Towards a fundamental theory of taxon transitions in microbial communities

Technological breakthroughs have fostered an appreciation for the microbiome and revolutionized our understanding of many parts of the tree of life. These discoveries have implications as far-ranging as one can imagine, from how we diagnose and treat ailments at the clinical bedside to questions regarding how life emerged and diversified throughout the biosphere. While the early study of the microbiota focused mostly on identifying its composition, recent years have explored the fundamental forces that shape the microbiota and how communities are structured (1). In a recent study published in PNAS, Long et al. develop a theoretical framework that can help to explain the transient dynamics of microbial communities, where the composition of the microbes in a community undergoes rapid change (2).

The study of the microbial systems can be broadly categorized around the following foci (Fig. 1): a) the assembly and emergence of community structure, b) the functional underpinnings of that structure, and c) the dynamic composition of community membership. This framing emphasizes that the greater mystery of the microbiota lives in answers to three related questions: What are the individual taxa that compose the microbiota, how do they function, and what forces govern the dynamics of the microbiota? This third pillar is an especially elusive one, as it deals with a dichotomy that has been challenging to reconcile: While the microbiota is nearly ubiquitous across the biosphere, it is constantly changing in the microbial constitution, which makes it a moving target for our investigations.

Fig. 1.

In this simplified framing, the microbiota can be partitioned into the following components: emergence and assembly, the mechanistic bases for its function, and the dynamic properties of the composition and structure of microbial communities (dashed line). Bidirectional arrows highlight that forces influencing one pillar likely operate on others. In PNAS, Long et al. focus their attention on theory pertaining to the dynamics underlying the composition of taxa in microbial communities.

Recent breakthroughs have added substantial rhyme and reason to how microbial communities are constructed. For example, many have focused on the metabolic functions that arise out of microbial interactions in communities (3, 4) or proposed principles of community assembly (5–7). Others have utilized theory on social interactions and related notions that underlie competition and coexistence to explain the composition of the microbiota (8, 9) and applied ideas from genetics detailing nonlinear interactions between genes and mutations (epistasis) to explain microbial community structure (10, 11). These findings have equipped those who think about this topic with the hope that laws underlie community composition. And still, large questions remain regarding how and why this composition changes. Not unlike evolutionary theory as applied to general problems of adaptation on “rugged” fitness landscapes (an analogy for how evolution must solve complex problems, where adaptive solutions rest on “adaptive peaks”) (12, 13), the study of microbial communities has embraced the notion that evolving systems functions in environmentally static settings or assemblages, but rather, in highly dynamic contexts.

Recently, scientists have studied the dynamics of the microbiota both empirically (14, 15), and theoretically (16). In PNAS, the authors seek an underlying theory that governs dynamism in microbial communities. Critically, the authors use both mathematical formalism and real-world data to validate their model. To accomplish this, they use—and partly reconcile—theoretical frameworks from physics and information theory that have been previously applied to different domains, including biology.

The authors invoke the structuralist approach, a framework that explains how the construction of complex biological ensembles is influenced by aspects of their environment. Its earliest uses were in genetic systems, where theoretical biologists used it to examine how phenotype arises from ensembles of genetic information, with implications for genotype-phenotype mapping, and evolvability (17). More recently, it has been used to describe other complex biological assembly problems, like species distributions in an ecosystem (18). The structuralist school suggests that qualitative changes in the composition of an ensemble arise from descriptive laws governing the biological actors in the system, and quantitative changes are driven by features of the environment (encoded as “environmental parameter space”). Critically, the theory makes a distinction between unstructured and structured transitions between objects in a system. Unstructured transitions happen when a system’s dynamic structure is dictated solely by random features of the environment. In structured transitions, features are shaped by the environment but are constrained by the composition and membership of the actors in that system. For example, take a hypothetical community of snakes, fungi, or microbes. In an unstructured transition, the membership of those communities is influenced by a nearly infinite space of possibility, limited only by features of the world in which they operate. On the other hand, in structured transitions, membership possibility is constrained by the interactions between the members such that the future state of the communities depends on the current membership. Their work builds on a literature in ecology that examines the different sorts of transitions that occur in ecological systems, for example, previous research on alternative stable states (16, 19, 20).

