Gut Microbiota in Hepatocellular Carcinoma: Etiology-Specific Evidence, Mechanisms, and Translational Challenges

Abstract

Hepatocellular carcinoma (HCC) usually develops in patients with chronic liver disease and remains a major cause of cancer-related death worldwide. Increasing evidence suggests that dysbiosis of the gut microbiota is related to hepatocarcinogenesis through the gut-liver axis. Patients with HCC often show enrichment of pro-inflammatory or endotoxin-producing bacteria and loss of short-chain fatty acid-producing and barrier-protective taxa. These changes may increase intestinal permeability and microbial translocation and may contribute to chronic hepatic inflammation, immune disturbance, and metabolic changes. Lipopolysaccharide, secondary bile acids, and trimethylamine-N-oxide are among the microbial products that may connect intestinal dysbiosis with liver fibrosis and tumor progression. Clinical studies have also examined microbial signatures for early HCC detection, risk stratification, recurrence, and response to immune checkpoint inhibitors. However, most cohorts are small and differ in etiology, region, medication exposure, and sequencing method. Animal studies indicate that antibiotics, probiotics, fecal microbiota transplantation, and diet can change tumor development or treatment response, but direct clinical evidence in HCC remains limited. This review summarizes the compositional, mechanistic, and clinical evidence and discusses the present status of microbiota-targeted strategies in HCC.

Plain Language Summary

Liver cancer usually develops after many years of liver injury. Bacteria in the intestine may be involved because bacterial products and metabolites can reach the liver through the portal vein. Studies have found that the gut bacteria of patients with hepatocellular carcinoma differ from those of people without cancer. Some of these differences may be useful for early detection or for predicting response to immunotherapy. However, the results are affected by region, the cause of liver disease, medicines, diet, and laboratory methods. Microbiome tests and treatments that change gut bacteria are therefore not part of routine HCC care at present. Further clinical studies are needed to determine whether they are useful and safe.

Introduction

Hepatocellular carcinoma (HCC) accounts for most primary liver cancers and remains a major cause of cancer-related death worldwide.^1,2 Most cases occur in chronic liver disease caused by hepatitis B or C virus (HBV/HCV), excessive alcohol intake, or non-alcoholic steatohepatitis (NASH).² Antiviral treatment has reduced the contribution of viral hepatitis in some regions, whereas NASH is the most rapidly increasing indication for liver transplantation in the United States and was the second leading indication among transplant candidates with HCC in 2019.³ The liver is continuously exposed to intestinal products through the portal circulation, which forms the anatomical basis of the gut-liver axis.⁴ Gut bacteria help maintain the intestinal barrier, metabolize bile acids and dietary fiber, and produce short-chain fatty acids (SCFAs), trimethylamine, and indole derivatives that can reach the liver.⁵ When this balance is disturbed, barrier dysfunction, inflammation, immune changes, and altered metabolism may contribute to HCC development.⁶ Human sequencing studies have described dysbiosis in HCC, but the reported taxa vary among cohorts. Escherichia-Shigella enrichment has been observed in several studies,^7–9 whereas lower abundance of Akkermansia or SCFA-producing taxa has mainly been reported in selected metabolic HCC and cirrhosis cohorts.^10–13 Mendelian randomization suggested associations for several taxa, including Desulfovibrio piger (OR = 1.304), although these findings still depend on the validity of the genetic instruments.¹⁴ Viable bacteria and bacterial DNA have also been detected in HCC tissue.^15,16 This review therefore focuses on findings that are sufficiently consistent to support further biomarker or treatment studies and on those that remain associative or preclinical. We pay particular attention to differences among HCC etiologies, confounding in human cohorts, and the distinction between human association, experimental causality, and clinical intervention.

Literature Search Strategy and Scope

For this narrative review, we searched PubMed, Web of Science, and Scopus, with the last search performed on July 31, 2026. Search terms included hepatocellular carcinoma, HCC, gut microbiota, gut microbiome, gut-liver axis, microbial metabolite, biomarker, immunotherapy, probiotics, diet, antibiotics, fecal microbiota transplantation, bacteriophage, and engineered bacteria. We mainly selected human HCC cohorts, prospective studies, studies with cirrhosis or etiology-matched controls, and experiments that directly examined a proposed mechanism. Evidence from other solid tumors or chronic liver diseases was used only when HCC-specific evidence was unavailable and is described as indirect evidence. This search was used to define the scope of a narrative review; it was not performed as a systematic review or meta-analysis.

Compositional Changes of Gut Microbiota in HCC

The use of 16S rRNA gene sequencing and shotgun metagenomics has expanded the study of gut microbial communities in liver disease.¹⁷ In HCC, dysbiosis is generally described as a change in microbial diversity, composition, or function compared with healthy individuals or patients with cirrhosis without HCC.¹⁸ However, direct comparisons among HCC etiologies remain limited. An analysis of cirrhosis datasets found no robust etiology-specific microbial signal after geographic variation was considered.¹⁹ Therefore, tumor stage, geographic region, treatment history, and the choice of comparator group should be considered together when interpreting individual studies. The following section summarizes diversity, enriched and depleted taxa, etiology-related differences, and the available evidence on intratumoral bacteria. Representative microbial signatures reported in HCC are summarized in Table 1.

