The review, led by researchers at Inner Mongolia Agricultural University, organizes decades of evidence into a Mechanism-Technology-Application framework designed to separate what has actually been demonstrated from what has merely been inferred. The authors are blunt about the distinction. Cross-feeding, as they define it, is any interaction in which a resource produced, released, or made accessible by one microorganism is acquired or further transformed by another. It is not inherently cooperative, not always beneficial to the host, and not a universal organizing principle of the gut ecosystem. The same resource transfer that sustains a health-associated butyrate producer can, in a different context, feed a dangerous pathogen.

The clearest examples come from sequential degradation, sometimes called a trophic relay. Complex dietary polysaccharides and host-derived mucus glycans are too large for most gut bacteria to touch. Specialist degraders such as Bacteroides species, Akkermansia muciniphila, and Bifidobacterium bifidum cleave these polymers into sugars and oligosaccharides that downstream consumers can use. In one well-characterized relay, Bacteroides ovatus releases breakdown products during inulin degradation that support B. vulgatus, B. thetaiotaomicron, and Parabacteroides goldsteinii. In another, Akkermansia muciniphila liberates mucin-derived sugars and acetate that fuel the butyrate producer Anaerostipes caccae, linking mucus turnover to the production of a short-chain fatty acid that nourishes the colonic lining and modulates immune signaling through receptors such as GPR43.

But relays have a dark side. In antibiotic-treated mice, the same liberation of host sugars becomes a lifeline for pathogens: sialic acid and fucose released from mucus by the depleted community boost the competitive fitness of Salmonella enterica serovar Typhimurium, while sialic acid catabolism supports the post-antibiotic expansion of Clostridioides difficile. Amino acid exchange tells a similar story. In antibiotic-perturbed mice, enterococci supply leucine and ornithine that enhance C. difficile fitness, and their arginine catabolism simultaneously generates metabolic signals associated with increased virulence. Metagenomic surveys suggest that more than 40 percent of gut microbes lack complete pathways for making branched-chain amino acids, implying that such dependencies are widespread, though predicted auxotrophy alone does not prove that transfer actually occurs.

Diffusible metabolites add another layer of coupling. Hydrogen produced by fermentative bacteria such as Roseburia intestinalis can be consumed by acetogens like Ruminococcus hydrogenotrophicus or by the methanogen Methanobrevibacter smithii, keeping fermentation thermodynamically favorable while shifting the end products toward acetate or methane. Elevated methane has been associated with delayed transit and constipation-predominant irritable bowel syndrome, although the review stresses that these associations do not establish hydrogen cross-feeding as a direct cause of disease. Perhaps the most elegant demonstrated exchange is bidirectional: Bacteroides xylanisolvens supplies folate to Clostridium butyricum, which in turn provides p-aminobenzoic acid that B. xylanisolvens needs to synthesize folate, a reciprocal loop shown under controlled culture conditions but not yet traced in a living gut.

Micronutrients reveal an even more competitive dimension. Bacteroides thetaiotaomicron makes no siderophores of its own yet steals iron from enterobactin and salmochelin secreted by Enterobacteriaceae using its XusABC system, a capacity that improves its colonization fitness during inflammation. Remarkably, the XusB protein captures enterobactin and shields it from the host’s iron-sequestering lipocalin-2, but Salmonella can then reacquire the protected complex. A single iron-trafficking route can therefore bolster a commensal, blunt host nutritional immunity, and still hand iron to a pathogen. Vitamin provisioning shows similar ambiguity: auxotrophic strains in humanized germ-free mice maintain stable abundances despite dietary swings, consistent with community-level sharing, yet the actual donors and transfer routes remain unidentified.

