Deterministic and stochastic processes generating alternative states of microbiomes

Abstract

The structure of microbiomes is often classified into discrete or semi-discrete types potentially differing in community-scale functional profiles. Elucidating mechanisms that generate such “alternative states” of microbiome compositions has been one of the major challenges in ecology and microbiology. In a time-series analysis of experimental microbiomes, we here show that both deterministic and stochastic ecological processes drive divergence of alternative microbiome states. We introduced species-rich soil-derived microbiomes into eight types of culture media with 48 replicates, monitoring shifts in community compositions at six time points (8 media × 48 replicates × 6 time points = 2,304 community samples). We then confirmed that microbial community structure diverged into a few state types in each of the eight medium conditions as predicted in the presence of both deterministic and stochastic community processes. In other words, microbiome structure was differentiated into a small number of reproducible compositions under the same environment. This fact indicates not only the presence of selective forces leading to specific equilibria of community-scale resource use but also the influence of demographic drift (fluctuations) on the microbiome assembly. A reference-genome-based analysis further suggested that the observed alternative states differed in ecosystem-level functions. These findings will help us examine how microbiome structure and functions can be controlled by changing the “stability landscapes” of ecological community compositions.

Basins of attraction of microbiome structure and soil ecosystem functions

“…The energy landscape analysis allows systematic analyses of taxon-rich community datasets by incorporating the information of multiple environmental factors (Dakos and Kéfi, 2022;Sánchez-Pinillos et al, 2024;Suzuki et al, 2021). While classic studies on community multistability have discussed ecological processes spanning a few intuitively distinguishable community states [high/low tree cover in forest-savanna transitions (Hirota et al, 2011;Staver et al, 2011aStaver et al, , 2011b or macrophyte-/phytoplankton-dominated state in shallow lakes (Ibelings et al, 2007;Scheffer and Carpenter, 2003)], it is now made possible to define basins of attraction based on highdimensional community datasets involving hundreds of species/taxa (Arumugam et al, 2011;Costea et al, 2017;Guim Aguadé-Gorgorió et al, 2023;Hayashi et al, 2024). Application of the general statistical platform will enhance our understanding of how stability landscape properties differ among diverse microbial and non-microbial systems.…”

“…Because our present data lacked the information of temporal changes in community structure, we are unable to discuss the frequency and pace of community structural transitions between basins of attraction. Monitoring of microbiome compositions (Faust et al, 2015; Hayashi et al, 2024; Yajima et al, 2023) is necessary for filling the gap between theoretical and empirical studies (Long et al, 2024).…”

“…Based on amplicon and shotgun sequencing technologies, large datasets of microbial species/taxonomic compositions have been made available, providing a basis for exploring reproducible states in microbiome community structure (Amor et al, 2020;Hayashi et al, 2024). Such high-throughput DNA sequencing studies in medicine, for example, have shown that human individuals can be classified into three or four semi-discrete clusters in terms of their intestinal microbiome compositions (Arumugam et al, 2011;Wu et al, 2011) [see also (Jeffery et al, 2012;Knights et al, 2014)].…”

“…While classic studies targeting freshwater and terrestrial biomes have explored basins of attraction based on simple characterization of community states (e.g., tree cover percentages), recent technical advances in microbial community (microbiome) research have come to provide opportunities for deepening our knowledge of biological community stability (Amor et al, 2020; Costea et al, 2017; Faust et al, 2015; Shaw et al, 2019; Toju et al, 2018; Zaneveld et al, 2017). Based on amplicon and shotgun sequencing technologies, large datasets of microbial species/taxonomic compositions have been made available, providing a basis for exploring reproducible states in microbiome community structure (Amor et al, 2020; H Fujita et al, 2023; Hayashi et al, 2024). Such high-throughput DNA sequencing studies in medicine, for example, have shown that human individuals can be classified into three or four semi-discrete clusters in terms of their intestinal microbiome compositions (Arumugam et al, 2011; Wu et al, 2011) [see also (Jeffery et al, 2012; Knights et al, 2014)].…”