Priority effects are weakened by a short, but not long, history of sympatric evolution

Abstract

Priority effects, or the effects of species arrival history on local species abundances, have been documented in a range of taxa. However, factors determining the extent to which priority effects affect community assembly remain unclear. Using laboratory populations of the bacterium Pseudomonas fluorescens, we examined whether shared evolutionary history affected the strength of priority effects. We hypothesized that sympatric evolution of populations belonging to the same guild would lead to niche differentiation, resulting in phenotypic complementarity that weakens priority effects. Consistent with this hypothesis, we found that priority effects tended to be weaker in sympatrically evolved pairs of immigrating populations than in allopatrically evolved pairs. Furthermore, priority effects were weaker under higher phenotypic complementarity. However, these patterns were observed only in populations with a relatively short history of sympatric evolution, and disappeared when populations had evolved together for a long time. Together, our results suggest that the evolutionary history of organismal traits may dictate the strength of priority effects and, consequently, the extent of historical contingency in the assembly of ecological communities.

1. Introduction

Priority effects, in which the order of species arrival at local habitat patches dictates the outcome of local species interactions, can result in historical contingency in community assembly, altering community structure and function [1–3]. Historical contingency caused by priority effects has been found in a range of organisms, including bacteria (e.g. [4]), fungi (e.g. [5]), plants (e.g. [6]) and animals (e.g. [7]), and increasing evidence indicates that the extent of historical contingency can be partly predicted from environmental conditions such as nutrient availability, disturbance frequency and temperature variability [8,9]. However, the strength of priority effects must be modulated not just by environmental conditions, but also by organismal traits that determine how species interact with one another [9]. These traits are often shaped by the evolutionary history of species [10], although only a limited number of studies have linked evolutionary history and priority effects [11–13]. Consider, for example, the evolutionary history of immigrants that colonize island communities. If all immigrants came to an island from the same mainland, immigrants might have traits that reflect a long history of shared, sympatric evolution. By contrast, if immigrants came from different, allopatric regions, some of the immigrants might encounter one another for the first time on the colonized island, with their traits having little shared evolutionary influence. These differences in the amount of prior evolutionary history that shapes species traits can determine the strength of priority effects, but the evidence needed to test this possibility is largely lacking.

In theory, shared evolutionary history can weaken priority effects if sympatric evolution results in niche differentiation, as expected from the concept of character displacement [14]. Niche differentiation can diminish priority effects because an early-arriving population might then exploit only the niche space it has specialized on, allowing other, late-arriving populations to use their own niche space. It is also possible, however, that shared, sympatric evolutionary history strengthens priority effects because sympatric evolution can cause populations to become similar in competitive ability, as expected from the neutral theory of biodiversity [15]. In this case, an early-arriving species could pre-empt local niche space, making it more difficult for subsequent species to establish. Which scenario is more likely may depend on the duration of evolutionary history. For example, niche partitioning can evolve rapidly in new environments (e.g. [10,16]), whereas trait convergence has been shown to evolve when species interact over longer time periods (e.g. [17,18]). Alternatively, either of these processes could operate exclusively, leading to either extreme niche differentiation or complete ecological neutrality, respectively.

In this paper, we report the first experimental test, to our knowledge, of the effects of both the presence and duration of shared evolutionary history on the strength of priority effects. Specifically, using the bacterium Pseudomonas fluorescens as an experimental system, we ask the following questions: does a history of sympatric evolution weaken or strengthen priority effects? If so, what is the biological mechanism underlying the role of evolutionary history and does the influence of evolutionary history on priority effects depend on the duration of prior evolution? Experimental populations of P. fluorescens are uniquely suited for asking these questions because of extensive knowledge on adaptive evolution in this bacterium. When founded from a single ancestral genotype, P. fluorescens populations rapidly diversify under static culture conditions into genetically based niche specialists, including multiple ‘wrinkly spreader' (WS) types, which colonize the air–liquid interface to form a biofilm mat [19–24]. We used these multiple WS types for our experiments.

