Experimental and theoretical support for costs of plasticity and phenotype in a nematode cannibalistic trait

Abstract Developmental plasticity is the ability of a genotype to express multiple phenotypes under different environmental conditions and has been shown to facilitate the evolution of novel traits. However, while the associated cost of plasticity, i.e., the loss in fitness due to the ability to express plasticity in response to environmental change, and the cost of phenotype, i.e., the loss of fitness due to expressing a fixed phenotype across environments, have been theoretically predicted, empirically such costs remain poorly documented and little understood. Here, we use a plasticity model system, hermaphroditic nematode Pristionchus pacificus, to experimentally measure these costs in wild isolates under controlled laboratory conditions. P. pacificus can develop either a bacterial feeding or predatory mouth morph in response to different external stimuli, with natural variation of mouth-morph ratios between strains. We first demonstrated the cost of phenotype by analyzing fecundity and developmental speed in relation to mouth morphs across the P. pacificus phylogenetic tree. Then, we exposed P. pacificus strains to two distinct microbial diets that induce strain-specific mouth-form ratios. Our results indicate that the plastic strain does shoulder a cost of plasticity, i.e., the diet-induced predatory mouth morph is associated with reduced fecundity and slower developmental speed. In contrast, the non-plastic strain suffers from the cost of phenotype since its phenotype does not change to match the unfavorable bacterial diet but shows increased fitness and higher developmental speed on the favorable diet. Furthermore, using a stage-structured population model based on empirically derived life history parameters, we show how population structure can alleviate the cost of plasticity in P. pacificus. The results of the model illustrate the extent to which the costs associated with plasticity and its effect on competition depend on ecological factors. This study provides support for costs of plasticity and phenotype based on empirical and modeling approaches.

space in a given environment accessible to a genotype without the need for genetic change 91 (Pigliucci 2001;West-Eberhard 2003;Pfennig 2021). Indeed, many case studies in plants, 92 insects, vertebrates, and nematodes have indicated the importance of phenotypic plasticity for 4 promoting adaptations across environments and for the evolution of novelty (West-Eberhard 94 5 fitness by a highly-plastic genotype compared to a less plastic one (Callahan et al. 2008;Murren 125 et al. 2015). Understandably, the terms 'cost of phenotype' and 'costs of plasticity' are, by virtue 126 of their definitions, ripe for confusion (Auld et al. 2010;Pfennig 2021), and they can be only 127 applied within a comparative framework (Callahan et al. 2008). A comprehensive analysis of 128 these constraints would adequately improve our understanding of the role of plasticity in 129 adaptive evolution. However, empirical studies on the costs of phenotype and plasticity remain 130 scarce, especially in metazoans. This has been largely due to two reasons: Firstly, in the wild, 131 conditions often cannot be properly controlled, nor can the effects of various factors be 132 delineated. Secondly, laboratory experiments are time consuming and large organisms cannot be 133 easily investigated. To study the constraints of plasticity, we make use of natural isolates of 134 nematodes that, given their small size and rapid reproduction, can be examined under laboratory 135 conditions. 136 137 The nematode P. pacificus is an established model system for studying phenotypic plasticity 138 (Sommer & McGaughran 2013;Sommer et al. 2017). The developmentally plastic mouth of P. 139 pacificus can exhibit two distinct forms; the eurystomatous (predatory) morph with a wide stoma 140 and hooked-like teeth, or the stenostomatous (non-predatory) morph with a narrow stoma and a 141 single tooth (Fig.1a) (Bento et al. 2010). P. pacificus is a hermaphroditic nematode and the use 142 of isogenic cultures has facilitated the elucidation of genetic and epigenetic mechanisms 143 underlying this irreversible switch. Specifically, the sulfatase-encoding eud-1 gene was identified 144 as the key developmental switch that is regulated by various environmental factors and 145 epigenetic mechanisms, and directs a downstream gene regulatory network consisting of more 146 than 20 identified proteins including structural components of mouth formation (Ragsdale et al. 147 2013;Kieninger et al. 2016;Bui et al. 2018;Namdeo et al. 2018;Sieriebriennikov et al. 2018;148 Sieriebriennikov et al. 2020;Sun et al. 2022). Importantly, worms respond to surrounding 149 environmental cues to adopt their mouth form in a strain-specific manner and various 150 environmental stimuli, including temperature, culturing condition, crowding and diet have been 151 shown to influence mouth-morph ratios (Werner et al. 2017;Werner et al. 2018;Lenuzzi et al. 152 2021). Principally, three major features assist in studying P. pacificus mouth-form plasticity. 