A test of the evolution of increased competitive ability in two invaded regions

Non-native plant species invasions can have significant ecological and economic impacts. Finding patterns that predict and explain the success of non-native species has thus been an important focus in invasion ecology. The evolution of increased competitive ability (EICA) hypothesis has been a frequently used framework to understand invasion success. Evolution of increased competitive ability predicts that (1) non-native populations will escape from coevolved specialist herbivores that were present within the native range and this release from specialist herbivores should result in relaxed selection pressure on specialist-related defense traits, (2) there will be a trade-off between allocation of resources for resistance against specialist herbivores and allocation to traits related to competitive ability, and (3) this shift will allow more allocation to competitive ability traits. We tested the predictions of EICA in the model plant Mimulus guttatus, a native of western North America (WNA). We compared how well the predictions of EICA fit patterns in two non-native regions, the United Kingdom (UK), an older more successful invasion, and eastern North America (ENA), a younger less successful invasion. We completed extensive herbivore surveys and grew plants derived from multiple populations in each region in a common greenhouse environment to test adherence to the predictions of EICA. We found evidence of specialist herbivore escape in the UK, but not the ENA plants. Compared to native plants the UK plants had lower levels of resistance traits, were taller, and produced larger and more flowers, while the ENA plants had mostly equivalent traits to the WNA plants. Plants from the UK conformed to the predictions of EICA more closely than those from ENA. The UK invasion is an older, more successful invasion, suggesting that support for EICA predictions may be highest in more successful invasions.


3.
Herbivore communities differed significantly between WNA, UK, and ENA populations with evidence 23 of specialist herbivore escape in the UK, but not necessarily the ENA plants. Compared  The UK invasion is an older, more successful invasion, suggesting that support for EICA may be highest in 31 more successful invasions. The lack of comprehensive conformity of either non-native region to the 32

Introduction 42
The translocation of non-native species into areas outside of their native range provides unique 43 opportunities for the study of evolution (Cox, 2004), including how selection pressures from herbivores 44 can shape plant defense evolution (Callaway & Maron, 2006). Comparisons between divergent biotic 45 and abiotic factors in the native and non-native habitats can aid understanding of how these variables 46 shape evolution in non-native plants (Whitney & Gabler, 2008). The testing of theories blending 47 ecological and evolutionary explanations can provide important insight into how non-native plants are 48 successful and how defense traits evolve; these tests often involve comparison of genotypes from the 49 native and non-native ranges (Orians & Ward, 2010). Better understanding of the mechanisms of non-50 native plant success may allow improved control and/or more accurate predictions of the impacts that 51 non-native species can have on native ecosystems. 52 Many hypotheses have been proposed to find a common reason for why plants become invasive 53 (Catford et al., 2009), dating back to Darwin's naturalization hypothesis (Diez et al., 2008). Hypotheses 54 have speculated on the potential for non-native plants to more efficiently use resources than native 55 plants (Coley et al., 1985), or have proposed that non-native plants are able to exploit an empty or less-56 crowded niche in the invaded habitat (Mack et al., 2000;Hierro et al., 2005,). Many of these hypothesis 57 also incorporate the idea that a competitive advantage is gained through an enemy release in a non-58 native habitat from co-evolved, specialist herbivores present in the native species range (Keane & 59 seed production and biomass were correlated with a decline in defenses against two specialist 67 herbivores that, at the time, were not present in eastern North America. By generating testable 68 predictions of the role that ecology plays in shaping the evolution of non-native plants, EICA hypothesis 69 has become one of the most widespread frameworks to explore the ability of non-native plants to 70 succeed (Bossdorf et al., 2005;Orians & Ward, 2010). 71 Two specific, testable predictions of EICA to explain the success of non-native plant populations include: 72 Firstly, non-native populations will escape from coevolved specialist herbivores that were present within 73 surveys in both ranges, we conducted a literature review on reports of herbivores and noted their 160 geographic range. We also looked at feeding records in the literature for herbivores of plant species 161 closely related to M. guttatus (i.e., Scrophulariaceae sensu lato) to see if there were any specialist 162 herbivores that may be able to shift hosts onto M. guttatus in the non-native regions. 163

