Exploratory Analysis of Gut Microbiome and Metabolic Profile Changes Following Lenvatinib and Anti-PD-1 Combination Therapy in Liver Cancer

TL;DR

This study explores how gut bacteria and metabolism change in mice with liver cancer before and after treatment with lenvatinib and anti-PD-1 therapy.

Contribution

The study identifies specific gut microbiota and metabolite changes associated with lenvatinib and anti-PD-1 combination therapy in liver cancer.

Findings

Combination therapy reduced certain gut bacteria like Lactobacillus and Clostridium while increasing Prevotella_7.

Metabolite levels such as pyridoxic acid and deoxycholic acid increased, while myristic acid and uric acid decreased.

The treatment partially reversed gut microbiome and metabolic changes caused by liver cancer in mice.

Abstract

Background/Objectives: Lenvatinib combined with anti-PD-1 therapy has shown promise in the treatment of hepatocellular carcinoma (HCC). The study evaluates changes in gut microbiota (GM) and metabolites during HCC treatment with lenvatinib combined with anti-PD-1. Methods: An HCC mouse model was established via diethylnitrosamine (DEN) injection, and the mice were then treated with lenvatinib, anti-PD-1, or their combination. GM composition and structural changes were assessed by 16S rDNA sequencing, and metabolite abundance by liquid chromatography–mass spectrometry (LC–MS). Results: Significant alterations in GM and metabolites were observed in the HCC group compared to the control group, and compared with the HCC group, both monotherapy and combination therapy resulted in varying degrees of GM and metabolites rebalancing. Specifically, compared to the HCC group, lenvatinib combined with anti-PD-1 therapy decreased the abundance of GM, including p_Patescibacteria, g_Lactobacillus, g_Clostridium_sensu_stricto_1, g_Eubacterium_siraeum_group, and g_Desulfovibrio, while the abundance of g_Prevotella_7 increased. Metabolite changes included increased 4-pyridoxic acid, deoxycholic acid, and taurochenodesoxycholic acid, and decreased myristic acid, oleic acid, riboflavin, and uric acid. Conclusions: HCC induces substantial alterations in the GM and metabolic profile of mice. Lenvatinib combined with anti-PD-1 treatment partially modulates these dysregulations. The relevant GM and metabolites may be associated with the efficacy of combined therapy and could serve as potential markers for further investigation.

Figures

Full text

1. Introduction

Hepatocellular carcinoma (HCC) poses a substantial and escalating global health challenge, with both incidence and mortality rates continuing to rise [1]. For HCC, multiple treatment options are currently available, including surgical resection, ablation therapy, transarterial chemoembolization, transarterial radioembolization, tyrosine kinase inhibitors, and monotherapy immunotherapy. However, current guidelines consistently recommend targeted therapy combined with immunotherapy as the preferred systemic standard treatment regimen for advanced liver cancer [2,3,4,5]. Although recent advances in treatment, including immune checkpoint inhibitors and targeted agents such as lenvatinib, have improved survival outcomes for patients with advanced HCC to a certain extent, significant limitations persist. Pronounced interpatient heterogeneity and the development of drug resistance continue to constrain the overall efficacy of these therapies [6,7]. Additionally, there are currently no definitive biomarkers to identify suitable patient populations for targeted or immune therapy, which may also influence treatment outcomes and even the development of drug resistance [6,8]. As a result, identifying mechanisms beyond direct immune modulation that contribute to therapeutic success remains a critical research priority.

