Rédiger un PREreview

Microbial communities are structured by interactions spanning antagonism, inducible responses, metabolic modulation, and morphogenetic effects, yet these processes are often reduced to a single antagonistic dimension. Here, we characterized pairwise interactions among 60 soil-derived actinobacterial strains, predominantly Streptomyces, across 12 phenotypes represented as a binary, directed multilayer network. The layers were neither reducible to a single ecological axis nor globally independent. Degree-preserving null models identified significant excess overlap in 8 of 65 layer pairs. Although antimicrobial release and growth inhibition formed the principal connectivity backbone, strongly connected component sizes were consistent with degree-based null expectations. Strain-level strategies formed a heterogeneous multidimensional space. Two principal components exceeded the 95th percentile of Horn’s parallel-analysis null, explaining 49.8% of variance; a third axis exceeded only the weaker mean-eigenvalue criterion and is reported as marginal. Gaussian mixture modelling gave no stable component count across covariance parameterizations, leaving the number of ecological groups unresolved rather than excluded. Dominance was phenotype dependent, with weak cross-layer correlations of David’s scores and no universal sender. Interaction responses also showed phenotype-specific spatial dependence. Ecological strategy was unrelated to 16S rRNA phylogeny. Pagel’s λ showed no significant signal across 21 traits (all q ≥ 0.39), and multilayer ecological distance was unrelated to 16S rRNA distance ( ρ= -0.04, P= 0.59), although inference was constrained by marker resolution. These results reveal a heterogeneous, multidimensional interaction landscape structured by partially coupled functional phenotypes rather than a single antagonistic hierarchy, potentially reflecting semi-independent evolution of secondary metabolism, development, sensing and stress-response pathways.