The starting point for the framework is a deceptively simple observation: the gut microbiome of social bees is remarkably small and specific compared with that of mammals. Honey bees and bumble bees harbour a core set of bacterial symbionts that have coevolved with their hosts and are transmitted socially, passing between nestmates through shared food, faecal contact and the communal environment of the hive. These microbes contribute to nutrition by aiding the digestion of pollen, to detoxification of harmful compounds, to immune priming against pathogens, and even to behaviour, with measurable consequences for colony performance and resilience. Yet despite the field’s rapid growth, most studies continue to analyse host genetics and environmental exposures separately, and many rely on simple correlations between the abundance of particular microbes and health traits such as parasite resistance or longevity.

That reliance on correlation is where the trouble begins, the authors argue. When researchers observe that bees with a certain microbiome profile are healthier, it is difficult to know what is driving what. The microbiome differences may reflect underlying host genetics, they may be a consequence of environmental exposure, or they may genuinely contribute mechanistically to the health outcome. Disentangling these possibilities is one of the central challenges in bee microbiome research, and the GEM–P framework is designed precisely to make this disentangling more systematic. It treats the genome, the exposome and the gut microbiome as interacting components that jointly shape phenotypes, while phenotypes themselves feed back to modify subsequent exposures and microbiome states, creating a dynamic, looping system rather than a one-directional chain of cause and effect.

Each component of the framework carries specific technical meaning. The genome encompasses host genetic variation, which can filter which microbes are able to colonise the gut, a process the authors describe as genetic filtering. The exposome captures the totality of lifetime exposures, including diet, pesticides, pathogens, temperature and landscape characteristics, all of which can reshape the microbial community or act directly on bee physiology. The microbiome refers to the gut symbionts themselves and their assembly dynamics and functions. The phenome covers the full suite of observable traits, from individual characteristics such as immune gene expression and nutrient processing to colony-level outcomes such as brood production, overwintering success and disease burden. Crucially, the framework operates at both scales simultaneously, recognising that colony phenotypes emerge from the actions and interactions of thousands of individuals.

Perhaps the most intellectually important contribution of the review is its treatment of context dependence. The authors highlight three major sources of it: scale, timing and social transmission. Similar microbiome shifts can be associated with different trait outcomes depending on the life stage and task of the bee, since nurse bees and foragers, or larvae and adults, have distinct physiologies and diets. Exposure history matters too, because a bee that has previously encountered a pesticide may respond differently to a microbial change than a naive individual. And the colony’s transmission structure, the way microbes are shared among nestmates, means that an individual’s microbiome is inseparable from its social environment. A microbial pattern that signals dysbiosis in one context may be an adaptive response in another, which helps explain why apparently contradictory findings recur across the literature.

To illustrate the framework’s utility, the authors use GEM–P to map existing evidence on genetic filtering, exposomal drivers, microbiome assembly and function, and phenotypic outcomes. This mapping exercise reveals where the evidence is strong and where it remains thin. For example, host genetics clearly constrain which symbionts can establish in the bee gut, and dietary exposures demonstrably alter the relative abundance of core taxa, but the mechanistic links connecting specific microbial functions to specific colony outcomes are far less well established. By laying the components out in a single organisational scheme, researchers can identify which pathway connecting genome, exposome, microbiome and phenotype is being invoked in any given hypothesis, and design sampling strategies that test that pathway specifically rather than relying on broad correlational surveys.

Importantly, the authors stress that generating pathway-specific hypotheses does not require comprehensive multi-omics approaches, which remain expensive and technically demanding for many laboratories. Instead, GEM–P supports relatively simple, targeted study designs. Sampling across life stages and castes, recording exposure histories, accounting for colony-level transmission, and measuring phenotypes at matched individual and colony scales can all help distinguish among alternative biological explanations for a microbiome–phenotype association without sequencing everything. The framework thus serves as a guide for experimental economy, pointing researchers toward the comparisons that are most informative for ruling out competing interpretations, such as whether a microbiome shift precedes or follows a change in host state, or whether it is shared across related bees in a way consistent with genetic filtering.

The implications extend beyond basic science into the practical world of beekeeping and pollination services. Social bees underpin agriculture through their pollination of crops and wild plants, and colony losses have economic and ecological consequences that are now felt worldwide. If microbiome-based interventions, such as probiotic supplements or management practices that support beneficial symbionts, are to be developed responsibly, they must rest on an understanding of when microbial change actually causes improvements in health and when it is merely a bystander. GEM–P offers a way to formulate those questions rigorously, linking interventions to specific points in the genome–exposome–microbiome–phenome network and specifying the outcomes and contexts in which effects should be expected. The work also contributes to the objectives of the BeeSustain project, an initiative on integrative modelling for enhanced beekeeping carrying capacity funded through Xjenza Malta’s Research Excellence Programme.

The review, published open access on 3 September 2026, arrives at a moment when the bee microbiome literature has grown rapidly but interpretively fragmented, with different subfields emphasising different components of the system and rarely integrating them. By providing a shared vocabulary and a common structural map, the authors hope the framework will help researchers interpret microbiome–phenotype associations across contexts, generate hypotheses that are explicitly pathway-specific, and design studies whose results genuinely advance mechanistic understanding. For a field in which correlation has often been mistaken for causation, and in which the same microbial signal can mean different things in a larva, a nurse bee or an overwintering colony, that kind of conceptual discipline may prove as valuable as any single experiment. The GEM–P framework does not claim to answer how social bee health is determined, but it tells the research community, with unusual clarity, exactly which questions to ask next.

Subject of Research: An integrative framework linking genome, exposome, gut microbiome and phenome in social bees

Article Title: The Genome Exposome Microbiome–Phenome (GEM–P): a conceptual framework for social bees

Article References: Cutajar, S., Alberoni, D., Di Gioia, D., & Mifsud, D. (2026). The Genome Exposome Microbiome–Phenome (GEM–P): a conceptual framework for social bees. Applied Microbiology and Biotechnology. https://doi.org/10.1007/s00253-026-14016-4

Image Credits: AI Generated

DOI: 10.1007/s00253-026-14016-4

Keywords: social bees, gut microbiome, exposome, genome, phenome, GEM–P framework, colony health, microbial transmission, symbiosis, bee resilience, host genetics, context dependence