Given this complexity, the review proposes a tiered evidence framework that mirrors a Design-Build-Test-Learn engineering cycle. Multi-omics and computational platforms such as COMETS, MICOM, and the Gut Cross-feeding Predictor generate candidate donor-resource-recipient hypotheses, but correlation networks cannot establish causality. Causal attribution demands co-culture experiments, targeted metabolite perturbation, and genetic disruption of production or uptake pathways, exemplified by a study in which paired mutants of Bifidobacterium breve and Limosilactobacillus reuteri confirmed that a nutritional interaction depended on the intermediate 1,2-propanediol. Stable-isotope tracing then resolves metabolite fate directly: labeled acetate has been shown to be incorporated into butyrate by Faecalibacterium prausnitzii and Roseburia strains, and labeled lactate and acetate traced the carbon route from Bifidobacterium adolescentis into butyrate synthesis by Eubacterium hallii via acetyl-CoA.

Even validated pairwise exchanges, however, may not survive contact with a real community. In a twelve-member synthetic gut community, pairwise interactions explained only part of the observed dynamics, and initial abundances and negative interactions imposed strong historical contingency. Functional redundancy complicates matters further: depleting specific Bacteroides members triggers compensatory expansion of relatives that partially restores glycan degradation. Host-associated models add spatial architecture, dietary fluctuation, and colonization history, all of which can attenuate or redirect interactions that looked stable in a flask. The authors argue that no single method suffices; only convergent evidence across candidate identification, causal validation, and ecological robustness testing can support claims of functional cross-feeding.

The translational implications are sobering but instructive. Cross-feeding-informed synbiotics would require a primary responder to release a defined intermediate that a secondary consumer then transforms, with a combination-specific benefit demonstrated against substrate-only and microbe-only controls, a standard that current human evidence does not meet. Among live biotherapeutic products, the dual-strain pairing of B. xylanisolvens and C. butyricum improved metabolic and barrier parameters in high-fat-diet mice, but the folate exchange was never traced in vivo. The seventeen-strain consortium GUT-103 attenuated colitis in gnotobiotic mice without resolving any donor-metabolite-recipient chain, and a thirty-seven-strain synthetic microbiota suppressed C. difficile through competitive proline depletion rather than resource transfer at all. Engineered examples cut both ways: redirecting polysaccharide-utilization loci in Bacteroides uniformis altered glycan flow to downstream consumers, a relatively direct case of engineered cross-feeding, while the clinical candidate NOV-001 achieved porphyran-dependent, largely reversible engraftment in a Phase 1/2a trial yet showed variable colonization, genetic instability, and no significant benefit over placebo for urinary oxalate. The Phase 2b ECO-RESET trial of SER-287 similarly failed its primary endpoint despite confirmed engraftment, demonstrating that colonization is not efficacy.

The review’s conclusion is neither dismissive nor utopian. Cross-feeding offers a rigorous, mechanistically grounded framework for prioritizing dietary substrates, strain combinations, and engineered resource-allocation strategies, and emerging tools, from spatially resolved transcriptomics to gut-on-chip platforms and digital-twin modeling, should progressively narrow the gap between inference and demonstration. But a reproducible, closed engineering cycle for gut microbial communities has not been achieved across heterogeneous human hosts, and the immediate priority is prospective validation of the full causal chain linking substrate availability, microbial activity, metabolite transfer, ecological persistence, safety, and clinically meaningful benefit. Cross-feeding, the authors emphasize, must be evaluated within the competitive, host-regulated ecology of the gut rather than as an isolated, programmable switch. The microbial economy of the intestine is real, increasingly legible, and, in carefully defined cases, already manipulable. Turning that legibility into reliable medicine is the task the field now faces.