Based on previous research, we expected that the WS genotypes that evolved sympatrically in static culture conditions would display character displacement. If the air–liquid interface is an environment that can be divided into multiple niches, character displacement among WS genotypes may occur within biofilm mats through differentiation of surface attachment mechanisms ([11,22,25], see also [26]). For this reason, we predicted that sympatric pairs of WS genotypes would, on average, exhibit weaker priority effects than pairs that had evolved allopatrically. Alternatively, it is conceivable that WS genotypes have similar competitive abilities if they have evolved in sympatry [23]. If sympatrically evolved pairs were competitively more similar than allopatrically evolved pairs, one might expect stronger priority effects via niche pre-emption in sympatrically evolved pairs. As detailed below, our results support the first hypothesis, but only when the history of sympatric evolution was relatively short.

2. Material and methods

(a). Generation of sympatrically and allopatrically evolved pairs

To generate sympatrically and allopatrically evolved pairs of strains, we independently propagated 12 replicated pairs of P. fluorescens in static 6 ml cultures of standard King's B (KB) liquid media at 28°C for seven weeks, with weekly transfers of 60 µl of homogenized culture to fresh media. Each of these 12 microcosms was founded with the same pair of two ancestral clones: a wild-type SBW25 clone [19] and a SBW25::lacZ mutant clone [27]. Use of the mutant with a neutral lacZ marker ensured that derived strains were easily distinguishable when plated on KB agar supplemented with 50 µg ml⁻¹ 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-gal) [23,27].

Each week, immediately after the transfer to fresh media, the cultured medium of each microcosm was plated, and two WS strains (one strain with LacZ and one without) isolated and stored in 70% glycerol at −80°C. This treatment resulted in 24 populations with neutral markers (e.g. one strain with LacZ, and one without; two strains per selection line). Pairings of these lacZ+ and lacZ− WS types were used in the community assembly experiments described below, to determine how shared evolutionary history influenced the strength of priority effects during community assembly. Since reproduction is completely asexual in P. fluorescens, different genetically based morphotypes are analogous to species in a community [28]. As such, we will refer to vials with multiple morphotypes as communities. See the electronic supplementary material for a diagrammatic representation of the design of the generation of the strain used in the immigration experiments (electronic supplementary material, figure S1a), and the pairing of strains for sympatric and allopatric pairs (electronic supplementary material, figure S1b).

(b). Community assembly

We used replicated pairs of strains to assess how evolutionary history influences priority effects after one, two and seven weeks of prior evolution. This experiment followed the general methods of Fukami et al. [23] and Knope et al. [12]. The strains for these pairs were isolates from the selection lines described above. For each pairing, microcosms for community assembly were independently initiated with one lacZ+ and one lacZ− strain for easy enumeration of ancestral state. For each duration of evolution, the two treatments were pairs with and without history of sympatric evolution. Sympatrically evolved pairs consisted of alternatively marked WS isolates (i.e. lacZ+ and lacZ−) from the same microcosm of the experiment described above. Allopatrically evolved pairs consisted of alternatively marked WS isolates from separate microcosms. For both sympatrically and allopatrically evolved pairs, all strains used were derived from the evolution experiment described above (electronic supplementary material, figure S1). In other words, for each duration of evolutionary history, all strains used for the community assembly experiment were evolved for the same amount of time.

For both sympatrically and allopatrically evolved pairs, each replicated 12 times for a total of 24 pairs, we had two treatments of immigration order: the strain of one lacZ status was introduced to 6 ml of sterile, static KB media on day 0 (first strain), with the opposite lacZ marker introduced to the same microcosm 24 h later (second strain). After both strains were introduced, abundance (colony forming units, or CFU) was quantified through destructive harvesting and dilution plating on days 0, 1, 2, 4, 6, 8 and 10. Each of the 24 replicated pairs were sampled at each of these time points. For each duration of evolutionary history (i.e. one, two or seven weeks), the strain pairs used were independent of other weeks. In other words, at each week, strain pairs were isolated from different propagated selection lines.

