Host Sensing of Microbiota-Associated Metabolites in Gastrointestinal Graft-Versus-Host Disease: Cell-Specific Mechanisms and Pharmacological Opportunities

Abstract

Background: Gastrointestinal graft-versus-host disease (GI-GVHD) remains a major cause of morbidity and mortality after allogeneic hematopoietic cell transplantation. Conditioning directly injures the intestinal epithelium, whereas antibiotics, nutritional disruption, and immunosuppression remodel the intestinal ecosystem; together, these exposures are associated with loss of obligate anaerobes and altered availability of microbiota- associated metabolites. This review examines how host tissues sense these metabolic changes and translate them into divergent biological outcomes. Methods: We integrate evidence on directly produced microbial metabolites, including short-chain fatty acids, tryptophan-derived indoles, and desaminotyrosine, as well as microbially modified bile acids, host–microbial co-metabolites such as trimethylamine N-oxide, and emerging microbiota-associated compounds. We then examine receptor-mediated pathways involving FFAR2/3, HCAR2, AhR, PXR, FXR, TGR5, and VDR, as well as receptor-independent mechanisms such as HDAC inhibition and mTORC1-STING-dependent regeneration. Results: We propose that metabolite activity in GI-GVHD is determined not by abundance alone, but by chemical form, concentration, tissue distribution, sensor identity, responding cell type, and disease stage. This framework explains how a metabolite may reinforce epithelial tolerance while promoting inflammatory myeloid or donor T-cell programs in another compartment. We further evaluate postbiotic replacement, receptor modulation, precision nutrition, defined microbial consortia, and fecal microbiota transplantation, highlighting pharmacokinetic, pharmacodynamic, infectious, and graft-versus-leukemia considerations. Conclusions: Cell- and compartment-resolved metabolite pharmacology may enable tissue-protective strategies that limit GI-GVHD without broadly suppressing antitumor immunity.

1. Introduction

1.1. Clinical Relevance of Gastrointestinal GVHD

Allogeneic hematopoietic cell transplantation (allo-HCT) can provide definitive treatment for numerous malignant and nonmalignant hematologic diseases, yet graft-versus-host disease (GVHD) continues to limit its overall benefit. This review specifically addresses gastrointestinal graft-versus-host disease occurring after allogeneic hematopoietic cell transplantation; GVHD arising after solid-organ transplantation or transfusion is outside its primary scope. In GVHD, immunocompetent donor cells recognize recipient antigens and mediate tissue injury [1]. The gastrointestinal tract is both a major target of acute GVHD and a site at which local damage can intensify systemic alloimmunity [1,2]. Injury caused by conditioning compromises tight junctions, the mucus layer, intestinal stem cells (ISCs), and Paneth cells. Damage-associated signals and translocated microbial products then activate host and donor antigen-presenting cells (APCs), expand alloreactive donor T-cells, and increase inflammatory cytokine release. Cytotoxic donor T-cells further injure the crypt epithelium, creating a feed-forward process in which barrier disruption, microbial translocation, and immune activation perpetuate one another. Concurrent loss of regenerative pathways, including interleukin-22-mediated support of ISCs, further weakens epithelial resistance to alloimmune injury [3,4]. Systemic corticosteroids remain the accepted first-line therapy for acute GVHD, although responses are frequently inadequate, especially in severe gastrointestinal involvement [1]. Intensifying immunosuppression may increase opportunistic infection, delay immune reconstitution, facilitate malignant relapse, and raise treatment-related mortality. Nonspecific inhibition of donor immunity can also attenuate the therapeutically desirable graft-versus-leukemia (GVL) response. These trade-offs motivate approaches that reinforce epithelial repair and tissue tolerance without broadly disabling antitumor immunity. The intestinal microbiota and its metabolites are particularly relevant in this context because they jointly influence barrier restoration, innate activation, antigen presentation, and donor T-cell behavior.

1.2. From Microbial Composition to Microbial Metabolic Function

Allo-HCT imposes profound and sustained ecological stress on the intestinal microbiota. Conditioning, broad-spectrum antimicrobials, diminished enteral feeding, dietary change, mucosal injury, and immunosuppressive treatment together reshape community structure. Across clinical cohorts, low microbial diversity, depletion of obligate anaerobes, and overgrowth by facultative pathogens have been linked to bacteremia, severe GVHD, and transplant-related death [5,6,7,8,9]. By contrast, a greater abundance of commensals such as Blautia correlates with lower GVHD-related mortality, whereas lactose-dependent Enterococcus expansion exacerbates GVHD in experimental models and is associated with adverse outcomes in clinical cohorts [7,8]. These observations identify dysbiosis as an important feature of allo-HCT outcomes, but taxonomic profiles alone do not capture the full biology of microbiota–host communication. Distantly related organisms may share metabolic capabilities, while strains assigned to one species can differ markedly in substrate use and metabolite output.

Strain-level functional variation is further shaped by genomic plasticity and ecological adaptation. Accessory genes, plasmids, prophages, pathogenicity islands, and other mobile genetic elements can confer distinct metabolic, antimicrobial-resistance, colonization, and virulence traits among strains assigned to the same species. Horizontal gene transfer contributes to the acquisition and dissemination of these functions, whereas phase variation, epigenetic regulation, and the rewiring of stress-response or two-component regulatory systems can alter microbial phenotypes without requiring conventional gene acquisition. Niche specialization further determines whether these traits support dietary glycan utilization, mucosal persistence, or exploitation of an inflamed intestinal environment. Consequently, commensal and pathogenic behavior should be understood as a context-dependent continuum rather than as a fixed taxonomic property. In allo-HCT recipients, antibiotics, epithelial injury, altered nutrient availability, inflammation, and immunosuppression may act as ecological filters that select for or unmask pre-existing pathobiont-associated functions. Accordingly, strain-resolved genomic and functional analyses are required to distinguish taxonomic persistence from preservation of beneficial metabolic activity. However, direct evidence that de novo horizontal gene transfer or strain-level genomic evolution causally drives GI-GVHD remains limited; these concepts should therefore be regarded as an ecological and evolutionary framework for future investigation rather than as established GI-GVHD mechanisms [10].

Here, “microbiota-associated metabolites” encompasses directly produced microbial metabolites, microbially modified host molecules, and host–microbial co-metabolites. These categories distinguish microbial production from the microbial transformation of host substrates and subsequent host conversion of microbial precursors. Short-chain fatty acids (SCFAs), especially butyrate and propionate, contribute to colonocyte bioenergetics, histone acetylation, junctional stability, and immune regulation through receptor-mediated and intracellular mechanisms [11,12]. Bacterial tryptophan products can engage metabolite sensors and protective interferon-related programs, thereby reducing epithelial leakiness and experimental GVHD [13]. In parallel, microbial conversion of primary bile acids produces unconjugated and secondary species with distinct activity at nuclear and membrane receptors. GVHD-associated T-cell inflammation suppresses these bacterial transformations, and the altered bile acid milieu can in turn reinforce alloimmune activation [14,15]. Dysbiosis therefore encompasses more than shifts in individual taxa; it also involves depletion, redistribution, or the maladaptive remodeling of community metabolic functions.

Recent mechanistic work further demonstrates that host responses to microbial cues are organized by cellular compartment. Microbiota-dependent regulation of epithelial major histocompatibility complex class II can alter GVHD independently of donor–recipient genetic disparity [16], and microbial antigenic inputs can influence the clonal repertoire selected from donor T-cells after transplantation [17,18]. The aromatic amino acid product desaminotyrosine, meanwhile, stimulates intestinal repair through host mTORC1- and STING-dependent programs rather than a conventional metabolite receptor [19]. These findings caution against assigning an invariant protective or harmful label to a metabolite. Biological activity instead reflects its chemical species and local exposure together with sensor availability, target-cell identity, anatomical distribution, and phase of GVHD.

This review examines how allo-HCT-associated ecological disruption changes host exposure to microbiota-associated metabolites, how intestinal epithelial, myeloid, innate lymphoid, and donor T-cell compartments sense these signals, and how the resulting pathways may be pharmacologically manipulated to prevent or treat acute GI-GVHD. We specifically distinguish human clinical associations, causal evidence from GVHD models, direct receptor or pathway validation, and mechanisms extrapolated from nontransplant disorders.

1.3. Scope and Evidence Framework

Throughout this review, the language used to describe microbiota–metabolite–GVHD relationships is aligned with the strength of the underlying evidence. Findings from human cohorts and multi-omics studies are described as associations unless supported by an interventional design. Metabolite supplementation, depletion, microbial transfer, or pharmacological perturbation in GVHD models is interpreted as evidence that an intervention modifies disease severity, but does not by itself establish that endogenous metabolite deficiency initiates GVHD. Statements that a receptor or signaling pathway is required for, or mediates, a biological effect are reserved for receptor-specific genetic manipulation, pharmacological blockade with evidence of target engagement, or rescue and epistasis experiments. Evidence derived from non-GVHD intestinal models is used only to support biological plausibility and is explicitly identified as mechanistic extrapolation or a candidate pathway. Accordingly, reduced metabolite abundance alone is not considered evidence that metabolite loss causes GI-GVHD.

This narrative review primarily addresses acute gastrointestinal GVHD because the most substantial mechanistic and clinical evidence linking the microbiota, microbiota-associated metabolites, and host-sensing pathways has been generated in this setting. Evidence from chronic GVHD is included only when a direct relationship with microbial composition or metabolite exposure has been demonstrated, such as the association between circulating or fecal SCFAs and protection from chronic GVHD. Mechanistic observations from inflammatory bowel disease, infection, metabolic disease, or steady-state intestinal homeostasis are used only to clarify plausible signaling pathways and are explicitly distinguished from direct GVHD evidence.

The relevant literature, published through 31 July 2026, was considered using combinations of terms related to “graft-versus-host disease,” “gastrointestinal GVHD,” “gut microbiota,” “microbial metabolites,” “short-chain fatty acids,” “bile acids,” “tryptophan metabolites,” “indoles,” “metabolite-sensing receptors,” and “postbiotics.” Evidence is summarized along three complementary dimensions: human clinical findings, experimental disease-modification evidence, and sensor or pathway validation. These dimensions are considered separately rather than combined into a single hierarchy.

Accordingly, this review traces the transition from allo-HCT-associated ecological disruption to microbial metabolic dysfunction, defines the major receptor-dependent and receptor-independent sensing pathways, and examines how these signals are integrated by intestinal epithelial cells, myeloid cells, innate lymphoid cells, and donor T-cells. The central proposition is that the biological effects of microbiota-associated metabolites in gastrointestinal GVHD are determined by the integration of metabolite exposure, receptor identity, responding cell type, tissue compartment, and disease stage. This cell- and context-resolved framework provides a pharmacological basis for developing postbiotics, metabolite-receptor modulators, and microbiota-directed interventions that promote intestinal tolerance while preserving antimicrobial defense and GVL activity.

2. From Intestinal Dysbiosis to Metabolic Dysfunction After Allogeneic Hematopoietic Cell Transplantation

2.1. Conditioning-Induced Epithelial Injury and Ecological Disruption

Intestinal dysbiosis after allo-HCT is initiated before gastrointestinal GVHD is clinically evident. Cytotoxic conditioning and total-body irradiation injure rapidly cycling epithelial cells, distort crypt architecture, and weaken mucus and tight-junction barriers [1,2,3]. Because ISCs and their niche are highly susceptible, regenerative capacity is already reduced when subsequent alloimmune injury develops. Loss of Paneth cells further diminishes α-defensins and other antimicrobial peptides, eroding the normal spatial separation between mucosal surfaces and luminal organisms and enabling expansion of potentially pathogenic taxa [3]. The intestine is thereby converted from a homeostatic microbial habitat into an inflammatory niche that favors translocation.

