Variable effects on growth and defense traits for plant ecotypic differentiation and phenotypic plasticity along elevation gradients

Abstract Along ecological gradients, phenotypic differentiation can arise through natural selection on trait diversity and magnitude, and environment‐driven plastic changes. The magnitude of ecotypic differentiation versus phenotypic plasticity can vary depending on the traits under study. Using reciprocal transplant‐common gardens along steep elevation gradients, we evaluated patterns of ecotypic differentiation and phenotypic plasticity of several growth and defense‐related traits for two coexisting but unrelated plant species, Cardamine pratensis and Plantago major. For both species, we observed ecotypic differentiation accompanied by plasticity in growth‐related traits. Plants grew faster and produced more biomass when placed at low elevation. In contrast, we observed fixed ecotypic differentiation for defense and resistance traits. Generally, low‐elevation ecotypes produced higher chemical defenses regardless of the growing elevation. Yet, some plasticity was observed for specific compounds, such as indole glucosinolates. The results of this study may suggest that ecotypic differentiation in defense traits is maintained by costs of chemical defense production, while plasticity in growth traits is regulated by temperature‐driven growth response maximization.


| INTRODUC TI ON
Species with wide distributions tend to exhibit large intraspecific variation in most functional and phenotypic traits. This geographical variation in biotic and abiotic factors across species distributions can lead to the evolution of morphologically and functionally different ecotypes (Hufford & Mazer, 2003;Kawecki & Ebert, 2004;Savolainen, Pyhäjärvi, & Knürr, 2007). Ecotypes are genetically distinct populations of a given species, displaying phenotypic traits that maximize fitness within a particular local abiotic and biotic conditions (Kawecki & Ebert, 2004). Along environmental gradients, trait-mediated local adaptations of plant ecotypes are the result of selection through phenotypic plasticity.
In particular, plant adaptation to habitat-specific abiotic and biotic factors can be studied along elevational transects with homogenous macroclimatic conditions, minimizing the effect of biogeographical history and barriers to gene flow Sundqvist, Sanders, & Wardle, 2013).
Plant growth and defense related traits have been shown to vary in response to different abiotic and biotic conditions. Therefore, it is expected that biogeographical gradients should select for clinal adaptation in such traits (Woods et al., 2012). Furthermore, growth and defense traits can be subjected to resource allocation tradeoffs, and the correlated expression of these traits should serve to maximize plant fitness within a given herbivory and climatic environment (Agrawal, Conner, & Rasmann, 2010). For instance, high and low-elevation Plantago lanceolata ecotypes growing at two temperature regimes (12 and 20°C to simulate cold and warm environment of different elevation gradients) showed strong plasticity in growth (i.e. both genotypes grew similarly within each environment), while their resistance to generalist herbivores reflected genetically-fixed patterns; high-elevation ecotypes were always less resistant, independently of the temperature regimes . Such differences would suggest that ecotypes growing at high elevation were selected to produce lower amounts of constitutive defenses because of lower amount of herbivory, while retaining a high degree of plasticity of growth-related responses to temperature. Such reciprocal transplant experiments have been used to measure the extent of ecotypic differentiation and phenotypic plasticity (Nahum, Inbar, & Ne'eman, and Ben-Shlomo., 2008), with the prediction that ecotypes adapted to one environment should change their phenotypes when placed in a novel environment, within their genetic constraints. Therefore, coupling reciprocal transplant with common garden experiments is critical because phenotypic plasticity of growth and defense traits in response to growing conditions can also generate clines, and such plasticity can obscure genetically based trait expression.
Here, we aim to measure the magnitude of ecotypic differentiation and plasticity in growth and defense traits for two unrelated plant species with similar geographical distribution along elevation gradients in the Alps (Supporting information Appendix S1: Figure   S1). Specifically, we will address the following questions: (a) is there ecotypic differentiation in plant growth and defense-related traits across an ecological gradient? (b) is there phenotypic plasticity in growth and defense-related traits across different plant ecotypes, and (c) what is the magnitude of phenotypic plasticity for both growth and defense-related traits along elevation gradients? To this end, we collected seeds of four populations of Cardamine pratensis (Brassicaceae) and six populations of Plantago major (Plantaginaceae); half of the populations originated from low elevation and the other half from high elevation (Supporting information Table S1). We reciprocally transplanted the high and low-elevation ecotypes at both their elevation of origin or at the opposite elevation using two common gardens along a mountain transect and assessed variation in growth and defense (secondary metabolite) related traits.

