Predicting community responses to ecosystem changes is essential for improving ecosystem management. From an industrial perspective, we are dependent on stable microbial communities that perform well. Moreover, we live in a time where humans create disturbances at various levels in natural ecosystems. It is therefore important to comprehend the consequences of our activity. To predict the community response to external forces, we need to understand how different ecosystems affect the community assembly processes.
We aimed to fill the knowledge gap on how carrying capacity and periodical disturbances affect the community assembly. It has previously been shown that the carrying capacity affects the community composition [46]. However, its effect on the assembly processes has remained unclear. Ecosystems with a lower carrying capacity support lower community size. Because the outcome of drift is density-dependent [6], communities with a low carrying capacity should have more populations vulnerable to drifting to extinction. However, our five-times difference in carrying capacity between cultivation regimes did not result in apparent differences in community assembly. The only exception was for the disturbed communities in Period 2, where the low carrying capacity regime (UDL) indicated a stronger influence of selection than the high (UDH; Fig. 4b). This observation was surprising as we hypothesised that drift might be more pronounced in systems with lower carrying capacity. In conclusion, the minor effects of carrying capacity observed for the replicate similarity rate for the undisturbed communities suggest that the effect of carrying capacity should be investigated further, including larger differences in carrying capacity.
The effect of the disturbance regime on the microbial community assembly was more evident. The disturbance we investigated was a substantial dilution of the microcosm’s inoculum. The dilution has two significant effects: the community size is reduced, and the concentration of resources increases strongly for the remaining individuals. These two changes are relevant in natural and human-created ecosystems, where resource supply vary due to natural processes (e.g. patchiness and floods) and human activity (e.g. eutrophication and saprobiation).
Investigating the temporal community composition through ordinations can reveal overall successional trajectories [47]. We found that whereas the PCoA ordinations indicated an overall deterministic trajectory for the undisturbed communities, the replicate similarity rate indicated that drift dominated the community assembly. This was evident for the microcosms starting with undisturbed culture conditions (UD Δµ > 0; Fig. 5). However, the results were less evident for the communities going from disturbed to undisturbed conditions (DU) as the replicate similarity rate was around zero. Nonetheless, there was an apparent decrease in the replicate similarity rate when going from disturbed (Δµ 1.1 × 10−2) to undisturbed conditions (Δµ 5 × 10–4).
The strength and unique feature of our experiment is the crossed design of the disturbance regimes. This crossing considerably increases the robustness of the conclusions drawn from the data. First, during the first period, all microcosms were inoculated with the same community, but in the second period, the twelve communities had assembled individually for 28 days. We could therefore investigate the effects of our experimental variables on drift and selection with different starting conditions. The temporal trends in the data were found to be independent of the starting condition, substantially increasing the strength of our conclusion.
Second, subjecting the communities to the opposite disturbance regime in Period 2 supports that we had stable attractors in our systems. An attractor is a point or a trajectory in the state space of a dynamical system. If the attractor is locally stable, the system will tend to evolve toward it from a wide range of starting conditions and stay close to it even if slightly disturbed [48]. We observed locally stable attractors based on the disturbance regime and thus one stationary phase for each disturbance regime. Some ecological systems show dramatic regime shifts between alternative stationary states in response to changes in an external driver [49]. Such systems typically exhibit hysteresis in the sense that they will not return directly to the original state by an opposite change in the driver. We found that community composition was reversible and dependent on the disturbance regime, as highlighted by the Bray–Curtis ordinations (Fig. 4). This reversibility indicates that the community changes we observed were not catastrophic bifurcations or regime shifts and that it is unlikely that the systems contain multiple stationary states within the same disturbance regime. We think this gives strong support for assuming that drift is the main driver for divergence in the community composition and that selection towards alternative attractors probably plays a minor role. Thus, we can conclude that shifting from a disturbed to an undisturbed ecosystem increased the contribution of drift. Our observations corroborate other investigations of bioreactors [15, 50] and simulations [51] that report that stochasticity is fundamental for the assembly of communities. However, the finding that drift was important for structuring the undisturbed microcosms was unexpected.