Once the barrier is breached, lipopolysaccharide, peptidoglycan, bacterial DNA, and other microbial-associated molecular patterns gain access to the lamina propria and circulation. Together with danger signals released by injured host cells, they activate innate and antigen-presenting populations, increase inflammatory cytokines, and facilitate donor T-cell priming [1,2]. The local ecology changes concurrently. Disturbed epithelial oxidative metabolism and perfusion, inflammatory oxygen and nitrogen radicals, and impaired nutrient uptake alter both luminal redox conditions and available substrates. These changes preferentially support facultative anaerobes over obligate anaerobic commensals adapted to a stable hypoxic lumen and complex-carbohydrate fermentation. Epithelial damage is therefore not only an outcome of GVHD; it is an upstream ecological event that permits dysbiosis and metabolic failure to emerge.

2.2. Ecological Selection by Antibiotics, Diet, and Transplantation-Related Therapies

Among transplantation-related exposures, antibiotics exert one of the strongest effects on the intestinal microbiome. Regimens with broad anaerobic coverage remove obligate anaerobes that normally supply colonization resistance and key metabolites. Early use of these agents has been associated with greater intestinal injury, GVHD-related death, and unfavorable transplant outcomes in animal models and patient cohorts [5,20]. Community collapse often culminates in domination by a single organism, commonly Enterococcus species or Enterobacteriaceae, and such domination may precede bacteremia caused by the same bacterial group [21]. These data support microbiome-conscious antimicrobial stewardship, but interpretation must account for confounding by indication, infection severity, neutropenia, center-level prescribing, and other time-varying clinical factors.

Nutrient availability constitutes a second, potentially adjustable ecological pressure. Poor oral intake, parenteral feeding, mucositis, and dietary restriction reduce fermentable carbohydrate delivery to the colon and thereby disadvantage SCFA-producing anaerobes. Conversely, particular substrates can selectively promote disease-associated taxa: lactose supports Enterococcus growth, and either dietary or genetic limitation of lactose reduces enterococcal expansion and mitigates experimental GVHD [8]. Experimental pretransplant obesity is likewise accompanied by microbiota alteration, barrier leakage, and more severe acute GVHD [22]. Excessive fructose consumption has more recently been reported to worsen acute GVHD by perturbing the microbiota, damaging the epithelial barrier, and inducing HIF-1α-associated inflammatory macrophage programs [23]. Resistant starch or prebiotic oligosaccharides may instead sustain butyrate-producing organisms and fecal butyrate, although available clinical observations remain exploratory and do not yet prove preventive efficacy [24]. Non-nutritive sweeteners may constitute an additional, but insufficiently characterized, dietary selection pressure. Experimental models and studies in nontransplant populations have reported compound-specific effects of saccharin, sucralose, aspartame, acesulfame-K, and related sweeteners on microbial composition, SCFA production, and host metabolic responses; however, the direction and magnitude of these effects vary according to sweetener formulation, dose, exposure duration, background diet, and experimental model [25]. Human studies have generally demonstrated smaller or inconsistent effects, and no direct evidence currently establishes non-nutritive sweetener exposure as a determinant of microbiome dysfunction or GI-GVHD after allo-HCT.

Transplant-associated treatments should consequently be analyzed as interacting ecological forces. Antibiotics, glucocorticoids, calcineurin inhibitors, gastric acid suppression, opioids, and changes in nutritional support can each modify microbial proliferation, motility, mucosal immunity, or substrate access. The microbiota can reciprocally influence drug transformation and host pharmacokinetics. Because patients with the most severe illness generally receive the most intensive treatment, studies require longitudinal sampling and time-updated exposure models to separate therapy-driven changes from dysbiosis produced by GVHD itself.

2.3. Loss and Reconfiguration of Microbial Metabolic Functions

Dysbiosis has functional consequences that cannot be reduced to the loss of named taxa. One prominent defect is diminished microbial fermentation. After allo-HCT, butyrate availability falls in part because obligate anaerobic producers are depleted. Replenishing butyrate increases epithelial histone acetylation and junctional protein expression and attenuates experimental GVHD; signaling through epithelial GPR43 adds a receptor-dependent layer of protection [11,12]. Thus, the aggregate capacity of a community to generate SCFAs may carry more biological information than the relative abundance of any individual producer.

The network of microbial transformations of host-derived bile acids is similarly remodeled. Bile salt hydrolases (BSH) deconjugate glycine- or taurine-conjugated primary bile acids, whereas bile acid-inducible (bai) enzymes and hydroxysteroid dehydrogenases (HSDH) catalyze dehydroxylation, oxidation, and epimerization. Depletion of these activities changes both bile acid abundance and the receptor activity encoded by the pool. Experimental and integrated multi-omics studies show that GVHD inflammation impairs bacterial bile acid conversion and that the resulting imbalance can reciprocally intensify epithelial and donor T-cell pathology [14,15]. Longitudinal patient studies likewise report fewer secondary bile acids and reduced representation of microbial transformation genes in individuals who develop GI-GVHD, although diarrhea, antibiotics, hepatic dysfunction, and nutritional status can influence these measures [26].

Tryptophan metabolism is disrupted in parallel. Indoxyl sulfate is a host–microbial co-metabolite formed through host conversion of microbiota-derived indole. Low urinary indoxyl sulfate after allo-HCT has been used as an indirect marker of bacterial tryptophan metabolism and is associated with inferior clinical outcomes [27], although urinary levels also depend on host biotransformation and renal handling. Human metabolomics also demonstrates coordinated changes across microbiota- and host-derived tryptophan products during acute GVHD [28]. In experimental systems, bacterial indoles engage type I interferon-associated tissue-protective programs, implying that impaired indole production may directly reduce epithelial resilience rather than simply accompany declining community diversity [13].

Other aromatic amino acid metabolites may participate in mucosal restitution. Desaminotyrosine activates regenerative programs that require mTORC1 and STING, illustrating metabolite action through intracellular stress and repair pathways rather than a classical receptor [19]. Dysbiosis can also favor inflammatory products. Trimethylamine N-oxide (TMAO), a host–microbial co-metabolite generated through the host oxidation of bacterially produced trimethylamine, drives M1-like macrophage polarization and worsens experimental GVHD through nuclear factor κB and NLRP3 inflammasome signaling [29]. Post-transplant metabolic disturbance therefore includes two complementary processes: disappearance of metabolites required for tissue protection and increased exposure to products that amplify innate inflammation.

Taxonomic resemblance should not be equated with metabolic equivalence. Strains within a single species may differ in carbohydrate fermentation, bile acid conversion, aromatic amino acid pathways, or antimicrobial resistance, whereas unrelated taxa can provide redundant functions. Establishing a biologically important metabolic defect consequently requires functional metagenomics, quantitative community measurements, and both targeted and discovery metabolomics rather than compositional profiling alone.

2.4. A Self-Amplifying Metabolic–Immune Circuit

Taken together, available data favor a reciprocal model of microbial involvement in GI-GVHD. Conditioning first damages the epithelium and destabilizes its microbial niche. Antibiotics, reduced enteral substrates, and inflammatory selection then remove metabolically active anaerobes and enrich facultative or pathogenic organisms. Reduced concentrations of SCFAs, secondary bile acids, indole derivatives, and reparative aromatic amino acid metabolites are associated with impaired epithelial recovery and diminished mucosal immune restraint after allo-HCT. Supplementation experiments in GVHD models indicate that selected metabolites can attenuate tissue injury or inflammatory responses; however, these interventions do not establish that endogenous depletion of the corresponding metabolites initiates disease. Conversely, translocated microbial components and selected inflammatory metabolites can amplify myeloid- and donor T-cell responses, thereby aggravating established epithelial injury [11,13,14,15,19,29].

Experimental transfer of disease-associated microbial communities can reproduce or intensify features of GVHD in murine models, and continuous exposure across the pre- and post-transplantation periods may be required to generate the full phenotype [30]. Nevertheless, temporal ordering and transferability alone cannot determine whether an individual metabolite initiates disease, amplifies established inflammation, or serves as a marker of reduced intake, altered transit, or tissue injury. The proposed self-amplifying metabolic-immune circuit should therefore be interpreted as a conceptual synthesis containing both experimentally validated links and association-based connections, rather than as a fully resolved linear causal pathway. (Figure 1).

Figure 1. Intestinal dysbiosis and metabolic dysfunction after allo-HCT. Conditioning, antibiotics and reduced enteral nutrition can injure the epithelial niche and alter microbial ecology. Changes in SCFA, indole, DAT and bile-acid profiles may alter epithelial, donor T-cell and myeloid responses, impair repair and amplify intestinal immune injury. These pools include microbial products, microbially modified host molecules and host–microbial co-metabolites. The central crypt illustrates the barrier and stem-cell niche. Human associations between microbial communities or metabolite profiles and transplant outcomes are distinguished from experimental GVHD-model findings involving epithelial injury, metabolite responses and immune damage. Gray dashed arrows indicate a conceptual sequence and feedback, with immune injury returning to epithelial injury. They do not establish that every link, or the complete circuit, has been causally validated in patients. Gray solid arrows in the origin schematic indicate microbial production or subsequent host conversion. Dose, compartment, responding cell and disease stage modify exposure and response. Blue identifies transplant exposures, rose/red injury, ochre dysbiosis, purple metabolite dysfunction and teal cellular responses. Green niche cells represent stem cells. Other cell, microbial and particle colors distinguish schematic biological elements, not evidence strength or specific taxa. Upward/downward symbols indicate increases/decreases, and ≠ means ‘does not establish.’ Metabolite depletion alone does not demonstrate disease causation. allo-HCT, allogeneic hematopoietic cell transplantation; GVHD, graft-versus-host disease; GI-GVHD, gastrointestinal graft-versus-host disease; ISC, intestinal stem cell; T-cell (T), T lymphocyte. DAT, desaminotyrosine; SCFA/SCFAs, short-chain fatty acid/acids.

3. Molecular Pathways for Host Sensing of Microbiota-Associated Metabolites

Host sensing of microbiota-associated metabolites involves G protein-coupled receptors, nuclear receptors, ligand-regulated transcription factors, and receptor-independent intracellular mechanisms. This section distinguishes direct ligand–sensor interactions from metabolic actions and downstream signaling requirements. The cellular and compartmental consequences of these mechanisms are examined in Section 4.

3.1. Short-Chain Fatty Acid Sensing

Acetate, propionate, and butyrate are the predominant SCFAs generated by the bacterial fermentation of nondigestible carbohydrates, with a smaller contribution from amino acid fermentation. Their host effects involve extracellular receptor signaling, intracellular enzyme inhibition, and metabolic utilization [11,31,32,33].

SCFAs are interpreted through extracellular receptors and intracellular metabolic mechanisms. The principal G protein-coupled sensors are free fatty acid receptor 2 (FFAR2/GPR43), free fatty acid receptor 3 (FFAR3/GPR41), and hydroxycarboxylic acid receptor 2 (HCAR2/GPR109A) [34]. Their ligand spectra overlap but are not identical. FFAR2 responds to acetate, propionate, and butyrate. FFAR3 is relatively more responsive to propionate and butyrate, whereas HCAR2 is activated by butyrate, β-hydroxybutyrate, and nicotinic acid. These receptors also differ in their distribution across epithelial, myeloid, neuronal, adipose, and other immune compartments. SCFA biology therefore cannot be represented by a simple one-ligand–one-receptor scheme.

FFAR2/GPR43 has direct mechanistic support in experimental GVHD. GPR43-dependent protection involves ERK signaling and an NLRP3–IL-18 program, and receptor deficiency worsens gastrointestinal injury and mortality [12]. Evidence for HCAR2-mediated intestinal regulation derives mainly from nontransplant models [32,35]. whereas a cell-defined causal role for FFAR3 in GI-GVHD remains unestablished. The consequences for epithelial and immune compartments are discussed in Section 4.

