Gram-negative bacteria and endotoxins promote hepatocellular carcinoma progression through the recruitment of myeloid-derived suppressor cells

Abstract

BACKGROUND

Bacteria were detected in several human malignancies; however, their presence and biological significance in hepatocellular carcinoma (HCC) remain incompletely characterized.

AIM

To investigate the role of bacterial components in HCC and to elucidate their effects on the tumor.

METHODS

Different experimental techniques have proved the presence of bacteria in HCC. Based on lipopolysaccharide (LPS) staining intensity, patients were stratified into LPS^low and LPS^high groups, and overall survival was analyzed using Kaplan-Meier analysis and Cox proportional hazards models. Associations among TLR4, CCL2, CCR2, and myeloid-derived suppressor cells (MDSCs) were explored using transcriptomic data from The Cancer Genome Atlas cohort. Functional experiments included LPS and lipoteichoic acid stimulation or inhibition in HCC cell lines and animal models, as well as transcriptome sequencing to identify differentially expressed genes and enriched signaling pathways.

RESULTS

LPS positive signals were detected in a significant proportion of 93 HCC pathological specimens. Compared with matched controls, fresh HCC tissues exhibited a significantly increased bacterial load. Patients in the LPS^high group demonstrated a significantly poorer prognosis, accompanied by elevated CD33 expression level and reduced CD8 expression level. Functional inhibition of TLR4 in HCC models revealed that LPS promoted the accumulation of monocytic MDSCs (MMDSCs) in the tumor microenvironment. In addition, neomycin treatment in HCC models reshaped both intestinal and intratumoral microbial composition, reduced MMDSC accumulation, and suppressed tumor growth.

CONCLUSION

Bacterial components, particularly LPS, play a significant role in HCC progression by modulating immune responses. Targeting the TLR4-CCL2-CCR2-MDSCs axis may provide a novel therapeutic approach for HCC.

Key Words: Hepatocellular carcinoma; Intra-tumor bacteria; Endotoxins; Gram-negative bacteria; Myeloid-derived suppressor cells

Core Tip: The presence of intratumoral bacteria was identified in a subset of hepatocellular carcinoma (HCC) cases. Gram-negative bacteria and their associated endotoxins were found to promote the recruitment of monocytic myeloid-derived suppressor cells through activation of the TLR4/CCL2/CCR2 signaling axis. This process may contribute to the establishment of an immunosuppressive tumor microenvironment, thereby influencing both tumor progression and clinical prognosis in HCC. Collectively, our findings identified lipopolysaccharide (LPS) as a potential prognostic biomarker for HCC and indicated that therapeutic targeting of the LPS/TLR4/CCL2/CCR2/monocytic myeloid-derived suppressor cell axis may represent a promising strategy for future HCC treatment.

INTRODUCTION

Primary liver cancer ranks as the sixth most prevalent cancer globally and constitutes the fourth leading cause of cancer-related mortality[1,2]. Hepatocellular carcinoma (HCC) accounts for more than 80% of primary liver cancer cases and is responsible for the majority of liver cancer-related fatalities worldwide[3]. The incidence of HCC, particularly those with hepatitis B virus (HBV) and hepatitis C virus infections, can be significantly mitigated through effective HBV vaccination and antiviral therapies that achieve sustained virological response in affected individuals[4,5].

Notably, infectious agents account for over 16% of global cancer incidence[6], with bacteria implicated in various tumor types[7]. Despite the recognized role of bacterial factors in cancer initiation and progression, a comprehensive characterization of these bacteria remains lacking, and the mechanisms by which they influence solid tumors were poorly understood[8,9]. Myeloid-derived suppressor cells (MDSCs) are a heterogeneous population of immature myeloid cells that are activated and mobilized under pathological conditions, such as cancer, exhibiting potent T-cell suppressive functions. Under physiological conditions, hematopoietic stem cells (HSCs) undergo a four-stage differentiation process into multipotent progenitors, lymphoid and myeloid progenitors, immune cell precursors, and finally mature immune cells, each performing distinct physiological roles. In cancer patients, tumor-derived factors, including cytokines and chemokines, remarkably disrupt normal hematopoiesis by impairing the progressive differentiation of HSCs while promoting the expansion of immature myeloid progenitors. This dysregulation results in the accumulation of aberrant immature myeloid cells, collectively referred to as MDSCs. These cells have exhibited to deplete L-arginine and L-cysteine nutrients being essential for T-cell function via inducible nitric oxide synthase and arginase, thereby suppressing the proliferation and activation of CD8⁺ T cells. Additionally, MDSCs promote tumor growth by inducing regulatory T cell (Treg) development and polarizing macrophages into an M2-like tumor-associated macrophage (TAM2) phenotype[10,11]. MDSCs, characterized in humans as CD33⁺CD11b⁺HLA-DR^-, can be further classified as polymorphonuclear (PMN)-MDSCs or monocytic MDSCs (MMDSCs) subtypes based on the expression levels of CD15⁺ or CD14⁺ markers[12]. In murine models, the corresponding phenotypes of PMN-MDSCs and MMDSCs are characterized as CD11b⁺Ly6G⁺Ly6C^- and CD11b⁺Ly6G^-Ly6C⁺, respectively[11,13,14]. The intratumoral accumulation of MDSCs is predominantly modulated by chemokine-driven recruitment mechanisms, and CCL2-CCR2 signaling pathway plays a central role in MMDSC trafficking, whereas CXCL1-CXCR2 interactions are critical for the recruitment and enrichment of PMN-MDSCs in the tumor microenvironment (TME)[15,16]. The presence of MDSCs in cancer patients and their tumor-promoting functions were well verified. Notably, MMDSCs exhibit greater immunosuppressive capabilities compared with their PMN counterparts[17,18].

In the HCC microenvironment, the mechanisms underlying the activation and recruitment of MDSCs are both similar to those in other cancer types and unique in their context. Various chemokines secreted by the TME stimulate the activation and chemotaxis of MDSCs into HCC tissues. Similar to other tumors, chemokines are the primary regulators of MDSC migration. However, MDSCs represent a highly heterogeneous population, and their specific impacts and mechanisms associated with different MDSC subtypes require more precise elucidation[19]. Bacterial colonization of the liver plays a pivotal role in the initiation and progression of HCC through multiple mechanisms. Firstly, it triggers chronic inflammation. Colonizing bacteria and their products activate pathways, such as TLR2/4, promote the infiltration of TH1/TH17 cells, and facilitate the transition from liver fibrosis to carcinogenesis. Secondly, bacterial products directly activate oncogenic signaling pathways. For instance, quinolinic acid secreted by Clostridium mitsuokai binds to the TIE2 receptor, thereby initiating the PI3K/AKT pathway, promoting tumor proliferation. Thirdly, they modulate the TME by downregulating anti-inflammatory cells and activating anti-apoptotic genes, thereby suppressing immune surveillance[20-22]. The role of TLR4 in mediating inflammatory and immunosuppressive responses further complicates the understanding of HCC pathogenesis[23]. Disruption of the intestinal mucosal barrier can enhance microbial translocation, potentially leading to microbial products entering the liver via the portal vein. However, the oncogenic signaling pathways and molecular mechanism governing MDSC recruitment in response to bacterial stimuli remain inadequately elucidated in HCC. Various experimental methodologies were employed to highlight the significance of bacteria and their products in HCC. Notably, Kaplan-Meier survival analysis and Cox multivariate analysis revealed that LPS was an independent prognostic factor for HCC. In human HCC tissues, the relative abundance of Proteobacteria at the phylum level increased in the LPS^high group compared with the LPS^low group, while the relative abundance of Lactobacillus decreased. Based on the differentially expressed genes (DEGs) and enriched signaling pathways identified through transcriptomic profiling following lipopolysaccharide (LPS) stimulation, accompanying by integrative analysis of The Cancer Genome Atlas (TCGA) dataset, the TLR4-CCL2-CCR2 signaling pathway was proposed as a putative mechanistic pathway. To further delineate how Gram-negative (G^-) bacteria and endotoxin promote the recruitment of MMDSCs and thereby facilitate HCC progression, a series of targeted interventions was administered to hepatic fibrosis-associated HCC mouse models. This multifaceted approach may deepen our understanding of the interaction among microbial composition, immune modulation, and tumorigenesis in the context of liver cancer.