(c). Quantifying priority effects

We calculated the strength and direction of priority effects, P_ij, by comparing how much a specific strain grew when it was introduced before another strain to how much it grew when it was introduced after. Following Vannette & Fukami [29], we quantified P_ij as the log of the ratio between the abundance of strain i, time-averaged over days 4 through 10, when introduced after strain j, D(i)_ji, and the abundance of strain i, also time-averaged over days 4 through 10, when it was introduced before j, D(i)_ij:

where P_ij values of zero indicate the absence of a priority effect, positive P_ij values indicate a facilitative priority effect, and negative P_ij values indicate an inhibitory priority effect [29].

(d). Quantifying complementarity: biofilm phenotypes

We expected sympatric evolution to result in phenotypic complementarity (a form of character displacement), and that pairs with more complementary phenotypes would in turn show weaker priority effects. To test this expectation, we characterized the phenotype of the biofilm formed at the air–liquid interface in static culture of each strain in independent monoculture. We measured biofilm thickness and the presence of a webbed appearance in the biofilms. The emergence of biofilms with this webbed appearance in P. fluorescens has previously been described [22,25]. Spiers & Rainey [22] found that cellulose increases cell–cell attachment and that a proteinacous attachment factor results in attachment to the glass walls of the vial. The webbed phenotype corresponds to more cell–cell adhesion, whereas the non-webbed phenotypes are more adhesive to the glass [25]. When inoculated together, the resulting biofilms are stronger than either phenotype alone [22,25]. We scored thickness as one of three categories of thin (scored as 0), medium (1) and thick (2), and webbing as either zero (not webbed) or one (webbed). For each strain, mat phenotypes were averaged from two replicates of independently formed air–liquid interface mats. From these two mat characteristics, we assigned a mat phenotype from which all pairwise distances among strains could be calculated using Euclidean distance. For any given pair of strains, we used values of this distance as an index of mat phenotypic dissimilarity and, more generally speaking, a proxy for phenotypic complementarity.

(e). Statistical analyses

Using general linear mixed models (GLMMs), we tested for the effects of evolutionary history, i.e. sympatric versus allopatric evolution (referred to as Evolutionary Treatment), the duration of evolutionary history, i.e. one, two or seven weeks (referred to as Week), and the timing of observation, i.e. days 4 through 10, during community assembly (referred to as Day) on the difference in abundance between first- and second-arriving strains. Similarly, we also tested for the effects of Evolutionary Treatment, Week and the biofilm phenotypic distance (referred to as Distance) on the strength of priority effects (P_ij values). In addition, we tested for the effects of Evolutionary Treatment and Week on the difference in the strength of priority effects between sympatrically evolved and allopatrically evolved pairs (ΔP). In each of these analyses, we included all possible two- and three-way interactions among factors in a full initial model. In each of the models, we included the identity of the selection lines from which strains were isolated as random effects. We performed model selection based on corrected AIC (AICc) rankings to determine the best-fit models [30] (see electronic supplementary material for AICc tables).

We complemented these analyses with two additional approaches: (i) t-tests that tested for the differences in abundances between the first- and second-arriving strains, (ii) t-tests that tested for the differences in the strength of priority effects between sympatrically and allopatrically evolved pairs and (iii) linear regressions that tested for significant relationships between biofilm phenotypic differences and the strength of priority effects.

For all analyses, we assumed that the density of any strain for which we found no colonies on the dilution plates was 10⁶ CFU ml⁻¹, a value that is roughly an order of magnitude lower than the detection threshold for our plates. The actual density could have been lower than 10⁶ CFU ml⁻¹ in some cases, but we used this value as a conservative method that could underestimate, but not overestimate, the strength of priority effects. As an alternative method, we also did the analyses after discarding all of the replicates where no colony was found for a strain. This alternative method yielded qualitatively identical results to the ones we present here. For each of the three weeks (one, two and seven weeks of prior evolution), the tested strain pairs were not the same continued evolutionary lineages. Instead, independent sets of strain pairs were used for each of the three weeks. All data analyses were performed in R [31].