Transport through monocarboxylate carriers permits intracellular SCFA actions distinct from surface-receptor signaling. Butyrate, and propionate at higher exposure, inhibit histone deacetylases and alter histone acetylation and chromatin accessibility. Experimental GVHD studies support an association between epithelial butyrate availability and histone acetylation [11], whereas the resulting epithelial and T-cell responses are considered in Section 4.1 and Section 4.3.

In differentiated colonocytes, butyrate also serves as a dominant oxidative fuel. Mitochondrial utilization generates adenosine triphosphate, consumes oxygen, and helps sustain a hypoxic epithelial interface favorable to obligate anaerobes. Impaired butyrate oxidation may therefore oxygenate the epithelial surface and support facultative anaerobic expansion, linking host energy metabolism back to community ecology.

SCFAs can therefore act as receptor ligands, epigenetic enzyme inhibitors, and respiratory substrates. Distinguishing these mechanisms is necessary before attributing an observed response to a specific SCFA receptor. SCFA exposure differs markedly between anatomical compartments. In a human postmortem study, total SCFA concentrations ranged from 80 ± 11 mmol/kg in descending-colon contents to 131 ± 9 mmol/kg in caecal contents, whereas portal and peripheral venous concentrations were 375 ± 70 and 79 ± 22 μmol/L, respectively [36]. By comparison, Mathewson et al. used 1 mM butyrate in primary intestinal epithelial-cell experiments associated with their GVHD study [11]. This nominal concentration exceeds the reported peripheral concentration of total SCFAs, supporting caution when extrapolating epithelial culture findings to circulating cells. However, total SCFA measurements cannot be equated with butyrate concentrations, and neither colonic contents nor culture medium directly represents mucosal cellular exposure. These postmortem data provide compartmental context rather than reference ranges for allo-HCT recipients.

3.2. Sensing of Tryptophan-Derived Indoles

Tryptophan is partitioned among host kynurenine and serotonin metabolism and bacterial pathways that produce indole, indole-3-aldehyde, indole-3-acetic acid, indole-3-propionic acid, indoleacrylic acid, and related molecules. Some of these products remain concentrated in the intestine, whereas others enter portal blood or undergo additional host conversion. Microbiota-derived indole undergoes host-mediated oxidation to indoxyl and subsequent sulfation, predominantly in the liver, to form indoxyl sulfate. Indoxyl sulfate is therefore a host–microbial co-metabolite, and its urinary concentration provides an indirect readout of bacterial tryptophan metabolism that also depends on host biotransformation and renal handling. For unsubstituted indole, a study of 53 healthy adults reported fecal concentrations of 0.30–6.64 mM, with a mean of 2.59 mM [37]. In an epithelial-cell study, HCT-8 cells were exposed to 1 mM indole, with transcriptional responses examined after 4 and 24 h [38]. This experimental concentration falls within the reported fecal range, providing a quantitative rationale for investigating local intestinal effects. Nevertheless, fecal measurements do not establish epithelial or lamina propria exposure, and these values should not be extrapolated to other indole derivatives or interpreted as reference concentrations after allo-HCT.

A central sensor for bacterial indole derivatives is the aryl hydrocarbon receptor (AhR). Without ligands, AhR is retained in a cytoplasmic protein complex. Ligand engagement promotes its nuclear entry, association with AhR nuclear translocator, and regulation of xenobiotic response elements. The resulting programs affect IL-22 production, antimicrobial defenses, epithelial regeneration, immune lineage differentiation, and xenobiotic handling. Indole-3-aldehyde provides an example of a microbial ligand engaging AhR in experimental intestinal models [39]. Its implications for epithelial protection and innate-cell-mediated responses are discussed in Section 4.1.

AhR ligands differ in affinity, metabolic persistence, tissue distribution, and their effects on receptor turnover. AhR activation should therefore be characterized at the level of individual ligands rather than treated as a uniform pharmacological state. Cell-specific differentiation outcomes are discussed in Section 4.

Indole compounds can additionally engage the pregnane X receptor (PXR), although activity depends on the metabolite, species, and ligand context. In reporter assays, indole-3-propionic acid (IPA) alone activated human PXR with an EC₅₀ of 120 μM, compared with 0.55 μM for mouse PXR; human PXR activation was enhanced when IPA was combined with 1 mM indole [40]. The same study linked this signaling axis to reduced TLR4-associated epithelial dysfunction. These results highlight substantial species differences, but assay-derived EC₅₀ values do not establish effective tissue concentrations or therapeutic thresholds in GI-GVHD. Supporting evidence derives principally from non-GVHD intestinal models, and receptor-specific validation in allo-HCT remains limited. The potential epithelial consequences are considered in Section 4.1.

In experimental GVHD, indole-mediated protection requires intact host type I interferon signaling [13]. This establishes a downstream signaling requirement rather than the direct binding of indoles to the interferon receptor. Tissue protection and preservation of GVL activity are discussed in Section 4.1 and Section 4.3, respectively.

3.3. Bile Acid Sensing

Bile acids originate from host cholesterol metabolism, but intestinal microorganisms greatly expand their chemical repertoire. Secondary bile acids and their derivatives are therefore described here as microbially modified host-derived molecules rather than exclusively microbial products. Hepatic primary bile acids are usually conjugated to glycine or taurine before entering bile. In the intestine, BSH remove these conjugates, HSDH catalyze oxidation and epimerization, and enzymes encoded by the bai operon mediate 7α-dehydroxylation. These reactions generate secondary and other microbially modified bile acids whose receptor profiles differ from those of the original host molecules.

Remodeling of the bile acid pool changes the balance of agonists, partial agonists, and antagonists available to host receptors [14,15,26]. The farnesoid X receptor (FXR/NR1H4) is a bile acid-activated nuclear receptor. After ligand binding, FXR partners with retinoid X receptor to control bile acid synthesis and transport, detoxification, lipid handling, and epithelial homeostasis. Donor T-cell FXR has direct experimental support in GVHD [14]. Its cellular consequences are discussed in Section 4.3.

Takeda G protein-coupled receptor 5 (TGR5/GPBAR1) is a membrane bile acid sensor distributed across epithelial, enteroendocrine, myeloid, neural, and other cell types. Its activation raises intracellular cyclic adenosine monophosphate and can suppress nuclear factor κB-dependent inflammation. In intestinal stem-cell models, TGR5 also promotes regeneration and epithelial recovery [41,42].

The vitamin D receptor (VDR) and PXR provide additional nuclear sensing routes for bile acids [43]. Lithocholic acid and selected derivatives activate VDR, an intestinal bile acid sensor that coordinates detoxification and barrier-associated transcription [44]. PXR recognizes several hydrophobic bile acids and regulates xenobiotic processing and epithelial defense. Because microbial indoles also activate PXR, this receptor may integrate signals from distinct microbial metabolic networks rather than report bile acid exposure alone [40].

Bile acids may also influence cellular stress through mechanisms not fully explained by canonical receptor activation. Accordingly, the efficacy of TUDCA should not be assigned to FXR or TGR5 without receptor-specific validation [45]. Its epithelial effects are discussed in Section 4.1.

The receptor signal encoded by a bile acid pool depends on molecular composition rather than total concentration. Conjugation controls passive permeability and transporter requirements, whereas hydroxylation and epimerization modify affinity and efficacy. Individual bile acids may act at several receptors; a weak FXR ligand may, for example, be a more effective agonist at TGR5 or VDR. Exposure also differs among the lumen, mucosa, portal blood, liver, and systemic circulation. Fecal bile acid data alone therefore cannot identify receptor engagement in a particular host compartment. In addition, substantial differences between murine and human bile acid pools limit direct quantitative extrapolation from experimental models. Mechanistic studies should therefore combine targeted bile acid profiling with transporter abundance, receptor activity, anatomical localization, and downstream pharmacodynamic responses.

Quantitative measurements are particularly important for individual bile acid derivatives. Hang et al. reported mean fecal levels of 3-oxoLCA and isoalloLCA of 23 and 2 pmol/mg, respectively, in patients with ulcerative colitis; isoalloLCA values ranged from 0 to 17 pmol/mg [46]. In the same study, representative T-cell differentiation experiments used 20 μM 3-oxoLCA or isoalloLCA. These findings demonstrate endogenous occurrence and experimental activity, but fecal mass-normalized measurements cannot be directly equated with culture-medium molarity or free concentrations at intestinal T-cells. Moreover, these human measurements were obtained outside allo-HCT and do not establish whether comparable cellular exposure occurs in GI-GVHD.

3.4. Noncanonical and Emerging Metabolite-Sensing Pathways

A recognized membrane or nuclear receptor is not required for every effect of microbiota-associated metabolites. Some products are used as fuels or biosynthetic substrates, some inhibit enzymes, and others alter intracellular stress or repair networks without directly binding the signaling protein implicated downstream. Correctly distinguishing these modes of action is essential when a sensor is proposed for a newly studied metabolite.

Desaminotyrosine illustrates the distinction between a downstream signaling requirement and a direct molecular sensor. Its protective response in experimental GVHD requires mTORC1 and STING, but direct binding to either component has not been established [19]. The regenerative phenotype is discussed in Section 4.1.

TMAO is a host–microbial co-metabolite whose proximal recognition mechanism remains uncertain [47]. Its formation involves microbial trimethylamine production and host oxidation, while circulating exposure also depends on renal elimination. Its inflammatory effects in experimental GVHD are discussed in Section 4.2.

Other candidate mechanisms are derived mainly from intestinal models outside GVHD. Spermine, spermidine, and related polyamines affect chromatin, autophagy, mitochondrial activity, and inflammasome function; microbiota-dependent polyamine signals have been connected to NLRP6 regulation and epithelial antimicrobial defenses [43,48], Because polyamines can arise from dietary, microbial, and host sources, their abundance alone does not establish microbial origin. Bacterial lactate can stimulate intestinal stem-cell proliferation and development through combined receptor and metabolic effects [42,49]. Whether either pathway materially contributes to crypt recovery after allo-HCT remains unknown.

Succinate may rise during dysbiosis, tissue hypoxia, or impaired mitochondrial metabolism. Extracellular succinate activates succinate receptor 1, while intracellular accumulation can stabilize HIF-1α and enhance interleukin-1β production [50,51]. Because microbial and host pathways both generate this intermediate, source-resolved approaches such as isotope tracing are necessary before a GVHD phenotype can be attributed specifically to bacterial succinate.

Bacterial sphingolipids and vitamin-derived antigens represent further signaling categories. Sphingolipids synthesized by symbiotic organisms alter host lipid networks and help regulate intestinal natural killer T-cell homeostasis [52]. Riboflavin intermediates can be displayed by the nonclassical antigen-presenting molecule MR1 and recognized by mucosal-associated invariant T-cells, providing a direct connection between microbial vitamin metabolism and lymphocyte activation [53]. Evidence directly linking either mechanism to GI-GVHD is currently sparse; they should therefore be treated as testable hypotheses rather than validated intervention targets.

Emerging metabolite pathways should be categorized by the type of evidence supporting them: verified binding to a receptor, transporter-dependent intracellular action, use as an energetic or anabolic substrate, modulation of a metabolic enzyme, or requirement for a downstream signaling module. This scheme avoids describing downstream dependencies as receptors and keeps GVHD-specific causality separate from mechanisms inferred from other inflammatory disorders. It also provides the necessary foundation for comparing how epithelial, myeloid, innate lymphoid, and donor T-cell compartments interpret the same metabolic exposure.

4. Cell-Specific Integration of Metabolite Sensing in Gastrointestinal GVHD

Building on the molecular pathways described in Section 3, this section examines how epithelial, stem-cell, myeloid, and donor T-cell compartments integrate metabolite signals. Particular attention is given to differences between tissue protection and immune suppression, the cellular attribution of experimentally observed effects, and the distinction between direct GVHD evidence and mechanistic extrapolation (Figure 2).