The gut microbiota (GM), a core element of the human microbiome, serves as a key regulator of tumorigenesis and treatment response [9]. This relationship is particularly relevant in the liver, an organ uniquely positioned to receive direct inputs from the intestine via the portal circulation [10]. Through this gut–liver axis, microbe-associated molecular patterns and microbial metabolites are translocated to the liver, profoundly influencing hepatic immune response, metabolic homeostasis, and the tumor microenvironment (TME) [11]. The gut–liver axis is characterized by close bidirectional communication between the GM and the liver. Research has revealed that the GM and metabolites enter liver tissue via the gut–liver axis, promoting liver injury and thereby contributing to the development of HCC [12,13,14]. Notably, patients with HCC often exhibit characteristic dysbiosis, marked by enrichment of pro-inflammatory bacterial species and depletion of commensal bacteria. This imbalance is implicated in driving metabolic disorders and establishing an immunosuppressive TME conducive to tumor progression [15]. Critically, microbiota-derived metabolites, such as short-chain fatty acids (SCFAs), vitamins, and bile acid derivatives, not only modulate local intestinal immunity but also exert systemic effects, impacting antigen presentation, T-cell function, and, ultimately, the anti-tumor response [16,17]. Multiple studies indicate that the GM can influence the efficacy and side effects of immunotherapy for liver cancer by modulating host immune function [18,19]. In a human HCC study, the GM was suggested to have a crucial impact on the efficacy of anti-PD-1 immunotherapy [20]. A recent study confirmed that, in mice, tumor-suppressing multi-enterobacteria combined with PD-1/PD-L1 inhibitors exerted a synergistic anti-tumor effect on immunotherapy-resistant HCC [21]. However, research on GM and its metabolites in the field of target–immunity combination therapy remains limited. Therefore, the GM and its metabolic output serve as essential mediators linking host metabolism and immunity, potentially playing an important role in determining therapeutic outcomes.

We hypothesized that the anti-tumor efficacy of the lenvatinib combined with anti-PD-1 (LP) therapy may involve modulation of the GM and associated metabolites. To systematically investigate this, we employed a diethylnitrosamine (DEN)-induced mouse model of HCC. Although the DEN model has certain limitations in the study of immune checkpoints [22,23], it has been widely applied in HCC-related research because it can simulate angiogenesis, metabolic reprogramming, immune exhaustion, and metastatic potential during the progression of human HCC [24,25]. The primary aim of this study was to comprehensively evaluate the effects of LP on the GM structure and the metabolome. Using integrated 16S rDNA gene sequencing and liquid chromatography–mass spectrometry (LC–MS)-based untargeted metabolomics, we characterized the composition, diversity, and changes in key bacterial genera within the cecal microbiota, and profiled associated metabolic alterations. Furthermore, the correlation analysis was employed to explore potential interactions between specific bacterial taxa and differential metabolites, assessing whether this microbiota–metabolite axis might constitute a mediating pathway underlying the therapy’s immune-activating effects.

2. Materials and Methods

2.1. Chemicals and Reagents

HPLC-grade methanol and acetonitrile were obtained from Merck (Darmstadt, Germany). Deionized water was purified using a Milli-Q system (Millipore, Billerica, MA, USA). Formic acid (FA), ammonium bicarbonate, and 11 deuterated internal standards (ISs), including leucine-d3, tryptophan-d5, phenylalanine-d5, fatty acid C18:0-d3, fatty acid C16:0-d3, carnitine C2:0-d3, carnitine C8:0-d3, carnitine C16:0-d3, chenodeoxycholic acid-d4, cholic acid-d4, and lysophosphatidylcholine 19:0, were purchased from Sigma-Aldrich (St. Louis, MO, USA).

2.2. Experimental Animals

In this study, a diethylnitrosamine (DEN)-induced spontaneous HCC mouse model was established by Saiye Biotechnology Co., Ltd. (Suzhou, China). C57/B6L male mice were housed in specific pathogen-free conditions at 25–26 °C and 60–70% humidity, with a standard 12 h light/dark cycle. To induce HCC, mice received an intraperitoneal (i.p.) injection of DEN (25 mg/kg) at 2 weeks of age, followed by periodic administration. At 9 months of age, DEN-treated mice were randomly divided into four groups: a model group, a lenvatinib monotherapy (LEN) group, an anti-PD-1 monotherapy (PD-1) group, and a combination therapy group of lenvatinib and anti-PD-1 (LP).

The mice were randomly assigned to different experimental groups, and the investigators were blinded. The mice in different subgroups were treated as follows: the model group was treated with 0.5% carboxymethylcellulose sodium (CMC-Na) via oral gavage (10 mg/kg/2 days) combined with anti-IgG (200 ug, i.p.) every 4 days; the LEN group was treated with lenvatinib via oral gavage (10 mg/kg in 0.5% CMC-Na) every 2 days combined with anti-IgG (200 ug, i.p.) every 4 days; the PD-1 group was treated with 0.5% CMC-Na via oral gavage (10 mg/kg) every 2 days combined with anti-PD-1 (200 ug, i.p.) every 4 days; and the LP group was given lenvatinib via oral gavage (10 mg/kg in 0.5% CMC-Na) every 2 days combined with anti-PD-1 (200 ug, i.p.) every 4 days. After 6 cycles of anti-PD-1 or isotype control IgG antibody treatment, the mice were euthanized via cervical dislocation for sampling.