153 First, the vast collection of naturally occurring wild isolates with hundreds of P. pacificus strains 154 being sequenced, accordingly resulted in a highly resolved phylogeny of diverse populations 6 ( Fig.1b) (Rödelsperger et al. 2017). Interestingly, culturing these isolates on the laboratory 156 bacterium E. coli displays a range of mouth-morph ratios, while some express intermediate 157 mouth-form ratios (see below) . While the parallel formation of both 158 mouth forms under the same environmental condition represents an unusual type of plasticity 159 that is not seen in the majority of plastic traits, it has allowed unprecedented insight into 160 associated molecular mechanisms and the identification of a large gene regulatory network 161 Kieninger et al. 2016;Bui et al. 2018;Namdeo et al. 2018;162 Sieriebriennikov et al. 2018;Sieriebriennikov et al. 2020;Sun et al. 2022). Second, 163 morphological mouth-form plasticity is coupled to behavioral plasticity. Specifically, the 164 predatory form enables predation and cannibalism on other nematodes, while such animals can 165 still feed on bacteria. In contrast, the non-predatory form obligates worms to feed on bacteria 166 ( Fig.1c) (Wilecki et al. 2015). This extension of morphological plasticity to behavior is thought 167 to eliminate resource competitors via predation and the expansion of nutrition (Quach & 168 Chalasani 2020). Finally, P. pacificus is a thoroughly-studied soil nematode that is reliably found 169 in association with scarab beetles with recent studies describing the dynamics and succession of 170 nematodes on the beetle carcass after the insect´s death ( Fig.1d)  . 171 172 Switching between the predatory and non-predatory mouth forms is a specific example of 173 phenotypic plasticity, in which an irreversible decision that occurs during the development of P. 174 pacificus via a bi-stable developmental genetic switch (Sieriebriennikov et al. 2018). The bias of 175 the developmental switch determines the ratio of mouth morphs with substantial natural variation 176 between populations of P. pacificus . In this respect, mouth-form plasticity 177 in P. pacificus is more akin to the switch between lytic and lysogenic cycles in bacteriophage ߣ 178 (Ptashne 1986) than wing pattern polyphenism in butterflies that is seasonally controlled 179 (Nijhout 1994). Importantly, the relative simplicity of mouth-morph plasticity in P. pacificus, as 180 well as its isogenic husbandry of genetically diverse strains, makes it ideal to study different 181 facets of phenotypic plasticity. Here, we took advantage of these features to perform a systematic 182 analysis of mouth morphs and their associated costs and extended our empirical findings via 183 simulating ecologically relevant scenarios in spatially-homogeneous and spatially-structured 184 populations using empirically-derived life history parameters. 185

Methods 187
188 Please see the Supporting Information. 189

Results 190
Within and between strains comparisons reveal a cost of phenotype 191 To measure the cost of phenotype, we took advantage of the extensive collection of P. pacificus 192 natural isolates and selected seven strains with intermediate mouth-morph ratios from across the 193 P. pacificus phylogeny (Fig.1b) (Rödelsperger et al. 2017;Lightfoot et al. 2021). The ratios of 194 these strains were considered intermediate since they were neither predominantly predatory nor 195 predominantly non-predatory on the standard laboratory food source E. coli (Fig. 2b,SI 196 Appendix; Table. S4). It should be noted that "intermediate" in this context does not imply a 50-197 50 chance of expressing either of the mouth morphs, but merely indicates that both alternative 198 morphs can be easily found in the lab. These strains allow testing whether there is a cost of 199 phenotype within the same genetic background. Note that such strains are rare in nature and were 200 obtained only because of the large available collection of P. pacificus wild isolates. 201 To measure the cost of phenotype, we selected overall individual fecundity as primary fitness 202 parameter to capture reproductive capacity of P. pacificus hermaphrodites via selfing (Haldane 203 1937;Orr 2009). Testing for daily fecundity showed that the majority of progeny were laid 204 within a window of 62 hours after maturation (nearly 91%) in an overall window of appr. 158 205 hours of total egg-laying (SI Appendix; Fig. S1a& Table. S2). The number of eggs laid in this 206 window provide a reasonable estimate of the life-time fecundity. In these intra-genotype 207 comparisons, we found a tendency in non-predatory animals to have more progeny than 208 predatory worms. Specifically, the estimated differences in the mean value of fecundity between 209 non-predatory and predatory animals based on the data showed strong and/or partial support in 210 four out of seven comparisons (Fig. 2c, SI Appendix; Table