MATERIALS AND METHODS

Patients’ selection and specimen processing

In this study, paraffin-embedded pathological tissue specimens from 93 patients with HCC treated at the Second Affiliated Hospital of South China University of Technology between January 2017 and March 2021 were retrospectively collected. In addition, paired tumor and adjacent non-tumorous fresh tissue samples were prospectively obtained from 30 HCC patients between March 2021 and November 2022. The inclusion criteria were as follows: (1) Patients with HCC who underwent the first operation; and (2) Patients who met the ethical review standards. The exclusion criteria were as follows: (1) Patients who received chemoradiotherapy or immunotherapy before and after surgery; (2) Patients who received antibiotic treatment before hospitalization; or (3) The absence of follow-up information. Normal liver tissues obtained from 7 patients with hepatic hemangioma were collected as control samples for comparative analysis of DNA expression levels in HCC tissues. In addition, fresh tumor specimens from 14 patients with HCC were collected for flow cytometry analysis. Given the relatively low microbial biomass in tumor tissues, stringent contamination-control measures were implemented throughout specimen collection and processing to minimize exogenous microbial interference from the patient, environment, or reagents. All personnel adhered to strict aseptic procedures, including hand hygiene and the use of sterile surgical attire and gloves prior to sampling. Surgically excised specimens were initially immersed in sterile saline, and surface blood was gently removed using sterile gauze before tissue sampling. To further reduce contamination risk, specimens were placed on more than four layers of sterile drapes during processing. Sampling instruments were prepared in advance and subjected to sequential decontamination, including treatment with sodium hypochlorite (84 disinfectant), immersion in 70% ethanol, thorough rinsing with sterile water to eliminate residual DNA, and subsequent autoclave sterilization. Separate sterile instruments were used for the collection of tumor and adjacent nontumorous tissues. The study protocol was reviewed and approved by the Ethics Committee of the Second Affiliated Hospital of South China University of Technology (Approval No. K-2022-031-02). Detailed experimental procedures are presented in the Supplementary material.

Animal experiment

After adaptive feeding, male C57BL/6 mice that aged 6-8 weeks (18-20 g) were administered 20% carbon tetrachloride (CCl₄) dissolved in olive oil for 4 weeks at a dose of 150 μL per mouse to induce hepatic injury. Age-matched male C57BL/6 mice receiving olive oil alone (150 μL per mouse) served as controls[14]. Four weeks later, 2-4 × 10⁶ double luciferase-labeled Hepa1-6 (Luc⁺/GFP⁺) cells were orthotopically injected into the liver. Tumor establishment and progression were assessed one week after cell implantation using an in vivo Xtreme imaging system (Bruker, Billerica, MA, United States), and longitudinal bioluminescence imaging was performed at weekly intervals. To more accurately replicate HCC development, DEN plus CCL₄ was additionally employed to generate an orthotopic HCC model through modulation of neomycin sensitive bacterial populations. Detailed experimental procedures are presented in the Supplementary material.

Data analysis

The data were presented as the mean ± SD. Statistical analysis was performed using SPSS 19.0 software (IBM, Armonk, NY, United States). Independent-samples t-test (n > 10) and Mann-Whitney U test (n < 10) were employed to compare two groups. For making comparisons involving more than two groups, one-way analysis of variance (ANOVA) was applied when data met assumptions of normality and homogeneity of variance; otherwise, the non-parametric Kruskal-Wallis test was used. Graphical illustrations were generated using GraphPad Prism 8.0 and Adobe Illustrator CC 2018. Bioinformatic visualization and data plotting were additionally performed using the Novogene cloud platform (https://magic.novogene.com/customer/main#/homeNew) and the OmicShare (Dior Biological) cloud platform (https://www.omicshare.com/).

RESULTS

Bacterial components in human HCC

LPS is a structural component of the outer membrane of G^- bacteria, whereas lipoteichoic acid (LTA) is a major constituent of the cell wall of Gram-positive (G⁺) bacteria. For immunohistochemical (IHC) analysis, paraffin embedded appendicitis tissues were used as positive controls for the detection of bacterial components, including LPS and LTA. For blank controls, phosphate-buffered saline (PBS) was substituted for the primary antibody. IHC analysis of appendicitis samples demonstrated the presence of both LPS-positive and LTA-positive (Figure 1A). Among 93 paraffin-embedded human HCC specimens, an LPS-positive cohort was identified using a positive area threshold exceeding 0.2%, as defined by the method presented by Nejman et al[7], accounting for 49.5% of the samples. In contrast, no LTA-positive samples were detected. To further exclude nonspecific staining, LPS immunohistochemistry was performed on aseptically cultured HepG2 cells as a negative control, and no LPS-positive signal was found. These findings indicate that the experimental environment and procedures did not introduce false positive LPS signals (Figure 1B). CD45, a pan-leukocyte surface marker, was used to assess the spatial relationship between bacterial components and immune cells. Immunofluorescence colocalization analysis of CD45 and LPS revealed that among 46 HCC patients with high LPS expression, 87% exhibited LPS localization adjacent to immune cell nuclei, while 100% showed LPS localization adjacent to non-immune cell nuclei (Supplementary Figure 1A).

Figure 1 Lipopolysaccharide was found present in both tumor and non-tumor tissues in paraffin sections of hepatocellular carcinoma. A: Hematoxylin-eosin staining and immunohistochemistry (50 μm) revealed the presence of lipopolysaccharide (LPS) in both tumor tissue regions and non-tumor tissue regions; B: LPS immunohistochemistry in different batches of HepG2 cells (100 μm). T: Tumor tissue region; N: Non-tumor tissue region; HE: Hematoxylin-eosin; PBS: Phosphate-buffered saline; LPS: Lipopolysaccharide; LTA: Lipoteichoic acid.

To further confirm the presence of bacteria in HCC tissues, fluorescence in situ hybridization (FISH) was performed using the universal bacterial probe EUB338 on tissue sections from the same 93 HCC patients. A positive signal area exceeding 0.799%, according to the criteria of Nejman et al[7], was found in 45.2% of samples. Red fluorescent bacterial probe signals were detected in paraffin-embedded HCC tumor sections, whereas no signal was identified in continuous sections hybridized with the negative control probe (non-EUB338). High-magnification imaging using an oil-immersion objective demonstrated that bacterial probe signals were predominantly localized in close proximity to the nucleus (Figure 2A). Consistent bacterial probe signals were also detected in corresponding regions of consecutive tissue sections (Supplementary Figure 1B).