| Studied species
Cardamine pratensis is a rhizomatous perennial herb that grows in a variety of habitats including nutrient-rich meadows, pastures, and forests and is common throughout Europe and in Central and Eastern Asia (Hultén & Fries, 1986). C. pratensis populations cover a wide elevation range, from sea level to about 1600 meters above sea level (Aeschimann, Lauber, Moser, & Theurillat, 2004), flowering from April to June. Flowers are self-incompatible, and plants generally produce clonal offspring as new rosettes, especially under moist conditions (Lövkvist, 1956), and are considered hemicryptophyte (i.e. a long-lived geophyte with overwintering green leaves). All Cardamine pratensis tissues, including leaves, contain glucosinolates (GLS), which, when in contact with myrosinases enzymes, are degraded into glucose and sulfate, along with various nitrile, isothiocyanate, and thiocyanate molecules that are toxic or deterrent to both herbivores and pathogens (Giamoustaris & Mithen, 1995;Hopkins, Ekbom, & Henkow, 1998;Kliebenstein, Pedersen, Barker, & Mitchell-Olds, 2002;Lambrix, Reichelt, Mitchell-Olds, Kliebenstein, & Gershenzon, 2001). Glucosinolates are often classified into three classes of compounds depending on their side-chain: aliphatic, indole and aromatic, several of which have been shown to be effective against generalist and, to some extent, against specialist herbivores (Daxenbichler et al., 1991;Louda & Rodman, 1983;Montaut & Bleeker, 2011). Glucosinolates are known to vary quantitatively and qualitatively, across both individuals and populations of same species Mauricio, 1998). In addition, phenotypic plasticity in GLS production has been previously observed in wild brassicaceous species (Agrawal, Conner, Johnson, & Wallsgrove, 2002). For instance, GLS profiles of Boechera stricta were strongly plastic, both among habitats and within habitats, and patterns of GLS plasticity varied greatly among genotypes (Wagner & Mitchell-Olds, 2018).
Plantago major is a perennial (or facultatively perennial depending on environmental conditions) rosette-forming herbaceous plant. As a poor competitor, P. major generally grows in ruderal areas, especially along paths or roadsides and near gateways where grass is short or absent (Warwick & Briggs, 1980). Native to Eurasia, P. major is a cosmopolitan species. It reproduces both sexually (self-compatible wind pollinated) and asexually through rosette formation. Generally low genetic diversity among populations of P. major has been shown to favor ecotypic and phenotypic differentiation (Van Dijk, Wolff, & Vries, 1988;Halbritter, Billeter, Edwards, & Alexander, 2015;Warwick & Briggs, 1980). P. major can cover a very wide elevation range: from the sea level to alpine ecosystems up to 3,000 meters above sea level (Ren, Wang, Chen, & Zhu, 1999). P. major also produce notable amounts of secondary metabolites belonging to the class of cyclopentanoid monoterpenes, namely iridoid glycosides (IGs) and caffeoyl phenylethanoid glycosides (CPGs) (Pankoke, Buschmann, & Müller, 2013), which act as herbivore deterrents against generalist chewing insect (Fuchs & Bowers, 2004 Pellissier et al., 2014), and their production have been shown to display plasticity (Bowers & Stamp, 1992;Kuiper & Smid, 1985;Lotz & Blom, 1986).

| Experimental design
Cardamine pratensis seeds were collected from two low-elevation and two high-elevation populations along two elevation gradients of the Jura Mountains in Switzerland in 2016. Plantago major seeds where collected from three low-elevation and three high-elevation population along three elevation gradients in the Swiss Alps during summer 2016 (Supporting information Table S1). Seeds were collected on randomly selected plants (C. pratensis, n = 6 plants/population; P. major, n = 10 plants/population) within a 100 m radius for each population.
While we acknowledge that we have not measured plasticity in the strict sense across genotypes, we here assumed that within a