Subject of Research: Metabolic cross-feeding interactions in the gut microbiome and their application to live biotherapeutic development

Article Title: Programming Gut Microbiome Function Through Cross‐Feeding: From Ecological Mechanisms to Live Biotherapeutics

Article References: Sun, C., Ma, T., Jin, H., Li, Z., Cheng, Y., Li, J., Kwok, L.-Y., & Sun, Z. (2026). Programming Gut Microbiome Function Through Cross‐Feeding: From Ecological Mechanisms to Live Biotherapeutics. Advanced Science, 13(55), Article e77267. https://doi.org/10.1002/advs.77267

Image Credits: AI Generated

DOI: 10.1002/advs.77267

Keywords: gut microbiome, cross-feeding, microbial metabolism, live biotherapeutics, synbiotics, butyrate, short-chain fatty acids, synthetic communities, stable-isotope tracing, microbiome therapeutics, colonization resistance, precision nutrition

The research team focused on the concept of a phage replication threshold, the minimum density of susceptible host cells required for a lytic phage population to sustain itself. Below this density, the rate at which new phage particles are produced through infection and lysis falls short of the rate at which virions are lost through clearance, diffusion barriers, and failed encounters with hosts. Although this threshold concept was established decades ago in simplified liquid cultures, its value in the complex, spatially structured, and immunologically active environment of the intestinal tract had never been rigorously quantified. Because the density of any invading bacterial species in the gut depends heavily on the composition of the resident microbiota, the researchers hypothesized that increasing microbial complexity should suppress phage replication and, in turn, alter the selection for phage resistance.

To test this hypothesis, the team assembled an elegant experimental system built around two clinically relevant, multidrug-resistant Escherichia coli strains. The first, designated EcST73, was isolated from a healthy traveler and belongs to sequence type ST73; the second, EcST1193, came from a case of neonatal sepsis. Each strain produces a distinct polysaccharide capsule, or K-antigen, which serves as the receptor for a specific lytic phage: φYodit, a Kayfunavirus that binds the K5 capsule of EcST73, and φMoritz, a Vectrevirus that binds the K1 capsule of EcST1193. Both phages belong to the order Autographivirales and possess short, non-contractile tails. The choice of capsule-targeting phages carried a practical advantage: phage-resistant mutants that shed their capsules form visibly translucent colonies on salt-free agar plates, allowing the researchers to track the rise and fall of resistance directly from fecal samples without any genetic engineering of the target population.

The experimental backbone consisted of three mouse lines harboring microbiotas of increasing complexity. Antibiotic-treated mice, whose colonization resistance had been destroyed by ampicillin, permitted the E. coli strains to reach enormous densities, averaging around 10¹⁰ colony-forming units per gram of feces. Gnotobiotic mice carrying a low-complexity microbiota of only a few species supported intermediate colonization levels in the range of 10⁶ to 10⁷ CFU per gram. Conventional specific-pathogen-free mice, with their full complement of resident microbes, mounted strong colonization resistance and kept the invading strains at or below roughly 10⁵ CFU per gram, with EcST1193 in fact being cleared from most animals entirely. Bacterial populations stabilized about 48 hours after oral infection, so the researchers administered a single oral dose of 10⁹ plaque-forming units of the matching phage at precisely that time point.

The results were unambiguous. In antibiotic-treated and low-complexity microbiota mice, phage titers rose and remained detectable for several days, evidence of active replication. In conventional mice, by contrast, phages became undetectable within 48 hours in most animals, indicating that the small E. coli population could not sustain viral reproduction. Strikingly, the phages had no significant effect on total E. coli loads in any of the models. Where replication did occur, resistance followed rapidly. In antibiotic-treated animals, capsule-defective mutants appeared within a day of phage administration and soon dominated the population, after which phage titers declined as susceptible hosts vanished. Sequencing of these resistant clones revealed mutations in genes governing capsule synthesis, transport, and regulation, including emrR, a transcriptional regulator of capsule production, kpsM, required for capsule polysaccharide transport, and, in EcST1193, the neuS and neuE genes involved in polysialic acid biosynthesis for the K1 capsule. Many clones also carried mutations in maltose metabolism genes, a signature of E. coli adaptation to the antibiotic-treated gut.