Figure 2 Bacteria and their products are components of human hepatocellular carcinoma. A: Through fluorescence in situ hybridization detection, bacterial 16S rRNA was found adjacent to the cell nucleus (25 μm), and its localization was more clearly observed under oil immersion microscopy (10 μm); B: The presence of bacterial DNA in human hepatocellular carcinoma (HCC) was assessed by reverse transcription-quantitative PCR; C: Comparison of peripheral blood leukocyte count between HCC patients and healthy controls (no-template control, n = 8; environmental blank control, n = 9; sample solution, n = 30; normal, n = 7; HCC, n = 30); D: Schematic representation of the analysis pipeline applied to 16S rDNA sequencing data; E: Major genus average distribution in cancer and para-cancer groups; F: Alpha diversity assessed using the Simpson index and β diversity evaluated by principal coordinate analysis comparing cancerous and paracancerous tissue groups (n = 30); G: Sample homogenates and HCC tissues were cultured on LB agar without antibiotics. ^aP < 0.0001; ^bP < 0.01; NS: Not significant. NTC: No-template control; EBC: Environmental blank control; HCC: Hepatocellular carcinoma; PCoA: Principal coordinate analysis; OTU: Operational taxonomic unit; G⁺: Gram-positive; G^-: Gram-negative.

To identify and control for potential contamination arising from the laboratory environment, multiple negative controls were incorporated into the experimental workflow, including a no-template control for quantitative PCR and an environmental blank control for DNA extraction. Quantitative analysis demonstrated a significantly higher bacterial DNA load in HCC tissue samples (Figure 2B). In contrast, no significant difference in total white blood cell count was identified between patients with HCC and 7 patients with hepatic hemangioma (Figure 2C). To verify the presence of viable bacteria in HCC tissues, bacterial cultures were established from lysates of HCC samples using LB and BHI agar media. Following gradient dilution, single-colony isolation and bacterial identification were performed by mass spectrometry under standardized culture conditions. In addition, 30 pairs of aseptically collected fresh HCC tissues and matched paracancerous tissues were subjected to 16S rDNA sequencing targeting the V3-V4 hypervariable regions. All procedures were conducted under strict aseptic conditions in accordance with RIDE criteria[24]. Sequencing was performed using the Illumina NovaSeq platform, yielding a total of 41146 operational taxonomic units (OTUs) prior to decontamination. Contaminant sequences were subsequently removed through stringent control of DNA extraction, amplification, and sequencing processes, in combination with the R package decontam[7], resulting in 40956 high-confidence OTUs (Figure 2D). Analysis of the top 10 taxa by relative abundance revealed that G^- bacteria predominated in HCC tissues, with relative abundances of 0.105 for G^- bacteria and 0.072 for G⁺ bacteria. In paired paracancerous tissues, the corresponding values were 0.38 for G^- bacteria and 0.068 for G⁺ bacteria. These findings indicated that the majority of the top 10 most abundant bacterial taxa in HCC tissues were G^- bacteria (Figure 2E). No significant differences in α-diversity or β-diversity were found between cancerous and paracancerous tissues (Figure 2F). Furthermore, 16S rDNA sequencing of cultured bacterial isolates demonstrated that 57% (17/30) of fresh HCC tissue samples contained viable bacteria. In contrast, no viable bacteria were recovered from the corresponding control sample solutions (Figure 2G).

G^- bacteria and endotoxins affect the survival of patients with HCC through the recruitment of MMDSCs

Using a 0.2% DAB-positive area cutoff of the total area assessed by IHC[7], 93 HCC patients were classified into two categories of LPS^low and LPS^high (Table 1), exhibiting significant differences in LPS expression level between these two categories (Figure 3A). Kaplan-Meier survival analysis revealed that the median overall survival (OS) in the LPS^high group was 95 weeks (95%CI: 73.6-119 weeks), while the median OS in the LPS^low group was 149 weeks (95%CI: 91-207 weeks), and a significant difference was noted between the two groups (χ² = 10.161, P = 0.001; Figure 3B). Variables with P < 0.05 from the univariate Cox analysis were involved in the multivariate Cox analysis, which identified LPS as an independent prognostic factor for HCC (HR = 2.41, 95%CI: 1.13-5.15, P = 0.023; Table 2). FISH analysis of EUB338 revealed that the bacterial load in the LPS^high group was greater than that in the LPS^low group (Figure 3C). The RNA-seq data from TCGA cohort for HCC indicated that TLR4 was positively correlated with CCL2 expression level, with a correlation coefficient of 0.5 (n = 371). TLR4 was also positively correlated with CD33 expression level, with a correlation coefficient of 0.57 (n = 371). Additionally, HCC RNA-seq data revealed that TLR4 was positively correlated with CCR2 expression level, with a correlation coefficient was 0.52 (n = 370; Supplementary Figure 1C). Subsequently, IHC was employed and higher CD33 expression level in the LPS^high group was identified compared with the LPS^low group. In contrast, CD8 expression level was reduced in the LPS^low group compared with the LPS^high group (Supplementary Figure 1D), and expression levels of both CD33 and CD8 were predominantly localized to noncancerous tissues. LPS expression level was positively correlated with CD33 expression level, whereas CD33 expression level was negatively correlated with CD8 expression (Supplementary Figure 1D).

Figure 3 Gram-negative bacteria and endotoxins affected the survival of hepatocellular carcinoma patients through recruitment of monocytic myeloid-derived suppressor cells. A: Relative expression of lipopolysaccharide (LPS) in LPS^low and LPS^high groups; B: Kaplan-Meier survival analysis of LPS^low and LPS^high groups; C: EUB338 fluorescence in situ hybridization of LPS^low and LPS^high groups (LPS^low, n = 47; LPS^high, n = 46); D and E: The relative expression levels of CD33 and CD8 in LPS^low and LPS^high groups (LPS^low, n = 16; LPS^high, n = 14); F: Representative images of IF staining with CD33⁺CD14⁺ cells in group of LPS^low and LPS^high (LPS^low, n = 16; LPS^high, n = 14); G: The flow cytometry data of MDSCs in LPS^low and LPS^high groups (LPS^low, n = 8, LPS^high, n = 6). ^aP < 0.05, ^bP < 0.01; ^cP < 0.0001. LPS: Lipopolysaccharide.

Table 1 List of clinical characteristics for 93 validation hepatocellular carcinoma cohort.

Characteristic	LPSlow (n = 47)	LPShigh (n = 46)	P value1
Age (year)	56.39 ± 2.48	58.45 ± 2.06	0.409
Gender
Female	6	9	0.373
Male	41	37
OS status
Alive	36	24	0.014a
Death	11	22
RFS status
None-recurrence	28	17	0.029a
Recurrence	19	29
TNM stage
I-II	39	29	0.03a
III-IV	8	17
Grade
I-II	26	15	0.027a
III-IV	21	31
MVI
No	33	25	0.114
Yes	14	21
Size (cm)
≤ 5	14	12	0.691
> 5	33	34
Child-Pugh
A	41	32	0.038a
B	6	14

^aP < 0.05.

¹χ² test used to calculate the P values.

LPS: Lipopolysaccharide; OS: Overall survival; RFS: Recurrence-free survival; MVI: Microvascular invasion.

Table 2 Independent prognostic factors predicting overall survival in 93 validation hepatocellular carcinoma cohort.