All animal experiments were carried out with approval of the Ethics Committee of Dalian Institute of Chemical Physics, Chinese Academy of Sciences (Approval Number: DICPEC2309), and performed according to the Animal Ethics Procedures and Guidelines of the People’s Republic of China.

2.3. DNA Extraction and 16S rDNA Sequencing

DNA Sequencing was performed with the assistance of LC-Bio Technologies Co., Ltd. (Hangzhou, China). DNA from various samples was extracted using the CTAB method, following the manufacturer’s protocol, with nuclease-free water used as a blank control. The extracted DNA was eluted in 50 μL of elution buffer and stored at −80 °C. The V3-V4 region of the bacterial 16S rDNA was amplified using the primers 341F (5′-CCTACGGGNGGCWGCAG-3′) and 805R (5′-GACTACHVGGGTATCTAATCC-3′) [26]. The primers were tagged at the 5′ ends with a unique barcode for each sample, along with universal sequencing primers. PCR amplification was performed in a 25 μL reaction mixture containing 25 ng of template DNA, 12.5 μL of PCR Premix, 2.5 μL of each primer, and PCR-grade water to a final volume of 25 μL. The amplification conditions included an initial denaturation at 98 °C for 30 s, followed by 32 cycles of denaturation at 98 °C for 10 s, annealing at 54 °C for 30 s, and extension at 72 °C for 45 s, with a final extension at 72 °C for 10 min. The PCR products were verified by 2% agarose gel electrophoresis. To exclude potential false-positive results, ultrapure water was used as a negative control throughout the DNA extraction process. PCR products were purified using AMPure XT beads (Beckman Coulter Genomics, Danvers, MA, USA) and quantified with a Qubit fluorometer (Invitrogen, Waltham, MA, USA). The amplicon libraries were prepared for sequencing, with both size and concentration assessed using an Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA) and the Library Quantification Kit for Illumina (Kapa Biosciences, Woburn, MA, USA), respectively. Sequencing was performed on the NovaSeq PE250 platform. Paired-end reads were assigned to their respective samples based on unique barcodes and trimmed to remove barcode and primer sequences. Merging of paired-end reads was performed using FLASH. Raw reads were subjected to quality filtering under specific conditions using fqtrim (v0.94), and chimeric sequences were removed using Vsearch software (v2.3.4). After dereplication using the Divisive Amplicon Denoising Algorithm (DADA2) (--p-trunc-len 400), feature tables and sequences were generated. Alpha and beta diversity were calculated after rarefying the sequence data to an even depth of 46,262 sequences per sample, followed by analysis with the SILVA (release 138) classifier to normalize feature abundances relative to each sample. This normalization step minimizes biases in diversity estimates arising from differences in library sizes. Alpha and beta diversity analyses were conducted in QIIME2, with visualizations generated in R. Sequence alignment was performed with Blast, and feature sequences were annotated using the SILVA database. Additional visualizations were produced using R (v3.5.2).

2.4. Metabolomics Analysis

After 48 h of lyophilization, an approximately 10 mg fecal sample was weighed into a 2 mL Eppendorf tube (Eppendorf AG, Hamburg, Germany) containing 6 mm zirconia grinding beads. To this, 300 μL of ultrapure water was added, followed by vortexing. Subsequently, 300 μL of methanol containing internal standards (ISs) was added to the mixture. Homogenization was performed using a mechanical grinding apparatus (MM400, Retsch Technology, Haan, Germany) at 30 Hz for 1 min, and the procedure was repeated 5 times. The samples were incubated at room temperature for 5 min, then centrifuged at 18,000 g and 4 °C for 15 min. The supernatant was carefully collected and filtered through a 0.22 μm nylon membrane. A 280 μL aliquot of the filtrate was evaporated to dryness and resuspended in 70 μL acetonitrile/water (20:80, v/v) for further analysis.

