Host–Microbe Interactions: Winning the Colonization Lottery

Abstract

Stochastic bottlenecks during bacterial colonization of animal hosts lead to reduced genetic diversity in the resulting microbiota and, at low-inoculation doses, can result in hosts that remain uncolonized. Bacterial strains vary in their colonization efficiency and resistance to displacement.

The importance of microbial associations for the growth and development of eukaryotic hosts is no longer in question. However, the details of how bacteria–host associations form and perpetuate are unknown in the vast majority of animal and plant symbiotic systems. The microbial communities associated with animals in their natural settings are often predictable at a coarse level and yet are surprisingly inconstant and seemingly somewhat random at finer focus (e.g. [1-3]). Two recent papers have tackled the question of stochasticity in bacterial colonization of the guts of Drosophila melanogaster [4] and Caenorhabditis elegans [5]. Both papers find that individual bacterial cells have a vanishingly small probability of successfully establishing themselves in the gut — on the order of 1 successful establishment per 1,000 bacterial cells ingested in C. elegans — but that those successful few then go on to establish the community for the entire gut. In one of these papers, which appears in this issue of Current Biology, Obadia et al. [4] fit this to a ‘lottery model’, wherein a few lucky winners out of a large number of participants reap large rewards.

The lottery model of colonization means that some individual hosts may remain uncolonized even though they have ingested millions of bacteria. Obadia et al. [4] use D. melanogaster to experimentally demonstrate that these colonization failures are not because the host fly is refractory to being colonized, but instead because every ingestion yields a finite number of bacterial cells playing the colonization lottery and in some cases none win. A further implication is that a single colonized host will sustain only a subset of the bacterial genetic diversity ingested. Vega and Gore [5] provide evidence in support of this prediction using a neutral competition experiment with labelled Escherichia coli bacteria to demonstrate that individual C. elegans fed on low doses of a mixed culture become arbitrarily colonized by only one of the two bacterial strains. However, both of these properties manifest only when the inoculation dose is relatively low. At higher bacterial inoculations, most or all hosts become colonized [4] and genetic diversity from the inoculum is retained in most individual hosts [5]. This is because even though each individual bacterium has a low probability of successfully colonizing the host, a sufficiently large inoculum will introduce enough cells to produce multiple simultaneous winners (Figure 1).

Figure 1. Stochasticity during bacterial colonization of animal hosts.

Because of stochastic bottlenecks, animal hosts that are inoculated with few bacteria become colonized by a single dominant strain or may entirely fail to become colonized. Animal hosts that are inoculated at a high bacterial dose are more likely to become colonized by a mosaic of strains. (A) Obadia et al. [4] demonstrated the dose dependency of colonization likelihood by feeding strains of L. plantarum to D. melanogaster. (B) Vega and Gore [5] fed E. coli to C. elegans to reveal that mosaic colonization occurs only at high inoculation density.

The findings of Vega and Gore [5] and Obadia et al. [4] are made possible because both studies make use of small, inexpensive and easily manipulated model hosts, where each individual worm or fly represents an independent play of the colonization lottery. Because the experiments are high-throughput, inoculation parameters can be readily varied to test the robustness of the colonization outcomes. Both papers then fit the empirical data to mathematical models that capture the quantitative dynamics in a fairly straightforward conceptual framework. The models posit that each individual bacterial cell that is ingested has a low probability of successfully colonizing the gut, meaning that the expected time between successful colonizations is exponentially distributed with an increasing rate of success at larger initial inoculation sizes. Since successful colonists divide and leave descendants, the physical space they occupy is assumed to be unavailable for subsequent invasion. After adding in parameters for post-ingestion rates of bacterial proliferation and cell death or shedding, the models accurately predict the mosaic colonization resulting from large inocula versus the monomorphic colonizations observed with low-inoculation doses.