When the researchers plotted phage titers against bacterial densities across all animals, a clear threshold emerged. Phage replication and the consequent selection for capsule-defective mutants occurred only when susceptible bacteria exceeded approximately 10⁶ CFU per gram of feces, equivalent to roughly 1.3 × 10⁶ bacteria per milliliter of cecal content. Below this line, no resistant mutants were ever observed, because phages simply could not replicate and therefore exerted no selective pressure. Notably, this in vivo threshold is about a hundred times higher than the proliferation threshold of 10⁴ bacteria per milliliter previously determined in vitro for phage T4 and E. coli. The authors attribute this discrepancy to the harsh realities of the intestinal environment: viscous contents impede phage diffusion, structural heterogeneity creates microenvironments where susceptible cells hide, off-target adsorption diverts virions, intestinal flow washes particles away, epithelial cells uptake phages, and digestive enzymes and bile salts inactivate them. Furthermore, fecal counts may overestimate the fraction of cells that are actively growing and expressing the phage receptor, since capsule production is often heterogeneously expressed in Enterobacteriaceae.

Perhaps the most consequential finding came from a competition experiment designed to expose a hidden subpopulation of susceptible cells. The researchers co-infected antibiotic-treated mice with a one-to-one mixture of wild-type EcST73 and a phage-resistant capsule-knockout mutant, each tagged with a distinct antibiotic-resistance marker so the two could be quantified separately. Without phage, the two strains coexisted stably. With phage φYodit, the wild-type population crashed by several orders of magnitude but then stabilized at around 2.4 × 10⁶ CFU per gram, remarkably close to the replication threshold. The phage went extinct in six of ten animals, and after viral extinction the susceptible population rebounded in some mice. Crucially, mini-plaque assays confirmed that most surviving wild-type cells remained fully phage-susceptible even after ten days of coexistence with the virus. In other words, the microbiota is not the only force that can shelter susceptible bacteria; a dense population of resistant cells is sufficient on its own to dilute phage spread below the critical threshold.

This observation carries sobering implications for phage therapy. The researchers showed that flooding the gut with phages at concentrations hundreds of times higher than the susceptible population, roughly 2 × 10⁹ virions against 5 × 10⁶ bacteria per gram of feces, still failed to eliminate the targets. Passive therapy, in which an overwhelming excess of phages drives host extinction without secondary rounds of replication, is therefore unlikely to achieve complete decolonization of the intestine. Instead, phages reduce susceptible populations to a residual level just beneath the replication threshold, where they persist indefinitely and can regrow once phages disappear. The characteristics of this refuge population and the niche it occupies remain open questions, but its existence helps explain why phage cocktails in previous mouse and calf studies reduced, yet never eradicated, intestinal loads of E. coli and Klebsiella pneumoniae.

For clinical translation, the threshold offers a practical decision rule. Human intestinal loads of E. coli vary between 10² and 10⁹ CFU per gram of feces, so phages should not be expected to replicate, or to select for resistance, in patients carrying fewer than 10⁶ bacteria per gram. Conversely, patients whose microbiotas have been disturbed by antibiotics or pathogen-driven dysbiosis often harbor densities well above the threshold, making them plausible candidates for successful phage treatment. Even a transient reduction in bacterial load may suffice to improve outcomes and reduce mortality, as multiple animal studies have demonstrated. The authors acknowledge limitations: only two phage-bacterium pairs and a single phage dose were tested, and both phages were capsule-specific Autographivirales whose properties might influence the threshold value. Nevertheless, comparable effects have been reported with diverse phages using other receptors, suggesting that the microbiota-constrained replication threshold may be a general feature of phage-bacteria interactions in the gut, one that must be respected if phage therapy is to fulfill its promise as a precision tool for microbiota engineering.