2.5. LC–MS Untargeted Metabolome Sequencing Analysis

Metabolic profiling of GM was performed using a vanquish UHPLC system coupled with a Q Exactive mass spectrometer (Thermo Fisher Scientific, Rockford, IL, USA). For positive electrospray ionization (ESI+) mode, separation was performed on a Waters BEH C8 column (50 mm × 2.1 mm, 1.7 μm) (Waters, Milford, MA, USA) maintained at 60 °C with a flow rate of 0.4 mL/min. Mobile phases were 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B). The gradient elution program started with 5% B for 0.5 min, increased linearly to 40% B within 1.5 min, then increased linearly to 100% B within 6 min, and was maintained for 2 min before returning to 5% B at 10.1 min, with a re-equilibration of 1.9 min. For negative electrospray ionization (ESI-) mode, an ACQUITY UPLC HSS T3 (50 mm × 2.1 mm, 1.8 μm) (Waters, Milford, MA, USA) was used, with a column temperature of 60 °C, a flow rate of 0.4 mL/min, and mobile phases of 6.5 mM NH_4_HCO_3_/water (A) and 6.5 mM NH_4_HCO_3_ 95% methanol/water (B). The gradient elution program started with 2% B for 0.5 min, increased to 40% B over 2 min, then to 100% B over 6 min, and was maintained for 2 min before returning to 2% B at 10.1 min, with a re-equilibration time of 1.9 min.

Mass spectrometry conditions were as follows: 45 arbitrary units of sheath gas, 10 arbitrary units of auxiliary gas, probe heater at 350 °C, and capillary temperature of 300 °C. The spray voltages in full-scan mode were 3500 V (ESI+) and 3000 V (ESI−), and the ion scanning range was 70–1050 Da. The spray voltage in full negative ion scanning mode was 3000 V, and the ion scanning range was 70–1050 Da. Peak extraction and integration were performed using TraceFinder software version 3.2.512.0 (Thermo Fisher Scientific Inc., Waltham, MA, USA). All identified metabolites were classified according to the confidence level recognized by the Metabolomics Standards Initiative (MSI) of the Chemical Analysis Working Group (CAWG) [27].

2.6. Statistical Analysis

All data were presented as mean ± standard deviation (SD). Comparisons between two groups were performed using the Mann–Whitney U test. The Kruskal–Wallis test was used for multiple group comparisons. Spearman’s rank correlation was used to assess the association between differential microbiota and metabolites. The Benjamini–Hochberg method controls the false discovery rate when testing multiple hypotheses. A corrected p-value below 0.05 means the difference is statistically significant. An uncorrected p-value below 0.05 indicates significance.

Statistical analyses and data visualization were performed using GraphPad Prism 9.5.1 software (GraphPad, San Diego, CA, USA) and the Lianchuan BioCloud platform (https://www.omicstudio.cn/home). Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) and variable importance projection (VIP) score analysis were conducted using MetaboAnalyst 6.0.

3. Results

3.1. Effects of LP on the Composition and Structure of GM in DEN-Induced HCC Mice

To evaluate the impact of LP on HCC, we constructed a DEN-induced HCC mouse model (Figure S1A). Experimental results demonstrated that the combination treatment exhibited superior efficacy compared to either monotherapy (Figure S1B–D). We performed 16S rDNA gene sequencing to study the effect of LP therapy on the GM of DEN-induced HCC mice.

The α-diversity indices, including Chao1, Observed_species, and Shannon, were not statistically different among these groups (Figure S1E–J). Principal coordinate analysis (PcoA) of β-diversity (Figure 1A) showed a clear separation between the control and model groups, as well as between the model group and each treatment group (LEN, PD-1, and LP groups), suggesting that the composition of the GM among these groups was distinct (R^2^ = 0.8243, p-value = 0.001). To further elucidate the GM composition across groups, we analyzed its relative abundance at the phylum and genus levels. At the phylum level, the dominant phyla in all groups included p_Firmicutes, p_Bacteroidota, p_Actinobacteriota, p_Desulfobacterota, and p_Patescibacteria (Figure 1B–D). At the genus level, the major genera consisted of g_Dubosiella, g_Desulfovibrio, g_Ligilactobacillus, g_Candidatus_Saccharimonas, and g_Olsenella (Figure 1E–G).

