Abstract
The priming effect (PE) occurs when fresh organic matter (FOM) supplied to soil alters the rate of decomposition of older soil organic matter (SOM). The PE can be generated by different mechanisms driven by interactions between microorganisms with different live strategies and decomposition abilities. Among those, stoichiometric decomposition results from FOM decomposition, which induces the decomposition of SOM by the release of exoenzymes by FOM-decomposers. Nutrient mining results from the co-metabolism of energy-rich FOM with nutrient-rich SOM by SOM-decomposers. While existing statistical approaches enable measurement of the effect of community composition (linear effect) on the PE, the effect of interactions among co-occurring populations (non-linear effect) is more difficult to grasp. We compare a non-linear, clustering approach with a strictly linear approach to separately and comprehensively capture all linear and non-linear effects induced by soil microbial populations on the PE and to identify the species involved. We used an already published data set, acquired from two climatic transects of Madagascar Highlands, in which the high-throughput sequencing of soil samples was applied parallel to the analysis of the potential capacity of microbial communities to generate PE following a (13)C-labeled wheat straw input. The linear and clustering approaches highlight two different aspects of the effects of microbial biodiversity on SOM decomposition. The comparison of the results enabled identification of bacterial and fungal families, and combinations of families, inducing either a linear, a non-linear, or no effect on PE after incubation. Bacterial families mainly favoured a PE proportional to their relative abundances in soil (linear effect). Inversely, fungal families induced strong non-linear effects resulting from interactions among them and with bacteria. Our findings suggest that bacteria support stoichiometric decomposition in the first days of incubation, while fungi support mainly the nutrient mining of soil's organic matter several weeks after the beginning of incubation. Used together, the clustering and linear approaches therefore enable the estimation of the relative importance of linear effects related to microbial relative abundances, and non-linear effects related to interactions among microbial populations on soil properties. Both approaches also enable the identification of key microbial families that mainly regulate soil properties.