Resistance traits 164
Following a release from specialist herbivores, EICA predicts the evolution of lower levels of some 165 herbivore resistance traits. To test this part of EICA we used plants derived from native and non-native 166 populations to assess patterns of genetic-based trait variation. We assessed specific leaf area (SLA), leaf 167 water content, leaf dry matter content (LDMC), trichome density, and foliar phytochemistry. After 168 growing the plants in a common greenhouse environment for one month, we harvested one leaf from 169 the fourth true-leaf pair. We weighed the leaf to get wet mass and then scanned the leaf (Epson 170 Perfection V19) to find leaf area using Image J (Rueden et al., 2017). Freeze dried leaves (see below) 171 were used to estimate dry weight and calculate specific leaf area (SLA), leaf water content, and leaf dry 172 matter content (LDMC). Leaf water content and LDMC are associated with performance of some 173 generalist herbivores consuming native M. guttatus (Rotter unpublished data). Trichome density was 174 measured by counting all the trichomes at the basal section of the adaxial side of each leaf within the 175 field of view of a dissecting microscope at 25x magnification. This density was converted to trichome 176 density per cm 2 (Holeski, 2007). 177 For phytochemical analysis, we quantified phenylpropanoid glycosides (PPGs), the predominant foliar 178 bioactive secondary compound in the species (Holeski et al., 2013; Keefover-Ring et al., 2014). The leaf 179 opposite the leaf in feeding trials (detailed below) was cut at the base of the petiole with scissors and 180 flash frozen in liquid nitrogen before being transferred to a -20 degree C freezer. Tissue was then 181 lyophilized using a pre-chilled FreeZone triad freeze dry system (Labconco; Kansas City, USA). We finely 182 ground the freeze-dried tissue in a small capacity ball mill (dental amalgamator with steel bearings). 183 Samples were stored and extracted as described in Holeski et al. 2013. We quantified the PPG content 184 of each sample via high-performance liquid chromatography [HPLC; Agilent 1260 HPLC with a diode 185 array detector and Poroshell 120 EC-C18 analytical column (4.6 · 250 mm, 2.7 μm particle size); Agilent 186 Technologies, Santa Clara, CA] maintained at 30°C, as described in Kooyers et al. (2017). The seven PPGs 187 analyzed in this study represent the PPGs present in detectable levels in the populations used in this 188 study. 189

Herbivore feeding trials 190
Herbivore response to plant resistance traits are often diffuse and vary depending on many different 191 factors. In addition to quantifying resistance traits, we thus also measured resistance though two 192 performance trials. For these trials, we used a subset of plant populations that represent the range of 193 native and non-native populations (Table S1). We conducted no-choice performance trials with neonate 194 Lepidopteran larvae of the specialist herbivore Junonia coenia and the generalist herbivore Trichoplusia 195 ni (Rotter & Holeski, 2017;. One leaf from a leaf pair was placed in an envelope and 196 treated as described above for analysis of PPGs. We assessed trichome density on the second leaf of the 197 leaf pair, as described in Holeski (2007). The leaf scored for trichomes was then placed into a water pic 198 and placed in a plastic container. In each container, we placed a single recently emerged first instar 199 caterpillar. Leaves were immediately replaced with leaves from the same plant (with the opposite leaf 200 harvested for phytochemical analysis) if/when the caterpillar consumed the entire leaf or if the leaf 201 wilted. After larvae had fed for 10 days, we ended each trial, froze the caterpillars, and then dried and 202 weighed them to determine caterpillar final dry mass. Larval initial (wet) weights were all within 0.001μg 203 of each other for a particular species, so we assumed that initial dry mass was identical across larvae Additionally, a more rapid growth rate allows greater survival when faced with pressure from predators 207 and parasitoids (Feeny, 1976;Benrey & Denno, 1997). 208