Table 1.

Core Microbial Signatures in HCC

Category	Taxa	Change	Key Function	Evidence Basis*
Pro-tumorigenic	Escherichia-Shigella	↑↑	LPS-associated inflammatory potential	Repeated human associations7–9,17
	Klebsiella	↑↑	PBP1B-TLR4 signaling and EMT	Human association + mechanistic models20
	Veillonella	↑	Metabolic/inflammatory association; mechanism uncertain	Human association7
	Fusobacterium	↑	FadA-mediated endothelial barrier disruption	Human fecal association + non-HCC mechanistic data7,21
Protective	A. muciniphila	↓↓	Barrier support and immune modulation	Human ICI association + preclinical data6,10,22,23
	SCFA-producing taxa	↓↓	Butyrate/propionate-related homeostasis	Human association + functional evidence11,12,17,24
	Bifidobacterium	↓	SCFA-associated immune regulation	Human association; species-level causality uncertain11,13,25

Notes: Symbols: ↑, increased abundance or activity; ↓, decreased abundance or activity; double arrows indicate a larger or more consistent directional change across the cited evidence. *Evidence basis describes the main source of support for each statement. Human association does not establish causality, and preclinical evidence does not demonstrate clinical benefit.

Abbreviations: LPS, lipopolysaccharide; PBP1B, penicillin-binding protein 1B; TLR4, Toll-like receptor 4; EMT, epithelial–mesenchymal transition; MAPK, mitogen-activated protein kinase; TAM, tumor-associated macrophage; FadA, Fusobacterium adhesin A; SCFA, short-chain fatty acid; ICI, immune checkpoint inhibitor.

Alpha and Beta Diversity Alterations

Alpha diversity describes the number and distribution of taxa within a sample. Findings in HCC have not been uniform and depend strongly on the comparator group. In the discovery cohort reported by Ren et al (2019), 75 patients with early HCC and cirrhosis were compared with 40 patients with cirrhosis and 75 healthy controls. Fecal microbial diversity was lower in cirrhosis than in healthy controls but was higher in early HCC than in cirrhosis (Shannon index, P = 0.0234; Simpson and inverse Simpson indices, both P = 0.0068).¹⁷ Weighted UniFrac analysis showed broadly overlapping community distributions rather than complete separation of the three groups. Other cohorts have reported lower diversity in HCC, and a smaller study of patients with cirrhosis also identified a microbiota profile associated with the presence of HCC.^8,12,13,26 These findings indicate that alpha diversity alone is not a stable diagnostic feature.

Enrichment of Pro-Tumorigenic Taxa

Several cohorts have reported enrichment of Gram-negative bacteria, particularly members of Proteobacteria, in patients with HCC.^8,9 Escherichia-Shigella, Klebsiella, and Enterobacter are among the genera reported in these cohorts. Wang et al found that pathogenic Klebsiella pneumoniae could translocate from the intestine to the liver. Its surface protein penicillin-binding protein 1B (PBP1B) interacted with TLR4 on HCC cells and activated PI3K/AKT and NF-kB signaling, with increased proliferation, epithelial-mesenchymal transition, and metastasis.²⁰ In an advanced-HCC cohort, Enterobacteriaceae abundance correlated with markers of liver injury, but this association did not establish a direct barrier mechanism.⁸ Streptococcus and Enterococcus have also been reported to increase in liver disease. Metagenomic analyses identified lantibiotic-encoding Streptococcus as a possible feature of NAFLD and cirrhosis, but its HCC-specific prognostic relevance has not been established.²⁷ Findings for Bacteroidetes are less consistent. Mendelian randomization suggested positive associations for Bacteroides ovatus, B. stercoris, and Paraprevotella xylaniphila.¹⁴ Enterotoxigenic B. fragilis induces an IL-17-dependent tumor-promoting response in a mouse model of colon tumorigenesis; its relevance to HCC remains indirect.²⁸ Fusobacterium nucleatum can disrupt endothelial junctions through FadA, but this mechanism was not established in HCC.²¹ Data on the fungal microbiome are more limited. Candida albicans was increased in an HCC cohort and promoted hepatocarcinogenesis in mice through an NLRP6-dependent mechanism.²⁹

Depletion of Protective, Anti-Inflammatory Taxa

Several bacteria that help maintain the intestinal barrier or produce SCFAs are less abundant in selected HCC cohorts. Akkermansia muciniphila was reduced in NAFLD-related HCC and in experimental HCC models.^6,10,13 In experimental models, its administration improved barrier function and reduced inflammatory or immunosuppressive changes.^6,10 In patients receiving immune checkpoint inhibitors, higher Akkermansia abundance was associated with disease control or a better response to PD-1-based therapy,^10,23 although these studies were small. Bifidobacterium was also reduced in NAFLD-related HCC cohorts.^13,25 The NAFLD-HCC microbiome showed altered SCFA-related functions and induced Treg expansion with attenuation of CD8+ T cells in an ex vivo assay, but these findings cannot be assigned to a single Bifidobacterium species.¹¹ Other HCC and cirrhosis cohorts reported lower abundance of several SCFA-producing genera, including Blautia and Agathobacter.^12,17

HCC etiology may account for part of the variation among studies, but direct etiology-stratified evidence remains limited. Liu et al compared HBV-related HCC with non-HBV/non-HCV HCC and reported differences in taxonomic and ecological-network patterns.³⁰ Jinato et al also found distinct microbial features between viral- and non-viral-related HCC.³¹ By contrast, an analysis of cirrhosis datasets found no robust microbial signal that separated different etiologies after geographic variation was considered.¹⁹ These findings indicate that etiology-specific patterns require direct comparison in adequately matched HCC cohorts and should not be generalized from cirrhosis or from a single geographic population.