S1; Table S6). These findings 211 suggest that the production of the predatory mouth morph can incur a fitness cost. This 212 observation is in concert with a previously report on the slower rate of development in 213 nematodes exhibiting the predatory morph in comparison to non-predatory worms (Serobyan et Afterwards, we measured fecundity and developmental speed in P. pacificus natural isolates that 216 show a biased mouth-morph ratio, i.e., strains that would produce an abundance of non-predatory 217 or predatory mouth morphs on the standard laboratory food source E. coli (Fig. 2d, SI Appendix; 218 Table S4). We selected two pairs of closely-related strains from the diverged clades B and C of 219 P. pacificus from La Réunion island (Rödelsperger et al. 2014). We found that in both pairs, the 220 non-predatory-biased strains produce more overall progeny than the predatory-biased strains 221 ( Fig. 2e, SI Appendix; comparisons of fecundity and developmental speed in four biased strains from two different P. 231 pacificus clades clearly illustrate the cost of producing the predatory phenotype. 232

Across-conditions testing indicates a cost of plasticity 233
Next, we wanted to determine if a cost of mouth-form plasticity exists in P. pacificus. Such a 234 cost of plasticity would be eminent when testing a non-plastic genotype relative to a plastic 235 genotype under different conditions (DeWitt et al. 1998;Pigliucci 2001;Callahan et al. 2008;236 Murren et al. 2015). Therefore, we performed a cross condition test by conducting experiments 237 on two distinct food sources, the standard E. coli condition used in the previous section, and a 238 Novosphingobium diet. The bacterial species Novosphingobium was found to be naturally 239 associated with P. pacificus and was proven to increase intraguild predation in the P. pacificus 240 reference strain PS312 (Akduman et al. 2018;Akduman et al. 2020). However, this association 241 was never studied in non-domesticated wild isolates of P. pacificus. Therefore, we grew two of 242 the biased strains with different mouth-morph ratios on Novosphingobium; the highly non-243 predatory-biased strain RSC017 and the highly predatory-biased strain RS5405. Indeed, RSC017 244 showed a substantial increase of the predatory morph of 84% on Novosphingobium, indicating strong plasticity. In contrast, the predatory-biased strain RS5405 remained highly predatory in 246 the new condition (Fig. 3a, SI Appendix; Table. S4). Thus, we established two distinct food 247 conditions that differentially affect plasticity levels of the two isolates. Henceforth, we refer to 248 RSC017 and RS5405 as plastic and non-plastic strains, respectively. 249 Theoretically, the cost of plasticity would be displayed in the strain that exhibits a change in 250 mouth-morph ratio upon altering food conditions. Accordingly, we would expect to detect the 251 highest effect on fitness in the plastic strain, and vice versa. Indeed, we found that the plastic 252 strain has lower fecundity and slower developmental speed on Novosphingobium when 253 compared to the non-plastic strain ( Fig. 3b-c, SI Appendix; Table. S1,3, Table S8). Thus, a strain 254 that plastically responds to a dietary change with the formation of the predatory mouth morph 255 exhibits reduced fitness under these novel conditions indicating a cost of plasticity. In contrast, 256 the non-plastic strain exhibits higher levels of fecundity and developmental speed on 257 Novosphingobium ( Fig. 3b-c, SI Appendix; Table. S1,3). Thus, a strain that is preferentially 258 predatory under both food conditions exhibits increased fitness when exposed to this new diet. 259 Taken together, these findings indicate a cost of mouth-morph plasticity in response to dietary 260 induction. These observations raise a fascinating question: which cost plays a larger role in 261 shaping the population dynamics and, consequently, the evolution of mouth-morph ratios? 262 The cost of phenotype maximizes the benefits of plasticity 263 To investigate how the cost of plasticity and the cost of phenotype would manifest in the wild, 264 we constructed a stage-classified model to simulate population dynamics of the plastic and the 265 non-plastic strains on both tested food sources (Fig. 4a). For modeling, we used the fecundity 266 measurements from the lab and scaled the developmental rates of the model based on the 267 laboratory estimates of developmental speed of P. pacificus (see Supplementary Methods). First, 268 we tested population dynamics of the selected strains in separation, i.e., without interactions or 269 competition. Surprisingly, the change from E. coli to Novosphingobium has only a minor effect 270 on the final population size of the plastic strain (Fig. 4b). The reduction in fecundity on 271 Novosphingobium relative to E. coli is presumably compensated by the increase in 272 developmental speed on Novosphingobium. To test the hypothesis that faster developmental 273 speed was indeed compensating for the cost of plasticity, (i.e., lower fecundity), we simulated this simulation confirmed this expectation (Fig S6). In contrast, in the non-plastic strain the 276 increase in fecundity and developmental speed on Novosphingobium results in a higher 277 frequency of all developmental stages compared to its dynamic on E. coli (Fig. 4c). Importantly, 278 the between strains cost of phenotype is clearly displayed when comparing the frequencies of the 279 two strains on E. coli (Fig. 4b-c). Thus, comparing both populations' trajectories without 280 involving interactions reveals that the cost of phenotype has a larger effect on the population 281 dynamics than the cost of plasticity. 