Invasion in a Community Context: Genetic Variation in the Strength of Direct and Indirect Effects


 Predicting invasion success requires understanding how abiotic and biotic factors in the local environment interact with the particular traits of potential invaders. Relative to studies of direct antagonistic effects, fewer studies have examined how positive species interactions, such as facilitation or mutualism, or indirect interactions in multispecies communities, can affect invasion success. We examined the effects of drought and mutualisms with rhizobia bacteria on the performance of a widely invasive legume, Medicago polymorpha. In a greenhouse experiment, we found that watering regime affected plant performance, but non-linear patterns in response to decreasing water were dependent on the specific plant genotype. In a second experiment, we found that the effects of drought on plant performance were dependent on the presence of rhizobia, particularly for genotypes collected from the invasive range. This suggests that indirect ecological effects may have important consequences for invasion success. We contextualize the strength of these direct and indirect effects by comparing this study to effect sizes in other studies of the same species. In this species, predicting invasion into a natural community context will require understanding multiple direct and indirect effects in the local environment, as well as their effects on the specific genetic composition of the invading population.


Introduction 34
Invasive species are an ongoing and increasing threat to ecosystems and economies Fewer studies have examined how positive species interactions, such as facilitation or 47 -7 -traits adaptive in the invasive range, which encompasses a broad array of different 134 environmental conditions. This suggests that invasion success depends not only on the 135 combination of direct and indirect effects at a potential invasion site, but also on the 136 particular genetic composition of the introduced population. 137 In this paper, we seek a more comprehensive approach to understanding plant 138 invasions that accounts for the community context in which invasions occur and intraspecific 139 variation in responses to potential barriers to invasion. We ask whether such a framework

Direct effects of drought 152
We used a greenhouse study to test the effects of drought on different genotypes of M. 153 polymorpha. We used 5 haphazardly selected M. polymorpha ascensions from the U.S. 154 National Plant Germplasm System. Because this species reproduces almost entirely through 155 selfing, resulting in homozygous clones, we refer to these ascensions as genotypes. These 156 genotypes were collected from different parts of the species range, with three from the native 157 range (292427 from Israel, 469263 from Egypt, and 543039 from Spain) and two from the -8 -invasive range (478530 from Peru and w65527 from Australia). Genotypes were categorized 159 as invasive or native using the CABI Invasive Species Compendium 160 (https://www.cabi.org/ISC). All genotypes were grown in a common garden greenhouse 161 environment for one generation prior to the experiment to ensure that any observed 162 differences among genotypes were due to genetic differences, rather than maternal or 163 epigenetic effects. Seeds collected from this common garden generation were used in the 164 experiment. 165 M. polymorpha seeds were scarified and planted in 164 mL conetainers (Stuewe and 166 Sons, Tangent, OR, USA) containing low-nitrogen LP5 Grower Mix soil (SunGro, Agawam, 167 MA, USA), composed of 90% topsoil, 9% sand (DecoRock All Purpose), and 1% perlite. 168 Plants were grown in the California State University, Northridge greenhouse with conetainers 169 haphazardly distributed among racks, with racks rotated weekly to reduce any microclimate 170 effects. We planted three seeds of each of the five genotypes in replicate conetainers and after 171 two weeks, thinned the seedlings to one per pot. We transplanted removed seedlings to new 172 conetainers when possible. 173 Plants were watered daily for 28 days and watering treatments were imposed 174 thereafter. We watered plants in all treatments approximately every two days, with some 175 adjustments based on the local temperature. On the few occasions where the outside 176 temperature exceeded 29°C, plants were watered daily. All plants were watered with the 177 same frequency, but with different volumes: Level 1 = 100 mL, Level 2 = 50 mL, Level 3 = 178 25 mL, Level 4 = 12.5 mL per conetainer. These volumes were based on preliminary data 179 showing that plants experienced signs of water stress after two days in the smallest volume 180 treatment, but did not die. The ecological relevance of these watering regimes differs 181 depending on the potential invasion site and the origin of each genotype.