3. Results

(a). Abundance through time

In all replicates of one week of evolutionary history, the first strain rose to high density by day 1, and the second strain remained at a lower abundance for the majority of the experiment (figure 1a). However, by the end of the experiment (day 10), the second strain of sympatrically evolved pairs reached higher density than the first strain, whereas in allopatrically evolved pairs, the second strain remained at a lower density (figure 1a). Surprisingly, in sympatrically evolved pairs, all second-arriving strains rose to higher density by day 10 of the experiment.

Figure 1.

Abundance dynamics in assembled communities. Points represent the abundances of first- (closed circles) and second- (open circles) arriving strains for sympatrically evolved (red, left) and allopatrically evolved (blue, right) pairs of strains. Each row of panels shows results for strain pairs that had the following duration of sympatric or allopatric evolution: (a) one week, (b) two weeks or (c) seven weeks.

For each time point, the difference in abundance between the first- and second-arriving strains provides a metric of establishment success of the second immigrant (figure 2). Higher positive values indicate that the first species remains dominant in the community, while smaller (or negative) values indicate more successful establishment by the second immigrant. This metric was higher for allopatrically evolved pairs than for sympatrically evolved pairs at day 10 (t-test, p = 0.0002), confirming that late-arriving immigrants attained higher relative abundance when they had evolved sympatrically with the first immigrant (figure 2a).

Figure 2.

Differences between the log-transformed abundances of first- and second-arriving strains. Points represent the mean differences between first- and second-arriving strain abundances for sympatrically evolved (red) and allopatrically evolved (blue) pairs of strains. Duration of sympatric or allopatric evolution: (a) one week, (b) two weeks or (c) seven weeks.

For all replicates of both two and seven weeks of prior evolution, the first-arriving strain rose to high density, and the second-arriving strain remained at relatively low density (figure 1b,c). For all time points, we found no significant difference in the abundances of the first and second immigrants for sympatrically evolved and allopatrically evolved pairs (t-tests, p > 0.05; figure 2b,c). The week two replicates showed an apparent trend towards reduced priority effects at day 10 in sympatrically evolved pairs, but this trend was not statistically significant (t-test, p > 0.05; figures 1b and 2b).

Analyses of GLMMs suggest differences in abundance between first- and second-arriving strains were significantly predicted by both Day and Week. In models with interactions among Day, Week and Evolutionary Treatment, the Evolutionary Treatment alone was not a significant predictor, but the interactions between Day and Week and between Day and Evolutionary Treatment were significant (electronic supplementary material, tables S1–S3). These GLMM analyses corroborate the observation that evolutionary history influenced abundances, but that this effect was evident only at late time points during community assembly and only when the strains had undergone a short duration of sympatric evolution (figure 2).

(b). Strength of priority effects

To compare the strength of priority effects between the two evolutionary history treatments, we calculated the difference in P_ij between allopatrically evolved and sympatrically evolved pairs for each given strain i (i.e. ΔP = [P_{{i,allopatrically evolved j}} − P_{{i,sympatrically evolved k}}]) for one, two and seven weeks of evolutionary history. Subscripts j and k correspond to the second strain in the experiment from the allopatric or sympatric evolution treatment, respectively. After one week, mean ΔP was marginally lower than zero (two-tailed t-test, p = 0.065), indicating that priority effects in allopatrically evolved pairs tended to be more negative than in sympatrically evolved pairs (figure 3a). The general linear mixed model analysis did not reveal significant effects of either Week or Evolutionary Treatment (electronic supplementary material, table S4). However, after two and seven weeks of evolutionary history, ΔP did not differ significantly from zero (t-tests, p > 0.1; figures 3b,c). The GLMM did not show significant treatment effects for Week or Evolutionary Treatment, suggesting that the lack of treatment effect at weeks 2 and 7 overwhelmed the signal that we found for treatment differences in the strength of priority effects at week 1.

