Quantifying the contribution of the rare biosphere to natural disturbances

Abstract

Understanding how populations respond to disturbances represents a major goal for microbial ecology. While several hypotheses have been advanced to explain microbial community compositional changes in response to disturbance, appropriate data to test these hypotheses is scarce, due to the challenges in delineating rare vs. abundant taxa and generalists vs. specialists, a prerequisite for testing the theories. Here, we operationally define these two key concepts by employing the patterns of coverage of a (target) genome by a metagenome to identify rare populations, and by borrowing the proportional similarity index from macroecology to identify generalists. We applied these concepts to time-series (field) metagenomes from the Piver's Island Coastal Observatory to establish that coastal microbial communities are resilient to major perturbations such as tropical cyclones and (uncommon) cold or warm temperature events, in part due to the response of rare populations. Therefore, these results provide support for the insurance hypothesis [i.e. the rare biosphere has the buffering capacity to mitigate the effects of disturbance]. Additionally, generalists appear to contribute proportionally more than specialists to community adaptation to perturbations like warming, supporting the disturbance-specialization hypothesis [i.e. disturbance favors generalists]. Several of these findings were also observed in replicated laboratory mesocosms that aimed to simulate disturbances such as a rain-driven washout of microbial cells and a labile organic matter release from a phytoplankton bloom. Taken together, our results advance understanding of the mechanisms governing microbial population dynamics under changing environmental conditions and have implications for ecosystem modeling.

Keywords: disturbance; generalists and specialists; metagenomics; metagenomics assembled genomes (MAGs); rare biosphere; resilience.

Figure 1

Characteristics of the disturbance events studied here and their effects on microbial community diversity. (A) Details of the PICO time series samples. Sample names are labeled chronologically within boxes, the color of which corresponds to the season along the three-year sampling period (see key). Continuous numbers in the name of samples indicate samples taken one week apart [e.g. pico281 and pioc282 samples are taken 1 week apart, while pico284 is sampled 2 weeks after pico282]. Each vertical line indicates a disturbance event and is labelled with a number for convenience. Description below each vertical line shows the details of each disturbance; additional details are provided in Table S1 and S2. (B) Chao1 diversity index for both amplicon 16S rRNA genes (blue) and extracted 16S rRNA gene reads from metagenomes (orange) are shown [using OTUs defined at 97% identity, see methods; similar patterns were observed with the Shannon index (data not shown)]. (C) Nonpareil diversity (N_d). (D) Mash distance based NMDS of metagenomes subsampled at the same sequencing effort. Colored arrows in (B), (C) and (D) show how Nd diversity and metagenomic composition changed by each disturbance event. Grey arrows in (D) represent the change in community composition following the disturbance event caused by either natural variation and/or recovery (the beginning of the arrow denotes pre-disturbance and the end of it denotes during disturbance samples). Note that not all events included post-disturbance samples and that some of the compositional shifts observed based on whole-metagenomes may not match perfectly the shifts observed based on 16S rRNA gene amplicon data (Fig. S3), which were used to defined disturbance events (see text for additional discussion). Also, events that affected small parts of the community, and thus did not result in substantial changes in community composition to be discernible by our analysis represented in the graph of panel (D), were not included in our study.

Figure 2

Graphical representation of our approach to define abundant vs. rare MAGs. MAG coverage depth (left y axis, blue bars) and coverage breadth (right y axis, orange line, shown as 1- coverage breadth) distribution for one metagenomic sample (pico127). The X axis is MAG rank by abundance, estimated as coverage depth, [i.e. TAD80 values normalized by genome equivalents]. Dashed blue and orange lines represent the normalized coverage depth of 0.1% and the coverage breadth of 0.1, respectively. The shaded grey region and vertical line (center of area on the x axis) indicate where both coverage depth and breadth drop sharply as abundance rank increases. The vertical line was therefore used to delineate abundant (left of the line) vs. rare (right of the line) MAGs. The green line is a log fitting of coverage depth vs. rank abundance with the function shown above the line. For detailed model fitting of coverage depth distribution, see Supplementary Fig. S6.

Figure 3

Response of abundant vs. rare MAGs to disturbance events. (A) The figure shows the number of MAGs for each of the three categories assessed: abundant MAGs that remained abundant after the event, MAGs transitioning to abundant from rare, and MAGs becoming rare from abundant for each of the disturbance events. For one given event, if MAG’s relative abundance and coverage breath fall below the threshold of being abundant in the pre-disturbance sample and fall above the threshold of abundant in the disturbed sample, this MAG will be in the category of Rare_Abun. Similar rules applied for other two categories based on MAG abundances before and after each disturbance event. Disturbance events are numbered as in Fig. 1. (B) Total cumulative relative abundance of MAGs assigned to each category.

Figure 4

Fraction of specialists vs. generalists selected by each disturbance event. (A) The graph shows the number of MAGs that became abundant from rare [i.e. favored by each disturbance event]. (B) Fraction of specialists and generalists disfavored by each disturbance event. Disfavored MAGs are those that become rare from abundant [i.e. similar to (A) but the opposite pattern]. Disturbance events are numbered as in Fig. 1A.