Environmental stress is increasing worldwide, yet we lack a clear picture of how stress disrupts the stability of microbial communities and the ecosystem services they provide. Here, we present the first evidence that naturally-occurring microbiomes display network properties characteristic of unstable communities when under persistent stress. By assessing changes in diversity and structure of soil microbiomes along 40 replicate stress gradients (elevation/water availability gradients) in the Florida scrub ecosystem, we show that: (1) prokaryotic and fungal diversity decline in high stress, and (2) two network properties of stable microbial communities—modularity and negative:positive cohesion—have a clear negative relationship with environmental stress, explaining 51–78% of their variation. Interestingly, pathogenic taxa/functional guilds decreased in relative abundance along the stress gradient, while oligotrophs and mutualists increased, suggesting that the shift in negative:positive cohesion could result from decreasing negative:positive biotic interactions consistent with the predictions of the Stress Gradient Hypothesis. Given the crucial role microbiomes play in ecosystem functions, our results suggest that, by limiting the compartmentalization of microbial associations and creating communities dominated by positive associations, increasing stress in the Anthropocene could destabilize microbiomes and undermine their ecosystem services.

The Stress Gradient Hypothesis (SGH), which predicts increasing ratios of facilitative:competitive interactions with increasing stress, has long been a guiding framework for conceptualizing plant–plant interactions. Recently, there has been a growing recognition of the roles of microbes in mitigating or exacerbating environmental stress for their plant hosts. As such, we might predict, based on the SGH, that beneficial microbial effects on plant performance should be positively associated with stress. Specifically, we hypothesized that support for the SGH would depend on the host plant's habitat specialization such that species that specialize in high stress habitats and thus likely coevolved with the resident microbes would exhibit stronger support for the SGH than non‐specialist plant species. We further hypothesized that support for the SGH would vary with germination frequency, since boosting germination of low‐frequency germinators is one effective means by which microbes can benefit plant species performance. Here, we explore whether plant–microbial interactions support the SGH using 12 plant species native to the Florida rosemary scrub. We conducted factorial experiments that manipulated the presence of microbes in nine soils collected along an elevational stress gradient, and recorded germination frequency and biomass. Microbes increased the germination frequency of four species, all of which had relatively low germination rates. Furthermore, we found support for the SGH in nearly one‐half of the species examined, with soil microbes facilitating germination with increasing stress for 5 of the 12 species tested, and none of the species exhibiting the opposite trend. Support for the SGH was not predicted by either the plant hosts' habitat specialization or germination frequency. In contrast to germination, biomass results showed little support for the SGH, with four of 12 species refuting and one species supporting SGH predictions. Taken together, our study documents that interactions between the soil microbial community and plant species along a stress gradient can support the SGH, but emphasizes that these effects are life‐history‐stage dependent. This work also identifies a common mechanism (germination facilitation) by which microbes can benefit plant species in stressful habitats.

Summary: Allelopathy is a common and important stressor that shapes plant communities and can alter soil microbiomes, yet little is known about the direct effects of allelochemical addition on bacterial and fungal communities or the potential for allelochemical‐selected microbiomes to mediate plant performance responses, especially in habitats naturally structured by allelopathy. Here, we present the first community‐wide investigation of microbial mediation of allelochemical effects on plant performance by testing how allelopathy affects soil microbiome structure and how these microbial changes impact germination and productivity across 13 plant species. The soil microbiome exhibited significant changes to ‘core’ bacterial and fungal taxa, bacterial composition, abundance of functionally important bacterial and fungal taxa, and predicted bacterial functional genes after the addition of the dominant allelochemical native to this habitat. Furthermore, plant performance was mediated by the allelochemical‐selected microbiome, with allelopathic inhibition of plant productivity moderately mitigated by the microbiome. Through our findings, we present a potential framework to understand the strength of plant–microbial interactions in the presence of environmental stressors, in which frequency of the ecological stress may be a key predictor of microbiome‐mediation strength.

Microorganisms underpin numerous ecosystem processes and support biodiversity globally. Yet, we understand surprisingly little about what structures environmental microbiomes, including how to efficiently identify key players. Microbiome network theory predicts that highly connected hubs act as keystones, but this has never been empirically tested in nature. Combining culturing, sequencing, networks and field experiments, we isolated ‘central’ (highly connected, hub taxa), ‘intermediate’ (moderately connected), and ‘peripheral’ (weakly/unconnected) microbes and experimentally evaluated their effects on soil microbiome assembly during early succession in nature. Central early colonisers significantly (1) enhanced biodiversity (35%–40% richer communities), (2) reshaped trajectories of microbiome assembly and (3) increased recruitment of additional influential microbes by > 60%. In contrast, peripheral microbes did not increase diversity and were transient taxa, minimally affected by the presence of other microbes. This work elucidates fundamental principles of network theory in microbial ecology and demonstrates for the first time in nature that central microbes act as keystone taxa.