Temporal dynamics of hot desert microbial communities reveal structural and functional responses to water input

Abstract

The temporal dynamics of desert soil microbial communities are poorly understood. Given the implications for ecosystem functioning under a global change scenario, a better understanding of desert microbial community stability is crucial. Here, we sampled soils in the central Namib Desert on sixteen different occasions over a one-year period. Using Illumina-based amplicon sequencing of the 16S rRNA gene, we found that α-diversity (richness) was more variable at a given sampling date (spatial variability) than over the course of one year (temporal variability). Community composition remained essentially unchanged across the first 10 months, indicating that spatial sampling might be more important than temporal sampling when assessing β-diversity patterns in desert soils. However, a major shift in microbial community composition was found following a single precipitation event. This shift in composition was associated with a rapid increase in CO₂ respiration and productivity, supporting the view that desert soil microbial communities respond rapidly to re-wetting and that this response may be the result of both taxon-specific selection and changes in the availability or accessibility of organic substrates. Recovery to quasi pre-disturbance community composition was achieved within one month after rainfall.

Figure 1. Non-metric multidimensional scaling ordination plots.

(a) Soil chemistry profiles (Euclidean distances with standardized data), excluding moisture content. To appreciate the changes induced by the addition of soil moisture content in the grouping of the samples see Supp. Figure 1. (b) Sequencing data (Bray-Curtis distances after Hellinger transformation). D328, microbial communities 3 days after rainfall; D355, microbial communities 30 days after rainfall. Samples that are closer together are more similar in soil chemistry or microbial community composition.

Figure 2. Global trend analysis in α-diversity (richness).

A linear trend is fit (red line) using generalized least squares (GLS) applying a model with AR1 temporal autocorrelation of errors. The average value on each sampling date is plotted with ± S.D. of the mean.

Figure 3. Temporal decay curves for microbial communities.

The blue line denotes the regression for communities sampled before the rainfall event. The red line denotes a separate regression for all sampling points. Both lines were obtained using the generalized dissimilarity model of Millar et al.. Only the slope of the solid red line, including the samples after the rainfall event, is significantly less than zero.

Figure 4. Taxonomic distribution, phylum level, of bacterial OTUs (97% cutoff).

Affiliation was performed using the Ribosomal Database Project Classifier with a confidence threshold of 80%. Pre-rainfall, microbial communities before the rainfall event; D328, microbial communities 3 days after rainfall; D355, microbial communities 30 days after rainfall.