The authors augment the structuralist approach with another theoretical framework known as the gravity model, which dictates rules of attraction (or interaction) between objects, parcels of information, or biological actors. In this case, the model establishes the force of transition between microbial communities (in this sense, the “gravity” in the gravity model fits the common understanding of gravity, the force of attraction between objects). Last, the authors use Markov chain processes—where the transition between states is determined by the existing membership—to describe the transition probability between any two communities. They utilize “model-free assumptions,” which means that the findings can be applied to systems of various kinds.

In PNAS, Long et al. take the courageous step of establishing a model for community transition that incorporates existing methods and innovates on them.

In PNAS, Long et al. take the courageous step of establishing a model for community transition that incorporates existing methods and innovates on them. The theory they establish predicts that structured transitions have a larger switching capacity than unstructured transitions, which may intuitively fit what happens in the microbiota, where changes are frequent, but not entirely random. Though the theoretical formalisms alone offer a lens on how microbial community structure may arise, the authors further attempt to validate the model and demonstrate its broad applicability. Specifically, they test their predictions on real-world data from the microbiota of the human gut, vaginal, oral, and ocean microbiota. They find that the rate at which switching occurs is higher than expected from an unstructured transition system (the null model), where the transitions are independent of the composition of those microbiota. Furthermore, the authors highlight that different microbiota contexts demonstrated features of the high-switching structured transitions that they theorized. Communities with a high-level of periodicity in their composition, like the vaginal and ocean microbiota, displayed high-switching within a small range of population sizes. Alternatively, settings with low periodicity—like the human gut and oral microbiota—demonstrated high-switching capacity across a wide range of community sizes.

These results are successful in a rare feat: They both highlight rules governing transition dynamics applying to microbiota of varying sorts and demonstrate the uniqueness of the biologies that define certain systems. Doing so constitutes an example of how a “soft physics” approach can be useful in the study of complex biological systems—fundamentals that describe the features of a system, while simultaneously appreciating the uniqueness of individual contexts. Such an approach breaks the classical tradeoff between generality and peculiarity in theoretical biology (21). In PNAS, Long et al. are successful in generating a robust theory with both features.

How do these findings contribute to our grander picture of the microbiota? As described in Fig. 1, the findings are a step forward in resolving one key dimension—dynamics—of the triad of knowledge pillars that help to define the mystery of microbiota. If we are to understand the rules dictating how community membership changes, we might be better equipped to engineer a given microbiota with the composition that we desire or predict when and how transitions will occur.

The ambitious goal of engineering the microbiota aims to utilize a range of methods and knowledge bases to construct microbial communities (22–24). But the findings in Long et al. suggest that engineering requires more than knowledge of how communities emerge, or how they function, as engineering a system of interactors is not a static endeavor. Building a microbiota is not like building most static constructs. Dynamics are important in all systems because the engineered assembly must survive the capricious nature of the world where it lives and operates. Such is the rationale for stress tests and sensitivity analyses, both of which examine whether a system is robust to perturbations of various kinds.

In PNAS, Long et al. take it a step further by offering a theory relevant to how one might engineer microbial communities that experience environmental perturbations, where the actors in the system change over time. And moving past engineering, one can apply these results to the challenge of predicting when microbial transitions will happen, near the point of a community transition, or “early warning systems” as it has been described in infectious diseases (25).

When considering the broader implications, we can use an analogy: Think of a hypothetical piece of important software where the syntax changes in an unpredictable fashion. While each iteration of the code is functional in the strict sense, it is nonetheless challenging to work with. If the software is to be ever used properly (and improved upon by computer scientists), then the user–engineer must learn the pattern underlying how and when coding changes happen.

In PNAS, Long et al. offer that the objects (code in our analogy and member taxa in microbial communities) can change according to a set of organized principles. And while the results speak directly to one of nature’s biggest enigmas in the microbiota, future efforts might explore how this sort of theory may apply to complex biological systems of various kinds, perhaps even interactions between agents in social systems.