Characteristic	Univariate analysis [HR (95%CI)]	P value	Multivariate analysis [HR (95%CI)]	P value
Age (year)
< 57/≥ 57	0.71 (0.36-1.4)	0.24
Gender
Female/male	1.56 (0.47-5.18)	0.464
Child-Pugh
Grade A/B	2.25 (1.09-4.67)	0.029a	0.6 (0.27-1.32)	0.205
Tumor diameter (cm)
≤ 5/> 5	2.73 (1.05-7.09)	0.039a	0.64 (0.22-1.84)	0.405
MVI
Yes/no	3.32 (1.57-7.01)	0.002a	0.32 (0.15-0.7)	0.004b
TNM stage
I-II	3.09 (1.55-6.15)	0.001a	0.59 (0.26-1.33)	0.201
III-IV
Edmondson-Steiner grading
I-II
III-IV	2.73 (1.28-5.82)	0.01a	0.63 (0.27-1.46)	0.282
LPS
Low/high	0.32 (0.15-0.67)	0.002a	2.41 (1.13-5.15)	0.023a

^aP < 0.05.

^bP < 0.01.

LPS: Lipopolysaccharide; MVI: Microvascular invasion.

Among the 30 fresh tissue samples, two subgroups of LPS^low (n = 16) and LPS^high (n = 14) were identified (Table 3). The 57% (17/30) of fresh HCC tissue samples contained viable bacteria (Supplementary Table 1). Western blot analysis indicated that the relative expression level of CD33 in the LPS^high group was significantly higher than that in the LPS^low group (Figure 3D). Conversely, the relative expression level of CD8 was significantly lower in in the LPS^high group compared with that in the LPS^low group (Figure 3E). Immunofluorescence analysis indicated that the number of CD33⁺CD14⁺ cells in the LPS^high group was significantly higher than that in the LPS^low group (Figure 3F). Among 30 HCC patients, 7 from the LPS^high group and 2 from the LPS^low group experienced recurrence, resulting in recurrence rates of 50% and 12.5%, respectively. Furthermore, the Fisher’s exact test revealed a significantly increased risk of recurrence in the LPS^high group (Table 3). To further demonstrate the difference in MDSCs (CD33⁺CD11b⁺HLA-DR^-) between the two groups, 14 fresh samples were collected from the LPS^low (n = 8) and LPS^high (n = 6) groups (Supplementary Table 2). Flow cytometry analysis revealed that the proportion of total MDSCs in the LPS^high group was significantly higher than that in the LPS^low group (Figure 3G).

Table 3 List of clinical characteristics for 30 validation hepatocellular carcinoma cohort.

Characteristic	LPSlow (n = 16)	LPShigh (n = 14)	P value1
Age (year)	53.88 ± 3.8	60 ± 2.47	-
Gender
Female	4	3	-
Male	12	11
OS status
Alive	14	10	-
Death	2	4
RFS status
None-recurrence	14	7	0.046a
Recurrence	2	7
TNM Stage
I-II	12	9	-
III-IV	4	5
Grade
I-II	10	7	-
III-IV	6	7

^aP < 0.05.

¹χ² test used to calculate the P values.

LPS: Lipopolysaccharide; OS: Overall survival; RFS: Recurrence-free survival.

In the 16S rRNA sequencing-based gut microbiota analysis, the relative abundance rate of Proteobacteria was elevated in the LPS^high group (Supplementary Figure 2A). The bacterial distribution at the genus level and LEfSe analysis revealed an increase in the relative abundance rates of the G^- bacteria Pseudomonas and Bacteroides in the LPS^high group. Analysis of the bacterial distribution and LEfSe analysis at the genus level revealed an increase in the relative abundance of Faecalibacterium and a decrease in the abundance of G⁺ Bacillus in the LPS^high group (Supplementary Figure 2B and C). The bacterial distribution at the genus level showed a decline in the relative abundance of G⁺ Lactobacillus, which is considered as a human probiotic[25] and may be associated with lower mortality rates in patients with HCC[26]. The relative abundance of G^- bacteria among the top 10 taxa was higher in the LPS^high group than that in the LPS^low group (0.162 vs 0.055). The bacterial culture results indicated differences between the two groups (Figure 2E).

Impairment of the intestinal mucosal barrier increased hepatic exposure to gut-derived microbiota and promoted HCC progression

Sirius red staining demonstrated that C57BL/6 mice developed liver fibrosis following 4 weeks of CCl₄ administration (Supplementary Figure 3A), accompanied by increased mRNA expression levels of α smooth muscle actin and collagen I in the HCC model (Supplementary Figure 3B), and primer sequences are presented in Supplementary Table 3. Intestinal microbiome sequencing has reportedly indicated that the symbiotic intestinal bacterial profile of C57BL/6 mice with hepatic fibrosis is different from that of nonfibrotic C57BL/6 mice[27]. The present study revealed that compared with that in the control group, the relative abundance of Proteobacteria increased in the CCL₄ group, whereas the relative abundance of Verrucomicrobiota decreased (Supplementary Figure 3C). The ratio of Bacteroidetes to Firmicutes was reduced (0.648 vs 1.012; Supplementary Figure 3C), aligning with the alterations in the intestinal flora at the phylum level in the fibrotic livers of mice, as evidenced by Li et al[28]. The relative abundance rates of Akkermansia and Bifidobacterium, which are probiotic genera, decreased (Supplementary Figure 3D)[29,30]. The present study indicated that β diversity significantly differed between the two groups (Supplementary Figure 3D). Following liver fibrosis, intestinal mucosal barrier dysfunction and intestinal bacterial translocation from intestine to the liver were noteworthy[31]. Subsequently, variations in the ileum were examined. In the CCL₄ group, the expression level of the tight junction protein occludin in the intestinal epithelium was downregulated (Figure 4A), while mRNA levels of inflammatory cytokines (interleukin-1β and interleukin-17) were significantly upregulated (Figure 4B). These results demonstrate that the ileum exhibited an inflammatory response and impaired intestinal mucosal barrier function.

Figure 4 An impaired intestinal mucosal barrier renders the liver susceptible to exposure to gut microbiota and promotes hepatocellular carcinoma growth. A: Occludin relative expression in ileum in control (Ctrl) and CCL₄ groups; B: Interleukin-1β and interleukin-17 mRNA expression in ileum in Ctrl and CCL₄ groups (n = 5 for each group); C: Hepatocellular carcinoma (HCC) tissues from Ctr and CCL₄ mice were homogenized and cultured on a BHI medium without antibiotics. The clone formation units were calculated; D: The flow cytometry-based proportional changes of monocytic myeloid-derived suppressor cells and polymorphonuclear-myeloid-derived suppressor cells in the tumor microenvironment (TME) between Ctrl and CCl₄ groups; E: The Ly6C relative expression in TME in Ctrl and CCL₄ groups; F: Establishment of murine HCC models and monitoring tumor growth; G: Weekly in vivo imaging was performed to monitor tumor growth in Ctrl and CCL₄ groups; H: Following the schematic diagram of animal interventions, tumor size and liver weight changes were compared between the Ctrl and CCl4 groups post-intervention. ^aP < 0.05, ^bP < 0.01; ^bP < 0.001; ^cP < 0.0001; NS: Not significant. IL: Interleukin; CFU: Clone formation unit; PMN: Polymorphonuclear; MDSC: Myeloid-derived suppressor cell; MMDSC: Monocytic myeloid-derived suppressor cell; Ctrl: Control.

