Strong Dispersal Limitation of Microbial Communities at Shackleton Glacier, Antarctica

Abstract

Microbial communities can be structured by both deterministic and stochastic processes, but the relative importance of these processes remains unknown. The ambiguity partly arises from an inability to disentangle soil microbial processes from confounding factors, such as aboveground plant communities or anthropogenic disturbance. In this study, we characterized the relative contributions of determinism and stochasticity to assembly processes of soil bacterial communities across a large environmental gradient of undisturbed Antarctic soils. We hypothesized that harsh soils would impose a strong environmental selection on microbial communities, whereas communities in benign soils would be structured largely by dispersal. Contrary to our expectations, dispersal was the dominant assembly mechanism across the entire soil environmental gradient, including benign environments. The microbial community composition reflects slowly changing soil conditions and dispersal limitation of isolated sites. Thus, stochastic processes, as opposed to deterministic, are primary drivers of soil ecosystem assembly across space at our study site. This is especially surprising given the strong environmental constraints on soil microorganisms in one of the harshest environments on the planet, suggesting that dispersal could be a driving force in microbial community assembly in soils worldwide. IMPORTANCE Because of their diversity and ubiquity, microbes provide an excellent means to tease apart how natural communities are structured. In general, ecologists believe that stochastic assembly processes, like random drift and dispersal, should dominate in benign environments while deterministic processes, like environmental filtering, should be prevalent in harsh environments. To help resolve this debate, we analyzed microbial community composition in pristine Antarctic soils devoid of human influence or plant communities for eons. Our results demonstrate that dispersal limitation is a surprisingly potent force of community limitation throughout all soil conditions. Thus, dispersal appears to be a driving force of microbial community assembly, even in the harshest of conditions.

Keywords: community assembly; determinism; dispersal; niche; stochasticity.

FIG 1

Map of the Shackleton Glacier, with sampling sites denoted as colored point. Map by Mike Cloutier, Polar Geospatial Center (Imagery © 2021 Maxar; reproduced with permission).

FIG 2

Relationships between soil salt concentrations, exposure age, and microbial ASV richness. (A) The relationship between soil principal component (PC 1) and soil NO₃- concentrations. The thick blue line shows the linear model fit to the observed data, and thin blue lines show 1,000 bootstrapped linear regressions. (B) Relationship between ASV richness and soil NO₃- concentrations. Thick blue line shows the LOWESS fit to the observed data, while thin blue lines show 1,000 bootstrapped LOWESS estimates.

FIG 3

Redundancy analyses of microbial community composition. (A) Redundancy analysis (RDA) biplot of microbial community composition in relation to physical variables and spatial arrangement of sites. The cluster of points in the upper left corner are Robert’s Massif. (B) Relationship between RDA Axis 2 and soil NO₃- concentrations. Thick blue line shows the linear model fit to all observed data, and thin blue lines show 1,000 bootstrapped linear regressions. (C) Relationship between Soil PC 1 and distance from the coast (km) for each site. Thick blue line shows the LOWESS fit to all observed data, and thin blue lines show 1,000 bootstrapped LOWESS fits.

FIG 4

Microbial communities were largely structured by stochastic processes across all soil wetting ages. (A) Contribution of stochastic and deterministic processes to microbial community composition based on Sørenson dissimilarity, which records presence-absence dissimilarities. We calculated the average variance component for each site. Thick lines show a LOWESS regression against NO₃- concentrations. Thin lines show 1,000 bootstrapped LOWESS regressions. (B) Contribution of stochastic and deterministic processes to microbial community composition based on Bray-Curtis dissimilarities, which record relative abundance dissimilarities. In all simulations, phylum richness was held fixed at the observed level for each site. Thick lines show a LOWESS regression against NO₃- concentrations. Thin lines show 1,000 bootstrapped LOWESS regressions.

FIG 5

Dissimilarity of modeled communities to observed communities when modeled communities were structured by neutral, dispersal, or environmental processes. (A) Averaged Sørenson dissimilarity of simulated communities from observed communities. For each site, we took the average pairwise dissimilarities to the observed site from 1,000 model simulations. Thick lines show the LOWESS fit to the averaged dissimilarities as a function of soil NO₃- concentrations. (B) For each site, we took the average pairwise dissimilarities to the observed site from 1,000 model simulations. Thick lines show the LOWESS fit to the average dissimilarities as a function of soil NO₃- concentrations. Thin lines show 1,000 bootstrap simulations. In all simulations, phylum richness was fixed to the observed level for each site.