-9 -
We started the experiment with 20 replicates per treatment combination (400 183 conetainers = 5 genotypes x 4 watering treatments x 20 replicates), but differential 184 germination success among genotypes, seedling deaths, and transplantation of excess 185 seedlings led to 14-26 replicates per treatment combination (N=411). In June 2016, 29 days 186 after the watering levels were imposed, we harvested the plants and separated above-and 187 belowground biomass of each plant and dried these at 50°C for three days, and then weighed 188 dry biomass using an analytical balance. We used total plant biomass (aboveground + 189 belowground biomass) as a measure of plant size and performance and the ratio of . We crushed the surface-sterilized nodules in sterile deionized water and plated the 208 solution on tryptone yeast (TY) agar plates. We incubated plates at 25°C and found only 209 colonies with the same morphology as E. medicae (Somasegaran & Hoben, 1985). We used 210 multiple colonies to inoculate a heat-sterilized TY nutrient broth. We simultaneously started a 211 sterilized broth of TY media without the rhizobia inoculant. Both media were cultured at 212 25°C for 24 hours before use. We added 1 mL of media (either rhizobia-inoculated or the 213 sterile control) around the base of 2-week-old M. polymorpha seedlings (terHorst et al., 214 2018). We assumed rhizobia to be absent or in low abundance in the commercial soil prior to 215 inoculation and did not control for colonization of rhizobia from the environment during the 216 experiment. We did not quantify rhizobia abundance in the soil, but assume that the addition 217 of rhizobia greatly increased their abundance. We refer to these as Enhanced or Background 218 rhizobia treatments. 219 We randomly distributed replicates among two watering treatments that we imposed 220 six days after rhizobia inoculation. All plants received 100 mL of water each time they were 221 watered. Plants in the high watering treatment were watered approximately four times more 222 often than those in the low watering treatment. Plants in the high water treatment were 223 watered approximately every two days, but occasionally daily when the temperature was 224 particularly high; watering was delayed a day following rare precipitation events. In the low 225 watering treatment, we only watered plants following signs of water stress, such as leaf 226 folding or mild wilting. In April 2015, 28 days after the watering treatments were imposed, 227 we harvested the plants. Plant biomass was divided into above-and belowground components 228 and weighed as dry mass, as above. 229

Statistical analysis 230
For the greenhouse experiment, we fit generalized linear mixed models using R to 231 examine the fixed effects of watering level, plant range (native or invasive), and their -11 -interaction on total biomass and above:belowground biomass ratio. We included genotypes 233 nested within range, and the interaction between watering level and genotype as random 234 effects in the models. We found the best-fit model and error distribution for each dependent 235 variable (gamma distribution in each case) by comparing Akaike Information Criteria (AIC) 236 values. Interaction terms and random effects were dropped from the model if their inclusion 237 did not result in a lower AIC value. Genotype was retained in all models to avoid 238 pseudoreplication in the experimental design. We performed likelihood ratio tests to 239 determine the significance of random factors. We conducted similar analyses for the 240 botanical garden experiment, but included rhizobia treatment as a fixed effect and the 241 genotype*rhizobia interaction as a random effect. 242