Furthermore, the portal vein, which transports blood from the gastrointestinal tract to the liver, was assessed to establish a physical link between the liver and the gut microbiota. In a murine model of HCC following liver fibrosis, the elevated endotoxin level was detected in portal vein blood (Supplementary Figure 3E), accompanied by an increased bacterial DNA burden in HCC tissues (Supplementary Figure 3F). Translocation of bacteria into HCC tissues was further confirmed by FISH using a bacteria specific probe (EUB338; Supplementary Figure 3G). In addition, the presence of viable translocatable bacteria was evaluated using culture-based assays, which demonstrated that HCC tissue lysates from mice with hepatic fibrotic lesions yielded a greater number of bacterial colonies than those from control mice (Figure 4C). These results demonstrated that intestinal barrier function was impaired, thereby enabling bacteria and their products to circulate through the portal vein into HCC tissues. Additionally, immune cell subsets in the liver cancer microenvironment of mice with liver fibrosis were investigated. Flow cytometry detected higher levels of CD11b⁺Ly6G^-Y6C⁺ mononuclear bone marrow immunosuppressive cells, whereas the levels of CD11b⁺Ly6G⁺Ly6C^- polymorphic mononuclear bone marrow immunosuppressive cells were not significantly elevated (Figure 4D; Supplementary Table 4). IHC confirmed these findings, revealing an increase in Ly6C⁺ cells in the HCC microenvironment of mice with liver fibrosis (Figure 4E). CD11b⁺Ly6G^-Ly6C⁺ cell accumulation was also evident in the spleen (Supplementary Figure 3H). To assess whether this population of bone marrow-derived cells has an immunosuppressive effect, splenic CD11b⁺Gr1⁺ myeloid cells were isolated, and their ability to inhibit CD8⁺ T-cell proliferation was examined in vitro. CD11b⁺Gr1⁺ myeloid cells inhibited CD8⁺ T-cell proliferation, demonstrating that these CD11b⁺Gr-1⁺ cells are MDSCs (Supplementary Figure 3I).

The livers of mice with liver fibrosis were exposed to additional microorganisms. To assess the effects of liver fibrosis on HCC growth, live imaging was employed to weekly monitor tumor growth in mice, starting one week after in situ injection of cells into the liver. Notably, compared with that in the olive oil administration group, the CCL₄ group exhibited a high tumor fluorescence intensity (Figure 4F), along with a significant increase in liver weight and a larger tumor size (Figure 4G and H). These findings indicate the occurrence of liver fibrosis and a faster tumor growth following CCL₄ administration.

The roles of LPS/TLR4/CCL2 in symbiotic G^- bacteria and endotoxin-mediated MMDSC accumulation

According to previous results, LPS is present in both cancerous and noncancerous tissue sites in HCC. Moreover, in human HepG2 and MHCC-97H cells, TLR4 and CCL2 mRNA levels were upregulated after overnight incubation with 1 ng/mL LPS. In human LX-2 cells, TLR4 and CCL2 mRNA levels were upregulated following overnight incubation with 100 ng/mL LPS, while no changes were found in TLR4 and CCL2 mRNA levels after overnight incubation with the same dose of LTA (Figure 5A). In animal models, LPS induced CCL2 mRNA upregulation (Figure 5B) and MMDSC accumulation (Figure 5C), whereas no changes were identified in the LTA group. In vivo, the tumor fluorescence intensity was greater in the LPS group (Figure 5D). Accordingly, the increases in liver weight and tumor size were more significant. However, no such changes were found in the LTA group (Figure 5E), indicating that the same dose of LPS, rather than LTA, led to MMDSC accumulation in the TME. To evaluate the accumulation mechanism of MMDSCs recruited by LPS, the DEGs and enrichment pathways of C57BL/6 mice treated with LPS and PBS for 2 weeks were compared, as determined by transcriptome sequencing. It was found that 3919 DEGs were upregulated, while 263 DEGs were downregulated (Figure 5F). The results of Kyoto Encyclopedia of Genes and Genomes pathway analysis revealed that DEGs were enriched in the chemokine signaling pathway, and the bubble map showed 10 pathways with significant enrichment (Figure 5G). The results of Gene Set Enrichment Analyses indicated that DEGs were enriched in the Toll-like receptor signaling pathway (Figure 5H). CCR2 is the only receptor for CCL2, and transcriptome sequencing revealed significant differences in the expression levels of CCL2, CCR2, and Ly6C (Figure 5I). However, no significant differences in the gene expression levels of CXCL1, CXCR2, and Ly6G were found. To further verify the DEGs identified using transcriptomic sequencing, the CCL2, CCR2, and Ly6C genes were selected for verification, and reverse transcription-quantitative PCR (RT-qPCR) confirmed that the gene expression trend was consistent with the sequencing results (Figure 5B and I).

Figure 5 Gram-negative bacteria and endotoxins mediate monocytic myeloid-derived suppressor cells aggregation through lipopolysaccharide/TLR4/CCL2. A: The mRNA expression levels of TLR4 and CCL2 in human HepG2, MHCC-97H and LX-2 cells were detected by phosphate-buffered saline (PBS), lipopolysaccharide (LPS), and lipoteichoic acid (LTA) after stimulation, respectively (n = 3 for each group); B: After the intervention with PBS, LPS, and LTA, the mRNA expression level of CCL2 was detected by reverse transcription-quantitative PCR (RT-qPCR); C: The aggregation of monocytic myeloid-derived suppressor cells (MMDSCs) was detected by flow cytometry (n = 4 for each group); D: Weekly in vivo imaging was performed to monitor tumor growth in control (Ctrl), LPS, and LTA groups; E: The changes in tumor size and liver weight in Ctrl, LPS, and LTA groups were assessed in animals that received different interventions; F: Volcano map of differential gene expression in Ctrl and LPS groups (n = 3); G: Kyoto Encyclopedia of Genes and Genomes enrichment pathway; H: Gene Set Enrichment Analyses enrichment pathway; I: Transcriptome sequencing showed the expression levels of interest genes. The mRNA expression levels of CCR2 and Ly6C in Ctrl and LPS group were detected by RT-qPCR (n = 4 for each group); J: The expression of CCL2 mRNA was detected by RT-qPCR in Ctrl and TAK-242 groups; K: The aggregation of MMDSCs was detected by flow cytometry in Ctrl and TAK-242 groups; L: Weekly in vivo imaging was performed to monitor tumor growth in Ctrl and TAK-242 groups; M: The changes in tumor size in Ctrl and TAK-242 groups were assessed in animals that received different interventions. ^aP < 0.05, ^bP < 0.01; ^cP < 0.001; ^dP < 0.0001; NS: Not significant. LPS: Lipopolysaccharide; LTA: Lipoteichoic acid; MMDSC: Monocytic myeloid-derived suppressor cell; GSEA: Gene Set Enrichment Analyses; Ctrl: Control.