Although the lottery model of colonization is explicitly stochastic, this does not imply that all bacteria are equivalent in the game. Obadia et al. [4] demonstrate that a Lactobacillus plantarum strain isolated from a wild-caught fly (WF) is considerably more effective at colonizing the Drosophila gut than are strains of the same species isolated from D. melanogaster in culture (CS) or from a human gastrointestinal tract (HS). WF successfully colonizes 100% of D. melanogaster hosts even when the inoculation dose is as few as 10 bacterial cells, whereas CS and HS strains successfully colonized only 50–70% of hosts at inoculation doses of millions of cells. These outcomes are stereotypical, with 5 out of 6 bacterial species and strains derived from wild flies colonizing at a success rate higher than 90%, compared with only 4 out of 10 species isolated from D. melanogaster in culture [4]. Both the Obadia et al. and Vega and Gore studies demonstrate priority effects, such that once a bacterial lineage becomes established in the gut, it is not easily displaced by subsequently ingested bacterial cells. Even in this regard, however, not all bacteria are equivalent. Bacterial strains [4] and species [5] vary both in their capacity to invade a colonized gut and in their ability to resist displacement, with the WF strain of Obadia et al. [4] showing particular ability to supplant previously introduced bacteria and to avoid being displaced itself.

The WF strain of Lactobacillus plantarum described in Obadia et al. [4] would appear to be highly adapted to colonization and persistence in the D. melanogaster gut, exhibiting both exceptional capacity to colonize and strong resistance to displacement compared with the CS strain isolated from D. melanogaster in culture. An intriguing alternative explanation, however, is that the CS strain is instead exceptionally well adapted to life in the vial. Obadia et al. [4] suggest that the WF strain of L. plantarum may have greater capability for adhesion to the gut lining, whereas flies inoculated with CS strain shed live bacterial cells at the highest rate. It is reasonable to infer that adhesion is superfluous for strains that experience continuous oral–faecal transmission in a closed space. The most adaptive strategy for bacteria associated with flies in culture may beto quickly flush through the gut and predominantly grow and divide on the Drosophila rearing medium [6]. Consistent with this interpretation, Acetobacter isolated from laboratory Drosophila consistently lose genes encoding flagellar motility and are enriched in the capacity for uric acid metabolism compared with their wild counterparts (P.D. Newell and A.E. Douglas, personal communication). These adaptations are consistent with life in an environment rich in uric acid excreted by flies and where the bacteria can expect passive carriage from vial to vial in perpetuity. Lactobacillus and Acetobacter strains have previously been shown to be rapidly lost from flies that are frequently transferred to sterile medium and therefore are not able to re-inoculate themselves [7], demonstrating that growth in the medium is crucial for stable association of bacterial strains that evolve with flies in culture.

The differences among bacterial strains and species in colonization efficiency and resistance to displacement raise an important caveat to the colonization models described by Vega and Gore [5] and Obadia et al. [4]. These models are defined under the assumption that a naïve, axenic gut is being colonized with a virgin inoculation of bacteria. While the first introduction of bacteria to a sterile host is a defining moment [8,9], what follows is a lifetime barrage of microbes into an established community. Priority effects, facilitation, microbe–microbe warfare, and impacts of host immunity and physiology are likely to be important factors influencing whether an ingested microbe will successfully colonize a mature gut. These will not negate the underlying stochasticity of colonization, but they may bend the odds. Empirical extension of the base lottery model to include direct and indirect interactions among microbes and between the host and potential colonists will require years of additional experimentation.

Nevertheless, an element of randomness in bacterial colonization of eukaryotic hosts is likely to prove to be a universal rule. In addition, stochastic bottlenecks during colonization are unlikely to be limited to the gut. The same principles are likely to apply to colonization of any part of an animal (e.g. [9]) or plant (e.g. [10]), including during pathogenic infections [11]. Indeed, in a parallel to the experimental structure of Vega and Gore [5], we used the standard pinprick assay for systemic infection of D. melanogaster [12] to introduce mixed-genotype bacterial infections at low, medium or high dose (~300, 3,000 or 30,000 cells of Providencia rettgeri, respectively). Consistent with the results observed by Obadia et al. and Vega and Gore, one of the two P. rettgeri genotypes randomly dominated each infected fly at the low-dose and medium-dose infections, although both genotypes became represented within individual flies at higher doses of infection (G.M. Fox and B.P. Lazzaro, unpublished data).

It is important to appreciate that bottlenecks in the colonization of the gut or any other body compartment may have diverse mechanistic bases, for example relating to the physics of how bacteria are introduced to hosts, the structural properties of the host or microbe, immunological interactions and resistance, or elements of microbial ecology. These bottleneck forces are certain to vary among diverse hosts and microbes and, crucially, may differ between laboratory and field settings. As phrased by Vega and Gore [5], this “demographic noise” is likely to be very important in the formation of microbial communities as thousands of bacterial cells compete in a colonization lottery that only few can expect to win.