Evolution has the capacity to alter the course of biological invasions, although such changes remain mostly unexplored by experiments. Integrating evolution into studies of invasions is important, because species traits can potentially evolve in ways that either moderate or exacerbate the impacts of invasions. We have assessed whether species evolved during experimental invasions by comparing the performance of founder populations and their potentially evolved descendants in communities of ciliates and rotifers. Residents (analogous to native species) that have previous experience with invaders consistently reduced the performance of naive invaders, supporting the emergence of increased biotic resistance as one consequence of evolution during invasions. Experienced invaders exhibited both increased and decreased performance depending on the invader species considered. Through its influence on performance and species abundance, evolution also changed community composition during the course of invasions. The idiosyncratic patterns of evolutionary changes in invading and resident species complicate predictions about the long-term consequences of invasions from initial post-invasion dynamics.
Biological invasions pose a major threat to biodiversity1, yet research on the eco-evolutionary dynamics of biological invasions is still in its earliest stages2,
Evolutionary changes (occurring over fewer than 100 generations) in morphology, behaviour and genotype occur in both invaders and native species during biological invasions10,
Strong evidence for evolution during invasions requires demonstration of stable changes in traits in response to an imposed selective regime. In the context of biological invasions, this requires measuring traits before and after invaders and residents have interacted for some time. Studying eco-evolutionary dynamics in invaded communities is especially challenging; resident and exotic species are often studied early in the course of interaction, before much evolution can occur, or much later, when species may have evolved, but initial trait states are no longer available for comparison. Often the detailed history of introductions and spread for invading exotic species is poorly known22,23, which further complicates inferences about potential evolution during the course of invasions. In an ideal scenario the traits of both resident and invading species early and later in the course of invasions would be measured to definitively assess whether evolution affects long-term invasion dynamics22. Such comparisons are possible in laboratory communities that are composed of organisms with short generation times where evolution can occur over experimentally tractable time frames. Experiments using these systems can alter the duration of post-invasion interactions and manipulate opportunities for potential evolution24.
We compared the performance of populations of resident and invading species before and after they had interacted, and potentially evolved, for approximately 200–400 generations. We used two different resident species assemblages (hereafter termed A and B) composed of aquatic bacteria, protists and rotifers. One species from each assemblage was designated as an experimental invader of the other. Assemblage A contained five ciliates and a rotifer, and assemblage B contained three different ciliates and a different rotifer (Supplementary Fig. 1). Paramecium bursaria (a resident in assemblage A) and Euplotes daidaleos (a resident in assemblage B) were selected as invaders that were experimentally introduced into the other assemblage (Fig. 1). The designated invaders were functionally similar (mixotrophic bacteriovores that contain symbiotic chlorellae) and the resident assemblages that they invaded contained congeners (E. patella and P. caudatum) with which they might be expected to interact strongly. Theory suggests that the consequences of evolution between invaders and natives may depend on the type of interspecific interaction involved25,26 (Supplementary Table 1), even leading to reduced performance or extinction27 in some circumstances.
Our experiment addressed three general questions. (1) Does evolution alter population abundances and community composition when one species invades a community? (2) Are patterns of changed abundance consistent with altered biotic resistance of residents or invasiveness of invaders? (3) Are evolutionary responses idiosyncratic or consistent across different communities and invaders? Post-invasion abundance was the trait used to assess performance and infer evolution in experimental treatments with different interaction histories (Fig. 1). By varying whether or not invaders and residents interacted for many generations before assessing performance in a second round of invasion, we determined whether performance changed over time. Treatments differed in the experienced (those that have a recent interaction history of approximately 200–400 protist generations) or naive status (no recent interaction history) of invading and resident species: (1) invading and resident species both experienced (+I/+R); (2) experienced invaders and naive residents (+I/−R); (3) invaders and residents both naive (−I/−R); (4) naive invaders and experienced residents (−I/+R); and (5) an uninvaded control containing only resident species originally present before invasions. We compared abundances of species against multiple a priori predictions for both invaders and residents (Supplementary Fig. 2). We expected that: (1) if prolonged interaction resulted in increased invader performance, experienced invaders (+I) would outperform naive invaders (−I) when grown with naive residents (+I/−R > −I/−R, Fig. 1a), and residents (+R or −R) would have lower abundances when grown with experienced versus naive invaders (+R/+I < +R/−I and −R/+I < −R/−I, Supplementary Fig. 2); (2) if experience increased resident performance, naive invaders (−I) grown with experienced residents (+R) would have lower abundances compared with those grown with naive residents (−I/+R < −I/−R, Fig. 1b). We expected the abundances of experienced residents (+R) to be higher than naive residents (−R) when interacting with experienced or naive invaders(+R/+I > −R/+I and +R/−I > −R/−I, Supplementary Fig. 2).