| Plant growth-related traits
After 8 weeks of growth in the field for both study species, aboveground biomass was separated from roots, oven-dried at 40°C for 48 hr and weighed to determine their dry biomass. Furthermore, in P. major plants, two additional growth-related traits were measured: (a) the chlorophyll content of the plant, which was measured as the average of three fully expanded leaves per plant using a SPAD-502Plus chlorophyll meter (Konica Minolta (China) Investment Ltd), (b) the specific leaf area (SLA), which was measured as the oneside area (calculated using ImageJ software) of the youngest fresh fully expanded leaf per plant divided by their oven-dried (40°C for 48 hr) biomass (mm 2 mg −1 DW) (Cornelissen et al., 2003). Higher SLA levels and chlorophyll content tend to positively correlate with potential relative growth rate, photosynthetic rate, or leaf nitrogen (N) across species (Garnier & Laurent, 1994;Poorter & Garnier, 2007).
Generally, species in resource-rich environments tend to have a higher SLA than those in resource-poor environments (Garnier & Laurent, 1994;Poorter & Garnier, 2007).

| Chemical analysis
For chemical analyses, sample preparation for each species followed different methods due to the different secondary metabolite extractions and analyses. ionization, using the method as described in (Glauser, Schweizer, Turlings, & Reymond, 2012).
Plantago major: at the end of the experiment, one young fully expanded leaf was oven-dried at 40°C for 48 hr prior being ground to powder using stainless steel beads in the tissue lyser. Then, 10 mg aliquots were weighed and 1.5 ml methanol was added to each tube along with 5 glass beads. The tubes were shaken 4 min at 30 Hz and centrifuged at 31,800 g for 3 min. The supernatant was diluted five times by adding 800 µl of MilliQ water to 200 µl of pure extract.
The injection volume was 1 μl. Detection was achieved in negative electrospray using deprotonated ions or formate adducts as quan- and 20 μg/ml. Concentrations were normalized to plant weight and expressed as μg/mg. Other Iridoid glycosides and caffeoyl phenylethanoid glycosides were putatively identified based on their retention time and chemical formula by comparing them to previous detection in P. major or in species of Plantago genus (Rønsted, Göbel, Franzyk, Jensen, & Olsen, 2000) and database (Dictionary of Natural Products, CRC Press, USA, version 6.1. on DVD) containing information on known IGs and CPGs and quantified as aucubin or verbascoside equivalents. Iridoid glycosides named with the code IG followed by numbers (Supporting information Figure S2) represent molecular formula corresponding to potential IG for which several isomers exist in the literature and thus cannot be unequivocally annotated.

| Herbivore bioassay
To measure plant resistance against insect herbivores (defined as the effect of plant defense traits on herbivore performance (Karban & Baldwin, 1997)), we used the generalist herbivore, Spodoptera littoralis (Lepidoptera: Noctuidae; obtained from Syngenta, Stein AG, Switzerland). S. littoralis is known to feed on species belonging to more than 80 families of plants (Brown & Dewhurst, 1975),

| Statistical analyses
All statistical analyses were performed within the R environment (R Development Core Team, 2017). For chemical data, we calculated the sum of glucosinolate compounds (GLS total) for C. pratensis and the sum of iridoid glycosides (IGs total) and caffeoyl phenylethanoid glycosides (CPGs total) for P. major, as well as a measure of chemical diversity for both plant species using the Shannon-Weaver diversity indices (Hill, 1973) with the diversity function in the vegan package in R (Oksanen et al., 2017).
To measure the interactive effects of elevation of origin and elevation of growth on plant growth and defense traits, we used two-way ANOVAs, including transplant sites (high and low), elevation ecotypes (high and low), and their interaction as fixed factors.
We also included the term population nested within elevation ecotypes in the model to assess variability across populations within a given elevation of origin. The response variables were aboveground biomass (AG biomass), larval weight gain, total GLS, total indole, total aliphatic, and chemical diversity for C. pratensis, and AG biomass, chlorophyll content, SLA, larval weight gain, total chemistry, total IGs, total CPGs and chemical diversity for P. major. All chemical traits were log-transformed prior analyses to meet normality and homoscedasticity assumptions. A significant effect of site of growth

