Disturbance, productivity, and natural enemies are significant determinants of the evolution of diversity, but their interactive effect remains unresolved. We develop a simple, qualitative model assuming trade-offs between growth rate, competitive ability and parasite resistance, to address the interactive effects of these variables on the evolution of host diversity. Consistent with previous studies our model predicts maximum diversity at intermediate levels of disturbance and productivity in the absence of parasitism. However, parasites break down these unimodal diversity relationships with productivity and disturbance, as selection for parasite resistance reduces the importance of growth rate-competitive ability trade-offs. We tested these predictions using the bacterium Pseudomonas fluorescens, which undergoes an adaptive radiation into spatial niche specialists under laboratory conditions. This is the first study of adaptive radiation in response to experimental manipulation of the three-way interaction between productivity, disturbance, and natural enemies. As hypothesized, unimodal diversity relationships with disturbance and productivity were weakened or disappeared in the presence of parasitic phages. This was the result of phages increasing diversity at environmental extremes, by imposing selection for phage-resistant variants, but decreasing diversity in less stressful environments, probably through reductions in resource competition. Phages had a net effect of increasing host diversity. Parasites and other natural enemies are therefore likely to have a large effect in mitigating the influence of other environmental variables on the evolution and maintenance of diversity.

Explaining patterns of genetic and ecological diversity remains one of the biggest challenges in evolutionary ecology. Both theory and experiments suggest parasites (and other natural enemies) are likely to play an important role in the evolution of diversity of their host species, potentially causing both increases and decreases in levels of diversity (reviewed in Holt and Lawton 1994; Chase et al. 2002; Vamosi 2005). Here we develop a generalized model and carry out experimental studies to determine how two key ecological variables, disturbances (periodic mass mortality events) and productivity, interact with parasites to affect the evolution of host diversity. Not only is this the first thorough examination of the interactive effect of natural enemies and productivity on adaptive radiations, but our model and experiments also extend to the factorial interaction of these variables with environmental disturbance.

Theoretical work predicts that diversity should be maximized at intermediate levels of both disturbance (intensity and frequency) and productivity (Connell 1978; Tilman 1982; Kassen et al. 2000; Kondoh 2001; Worm et al. 2002; Shea et al. 2004). Explanations for unimodal diversity relationships in these models rely on the existence of fitness trade-offs between ecological strategies (but see Kadmon and Benjamini 2006). For example, trade-offs may exist between colonization ability (or growth rate) and competitive ability (e.g., Tilman 1994), or between the abilities to use different resources (e.g., Tilman 1977). Ultimately, intermediate levels of disturbance or productivity maximize the likelihood of species coexistence, and hence diversity, by increasing intra- relative to interspecific competition (stabilizing mechanisms), and by equalizing fitness between species (equalizing mechanisms) (Chesson 2000). These mechanisms result in negative frequency-dependent selection (Clarke 1979), such that a rare species has a fitness advantage relative to a common species. Theoretical work investigating the interactive effect of disturbance and productivity also predicts unimodal relationships, although maximal diversity for one variable may be dependent on the magnitude of the other ecological variable (Kondoh 2001; Worm et al. 2002).

A large body of empirical work supports these predictions that species diversity is maximized at intermediate disturbance frequencies (Connell 1978; Sousa 1979; Petraitis et al. 1989; Floder and Sommer 1999; Buckling et al. 2000; Mackey and Currie 2001) and intermediate productivity (Tilman 1982; Rosenzweig 1995; Bohannan and Lenski 2000a; Kassen et al. 2000; Mittelbach et al. 2001; Kassen et al. 2004; Hall and Colegrave 2007). Furthermore, studies addressing the interactive effects of productivity and natural enemies are broadly consistent with the prediction that greater levels of disturbance or predation are required to maximize diversity with increasing productivity (Worm et al. 2002). However, unimodal relationships are by no means ubiquitous, with studies in natural communities reporting positive, negative, or no relationship between diversity and ecological variables (Rosenzweig 1995; Mackey and Currie 2001; Mittelbach et al. 2001). There are a number of possible explanations for this inconsistency, some of which we outline below. First, with correlational studies of natural populations, other environmental factors that affect diversity may covary with disturbance and productivity. For example, patterns of diversity could vary at differing spatial (local vs. regional) and temporal scales (Mittelbach et al. 2001). Second, some studies may only detect the increasing or decreasing phase of a unimodal relationship. Third, the effect of productivity and disturbance on diversity may be contingent on other ecological variables, in particular the species-specific, population-dynamic, and coevolutionary impact of mortality rates imposed by parasites and predators (Bonsall and Holt 2003; Morgan and Buckling 2004). We consider that the constant effect of productivity, the instantaneous effect of mass-mortality disturbances, and the perturbations imposed by interactions with natural enemies, can be fundamentally different kinds of environmental pressures on host diversity and diversification.

The impact of generalist parasites on the evolution and maintenance of host diversity is complex, with evidence of both diversity-enhancing and -inhibiting effects. Parasites are likely to affect the evolution of within-population host diversity via three main processes. Parasites will reduce host diversity if there is convergent (directional) host evolution to resist parasites. Second, parasites may promote diversification via competition for enemy-free space, if there exist alternative resistance strategies or trade-offs between competitive ability or growth rate and resistance to parasites (Holt and Lawton 1994; Abrams 2000; Chase et al. 2002; Vamosi 2005). Third, parasites can also reduce host diversity via host density reductions: genotypes or species can be lost stochastically and diversifying selection driven by competition for resources will be weakened (Buckling and Rainey 2002b). Note that this latter effect of parasites on host diversity does not result from direct parasite-imposed selection.

Here, we are interested in addressing the interactive effect of productivity, disturbance, and parasitism (or more generally, shared enemies) on the evolution of host diversity. There is unfortunately no explicit theory (either ecological or evolutionary) that addresses this issue. We are aware of one empirical ecological study that addressed the interactive effect of these variables (substituting parasites with predators) on protozoan and rotifer diversity in artificial treeholes, but despite complex interactions, no specific conclusions were drawn (Kneitel and Chase 2004). By using experimental populations of bacteria in microcosms, it is possible to extend this work by considering the evolution (and ecological maintenance) of diversity under very precisely controlled conditions, and across a much wider range of disturbance and productivity regimes. In so doing, it should be possible to determine clearer patterns of diversity, and allow us to identify the likely mechanisms responsible for these patterns.

Experimental populations of Pseudomonas fluorescens, a common, plant-colonizing bacterium provide a useful system to study the evolution of diversity. When propagated in spatially heterogeneous environments (a static, media-filled tube), P. fluorescens rapidly diversifies into morphologically (Rainey and Travisano 1998) and genotypically (Spiers et al. 2002) distinct spatial niche specialists. Wrinkly spreader (WS) morphs create mats at the air-broth interface, smooth (SM) colony morphs occupy the broth phase, whereas fuzzy spreaders (FS) occupy the base of the tube, and several, ecologically distinct variants occur within these categories (Brockhurst et al. 2006). Given the asexual nature of reproduction each variant within each morph is comparable to a species. The diversity of these communities is stably maintained through negative frequency-dependent selection in which rare morphs have a competitive advantage (Clarke 1979).

This system has been used to investigate both univariate and bivariate effects of disturbance and productivity on the evolution of diversity. Diversity is maximized at intermediate levels of disturbance frequency and productivity (Buckling et al. 2000; Kassen et al. 2000), with no significant interaction between these environmental variables (Kassen et al. 2004). The general mechanism explaining this pattern is that fitness differences between the different spatial niche specialists are reduced under intermediate disturbance frequencies and productivities (Buckling et al. 2000; Kassen et al. 2000). However, the dynamics of P. fluorescens diversification in these contexts is not simply driven by changes in the quality of the different spatial niches. For example, mat formation by WS requires the cooperative production of a polymer; and this is also likely to be maximized at intermediate disturbance (Brockhurst et al. 2007) and high productivity (Travisano and Rainey 2000).

The effect of parasitism on P. fluorescens diversification has been investigated by the addition of parasitic lytic phages, which invade, replicate, and then lyse susceptible bacteria (Buckling and Rainey 2002b). Their impact on within-population diversity is contingent on other ecological variables. Phages have been shown to increase diversity in spatially homogeneous environments, but decrease diversity in heterogeneous environments (Buckling and Rainey 2002a; Brockhurst et al. 2004). Furthermore, the addition of phages flattened the relationship between diversity and disturbance (Morgan and Buckling 2004). In both cases, phages increased diversity when it is otherwise low, by selecting for intrinsically resistant FS morphs (and very occasionally resistant mucoid morphs); and decreased diversity when it is otherwise high, by reducing density and hence reducing selection to occupy novel spatial niches. It therefore seems likely that phages will reduce the variance in diversity across gradients of both productivity and disturbances.

Given the complexity of the system, i.e. the simultaneous operation of spatial niche specialization (Rainey and Travisano 1998), cooperation and cheating (Rainey and Rainey 2003), and the possibility of negative interactions caused by toxin production (Rainey and Travisano 1998; Hodgson et al. 2002), developing realistic mathematical models to predict the interactive effect of productivity, disturbance and phages on the evolution of P. fluorescens diversity is simply not possible. Instead, we considered very general life-history differences between the different morph classes (WS, SM, and FS): WS tends to be the dominant morphotype when productivity is high or when total population sizes are greatest (Rainey and Travisano 1998; Travisano and Rainey 2000; Brockhurst et al. 2007), suggesting that they are better competitors in productive environments; SM have the highest intrinsic growth rate (Rainey and Rainey 2003); and FS are resistant to the ancestral phage (and to predators; Gallet et al. 2007). Taken together, these trade-offs between growth rate, competition, and phage resistance suggest that SM will gain advantage in disturbed, unproductive environments, WS will dominate in undisturbed, productive environments, and FS will gain advantage in the presence of phages. We therefore constructed simple qualitative models assuming strategies have trade-offs between being a good competitor (WS), a fast grower (SM) and resistant to phages (FS). We then tested our simple verbal and mathematical models by evolving initially isogenic populations of P. fluorescens under a spectrum of productivity and disturbance regimes in the presence or absence of a lytic bacteriophage, in a fully factorial design, over the course of 16 days.

Materials and Methods


To formalize our verbal arguments given above, in this section we present a resource–consumer–natural enemy model to explore the effects of disturbance and productivity on the evolution of diversity. From the outset, our model does not attempt to represent the full complexities of the bacteria–phage system, however we incorporate essential trade-offs in resource utilization and/or resistance to phages to capture the critical aspects of the evolutionary biology of the system. In particular, we include competition-growth and (when phages are present) competition-resistance trade-offs to explore how changes in productivity and disturbance affect the invasion and coexistence of different bacteria strategies in the presence or absence of phages.

To explore the effects of productivity on invasion (and hence potential for adaptive radiation) we incorporate trade-offs into resource (R)–bacteria (Bi)–phage (P) interaction and explore the conditions under which two alternative bacterial strategies (i and j) can invade and coexist. The single resource—bacteria–phage interaction can be described by


Here, in equation (1), resource renewal rate is constant (λ) so that as λ increases the productivity of the system increases (Bonsall and Holt 2003). The following description of subscripted parameters refer to the specific example of strategy i: parameters for strategy j are simple conversions. Resources are consumed by bacteria at rate αi · R(t) and unconsumed resources decay at rate inline image

In equation (2), resource consumption i · Bi(t) · R(t)) affects the ith bacterial intrinsic rate of increase (ri). However, the underlying rate of population increase is also a function of competitive ability such that a trade-off is established between growth rate and competitive ability


where aij is the strength of competition of the jth strategy on the ith strategy. Overall, bacteria strain population growth rate is affected by both intraspecific (logistic) and interspecific competition inline image; where Ki is the carrying capacity of the ith strategy.

In the presence of phages, bacteria are infected at rate, gi) · P(t). However, individual bacteria strains may be resistant to attack and resistance may also be traded off with competitive ability


where βi is the attack coefficient by phages on the ith bacteria strategy and ɛ is the strength of the effect of the trade-off on resistance to predation. Small values of ɛ implies a weak trade-off whereas high ɛ leads to strong interaction between competitive ability and resistance such that highly competitive strategies are poorly defended against phages. Finally, bacteria strains may die due to other (density-independent) processes at rate inline image.

When phages are present (eq. 3), they increase by attacking bacteria (at rate gi) · Bi (t) +gj) · Bj (t)), which yields γ new phages. Phages decrease in numerical abundance through natural mortality at rate μP.