Intratumoral Microbiota and Tumor-Gut Axis

Intratumoral bacteria have been described across several cancer types, although the distribution is tumor-type specific.³² In HCC, fluorescence in situ hybridization, tissue culture, and 16S rRNA sequencing detected bacterial DNA and viable bacteria in tumor and peritumor tissues.¹⁶ The predominant phyla included Patescibacteria, Proteobacteria, Bacteroidota, Firmicutes, and Actinobacteriota, and several taxa were associated with clinicopathological features.¹⁶ However, the study did not establish that individual intratumoral species directly caused DNA damage, recurrence, or poor survival. The origin, viability, and functional importance of these organisms therefore require confirmation with strict low-biomass contamination controls.

Diagnostic and Prognostic Potential

Fecal microbial profiles have been studied as possible non-invasive biomarkers. In one Chinese cohort, a 30-OTU model distinguished early HCC from non-HCC samples with an internally cross-validated AUC of approximately 0.81.¹⁷ Models combining bacterial genera with alpha-fetoprotein also showed good discrimination in individual studies.³³ However, most estimates were obtained in single-region cohorts and external validation was limited. Diet, geography, HCC etiology, severity of cirrhosis, medication use, and analytical platform can all change the selected taxa and model performance. These models should therefore be regarded as exploratory rather than universal diagnostic tests.

Pathogenic Mechanisms via the Gut–Liver Axis

The gut-liver axis provides a direct route by which intestinal bacteria and their products may influence HCC.^6,34 Microbial products and, in some studies, viable bacteria can reach the liver through the portal circulation, whereas bile acids and other hepatic signals can alter the intestinal community.³⁵ In HCC, loss of this balance has been linked to barrier dysfunction, hepatic inflammation, immune suppression, metabolic changes, and direct oncogenic signaling.³⁶ The main mediators and their effects on liver and immune cells are summarized below and in Figure 1.

Figure 1.

Gut-liver pathways involved in hepatocellular carcinoma (HCC). Left panel: in the healthy intestine, the mucus layer, tight junctions, Akkermansia, and Faecalibacterium help maintain the barrier. During dysbiosis, the barrier is disrupted, Escherichia-Shigella and Klebsiella increase, and bacterial products, including lipopolysaccharide (LPS), B. fragilis toxin (BFT), deoxycholic acid (DCA), and trimethylamine (TMA), enter the portal circulation. Middle panel: gut-derived molecules are transported to the liver through the portal vein. Hepatic FMO3 oxidizes TMA to trimethylamine-N-oxide (TMAO). Right panel: LPS activates TLR4 on kupffer cells and induces NF-kB/STAT3 signaling and TNF-alpha and IL-6 production. Hepatic stellate cells produce collagen and promote fibrosis. Altered secondary bile acids and TMAO may influence immune and metabolic signaling, but these links are supported mainly by experimental or associative evidence.

Intestinal Barrier Dysfunction and Microbial Translocation

Increased intestinal permeability, often described as a leaky gut, is common in advanced cirrhosis, which is the background liver disease in many patients with HCC.^37,38 Under normal conditions, tight-junction proteins such as zonula occludens-1 (ZO-1), claudins, and occludin work together with the mucus layer and antimicrobial peptides to limit passage of luminal contents.³⁹ Dysbiosis and chronic inflammation may weaken these defenses in several ways.

Mucin and Barrier Regulation

In experimental HCC models, loss of Akkermansia muciniphila was associated with impaired barrier function, whereas reintroduction of the bacterium improved barrier integrity and reduced hepatic inflammation.^6,10 A consistent HCC-specific role for individual mucin-degrading taxa has not yet been demonstrated.

Tight-Junction Disruption

Inflammatory signaling and bacterial translocation can reduce tight-junction integrity in chronic liver disease.^37–39 A high-salt challenge altered the gut microbiota, particularly by reducing Lactobacillus, and increased Th17 responses in mice and in a small human pilot study.⁴⁰ Its direct contribution to barrier failure or HCC has not been established.

Epithelial Injury and Translocation

Enterotoxigenic Bacteroides fragilis secretes B. fragilis toxin, which disrupts epithelial junctions and induces IL-17-related inflammation in colon tumor models.²⁸ This mechanism is biologically relevant to barrier injury but has not been directly demonstrated in HCC. Endotoxemia is common in advanced cirrhosis and is related to bacterial translocation and portal hypertension.³⁷ In a DEN-induced mouse model, gut sterilization or TLR4 deficiency reduced tumor promotion.³⁴ Thus, microbial translocation has strong experimental support, although its contribution in individual patients is difficult to separate from advanced cirrhosis.