| Plant chemical defenses and resistance
The glucosinolate profiles of C. pratensis leaves consisted of six GLS compounds (two aliphatic, three indoles and one aromatic), and the secondary metabolites profile of the P. major leaves F I G U R E 2 Cohen's d standardized effect sizes (±95% CI) for the influence of growing at opposite elevations of origin on plant growth and defense related traits, for high and low-elevation ecotypes of C. pratensis (a) and P. major (b) consisted of 13 IGs and 3 CPGs compounds (Supporting information Figure S2).
We found ecotypic effect in insect weight gain; larvae on lowelevation ecotypes grew 81% more compared to high-elevation ecotypes (Table 1, Figure 3b). Finally, we also found significant population-level effects for several traits (See Supporting information Figure S3 and Table 1), indicating that local differentiation in trait expression is also influenced by adaptation to different mountain transects.
In P. major, in terms of absolute compound quantities, low-elevation plants produced 17% more compounds in total, 17% more IGs, and 22% more CPGs (Figure 4, Table 1). The PERMANOVA (Table S2) revealed a plant ecotypic effect (elevation ecotype effect, F = 4.5; p = 0.001) and a growing elevation effect (F = 3.55; p = 0.006) (Figure 5c,d) in the abundance and diversity of secondary F I G U R E 3 Reaction norms of C. pratensis ecotypes for growth (a), larval weight gain (b) and defense (c, d, e, f) traits. Mean phenotypic values (mean ±1 SE for each elevation ecotype) are represented in black (low-elevation ecotypes) or gray (high-elevation ecotypes) across two contrasting growing elevations (high or low) metabolites in P. major. Additionally, we found that abundance of the total chemistry and diversity of the compounds were significantly affected by the AG biomass of P. major (F = 8.6; p = 0.001). For P.
major, we also observed signify cant effects of population-level effect on all the measured traits (marginal for SLA and chlorophyll content) (Supporting information Figure S4 and Table 1). Finally, we also found ecotypic differentiation for S. littoralis larval weight gain ( Figure 4d, Table 1): larvae on low-elevation ecotypes grew 8% more than on high-elevation ecotypes.

| D ISCUSS I ON
The major aim of this study was to elucidate on the variable responses of growth versus defense related traits using common gardens of plant ecotypes growing at different elevations. We observed ecotypic differentiation accompanied by plasticity in growthrelated traits, while we mainly observed ecotypic differentiation for defense traits for both P. major and C. pratensis. Below, we outline the potential causes for such divergence along elevation gradients.

| Plant biomass accumulation
Plasticity can be visualized as a change in the slope of the reaction norm between the ecotype at the elevation of origin and the same ecotype growing at opposite elevation (Doughty, 1995;Gotthard et al., 1995). In this regard, for both species, plant growth-related traits (AG biomass, leaf chlorophyll content and SLA) showed plasticity (Figures 2, 3a, 4a,b, c). Our results compliment other findings where the combination of ecotypic differentiation and F I G U R E 4 Reaction norms of P. major ecotypes of growth traits (a, b, c), larval weight gain (d) and defense traits (e, f, g (total chemistry), h). Mean phenotypic values (mean ±1 SE for each elevation ecotype) are represented in black (low-elevation ecotypes) or greay (highelevation ecotypes) across two contrasting growing elevations (high or low) TA B L E 1 Two-way ANOVA results, indicating interactions between the effects of high and low-elevation ecotypes and elevation of growth (in two common garden sites) on growth and defense traits phenotypic plasticity in growth-related traits such as biomass and flower size was shown for invasive species at their invasive range (Martín-Forés et al., 2017). More specifically, we observed that in both species, the AG biomass across both ecotypes was higher at low-elevation growing sites and lower at high-elevation growing sites (Figures 3a, 4a). Higher AG biomass production of both ecotypes at low-elevation growing site comes as no surprise, given the growing conditions at low-elevation are warmer and more favorable than at high elevation. Two reasons have been put forward for plants to reduce growth at high elevation. First, a decrease in the general metabolic activity as a function of colder temperature inhibits photosynthetic rate and biomass production (Boyer, 1982). Second, it has been proposed that because plants growing at higher elevations typically receive direct sunlight and higher ultraviolet radiation, and ultraviolet radiation destroys the auxins content at the apical shoots, they tend to grow much slower than lowland plants (Keller, Stahlberg, Barkawi, & Cohen, 2004). Furthermore, as both C. pratensis and P. major are perennial species, it could be argued that high-elevation ecotypes accumulated higher AG biomass than low-elevation ecotypes once placed in more favorable low-elevation conditions to compensate for the next year's growing season, when they would have to allocate more resource to flower and seed production. Such a scenario should be less likely for low-elevation plants growing at their elevation of origin. However, we make this argument with caution for P. major, since it is a facultative perennial plant.
Interestingly, we also observed that high-elevation ecotypes of both species always produced more biomass than low-elevation ecotypes (Figures 3a, 4a). This is somewhat surprising, since we expected alpine plants to grow smaller in harsher and colder environments (Atkin & Day, 1990;Körner, 2003). Plant size is negatively correlated with extremely cold temperatures (Squeo, Rada, Azocar, & Goldstein, 1991) and as a consequence, generally decreases with elevation (Körner, 2003). Plants adapted to high elevation, where growing season is short, should favor fast biomass accumulation (Körner, 2016). For instance, plants growing in colder conditions typically exhibit greater photosynthetic and respiratory capacities than their warmer-grown counterparts (Atkin, Loveys, Atkinson, & Pons, 2006). Therefore, highelevation ecotypes could benefit from faster development and high rates of metabolism (Körner, 2016), and, at equal growing conditions (same soil) and during the same growing timeframe, have actually accumulated more biomass than their low-elevation counterparts.
Finally, we also want to note that because we worked at the ecotypic level, one might argue that the plastic response we observed in growth-related traits might be driven by genotypic differences within each population. In other words, if a population is highly genetically differentiated, a random sampling would result in more likely piking highly plastic genotypes, which would drive the overall population mean change. If this were the case, larger (in our case lowland) populations should have shown higher levels of plasticity overall, but this was not the case (see Supporting information Figures S3 and S4).

