Isagi, Y. & Ohgushi, T. 2014. An Keller, S.R. & Taylor, D.R. 2008. History, chance and adapta- exotic herbivorous insect drives the evolution of resistance tion during biological invasion: separating stochastic pheno- in the exotic perennial herb Solidago altissima. Ecology 95: typic evolution from response to selection. Ecol. Lett. 11: 2569–2578. 852–866. Schaffner, U., Ridenour, W.M., Wolf, V.C., Bassett, T., Muller-€ Lankau, R.A. & Strauss, S.Y. 2011. Newly rare or newly com- Scharer,€ H., Sutherland, S. et al. 2011. Plant invasions, gen- mon: evolutionary feedbacks through changes in population eralist herbivores, and novel defense weapons. Ecology 92: density and relative species abundance, and their manage- 829–835. ment implications. Evol. Appl. 4: 338–353. Seastedt, T.R. 2015. Biological control of invasive plant species: Lawrence, D., Fiegna, F., Behrends, V., Bundy, J.G., Philli- a reassessment for the Anthropocene. New Phytol. 205: 490– more, A.B., Bell, T. et al. 2012. Species interactions alter 502. evolutionary responses to a novel environment. PLoS Biol. Simms, E.L. 2000. Defining tolerance as a norm of reaction. 10: e1001330. Evol. Ecol. 14: 536–570. Lehndal, L. & Agren, J. 2015. Latitudinal variation in resis- St. Louis, E. (2014) The impact of two introduced herbivores tance and tolerance to herbivory in the perennial herb on the population ecology of Lythrum salicaria: implications Lythrum salicaria is related to intensity of herbivory and plant for plant performance, reproduction, and community diver- phenology. J. Evol. Biol. 28: 576–589. sity. pp. 25–26. Masters Thesis, Faculty of Science, Univer- Leimu, R., Koricheva, J. & Larsson, S. 2006. A meta-analysis sity of Ottawa, Ottawa, ON. of tradeoffs between plant tolerance and resistance to herbi- Stastny, M., Schaffner, U. & Elle, E. 2005. Do vigour of intro- vores: combining the evidence from ecological and agricul- duced populations and escape from specialist herbivores con- tural studies. Oikos 112:1–9. tribute to invasiveness? J. Ecol. 93:27–37. McEvoy, P.B., Grevstad, F.S. & Schooler, S.S. (2012) Insect Stewart, A.J.A., Bantock, T.M., Beckmann, B.C., Botham, invasions: lessons from biological control of weeds. In: Insect M.S., Hubble, D. & Roy, D.B. 2015. The role of ecological Outbreaks Revisited, (P. Barbarosa, D. Letourneau & A.A. interactions in determining species ranges and range Agrawal, eds.). pp. 395–428. Blackwell Publishing Limited, changes. Biol. J. Lin. Soc. 115: 647–663. Oxford. Stinchcombe, J.R. 2002. Can tolerance traits impose selection Mitchell, C.E., Agrawal, A.A., Bever, J.D., Gilbert, G.S., Huf- on herbivores? Evol. Ecol. 16: 595–602. bauer, R.A., Klironomos, J.N. et al. 2006. Biotic interactions Thomsen, C.J.M. & Sargent, R.D. 2017. Evidence that an her- and plant invasions. Ecol. Lett. 9: 726–740. bivore response affects selection on floral traits and inflores- Mooney, H.A. & Cleland, E.E. 2001. The evolutionary impact cence architecture in purple loosestrife (Lythrum salicaria). of invasive species. Proc. Natl. Acad. Sci. USA 98: 5446–5451. Ann Bot.. https://doi.org/10.1093/aob/mcx026. Muller-Sch€ arer,€ H. & Schaffner, U. 2008. Classical biological Turley, N.E., Godfrey, R.M. & Johnson, M.T.J. 2013. Evolution control: exploiting enemy escape to manage plant invasions. of mixed strategies of plant defense against herbivores. New Biol. Invasions 10: 859–874. Phytol. 197: 359–361. Muller-Sch€ arer,€ H., Schaffner, U. & Steinger, T. 2004. Evolu- Turner, K.G., Hufbauer, R.A. & Rieseberg, L.H. 2013. tion in invasive plants: implications for biological control. Rapid evolution of an invasive weed. New Phytol. 202: Trends Ecol. Evol. 19: 417–422. 309–321. Nunes, A.L., Orizaola, G., Laurila, A. & Rebelo, R. 2014. Rapid Vandepitte, K., De Meyer, T., Helsen, K., Van Acker, K., Ron- evolution of constitutive and inducible defenses against an dan-Ruiz, I., Mergeay, J. et al. 2014. Rapid genetic adapta- invasive predator. Ecology 95: 1520–1530. tion precedes the spread of an exotic plant species. Mol. Ecol. Paynter, Q., Buckley, Y.M., Peterson, P., Gourlay, A.H. & Fowler, 23: 2157–2164. S.V. 2016. Breaking and remaking a seed and seed predator Vellend, M., Harmon, L.J., Lockwood, J.L., Mayfield, M.M., interaction in the introduced range of Scotch Broom (Cystisus Hughes, A.R., Wares, J.P. et al. 2007. Effects of exotic species scoparius) in New Zealand. J. Ecol. 104: 183–192. on evolutionary diversification. Trends Ecol. Evol. 22: 481– Prentis, P.J., Wilson, J.R.U., Dormontt, E.E., Richardson, D.M. 488. & Lowe, A.J. 2008. Adaptive evolution in invasive species. Wang, Y., Huang, W., Siemann, E., Zou, J., Wheeler, G.S., Trends Plant Sci. 13: 288–294. Carrillo, J. et al. 2011. Lower resistance and higher R Core Team. 2016. R: A Language and Environment for Statistical tolerance of invasive host plants: biocontrol agents reach Computing. R Foundation for Statistical Computing, Vienna, high densities but exert weak control. Ecol. Appl. 21: 729– Austria. http://www.R-project.org/. 738. Rapo, C., Muller-Sch€ arer,€ H., Vrieling, K. & Schaffner, U. Weiner, J., Martinez, S., Muller-Scharer, H., Stoll, P., Schmid, 2010. Is there rapid evolutionary response in introduced B. & Society, B.E. 1997. How important are environmental populations of tansy ragwort, Jacobaea vulgaris, when materal effects in plants? A study with Centaurea maculosa J. exposed to biocontrol? Evol. Ecol. 24: 1081–1099. Ecol. 85: 133–142. Renaud, S., Gomes Rodrigues, H., Ledevin, R., Pisanu, B., Cha- Wendling, C.C., Wegner, K.M. & Wendling, C.C. 2015. Adap- puis, J.-L. & Hardouin, E.A. 2015. Fast morphological tation to enemy shifts: rapid resistance to evolution to local response of house mice to anthropogenic disturbances on a Vibrio spp. in invasive Pacific oysters. Proc. R. Soc. B Biol. Sci. Sub-Antarctic island. Biol. J. Lin. Soc. 114: 513–526. 282: 20142244.

ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY. J. EVOL. BIOL. 30 (2017) 1042–1052 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 1052 M. STASTNY AND R. D. SARGENT

Wise, M.J. & Carr, D.E. 2008. On quantifying tolerance of Table S2 A priori orthogonal contrasts used in the anal- herbivory for comparative analyses. Evolution 62: 2429– yses of the factorial experiment. 2434. Figure S1 Comparison of initial plant size prior to Zangerl, A.R. & Berenbaum, M.R. 2005. Increase in toxicity experimental treatments. of an invasive weed after reassociation with its co- Figure S2 Comparison of regrowth capacity after evolved herbivore. Proc. Natl. Acad. Sci. USA 102: 15529– above-ground harvest. 15532. Zust,€ T., Rasmann, S. & Agrawal, A.A. 2015. Growth-defense Figure S3 Linear regression of resistance and tolerance tradeoffs for two major anti-herbivore traits of the common corrected for damage. milkweed Asclepias syriaca. Oikos 124: 1404–1415. Figure S4 Linear regression of resistance and tolerance relative to plant size.

Supporting information Data deposited at Dryad: doi:10.5061/dryad.k1mc5 Additional Supporting Information may be found online in the supporting information tab for this article: Received 8 September 2016; revised 17 March 2017; accepted 20 Table S1 List and geographic locations of sampled March 2017 Lythrum populations.

ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY. J. EVOL. BIOL. 30 (2017) 1042–1052 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2017 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY