Evolution of coexistence in a crowded microplate well

Understanding what causes the emergence and maintenance of biological diversity has been a central aim in biology ever since the dawn of the discipline. The realization that diversity in microbial communities, such as those living in our guts and on our bodies, may affect our health and that of the agricultural systems on which we depend has generated substantial recent efforts to describe the extraordinary diversity of microbial communities. However, efforts to understand the ecological and evolutionary mechanisms responsible for generating and maintaining all this diversity have severely lagged behind those to describe it. This is unfortunate and unnecessary. Microbes offer a great opportunity to rigorously test hypotheses about these mechanisms in real time in controlled experiments in the laboratory. As Frenkel et al. (1) show in PNAS, such experiments may also reveal novel mechanisms of diversification and coexistence—in their case one where the avoidance of crowding plays a key role.

The origin and maintenance of biological diversity requires natural selection operating in heterogeneous environments (2, 3). In homogeneous environments with a single limiting resource equally accessible to all individuals, natural selection will remove all but the most superior competitor (4). However, in environments with temporal or spatial variation in the type or amount of limiting resource, fitness trade-offs between adaptations to different niches may lead to the stable maintenance of different genotypes through negative frequency-dependent selection (5).

An important lesson from laboratory evolution studies with microbes is that virtually all environments, even those designed to be homogeneous, such as chemostats, are heterogeneous at the scale relevant for microbes. For instance, temporal variation in resource availability occurs in serial transfer experiments in batch cultures with the opportunity for specialist adaptation (6), potential resources inaccessible at the start of the experiment may become available after evolutionary changes in the population (7), and additional niches may be “constructed” by the evolving populations themselves (8), for instance, due to the excretion of metabolites allowing the evolution of cross-feeding ecotypes (9–11). Perhaps most fundamentally, experimental methods to create homogeneous conditions, such as mixing and a constant nutrient input, cannot remove heterogeneity introduced by the spatial structure of the environment, such as oxygen and other physicochemical gradients, including vessel walls to which cells may attach.

A classic example of adaptive diversification driven by competition in a spatially structured environment is exemplified by the work of Paul Rainey and Michael Travisano (12). Propagating the bacterium Pseudomonas fluorescens in unshaken (spatially structured) microcosms, they observed rapid diversification of clonal populations into multiple niche-adapted types (Fig. 1A). Each of the derived types was found to be adapted to different niches as evidenced by frequency-dependent trade-offs. Interestingly, recent work has shown that the trade-offs are a consequence of different strategies for colonizing the air–liquid interface (13). Each strategy is vulnerable to the effects of gravity—cellular mats colonizing the air–liquid interface collapse after a period, allowing invasion of new types. Not only are the interactions among types frequency dependent, but they are also time-lagged. The net effect is maintenance of multiple types on a single limiting resource.

Fig. 1.

Coexistence of ecotypes with differential access to limiting resources evolved in two experimental microcosms. (A) Pseudomonas fluorescens microcosm where three ecotypes coexist while competing for access to growth-limiting oxygen (gray gradient) at the air–liquid interface via different strategies (12). Trade-offs between the success in positioning cells at the air–liquid interface and the probability of failure for each strategy cause time-lagged frequency-dependent interactions that maintain the dynamical coexistence of the three types. Arrows indicate which type invades which other type when rare. (B) Saccharomyces cerevisiae microcosm where two ecotypes live near the bottom of the microplate well, causing the depletion of an unknown limiting resource (gray gradient). One ecotype (B type or “bottom dweller”) ends up at the very bottom of the well; the other ecotype (A type or “adherent”) attaches also to the sides of the wall of the well (1). Adherent cells initially grow slower than bottom dwellers upon transfer to fresh medium, but later avoid the crowding suffered by bottom dwellers, when only cells in the top layer (dark brown) have access to resources and cells below that layer (light brown) cannot reproduce. Here, a trade-off between the ability to grow at low and high cell density stabilizes coexistence of the two types.