Subject of Research: Phage replication threshold in the mammalian gut and its dependence on microbiota-constrained bacterial density

Article Title: Microbiota-constrained bacterial density limits phage infection in the gut and allows persistence of susceptible cells

Article References: Bertola, A., Wenner, N., Rocha, L. L., Keys, T. G., & Diard, M. (2026). Microbiota-constrained bacterial density limits phage infection in the gut and allows persistence of susceptible cells. iScience, 29(11), Article 117433. https://doi.org/10.1016/j.isci.2026.117433

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117433

Keywords: bacteriophage, phage therapy, gut microbiota, Escherichia coli, replication threshold, phage resistance, capsule, colonization resistance, antibiotic-resistant bacteria, Enterobacteriaceae, mouse model, microbiota engineering

The review is careful to separate what is established from what remains hypothetical, a distinction that matters in a field where observational data can easily be overinterpreted. The authors distinguish three clinical scenarios that are often conflated in the literature. The first is simultaneous co-infection, in which a patient harbors both SARS-CoV-2 and actively toxin-producing C. difficile at the same time. The second is CDI that develops during hospitalization for COVID-19, a situation in which antimicrobial exposure and disrupted infection-prevention practices during pandemic surges may play dominant roles. The third, and perhaps the most intriguing, is delayed or recurrent CDI after apparent recovery from acute infection, which points toward persistent alterations of the intestinal ecosystem rather than simple nosocomial exposure. Each scenario implies a different balance of contributing causes, and each demands a different study design to disentangle.

The most straightforward explanation for an association between the two diseases is iatrogenic. Patients hospitalized with severe COVID-19 frequently received broad-spectrum antibiotics, either because bacterial co-infection was suspected or as empirical coverage during the chaotic early phases of the pandemic, when distinguishing viral pneumonia from secondary bacterial infection was difficult. Antibiotic exposure is the single strongest risk factor for C. difficile infection because it dismantles colonization resistance, the collective ability of a diverse gut microbiota to suppress pathogen expansion through competition for nutrients, production of bacteriocins, and, critically, the metabolism of bile acids. Primary bile acids such as taurocholate stimulate germination of C. difficile spores, while secondary bile acids generated by commensal bacteria inhibit its growth. When antibiotics deplete the bile acid-metabolizing community, the gut becomes a permissive niche for the pathogen.

Yet the review argues that antibiotics and hospitalization alone cannot fully account for the observed patterns. SARS-CoV-2 is increasingly recognized as an agent capable of perturbing the gut directly. Viral RNA has been detected in intestinal tissue and stool, and the angiotensin-converting enzyme 2 receptor, the cellular entry gate for the virus, is abundantly expressed on intestinal epithelial cells, particularly mature enterocytes of the small intestine. The authors emphasize that ACE2 is not merely a viral docking site. In normal physiology, ACE2 participates in regulating amino acid homeostasis, antimicrobial peptide expression, and the composition of the gut microbiome. Loss or dysregulation of ACE2 function during infection could therefore weaken epithelial barrier integrity and shift the microbial community toward a dysbiotic state, independently of any antimicrobial exposure.

This leads to the renin-angiotensin system, one of the review’s central mechanistic themes. SARS-CoV-2 binding to ACE2 promotes internalization and degradation of the receptor, reducing its protective enzymatic activity. The consequence is an imbalance in the renin-angiotensin axis: less angiotensin II is converted into the vasoprotective angiotensin-(1-7), and unopposed angiotensin II signaling through the AT1 receptor increases. In the intestine, excessive angiotensin II/AT1R signaling is associated with inflammation, increased epithelial permeability, and impaired barrier repair. The authors propose that this signaling imbalance could amplify mucosal damage during COVID-19 and thereby lower the threshold at which residual C. difficile spores, persisting in the lumen after germination conditions improve, can establish infection and produce toxin.

Downstream of angiotensin signaling, the review highlights inflammatory cascades that connect viral injury to bacterial opportunity. Severe COVID-19 is characterized by activation of the transcription factors NF-κB and MAPK in epithelial and immune cells, driving robust production of interleukin-6 and other pro-inflammatory mediators. Persistent inflammatory signaling can disrupt epithelial tight junctions, alter mucus production, and change the metabolic landscape of the lumen in ways that favor pathogenic taxa over obligate anaerobes that maintain colonization resistance. The authors also draw attention to nutritional vulnerability, noting that malnutrition and micronutrient deficiencies, common in hospitalized and older patients, impair epithelial regeneration and immune defense. They further discuss apolipoprotein E-related immunometabolic pathways as a possible link between host lipid metabolism, inflammatory resolution, and susceptibility to intestinal injury, an area they flag as requiring direct experimental validation.

