Bacteroides species occupy central positions in gut microbial cross-feeding networks as dominant degraders of dietary and host-derived glycans. By releasing diffusible metabolites, Bacteroides can support beneficial commensals and host homeostasis, yet the same interactions may be co-opted under dysbiotic conditions to promote opportunistic expansion, disease progression, or therapeutic resistance. These divergent outcomes arise from context-dependent network structures, strain-level heterogeneity, and spatial organization rather than intrinsic microbial traits. Despite growing mechanistic insight, clinical translation remains limited by poor reproducibility, insufficient strain-level resolution, overreliance on association studies, and single-species frameworks that neglect network behavior. Here, we synthesize recent advances in the molecular architecture of Bacteroides-mediated cross-feeding, including polysaccharide utilization loci (PULs), spatially deployed surface enzymes, and outer membrane vesicles (OMVs), as well as their ecological dynamics and functional consequences, and critically evaluate why these interactions fail to yield predictable clinical outcomes. We propose a shift toward network-based microbiome therapeutics, highlighting predictive metabolic modeling, cross-feeding-guided consortium design, improved spatiotemporal resolution, and targeted modulation of host-microbe interfaces as key future directions. The principal novelty of this review lies in reframing Bacteroides cross-feeding networks, rather than individual species, as the fundamental unit of microbiome therapeutics, and in tracing this network logic from molecular mechanisms through ecological dynamics to a concrete, experimentally addressable translational roadmap. Embracing cross-feeding networks as therapeutic units may enable more reproducible, mechanism-driven, and personalized microbiome interventions.
Chen et al. (Fri,) studied this question.