The mechanism of diversification discovered by Frenkel et al. (1) also crucially depends on spatial structure, but is subtly different from the mechanism identified by Rainey and Travisano. In previous work, Frenkel et al. (14) evolved populations consisting of mixtures of two fluorescently labeled haploid strains of the yeast Saccharomyces cerevisiae for ∼1,000 generations in unshaken 96-well microplates via serial transfer. In most populations, natural selection caused the loss of one of the two strains, but in 13 of the ∼1,000 populations, both strains coexisted for hundreds of generations. In their new study, Frenkel et al. (1) address the causes of this seemingly stable coexistence. They find that, in all cases, coexisting strains exhibit markedly different growth phenotypes, with one strain occupying the round bottom of the well (termed B type for “bottom dweller”) while the other strain disperses more broadly upon transfer and also adheres to the sides of the wall surrounding the well bottom (termed A type for “adherent”; Fig. 1B). When mixed at different initial frequencies, coevolved A and B types typically equilibrate at the frequency at which they stably coexist, confirming that negative frequency-dependent fitness interactions stabilize coexistence.

What mechanism underpins coexistence in this case? The different distributions of A and B types within a well suggest the following hypothesis: At the expense of a lower maximum growth rate, adherent types may have better access to nutrients later during the growth cycle when cell densities are high and crowding among bottom dwellers limits nutrient access. The authors provide multiple lines of support for this explanation. First, coevolved A and B types no longer coexist when cultures are shaken or grown in flat-bottom wells, confirming the key role of spatial structure. Second, the A type has a growth disadvantage relative to the B type at low cell density but a relative advantage at high cell density, consistent with the hypothesized trade-off that allows the dynamically stable coexistence of both types. Third, a simple mathematical model assuming a lower maximum growth rate of the A type (reflecting the fitness cost of adherence) and a transition to linear growth at high cell density for the B type (reflecting growth only of the surface layer of bottom-dwellers; Fig. 1B) describes the competitive dynamics rather well. Interestingly, the model predicts that coexistence can be broken at evolutionary time scales when the difference in maximum growth rates of the two types becomes too large, consistent with observations that bottom dwellers were only temporarily present in some populations.

The reported mechanism of coexistence is not fully understood at the molecular level. Sequencing three independently evolved adherent clones revealed mutations in three different biosynthesis genes of ergosterol (a component of fungal cell membranes). These mutations probably reduce ergosterol production, because deletion of one of the mutated genes and chemical inhibition of the ergosterol pathway in the ancestral strain yields the adherent phenotype. However, why reduced ergosterol production incurs a fitness cost and how it allows cells to attach to the polysterene wall of the well is unclear. The authors speculate that increased hydrophobicity of the cell surface

Given the general importance of processes such as adherence, crowding, and dispersal, the mechanism revealed by Frenkel et al. will likely be relevant outside the laboratory.

may play a role, because adding detergent eliminated the adherent phenotype.

Competitive benefits due to crowding avoidance via dispersal have been suggested previously for situations where within-lineage competition is stronger than competition between lineages (15, 16). This is generally the case when access to resources is density-dependent, such as in colonies of nonmotile microbes growing in structured environments (17). Motility mechanisms are one obvious way to disperse and escape the negative effects of crowding, and provide benefits even when motility is random and not directed toward nutrients (18). For nonmotile microbes, the means of dispersal are limited, but as Frenkel et al. (1) show, not impossible. Interestingly, the proximate mechanism of dispersal adopted by the adherent type of yeast is similar to that used by the wrinkly spreader type of P. fluorescens (12): In both examples, adherence is the means—either to the wall of the vessel (adherent type) or to other cells to form a self-supporting mat (wrinkly spreader). A more complex flagella-independent dispersal strategy was recently reported in Bacillus subtilis colonies growing on agar, where two cell types, one producing surfactin and the other producing matrix, cooperated to form bundles of tightly aligned cell chains that push themselves away from the colony (19).

Given the general importance of processes such as adherence, crowding, and dispersal (15), the mechanism revealed by Frenkel et al. (1) will likely be relevant outside the laboratory. There is a clear need for further studies that explicitly consider the role of spatial environmental structure in microbial ecology and evolution, preferably combining experiments with modeling, as in Frenkel et al.’s study, to test and quantify the mechanisms involved. Such studies will help to resolve the ecological causes of diversification. Taking advantage of novel technologies to manipulate and monitor interactions among individual cells, such as microfluidics and time-lapse microscopy, future studies are certain to reveal novel mechanisms by which microbes ascertain privileged access to resources at the spatial scales relevant in their natural environments.