282 The cost of plasticity manifests in a competition setup 283 In nature, P. pacificus does not occur in isolation, rather it competes with other nematodes over 284 resources. Additionally, given the coupling between morphological and behavioral plasticity, 285 predatory worms are able to predate while non-predatory worms are not. Testing the costs of 286 plasticity and phenotype in a competition setup might shed light on the evolution of the 287 predatory mouth morph. Therefore, we first tested if predation rate positively correlates with the 288 proportion of predatory individuals in wild isolates. To avoid the compounding effect of 289 relatedness on predation , we selected C. elegans as prey for P. 290 pacificus predators. Indeed, testing nine P. pacificus wild isolates with different mouth morph 291 bias, shows that morphological and behavioral plasticity positively correlate (SI Appendix; Fig.  292 S3). Second, we measured predation rates of the plastic and the non-plastic strains against one 293 another by testing predation rates over the two food sources E. coli and Novosphingobium (Fig.  294 4d). 295 Next, we used the experimentally obtained predation values for each food source to simulate the 296 effect of interactions between strains on their dynamics in a spatially-homogeneous population 297 (see Supplementary Methods, SI Appendix; Fig. S4). Specifically, we used these estimates to 298 simulate the interactions between the two isolates in a population with an equal number young 299 adults form the plastic and the non-plastic strains at the start of the simulation. Notably, 300 simulated populations were completely dominated by the non-plastic strain for both food 301 conditions. In addition, rapid elimination of the plastic strains prevents the formation of its dauer 302 larvae, as J2 animals of this strain were completely eradicated by the non-plastic strain (Fig. 4e,  303 f). Thus, the cost of plasticity greatly affects the dynamics of the plastic strain in a spatially-304 homogenous population. 305

Spatial structure significantly affects population dynamics 306
While modeling the interaction of the plastic and the non-plastic strains in a population without 307 any spatial structure is informative, a more realistic scenario would involve dispersal from 308 different populations upon the depletion of food on the beetle carcass, and competition over the 309 nutrient-rich carcasses in the vicinity. Exploring such scenarios in the lab would be a tremendous 310 undertaking. Therefore, we extended our model to include a stepping-stone migration scenario to 311 illustrate the effect of costs of plasticity and phenotype on the competitive dynamics between P. 312 pacificus strains in a structured population. We constructed a simple structured population by 313 arranging n localities in one dimension. Each simulation starts with 50 young adults (YAs) of the 314 plastic strain in the first locality and 50 YAs of the non-plastic strain in the n th locality, with rest 315 of localities being empty. All the localities contain a fixed amount of resource and dauer larvae 316 migrate with a fixed rate from a food-poor locality to a neighboring food-rich locality (Fig. 5a). 317 The simulation concludes when all the food in every locality has been depleted. 318 Based on these simulations, on E. coli, higher fecundity of the plastic strain allows adults of this 319 isolate to completely dominate the structured population even in the face of predation (Fig. 5b, SI  320 Appendix; Fig. S5a). Although the predation rate of the non-plastic strain is higher, even dauer 321 larvae of the plastic strain continues to fully dominate the structured population (Fig. 5c). These 322 results support a considerable cost of phenotype for the non-plastic strain in the spatially-323 structured population. In addition, in a scenario without predatory interactions, the frequency of 324 the plastic strain decreases only marginally (Fig 5b vs. d, c vs. e). This finding results from the 325 change in the number of migratory dauer larvae of the non-plastic strain (SI Appendix; Fig.  326 S5a,b). Most importantly, the pace in which the plastic population grows, results in exceptionally 327 high numbers of predators belonging to the plastic strain, which outcompete the non-plastic 328 strain in the presence of interaction. Thus, the cost of phenotype substantially influences the 329 abundance of the non-plastic strain, in particular in the presence of interactions. 