Among the more forward-looking elements of the review is its treatment of extracellular vesicles as a potential biomarker framework. Extracellular vesicles, small membrane-bound particles released by cells into circulation and bodily fluids, carry proteins, lipids, and nucleic acids that reflect the physiological state of their cell of origin. The authors suggest that vesicle cargo could eventually serve as a readout of the combined inflammatory, epithelial, and microbial perturbations that characterize patients at heightened risk of CDI during or after COVID-19. Such biomarkers could, in principle, allow clinicians to identify which recovered COVID-19 patients warrant closer surveillance for delayed intestinal complications, although this remains a hypothesis-generating proposal rather than a validated clinical tool.

The authors are explicit that direct longitudinal evidence linking SARS-CoV-2 infection to subsequent CDI remains limited. Much of the available data come from single-center cohorts, retrospective chart reviews, and studies with heterogeneous testing practices, which complicates interpretation. Distinguishing true C. difficile infection from colonization in patients with diarrhea of other causes is a persistent diagnostic challenge, and pandemic-era disruptions to infection-prevention bundles and testing algorithms further muddy the picture. What the review offers instead is convergence: multiple independent lines of clinical and experimental evidence that all point in the same direction, supporting what the authors describe as a cautious, testable hypothesis rather than a demonstrated causal relationship.

To move the field forward, the review calls for prospective studies with standardized CDI testing protocols, careful quantification of antimicrobial exposure, and integrated microbiome and metabolomic profiling of patients followed from acute COVID-19 through recovery. Such studies would need to track bile acid pools, inflammatory markers, and microbial community structure over time to determine whether SARS-CoV-2 infection produces a durable shift in colonization resistance, and whether that shift translates into elevated risk of primary or recurrent CDI. Mechanistic biomarker discovery, including the extracellular vesicle framework, would ideally proceed in parallel. The authors also underline the importance of clarifying prognosis and long-term consequences, since recurrent CDI carries substantial morbidity and mortality in the same older and medically complex populations that bore the heaviest burden of the pandemic.

The broader significance of the work lies in its reframing of COVID-19 as a condition with potential gastrointestinal sequelae mediated through the microbiome, rather than a purely respiratory illness. If the hypothesis holds, the implications extend beyond the acute pandemic: clinicians managing patients recovering from severe COVID-19 may need to weigh antibiotic stewardship more carefully, monitor for gastrointestinal symptoms after discharge, and consider gut-protective strategies such as targeted microbiome restoration. At the same time, the review’s disciplined separation of established observations from mechanistic speculation serves as a model for how the field should proceed. The interaction between a global respiratory pathogen and an opportunistic intestinal spore-former is almost certainly multifactorial, involving antimicrobial exposure, hospitalization dynamics, dysbiosis, barrier dysfunction, altered bile acid metabolism, and persistent inflammation. Untangling those threads will require exactly the kind of rigorous, longitudinal, mechanistically informed research that the authors advocate.

Subject of Research: Potential mechanisms linking COVID-19 and Clostridioides difficile infection through gut dysbiosis

Article Title: COVID-19 and clostridioides difficile infection: clinical overlap, gut dysbiosis, and mechanistic hypotheses

Article References: de Carvalho Leitão, R. F., da Silva Costa, D. V., Nguyen, A., Vieira, L. L., Nunes, P. I. G., Warren, C. A., Oriá, R. B., & de Castro Brito, G. A. (2026). COVID-19 and clostridioides difficile infection: clinical overlap, gut dysbiosis, and mechanistic hypotheses. Gut Pathogens. https://doi.org/10.1186/s13099-026-00862-0