Figure 3.

Comparison of the strength of priority effects between allopatrically and sympatrically evolved pairs. Each data point represents the priority effect experienced by focal strain i when introduced after a sympatrically evolved strain j and after an allopatrically evolved strain k. Diagonal lines represent the case where the strength of priority effects in sympatrically and allopatrically evolved pairs were the same (ΔP = 0). Points falling below (ΔP > 0) or above (ΔP < 0) the diagonal line represent stronger priority effects in allopatrically and sympatrically evolved pairs, respectively. Duration of sympatric or allopatric evolution: (a) one week, (b) two weeks or (c) seven weeks.

(c). Phenotypic complementarity and priority effects

In the general linear mixed model analyses, we found that the best-fit model was one without interactions, revealing that Week was the most significant factor (electronic supplementary material, table S5). However, inclusion of the interaction between Evolutionary Treatment and Week did yield a well-fit model (electronic supplementary material, tables S5–S7).

With allopatrically and sympatrically evolved pairs for one week analysed together, inhibitory priority effects were weaker (i.e. P_ij was less negative) when the paired strains were more dissimilar (i.e. more complementary) in their biofilm characteristics (linear regression, p = 0.004). Analysed separately, this relationship was significant for both allopatrically evolved (linear regression, p = 0.03) and sympatrically evolved pairs (linear regression, p = 0.03) (figure 4a). After two weeks of shared evolutionary history, priority effects were weaker in more complementary pairs (linear regression, p = 0.0014) when sympatrically and allopatrically evolved pairs were analysed together. However, when analysed separately, only the relationship for allopatrically evolved pairs was significant (linear regression, p = 0.004) (figure 4b). No relationships (either combined or separately analysed) were significant between phenotypic complementarity and the strength of priority effects after seven weeks of evolutionary history (figure 4c). This observation that the evolutionary treatment effect was mitigated with longer periods of evolutionary history was supported by the GLMM showing a significant interaction between Distance and Week (electronic supplementary material, table S7).

Figure 4.

Relationship between phenotypic similarity (trait distance) and the strength of priority effects (P_ij). The x-axis is the Euclidean distance between strains i and j in mat phenotypes, as measured by webbing and thickness. The y-axis is the P_ij value for the corresponding strains i and j. Duration of sympatric or allopatric evolution: (a) one week, (b) two weeks or (c) seven weeks. For week 1 (a), the relationship was significant for both sympatrically and allopatrically evolved pairs (linear regression, p = 0.03). For week 2 (b), only allopatrically evolved pairs were significant (linear regression, p = 0.004). For week 7 (c), neither sympatrically nor allopatrically pairs showed a significant relationship.

4. Discussion

In this laboratory experiment, we initially found a pattern suggesting weaker priority effects in pairs of populations that had evolved sympatrically prior to immigration than in those that had evolved allopatrically (figure 3a). However, as the duration of evolutionary history increased, this weakening of priority effects by sympatric evolution disappeared (figure 3b,c). These results provide the first experimental evidence, to our knowledge, that short-term, but not long-term, history of sympatric evolution might weaken the extent of historical contingency in community assembly. Our data further suggest that, in our experiment, the strength of priority effects was in part determined by the amount of phenotypic complementarity in interacting populations (figure 4), a finding consistent with the functional guild hypothesis demonstrated in grassland plants [32] and the niche component hypothesis proposed with nectar yeasts [29]. Our results are also consistent with those of Brockhurst et al. [11], who showed that evolution of character displacement in P. fluorescens biofilms allowed for higher productivity and invasion resistance. The novel aspect of our study, however, is that we examined the effect of the duration of evolutionary history, which revealed the time-sensitive influence of the history of sympatric evolution (figures 2 and 3).