330 On Novosphingobium, higher fecundity and faster developmental speed of the non-plastic strain 331 turns this isolate into a formidable adversary for the plastic strain. Therefore, the frequencies of 332 adults and dauer larvae of the plastic strain are extremely reduced in the structured population 333 (Fig. 5b, c). However, when interactions are limited, in contrast to E. coli, the frequencies would 334 slightly increase (Fig 5b vs. d, c vs. e). This is due to the non-plastic strain profiting from a 335 higher growth rate and higher predation on Novosphingobium, but only higher growth when 336 interactions are eliminated (SI Appendix; Fig. S5c,d). Thus, the cost of plasticity would greatly 337 affect the abundance of the plastic strain when competing with a predator under this condition. 338

Initial food source also affects population dynamics 339
To capture how significantly the costs of plasticity and phenotype would affect the dynamics of 340 structured populations, we simulated two scenarios where each isolate would start with a 341 favorable food source; E. coli for the plastic strain, and Novosphingobium for the non-plastic 342 strain, or the unfavorable food source; Novosphingobium for the plastic strain, and E. coli for the 343 non-plastic strain (Fig. 5b-e). A pair of food sources were labeled "favorable" or "unfavorable" 344 for a strain given the relative fecundity of the strain on each source. Interestingly, the results 345 indicate that the initial condition in which each population starts dramatically affects which 346 strain would ultimately dominate the structured population. When the conditions are favorable 347 for both strains, the cost of plasticity of the plastic strain is greater than the cost of phenotype of 348 the non-plastic strain. In contrast, the relationship between the costs reverses under conditions 349 that are unfavorable to both strains. Thus, the interaction of the cost of phenotype and the cost of 350 plasticity is context dependent. Together, these simulations reveal that spatial structure and initial 351 food sources could affect the population dynamics with different consequences for the costs of 352 plasticity and phenotypes on the two isolates. However, such projections about the population 353 dynamics of these strains of P. pacificus should be taken with caution, as many aspects of P. 354 pacificus population dynamics and its dispersal patterns in the wild remain poorly understood. 355

Discussion 356
Experiential detection of the costs associated with plasticity, especially in metazoans, has proved 357 to be a daunting challenge. For instance, the predator-induced spine of Daphnia pulex was 358 reported to show mild support for both the costs of production and maintenance (Scheiner & Berrigan 1998). Similarly, in the Scandinavian frog, Rana temporaria, the costs of metamorphic 360 size were shown to exhibit a plasticity cost in southern populations, whereas northern 361 populations displayed no such costs (Merilä et al. 2004). Van Buskirk and Steiner (2009), in 362 their meta-analysis concluded that costs of plasticity are mostly low, if existing at all. However, 363 the same authors suggested that these costs may influence adaptive evolution under stressful 364 conditions. Additionally, meta-analysis on aquatic gastropods argued for further empirical 365 investigations to better quantify the energetic costs of plasticity of shell formation (Bourdeau et 366 al. 2015). A more recent study on the cannibalistic cane toads, signifies favoring canalized 367 defenses over plasticity, providing the high cost of plasticity rather than the cost of phenotype 368 (Devore et al. 2021). Together, this diversity of findings indicates the need for establishing a 369 comprehensive empirical framework to address both theoretical and conceptual asserts. It has been argued that resource polyphenism -i.e., the environmental induction of alternative 381 phenotypes to use different resources, such as the development of cannibalistic morphs as a 382 response to environmental stress (Pfennig & McGee 2010) -is the most relevant of discrete 383 plastic response. Cannibalism provides trophic and survival advantages by either extending 384 energy resources or eliminating competition (Church & Sherratt 1996;Claessen et al. 2004). It 385 has been suggested that the predatory mouth form in P. pacificus boosts survivorship under 386 severe conditions (Serobyan et al. 2014), and reduces competition on the basis of genomic 387 relatedness (Lightfoot et al. 2021). Nevertheless, various P. pacificus natural isolates are either 388 predominantly non-predatory or intermediately so. Our results suggest that, in isolation, the 389 fitness payoff incurred by the predatory-biased population makes it inferior to the non-predatory-be more detrimental when both isolates are interacting in a spatially-homogenous population. 392 The effect of growth rate, developmental speed, and predation are highly context dependent, as 393 shown by our simulations under different starting conditions, resulting in different population 394 dynamics (SI Appendix ; Fig. S5). 395