Hepatic Inflammation via PAMP-PRR Signaling

After reaching the liver, pathogen-associated molecular patterns (PAMPs) bind pattern-recognition receptors on resident cells and maintain inflammation during fibrosis and tumor development.³⁴

Kupffer Cells

These hepatic macrophages respond to lipopolysaccharide through TLR4. TLR4/MyD88/NF-kB signaling increases inflammatory cytokines, including TNF-alpha, IL-1beta, and IL-6.⁴¹ Repeated exposure may maintain a persistent inflammatory macrophage phenotype during chronic liver injury.

Hepatic Stellate Cells

TLR4 and TGF-beta-related signaling can activate hepatic stellate cells and promote their transition into collagen-producing myofibroblasts. This process contributes to fibrosis and cirrhosis, which provide the background for most HCC cases.⁴²

Immune Dysregulation and Tumor Immune Evasion

Gut dysbiosis may also alter hepatic and systemic immune responses and thereby facilitate immune escape by HCC.

Treg Expansion

In an obesity-associated mouse model, lipoteichoic acid triggered gasdermin D-dependent release of IL-1beta and IL-33 from senescent hepatic stellate cells. IL-33 then promoted HCC development through activation of ST2-positive Treg cells.⁴³ This study supports an LTA-IL-33-Treg pathway but does not establish DCA or PGE2 as part of the same mechanism. Separately, single-cell analysis of recurrent HBV-related HCC identified distinct T-cell, myeloid-cell, and dendritic-cell ecosystems, although that study did not assess a microbiota-driven mechanism.⁴⁴

Myeloid and Neutrophil-Related Pathways

TMAO increased periostin through ILK/AKT/mTOR signaling in inflammatory HCC cell models and was associated with neutrophil infiltration.⁴⁵ In a separate HCC study, neutrophil extracellular traps were increased in metastatic HCC and promoted metastasis through a TLR4/9-COX2 inflammatory pathway in experimental models.⁴⁶ These findings are mainly mechanistic and do not establish TMAO or NETs as validated clinical biomarkers.

Natural Killer T-Cell Modulation

In the STAM NASH-HCC model, a breast milk-derived Akkermansia muciniphila strain increased hepatic CXCR6+ natural killer T-cell infiltration and suppressed progression from NASH to HCC. This effect was strain specific and was lost in NKT-cell-deficient mice.²² Human immunotherapy cohorts have reported an association between higher Akkermansia abundance and treatment response,^10,23 but a causal immune mechanism has not been demonstrated in patients. In acute liver injury rather than HCC, commensal Lactobacillus promoted IL-22-related barrier function and recruitment of regulatory dendritic cells;⁴⁷ extrapolation of this mechanism to HCC should therefore be cautious.

Metabolic Reprogramming by Microbial Metabolites

Microbial metabolites provide another link between the intestine and HCC.⁴⁸ SCFAs are produced by anaerobic fermentation of dietary fiber and usually support barrier and immune homeostasis, although their effects depend on dose and disease context.

Butyrate

Butyrate can inhibit histone deacetylases and regulate inflammatory signaling, but its effect depends on the host and dietary context.⁴⁹ In dysbiotic mice, prolonged feeding with fermentable inulin induced cholestasis, hepatic inflammation, and HCC, whereas germ-free or antibiotic-treated animals were protected.⁵⁰ This model indicates that increased microbial fermentation is not uniformly beneficial and should not be interpreted as evidence that dietary fiber causes HCC in humans.

Propionate

Propionate signals through free fatty acid receptors and reduced NF-kB activity, IL-6 release, and Th1/Th17-related signaling in human immune-cell experiments.²⁴ Its direct role in HCC has not been established.

Bile Acids

Gut bacteria deconjugate primary bile acids and convert part of the bile-acid pool into secondary bile acids. These metabolites can influence barrier, metabolic, and immune signaling along the gut-liver axis.^5,48

Deoxycholic Acid (DCA)

In an obesity-associated mouse model, altered gut microbiota increased DCA. Enterohepatic DCA promoted a senescence-associated secretory phenotype in hepatic stellate cells, and reducing gut bacteria or DCA production suppressed tumor development.⁵¹ In a separate model, loss of TLR5 aggravated cholestatic injury and hepatocarcinogenesis in FXR-deficient mice.⁵² These studies support microbiota-bile acid interactions in experimental HCC but do not define a single linear pathway in patients.

Ursodeoxycholic Acid (UDCA)

In tumor-bearing mice, UDCA activated a TGR5-cAMP-PKA pathway, promoted CHIP-mediated TGF-beta degradation, and reduced Treg-related immunosuppression.⁵³ In a human HCC immunotherapy cohort, fecal UDCA and ursocholic acid were enriched in patients with an objective response and correlated with Lachnoclostridium abundance.⁵⁴ These findings do not establish clinical benefit from UDCA supplementation.