Why was the effect of phenotypic complementarity lost over longer evolutionary periods? There are two potential explanations. First, two phases of trait evolution in competing populations may have resulted from the following scenario. Initially, sympatrically evolving pairs may have quickly achieved niche partitioning [14,22]. Subsequently, a longer period of sympatric evolution may have gradually led to increased competitive similarity among populations. Beyond theoretical support [15,17,33–36], empirical evidence for gradual convergence has been found in several taxonomic groups, including cichlid fishes [37], protozoans [38] and ovenbirds [18]. Second, both sympatrically and allopatrically evolving pairs may have initially diverged in traits, but sympatrically evolving pairs may have converged at a faster rate than allopatric pairs. More fine-grained temporal sampling at earlier time points, and longer periods of evolutionary history would help to distinguish the two alternative explanations. Either way, findings from our experiment suggest that slow trait convergence in sympatrically evolved species could influence how species interact ecologically when they colonize a new habitat.

One limitation of this study concerns how biofilm phenotypes were characterized. As in many other efforts to relate phenotypic variation to biotic interactions, it is unclear in this study how relevant our phenotypic characterization may be to explaining the strength of priority effects. For example, given the results for priority effects, one would expect sympatrically evolved pairs to be phenotypically more dissimilar than allopatrically evolved pairs in the week one replicates, but we did not observe this difference (t-test, p = 0.6, figure 4). Therefore, phenotypic similarity, as estimated in this study, does not really explain the difference that we observed in the strength of priority effects between sympatrically and allopatrically evolved pairs. Likewise, another expectation would be for allopatrically evolved pairs of week 2 and week 7 replicates to be more phenotypically similar to each other than sympatrically evolved pairs of week 1 replicates, but this expectation was not observed, either (figure 4). One possible explanation for these apparent discrepancies between expected and observed patterns is that the biofilms developed originally through mutations affecting the type of adhesion that we did not measure [22,25]. The thickness and webbing phenotypes we measured may have been a result of subsequent mutations that were of secondary importance to determining interactions between sympatrically evolved pairs. It is striking that significant relationships were still detected between phenotypic distance and the strength of priority effects even with crude phenotypic measurements like ours (figure 4). However, a more thorough trait characterization may enable a more mechanistic explanation of the weakened priority effects in sympatrically evolved pairs than is possible with our data.

Two other aspects of our experiment point to future research directions. First, the abiotic environment for our evolutionary treatments was the same as the environment for our assembly experiment. If the environment for community assembly was different from the historical environment in which populations had evolved, prior sympatric evolution might not have the same effects that we found (see also [12]), an idea that we believe deserves experimental tests. Second, our experiment focused on immigration by pairs of populations as a simplest case to study the role of arrival order. In more diverse communities, sympatric evolution affecting species traits can involve not only competitive interactions, but also other types of antagonistic and mutualistic interactions (e.g. predator–prey, host–parasite, plant–herbivore, plant–pollinator, habitat modification). Sympatric evolution involving these other ecological interaction types should also be studied to better understand evolutionary influences on priority effects.

Although results of microbial experiments like ours should not be uncritically extrapolated to other systems, the basic idea we have considered here may be broadly applicable, particularly for communities that assemble on islands, lakes and other isolated habitat patches with discrete boundaries. As these communities assemble by immigration and in situ diversification [39–41], deterministic sets of ecomorphs are expected to emerge through ecological niche filling [42]. However, this expectation is not always met, as shown in Anolis lizards [43], cichlid fishes [44] and Hawaiian spiders [45]. Our results suggest that variation in the evolutionary history of immigrants relative to one another may sometimes explain why we see a pattern of more deterministic assembly in some cases and more historically contingent assembly in others.

Data accessibility

Data are available in the Dryad Digital Repository at https://doi.org/10.5061/dryad.63t3d [46].

Authors' contributions

P.C.Z. and T.F. conceived and designed the study; P.C.Z. performed the experiments and analyses; and P.C.Z. and T.F. wrote the manuscript.

Competing interests

We have no competing interests.

Funding

This work was funded by an NSF Postdoctoral Research Fellowship (P.C.Z.) and a Stanford University Terman Fellowship (T.F.).