To explore the effects of disturbance intensity (under different productivity regimes), we make use of an approach (the linear chain trick; MacDonald 1978; Mangel and Bonsall 2004) that distributes the effects of resource consumption and population growth rate on population change. In particular, we replace equation (2) with


where x0(t) and y0(t) are auxiliary state variables determined from:




Here, ηz and νz are parameters that describe the overall rate at which bacterial growth rate (f(ri) · [αi · Bi (t) · R(t)]) or death rate (inline image) cascade through a chain of linked equations, respectively: these are essentially time-delayed processes that allow disturbance to be described in terms of a set of ordinary differential equations. By introducing p+ 1 (or q+ 1) equations in which each equation links two successive members of the chain, a distribution of the effects of disturbance can be built up. Small values of ηz (or νz) imply highly disturbed regimes in which the response of the population to changes in population growth (or mortality) is long. Larger values of ηz (or νz) give rise to a quicker response of the population to disturbance and as ηz→∞ and νz→∞ we recover the standard differential equation for bacteria population change (eq. 2). This approach provides a straightforward way in which to describe differential responses to disturbance intensity (by stepping species dynamics through a chain of ordinary differential equations) that essentially mimics the effects of a distributed time delay acting on a species' numerical response.

The evolutionary analysis of the model proceeds by determining the fitness (invasion) criteria inline image for an initially rare strategy with trade-offs in competition growth (eq. 4), and when phages are present, competition resistance to predation (eq. 5). Mutual invasibility of strategies (into equilibrial resident assemblages of resource, bacteria and, when present, phages) implies that novel lineages (with different competitive abilities, growth rates, and resistance levels) can invade and coexist (Bonsall and Mangel 2004; Mangel et al. 2007). When sympatric coexistence is possible, it implies an adaptive radiation should occur as long as there are no constraints on differentiation. Boundaries of coexistence were determined using a Newton–Raphson root-finding algorithm, and relative areas of coexistence (a measure of the impact of disturbance, productivity and/or phages) were obtained by simply determining the area under each invasion curve and dividing by the coexistence area of a reference environment.


Productivity and disturbance gradients

Populations were initiated with inoculations of isogenic ancestral P. fluorescens (SBW25), either with or without the SBW25 bacteriophage Φ2 (Buckling and Rainey 2002a) (approximately 105 particles) into static microcosms at 28°C under four different productivity levels and five disturbance regimes. Based on prior work (Kassen et al. 2000), a productivity gradient was generated using a standard concentration of growth medium, ×5 concentration, and serial dilution to ×0.1 and ×0.01 to give a range of four nutrient concentrations. In line with previous studies (Buckling et al. 2000), global disturbances were induced by transferring 0.1% (6μl aliquot), from a vortexed microcosm to obtain a mixed population, into 6 mL of fresh medium every day, every second, every fourth, every eighth day, and not at all over the course of 16 days. This factorial experiment (presence/absence of phages × five disturbance frequencies × four productivities) was fully repeated in six blocks. At the end of each block, cultures were plated and frequencies of distinct colony morphologies (i.e., all recognizable variants across all three morphs) were scored for approximately 100 colonies. Species richness was calculated as the number of distinct colony-morphology variants observed in these samples. Diversity was then estimated as Simpson's index of diversity (adapted from Simpson 1949), 1 −λ= 1 −∑pi2, where pi is the proportion of each morph in a culture sample. Occasionally no viable colonies were obtained at ×5 standard nutrient concentration across productivity and disturbance gradients, despite bacterial growth, so data from these cultures were ignored in statistical analyses.

Statistical analysis of diversity relationships, densities, and morph frequencies

Indices of diversity were analyzed in a general linear model that included the covariates log2-transformed frequencies of disturbance (linear and quadratic), log10-transformed productivities (linear and quadratic), and the fixed factor presence/absence of phages. Block was included as a random effect. The full model, including all relevant interactions between explanatory variables, was simplified using the criterion of nonsignificance of terms, until only significant terms remained. Species richnesses were square-root transformed and bacterial densities were log10-transformed to improve normality of residuals. The statistical analysis of each of these response variables (species richness, bacterial densities) followed that outlined for the analysis of the diversity indices. Proportions of bacterial colonies represented by SM, FS or WS morphs were analyzed separately using generalized linear models with a quasi-binomial error structure to account for overdispersion. These models were simplified using F-tests. Standard model-checks verified normality of standardized residuals and homogeneity of variance (Crawley 2002).