Image Credits: AI Generated

DOI: 10.1186/s13099-026-00862-0

Keywords: COVID-19, Clostridioides difficile, SARS-CoV-2, ACE2, gut dysbiosis, renin-angiotensin system, bile acid metabolism, colonization resistance, inflammation, extracellular vesicles, hospitalization, antibiotic exposure

The empirical foundation for this view was laid by large-scale laboratory screens, most notably a 2018 study in which the team around Lisa Maier and Kiran Patil tested roughly 1,000 marketed drugs against representative strains of human gut bacteria and found that about a quarter inhibited at least one commensal microbe. Subsequent work extended the map beyond pharmaceuticals. Recent screens have shown that industrial and agricultural chemicals also suppress gut bacteria in vitro, and that some per- and polyfluoroalkyl substances, the persistent so-called forever chemicals, are actively bioaccumulated by human gut microbes. Even dietary xenobiotics, transformed by resident bacteria, can restructure microbial communities. The boundary between the pharmacy, the food supply and the chemical environment, the review makes clear, is biologically porous.

What distinguishes non-antibiotic antimicrobials from true antibiotics is not the fact of antibacterial activity but its spectrum and potency. Unlike antibiotics, which are typically active at low concentrations and target a broad range of organisms, non-antibiotic antimicrobials tend to act at higher doses and against a narrower set of microbes. Paradoxically, that narrow set often excludes the Enterobacteriaceae, the family containing many classic pathogens for which new antibiotics are urgently needed, while hitting beneficial commensals harder. This inverted selectivity means that everyday exposure to these compounds can quietly erode the protective, health-associated fraction of the microbiome while leaving opportunistic pathogens relatively unscathed, creating ecological vacancies that resistant or disease-associated organisms can exploit.

The molecular mechanisms underlying these effects are only beginning to be resolved, but several recurring strategies have emerged. Some non-antibiotics share drug targets with classical antibiotics or, more intriguingly, bind bacterial counterparts of the human proteins they were designed to hit. Others mimic essential metabolites, poisoning microbial enzymes through molecular masquerade, or hijack bacterial transport systems to gain entry into cells. Individual examples are illuminating: the artificial sweetener saccharin has been shown to disrupt bacterial cell envelope stability and interfere with DNA replication dynamics; nonsteroidal anti-inflammatory drugs target DNA replication; the antidiabetic drug acarbose impairs gut Bacteroides growth by inhibiting intracellular glucosidases; and the Parkinson’s drug entacapone disrupts gut microbial homeostasis through iron sequestration. Yet for the majority of non-antibiotic antimicrobials, the microbial targets remain unknown, a knowledge gap the authors identify as a central obstacle to rational risk assessment.

Crucially, the antimicrobial activity of these compounds cannot be understood in isolation from the host. Microbiome composition varies enormously between individuals, and a drug that devastates one person’s community may barely perturb another’s. Host conditions further modulate outcomes: intestinal pH, bile salt exposure, nutrient availability, body temperature and the community’s biogeography all alter both drug activity and microbial susceptibility. Proton-pump inhibitors, for example, appear to increase the risk of Clostridioides difficile infection primarily by altering gut pH rather than through direct effects on the microbiome, a reminder that host-conditioned mechanisms can masquerade as microbiome effects. Conversely, some drugs act on the host in ways that reshape the gut environment, as with corticosteroids that impair mucin production, or statins whose microbiome-linked effects on metabolic health are still being mapped.

The clinical consequences of these interactions are now measurable in populations. Population-level metagenomic studies have shown that medications leave distinct, reproducible signatures on the gut microbiome, and that drugs taken years before sampling can still act as hidden confounders in microbiome research. One 2025 study reported that non-antibiotics disrupt colonization resistance against enteropathogens, and a companion analysis identified medication-microbiome interactions that affect gut infection outcomes. At the same time, the microbiome can alter drug efficacy in both directions: gut bacterial tyrosine decarboxylases deplete levodopa in Parkinson’s disease, microbial metabolism of the anti-inflammatory mesalazine diminishes its benefit in inflammatory bowel disease, and microbiome-derived metabolites such as inosine and 3-indole-3-acetic acid modulate responses to cancer immunotherapy and chemotherapy. Drug, microbe and host form a three-body problem, the review argues, that medicine has traditionally treated as a two-body one.