396
The effect of population structure and the non-homogenous distribution of resources in the 397 environment on the outcome of competition between a plastic and non-plastic strain illustrates 398 the complex nature of the ecological consequences of the cost of plasticity. The role of 399 phenotypic plasticity, and dispersal, in invasions have long been appreciated (Sharma et al. 400 2005), based on the assumption that plasticity provides a "Jack-of-all-trades" strategy. This 401 assumption has been challenged (Hulme 2007), and does not explain the pattern we observe in 402 our model. The models proposed to predict the population-level consequences of plasticity thus 403 far (reviewed in Wennersten and Forsman (2012)) have been almost entirely conceptual. A 404 promising recent attempt by Brass et al. (2021) incorporates plasticity in a continuous-time 405 stage-structured model to predict the ecological effects of plasticity, but their model differs from 406 ours, since they include plasticity as maternally-determined phenotypic variation within a species 407 and do not explore the effect of population structure nor the non-homogenous resource 408 distribution on the cost of plasticity. The effects of spatial heterogeneity and dispersal on the 409 evolution of plasticity have been explored before, e.g., Scheiner and Holt (2012) and Edelaar et 410 al. (2017), but our results are not comparable to those, since our model lacks any evolutionary 411 component, e.g., mutation, recombination, etc., and is solely concerned with the short-term 412 ecological consequences of plasticity in P. pacificus. 413 414 Taken together, our results suggest a four-pronged explanatory framework, combining the cost of 415 plasticity, cost of phenotype, environmental influence, and population structure, each playing a 416 crucial role in adaptive plasticity. However, several questions remain to be answered. For 417 example, measuring predation dynamics and migration rates on beetle carcasses can increase the 418 accuracy of modeling approaches. Also, predator consumption might differ as a functional 419 response to prey density, given search, handling time, foraging efficiency, and predation risks 420 (Solomon 1949;Holling 1959a, b;Lima et al. 1985;Sentis et al. 2013). Additionally, in nature, 421 nematode mobility is not restricted to a one-dimensional dispersal. Thus, such parameters merits further empirical and theoretical analyses. Finally, a key question that was hardly identified in 423 other plastic systems is the molecular machinery underlying the production and maintenance of 424 plasticity (Pigliucci 2001;Murren et al. 2015). In P. pacificus, the readily available molecular 425 techniques permit such potential investigations. In conclusion, this study integrates empirical and 426 theoretical approaches to emphasize how different types of costs influence the evolution of 427 adaptive plasticity, while setting the stage for further investigations. 428 YA= young adults with no eggs inside the uterus; BA= breeding adults with eggs inside the 463 uterus. In (b) and (d), the 95%HDI for each strain was estimated using a Bayesian approach to 464 estimate the probability of expressing the predatory mouth morph based on the observations; the 465 95% HDI for the means and the difference in means in (c) and (e) was calculated using 466 Kruschke's BEST method. We used [-5,5] interval as our ROPE, i.e., differences of means 467 within this interval are practically equal to no difference; the same ROPE and was used for all 468 the analyses in this manuscript (see Supplementary Methods). 469 with 50 YAs of the plastic strain on the first locality and 50 YAs of the non-plastic strain on the 518 12 th locality. At each step, ߱ dauer larvae migrate from population i to j if j has more food than 519 The frequency of the plastic strain adults (YA, BA i , OA) (b) and dauer larvae (c) across 520 12 localities with interaction (i.e., predation) after 1000 steps. As previously noted, the end point 521 of 1000 steps represent an arbitrary endpoint, which roughly corresponds to 10 generations. 522 Food will be gone long before as evident by the production of dauer larvae. (d-e) The frequency 523 of the plastic strain adults (YA, BA i , OA) (d) and dauer larvae (e) across 12 localities assuming 524 no interaction between the two strains. In this scenario, the frequency of the plastic strain in the 525 metapopulation is a function of developmental speed and fecundity only. At the start of the 526 simulation, for each strain in localities 1 and 12, n E = n J2 = n J3 = n J4 = n BA = n OA = 0, and n YA = 527 50. while m = 0.1. The initial food supply, S 0 = 10 12 in the entire metapopulation. 528 529

Data availability 530
The software used to run all simulations and conduct all the data analysis was written in Python 531 3.10.4. For reproducibility, the code and the raw experimental data are available at 532