Assuming trade-offs between growth rate and competitive ability, the range of competitive abilities that satisfied the invasion criteria in the absence of a shared enemy was greatest for intermediate productivity and disturbance intensity (Fig. 1A). This result is consistent with previous theoretical studies (Connell 1978; Tilman 1982). High productivity and low disturbance strengthen the effect of interspecific competition, and hence allow good competitors to dominate; whereas fast growers dominate under low productivity and high disturbance. Both types are most likely to coexist under intermediate productivity and disturbance. We then explored the impact of a shared natural enemy on community dynamics, additionally incorporating simple trade-offs between competitive ability and resistance to parasitism. This broke down the clear, unimodal relationship between disturbance intensity, environmental productivity, and the range of competition parameter space that allows coexistence of hosts (Fig. 1B). This effect of predation altering the effects of environmental gradients is not dependent on the precise form of the trade-off as alternative relationships between competitive ability and resistance to phages give qualitatively similar results.

Figure 1.

The magnitude of competition-parameter-space that allows mutual invasion by new species or strains under different conditions of disturbance intensity and environmental productivity, in (A) the absence and (B) the presence of a shared natural enemy. In the absence of natural enemies, there is a greater range of novel phenotypes that can invade when the environment has intermediate values of both disturbance intensity (triangles, solid line) and productivity, compared to all other combinations of disturbance (low intensity, squares with dot-dash lines; high intensity, diamonds with dashed lines) and productivity. When natural enemies are present, this “intermediate” effect breaks down entirely, and the potential for coexistence declines with increasing productivity in very disturbed environments.

This effect of phages occurs as the low and intermediate productivity and high and intermediate disturbance allowed coexistence of resistant types and fast growers; and high productivity and low disturbance allowed coexistence of resistant types and good competitors. As such, the range of phenotypes that could coexist did not vary much over the productivity and disturbance gradients except under conditions of intense disturbance, and the introduction of phages leads to fitness-equalizing mechanisms across the range of productivities and disturbance regimes.


Consistent with previous work (Buckling et al. 2000; Kassen et al. 2000; Kassen et al. 2004; Morgan and Buckling 2004), unimodal relationships, peaking at intermediate levels (disturbance frequency of 4 days and ×1 productivity), were found between diversity and disturbance and productivity in the absence of phages (tests of nonzero quadratic terms in multiple regression analyses: productivity quadratic slope =−0.129, t212= 7.212, P < 0.001, Fig. 2; disturbance frequency quadratic slope =−0.049, t212= 4.750, P < 0.001, Fig. 3). These unimodal relationships were also true of the relationship between species richness, disturbance and productivity in the absence of phages (productivity quadratic slope =−0.254, t212= 8.618, P < 0.001; disturbance frequency quadratic slope =−0.107, t212= 6.314, P < 0.001).

Figure 2.

Diversity of bacterial communities evolved under different conditions of environmental productivity (measured as the log10-transformed concentration of nutrients in the bacterial growth medium, relative to standard conditions), for different levels of disturbance frequency (number of times that a 99.9% mortality event was imposed on the community over a 16-day adaptive radiation: (A) once, (B) twice, (c) four, (d) eight and (e) sixteen times) in the absence (open circles, dashed lines) and presence (filled circles, solid lines) of phages. Circles show mean Simpsons Index of Diversity ±1 SE; lines show the fitted values from minimal adequate general linear models, ignoring the effect of experimental block. In the absence of phages, diversity showed a humped relationship with productivity: this relationship was significantly weakened by the presence of phage. Phages decreased diversity in intermediate environments and increased it in extreme environments. Across all environments, phages increased mean and decreased variance in diversity.

Figure 3.

Diversity of bacterial communities evolved under different frequencies of disturbance (x-axis), and conditions of environmental productivity ((A) ×0.01 standard concentration of growth medium, (B) ×0.1, (C) ×1 and (D) ×5) in the absence (open circles, dashed lines) and presence (filled circles, solid lines) of phages. Circles show mean Simpsons Index of Diversity ± 1 SE; lines show the fitted values from minimal adequate general linear models, ignoring the effect of experimental block. Curvature of the diversity–disturbance relationship was nonsignificant in the presence of phages.