3.2. Altered GM of Cecal Contents Associated with LP Therapy

To identify differentially abundant microbial taxa across groups, LDA Effect Size (LEfSe) analysis was performed to find potential biomarker genera. The Linear Discriminant Analysis (LDA) score reflects the effect size of each bacterial genus, with high scores indicating greater importance in the group comparison. When using a threshold LDA score of 4, we found that among the top 30 taxa showing changes in relative abundance, 11, 6, 3, 4, and 6 were significant in the control, model, LEN, PD-1, and LP groups, respectively. The LP group exhibited a greater number of biomarkers than either the LEN or PD-1 monotherapy groups (Figure 2A,B, Table S1).

Among the top 10 differentially expressed bacterial phyla with significant differences in multiple groups of comparisons (Figure 2C, Table S2), the abundances of p_Desulfobacterota (Figure 2D), p_Patescibacteria (Figure 2E), and p_Proteobacteria (Figure 2F) were significantly upregulated in the model group compared to the control group. In contrast, these taxa were significantly downregulated in all treatment groups (LEN, PD-1, and LP), with the most pronounced decrease in p_Patescibacteria observed in the LP group.

At the genus level, among the top 15 differentially expressed bacterial genera with significant differences in multiple groups of comparisons (Figure 2L, Table S2), g_Desulfovibrio (Figure 2G), g_Candidatus_Saccharimonas (Figure 2H), g_Enterorhabdus (Figure 2I), g_Lactobacillus (Figure 2J), and g_Proteus (Figure 2K) were significantly elevated in abundance in the model group relative to the control group. In contrast, their abundance was significantly reduced across all treatment groups (LEN, PD-1, and LP), with particularly marked downregulation of g_Candidatus_Saccharimonas, g_Enterorhabdus, and g_Lactobacillus in the LP group.

Functional prediction analysis using Tax4Fun on the differential microbiota between the model and LP groups revealed that these microbial changes were mainly associated with alterations in multiple metabolic pathways (Figure 2M). This indirectly suggests that the modified GM and associated metabolites in the LP group may contribute to the anti-HCC mechanism of the combination treatment.

3.3. LP Therapy Partially Restores Altered Metabolites in DEN-Induced HCC Mice

To further investigate the metabolic effects of the LP therapy, non-targeted LC–MS metabolomics analysis was performed on intestinal content samples from all groups of mice. A total of 316 metabolites were identified in positive and negative ionization modes (Table S3). OPLS-DA was employed to evaluate treatment-related alterations in metabolic profiles and to identify significantly altered metabolites. The reliability of the OPLS-DA model was verified through permutation tests. Clear separation along the first predictive component (PC1) was observed between groups. The OPLS-DA model showed a separation rate of 19.8% between the model group and the control group (Figure 3A). Compared to the model group, the separation rates were 12% for the LEN group (Figure 3B), 14.5% for the PD-1 group (Figure 3C), and 16.4% for the LP group (Figure 3D), with the highest divergence in the model vs. control and model vs. LP comparisons. These results suggest that both monotherapy and combination treatment modulated endogenous metabolic disturbances in DEN-induced HCC mice to varying degrees.

VIP scores from the OPLS-DA model were used to identify the most representative metabolites in each group. VIP > 1.0 and p-value < 0.05 were used as thresholds to identify differentially expressed metabolites (DEMs) between groups. Compared to the control group, the model group exhibited 44 upregulated and 32 downregulated metabolites (Figure 3E). Relative to the model group, the LEN group showed 17 up- and 9 downregulated metabolites (Figure 3F); the PD-1 group had 13 up- and 21 downregulated metabolites (Figure 3G); and the LP group displayed 27 up- and 17 downregulated metabolites (Figure 3H). Interestingly, the LP group had the highest number of differential metabolites.

KEGG pathway enrichment analysis of the differential metabolites between the model group and the LP group indicated that they might be involved in metabolic pathways such as pantothenate and CoA biosynthesis; alanine, aspartate, and glutamate metabolism; and riboflavin metabolism (Figure 3I). Notably, metabolites including 10Z,12E-hexadecadienoic acid, heptadecanedienoic acid, and pelargonic acid were elevated in the model group compared to the control group. In contrast, these increases were significantly attenuated in the LP group (Figure 3J–L). Conversely, compared to the control group, ethyl alpha-glucopyranoside, 4-pyridoxic acid, 2-O-methylascorbic acid, tetracosahexaenoic acid, 3-methyl-2-oxobutyrate, 3-hydroxymethylglutaric acid, 2-oxoglutaric acid, and 1,5-anhydrosorbitol were reduced in the model group and significantly restored in the LP group (Figure 3M–T). Significantly, the levels of 10Z,12E-hexadecadienoic acid, heptadecanedienoic acid, pelargonic acid, 4-pyridoxic acid, 2-O-methylascorbic acid, and tetracosahexaenoic acid in the LP group were significantly reversed toward normal levels.