Taurocholic Acid

Hydrophobic bile acids, including taurocholate, accumulated in experimental NASH-HCC, and increasing their intestinal excretion reduced tumor development.⁵⁵ In DEN-injected mice, cholic acid supplementation accelerated a procarcinogenic transcriptional state, although histological HCC was not observed during the 10-week experiment.⁵⁶ Microbial taurocholate metabolism has also been studied in colorectal cancer,⁵⁷ but the available evidence does not establish a Klebsiella-taurocholate-NOX2 pathway in HCC.

TMAO

Gut bacteria convert dietary choline and related nutrients to trimethylamine, which is oxidized by hepatic FMO3 to TMAO.^58,59 In inflammatory HCC cell models, TMAO increased periostin through ILK/AKT/mTOR signaling and was associated with neutrophil infiltration.⁴⁵ Direct prospective evidence linking circulating TMAO to HCC progression remains limited.

Indole Derivatives

Indole-3-propionic acid is a microbial tryptophan metabolite that activates pregnane X receptor, reduces inflammatory signaling, and supports intestinal barrier integrity.⁶⁰ Its concentration and clinical importance in HCC have not been consistently established.

Endogenous Ethanol

High-alcohol-producing Klebsiella pneumoniae strains caused fatty liver in mice and were enriched in a subset of patients with NAFLD.⁶¹ Endogenous ethanol may therefore contribute to metabolic liver injury, but its direct role in HCC remains indirect.

Feedback Loop: Liver-To-Gut Signaling

HCC and cirrhosis can in turn alter the intestinal microbiota. Reduced bile-acid delivery during cholestasis may favor Clostridium overgrowth, portal hypertension can cause intestinal congestion, and systemic inflammation can change mucus and antimicrobial-peptide production.³⁷ These changes may maintain dysbiosis after liver disease has developed. Thus, the gut-liver axis includes barrier failure, PAMP-related inflammation, immune suppression, and altered microbial metabolism. Each pathway is biologically plausible, but the relative importance of these pathways in patients remains uncertain.

Table 2 and Figure 2 summarize the main mechanisms linked to individual bacterial taxa.

Table 2.

Key Gut–Liver Axis Mechanisms in HCC

Mechanism	Key Mediator	Microbial Source	Hepatic Target	Downstream Effect	Evidence Basis*
Barrier dysfunction	LPS, BFT	E. coli, B. fragilis	Epithelial junctions	Leaky gut and PAMP influx	Human cirrhosis association + DEN/colon models28,34,37–39
Inflammation	LPS, LTA	Klebsiella, Gram-positive bacteria	TLR4 (LPS); TLR2/caspase-11-related pathways (LTA)	NF-kB/IL-6/STAT3 and IL-1/IL-33 signaling	Preclinical mechanistic evidence20,34,41,43
Immune modulation	Akkermansia-related signals	A. muciniphila	CXCR6+ NKT cells; myeloid cells	Altered antitumor immunity and ICI association	Human association + strain-specific mouse data10,22,23
Metabolic shift	DCA and altered bile acids	Gut bacterial bile-acid transformation	HSC/SASP and FXR-related signaling	Inflammation and tumor promotion	Preclinical mechanistic evidence51,52,55
Direct tumor promotion	PBP1B	K. pneumoniae	TLR4 on HCC cells	PI3K/AKT, NF-kB, proliferation and EMT	HCC-derived bacterium + mechanistic mouse data20

Notes: Symbols: ↑, increased abundance or activity; ↓, decreased abundance or activity; double arrows indicate a larger or more consistent directional change across the cited evidence. *Evidence basis describes the main source of support for each statement. Human association does not establish causality, and preclinical evidence does not demonstrate clinical benefit.

Abbreviations: LPS, lipopolysaccharide; BFT, Bacteroides fragilis toxin; ZO-1, zonula occludens-1; PAMP, pathogen-associated molecular pattern; LTA, lipoteichoic acid; TLR4, Toll-like receptor 4; HSC, hepatic stellate cell; NF-κB, nuclear factor kappa B; IL-6, interleukin-6; STAT3, signal transducer and activator of transcription 3; TMAO, trimethylamine-N-oxide; DCA, deoxycholic acid; MDSC, myeloid-derived suppressor cell; Treg, regulatory T cell; CD8, cluster of differentiation 8; ICI, immune checkpoint inhibitor; mTOR, mechanistic target of rapamycin; FXR, farnesoid X receptor; FadA, Fusobacterium adhesin A.

Figure 2.

Microbial pathways that may promote hepatocellular carcinoma (HCC). Escherichia-Shigella-derived lipopolysaccharide (LPS) activates TLR4 on kupffer cells and hepatic stellate cells and increases NF-kB-related inflammation and fibrosis. Enterotoxigenic Bacteroides fragilis produces B. fragilis toxin (BFT), which can disrupt epithelial junctions and activate beta-catenin/wnt signaling. Clostridium species convert primary bile acids to secondary bile acids, including DCA; altered secondary bile acid signaling promoted inflammation and tumor development in experimental models. TMA-producing bacteria generate trimethylamine (TMA), which is oxidized to TMAO in the liver. TMAO-related immune and metabolic effects in HCC remain supported mainly by indirect or preclinical evidence.