Unimodal diversity relationships with both productivity and disturbance were significantly altered by the presence of phages (interaction between phage presence/absence and quadratic term for productivity, F1,212= 26.99, P < 0.001, Fig. 2; interaction between phage presence/absence and quadratic term for disturbance frequency, F1,212= 5.96, P= 0.015, Fig. 3). In the presence of phages, the unimodal relationship between diversity and productivity was weakened (quadratic slope reduced to −0.048, t212= 2.72, P= 0.007, Fig. 2) and was reduced to nonsignificance between diversity and disturbance (quadratic slope reduced to −0.012, t212= 1.12, P= 0.264, Fig. 3). Overall, a sign test on paired differences in diversity between phage and no phage treatments provided statistical support for the prediction that phages would decrease diversity when otherwise high, and increase it when otherwise low (P < 0.0001; 84 pairs in hypothesized direction, 14 in opposite direction, 6 showed no difference; direction based on whether phage-free diversity was higher or lower than the median phage-free diversity). This effect caused the total variance in diversity across all replicates without phages to be significantly higher than for those with phages (F116,108= 1.63, P= 0.011).

Consistent with the diversity results, presence of phages significantly altered the unimodal relationship between species richness, productivity, and disturbance frequency (interaction between phage presence/absence and quadratic term for disturbance frequency, F1,212= 12.27, P < 0.001; interaction between phage presence/absence and quadratic term for productivity, F1,212= 14.13, P < 0.001). These interactions represented a significant weakening of the quadratic relationships in the presence of phage (quadratic slope for productivity reduced from −0.254 to −0.186, t212= 4.87, P < 0.001; quadratic slope for disturbance reduced from −0.107 to nonsignificant slope of −0.024, t212= 1.30, P= 0.195). Presence of phages caused species richness to decrease when otherwise high, and to increase when otherwise low (sign test as for diversity results, 61 comparisons in hypothesized direction, 17 in opposite direction, 32 showed no difference; P < 0.001 if zero differences split equally between null and alternative hypotheses; P≪ 0.001 if zero differences ignored).

Overall, phages promoted diversity across the productivity and disturbance frequency gradients. Mean diversity was significantly higher in the presence versus the absence of phages (mean ± 1 SE Simpsons index of diversity: without phages: 0.284 ± 0.021; with phages: 0.391 ± 0.022; test of main effect of phage presence over and above unimodal effects of disturbance and productivity, F1,215= 14.24; P < 0.001). The same qualitative pattern was detected for species richness: phages caused an increased mean (F1,215= 4.37, P= 0.038, having accounted for unimodal relationship with environmental variables) and decreased variance in species richness (F119,110= 1.401, P= 0.037).

Phages did not have a net effect of reducing total bacterial density (test of main effect of phage on log10-transformed density, having accounted for unimodal response to productivity and disturbance; F1,215= 1.05, P= 0.306, Fig. 4). However, a sign test on paired differences in density between phage and no phage treatments within each disturbance, productivity, and block combination revealed that there was a tendency for phage to reduce density when density was otherwise “high” (higher than the median) in the absence of phages, and increase, or cause no change in density when it was otherwise low in the absence of phages (P < 0.0001; 83 pairs in hypothesized direction, 21 in opposite direction) (Fig. 4).

Figure 4.

Density, measured as number of colony forming units per milliliter, of radiated communities under varying conditions of (A) environmental productivity and (B) frequency of disturbance, in the absence (white columns) and presence (gray columns) of phages. Columns show mean ±1 SE calculated from raw data. Phages decreased bacterial densities in intermediate environments, but increased them in extreme environments. Across all environments there was no detectable net impact of phages.