Figure 2. Cell-specific sensing of microbiota-associated metabolites. The upper panels summarize experimental GVHD findings: GPR43-associated intestinal protection through ERK/NLRP3/IL-18, selected indole-dependent protection requiring host type I IFN signaling, and donor T-cell FXR signaling that can enhance inflammation. DAT requires mTORC1 and epithelial STING for the studied regenerative response, whereas TMAO promotes downstream NF-κB/NLRP3-associated inflammation. Direct DAT binding remains unproven, and the proximal TMAO sensor is unresolved. Solid arrows in response pathways indicate functional relationships tested in GVHD models, not necessarily direct ligand binding. Solid source-conversion arrows depict microbial production, microbial modification of host bile acids, or microbial TMA followed by host oxidation to TMAO. Gray dashed arrows identify mechanisms established outside GVHD: HCAR2/FFAR3 and T-cell programs, AhR/IL-22 and PXR-associated barrier regulation, and TGR5/VDR/PXR biology. These mechanisms are not presented as verified mediators of clinical GI-GVHD responses. Green result labels/arrows indicate repair or protection, red inflammation, and up arrow means an increase. Green, ochre and purple headings distinguish SCFAs, indoles and bile acids, respectively. Other illustration colors identify biological elements, not evidence strength. Epithelial icons indicate tissue protection without assigning every signal exclusively to epithelial cells. The asterisk restricts retained GVL to tested regimens. Human metabolite associations do not establish these causal pathways or clinical efficacy. AhR, aryl hydrocarbon receptor; DAT, desaminotyrosine; ERK, extracellular signal-regulated kinase; FFAR3, free fatty acid receptor 3; FXR, farnesoid X receptor; GPR43, G protein-coupled receptor 43; GVHD, graft-versus-host disease; GI-GVHD, gastrointestinal graft-versus-host disease; GVL, graft-versus-leukemia; HCAR2, hydroxycarboxylic acid receptor 2; IFN, interferon; IL-18/IL-22, interleukin-18/interleukin-22; ISC, intestinal stem cell; mTORC1, mechanistic target of rapamycin complex 1; NF-κB, nuclear factor kappa B; NLRP3, NLR family pyrin domain containing 3; PXR, pregnane X receptor; SCFAs, short-chain fatty acids; STING, stimulator of interferon genes; T-cell, T lymphocyte; TGR5, G protein-coupled bile acid receptor 1; TMA, trimethylamine; TMAO, trimethylamine N-oxide; VDR, vitamin D receptor.

4.1. Intestinal Epithelial and Stem Cells: Barrier Integrity and Tissue Tolerance

Differentiated epithelial cells and crypt stem/progenitor cells may respond differently to the same metabolite exposure. Butyrate restoration improves epithelial survival and barrier integrity in experimental GVHD [11]. However, non-GVHD crypt models show that the exposure of normally protected stem/progenitor cells to butyrate can inhibit proliferation [54]. This distinction highlights the importance of barrier architecture and epithelial differentiation state; it should not be interpreted as an established inhibitory effect on intestinal regeneration during GVHD.

FFAR2-associated protection and indole-dependent host responses provide additional evidence for tissue-protective metabolite signaling [12,13]. The molecular pathways are described in Section 3.1 and Section 3.2. A remaining question is whether protection reflects direct epithelial sensing, signals from neighboring innate cells, or coordinated responses across these compartments. In particular, AhR- and PXR-associated barrier mechanisms established outside transplantation require further cell-specific validation in allo-HCT [39,40].

TUDCA reduces epithelial apoptosis and MHC class II expression, thereby limiting intestinal antigen presentation [45], whereas desaminotyrosine supports intestinal regeneration [19]. TGR5-dependent stem-cell regeneration provides a further candidate mechanism, although the supporting evidence largely derives from non-GVHD models [41]. These findings identify epithelial survival, local antigen presentation, and regenerative capacity as distinct therapeutic outcomes. Their implications for preservation of GVL activity are considered in Section 4.3.

4.2. Myeloid Cells and APCs: Inflammatory Amplification

After conditioning, recipient macrophages and dendritic cells are exposed to microbial products in the lamina propria, draining lymph nodes, and bloodstream. Barrier failure permits translocation of microbial-associated molecular patterns, while damaged tissue releases extracellular adenosine triphosphate, uric acid, and other danger signals. Metabolite sensing consequently occurs against a highly activated inflammatory background and can either increase or reduce the threshold for myeloid-cell activation.

SCFAs and selected bile acid or AhR ligands can favor less inflammatory macrophage and dendritic-cell states in nontransplant models [32,42]. Their relevance to GI-GVHD requires demonstration that myeloid-cell sensing contributes to disease modification. Inflammasome responses also require cellular attribution: epithelial repair-associated signaling should be distinguished from inflammatory IL-1β production in recipient immune compartments. Thus, an anti-inflammatory effect in another disease model does not establish a validated myeloid therapeutic target in GVHD.

TMAO provides direct experimental evidence of a metabolite-associated inflammatory myeloid response in GVHD. Exposure promotes an M1-like macrophage state, NF-κB/NLRP3-associated signaling, and inflammatory cytokine release, accompanying increased disease severity [29]. Its host–microbial formation and determinants of systemic exposure are described in Section 3.4.

Professional APCs and intestinal epithelial cells contribute differently to alloantigen presentation. Recipient dendritic cells support donor T-cell priming, whereas epithelial MHC class II provides a local route for donor CD4⁺ T-cell activation within the intestine [55]. Metabolite-associated changes in these compartments should therefore be assessed separately: reduced APC activation, reduced epithelial antigen presentation, and improved barrier integrity represent distinct outcomes and may require different intervention windows.

4.3. Donor T-Cells: Differentiation, Cytotoxicity and GVL Activity

Donor T-cells can encounter microbiota-derived metabolites during priming in lymphoid organs, circulation, and subsequent entry into the intestine. SCFAs promote regulatory T-cell differentiation in nontransplant models [31,32,33,56], but this does not establish that the direct stabilization of donor-derived Tregs mediates protection in GVHD. Studies must distinguish direct donor T-cell effects from secondary changes caused by epithelial protection or altered antigen presentation.

Microbially modified bile acid derivatives exert structure-dependent control over T-cell fate. In non-GVHD models, 3-oxo-lithocholic acid suppresses RORγt-dependent Th17 differentiation, isoallo-lithocholic acid promotes regulatory T-cell development, and other secondary species support RORγt⁺ peripheral regulatory T-cells [46,57,58]. These experiments establish a mechanistic link between bacterial bile acid conversion and the Th17/Treg balance, but sufficient exposure of relevant tissues after allo-HCT has not been demonstrated for many of these molecules.

Donor T-cell FXR has direct mechanistic support in experimental GVHD. Altered microbial bile acid conversion changes the ligand environment and can enhance FXR-dependent inflammatory responses in donor T-cells [14]. This finding establishes the donor lymphocyte compartment as a relevant site of bile acid sensing, while leaving the effects of epithelial or hepatobiliary FXR manipulation to be evaluated separately.

AhR-associated T-cell differentiation is context dependent, with regulatory, Th17, and Tr1 outcomes reported under different experimental conditions. These findings should not be interpreted as evidence that a microbial indole produces the same differentiation program in donor T-cells during GVHD. Translation requires direct assessment of donor-cell phenotype and function rather than aggregate AhR activity alone.

Donor CD8⁺ T-cells warrant separate evaluation. Metabolite-dependent changes in mitochondrial metabolism, chromatin acetylation, cytokines, or homing receptors may alter their expansion and cytotoxicity. Excessive inhibition could alleviate GVHD while compromising pathogen defense or the GVL response. Notably, indole-mediated tissue protection and TUDCA each reduced experimental GVHD without abolishing GVL activity [13,45]. These findings suggest that increasing tissue tolerance may separate GVHD control from GVL more effectively than broad donor T-cell suppression. Even so, GVL preservation must be tested directly for each metabolite or receptor intervention.

4.4. Spatial, Temporal and Cellular Determinants of Biological Response

The examples discussed in Section 4.1, Section 4.2 and Section 4.3 show that metabolite activity cannot be inferred from ligand identity alone. Responding-cell identity, differentiation state, receptor expression, and barrier integrity influence which biological response predominates. An effect demonstrated in one compartment should therefore not be assumed to occur in another.

Fecal concentrations are incomplete surrogates for cellular exposure. Levels in the lumen, mucus, intestinal tissue, portal blood, and systemic circulation can differ because of microbial utilization, epithelial uptake, host biotransformation, and clearance. Paired metabolite measurements and cell-resolved pharmacodynamic analyses are needed to connect exposure with the proposed response.

The relevant target compartment also changes over the disease course. Donor T-cell priming and recipient APC activity are prominent during initiation, whereas epithelial injury, local antigen presentation, and inflammatory effector responses sustain established disease. Regenerative programs become particularly relevant during recovery. Metabolite-directed interventions should consequently specify the intended target cell, exposure compartment, and therapeutic window rather than assume a uniform effect throughout GI-GVHD (Table 1).

Table 1. Selected evidence for host sensing of microbiota-associated metabolites relevant to GI-GVHD after allo-HCT.