3.4. Correlation Analysis of Differential GM and Metabolites

To explore the potential associations between differential GM and differential metabolites in the model and LP groups, we first performed Spearman correlation analysis. Desulfovibrio showed a strong positive correlation with heptadecanedienoic acid (r = 0.909) and a strong negative correlation with 3-methyl-2-oxobutyrate (r = −0.846), 4-pyridoxic acid (r = −0.874), and ethyl alpha-glucopyranoside (r = −0.706). Enterorhabdus showed a strong positive correlation with heptadecanedienoic acid (r = 0.783) and a strong negative correlation with 3-methyl-2-oxobutyrate (r = −0.804), 4-pyridoxic acid (r = −0.839), 2-O-methylascorbic acid (r = −0.727), ethyl alpha-glucopyranoside (r = −0.860), and tetracosahexaenoic acid (r = −0.783). Lactobacillus showed a strong negative correlation with 2-O-methylascorbic acid (r = −0.734), 3-methyl-2-oxobutyrate (r = −0.755), 4-pyridoxic acid (r = −0.797), 1,5-anhydrosorbitol (r = −0.797), and ethyl alpha-glucopyranoside (r = −0.776). Proteus was strongly negatively correlated with ethyl alpha-glucopyranoside (r = −0.8252) (Figure 4A,B). The high value of the Spearman correlation coefficient r might be related to the limited sample size. Compared with the model group, at the phylum level of GM, the LP group exhibited decreased abundance of p_Patescibacteria (Figure 4C). At the genus level, decreased abundance was observed in g_Lactobacillus, g_Clostridium_sensu_stricto_1, g_Eubacterium_siraeum_group, and g_Desulfovibrio. At the same time, g_Prevotella_7 showed increased abundance (Figure 4D–H). Among metabolites, 4-pyridoxic acid, deoxycholic acid, and taurochenodesoxycholic acid showed increased abundance, while myristic acid, oleic acid, riboflavin, and uric acid exhibited decreased abundance (Figure 4I–O). Analysis using the gutMGene database revealed clear correlations between these six differential GM species and seven differential metabolites (Figure 4P). However, we acknowledge that these changes in relative abundance do not directly imply causality or a mechanistic link to the observed changes in metabolites. Further experimental validation is needed to establish the specific roles of these taxa in driving metabolic alterations.

4. Discussion

Systemic therapy for HCC is gradually entering an era of targeted–immunotherapy combination regimens [28,29,30,31]. In the treatment of unresectable HCC, the combination therapy of lenvatinib and anti-PD-1 has demonstrated encouraging efficacy and favorable tolerability [32,33,34]. However, clinical observations reveal substantial interpatient variability in treatment responses, even among those receiving identical therapies [35]. The GM and its metabolites regulate HCC growth, metastasis, and drug resistance through the “gut–liver axis” via various molecular mechanisms [20,36,37]. Identifying key GM species and metabolites as biomarkers to predict the efficacy of targeted immunotherapy combinations could yield novel therapeutic strategies for HCC treatment.

Using 16S rDNA sequencing of cecal contents from different treatment groups, we first assessed the modulatory effects of LP therapy on the GM. Notably, the abundances of p_Desulfobacterota, p_Patescibacteria, and p_Proteobacteria were significantly upregulated in the model group compared to the control group, and significantly downregulated following treatment with LEN, PD-1, and LP regimens, especially in the LP group. Studies have shown that following treatment of liver-related diseases with drugs that have liver-protective effects, the abundances of p_Desulfobacterota, p_Patescibacteria, and p_Proteobacteria are all downregulated [38,39,40]. Additionally, Parcubacteria (a lower taxon of Patescibacteria) are increased in rectal cancer [41]. The study by Ni et al. reported a significant increase in p_Proteobacteria in the GM of HCC patients [42], which is consistent with our findings. In summary, LP therapy may reduce the abundance of potential pathogenic bacteria [43,44,45].