Plant fitness traits 209
Finally, EICA predicts an increase in fitness/competitive ability traits with a release from specialist 210 herbivores and the decline of herbivore resistance traits. To test plant fitness traits we used the plants 211 from the resistance traits measurements. We grew all plants for a total of six months prior to harvest 212 with the exception of several populations of annual plants that were harvested after they stopped 213 producing flowers. We assessed traits related to reproductive development, reproductive fitness, and 214 vegetative fitness. We assessed reproductive development by counting the number of days until a plant 215 first flowered. We also measured the corolla width (bigger flowers have been associated with pollinator 216 preference; Martin, 2004) of the first flower on the day after it was fully emerged. We collected pollen 217 from the first two flowers. Pollen was then stained, counted, and evaluated for viability with a 218 hemocytometer following the procedure in Kearns and Inouye (1993). We self-pollinated each plant 219 with the next three flowers, saturating each stigma with as much pollen as possible. Seeds were 220 collected from these flowers and total seed was counted. Finally, the total number of flowers produced 221 by a plant were counted at the time of plant harvest. Plants that did not flower by the end of the six-222 month trial (n=32 plants) were excluded from these analyses. Vegetative traits quantified included 223 specific leaf area and leaf water content, which were measured as described above during our 224 quantification of resistance traits. At harvest, we measured the total height (length) of the plant, from 225 the root crown to the end of the largest shoot. We then dried all plants in a drying oven and measured 226 aboveground biomass, belowground biomass, and total (aboveground + belowground) biomass. 227

Statistical analysis 228
To compare herbivore communities, we used non-parametric multidimensional scaling (NMDS) to look 229 at herbivore family and functional feeding guild differences between native and non-native populations 230 of M. guttatus. The NMDS was performed using PC ORD v. 6 (McCune & Mefford, 2016). We used 231 Jaccard distance as the similarity measure, and the program was run on "Autopilot" mode under the 232 "slow and thorough" method, with principal axes rotation. Significance of the ordination was based on a 233 Monte Carlo test with 250 iterations. In addition to the NMDS we looked for differences between the 234 non-native populations and the native sub-regions in the above herbivore communities using multi-235 response permutation procedures (MRPP). We also used ANOVA (transformed with either a log or root 236 transformation as assessed by Q-Q plots; we used Kruskal-Wallis tests if we could not obtain a normal 237 distribution) to look at the differences of field measured herbivory and herbivore richness between 238 regions and sub-regions. Trait values, fitness and resistance traits, were analyzed using a nested ANOVA 239 (plant family nested within population and population as a factor) to look for differences between the 240 two non-native ranges and the native geographical clades. We further used Tukey post-hoc tests for 241 pairwise comparisons. Lastly, we wanted to test for the predicted tradeoffs between herbivore 242 resistance traits and competitive ability traits in the non-native populations. To narrow down important 243 traits as well as suits of traits we used PCA to find the two most important contributors to variation 244 (components) for resistance traits and then for fitness/competitive ability traits for the two introduced 245 regions. We took these components and used a linear regression (with population means of the 246 components to account for population structure) to look for the relationship between the PCA 247 components for resistance traits and the fitness/competitive ability PCA components. In addition to 248 using the PCA components, we used correlation matrices to look at all pairwise trait tradeoffs (using 249 population means) for each region.  