Metabolite and Pathway	Disease Context	Target Cell/Compartment	Main Finding and Limitation	Human Findings	Experimental Disease Evidence	Sensor/Pathway Evidence	Refs.
Butyrate (M) HDAC inhibition; oxidative substrate	Murine acute GVHD	Recipient IECs	Restored epithelial histone acetylation and barrier protection. Culture and luminal exposure do not establish human target-cell exposure.	Clinical efficacy of supplementation not established.	Supplementation reduced intestinal injury and GVHD severity.	Epithelial biochemical and functional evidence; not proof that all effects require a surface receptor.	[11]
Butyrate/propionate (M) FFAR2 (GPR43)–ERK–NLRP3–IL-18	Murine acute GVHD	Recipient nonhematopoietic compartment; IECs implicated	Reduced tissue injury; propionate protection required GPR43, whereas butyrate retained partial protection without it. Acetate did not benefit the tested regimen.	Human FFAR2-dependent GI-GVHD protection not established.	Metabolite supplementation and GPR43 agonism reduced disease in experimental transplantation.	Gpr43 loss, agonist/antagonist studies and compartment-resolved NLRP3 experiments; complete IEC-specific receptor attribution remains limited.	[12]
Butyrate/propionate (M) Sensor not established in the human association	Human chronic GVHD	Circulating metabolite measurements; target cell unresolved	Higher circulating levels associated with protection from chronic GVHD; does not support acute GI-GVHD treatment efficacy.	Lower day-100 plasma butyrate and propionate in patients developing chronic GVHD; observational cohorts.	No acute GI-GVHD intervention conclusion from this cohort.	No receptor dependence established by the clinical association.	[59]
SCFAs (M) FFAR2/HCAR2; HDAC-associated mechanisms	Non-GVHD intestinal models	Colonic Tregs; epithelial and myeloid compartments	Regulatory immunity and intestinal homeostasis; direct donor-Treg mediation in GVHD remains unproven.	Not clinical allo-HCT intervention evidence.	Treg and intestinal-inflammation experiments outside GVHD.	Receptor and epigenetic mechanisms evaluated in the specific non-GVHD models; mechanisms differ among studies.	[31,32,33,35,56]
Butyrate (M) Metabolic filtering; HDAC inhibition	Non-GVHD crypt injury/homeostasis	Crypt stem/progenitor cells	Exposure can suppress stem-cell proliferation when crypt protection is disrupted; differentiated IEC effects should not be generalized to ISCs.	No clinical GI-GVHD efficacy evidence in the cited study.	Crypt protection and injury models outside GVHD.	Experimental evidence for exposure-dependent stem-cell responses; no GVHD receptor validation.	[54]
Indole-related microbial products (M) Type I IFN signaling	Murine acute GVHD	Recipient tissue-protection compartment	Limited intestinal injury and experimental GVHD; GVL-related activity assessed in the tested models.	Clinical supplementation efficacy not established.	Metabolite-related interventions attenuated GVHD.	Type I IFN pathway dependence; does not validate AhR or PXR as the GVHD mediator.	[13]
Tryptophan-derived indole ligands (M) AhR–IL-22	Non-GVHD mucosal/infection models	ILC3 and mucosal immune compartment	Supported mucosal reactivity and IL-22 responses; ligand and disease context determine activity.	No clinical GI-GVHD efficacy evidence in the cited study.	Mechanistic mucosal experiments outside GVHD.	AhR-associated ligand–pathway evidence in non-GVHD systems.	[39]
Indole-3-propionic acid/indole (M) PXR–TLR4 axis	Non-GVHD intestinal barrier models	Intestinal epithelium	Barrier protection and reduced inflammatory responses; receptor potency is species- and ligand-context dependent.	Human receptor assays are not clinical treatment evidence.	Barrier experiments outside GVHD.	PXR activation assays and PXR/TLR4 mechanistic studies; allo-HCT cell-specific dependence not established.	[40]
Indoxyl sulfate (HM) Host oxidation/sulfation of microbial indole; biomarker context	Human allo-HCT outcomes	Urinary measurement; responding cell unresolved	Low early urinary levels associated with microbiome disruption and poor outcome; renal handling and host metabolism affect interpretation.	Observational association; no IS replacement efficacy demonstrated.	No therapeutic GI-GVHD conclusion from this cohort.	No protective metabolite-sensor mechanism established by this association.	[27]
Microbially modified bile acids (MM) FXR	Experimental GVHD; human translational observations	Donor T-cells	Loss of selected FXR-modulating bile acids can favor inflammatory T-cell signaling; effects are compound- and compartment-specific.	Human bile-acid/T-cell findings and UDCA-related observations; separate from causal mouse experiments.	FXR agonism worsened disease; donor-T-cell FXR deletion improved survival.	Donor-T-cell genetic validation and pharmacological perturbation; not a uniform effect of all bile acids.	[14]
Microbially modified bile acids (MM) Sensor not established by cohort measurements	Human acute GVHD/acute GI-GVHD after allo-HCT	Fecal/multi-omics measurements	Altered bile-acid profiles associated with disease; association does not establish direction of causality.	Multi-omics acute-GVHD findings [15]; longitudinal fecal GI-GVHD association [26].	Human associations shown here; interventional findings must be considered separately.	A measured bile-acid shift alone does not validate a receptor.	[15,26]
3-oxoLCA, isoalloLCA and other defined bile-acid derivatives (MM) RORγt and compound-specific Treg pathways	Non-GVHD T-cell differentiation/intestinal models	CD4+ T-cells/peripheral Tregs	Structure-dependent Th17/Treg regulation; relevant free exposure after allo-HCT remains unresolved.	Human metabolite measurements do not establish GI-GVHD efficacy.	T-cell and intestinal immune mechanisms outside GVHD.	Ligand/pathway validation is compound-specific and restricted to the cited models.	[46,57,58]
Lithocholic acid (MM) PXR	Non-GVHD bile-acid detoxification models	Hepatic xenobiotic-processing compartment	LCA sensing supported detoxification and protection from liver toxicity; not a GI-GVHD efficacy finding.	No GI-GVHD treatment evidence in this study.	Liver-toxicity experiments outside GVHD.	PXR activation and genetic mechanistic evidence; no allo-HCT-specific receptor validation.	[43]
Lithocholic acid (MM) VDR	Non-GVHD receptor/intestinal detoxification studies	Intestinal bile-acid sensing	LCA can engage VDR-dependent detoxification programs; protective GVHD efficacy remains unestablished.	No GI-GVHD treatment evidence in this study.	Receptor and intestinal studies outside GVHD.	Ligand and receptor assays support VDR sensing; do not establish a GVHD mechanism.	[44]
Administered TUDCA (A) Epithelial stress and antigen-presentation pathways	Murine acute GVHD; human cellular experiments	IECs/nonhematopoietic target tissues	Reduced epithelial damage and antigen presentation; GVL-related activity directly assessed in experimental systems.	Human cellular evidence; no clinical allo-HCT treatment trial in the cited study.	Systemic prophylaxis reduced GVHD in murine transplantation models.	Epithelial functional validation; benefit not assigned to FXR or TGR5 without receptor-specific evidence.	[45]
Administered UDCA (A) Mechanism not established by the prophylaxis trial	Human allo-HCT prophylaxis	Hepatic and clinical GVHD outcomes	Reduced hepatic complications and severe acute GVHD, including intestinal outcomes; does not establish treatment of active GI-GVHD.	Randomized open-label multicenter study; n = 242.	Separate experimental FXR evidence appears in the preceding FXR row.	The clinical trial does not demonstrate a receptor-dependent mechanism or direct GVL preservation.	[60]
Secondary bile acids and experimental agonists (MM/S) TGR5	Non-GVHD regeneration models	ISCs	Supported epithelial regeneration; direct GI-GVHD efficacy remains unestablished.	No clinical GI-GVHD efficacy evidence in the cited study.	Stem-cell and intestinal regeneration experiments outside GVHD.	TGR5-dependent regeneration in non-GVHD systems.	[41]
TGR5 agonist signaling (S) TGR5-associated anti-inflammatory pathways	Non-GVHD atherosclerosis/inflammation models	Macrophages	Reduced macrophage inflammation and lipid loading; not evidence of GI-GVHD treatment.	No clinical GI-GVHD efficacy evidence in the cited study.	Myeloid effects evaluated outside GVHD.	TGR5 mechanistic evidence restricted to the non-GVHD model.	[42]
Desaminotyrosine (DAT) (M) mTORC1- and STING-dependent regeneration	Murine GVHD; human allo-HCT outcomes/organoids	ISC/IEC compartment	Supported regenerative tissue tolerance and experimental antitumor activity. Human survival/relapse associations do not prove treatment efficacy.	Higher fecal DAT associated with better survival and lower relapse in 50 recipients; human organoid activity.	Supplementation mitigated GVHD; GVL-related activity assessed in experimental models.	Epithelial STING genetic and organoid evidence; mTORC1 perturbation. Direct DAT binding to STING not established.	[19]
Trimethylamine N-oxide (TMAO) (HM) NF-κB/NLRP3-associated signaling	Murine acute GVHD	Macrophages	Promoted M1-like polarization and worsened GVHD; human exposure thresholds remain unresolved.	Clinical causality and intervention thresholds not established.	TMAO exposure aggravated experimental GVHD.	Inflammatory pathway evidence; NF-κB and NLRP3 are not identified as direct TMAO-binding receptors.	[29]
SCFAs (M) Metabolic and epigenetic programming	Non-GVHD antigen-activation models	Antigen-activated CD8+ T-cells	Enhanced memory potential; cannot be equated with preserved GVL after allo-HCT.	No clinical GI-GVHD efficacy evidence in the cited study.	CD8+ T-cell programming studied outside transplantation.	Cellular metabolic mechanisms; GVHD-specific sensor dependence not established.	[61]

Evidence interpretation: Human findings, experimental disease modification and sensor/pathway validation are independent dimensions; no single evidence ranking or formal certainty grade is assigned. Representative original studies discussed in Section 3, Section 4 and Section 5 are separated where chemical identity, disease setting or mechanism differs. Human association does not establish causality or therapeutic efficacy. Pathway dependence does not establish direct ligand binding. Non-GVHD and chronic-GVHD findings do not establish efficacy in acute GI-GVHD. Metabolic origin: M, microbiota-derived product in the intestinal pathways discussed here; MM, microbial modification of host-derived bile-acid substrates; HM, host–microbial co-metabolite requiring host conversion of a microbial precursor; A, administered compound, with intervention effects distinguished from endogenous origin; S, synthetic receptor agonist. Unconjugated UDCA and its taurine conjugate TUDCA are listed separately. References correspond to the numbered reference list. Abbreviations: AhR, aryl hydrocarbon receptor; allo-HCT, allogeneic hematopoietic cell transplantation; BA, bile acid; DAT, desaminotyrosine; ERK, extracellular signal-regulated kinase; FFAR2, free fatty acid receptor 2; FXR, farnesoid X receptor; GI-GVHD, gastrointestinal graft-versus-host disease; GVL, graft-versus-leukemia; HCAR2, hydroxycarboxylic acid receptor 2; HDAC, histone deacetylase; IEC, intestinal epithelial cell; IFN, interferon; IL, interleukin; ILC3, group 3 innate lymphoid cell; ISC, intestinal stem cell; IS, indoxyl sulfate; LCA, lithocholic acid; MHC-II, major histocompatibility complex class II; mTORC1, mechanistic target of rapamycin complex 1; NF-κB, nuclear factor kappa B; NLRP3, NLR family pyrin domain containing 3; PXR, pregnane X receptor; RORγt, retinoic acid receptor-related orphan receptor gamma t; SCFA, short-chain fatty acid; STING, stimulator of interferon genes; TGR5, G protein-coupled bile acid receptor 1; Th17, T helper 17 cell; TLR4, toll-like receptor 4; TMAO, trimethylamine N-oxide; Treg, regulatory T-cell; TUDCA, tauroursodeoxycholic acid; UDCA, ursodeoxycholic acid; VDR, vitamin D receptor.

5. Pharmacological Exploitation of Microbiota–Metabolite–Host Sensing Pathways

The microbiota–metabolite–host-sensor axis offers several distinct intervention points: replacing a depleted metabolite, modulating its host sensor, restoring microbial metabolite production and transformation of host-derived substrates, or altering the ecology that supports the relevant pathway. These approaches differ in controllability and biological breadth. A chemically defined postbiotic permits reproducible dosing, but may not recapitulate the location and kinetics of community production. Fecal microbiota transplantation (FMT) and live biotherapeutics can restore multiple functions at once, yet engraftment, metabolite output, and biosafety are less predictable. Translation should therefore emphasize correction of a measurable metabolic and sensing defect rather than restoration of taxonomy as an end in itself.

5.1. Direct Metabolite Replacement and Postbiotics

SCFAs are among the most extensively investigated candidates for postbiotic therapy. In murine acute GVHD, butyrate supplementation or restoration of epithelial butyrate exposure increases histone acetylation, strengthens tight-junction integrity, and attenuates intestinal injury [11]. Propionate attenuated experimental GVHD through GPR43-dependent signaling, whereas acetate did not improve GVHD in the tested regimen [12]. By contrast, the principal human evidence linking butyrate and propionate to GVHD protection derives from observational studies of chronic GVHD. These chronic GVHD associations provide clinical relevance but do not demonstrate that SCFA supplementation prevents or treats acute GI-GVHD. Prospective intervention studies specifically enrolling patients with, or at risk of, acute GI-GVHD are therefore required [59].

Practical delivery of SCFAs is constrained by taste, rapid proximal absorption, extensive utilization by colonocytes and liver, and difficulty sustaining distal intestinal exposure. Oral sodium butyrate may therefore generate inconsistent mucosal concentrations. Potential alternatives include tributyrin or other esterified prodrugs, delayed-release and microencapsulated preparations, rectal formulations, and fermentable substrates that produce SCFAs locally. Formulation should follow the intended compartment: colon-targeted delivery may suit epithelial and lamina propria effects, whereas modulation of circulating or lymphoid cells may require systemic exposure. The latter may also produce metabolic effects not encountered during physiological local generation.

Bile acid-based strategies benefit from greater prior clinical use. Orally administered ursodeoxycholic acid (UDCA) has long been used to prevent hepatobiliary complications after allo-HCT; in a randomized multicenter trial, prophylaxis reduced hepatic events, severe acute GVHD, and nonrelapse mortality [60]. Mechanistic studies now suggest that UDCA can reduce FXR activity, inhibit proliferation of human T-cells, and associate with lower GVHD-related mortality [14]. In experimental models, TUDCA decreases epithelial apoptosis and MHC class II expression while preserving GVL activity. UDCA and TUDCA should not be treated as interchangeable [45]. Their effects may involve FXR modulation, reduction in endoplasmic reticulum stress, membrane stabilization, and broader remodeling of the bile acid pool to different extents.

Microbial indole derivatives represent another postbiotic opportunity. In experimental allo-HCT, they engage host type I interferon-dependent tissue protection, reduce GVHD, and retain GVL activity [13]. Translation is complicated by heterogeneity in AhR and PXR potency, rapid oxidation and conjugation by the host, and conversion into products with distinct biology. Colon-restricted analogues, controlled-release prodrugs, or more stable derivatives may provide more consistent exposure than the parent indole. Persistent systemic AhR stimulation should nonetheless be avoided until ligand-specific toxicology is adequately defined.

Desaminotyrosine is compelling because it promotes mTORC1–STING-dependent epithelial regeneration rather than imposing broad donor immune suppression [19]. Development will require the systematic characterization of oral absorption, metabolic stability, tissue distribution, the exposure range that supports repair, and possible interference with antiviral or antitumor STING functions. Current efficacy evidence remains preclinical.

A combination postbiotic could theoretically better approximate physiological microbial signaling by supplying complementary epithelial fuels, regenerative indoles, and immunoregulatory bile acids. Such combinations cannot be presumed additive. Their development requires factorial dose exploration and the direct assessment of synergistic or antagonistic receptor engagement. A fixed formulation should be justified by nonoverlapping mechanisms and compatible pharmacokinetics rather than by the independent efficacy of each component.

5.2. Pharmacological Modulation of Metabolite-Sensing Receptors

Direct receptor modulation may reduce dependence on variable microbial production. FFAR2 agonists could reproduce selected acetate- or propionate-associated epithelial effects, while FFAR3 and HCAR2 ligands may target immune, neural, or epithelial responses. The protective effect of GPR43 signaling in experimental GVHD provides direct validation of FFAR2 as a candidate target [12]. Studies outside allo-HCT also implicate FFAR2 and HCAR2 in inflammasome-regulated barrier homeostasis and regulatory immunity [32,35]. Synthetic FFAR2 agonists, including allosteric and dual FFAR2/FFAR3 compounds, have not yet been tested adequately in allo-HCT. Nicotinic-acid activation of HCAR2 is clinically feasible but may be limited by systemic lipid effects and prostaglandin-associated flushing.

AhR offers a promising yet unusually complex pharmacological target. Ligands vary in affinity, metabolic persistence, transcriptional bias, and capacity to promote receptor degradation. Depending on chemical structure and the cytokine environment, AhR can support epithelial repair, ILC3-derived IL-22, Th17 cells, conventional regulatory T-cells, or Tr1 differentiation [39,56,61]. Candidate modulators should therefore be defined by ligand-specific transcriptional signatures rather than the generic label of agonist. Local, rapidly metabolized, or partial agonists may provide greater control than durable high-affinity compounds with systemic distribution.

PXR activation may reinforce barrier integrity and limit TLR4-dependent epithelial inflammation [40]. PXR activation can induce both cytochrome P450 3A4 (CYP3A4) and ATP-binding cassette subfamily B member 1 (ABCB1), which encodes P-glycoprotein (P-gp). CYP3A4 catalyzes the oxidative metabolism of numerous transplantation-related drugs, whereas intestinal P-gp limits net oral absorption by transporting susceptible substrates back into the intestinal lumen [62]. Their coordinated induction may reduce systemic exposure to CYP3A4 and/or P-gp substrates, including tacrolimus, cyclosporine, and sirolimus; PXR-mediated CYP3A4 induction may also reduce ruxolitinib exposure. Conversely, concomitant CYP3A inhibition by azole antifungals may increase drug exposure and toxicity. PXR-directed interventions would therefore require prospective pharmacokinetic and drug–drug interaction assessment. Application in allo-HCT would require prospective drug–drug interaction studies and, preferably, an intestinally restricted ligand with minimal hepatic activation.

TGR5 agonism could combine stem-cell regeneration with the suppression of myeloid nuclear factor κB signaling. Experimental agents such as INT-777 are active in non-GVHD systems, and TGR5 stimulation can enhance intestinal stem-cell recovery [41,42]. Direct efficacy and safety in GI-GVHD have not been established. Effects on gallbladder filling, enteroendocrine hormone release, pruritus, and energy metabolism argue for development of gut-restricted or signaling-biased agonists.

FXR requires the greatest caution because therapeutic direction varies by compartment. The FXR agonist GW4064 worsened experimental GVHD, whereas deletion of FXR in donor T-cells improved survival [14]. Systemic agonists developed for cholestatic or metabolic liver disease therefore cannot be presumed beneficial in GVHD. Conversely, indiscriminate antagonism could disturb hepatic and intestinal bile acid homeostasis. More rational options include transient blockade during donor T-cell expansion, T-cell-selective inhibition, or delivery that spares epithelial and hepatic FXR.

5.3. Restoration of Endogenous Metabolite Production

Microbiome-directed interventions differ substantially in biological breadth, manufacturing control, therapeutic purpose, and evidentiary maturity. Studies in inflammatory bowel disease support the general biological plausibility that dietary substrates, probiotics, and fecal microbiota transplantation can modify microbial composition, metabolic activity, barrier function, and mucosal immune responses [63]. However, these findings cannot be directly extrapolated to allo-HCT recipients, who frequently have profound immunosuppression, epithelial injury, neutropenia, and concurrent broad-spectrum antimicrobial exposure. Accordingly, microbiome restoration strategies are considered separately below according to their intended purpose, stage of evidence, functional or pharmacodynamic endpoints, clinical endpoints, and principal safety limitations.

Restoring microbial metabolite production and transformation of host-derived substrates may provide more physiological and durable metabolite exposure than direct administration. The prerequisite is appropriate antibiotic stewardship, not withdrawal of necessary antimicrobial treatment from neutropenic or otherwise immunocompromised patients. In practice, this means avoiding redundant anaerobic coverage, shortening unnecessary courses, de-escalating when clinical data permit, and weighing infectious efficacy against collateral loss of metabolically important commensals.

Precision nutrition can provide substrates to surviving microbial pathways. Resistant starch, fermentable fibers, inulin-type fructans, and oligosaccharides may increase SCFA production, but only if compatible fermenters are present. In a study using historical controls, combined resistant starch and a glutamine–fiber–oligosaccharide formulation was associated with preservation of the gut microbiota and lower overall aGVHD incidence [24], The difference in GI-aGVHD incidence was not statistically significant, and the contribution of each component remains uncertain. Mucositis, nausea, diarrhea, ileus, and poor intake may nevertheless limit feasibility. Interventions should therefore be matched to enteral tolerance and baseline functional capacity; supplying a substrate to a community lacking the necessary genes may be ineffective or drive undesirable cross-feeding.

Probiotics or defined consortia could restore discrete functions such as butyrate synthesis, indole production from tryptophan, bile salt deconjugation, or secondary bile acid formation. Compared with conventional FMT, defined products offer tighter manufacturing control, although successful activity still requires engraftment and suitable substrates. A clinically informed experimental study linked the carbapenem-associated loss of Bacteroides ovatus to corticosteroid nonresponse; reintroduction of B ovatus limited mucus degradation and improved GVHD, illustrating restoration of an ecological function rather than delivery of a single metabolite [64]. Translation requires strain-level evaluation of virulence, antimicrobial susceptibility, translocation, and horizontal resistance-gene transfer.

Autologous FMT uses a recipient sample banked before intensive conditioning or antimicrobial exposure and may reduce exogenous pathogen transmission. After antibiotic treatment in allo-HCT, this approach restored diversity and recipient-derived strains [65]. Its use requires prospective stool collection, and the stored pretransplant community may already contain dysbiosis or multidrug-resistant organisms.

Third-party donor FMT can provide a broader microbial and metabolic repertoire than autologous FMT, but its clinical interpretation depends on the therapeutic objective and disease status of the enrolled population. An initial single-group study conducted after allo-HCT primarily established feasibility, donor-strain acquisition, and recovery of microbial diversity; it was not designed to demonstrate efficacy against GI-GVHD [66]. In a prospective single-center, single-arm study, van Lier et al. administered donor FMT to 15 allo-HCT recipients with steroid-refractory or steroid-dependent acute or late acute intestinal GVHD. Ten patients achieved a complete clinical response within one month, but the absence of a concurrent control group precludes definitive attribution of these responses to FMT [67]. MaaT013, a standardized pooled allogeneic microbiota ecosystem product, was subsequently evaluated in 24 patients with grade III–IV steroid-refractory GI-aGVHD in a multicenter, single-arm phase 2 study. The primary endpoint was gastrointestinal complete response or very good partial response at day 28, achieved by 7 of 24 patients (29%). The gastrointestinal overall response rate, including partial responses, was 9 of 24 (38%) [68].

Randomized FMT studies must also be interpreted according to their prespecified objectives. The randomized, double-blind trial conducted by Rashidi et al. primarily evaluated the four-month rate of all-cause infection in allo-HCT and acute myeloid leukemia populations rather than the prevention or treatment of GI-GVHD [69]. Similarly, the preplanned interim analysis reported by Reddi et al. primarily evaluated donor-dependent differences in microbial engraftment in the aGVHD-prevention setting [70]. Neither study should be cited as definitive evidence that FMT prevents or treats GI-GVHD. Donor selection should therefore incorporate colonization fitness, metabolic function, and biosafety rather than microbial diversity alone.

Microbial engraftment, increased alpha diversity, and the restoration of metabolite-producing taxa should be interpreted as functional or pharmacodynamic endpoints rather than evidence of clinical benefit. Clinical efficacy requires prespecified patient-centered outcomes, including GI-GVHD incidence or severity, treatment response, corticosteroid exposure, infection, nonrelapse mortality, overall survival, and preservation of immune reconstitution and GVL activity. Prevention studies in patients without established GI-GVHD should therefore be evaluated separately from therapeutic studies enrolling patients with active or steroid-refractory disease.

5.4. Therapeutic Timing, Delivery and Patient Stratification

Intervention goals change across the transplant timeline. Before conditioning, the priorities are nutritional optimization, avoidance of unnecessary antibiotics, preservation of ecological resilience, and, where appropriate, collection of autologous stools. At this stage, interventions should emphasize the safety and maintenance of baseline microbial function rather than the suppression of alloimmunity.

During the early period after conditioning, barrier preservation becomes the dominant objective. Locally delivered SCFAs, indole derivatives, TUDCA, or desaminotyrosine could in principle limit apoptosis and support repair before donor T-cell expansion reaches its maximum [11,13,19,45]. Around neutrophil engraftment, enteral nutrition, function-directed prebiotics, and cautious microbial restoration may prevent the prolonged depletion of obligate anaerobes. FMT in this interval requires clinical stability, control of active infection, and coordination with antibiotics, because ongoing broad-spectrum therapy may block donor-community engraftment.

Once GI-GVHD is established, treatment should be aligned with the prevailing mechanism. Predominant epithelial injury with limited systemic inflammation may be suited to tissue-protective postbiotics, whereas established donor T-cell activation is likely to require concomitant glucocorticoid or JAK-pathway therapy. In steroid-refractory disease, FMT, pooled microbial products, and regenerative metabolites should be evaluated as adjuncts to evidence-based immunomodulation rather than replacements for it.

