Dormancy contributes to the maintenance of microbial diversity

Abstract

Dormancy is a bet-hedging strategy used by a variety of organisms to overcome unfavorable environmental conditions. By entering a reversible state of low metabolic activity, dormant individuals become members of a seed bank, which can determine community dynamics in future generations. Although microbiologists have documented dormancy in both clinical and natural settings, the importance of seed banks for the diversity and functioning of microbial communities remains untested. Here, we develop a theoretical model demonstrating that microbial communities are structured by environmental cues that trigger dormancy. A molecular survey of lake ecosystems revealed that dormancy plays a more important role in shaping bacterial communities than eukaryotic microbial communities. The proportion of dormant bacteria was relatively low in productive ecosystems but accounted for up to 40% of taxon richness in nutrient-poor systems. Our simulations and empirical data suggest that regional environmental cues and dormancy synchronize the composition of active communities across the landscape while decoupling active microbes from the total community at local scales. Furthermore, we observed that rare bacterial taxa were disproportionately active relative to common bacterial taxa, suggesting that microbial rank-abundance curves are more dynamic than previously considered. We propose that repeated transitions to and from the seed bank may help maintain the high levels of microbial biodiversity that are observed in nearly all ecosystems.

Fig. 1.

Summary of multispecies microbial dormancy model structure. A portion of each taxon can occupy an active and dormant state. Active portions of populations have logistic growth and experience density-independent mortality. Dormant individuals are inactive and can therefore not reproduce, but are subject to mortality. Transition between the active and dormant states is controlled by environmental cues at local and regional scales.

Fig. 2.

Mean local richness (number of species) of model equilibrium solutions across a range of values for the strength of dormancy, manipulated using the parameter for mortality of dormant individuals (m_D), and strength of regional environmental cues, manipulated by a weighting parameter (W).

Fig. 3.

UPGMA dendrograms depicting the degree of coupling between active and total (active + dormant) composition in simulated and lake microbial communities. By varying the mortality rate of dormant individuals (m_D) and the relative strength of regional and local environmental cues (W), we explored how dormancy and regional cues influenced coupling or decoupling of active and total communities. Vertical scales indicate Bray-Curtis distance. (A) Simulated communities with low mortality of dormant individuals (m_D = 0.1) and strong regional environmental cues (W = 0.95) resulted in decoupling of active and total communities. (B) In contrast, simulated communities with low persistence of dormant individuals (m_D = 0.9) and weak regional control (W = 0.05) resulted in coupling between active and total communities. (C) The composition of active (rRNA) and total (rDNA) bacterial communities in eight lakes was decoupled. (D) In contrast, the composition of active and total eukaryotic communities was more strongly coupled.

Fig. 4.

The contribution of dormant taxa to microbial richness along a total phosphorus (TP) gradient in our lake survey. Dormant taxa were identified from terminal restriction fragments that were recovered in rDNA fingerprint profiles but not in rRNA fingerprint profiles. The proportion of dormant bacterial taxa decreased with increasing TP (a proxy for ecosystem productivity), but this was not the case for eukaryotic microbes. Together, the proportion of dormant taxa (D_T) could be predicted for different microbial domains (M; 1 for bacteria, 0 for eukaryotes) with the following multiple regression model: D_T = 0.12 + 0.25M − 0.005(M)(TP) (R² = 0.85, n = 16, P < 0.001).

Fig. 5.

Rank-abundance curves for the total (rDNA-based) community of bacteria from two lakes. Points indicate the relative recovery of the corresponding OTUs (97% sequence identity) in the active community (rRNA-based). Gray squares indicate OTUs that were scored as disproportionately active, and black squares indicate OTUs that were scored as disproportionately inactive. Logistic regression revealed that with each increasing rank, the probability of a taxon being active increased by 6% and 2% for Lake 1 and Lake 2, respectively.