Figure 2B). This was also true when comparing the non-native 272 regions to the native sub-regions (F5,34=2.33, p = 0.063). 273 Herbivore communities, at the family level, differed between the native subregions and the non-native 274 populations (MRPP A = 0.085, p < 0.001, Table 1) with the two non-native regions (ENA and the UK) 275 having similar herbivore families to one another (A = -0.017, p = 0.86; Figure 3). The similarity in 276 herbivore communities in ENA and the UK was generally driven by families dominated by generalist 277 herbivores such as terrestrial gastropods and mammals. Differences between the UK populations and 278 the native Cordilleran populations (which includes Alaska and is thus from which the UK populations are 279 thought to be derived; A = 0.092, p <0.001), were driven in part by the lack of leaf mining Agromyzidae 280 in the UK. We also found substantial geographic variation in herbivore community composition within 281 the native subregions. Native subregions were generally separated because of specialist insects that 282 dominated in particular subregions. For instance, leaf mining Agromyzidae flies were common in the 283 Cordilleran subregion as a dominant herbivore while the more southern subregions were dominated by 284 specialist caterpillar species. Herbivore functional feeding guild differences across regions were similar 285 to these herbivore community patterns (Table 1), and were driven by generalist chewers being more 286 common in the non-native regions. 287 Herbivore resistance traits 288 In comparing traits between non-native and native regions, we focus on trait comparisons between 289 populations from the non-native ENA and the native WNA regions and between the non-native UK 290 populations and their likely ancestral WNA Cordilleran subregion. 291 We found mixed evidence of an overall relaxation of selection on resistance traits predicted by EICA in 292 the non-native M. guttatus populations. Physical resistance traits varied between native and non-native 293 regions. Trichome density was significantly different between all regions (F2, 518 = 86.63, p < 0.001, Figure  294 4). In support of EICA, native WNA populations had, on average, three and a half times higher trichome 295 density than the non-native ENA plants, which was similar when using the native sub-regions (F5, 516 = 296 56.62, p < 0.001, Figure 4). In contrast to the predictions of EICA, the UK population had one and half 297 times higher trichome density than the native Cordilleran sub-region (Tukey post hoc: p = 0.002). 298 Specific leaf area was not significantly different between any of the native and non-native regions 299 (F2,518= 1.82, p = 0.121, Figure 4). Leaf water content in the UK populations was slightly higher than the 300 Cordilleran populations and the non-native ENA populations was slightly higher in than the native WNA 301 populations (F2,517= 4.53, p = 0.011, Figure 4), suggesting a relaxation in herbivore defense. Leaf dry 302 matter content did not differ significantly across any of the native and non-native regions (F2.517 = 0.93, p 303 = 0.392, Figure 4). We found no evidence that specialist herbivores performed better on plants from non-native regions 311 than from native, as predicted by EICA. We found no difference in performance of a generalist or a 312 specialist herbivore feeding on tissue from native vs. non-native regions. The generalist caterpillar 313 second component (18.8%) was associated with root mass, pollen viability and seed production. All the 378 components had non-significant relationships to one another (Table 2). 379 The native region (WNA) also showed evidence of resistance vs. fitness/competitive ability tradeoffs.  height, and number of flowers produces, the second fitness/ competitive ability component (23.4%) was 393 associated with root mass, seed count, and percent pollen viability. We found no evidence of tradeoffs 394 between these components (Figure 8, Table 2). 395