| Plant chemical defenses and resistance
We observed ecotypic differentiation across most plant defense and resistance measures in both species. First, the ordination showed ecotypic differentiation for the overall secondary metabolite blend for both species (see Supporting information Table S2 and ecotypic segregation in the NMDS plot in Figure 5) despite the pattern Note. Signif. Codes for p-value: 0 "***" 0.001 "**" 0.01 "*" 0.05 " †" 0.1.
TA B L E 1 (Continued) of production (increase or decrease in concentration). Similarly, aliphatic GLS, chemical diversity, total IGs, total CPGs, and larval weight also clearly showed ecotypic differentiation for both species.
( Figures 3e,f, 4d,e,f). Generally, regardless of the growing elevation, low-elevation ecotypes produced more chemical defenses (Figures   3c, 4g). These results are in line with other findings showing cold temperature-driven suppression of plant secondary metabolites , and a general decrease in secondary metabolite production at high elevation (Kergunteuil et al., 2018). However, a decrease in secondary metabolite production in high-elevation ecotypes could also be attributed to a decrease in herbivory pressure at high elevation. To date, we have no data that allows disentangling biotic and abiotic effects of defense decline at high elevation, but likely both synergistically interact for selecting such a chemical phenotype .
Interestingly, however, indole GLS showed no ecotypic differentiation: high-elevation ecotypes produced more of these compounds when placed at low-elevation (see ecotype × environment effect in Table 1). Unlike aliphatic GLS, for which induction has been rarely observed (Koritsas, Lewis, & Fenwick, 1991;Li, Kiddle, Bennett, & Wallsgrove, 1999), induction of indolic GLS has been widely documented in several systems (Agrawal, Strauss, & Stout, 1999;Doughty, Kiddle, Pye, Wallsgrove, & Pickett, 1995;Griffiths, Birch, & Macfarlane-Smith, 1994;Moyes, Collin, Britton, & Raybould, 2000;Raybould & Moyes, 2001;Siemens & Mitchell-Olds, 1998), including in the closely related Cardamine hirsuta (Bakhtiari, Glauser, & Rasmann, 2018). In addition, indole GLS have been previously shown to be strongly influenced by environmental factors, suggesting favorable selection pressures for plasticity in this class of secondary metabolites. If plasticity is a means of F I G U R E 5 Non-metric multidimensional scaling (NMDS) plot of Cardamine pratensis plant ecotype of (a: high elevation and b: lowelevation common gardens) and Plantago major (c: high elevation and d: low-elevation common gardens). Distance matrices were generated using secondary metabolite (glucosinolates in C. pratensis and iridoid glycosides and caffeoyl phenylethanoid glycosides for P. major) concentrations and diversity. The 95% confidence interval ellipses are represented based on the two elevation ecotypes (high-elevation ecotype in gray and low-elevation ecotype in black). Stress values: (a) and (b) = 0.12, (c) and (d) = 0.2, K = 2 saving energy (Bidart-Bouzat, Mithen, & Berenbaum, 2005;Traw, 2002), this could indicate that the production of indole GLS might be more costly than the production of other GLS in C. pratensis at high elevation. On the other hand, it might also indicate that temperature dictates indole GSL production more than other classes of GSLs, because indole GSL compounds are intrinsically more inducible. In other words, we could imagine a scenario in which energy-saving plasticity of induction has evolved in response to variable herbivory pressure (i.e. optimal defense hypothesis Zangerl and Rutledge (1996)) (Agrawal et al., 2002;Humphrey et al., 2018;Wagner & Mitchell-Olds, 2018), and it has been retained during range expansion toward higher elevations. Therefore, plasticity in defense-related traits is a