Preclinical Evidence

Animal models, including chemically induced, dietary, genetic, germ-free, and microbiota-transfer models, allow direct testing of microbiota-related mechanisms in HCC. They reproduce selected features of human disease, including inflammation, fibrosis, dysplasia, and tumor formation, but no single model represents the full clinical course. DEN and DEN/CCl₄ models showed that bacterial PAMPs, particularly lipopolysaccharide, can promote HCC.⁶² TLR4-deficient or antibiotic-treated mice developed fewer DEN-induced tumors and showed lower IL-6/STAT3 signaling and hepatic stellate-cell activation. Restoration with Escherichia coli-derived lipopolysaccharide increased tumor formation again.³⁴ These experiments support a causal role for translocated bacterial products in these models.

Bacterial strains and transferred microbiota can affect tumor development in experimental models. Klebsiella pneumoniae accelerated HCC through the PBP1B-TLR4-PI3K/AKT pathway, and targeted suppression of intestinal colonization reduced this effect.²⁰ In a choline-deficient high-fat diet with DEN, several bacterial genera correlated with fibrosis, inflammation, or tumors, but the study did not by itself establish causality for individual taxa.⁶³ In another diet-induced NASH-HCC model, antibiotic depletion of gut microbiota reduced liver injury and tumor number.⁶⁴ In an obesity-associated model, reducing gut bacteria or DCA production suppressed the DCA-SASP pathway and liver tumor development.⁵¹ Transfer of microbiota from patients with HCC promoted liver inflammation, fibrosis, dysplasia, and tumor progression in mice, whereas microbiota from healthy donors did not produce the same effect.²⁰

Microbiota manipulation also changed antitumor immunity in experimental studies. In the STAM NASH-HCC model, administration of a breast milk-derived Akkermansia muciniphila strain increased hepatic CXCR6+ natural killer T-cell infiltration and suppressed progression from NASH to HCC. The effect was strain-specific and was not evident in NKT-cell-deficient mice.²² Human HCC cohorts have found associations between baseline or longitudinal microbiota features and the response to immune checkpoint inhibitors.^23,54,65 By contrast, responder-derived FMT has mainly been evaluated in other malignancies. A study of 13 patients with anti-PD-1-refractory advanced solid tumors reported microbiome changes and clinical benefit in a subset of patients, but it was not HCC-specific.⁶⁶ These findings support further investigation, but they do not establish benefit from FMT or bacterial supplementation in patients with HCC.

Clinical Evidence and Microbiota-Targeted Interventions

Available human evidence is mainly derived from cross-sectional case-control studies and small prospective biomarker cohorts. Several studies developed microbial models for separating early HCC from cirrhosis or healthy controls, while immunotherapy cohorts reported associations of baseline or longitudinal bacteria, fecal inflammation, and circulating metabolites with treatment outcome.^17,23,54,65 Community structure also differed among HBV-related, non-viral, and metabolic dysfunction-associated HCC.^11,30 These findings suggest that the microbiome may provide additional clinical information, but sample sizes were generally small, control groups differed, and external validation was uncommon. Prior locoregional or systemic therapy, proton pump inhibitors, antibiotics, diet, liver function, and tumor stage may affect both the microbiota and clinical outcome. Reverse causality is also possible because advanced cancer, portal hypertension, hospitalization, and treatment-related nutritional changes can themselves alter the microbiota. Representative human clinical studies, validation approaches, and major limitations are summarized in Table 3.

Table 3.