Furthermore, the effects of LPS/TLR4 signaling on CCL2 mRNA expression level in HCC were examined. The results indicated that the administration of TAK-242, a TLR4 inhibitor, inhibited the increase in CCL2 mRNA expression level in LPS-stimulated C57BL/6 mice and further reversed the accumulation of MMDSCs in LPS-stimulated C57BL/6 mice (Figure 5J and K). These results indicate that TLR4 is necessary for LPS-induced CCL2 mRNA expression level in HCC. In vivo imaging indicated that after intraperitoneal injection of the TLR4 inhibitor TAK-242, no significant change in tumor luciferin intensity was found in the LPS group compared with that in the PBS group (Figure 5L), and the effect of LPS on tumor growth was also reversed (Figure 5M).

Targeting CCL2/CCR2/MMDSCs inhibited tumor growth

CCR2 is the sole known receptor for CCL2[32], and the majority of MMDSCs express CCR2. Additionally, the CCL2/CCR2 axis plays a crucial role in recruiting MMDSCs into the TME. The CCL2/CCR2 axis affects tumor growth, invasion, and metastasis, as well as angiogenesis in the TME[33]. In some cases, targeting CCL2 has been studied in multiple preclinical cancer models, and CCL2-targetting agents have been assessed for the treatment of prostate cancer in clinical trials[34]. However, the role of the CCL2/CCR2 axis in HCC tumors and its potential as a therapeutic target remain poorly understood. Therefore, CCL2 expression level in the liver TME was assessed in the presence of LPS, and it was revealed that CCL2 overexpression in the liver TME led to the accumulation of MMDSCs. Neutralization with CCL2 antibodies reduced the accumulation of MMDSCs in the TME (Figure 6A). Compared with the PBS control, the administration of a CCR2 antagonist also reversed the accumulation of MMDSCs in the TME (Figure 6A), which significantly decreased the tumor luciferase intensity (Figure 6B), and significantly slowed tumor growth (Figure 6C). Furthermore, the role of MMDSCs in HCC was assessed, and the markers of MMDSCs in mice were CD11b⁺Ly6G^-Ly6C⁺. PBS served as the control treatment, whereas MMDSCs were exposed to a neutralizing anti-Ly6C antibody. After neutralizing MMDSCs, tumor growth was monitored weekly, and it was found that the tumor luciferase activity in the Ly6C Ab group was significantly reduced (Figure 6D) and tumor growth was significantly inhibited (Figure 6E). In the lymphocyte isolation fluid of the TME, the proportion of CD45⁺IFN-γ⁺CD8⁺ T cells in the Ly6C Ab group was significantly greater than that in the control group (Figure 6F). Tumor development, bacterial infection, and inflammatory processes are associated with the robust expansion and recruitment of MMDSCs[35]. To further clarify the role of LPS, CCL2/CCR2/MMDSCs were treated with exogenous LPS in mice (Supplementary Figure 4A), as inspired by previous results showing that endogenous LPS targets CCL2/CCR2. The accumulation of MMDSCs in the TME was reduced (P < 0.05; Supplementary Figure 4B), the proportion of MMDSCs targeted by IFN-γ⁺CD8⁺ T cells in the Ly6C Ab group was significantly greater than that in the control group of CD45⁺ cells (Supplementary Figure 4C), and targeting CCL2/CCR2/MMDSCs was reduced. The intensity of tumor luciferase significantly decreased (Supplementary Figure 4D), and tumor growth was significantly inhibited (Supplementary Figure 4E and F). To determine whether LPS relies on intact G^- bacteria to recruit MMDSCs, 0.5 g/L neomycin was administered for two weeks. After treatment, MMDSC accumulation in the TME, tumor luciferase activity, and tumor growth did not significantly differ compared with the control group. These results indicated that neomycin did not alleviate the growth-promoting effect of exogenous LPS in HCC (Supplementary Figure 4E and F).

Figure 6 Targeting CCL2/CCR2/monocytic myeloid-derived suppressor cell inhibits tumor growth. A: The aggregation of monocytic myeloid-derived suppressor cells was detected by flow cytometry in control (Ctrl), CCL2 Ab and CCR2 antag groups; B: Weekly in vivo imaging was performed to monitor tumor growth in Ctrl, CCL2 Ab and CCR2 antag groups; C: The Change of tumor size in Ctrl, CCL2 Ab and CCR2 antag groups by giving animals different interventions (n = 5 for Ctrl and CCR2 antag, 4 for CCL2 Ab); D: Weekly in vivo imaging was performed to monitor tumor growth in Ctrl and Ly6C Ab groups; E: The change of tumor size in Ctrl and Ly6C Ab groups by giving animals different interventions; F: The aggregation of IFN-γ⁺CD8⁺T was detected by flow cytometry in Ctrl and Ly6C Ab groups (n = 4 for each group); G: Molding diagram of DEN combination with CCL₄ (n = 5 for each group); H: Major phylum average distribution in Ctrl and neomycin groups; I: The ratio of Bacteroidetes/Firmicutes in Ctrl and neomycin groups; J: Major genus average distribution in Ctrl and neomycin groups; K and L: Alpha diversity was evaluated using the Simpson and Chao1 indices, while beta diversity was assessed by principal coordinate analysis, revealing distinct microbial community structures between the Ctrl and neomycin groups. ^aP < 0.05, ^bP < 0.01. PBS: Phosphate-buffered saline; PCoA: Principal coordinate analysis; MMDSC: Monocytic myeloid-derived suppressor cell; Ctrl: Control.

Neomycin-sensitive G^- bacteria could promote HCC growth

To illustrate the impact of G^- bacteria on HCC, mice from the HCC model, generated using DEN combined with CCL₄, were randomly divided into two groups. The experimental group received drinking water containing 0.5 g/L neomycin orally for 5 weeks, while the control group received sterilized water orally for the same duration (Figure 6G). The 16S rDNA V3-V4 sequencing of fecal stool collected from the ileocecal region revealed a decrease in the relative abundance of Proteobacteria in the neomycin group compared with the control group. The relative abundance rates of Verrucomicrobiota, Bacteroidetes, and Firmicutes increased (Figure 6H), whereas the relative abundance rate of Desulfobacterota decreased. The ratio of Bacteroidetes to Firmicutes increased (0.407 vs 0.216; Figure 6I). The relative abundance rates of Lactobacillus and Desulfovibrio decreased, whereas the relative abundance of Dubosiella and the abundance of Akkermansia increased (Figure 6J). In contrast to the findings found in the liver fibrosis-associated HCC model described in Supplementary Figure 3C, the relative abundance rate of Proteobacteria at the phylum level was elevated, whereas the relative abundance rates of Bacteroides, Firmicutes, and Akkermansia were reduced (Figure 6J). Additionally, both α diversity and β diversity exhibited significant differences between the control group and the neomycin group (Figure 6K and L).

Subsequently, the portal vein was examined, and lower levels of endotoxin in portal vein blood of a mouse model of HCC were found following neomycin treatment (Figure 7A). Finally, the presence of viable translocated bacteria in HCC tissues was investigated. The HCC tissue lysates from neomycin-treated mice contained a reduced number of bacterial colonies compared with those from control mice (Figure 7B). The 16S rDNA V3-V4 sequencing of HCC tissues revealed that the relative abundance rate of Proteobacteria in the neomycin group decreased compared with that in the control group, while the relative abundance rate of Firmicutes increased. The relative abundance rate of Bacteroidetes also increased (Figure 7C). According to the t-test and LEfSe analysis, the relative abundance rate of G⁺ Lactobacillus at the genus level increased, while the relative abundance rates of G^- Aeromonas and Bacteroides decreased. In addition, the relative abundance rate of G⁺ Cutibacterium decreased (Figure 7D-F). TLR4 and CCL2 mRNA expression levels were significantly reduced in the neomycin group (Figure 7G). MMDSC accumulation in the TME was significantly reversed in the neomycin group (Figure 7H). Ly6C expression level was also significantly reduced in the neomycin group (Figure 7I), and the tumor number and maximum tumor diameter in the neomycin group were significantly lower than those in the control group. These findings indicated that neomycin-sensitive bacteria could promote HCC growth by recruiting MMDSCs (Figure 7I).