reflection of both biotic and abiotic environmental conditions that affect the expression of defenses. Conversely, the lack of plasticity in the majority of defense related traits in our study could be because the benefits of plasticity could not outweigh the costs affiliated with high herbivore pressure earlier in the season, or other potential costs of defense plasticity. For example, indolic GLS did not show plasticity, in contrast to non-indolic GLS, in Cardamine cordifolia plants growing in shaded-common gardens, that are characterized by low herbivory (Humphrey et al., 2018). In contrast to our results, Humphrey et al. (2018) also found plasticity in larval weight gain of a specialist herbivore (Scaptomyza nigrita).
Detailed analysis of the effect sizes (SESs) between growth and defense related traits in C. pratensis (Figure 2a) indicates that the plasticity displayed by high-elevation ecotypes is higher for AG biomass (very large SES) (Cohen, 1988) compared to indolic GLS production (large SES). In P. major ( Figure 2b) the magnitude of plastic responses in all growth-related traits were also very large, compared to the non-significant plastic responses for all defense-related traits (except for some the individual compounds, Supporting information Figure S2). Nevertheless, the lack of plastic response to elevation in defense-related traits does not completely discard the potential for plastic responses in chemical defenses. The environmental effects of growing elevation could influence plant chemistry at any time throughout the growing season; since chemistry was measured only at the end of the field season, plasticity in expression of such traits could have disappeared by the end of the season.
Moreover, the phytohormone activation machinery underlying expression of chemical defenses in response to herbivory is a very fast process (Mousavi, Chauvin, Pascaud, Kellenberger, & Farmer, 2013). In contrast, the detection of the potential plastic responses in plant defense to abiotic stimuli might be masked by the time-dependency of the growing season (Anderson, Lee, & Mitchell-Olds, 2011). Additionally, two studies on C. cordifolia and P. lanceolata showed phenological variation in plant tissue GLS and IGs content, respectively (Darrow & Deane Bowers, 1997;Rodman & Louda, 1984). Therefore, ontogeny should also be addressed when measuring plasticity, since plants have been shown to express different levels of plasticity in defense traits as they grow.

| CON CLUS IONS
Few studies have assessed phenotypic variation of plant growth versus defense traits in response to contrasting environments.
Here, we documented that plant growth traits displayed strong ecotypic differentiation accompanied by plasticity, but, in contrast, we found little support of phenotypically plastic defense and resistance traits in response to different growing habitat across steep elevation gradients. Future research on similar systems would require coupling the observed effects on plant phenotypes with genetically-explicit fitness measurements and selection gradient analyses in order to disentangle the fitness benefits of phenotypic plasticity versus fixed ecotypic differentiation at the population level.

ACK N OWLED G M ENTS
We thank Adrienne Godschalx and Mark Szenteczk for their valuable comments and proofread of the manuscript. This work was financed by a Swiss National Science Foundation grant 159869 to SR.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

AUTH O R CO NTR I B UTI O N S
MB, LF and VC performed the experiment, collected and analyzed the data. SR conceived the study, analyzed the data. GG assisted with chemical analysis. MB, LF and SR wrote the manuscript.

DATA ACCE SS I B I LIT Y
The data associated with this publication are deposited at Dryad data repository.