A post-invasion history of interaction with resident species changed the performance of both invaders, E. daidaleos invading community A (ANOVA F3,16 = 97.25, P < 0.0001) and P. bursaria invading community B (F3,16 = 54.80, P < 0.0001). E. daidaleos performance increased when populations were founded with individuals that had a previous history of interaction with residents Fig. 2a,c).The highest mean abundance occurred in the experienced invader/experienced residents (+I/+R) treatment. Experience with the residents of assemblage A also improved the performance of E. daidaleos over that in its source community, assemblage B (Fig. 2c). P. bursaria performance decreased when interacting with experienced residents (Fig. 2b,d), suggesting that changes in the residents of assemblage B resulted in increased biotic resistance to invaders. Changes in P. bursaria performance after interacting with residents of the invaded assemblage also reduced its average abundance relative to that in its source community, assemblage A (Fig. 2d). We also noted that initial invasion by naive P. bursaria into an experienced resident assemblage (−I/+R) required roughly twice the propagule pressure of invaders and lower abundances of resident species relative to invaders in other treatments.
Interaction history also influenced the abundance of resident species in both communities (assemblage A: MANOVA F20,54 = 16.43, P < 0.0001; assemblage B: F16,52.6 = 3.80, P = 0.0001), with responses varying idiosyncratically among species (Figs 3,4). Four species in assemblage A (Blepharisma americanum, E. patella, Prorodon niveus, and P. bursaria) became less abundant after prolonged interactions with an invader, while one species from assemblage A (Spirostomum teres) and three from assemblage B (Stentor coeruleus, P. caudatum, and E. daidaleos) became more abundant after prolonged interactions with an invader. Interaction history did not affect the performance of either rotifer species. The collective changes in species abundances associated with different interaction histories changed the overall species composition of both assemblages. In assemblage A, communities composed of experienced invaders and residents (+I/+R) had elevated abundances of E. daidaleos and S. teres relative to other treatments. In the remaining treatments assemblage A had higher abundances of B. americanum and E. patella (Fig. 5a). For assemblage B, the two treatments with naive residents (+I/−R, −I/−R) were most similar in composition (Fig. 5b). These treatments had higher abundances of one resident species, Monostyla sp., and the invader, P. bursaria, while treatments with experienced residents had higher abundances of S. coeruleus and lower abundances of the invader P. bursaria.
The history of interactions altered the performance of both resident and invading species in our experiment. We propose that evolution, rather than phenotypic plasticity, led to observed differences in performance28. Within assemblages, treatments differed only in the duration of potential preadaptation of the invaders and residents, and not in the identities of species present28. Attributing differences to phenotypic plasticity would require that: (1) different plastic (nongenetic) responses were induced by the same set species, with differences depending only on the history of interaction; and (2) that those plastic responses would persist and be transmitted over many generations of population growth, long after the differences in inducing conditions had ended. Studies of other organisms, including a ciliate, document that rapid evolutionary responses to interaction history and species identity can occur29,
Species in our assemblages can potentially interact directly and/or indirectly via multiple food web pathways (see Supplementary Fig. 1); consequently we cannot unambiguously attribute altered performance to changes in traits such as competitive ability, antipredator defence, or attack rates. At least two resident species, E. patella and P. niveus, became less abundant in the treatment containing experienced invaders and residents (+I/+R), so much so that they became difficult to detect (Fig. 3). This result would not have been apparent if only the naive invasion treatment (−I/−R), analogous to recent invasions in nature, was compared with the uninvaded resident community (control treatment). This result underscores the importance of considering long-term eco-evolutionary dynamics in biological invasions, since some responses may take many generations to become apparent.
Evolution can potentially enhance the performance of invaders and/or increase the biotic resistance of resident species. Evolution of increased invader performance (for example, E. daidaleos invading assemblage A) can exacerbate the impact of invasions over time. Other invaders may become less successful over time if increased invader performance is outstripped by increased resident performance (for example, assemblage B), leading to the appearance of increased biotic resistance. The amount of existing genetic variation and the strength of selection on traits that affect performance can both influence these differences in evolutionary potential. The long-term consequences of evolution among invaders and residents may be difficult to predict from early post-invasion patterns34, but our findings suggest that eco-evolutionary dynamics may alternately exacerbate or reduce community-level impacts over time. Because most biological invasions are difficult or impossible to reverse, it is important to recognize that phenotypic changes will sometimes ameliorate the impacts of invaders without human intervention, while other invasions will require continued intervention to minimize their impacts. The challenge remains in predicting when the initial impacts of invasions will become more or less severe over time.