Perhaps the most alarming dimension is the contribution of non-antibiotic antimicrobials to the spread of antibiotic resistance. Laboratory and animal studies have shown that antidepressants, antipsychotics, antiepileptic drugs and artificial sweeteners can all promote the horizontal transfer of antibiotic resistance genes, accelerate plasmid conjugation, induce mutagenesis or select for efflux-based resistance in Escherichia coli. Because these compounds are consumed chronically by billions of people and persist in the environment, they may exert a continuous, diffuse selection pressure that quietly amplifies the resistome in human guts and beyond, independent of any antibiotic prescription. The review frames this as a planetary-scale problem, noting that microbiomes across human, animal and environmental habitats are connected by gene flow, so resistance selected anywhere can eventually matter everywhere.

Dietary chemicals occupy a particularly contested middle ground in this landscape. Emulsifiers such as carboxymethylcellulose have been shown in controlled-feeding studies to damage the human gut microbiota and metabolome, while artificial sweeteners have been reported to induce glucose intolerance in mice and personalized, microbiome-dependent glucose responses in humans, though some trials of saccharin and sucralose in healthy adults have found no detectable microbiome effects at all. Pesticide exposure has been associated with altered gut microbiota and metabolites in observational studies of both the general population and occupationally exposed workers. This inconsistency, the authors suggest, reflects the context-dependence that pervades the entire field: identical exposures can yield different outcomes depending on a person’s baseline microbiome, diet, host physiology and co-exposures to other xenobiotics.

The translational implications are substantial. The authors call for integrated stewardship of all xenobiotics, not just antibiotics, across medicine, agriculture, industry and the environment. Practically, that means incorporating microbiome effects into drug development and safety testing, using new high-throughput anaerobic screening methods and defined microbial community models to identify antimicrobial activity before compounds reach patients, deploying chemical-genetic and proteomic tools to pin down microbial targets, and developing computational models of xenobiotic metabolism that can predict how individual microbiomes will respond to given drugs. For clinicians, the immediate lesson is more modest but no less important: non-antibiotic prescriptions are microbiome interventions, and polypharmacy should be recognized as a combinatorial, dose-dependent perturbation of the gut ecosystem.

What the review ultimately delivers is a reframing. Antimicrobial activity, once considered the defining property of a specialized class of drugs, is revealed to be a generic hazard of the chemical world that modern humans inhabit, arising because human and bacterial biochemistry share enough common machinery that molecules designed for one can wound the other. The benefits of these compounds, in treating psychiatric illness, diabetes, cancer and countless other conditions, are real and immense, and the authors are careful not to argue that non-antibiotic antimicrobials should be abandoned. The task, they conclude, is to move from unwitting to deliberate management: to know which compounds disturb the microbiome, through which mechanisms, in whom, and with what downstream consequences for infection risk, chronic disease and resistance, so that the hidden chemistry between our drugs, our food, our environment and our microbes can finally be conducted with open eyes.

Subject of Research: Antimicrobial effects of non-antibiotic compounds on the human gut microbiome and host health

Article Title: Drug–microbiome–host interactions: antimicrobial effects of non-antibiotic compounds

Article References: de la Cuesta-Zuluaga, J., Patil, K. R., & Maier, L. (2026). Drug–microbiome–host interactions: antimicrobial effects of non-antibiotic compounds. Nature Reviews Gastroenterology & Hepatology. https://doi.org/10.1038/s41575-026-01258-w

Image Credits: AI Generated

DOI: 10.1038/s41575-026-01258-w

Keywords: gut microbiome, non-antibiotic drugs, antimicrobial resistance, microbiome–drug interactions, artificial sweeteners, pesticides, colonization resistance, pharmacomicrobiomics, xenobiotics, dysbiosis, horizontal gene transfer, drug metabolism