Figure 7 Neomycin sensitive Gram-negative bacteria promote hepatocellular carcinoma growth. A: Portal endotoxin levels in control (Ctrl) and neomycin groups; B: Hepatocellular carcinoma tissues from Ctrl and neomycin mice were homogenized and cultured on a medium without antibiotics. The clone formation units were calculated; C: Major phylum average distribution in Ctrl and neomycin groups; D: Major genus average distribution in Ctrl and neomycin groups; E: The differential bacteria at genus level was detected in Ctrl and neomycin groups by t-test; F: The LEfSe analysis revealed genus-level differences in bacterial composition between the Ctrl and neomycin groups; G: The mRNA expression levels of TLR4, CCL2, and CCR2 were detected by reverse transcription-quantitative PCR in Ctrl and neomycin groups; H: The aggregation of monocytic myeloid-derived suppressor cells was detected by flow cytometry in Ctrl and neomycin groups; I: The relative expression of Ly6C was detected by immunohistochemical in Ctrl and neomycin groups. Tumor size, maximum tumor diameter, and liver weight were assessed in in Ctrl and neomycin groups (n = 5 for each group). ^aP < 0.05, ^bP < 0.01; ^cP < 0.001; ^dP < 0.0001. CFU: Clone formation unit; Ctrl: Control; Neo: Neomycin.

DISCUSSION

Disruption of gut microbiota homeostasis, in conjunction with the increased intestinal permeability, promotes the translocation of intestinal bacteria and their microbial products beyond the gut lumen. Furthermore, impairment of the liver’s intrinsic bacterial clearance mechanisms may facilitate bacterial persistence, thereby establishing a permissive microenvironment that provides both a niche and nutritional support for bacterial survival and proliferation in hepatic tissue[36]. The presence of bacteria in clinically and pathologically confirmed HCC paraffin-embedded tissue specimens was examined using IHC and FISH. In parallel, viable bacteria were isolated and cultured from freshly obtained aseptic HCC tissue samples. Furthermore, 16S rDNA sequencing was conducted to characterize and compare the microbial community composition of cancerous and paired paracancerous tissues from 30 fresh HCC cases. Furthermore, RT-qPCR was employed to assess bacterial density, confirming the presence of symbiotic bacteria in some HCCs. LPS was prevalent in both HCC and noncancer tissues, and it was detected in immune cells or nonimmune cells. Bacterial LPS and 16S rRNA signals were predominantly localized in close proximity to the nucleus. Notably, IHC analysis did not detect the G⁺ bacterial cell wall component LTA, which may discordant with the presence of G⁺ bacteria indicated by 16S rDNA sequencing and bacterial culture results. This discrepancy may be attributable to alterations in G⁺ bacterial cell wall composition following bacterial translocation into tissue[7,37].

Studies on the relationship between microbes and HCC primarily examined how the gut microbiota could contribute to the carcinogenesis and malignant transformation of HCC[38]. However, the mechanisms by which the microbiota associated with HCC modulates the tumor immune microenvironment remain poorly understood. HCC samples were classified as LPS^high and LPS^low based on the intensity of LPS staining[7], and the results of Kaplan-Meier survival analysis and Cox multivariate analysis indicated that LPS was as a risk factor, influencing HCC patients’ survival. TLR4, as a receptor of LPS, activates the expression levels of cytokines, such as TNF-α and interleukin-1β. It subsequently stimulates the secretion of chemokines, such as CXCL1 and CCL2, by hepatocytes, hepatic stellate cells, and interstitial macrophages. These two chemokines are responsible for the recruitment of PMN-MDSC and MMDSC to the HCC microenvironment. TLR4-induced aggregates of MDSCs following high-dose LPS exposure or bacterial infection have been reported to consist predominantly of PMN-MDSCs, as found in lung cancer and intrahepatic cholangiocarcinoma[39,40]. In HCC, positive correlations between TLR4 expression level and the chemokine axis components CCL2 and CCR2, as well as the myeloid marker CD33, were identified through analysis of TCGA dataset. Moreover, IHC analyses demonstrated a positive association between LPS and CD33 expression levels, alongside a negative correlation between CD33 and CD8 expression levels. Western blot analysis of fresh HCC tissues, combined with multiplex immunofluorescence staining and flow cytometric profiling of MDSCs from 14 freshly resected HCC specimens, further indicated that LPS could modulate the HCC immune microenvironment by promoting the recruitment of MMDSCs. CD8⁺ T cells are central effectors of antitumor immunity[41], and their functional impairment is characterized by diminished proliferative capacity and reduced production of cytotoxic effector molecules. Through suppression of CD8⁺ T cell proliferation and function, MDSCs contribute to an immunosuppressive TME, thereby adversely influencing the prognosis of patients with HCC[42].

Sequencing of the intestinal microbiome has reportedly shown that the symbiotic intestinal bacterial profile of C57BL/6 mice with hepatic fibrosis differs from that of nonfibrotic C57BL/6 mice[27]. A detailed investigation into liver fibrosis revealed that the association between liver fibrosis and bacteria, along with their products, could stem from the gut-liver axis. In this axis, pathological bacterial translocation and dysbiosis could trigger or exacerbate inflammation and cellular damage, thereby accelerating the progression of liver fibrosis, subsequently progressing to cirrhosis[36]. The pathological structure of the intestinal mucosa in patients with liver cirrhosis is characterized by the widening of intercellular spaces and vascular edema, leading to the increased intestinal permeability[43,44]. This imbalance in the intestinal microbiota, along with altered intestinal permeability, further disrupts the intestinal barrier. As a result, pathogen-associated molecular patterns, such as bacterial fragments and products, enter the liver via the portal vein system, promoting the release of proinflammatory cytokines and accelerating the progression of liver fibrosis[45]. In addition to dependence on the portal venous system, bacterial translocation may also occur via the mesenteric lymphatic pathway to hepatic lymph nodes or through direct ascending migration from the intestine[46]. Leinwand et al[47] demonstrated that bacteria detected in the murine liver originate from the intestine by orally administering fluorescently labeled Porphyromonas gingivalis and tracking in vivo translocation pathway. Although bacterial abundance in the upper gastrointestinal tract was relatively low, microbial density significantly increased from the jejunum and ileum toward the ileocecal region, rising from approximately 10⁵-10⁸ colony forming units per milliliter. In cirrhosis, alterations in intestinal permeability occur predominantly in the small intestine. The mucus layer of the distal ileum is relatively thin, thereby promoting bacterial penetration and translocation[27,48]. Experimental evidence further indicates that hepatic fibrosis and dysbiosis disrupt the ileal mucosal barrier, enabling intestinal bacteria and endotoxins to reach HCC tissue via the portal circulation. In comparison with an HCC model established by olive oil gavage combined with in situ injection, a hepatic fibrosis-associated HCC model exhibited a significant increase in MMDSCs, whereas PMN-MDSCs did not show a comparable expansion in the TME. To delineate the specific role of LPS in MMDSC recruitment, LTA was employed as a G⁺ bacterial control. At equivalent doses, only LPS induced marked overexpression of the TLR4/CCL2 axis in both in vitro and in vivo models. Transcriptomic profiling was subsequently performed to compare DEGs in C57BL/6 mice treated with LPS or PBS for two weeks. Functional enrichment analysis revealed that the DEGs were predominantly associated with the chemokine signaling pathway and the Toll-like receptor signaling pathway. As CCR2 is the sole known receptor for CCL2[32], the expression analyses of Ccl2, Ccr2, and Ly6c significantly differed between groups, whereas no significant changes were identified in the expression analyses of Cxcl1, Cxcr2, or Ly6 g. RT-qPCR validation of Ccl2, Ccr2, and Ly6c confirmed expression patterns being consistent with the transcriptomic data. Consistently, Dapito et al[49] demonstrated that intestinal G^- bacteria and endotoxins could promote HCC growth through Toll-like receptor signaling, an effect dependent on TLR4 activation in non-bone marrow-derived resident liver cells. Although neither TLR4 nor the gut microbiota is essential for HCC initiation, both are closely associated with tumor progression. Furthermore, Orci et al[50] pointed out that liver-specific TLR4 knockout or pharmacological inhibition of TLR4 effectively reversed LPS-mediated enhancement of HCC growth and recurrence[50]. These findings are consistent with the findings of the present study, in which TLR4 inhibition attenuated LPS-driven tumor progression, thereby supporting the proposed TLR4/CCL2/CCR2 axis. In contrast, investigations into TLR2, which recognizes pathogen-associated molecular patterns derived from G⁺ bacteria, have not demonstrated a contributory role in HCC development. Specifically, no evidence supporting TLR2 mediated promotion of hepatocarcinogenesis was found in the DEN combined with CCl₄ induced HCC model[49].