In addition, we analyzed changes in metabolites within the cecal contents of HCC mice. In this study, the deoxycholic acid level in the LP group was significantly higher than that in the model group, while oleic acid and uric acid levels were significantly downregulated. Metabolite changes in the cecal contents of HCC involve multiple metabolic pathways. They can directly or indirectly affect the liver’s metabolic and immune microenvironment by regulating host signaling pathways and gene expression, thereby influencing the occurrence, development, and treatment of HCC [37,46,47]. Deoxycholic acid is known to be a secondary bile acid [48]. Research has found that the proportion of secondary bile acids within total bile acids in the serum of patients and DEN-HCC mice is significantly reduced, particularly deoxycholic acid, which may be closely associated with HCC [49]. Oleic acid, a significant product of fatty acid synthesis, has been implicated in tumor progression through multiple molecular mechanisms [50,51]. Elevated levels of oleic acid have been observed in HCC patients compared to healthy individuals, suggesting its potential as a biomarker for this malignancy [52]. Oleic acid can promote cell survival and proliferation by activating signaling pathways such as PI3K/AKT, which are crucial for cancer cell growth [53]. Additionally, oleic acid may influence HCC progression by modulating endoplasmic reticulum stress, a cellular response that can affect tumor cell viability and resistance to therapy. The interaction of adipocytes in the tumor microenvironment, which can influence lipid metabolism and availability, may also contribute to elevated oleic acid levels [54]. These findings collectively indicate that oleic acid may play a functional role in the metabolic adaptation of liver cancer, highlighting its importance in both the pathogenesis and potential therapeutic targeting of HCC. Elevated serum uric acid levels have been identified as a significant risk factor for HCC [55]. Moreover, in HCC patients, serum uric acid levels exhibit a strong negative correlation with survival duration [56]. Uric acid, as the final product of purine metabolism, exhibits antioxidant properties under physiological conditions. Elevated uric acid levels, however, may contribute to a pro-inflammatory microenvironment through mechanisms such as NLRP3 inflammasome activation, leading to increased release of inflammatory cytokines. This process can modulate immune cells’ function and potentially influence responses to immune checkpoint inhibitors [57]. Thus, alterations in uric acid concentration may not only reflect metabolic dysregulation in liver cancer cells but also serve as an indirect indicator of the tumor immune microenvironment. Therefore, the findings of this study preliminarily confirm that LP therapy may reduce harmful metabolites in the cecal contents while restoring beneficial metabolites.

By systematically resolving correlations between GM and metabolites, we constructed a mechanistic model of the GM–metabolite axis. This model reveals they may play a significant role in remodeling the intestinal–hepatic metabolic microenvironment via LP combination therapy. LP treatment significantly inhibited these pathogenic bacteria and modulated associated metabolic imbalances, suggesting a synergistic mechanism derived from the systematic regulation of colony–metabolism interactions. This study preliminarily identified potential biomarker combinations. These findings provide a theoretical foundation and practical directions for subsequent clinical translation and targeted microbial-metabolic interventions. However, this study has limitations. First, the mechanism model is based mainly on group correlation analysis. It has not yet been verified through causal experiments. Thus, this model represents a hypothesis that requires confirmation through future research. Furthermore, the differences in GM composition and metabolic profiles between responders and non-responders to immunotherapy should be further analyzed in future clinical research to clarify their correlation with treatment response.

5. Conclusions

Analysis using 16S rDNA sequencing and metabolomics demonstrated that HCC induces significant alterations in GM and metabolic profiles, while LP therapy partially reverses these changes. Correlation analysis indicated that six GM species and seven metabolites may be associated with the therapeutic efficacy of LP in HCC. However, we acknowledge that the claims regarding predictive fecal or blood tests are speculative at this stage. At the same time, our findings suggest potential biomarkers for further exploration. Additional research is needed to validate these associations and establish their predictive value in clinical settings. Additionally, the partial reversal of HCC-induced GM dysregulation by LP suggests that modulation of GM could serve as a promising adjunctive therapeutic approach. Future directions may include exploring interventions such as probiotics, prebiotics, or fecal microbiota transplantation to enhance the efficacy of combination therapy and address drug resistance. In summary, specific microorganisms and metabolites have been proposed that may directly or indirectly modulate the host immune response to tumors. Validation of these approaches could yield novel drug targets.