Our experimental protocol used protists and rotifers with short generation times so that evolutionary responses to different selective regimes imposed by different invasion scenarios could be measured over tractable periods of time. The experiment involved three phases to ensure that populations of species had ample time for potential evolutionary change and possible coadaptation to be measured. Phase 1 ensured that the species grown in each assemblage, A and B, had sufficient time to interact and potentially coadapt to one another, thus sharing a recent history of interaction that spanned 300–550 protist generations. Phase 2 established the actual invasion of replicates of each assemblage by an invader drawn from the other community. Species in the invaded treatment interacted for a sufficient number of generations (approximately 200–400 protist generations) so that the invaders and species in the resident communities (analogous to native species with a shared history of interaction and potential coadaptation) might have time to potentially evolve. Phase 3 compared the performance of different lines of potentially evolved invaders and residents with naive invaders and residents in all possible combinations to assess the effects of interaction history on species performance and community composition. Because the treatments within each community differed solely in terms of the length of interaction history, rather than species composition or external environmental factors, phenotypically plastic responses would be consistent across treatments within a community (and probably transient in nature). If the observed differences among treatments did involve an inducible defence, it would imply that some sort of evolutionary change drove differences in plastic traits. A detailed description of the experimental design follows.
The original ‘native’ species, termed residents, in each source assemblage consisted of naturally co-occurring bacterivorous and predatory ciliates and rotifers originally collected from the same pond (Bamboo Pond) on the campus of Rutgers University. Assemblage A contained five ciliates and a rotifer, B. americanum, S. teres, E. patella, P. niveus, P. bursaria, and Lecane sp. (Supplementary Fig. 1). Assemblage B contained three ciliates and a second rotifer, S. coeruleus, P. caudatum, E. daidaleos and Monostyla sp. (Supplementary Fig. 1). These assemblages were selected for their stability in species composition and were maintained in laboratory cultures for 3 years, equal to at least 550 (slowest growing protist species) to 1,095 (fastest growing protists) protist generations and approximately 275 generations for rotifers, before the start of the experiment. B. americanum, S. teres, E. patella, P. niveus and Lecane sp. were grown continuously together during this period, as were S. coeruleus, P. caudatum and Monostyla sp. Each invader was maintained in separate laboratory cultures from first collection until several months before the start of phase 1 of the experiment, at which time they were added to their respective source assemblages (P. bursaria to assemblage A and E. daidaleos to assemblage B). We initiated our experiment to provide 18 months, or about 300–550 protist generations, in which species in each assemblage further coadapted to laboratory conditions and one another (Fig. 1, phase 1). In this way we ensured a prolonged period of recent evolutionary history for each assemblage maintained separately in the lab. Documented microgeographic evolution suggests that even within a single geographic range of a population, different subpopulations of a species can differ both genetically and in terms of their interactions within a community, suggesting that the prolonged period of laboratory culture that our assemblages experienced allopatrically would allow them to potentially interact differently depending on our experimental treatment. If the prolonged recent evolutionary history in the lab was insufficient to overcome the previous history before collection from Bamboo Pond, we would not expect to see an effect of our treatment. Assemblages were initially inoculated with the same four bacterial species, Proteus vulgaris, Serratia marcesens, Bacillus subtilis, and B. cereus, as known edible resources for the bacterivores, but we did not quantify the number and abundances of bacterial species and others were probably present. Five treatments manipulated the duration of interaction history following invasion and possible evolution for assemblage A and B, for a total of 10 treatments, each replicated five times in separate microcosms.
Microcosms were loosely lidded 250-ml jars with 100 ml of medium (1 Protist Pellet, Carolina Biological Supply, plus 0.14 mg Herptivite vitamin supplement in 2.8 l of well water, autoclaved before the addition of organisms) and two sterile wheat seeds for additional nutrients. This phase created the starting conditions for each assemblage that was subsequently subjected to experimental invasions. Communities were subcultured every three weeks by gently mixing each jar and then adding approximately 3 ml of old culture to a new jar with fresh medium. To ensure that bacterial resource species remained consistent throughout the experiment, fresh medium was bacterized with the same stock strains of bacteria consistently grown individually in the absence of bacterivores. Bacteria in fresh medium grew for three days at room temperature and were highly abundant before the addition of 3 ml of old culture containing ciliates, rotifers, and bacteria. Jar location was randomized in a Percival incubator at 22 °C with a 5-h light/19-h dark photoperiod. Counts of the number of individuals in subsamples of a known volume from each jar were used to estimate species densities weekly.