Selected Human Clinical Studies of Gut Microbiota in HCC

Study (region)	Population and etiology	Comparator	Platform	Main finding	Validation	Main limitations
Ren et al, 201917 (China)	75 early HCC; HBV-endemic cohort	40 cirrhosis + 75 healthy controls	Fecal 16S rRNA sequencing; random-forest classifier	30-OTU model distinguished early HCC from the combined cirrhosis and healthy-control groups; AUC 0.8064 (95% CI 0.7447–0.8680)	Internal fivefold cross-validation	Single-country cohort; limited etiology stratification; no external multinational validation
Behary et al, 202111 (Australia)	32 NAFLD-HCC with cirrhosis	28 NAFLD cirrhosis + 30 non-NAFLD controls	Shotgun metagenomics, fecal metabolomics, ex vivo immune assays	NAFLD-HCC had a distinct microbial/metabolite profile and induced a more immunosuppressive T-cell phenotype ex vivo	Orthogonal metabolomic and ex vivo functional analysis	Cross-sectional; modest sample; no prospective clinical validation
Liu et al, 201930 (China)	35 HBV-HCC + 22 non-HBV/non-HCV HCC	33 healthy controls	Fecal 16S rRNA sequencing; predicted functional pathways	HBV-related and non-viral HCC showed different taxonomic and ecological-network patterns	Etiology-stratified internal comparison	No cirrhosis-only control; cross-sectional; predicted rather than measured functions
Lee et al, 202254 (Taiwan, China)	uHCC on ICI: 20 objective responders + 21 progressive disease; mostly viral HCC	33 consecutive Child-Pugh A patients for validation; 17 healthy controls	Baseline fecal microbiome and bile-acid/metabolite profiling	Lachnoclostridium enrichment with Prevotella 9 depletion was associated with longer PFS and OS	Independent consecutive validation cohort	Small, mainly viral cohort; heterogeneous ICI regimens; association does not establish causality
Wu et al, 202265 (China)	35 systemic-treatment-naive patients with uHCC receiving anti-PD-1-based therapy	Responders vs non-responders; serial samples during treatment	Fecal 16S rRNA sequencing and serum metabolomics	Responders had higher baseline alpha diversity; a serum-metabolite classifier had an AUC of 0.793 (95% CI 0.632–0.954)	Prospective longitudinal sampling; internal classifier	Single center; small cohort; no independent external validation
Ponziani et al, 202223 (Italy)	11 patients with cirrhosis and HCC receiving tremelimumab and/or durvalumab	Disease control (n = 6) vs no disease control; serial samples	Microbiota profiling, fecal calprotectin, and serum barrier/immune markers	Disease control was associated with lower baseline fecal calprotectin, higher Akkermansia, and lower Enterobacteriaceae	Longitudinal within-patient analysis	Very small exploratory cohort; no external validation; mixed ICI exposure
Wu et al, 202510 (China)	53 PD-1-treated HCC: 18 responders + 35 non-responders; 88 patients across clinical subcohorts	Normal, MAFLD-HCC, and viral-HCC subgroups	Metagenomics, qPCR, plasma LPS, mouse and immune profiling	Higher Akkermansia abundance was associated with PD-1 response and longer PFS	Clinical association supported by mechanistic mouse experiments	Single center; small MAFLD-HCC subgroup; prior RFA/TACE common; no intervention with Akkermansia in patients

Notes: Studies were selected to represent diagnostic, etiology-specific, metabolic HCC, and HCC immunotherapy cohorts with reported sample size and validation information; the table is not a systematic inventory.

Abbreviations: AUC, area under the curve; HCC, hepatocellular carcinoma; HBV, hepatitis B virus; ICI, immune checkpoint inhibitor; MAFLD, metabolic dysfunction-associated fatty liver disease; OTU, operational taxonomic unit; PFS, progression-free survival; OS, overall survival; qPCR, quantitative polymerase chain reaction; RFA, radiofrequency ablation; TACE, transarterial chemoembolization; uHCC, unresectable HCC.

The evidence for microbiota-targeted interventions differs substantially by modality (Table 4). Live bacterial supplementation, prebiotics, and dietary changes have altered inflammation or tumor burden in animal models, but the direction of effect has depended on the intervention and model. Human studies in metabolic liver disease and cirrhosis provide indirect information on biological effects and safety, but no HCC-specific randomized trial has shown prevention of HCC, longer survival, or improved efficacy of immune checkpoint inhibitors. Broad-spectrum antibiotics also reduced tumors in selected mouse models. They should not be used as a microbiome treatment for HCC because of non-specific effects on the microbial community, infection and resistance risks, and possible interference with immunotherapy. In clinical practice, their role remains the treatment of documented infection.

Table 4.

Current Evidence for Microbiota-Targeted Interventions Relevant to HCC

Modality	Current HCC Evidence	Clinical Stage/Sample	Endpoints and Safety	Present Interpretation
Probiotics, prebiotics, and diet	Animal HCC studies and indirect non-HCC mechanistic evidence22,24,50	No HCC-specific randomized trial with cancer or survival endpoint	Microbial/metabolic change and liver markers reported indirectly; HCC safety and oncologic benefit not established	Not recommended as an HCC treatment outside a trial
Antibiotics	Tumor reduction in selected mouse models; no therapeutic HCC trial34,51,64	No phase-defined HCC microbiome-treatment study	Non-specific depletion, infection and resistance risks; possible confounding of ICI outcomes	Use only for standard infectious indications
FMT	Early clinical evidence in other solid tumors; HCC protocol published66,67	FAB-HCC Phase II pilot, NCT05750030; completed; actual n = 12	Primary: safety. Secondary/exploratory: ORR, DCR, PFS, OS, quality of life, engraftment, immune and multi-omics endpoints	No HCC efficacy result posted; investigational
Phage therapy	Preclinical concept for selective depletion of pathogenic taxa68–70	No HCC clinical trial identified	Requires strain specificity, resistance monitoring, manufacturing and off-target safety assessment	Preclinical proof of concept
Engineered bacteria and postbiotics	Preclinical delivery or metabolite-modulation studies, mainly outside HCC68–70	No HCC clinical trial identified	Containment, dose control, biodistribution, immunogenicity and long-term safety remain unresolved	Preclinical proof of concept

Note: Trial status was checked against ClinicalTrials.gov through August 25, 2026.

Abbreviations: DCR, disease control rate; FMT, fecal microbiota transplantation; HCC, hepatocellular carcinoma; ICI, immune checkpoint inhibitor; ORR, objective response rate; OS, overall survival; PFS, progression-free survival.