Discussion 396
By comparing two different plant invasions of differing ages to their native counterparts we found some, 397 but not comprehensive, support for EICA. Support was strongest in the non-native UK, the older of the 398 two invasions. Both the non-native UK and the ENA plants had different herbivore communities than the 399 native WNA plants. However there was adherence to the EICA prediction of a reduction in herbivore 400 damage as well as clear evidence of specialist herbivore escape in only the UK range. We found 401 relatively minor support for the prediction that there would be a decline of herbivore resistance traits in 402 the non-native plants, with some changes in trait values in the non-native vs. native regions, but no 403 differences in herbivore performance in no-choice trials. The UK plants were larger, taller, and produced 404 more seeds and flowers than their native counterparts, in accordance with EICA predictions, while the 405 non-native ENA plants were generally smaller and had poorer pollen production than the native WNA 406 plants. Lastly the UK plants exhibited some tradeoffs between resistance traits and fitness/ competitive 407 ability while the ENA plants did not, confirming to the predictions that release from specialist herbivores 408 can result in allocational tradeoffs that allow for increases in fitness/ competitive ability. 409 Enemy release and resistance traits in the non-native populations 410 We found some evidence of escape from coevolved specialist herbivores in both of the non-native 411 regions. However, this did not translate to the same pattern of relaxed defenses in the two non-native 412 regions. Each non-native region had several resistance traits present at lower levels than in their native 413 ancestral regions. The non-native ENA populations had lower trichome density and higher leaf water 414 content than did the native WNA populations, while the non-native UK populations had higher leaf 415 water content and lower levels of total PPGs than the native Cordilleran region. However, levels of some 416 defenses were also higher in the non-native regions than the native, and we found no difference in 417 performance of generalist and specialist herbivores feeding on native vs. non-native plants. native range. If non-native plants are less palatable than native, but not toxic, this could explain the lack 436 of differences between caterpillar performance in our no-choice trials. The presence of overlap in 437 resistance traits, as some traits likely deter both generalists and specialists, could also result in the 438 overall maintenance of traits that defend against generalist herbivores. The end result would be the 439 maintenance of certain resistance traits that may deter specialist herbivore despite the absence of 440 specialists in the new habitat. For instance, the PPG conandroside has a negative impact on the 441 performance of the generalist herbivores Grammia incorrupta and Spodoptera exigua as well as a 442 negative impact on the specialist herbivore Junonia coenia . 443 Changes to competitive ability in non-native plants 444 The EICA prediction that trait values related to fitness and/or competitive ability will be higher in non-445 native regions was partially supported by our data. Like resistance traits, we did not see similar patterns 446 in fitness and/or competitive ability traits between the two non-native regions. Fitness/competitive 447 ability traits tended not to conform to the predictions of EICA for the non-native ENA region; these trait 448 values were generally very similar to those for the native WNA region. In contrast, fitness/competitive 449 ability trait values were greater in the non-native UK than the native Cordilleran region for many traits, that are limited by defending themselves may gain a significant advantage when these resources are in 476 abundance (Blumenthal, 2006). 477 This lack of clear tradeoffs, as predicted by EICA, has also been found in other reviews focused on EICA 478 (Bossdorf et al., 2005;Felker-Quinn et al., 2013). Both of these studies found overall that non-native 479 plant populations changed in their herbivore resistance traits as well as their fitness/ competitive ability 480 traits but these changes did not reflect EICA predictions of a tradeoff (a direct relationship between an 481 increase in fitness/ competitive ability and a decrease in herbivore resistance traits). These studies 482 proposed that more specific looks at relevant traits was needed in testing EICA predictions. Although it is 483 possible that we missed some of the key traits that are involved in tradeoffs, our study was relatively 484 comprehensive in our trait selection particularly for traits important to the ecology of M. guttatus. 485 Can EICA predict the success of M. guttatus invasions? 486 Finally, our prediction that the more successful invasion (the UK) would display more evidence of 487 adherence to EICA than the less successful invasion (ENA), was supported. The non-native UK 488 populations showed greater adherence to multiple predictions of EICA than the non-native ENA region. 489 Within the EICA framework, species that have become extremely successful invaders such as Triadica that were ranked as more invasive had lower rates of herbivory than those non-natives that were not 498 considered as invasive (Cappuccino & Carpenter, 2005). This supports the idea that the strongest 499 evidence for EICA may be found in more successful invasions. 500 There are many different frameworks for understanding the success of non-native organisms (Catford et 501 al., 2009) and it is likely that there is not a single one that can consistently and fully explain why a non-502 native species becomes successful across systems (Gurevitch et al., 2011;Lau & Schultheis, 2015). This is 503 the case with our results; although we found some evidence to support EICA, particularly in the non-504 native UK region, there were several patterns that were not necessarily compatible with EICA (e.g., 505 caterpillar performance was not different between the native and non-native plants and the sometimes 506 positive relationship between resistance traits and fitness/ competitive ability in the UK plants    Tables. Table 1. MRPP results for differences between the non-native regions and the native sub-regions herbivore community at the family level (on bottom and in grey) and functional feeding group (on top in white). The full model was significant for herbivore communities at the family level (A = 0.085, p < 0.001) and for functional feeding groups (A = 0.131, p < 0.001). Bolded results are significantly different pair wise comparisons.

Coastal
Cordilleran  Table S1. Locations of all populations used in this study. With plant life history, region and subregion (Stace 2010, Twyford andFreidman 2015). Populations with an * were additionally used for caterpillar feeding trials.