The second phase of the experiment involved invading replicates of A and B with an invader drawn from the other assemblage to create a situation in which the invaders and resident species had an opportunity to potentially evolve. We experimentally invaded five replicates of resident community A with 15–20 cells of E. daidaleos originally grown in source community B, and resident community B with 15–20 cells of P. bursaria originally grown in source community A (Fig. 1, phase 2, +I/+R treatments). Previous experience with these species indicated that this number of introduced individuals is sufficient to establish populations. We chose these mixotrophic species as invaders (both contain photosynthetic symbionts) because they are functionally similar. Both also invaded assemblages in which another congener was already established. This approach achieved some potential degree of niche separation between resident and invading species, while still allowing for interspecific interactions between organisms. Following these initial invasions, communities grew for another 13 months, or approximately 200–400 protist generations (Fig. 1, phase 2), allowing invaders and residents to potentially evolve over time. At the same time, replicates of the original uninvaded assemblages remained a source of naive species without a history of post-invasion interaction (Fig. 1).
The third phase involved testing whether populations that had an opportunity to evolve for 200–400 protist generations after invasion differed in performance from populations that did not have an opportunity to evolve. We established invasions during a three-week period to create the final treatments used to evaluate the performance of naive or potentially coevolved sets of species (Fig. 1, phase 3). For treatments with naive invaders and experienced residents (−I/+R), we first isolated inocula of invasion-experienced resident species (from experienced invader/experienced resident treatments for each assemblage) and reassembled them into communities with the original resident species composition for A and B. Next we invaded them with an inoculum of invaders derived from populations of naive invaders. For each of the invasions we again used an inoculum of 15–20 invading cells per replicate. Similarly, freshly subcultured assemblages of naive residents (−R) were inoculated with invaders derived from experienced (+I) or naive populations (−I) (with or without a long history of interaction with members of the invaded assemblages). All treatments were subsequently checked for successful establishment of the invaders. To assess the effect of interaction history on post-invasion community dynamics, we again estimated the abundances of all species weekly for six weeks.
We calculated species densities as number +1 per ml and then log10-transformed the values before analysis. We then averaged log10-transformed densities over the six-week time-period for each replicate following the final invasions to capture the post-invasion dynamics. These mean values became the response variables for our statistical analyses. For each assemblage, A and B, a multivariate analysis of variance (MANOVA SAS 9.435) evaluated the effect of potential evolutionary treatments on the mean of the log of each resident species’ abundance. In assemblage A, one native species, P. niveus, failed to establish in the −I/+R treatment, probably because of exceptionally small propagule size resulting from the difficulty of locating P. niveus individuals in the experienced invader/experienced residents (+I/+R) treatment when we established the −I/+R treatment. For this resident species and each invader, we instead conducted separate ANOVAs for dynamics. ANOVAs testing the response of invaders included the ‘source’ community treatment to determine how evolutionary experience affected performance relative to naive invaders in their source community. We evaluated significant differences between treatment means for each response variable using Tukey’s Honestly Significant Difference multiple comparison test (at α = 0.05) to help identify which treatments differed in abundance.
To evaluate the overall community response to differing evolutionary histories of invaders and resident species, we conducted principal component analyses of log10-transformed mean species abundances for each assemblage for the post-invasion compositions. For assemblage A, native species P. niveus was not included in the principal component analysis to prevent its lack of establishment in the −I/+R treatment from unduly affecting the analysis on community composition. We then used community scores for the first three principal components as response variables in separate MANOVAs for each assemblage for the post-invasion compositions.
The data from this study supporting our findings are available from the corresponding author upon reasonable request.
How to cite this article: Faillace, C. A. & Morin, P. J. Evolution alters the consequences of invasions in experimental communities. Nat. Ecol. Evol. 1, 0013 (2016).
This research was supported by Rutgers University (P.J.M.) and a Rutgers University Graduate Program in Ecology and Evolution Ted Stiles Grant (C.A.F). We thank T. Fukami, S. Lawler, S. Naeem, O. Petchey, and members of the Morin Laboratory Group for comments on the manuscript.
Supplementary Figures 1 and 2, Supplementary Table 1