Stratification must extend beyond relative taxonomic abundance. A useful clinical model would integrate absolute microbial load, genes for butyrate synthesis, BSH and bai activity, tryptophan metabolism, fecal and plasma metabolites, recent antimicrobial and nutritional exposures, and epithelial injury markers such as REG3α and ST2. Receptor activity could add a pharmacodynamic layer. For example, loss of secondary bile acids together with increased T-cell FXR activity could identify a candidate subgroup for FXR-directed intervention, whereas preserved SCFA biosynthetic genes in the setting of substrate deficiency could favor a prebiotic over direct SCFA replacement (Figure 3) [14,26,27,28].

Figure 3. Therapeutic translation after allo-HCT. The upper timeline locates selected human clinical studies within transplant phases. The gray solid arrow indicates temporal progression, not treatment efficacy, a recommended sequence or validated optimal treatment windows. Stage colors progress from blue through green and ochre to red. UDCA prophylaxis has clinical trial evidence. The RS + GFO study used historical controls and mucositis as its primary outcome. Early FMT studies assessed microbiota restoration or feasibility; donor FMT for active intestinal GVHD provided an uncontrolled treatment signal. MaaT013, a pooled donor-derived microbial ecosystem product, has single-arm clinical evidence in steroid-refractory disease. These human study designs provide different levels of evidence, and restoration or engraftment does not itself prove clinical benefit. The green middle row and mouse icon identify direct preclinical GVHD-model interventions: butyrate/propionate, selected indoles, DAT, TUDCA and GPR43 agonism. TUDCA is distinct from clinically studied UDCA. The ochre lower row and dashed arrows denote proposed local delivery, receptor modulation, microbial products and combinations requiring prospective validation. Illustrative particle colors distinguish interventions without implying chemical structure, evidence strength or clinical efficacy. Tissue/cell illustrations indicate disease context rather than proven mechanisms of FMT or MaaT013. ‘+’ indicates a proposed combination. Allo-HCT, allogeneic hematopoietic cell transplantation; DAT, desaminotyrosine; FMT, fecal microbiota transplantation; GFO, glutamine, fiber and oligosaccharides; GVHD, graft-versus-host disease; GI-GVHD, gastrointestinal graft-versus-host disease; GPR43, G protein-coupled receptor 43; RS, resistant starch; TUDCA, tauroursodeoxycholic acid; UDCA, ursodeoxycholic acid. MaaT013 is a product designation, not an abbreviation.

5.5. Pharmacokinetic, Pharmacodynamic and Safety Considerations

Translational status, intended therapeutic window, principal outcomes, and major safety limitations of these interventions are summarized in Table 2. Microbiota-associated metabolites require pharmacokinetic analysis by anatomical compartment. Stool concentration represents net production minus microbial utilization, epithelial uptake, and transit losses; it is not equivalent to mucosal exposure. Plasma values additionally reflect barrier permeability, portal extraction, hepatic biotransformation, and renal elimination. Translational studies should therefore obtain serial stool and blood samples and, whenever feasible, intestinal tissue. Across these comparisons, culture-medium concentrations represent nominal exposures rather than measured free concentrations at target cells. Fecal or tissue amounts expressed per unit mass cannot be directly converted to molarity without additional information, and animal dosing regimens should be distinguished from measured tissue concentrations. The values presented in Section 3.1, Section 3.2 and Section 3.3 therefore provide contextual benchmarks rather than established physiological or therapeutic ranges for GI-GVHD. Target-cell exposure after allo-HCT remains insufficiently characterized for many of the metabolites discussed.

Table 2. Clinical studies, preclinical interventions and proposed strategies targeting the microbiota–metabolite–host axis after allo-HCT.

Strategy/Study	Population and Disease Status	Purpose and Tested/Proposed Timing	Design/Evidence Stage	Main Endpoint and Finding	Microbiota/Functional Endpoint	Principal Limitation and GVL Assessment	Refs.
A. Clinical studies and human observational evidence
Antibiotic exposure/stewardship rationale	Allo-HCT recipients; antimicrobial exposure cohorts rather than active GI-GVHD treatment enrollment	Preserve commensal function while treating infection	Human observational comparisons with complementary murine experiments	Anaerobe-active broad-spectrum exposure associated with GVHD-related mortality; not a randomized stewardship efficacy trial.	Microbial disruption and loss of protective commensals.	Infection treatment cannot be withheld; confounding by indication. Direct GVL preservation not established.	[5,20]
Resistant starch + GFO	Allo-HCT recipients without established GVHD at intervention; n = 49 vs. 142 historical controls	Prevention; conditioning to day +28	Prospective supplementation study with historical controls	Primary: mucositis incidence/duration. aGVHD was secondary; overall aGVHD lower, but GI-GVHD incidence difference was not significant.	Diversity and butyrate-producing taxa preserved; fecal butyrate maintained/increased more often.	Nonrandomized historical comparison; intake/tolerance varied. GVL preservation not established.	[24]
UDCA prophylaxis	Allo-HCT recipients; n = 242; no active disease targeted at baseline	Hepatic-complication prevention; before conditioning to day +90	Randomized open-label multicenter trial	Hepatic-complication prevention was the trial focus. Severe aGVHD and intestinal involvement also reduced; not treatment of active GI-GVHD.	No microbial restoration mechanism established by the trial.	Trial does not establish an FXR mechanism; UDCA is distinct from TUDCA. Direct GVL preservation not established.	[60]
Autologous FMT	Antibiotic-exposed allo-HCT recipients; published microbiota analysis n = 25	Reconstitution after antibiotic-associated disruption; not active GI-GVHD therapy	Randomized auto-FMT vs. no-intervention study; microbiota analysis	Published analysis established microbiota reconstitution, not GI-GVHD clinical efficacy.	Improved diversity and recovery of pre-antibiotic community composition.	Requires banked stool and screening; stored pathogens possible. GVL preservation not established.	[65]
Third-party donor FMT	Post-allo-HCT microbiota restoration; not active GI-GVHD treatment enrollment; 13 received capsules	Feasibility and microbiome restoration after transplantation	Single-group pilot study	Primary endpoint: ability to receive oral FMT capsules; feasibility demonstrated.	Donor taxa acquisition and diversity recovery.	Not designed to establish GI-GVHD prevention efficacy; pathogen/AMR transmission. GVL preservation not established.	[66]
Third-party donor FMT	Established steroid-refractory/dependent acute or late acute intestinal GVHD; n = 15	Treatment of active intestinal disease	Prospective single-center single-arm study	Clinical intestinal GVHD response assessed; 10/15 achieved complete response within one month.	Increased diversity, partial donor engraftment and butyrate-producing taxa.	Concurrent treatment and no control group limit attribution; infection/AMR risk. GVL preservation not established.	[67]
Pooled donor microbiota MaaT013	Grade III–IV steroid-refractory GI-aGVHD; prospective cohort n = 24	Treatment of established refractory GI disease	Multicenter single-arm phase 2 study	Primary D28 GI CR + VGPR: 7/24 (29%). GI-ORR including PR: 9/24 (38%). The separate EAP cohort is not pooled with these results.	Microbiota/ecosystem restoration evaluated separately from clinical response.	No concurrent comparator; infections require adjudication [71]. Pooled donor product, not a defined-strain consortium; GVL unproven clinically.	[68]
Third-party donor FMT	Allo-HCT n = 74; separate AML n = 26; after neutrophil recovery; not active GI-GVHD treatment enrollment	Infection prevention/microbiota restoration	Randomized double-blind placebo-controlled phase 2 trial	Primary: 4-month all-cause infection rate; no significant reduction. A randomized design does not make this a GI-GVHD efficacy trial.	Microbiota restoration despite unmet clinical primary endpoint.	Keep AML and allo-HCT findings separate; endpoint cannot be replaced by diversity improvement. GVL preservation not established.	[69]
Third-party donor FMT	Preventive post-allo-HCT setting; 20 received FMT; 17 evaluable for engraftment	Donor selection/engraftment optimization; preventive context	Single-arm donor-selection run-in; preplanned interim analysis	Primary: donor engraftment. Severe aGVHD assessed as a clinical outcome; no randomized preventive efficacy conclusion from this report.	Donor-dependent differences in engraftment.	Interim/donor-cohort analysis is distinct from randomized preventive efficacy. GVL preservation not established.	[70]
B. Preclinical interventions and mechanistic extrapolation
Butyrate/propionate replacement	Murine acute GVHD; IEC experiments	Early intestinal protection in the tested transplant regimens	Preclinical GVHD intervention	Reduced intestinal injury/severity; butyrate also acted independently of GPR43.	Epithelial histone acetylation; GPR43–ERK–NLRP3/IL-18 engagement.	Dose and compartmental exposure; clinical efficacy and preservation of GVL under clinical dosing remain unresolved.	[11,12]
Indole-3-carboxaldehyde and indole-producing bacteria	Murine acute GVHD	Tissue protection around transplantation	Preclinical GVHD intervention	Reduced intestinal injury and GVHD; GVL retained in directly tested murine models.	Type I IFN-dependent protective responses.	Tested indole regimen supports experimental GVL preservation; cannot be generalized to all indoles or human use.	[13]
DAT supplementation	Murine GVHD; human organoids; observational patient biomarker data	Pretransplant regenerative protection in the tested regimen	Preclinical efficacy; human ex vivo/observational support	Improved regeneration and experimental GVHD; leukemia models showed retained/enhanced antitumor responses.	mTORC1- and epithelial STING-dependent regenerative program.	No human DAT dosing efficacy trial; PK unresolved. Human survival/relapse associations do not prove GVL preservation.	[19]
TUDCA administration	Murine allo-HCT models; human cellular experiments	Prophylactic epithelial protection	Preclinical GVHD intervention	Reduced epithelial injury and antigen presentation; cytotoxicity and in vivo GVL retained in tested models.	Epithelial stress and MHC-II-associated antigen-presentation changes.	No allo-HCT clinical efficacy trial in this study; distinct from UDCA. GVL evidence is model-specific.	[45]
Bacteroides ovatus supplementation	Murine dysbiotic GVHD; patient microbiome associations	Restore ecology after dysbiosis in experimental regimens	Preclinical intervention with clinical association	Bacterial reintroduction reduced experimental GVHD; human abundance linked to steroid response.	Reduced mucus-degrading microbial activity; ecological protection.	No human B. ovatus dosing efficacy trial; strain safety, translocation and antibiotic compatibility.	[64]
FFAR2/GPR43 agonism	Murine acute GVHD	Recipient tissue protection during experimental GVHD	Preclinical receptor-directed intervention	Agonist reduced disease; protection depended on recipient GPR43.	ERK and epithelial-associated NLRP3/IL-18 responses.	Receptor compartment and systemic effects require resolution; human efficacy/GVL preservation unresolved.	[12]
Donor-T-cell FXR inhibition	Murine GVHD; human T-cell experiments	Limit donor T-cell inflammatory amplification	Preclinical target validation; selective drug strategy remains investigational	Donor-T-cell FXR deletion improved survival; agonism worsened GVHD.	FXR-dependent T-cell signaling/proliferation.	Genetic deletion is not equivalent to a safe clinical antagonist; hepatic/intestinal BA homeostasis and GVL require assessment.	[14]
AhR/PXR/TGR5 modulation	Non-GVHD mucosal, barrier, regeneration or myeloid models	Candidate target-cell-specific protection	Mechanistic extrapolation; no established clinical GI-GVHD efficacy	Mucosal IL-22, barrier protection, stem-cell regeneration or myeloid anti-inflammatory responses, depending on the target.	Target-specific transcription, barrier or regeneration readouts.	Ligand/tissue specificity; PXR drug interactions and hepatobiliary effects; GVL preservation unestablished.	[39,40,41,42]
C. Proposed strategies requiring product-specific validation
Function-guided nutrition/prebiotics	Prospective allo-HCT populations selected by microbial function and enteral tolerance	Support residual metabolite production; preventive setting proposed	Proposed personalized strategy; specific tested RS + GFO regimen listed above	Requires prospective GI-GVHD endpoints; broad efficacy cannot be inferred from one formulation.	Candidate measures: metabolic genes, SCFAs, bile-acid transformation and indole output.	Substrate–community mismatch; gastrointestinal intolerance. GVL preservation unestablished.	[24]
Colon-targeted metabolite formulations	Future early post-allo-HCT/epithelial-injury populations	Increase local exposure; prodrugs/controlled release proposed	Proposed formulation development based on preclinical parent compounds	No formulation-specific GI-GVHD benefit established by the cited parent-metabolite studies.	Measured mucosal exposure and epithelial target engagement required.	Stability, release kinetics and systemic absorption; extrapolation from nominal culture concentrations.	[11,13,19,45]
Function-defined consortium proposed in this review	Future selected allo-HCT recipients	Restore specified microbial metabolic functions	Product-specific proposal; FMT and natural-strain studies provide rationale	No efficacy of a specified function-defined product is established by the FMT/natural-strain studies cited here.	Metabolite production, strain engraftment and functional pathway flux.	Distinguish this proposed product from MaaT013; manufacturing, engraftment and infection risk require specific testing.	[64,65,66,67,68]
Engineered microbial strains	Future selected allo-HCT recipients	Deliver defined metabolic activity locally	Proposed strategy; cited natural-strain evidence does not validate engineered products	No engineered-product GI-GVHD efficacy established by the cited B. ovatus study.	Expression stability, delivered metabolite output and target-cell engagement.	Translocation, persistence and genetic stability; no established GVL preservation.	[64]
Combined metabolite replacement	Future exposure- and pathway-selected allo-HCT populations	Combine complementary epithelial/immune effects	Proposed combination; single-agent evidence is not combination efficacy	Requires factorial dose and clinical outcome testing.	Compound-specific exposure and simultaneous pathway engagement.	Synergy or antagonism cannot be presumed; PK compatibility and GVL must be tested.	[11,13,19,45]