FMT is being studied clinically, but no HCC-specific efficacy result has been reported. The FAB-HCC Phase II pilot enrolled 12 patients and was completed on March 17, 2025; no results have been posted in the trial registry (NCT05750030). The protocol evaluates a single FMT from an HCC immunotherapy responder or a healthy donor, followed by continued atezolizumab plus bevacizumab in patients who did not achieve or maintain an objective response to prior treatment.⁶⁷ Safety is the primary endpoint, while response, survival, quality of life, engraftment, immune changes, and multi-omics measures are secondary or exploratory endpoints. Phage therapy, engineered bacteria, and defined postbiotics remain preclinical in HCC. These approaches require strain-level characterization, controlled manufacturing, long-term safety data, and prospective trials that account for HCC etiology. The clinical development pathway for microbiome-based approaches in HCC is summarized in Figure 3.

Figure 3.

Clinical development of microbiome approaches in HCC. Left panel: fecal and blood samples can be analyzed by multi-omics methods to identify candidate markers for early diagnosis or treatment response. Any model requires independent validation across regions and HCC etiologies. Middle panel: fmt is under early clinical investigation. Probiotics, postbiotics, phage therapy, and engineered bacteria are supported mainly by preclinical or indirect clinical evidence. Right panel: tumor response and changes in resistance to immune checkpoint inhibitors are investigational outcomes. A survival benefit from a microbiota-targeted intervention has not been established in HCC.

Abbreviations: AUC, area under the curve; FMT, fecal microbiota transplantation; HCC, hepatocellular carcinoma; ICI, immune checkpoint inhibitor; OTU, operational taxonomic unit; SCFA, short-chain fatty acid.

Challenges and Future Directions

Several issues still limit clinical interpretation of the microbiome in HCC.⁶⁸ Most human studies included fewer than 200 HCC cases and were not powered to analyze HBV-, HCV-, metabolic dysfunction-, and alcohol-related HCC separately. These groups differ in inflammation, fibrosis, diet, medication use, and treatment history. They should therefore be reported separately whenever possible. Antiviral treatment, antibiotics, proton pump inhibitors, lactulose, locoregional therapy, systemic therapy, and immune checkpoint inhibitors may also affect the microbiome or outcome. Cross-sectional samples cannot show whether dysbiosis preceded HCC or followed cirrhosis, portal hypertension, dietary change, hospitalization, or cancer treatment. This problem of reverse causality is particularly important for proposed diagnostic or prognostic markers. Sequencing platform, DNA extraction, and bioinformatics workflow add further variation among studies.^71–75 At present, no microbial signature has been reproduced sufficiently across populations to be considered universal. Future cohorts should collect standardized clinical metadata, use etiology-matched controls, obtain longitudinal samples, and include independent external validation.

Taxonomic differences do not always indicate different biological functions because unrelated bacteria may perform similar metabolic roles.^76,77 This functional redundancy complicates interpretation of microbiome-HCC associations. Mendelian randomization provides supportive evidence for causality but depends on the availability and validity of genetic instruments.¹⁴ Humanized gnotobiotic models, gene-edited bacterial strains, and controlled perturbation studies may help test individual mechanisms more directly. Microbial patterns also differ by etiology and disease stage, so one diagnostic or treatment strategy is unlikely to fit all patients.^30,31 The detection of bacteria in tumor tissue raises additional questions about the relative roles of intratumoral and gut-derived organisms. Circulating microbial DNA is also of interest, but contamination control and prospective clinical validation are required.¹⁶

Microbiota-based products also face regulatory and practical barriers. Live microbial interventions intended to prevent or treat disease are regulated as medicinal products and require defined manufacturing, quality, and safety assessment.⁶⁹ Multi-omics profiling, longitudinal sampling, and controlled FMT production also require specialized infrastructure. Targeted qPCR panels or small, defined biomarker sets may be more feasible if they retain performance across centers. Useful near-term priorities include shared sampling procedures, complete clinical metadata, multicenter longitudinal cohorts, and endpoints that measure both microbial change and patient outcome.^70,78 These steps are needed before a microbiome marker or intervention can be incorporated into HCC prevention, diagnosis, or treatment.

Conclusion

Studies in patients and experimental models consistently link the gut microbiota with HCC through the gut-liver axis. Frequently reported changes include enrichment of inflammatory or potentially pathogenic taxa, loss of barrier-protective bacteria, increased intestinal permeability, altered bile-acid metabolism, and changes in immune regulation. However, most human studies show association rather than causation. More direct causal evidence comes mainly from animal or ex vivo experiments, and mechanisms described in metabolic liver disease or other cancers cannot be assumed to occur in every patient with HCC.

Microbiome-based tests and treatments are therefore not ready for routine HCC care. Prospective studies should collect samples over time, analyze HCC etiologies separately, use standardized laboratory methods, and validate findings in independent populations. Intervention trials should report phase, sample size, safety, microbial engraftment or functional change, tumor response, and survival. A microbial or metabolite marker would be clinically useful only if it provides information beyond established variables. Microbiota-targeted treatment may eventually serve as an adjunct, but benefit and safety must first be shown directly in patients with HCC.

Funding Statement

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Disclosure

The author reports no conflicts of interest in this work.