We also considered changes in frequencies of the principal morphotypes. Irrespective of phage presence/absence, SM frequency was highest at environmental extremes and low in intermediate environments (positive quadratic relationship with disturbance frequency, F1,216= 23.71, P < 0.001; positive quadratic relationship with productivity, F1,216= 39.29, P < 0.001). Phages caused a significant reduction in the proportion of the community that retained the ancestral-like smooth morphotype (test of effect of phage presence on relative frequency of SM morphs, F1,216= 12.78, P < 0.001, Fig. 5A,B). This result implies that other, more resistant, morphotypes should be more frequent in communities with phage present. Consistent with this, phages dramatically increased the frequency of intrinsically resistant FS morphs (test of phage effect on frequencies of FS in radiated communities: F1,214= 68.62, P < 0.001, Fig. 5C,D). Furthermore, the frequency of FS was not related to environmental gradients (effects of disturbance frequency: linear term F1,214= 2.00, P= 0.159; quadratic term F1,214= 2.14, P= 0.145, Fig. 5C. Effects of productivity: linear term F1,214= 1.02, P= 0.313; quadratic term F1,214= 0.28, P= 0.599, Fig. 5D). Frequencies of WS showed a unimodal relationship with disturbance frequency (negative quadratic slope F1,214= 30.55, P < 0.001, Fig. 5E) and productivity (negative quadratic slope F1,214= 54.25, P < 0.001, Fig. 5F), but were not affected by the presence of phages (F1,214= 0.0125, P= 0.911, Fig. 5E,F).

Figure 5.

Proportion of bacterial colonies, from a sample of each radiated community, that were (A,B) SM morph, (C,D) FS morph, and (E,F) WS morph. Proportions graphed against (A,C,E) frequency of disturbance events over 16 days of radiation, and (B,D,F) environmental productivity relative to standard growth medium, in the absence (white columns) and presence (gray columns) of phage. Columns are mean proportions ±1 SE calculated from raw data. Proportions of SM morphs were reduced in intermediate environments; resistant FS morphs were more frequent in the presence of phage; WS frequency was highest in intermediate environments and was not influenced by phage presence.


We addressed the interactive effect of productivity, disturbance, and parasites on experimental adaptive radiations of bacterial populations. Both our mathematical model, and our verbal arguments based on previous empirical work, predict that parasites should reduce or break down the unimodal relationships between diversity and productivity and disturbance frequency. Our experimental tests using bacteria and parasitic phages supported this prediction, and also resulted in phages causing a net increase in bacterial diversity. Following a fixed period of adaptive radiation of bacterial populations, both species richness (the number of evolved colony morphology types) and diversity (the evenness of distribution of individuals between morphology types) showed unimodal relationships with both disturbance frequency and productivity. These results replicate previous work on this experimental system (Buckling et al. 2000; Kassen et al. 2000; Kassen et al 2004). However, parasitic bacteriophages negated any detectable unimodality in the relationship between richness/diversity and disturbance frequency, and significantly weakened the steepness of the humped relationship between richness/diversity and productivity. This bidirectional effect of phages on host diversification, causing the attenuation or “softening” of unimodally related diversity relationships, is consistent with previous work looking at interactions between parasites and other environmental variables (spatial heterogeneity and disturbance) in this system (Brockhurst et al. 2004; Morgan and Buckling 2004).

Despite the gross simplification of our model with respect to the experimental system, the mechanisms responsible for diversity patterns in the model also appeared to be operating to some extent in the experiments. The model assumed trade-offs between competitive ability, growth rate, and phage resistance. In the absence of phages, diversity was maximized at intermediate disturbance and productivity because these conditions favored the coexistence of both fast growers and good competitors. Previous work demonstrates that SM are the fastest growing morphotypes (Rainey and Rainey 2003) and that WS show their maximal fitness under conditions that maximize bacterial population density (Rainey and Travisano 1998; Travisano and Rainey 2000; Brockhurst et al. 2007), suggesting they are good competitors. Consistent with the model, we found that WS and SM were more evenly represented under intermediate disturbance frequencies and productivity (Fig. 5). In apparent contrast to the model predictions, SM tended to dominate at all environmental extremes, instead of just high disturbance and low productivity (Fig. 5). Note, however, that in this experimental system, extremely high productivity and low disturbance create toxic environments for the bacteria (as reflected by the density reductions relative to intermediate disturbance and productivity), rather than allowing higher growth rates and population densities, respectively, as assumed in the model.

The mechanism resulting in parasite-mediated breakdown of unimodal relationships in the model stems from the trade-off between parasite resistance and competitive ability. This trade-off becomes important relative to the trade-off between competitive ability and growth rate in the presence of parasites. In qualitative terms, low and intermediate productivity and intermediate disturbance allowed coexistence of resistant types and fast growers; and high productivity and low disturbance allowed coexistence of resistant types and good competitors. The combination of parasites, high disturbance, and high productivity was detrimental to coexistence in the model, as these conditions favor dominance of fast growers in the absence of a trade-off between growth rate and resistance to parasites. Experimentally, this prediction was confirmed by the low diversity of highly productive and highly disturbed environments, where fast-growing smooth morphs dominated.