Evidence interpretation: Stage describes the specific cited study, not regulatory approval or the entire intervention class. Prevention, active-disease treatment, microbiota reconstitution and infection prevention are distinct objectives. Engraftment and diversity are functional endpoints and do not by themselves demonstrate patient benefit. MaaT013 is a pooled donor-derived ecosystem product, distinct from a defined-strain consortium. Clinical relapse reporting is not equivalent to direct GVL validation. Experimental GVL findings apply to the compounds, doses and models tested. Safety: FMT and live microbial products require donor/product pathogen screening, antimicrobial-resistance surveillance and adjudication of infectious events [71]. Proposed-strategy citations provide biological rationale from parent compounds or related products; they do not establish efficacy of the proposed formulation, engineered strain, consortium or combination. References correspond to the numbered reference list. Abbreviations: aGVHD, acute graft-versus-host disease; AhR, aryl hydrocarbon receptor; allo-HCT, allogeneic hematopoietic cell transplantation; AML, acute myeloid leukemia; AMR, antimicrobial resistance; BA, bile acid; CR, complete response; D28, day 28; DAT, desaminotyrosine; EAP, expanded access program; ERK, extracellular signal-regulated kinase; FFAR2, free fatty acid receptor 2; FMT, fecal microbiota transplantation; FXR, farnesoid X receptor; GFO, glutamine, fiber and oligosaccharide mixture; GI, gastrointestinal; GI-GVHD, gastrointestinal graft-versus-host disease; GVL, graft-versus-leukemia; HCAR2, hydroxycarboxylic acid receptor 2; IEC, intestinal epithelial cell; IFN, interferon; IL, interleukin; MHC-II, major histocompatibility complex class II; mTORC1, mechanistic target of rapamycin complex 1; NLRP3, NLR family pyrin domain containing 3; ORR, overall response rate; PK, pharmacokinetics; PR, partial response; PXR, pregnane X receptor; RS, resistant starch; SCFA, short-chain fatty acid; STING, stimulator of interferon genes; TGR5, G protein-coupled bile acid receptor 1; TUDCA, tauroursodeoxycholic acid; UDCA, ursodeoxycholic acid; VGPR, very good partial response.

Chemical forms should also be resolved. SCFA salts, ester prodrugs, and protein-bound molecules differ in absorption and distribution. Bile acids undergo transporter-dependent enterohepatic cycling, microbial deconjugation, hepatic reconjugation, and structure-dependent receptor activation. Indoles are rapidly oxidized, sulfated, and glucuronidated. Hepatic or renal impairment may therefore change systemic exposure even if bacterial production remains constant.

Pharmacodynamic studies should demonstrate pathway engagement rather than rely on concentration alone. Potential measures include epithelial histone acetylation and IL-18 for SCFA treatment, CYP1A1-associated transcription for AhR ligands, FGF19- or SHP-related responses for FXR modulation, cyclic adenosine monophosphate signaling for TGR5, and regenerative epithelial programs for desaminotyrosine. Because opposite effects can occur in epithelium and lymphocytes, bulk blood or fecal markers should not substitute for target-cell-resolved measurements.

Prospective assessment of drug interactions is essential. AhR and PXR ligands may regulate CYP enzymes and transport proteins, while altered bile acids or barrier permeability can change oral absorption [62]. Tacrolimus, cyclosporine, sirolimus, ruxolitinib, glucocorticoids, and azole antifungals deserve particular attention because exposure is often CYP3A dependent or the therapeutic window is narrow.

FMT and live biotherapeutics carry hazards distinct from those of purified postbiotics. Potentially transmissible agents include enteric bacteria, fungi, parasites, viruses, multidrug-resistant organisms, and uncharacterized pathogenic strains. FMT has transmitted extended-spectrum β-lactamase-producing Escherichia coli bacteremia to immunocompromised recipients, including one fatal infection [71]. Donor qualification must therefore cover resistant organisms, enteric and blood-borne pathogens, relevant emerging infections, and the resistome. Traceable manufacturing, quarantine, batch testing, and prolonged safety monitoring are also required.

Defined metabolites are easier to control but are not inherently benign. Bile acid derivatives may cause hepatobiliary, cholestatic, or metabolic toxicity; sustained AhR stimulation can be immunotoxic or carcinogenic depending on ligand structure; and systemic TGR5 agonism may affect gallbladder and enteroendocrine physiology. Safety endpoints must extend beyond short-term improvement in GVHD to infection, organ toxicity, immune reconstitution, relapse, and maintenance of GVL. A time-limited, tissue-restricted intervention that replaces a documented missing signal is likely to offer the most favorable therapeutic index.

6. Translational Challenges, Future Directions and Conclusions

6.1. Causality Versus Consequence

A major translational problem is determining whether an abnormality in microbiota-associated metabolite profiles precedes GI-GVHD, amplifies inflammation after it begins, or simply reflects treatment and tissue damage. Most human studies associate fecal or circulating metabolite profiles with subsequent clinical outcomes. These analyses remain vulnerable to reverse causation. Important time-varying confounders include conditioning intensity, antibiotic exposure, diet, diarrhea, organ dysfunction, immunosuppression, and the underlying malignancy [26,27,28,72]. Depletion of a metabolite considered protective therefore does not by itself establish that the depletion contributes causally to disease.

The evidence instead supports a reciprocal and evolving relationship. Conditioning damage and donor T-cell inflammation alter the intestinal niche and suppress microbial functions such as SCFA generation and secondary bile acid conversion. The resulting loss of metabolic support can then impair repair or strengthen inflammatory signaling, converting an initial consequence into a feed-forward amplifier [14,15]. T-cell-driven inflammation, for example, reduces microbial BSH activity and remodels the bile acid pool; the altered ligand environment can subsequently enhance donor T-cell activation through FXR-dependent mechanisms [14].

Demonstrating causality requires concordant evidence from several complementary systems. Serial clinical cohorts should show that the metabolic change precedes symptomatic GI-GVHD and persists after adjustment for time-updated exposures. Germ-free or gnotobiotic animals bearing defined communities can isolate a pathway from broader taxonomic disruption. Stable-isotope experiments are needed to apportion dietary, bacterial, and host contributions. Depletion and replacement studies should establish exposure–response relationships in the lumen, mucosa, portal blood, and systemic circulation. Finally, rescue designs and conditional receptor deletion must test whether the proposed host pathway is required. Work on butyrate, indoles, bile acids, and desaminotyrosine provides strong precedents [11,12,13,14,19], but reproducibility across transplant protocols, microbial backgrounds, and clinically relevant treatments remains necessary.

6.2. Methodological and Model Limitations

Animal models do not fully reproduce human metabolite exposure. Bile acids illustrate this limitation: conventional mice possess abundant CYP2C70-derived muricholic acids, whereas humans have pools dominated by cholic acid, chenodeoxycholic acid, and their microbially modified derivatives. Because bile acid structures differ in activity at FXR, TGR5, PXR, and VDR, species composition can materially change receptor occupancy and drug response [73]. Humanized bile acid models may therefore be required before receptor-directed therapies are translated to GVHD trials.

Microbiota-dependent phenotypes also vary with supplier, cage, housing, diet, baseline community, and antibiotic protocol [74]. These factors can alter disease severity and metabolite exposure even in genetically identical animals. Reproducibility requires the transparent reporting of diet and housing, characterization of baseline microbiota and absolute bacterial burden, precise antimicrobial schedules, and replication in independently maintained colonies.

Human studies face additional technical constraints. Amplicon sequencing of the 16S rRNA gene has limited strain and functional resolution and generally reports relative rather than absolute abundance; an apparently enriched taxon may simply remain after the depletion of others [75,76]. Shotgun metagenomics, quantitative profiling, metatranscriptomics, and direct functional-gene assays should therefore complement compositional surveys. Even these methods measure genetic potential rather than pathway flux, so metabolomics and, when possible, isotope tracing remain indispensable.

Stool metabolites likewise cannot be treated as direct measures of the exposure experienced by host cells. Fecal values integrate synthesis, microbial consumption, host uptake, transit, and biotransformation, and may diverge from concentrations at the epithelial surface, in lamina propria, or in portal and peripheral blood. Direct comparisons of luminal and mucosa-associated communities further show that fecal samples incompletely represent mucosal colonization and host interaction [77]. Assigning a metabolite effect to a compartment therefore requires paired stool, plasma, and intestinal tissue analyses.

Mechanisms defined in acute GI-GVHD should not automatically be extended to chronic GVHD. Acute disease is characterized by epithelial damage, alloimmune activation, and inflammatory cytotoxicity, whereas chronic disease incorporates prolonged immune dysregulation, abnormal repair, and fibrosis. A metabolite that protects the acute epithelial barrier may therefore have a different function during chronic remodeling [59].

6.3. Conclusions

Microbiota-derived metabolites cannot be assigned fixed protective or pathogenic identities. Their activity in GI-GVHD emerges from chemical form, local concentration, tissue distribution, sensor identity, responding cell, and phase of disease. A pathway that repairs epithelium in one compartment may increase donor T-cell or inflammatory myeloid activity in another. Stool abundance alone therefore cannot predict biological effect, and systemic receptor modulation may not reproduce the actions of a locally generated microbial ligand.

The next phase of the field should move from nonspecific increases in microbial diversity to mechanism-defined, cell- and receptor-resolved postbiotic pharmacology. Progress will require causal models, anatomically resolved exposure data, validated markers of target engagement, and delivery systems that place a metabolite or receptor modulator in the intended intestinal compartment. Stratification should also incorporate antimicrobial history, nutritional state, baseline microbial functions, epithelial injury, and the current stage of GVHD.

The clinical value of modifying host responses to microbiota-associated metabolites will ultimately depend on protecting tissue without imposing unacceptable infectious, metabolic, or oncological risk. GI-GVHD control, GVL activity, immune reconstitution, infection, and malignant relapse should be evaluated as interconnected outcomes rather than as isolated efficacy and safety domains. Under these conditions, metabolite-sensing pathways offer promising but highly context-dependent targets for the precision prevention and treatment of GI-GVHD.