Comparing direct and indirect effect sizes to previous studies 243
Papers were selected for inclusion in this meta-analysis based on a Google Scholar, 244 Web of Science, and JSTOR search for keywords "Medicago polymorpha", "M. were calculated using Cohen's d (Ziegel et al., 1995). 249 Direct effects represented the standardized difference between a treatment and 250 control. For example, the direct effect of rhizobia in the second experiment was calculated as 251 the difference in a variable (X) between the Enhanced rhizobia treatment (X2) and the 252 Background rhizobia treatment (X1), standardized by the pooled standard deviation and 253 multiplied by a corrector for sample size (J): 254 We calculated the additive effect of the control treatment (X1) and both other 256 treatments (X2 and X3) as in terHorst et al. (2015). 257 We also calculated standardized indirect effect sizes using an equivalent of Cohen's d. We standardized this indirect, or non-additive, effect by the pooled standard 264 deviation, resulting in this standardized indirect effect comparable to direct effects: 265 Similarly, we calculated direct and indirect effect sizes from previous studies on M. 267 polymorpha in response to various ecological interactions. We took the absolute value of 268 each effect size and grouped these as direct effects of drought, disturbance, herbivory, 269 nutrient level, competition, and mutualism, as well as indirect effects (Table S2). We further the direction and magnitude of their responses (Fig. 1A). The genotype from Australia 281 (w65527), on average, grew to the largest size, but was the most affected by decreasing water 282 to the lowest level. In contrast, the genotype from Spain (543039) had the smallest plants, but genotypes had variable responses (Fig 1B). 294 In the botanical garden experiment, survival was similar to the greenhouse experiment 295 (95%). The effect of the watering treatment was dependent on the rhizobia treatment and the 296 range from which genotypes were collected (F1,396 = 4.68, P = 0.031). Invasive genotypes 297 grown in the high water and enhanced rhizobia treatment produced the most biomass (Fig.  298 2A). Although native genotypes grew larger with more water, there was no added benefit of 299 enhanced rhizobia ( Fig. 2A). However, native genotypes grew larger with enhanced rhizobia 300 in the low water treatment (Fig. 2A). Conversely, invasive genotypes grown in the low water 301 treatment were actually smaller in the enhanced rhizobia treatment, but benefitted from 302 rhizobia in the high water treatment. Beyond the effects of collection range, different 303 genotypes responded differently to watering treatments (C 2 = 11.6, df = 1, P<0.001). In the 304 -14 -background rhizobia treatment, most genotypes responded negatively to the low water 305 treatment, albeit to different extents, but nine genotypes had positive average responses to the 306 low water treatment (Fig. 3A). However, in the enhanced rhizobia treatment, the rank-order 307 of effects of watering on M. polymorpha biomass was quite different (Fig. 3B), reflecting the 308 fact that different genotypes responded to the watering and rhizobia treatments differently. compared to previous studies on this species (Fig. 4). In decreasing order of effect size, 326 mutualism, nutrients, competition, drought, disturbance, and herbivory had greater effects on 327 M. polymorpha abundance than those observed in this study. However, consistent with 328 previous studies, effect strengths were dependent on whether genotypes originated from the -15 -native or invasive range. Native range genotypes were affected more by interactions with 330 mutualists, whereas invasive range genotypes were affected more by nutrients, drought, and 331 herbivory. In past studies, indirect effects were smaller in magnitude than direct effects, 332 inconsistent with the results observed in this study.

Direct effects of water 352
We found that the amount of water provided to plants in both experiments affected 353 plant performance (total biomass), as well as growth allocation to above-and belowground -16 -biomass. In the greenhouse experiment, the effects of watering level were generally non-355 linear (Fig. 1), suggesting that certain genotypes of M. polymorpha may perform best at 356 intermediate water levels. Previous work found that piñon pine (Pinus edulis) genotypes 357 originating in dry habitats performed better in drier soils (Mitton & Duran, 2004). Similarly, 358 the genetic differences we observed could be the result of different evolutionary histories. We 359 minimized the effects of recent ecological environments using a common garden generation, 360 suggesting that the observed differences are not the result of phenotypic plasticity or maternal 361 effects, but rather inherited genetic differences. It is possible that the effects of water on our 362 different M. polymorpha genotypes may depend on long-term historic soil moisture in the 363 environments in which these genotypes evolved. Unfortunately, we do not have sufficiently 364 precise information about the original collection sites to obtain historic records to test this 365

hypothesis. 366
Previous studies suggest that plants grown under water stress will allocate more 367 energy toward belowground growth to increase root biomass and acquire more water Regardless of the mechanism, this result suggests that allocation to above-or belowground 377 growth is somewhat plastic in this species, although the variation among genotypes suggests 378 that it is also partially genetically determined. is certainly a very conservative subset of that found around the globe (Bromfield et al., 2010). 411

Selection for traits during invasion 412
We compared genotypes from the native and invasive range because we assumed that be in response to other factors besides competition, such as water, searching for rhizobia 428 mutualists, or acquiring nitrogen in the absence of mutualist partners.

-19 -
The benefits from rhizobia in our experiment varied widely among genotypes (Fig. 3). where precipitation and water availability are highly seasonal. We were surprised to see that 457 mutualisms ranked so highly in importance, but the generality of this result should be 458 that have yet to be quantified. 498 Overall, these findings suggest that the relative magnitudes of direct and indirect 499 effects are context dependent. Though general trends emerge, including the importance of 500 mutualism, nutrient level, and competition, the order and magnitude of these effects are 501 dependent upon the genetic composition of the invading M. polymorpha population. Our 502 review of previous studies suggests that indirect are unimportant relative to direct effects in 503 predicting invasion success of this species. However, the relatively large role that indirect 504   543039  w64614  577339  w619008  w619003  368953  w619534  I4  w65510  566873  w65524  566883  W1  577402  459130  469263  w65533  566880  535520  CVI30  458578  CVI50  493293  478440  577437  577391  I9  517214  478439  GP5  478466  566877  577392  478530  227025  w65527