Overall, the model suggested qualitatively that variance in bacterial diversity across environmental gradients should be reduced in the presence relative to the absence of parasites. Our experimental results support this prediction. Helping to drive this reduction in variance, phage-resistant FS consistently increased in frequency in the presence of the phages across all levels of productivity and disturbance (Fig. 5).

Previous work has identified another potentially important mechanism of phage-mediated changes in diversity that was not investigated in our model. Phages reduce bacterial density, which can in turn weaken selection imposed on the bacteria to occupy novel spatial niches (Buckling and Rainey 2002b). Crucially, phage-mediated reductions in bacterial density are most likely when phage transmission is maximized under conditions that maximize bacterial density in the absence of phages (Morgan and Buckling 2004). As such, this diversity-reducing mechanism should be most pronounced at intermediate disturbance frequencies (Morgan and Buckling 2004) and productivities, further contributing to the flattening of the relationships between diversity and disturbance and productivity in the presence of phages. Despite phages not having a net effect on bacterial density, our data remain consistent with the operation of this mechanism to some extent, because phages reduced densities in high-density environments. Note that the apparently paradoxical effect of phages increasing bacterial density at extreme values of environmental parameters may result because phage slowed growth, preventing population crashes due to overexploitation of limited resources.

Taken together with previous studies of experimental adaptive radiations (Brockhurst et al. 2004; Morgan and Buckling 2004), these results suggest that phages tend to promote diversity in equivalent environments that supported lower diversity in the absence of phages, but decrease diversity when diversity was high in the absence of phages. Interestingly, data from purely ecological studies also suggest a similar pattern: enemies are more likely to increase diversity when diversity is low under enemy-free conditions or decrease diversity when it is otherwise high (Worm et al. 2002). However, it is important to emphasize that this pattern may simply represent a statistical artifact, in that there is a higher probability of randomly increasing a measure when it is low than when it is high, and vice versa.

As emphasized in the introduction, numerous other uncontrolled variables are likely to contribute to the observed patterns of diversity in the present study. A large amount of unexplained variation is likely to be because of coevolution between bacteria and phage. Bacteria and phage undergo multiple rounds of reciprocal evolution of resistance and infectivity, respectively, during the time scale of the current experiment (Buckling and Rainey 2002b). Variation between populations in the rates of coevolution and the timing of the appearance of particular resistance and infectivity mutations is likely to have had a major impact on bacterial diversity and density (Buckling and Rainey 2002b; Brockhurst et al. 2004). Note that phages are able to overcome the intrinsic resistance of FS, and hence FS are likely to have decreased in frequency if the experiment had been continued for longer periods of time. Furthermore, longer-term coevolution can purge diversity in this system because of directional selection for resistance (Buckling and Hodgson 2007).

Although the evidence for the effects of shared natural enemies such as parasites on host adaptive radiations remains limited (Schluter 2000), we have demonstrated here how the ability to resist parasites leads to equalization in fitness and promotes diversity. In contrast to conventional expectations under shared enemies, where in the absence of trade-offs diversity is expected to decline (Holt 1977; Holt and Lawton 1994; Bonsall and Hassell 1997), we have illustrated here that natural enemies are influential in driving adaptive radiations. These data contribute to a growing body of work that directly identifies parasites and other natural enemies as key contributing factors in driving host diversification (Bohannan and Lenski 2000b; Schluter 2000; Buckling and Rainey 2002b; Brockhurst et al. 2004, 2005; Morgan and Buckling 2004; Vamosi 2005; Nosil and Crespi 2006; Meyer and Kassen 2007). Furthermore, we have shown that the addition of a trophic level breaks down unimodal relationships between host diversification and environmental parameters. As such, variance in host diversity is decreased in the presence of a shared parasite across a range of environmental conditions. These fitness equalizing mechanisms and the evolutionary implications clearly warrant further experimental and theoretical exploration.

Associate Editor: J. Koella


T. Tregenza and D. Hosken commented on earlier drafts. This work was funded by grants from the Leverhulme Trust, European Social Fund and the Royal Society.