In dispersal-limited communities where resources are supplied continuously, such as in the undisturbed communities examined here, the selective process competition has been hypothesised to be high [7]. However, our experimental environment offered little variation in the resources provided, as the medium provided was the same throughout the experiment. This may have led to populations becoming “ecologically equivalent”, meaning that their fitness difference was too small to result in competitive exclusion on the time scale of our experiment [5, 52]. Under these assumptions, community assembly is similar to the neutral model in which the growth rates of the community members are comparable [53].
During disturbances, we found that selection dominated community assembly. Our results support Zhou et al. hypothesis stating that determinism should increase due to biomass loss in dispersal-limited communities [24]. However, they oppose their other hypothesis stating that nutrient inputs should increase stochasticity [24], making low abundant populations vulnerable to local extinction [6, 7]. During the disturbances, the Sørensen similarity between replicates was stable or increasing, indicating that the periodical disturbance did not result in the extinction of low abundant populations. Instead, it appears that the dilution removed competition for some time, resulting in a phase where all populations got “a piece of the cake”. Several studies have observed increased stochasticity as a result of increased resource availability [7, 11, 24, 26]. However, we found that disturbances resulting in periods with exponential growth due to density-independent loss of individuals and high resource input suppressed the effect of stochastic processes. This exponential growth period without competition would enable more populations to stay above the detection limits of the 16S-rDNA-sequencing method.
More OTUs were enriched under the disturbed regime than under the undisturbed. During the disturbance, the microcosms were diluted ~2 day−1, whereas the dilution factor was 1 day−1 during the undisturbed regime. We cannot assume steady-state in the disturbed microcosms, but it was interesting to see a substantial increase in the abundance of OTUs classified as Gammaproteobacteria. Gammaproteobacteria include many opportunists [54] that appeared to exploit the resource surplus following the disturbance. This opportunistic lifestyle fits within the r- and K-strategist framework [55].
Organisms with high maximum growth rates but low competitive abilities are classified as r-strategists. These r-strategists are superior in environments where the biomass is below the carrying capacity. On the other hand, K-strategists are successful in competitive environments due to their high substrate affinity and resource specialisation [56]. Based on the taxonomic responses, it appears as disturbances in the form of dilutions selected for r-strategists, whereas the undisturbed regime selected for K-strategists. The r-strategists selected for during the disturbance periods included genera such as Vibrio and Colwellia [57], and the genus Vibrio includes many pathogenic strains [58]. Thus, our findings may have implications for land-based aquaculture systems where conditions favouring r-strategists is linked to high mortality and reduced viability of fish [56].
The DeSeq2 results pose some new questions regarding the link between phylogeny and niche fitness. Generally, ecologists assume that closely related taxa have similar niches, as they have a common evolutionary history and, thus, similar physiology [59, 60]. For example, here, OTUs belonging to Gammaproteobacteria co-occurred when the environment was disturbed. However, for other classes such as Alphaproteobacteria and Flavobacteria, the OTUs responded differently to the disturbance regimes, despite belonging to the same class. This lack of phylogenetically coherent response indicates that the paradigm of correlation between phylogeny and niche requires further studies.
This study was performed on complex marine microbial communities cultivated under controlled experimental conditions. We found that undisturbed environments enhanced the contribution of drift on community assembly and that disturbances increased the effect of selection. These observations might be different in more diverse ecosystems such as soils or the human gut. In such ecosystems, the microbes are more closely associated with, for example, soil particles or attached to the gut lining. It has been shown that the biofilm-associated and planktonic microbial communities have different community compositions [61]. Consequently, the community assembly processes may be affected differently by environmental fluctuations. Our experimental variables should therefore be tested in other ecosystem settings to verify our conclusions.
To our knowledge, this study is the first to experimentally estimate the effect of periodical disturbances and carrying capacity on community assembly in dispersal-limited ecosystems. We observed that carrying capacity had little effect on community assembly and that undisturbed communities were structured more by drift than disturbed systems dominated by selection. Using an experimental crossover design for the disturbance regime, we showed that these observations were independent of the initial community composition. Our experiment illustrates that cultivating complex natural microbial communities under lab conditions allowed us to test ecologically relevant system variables and draw robust conclusions.
Source: Ecology - nature.com