Inhibition of TLR4 function demonstrated that the accumulation of MMDSCs in the TME, induced by G^- bacteria and endotoxin exposure, could depend on TLR4 signaling. TLR2, acting in heterodimeric complexes with TLR1 or TLR6, recognizes lipoproteins and peptidoglycans derived from G⁺ bacteria, whereas bacterial flagellin is specifically detected by TLR5. In addition, intracellular TLR3 and TLR9 are activated by microbial nucleic acids, including double-stranded RNA and unmethylated CpG- containing DNA motifs, respectively[51,52]. As other Toll-like receptors and nucleotide-binding oligomerization domain-like receptors were not examined, the present findings support the conclusion that TLR4 is necessary for the observed immunomodulatory effects but may not be sufficient in isolation. The potential involvement of additional components of the TME, including stromal cells and paracrine signaling networks, in promoting HCC progression therefore cannot be excluded. Elucidation of the relative contributions of TLR4 expressed by hepatocytes vs hepatic stellate cells to HCC development may require further investigation using tissue-specific conditional TLR4 knockout models or experimental strategies, enabling the functional separation of these cell populations. At the cellular level, human HepG2 and MHCC-97H hepatoma cells exhibited activation of TLR4 signaling in response to a lower concentration of LPS (1 ng/mL) than that required to elicit a comparable response in human LX-2 hepatic stellate cells, indicating that LPS dose is a critical determinant of cell type-specific TLR4 activation. This dose dependency represents an important aspect for further exploration in subsequent stages of investigation. Targeting the CCL2/CCR2 signaling axis resulted in a significant reduction in MMDSC accumulation, while depletion of MMDSCs led to an increased proportion of IFN-γ⁺CD8⁺ T cells in the TME and suppression of HCC tumor growth. Moreover, inhibition of the CCL2/CCR2/MMDSC axis attenuated the tumor-promoting effects of exogenously administered LPS. In in situ HCC injection models, neomycin treatment failed to eliminate the biological effects of exogenous LPS, demonstrating that circulating LPS could drive systemic inflammation and metabolic dysregulation independently of viable bacteria through TLR4 signaling in immune and metabolic cell populations[53]. By contrast, neomycin administration in the DEN combined with CCl₄-induced HCC model significantly reduced MMDSC accumulation and suppressed both tumor burden and growth. These effects were accompanied by restoration of gut microbial balance, characterized by a reduced relative abundance rate of Proteobacteria and an increased abundance rate of Lactobacillus, thereby partially correcting dysbiosis in mice with hepatic fibrosis. Collectively, these findings indicate that modulation of gut microbiota composition and therapeutic targeting of the CCL2/CCR2/MMDSC axis represent promising strategies for HCC intervention and provide a conceptual framework for future studies employing fecal microbiota transplantation in Germfree models to identify microbiota-derived therapeutic targets for HCC.

The limited depth of 16S rDNA sequencing, together with constraints related to clinical sample availability, posed challenges for accurate taxonomic identification of microbial species and comprehensive characterization of microbial community structure, particularly regarding G^- bacteria in the present study. Alterations in intestinal mucosal permeability and increased bacterial burden have been reported to elevate endotoxin concentrations in both the portal and peripheral circulation of patients with HCC[54,55]. These findings are consistent with clinical evidence, indicating that G^- bacteria represent the predominant etiological agents of spontaneous bacterial peritonitis in patients with cirrhosis, in whom dysfunctional intestinal bacterial translocation constitutes the principal pathophysiological mechanism underlying infection. Consequently, intestinal G^- bacteria are widely recognized as the primary initiators of infectious complications in cirrhotic patients[33]. The majority of patients with HCC have underlying cirrhosis, a condition characterized by structural alterations of the intestinal barrier and remarkable immune dysfunction. Portal hypertension and disordered intestinal circulation compromise the integrity of the intestinal mucosal barrier, thereby reducing its capacity to prevent bacterial translocation. Impairment of immune surveillance further limits the clearance of translocated bacteria, promoting bacterial persistence and expansion. As a result, bacteria traversing the intestinal mucosal barrier may result either in clinically overt infectious processes or in persistent low-grade inflammation that coexists with HCC. In line with this pathophysiological framework, analysis of intratumoral microbial composition demonstrated that the 10 most abundant bacterial taxa detected in HCC tissues were predominantly G^-.

CONCLUSION

The presence of intratumoral bacteria was identified in a subset of HCC cases. G^- bacteria and their associated endotoxins were found to promote the recruitment of CCR2⁺ MMDSCs through activation of the TLR4/CCL2/CCR2 signaling axis. This process may contribute to the establishment of an immunosuppressive TME, thereby influencing both tumor progression and clinical prognosis in HCC. Distinct differences were found between LPS^high and LPS^low groups, and the LPS^high group exhibited an increased relative abundance of Proteobacteria and a reduced abundance of Lactobacillus. Consistently, in an HCC model induced by DEN in combination with CCl₄, neomycin treatment decreased the relative abundance of Proteobacteria while increasing the abundance of Lactobacillus. This intervention also attenuated the tumor-promoting effects mediated by neomycin-sensitive bacteria. Collectively, these findings identified LPS as a potential prognostic biomarker for HCC and indicated that therapeutic targeting of the LPS/TLR4/CCL2/CCR2/MMDSC axis may represent a promising direction for future HCC treatment strategies.

ACKNOWLEDGEMENTS

We thank the TCGA project. We thank chief physician Guo-Zhang Liu provided pathological technical guidance.