Introduction

The liver has long been considered sterile. But advances in sequencing technology have revealed the microbes in the liver. Furthermore, a number of recent studies have reported the presence of symbiotic bacteria and fungi in various human tumor tissues1,2,3, including liver tumors4,5,6. For example, one study reported a significant increase in alpha diversity of microbiota in liver tumors compared to normal liver tissue, and the intratumoral microbiome composition correlated with patient clinicopathological parameters6. Besides, one study characterized the microbes within liver tumors and found that the pattern of intratumoral microbes, known as hepatotype, significantly affected patient survival after surgery5. Another study confirmed that the liver tumor microbiome is closely related to changes in host metabolism, epigenetic and gene expression profiles, indicating strong interaction between the tumor microbiome and the host7. Jiang et al. suggested that the identification of Intestinimonas, Brachybacterium, and Rothiawas established as independent risk factors influencing the overall survival (OS) of HCC patients who underwent surgical resection8. Altogether, these studies have clarified the association between the microbial composition in the tumor microenvironment (TME) and the progression of liver cancer from different perspectives and using different techniques. However, only focusing on the bacterial composition in the tumor is a common shortcoming of these studies. The potential role of other microbial kingdoms, such as mycobiome, in the liver TME remains unknown.

Hepatocellular carcinoma (HCC), constituting an overwhelming 80–90% of all primary liver cancers, stands as the most prevalent form of liver malignancy. Despite remarkable strides made in therapeutic interventions, HCC continues to exhibit high mortality rates, both within the United States and on a global scale9,10. HCC ranks as the sixth most prevalent cancer worldwide and the fourth leading cause of cancer-related deaths, posing a substantial burden on public health globally11. A phase 1b study investigating the combination of atezolizumab and bevacizumab in patients with untreated unresectable HCC revealed a tolerable profile of side effects and encouraging antitumor efficacy, featuring an objective response rate of 36% and a median progression-free survival duration of 7 months12. In recent years, with the clinical progress of immunotherapy represented by immune checkpoint inhibitors (ICIs) in HCC, the effect of immunotherapy has been gradually recognized13. ICIs has become the most commonly used and most important treatment for unresectable HCC. However, due to the heterogeneity of the TME, a considerable number of patients are unable to benefit from immunotherapy14, and there is an urgent need for clear biomarkers to predict the efficacy of HCC immunotherapy.

The relationship between bile acid synthesis and HCC has received increasing attention. Bile acid is an important product of liver metabolism, and the abnormal synthesis of bile acid has been proved to be significantly related to the occurrence and development of HCC. Several studies have investigated the relationship between microbiome, bile acids, and the development of liver tumors15,16,17. For instance, one study showed that gut microbiome-mediated bile acid metabolism regulates liver cancer through natural killer cells15. Notably, due to the spatial structure of microorganisms located in the TME, it can more directly influence various activities in the liver, including bile acid synthesis. Therefore, elucidating the relationship between the microbiome in the tumor, bile acid synthesis, and HCC is helpful to offer new treatment options.

In this study, using the paired intratumoral microbiome and host transcriptome samples from The Cancer Genome Atlas (TCGA) HCC cohort, we characterized the intratumoral bacteriome and mycobiome composition and their interactions in HCC. Besides, we identified high abundance of genus Malassezia in tumors, and correlated the abundance of Malassezia with the TME signatures and bile acid synthesis. Furthermore, we developed a set of genes correlated with Malassezia to predict the patient survival. These results provide novel insights into the potential role of intratumoral mycobiome in HCC and offer guidance to improve the current landscape of HCC treatment.

Methods

Data collection

The RNA-sequencing data and clinical information of patients with HCC were downloaded from the TCGA database (https://portal.gdc.cancer.gov/). Summary of patient characteristics was provided in supplementary Table 1. The intratumoral microbiome abundance data was downloaded at https://github.com/knightlab-analyses/mycobiome2. In addition, four independent clinical cohorts, including GSE76427, GSE16757, GSE20017, and GSE56140, were downloaded from the gene expression omnibus (GEO) dataset (https://www.ncbi.nlm.nih.gov/geo/) and were included in this study.

Immune infiltration analysis

TIMER, CIBERSORT, quanTIseq, MCPcounter, xCell, and EPIC were used to determine the abundance of immune cell populations. ESTIMATE algorithm was used to calculate the stromal score, immune score, and tumor purity18. To evaluate the potential response to immunotherapy, the Tumor Immune Dysfunction and Exclusion (TIDE) score was calculated online (http://tide.dfci.harvard.edu/)19.

Microbial analysis

Alpha diversity measured by the richness, Shannon, and Simpson indices was calculated with “vegan” package. Beta diversity was measured by Bray-Curtis dissimilarity index with “vegan” package. Spearman correlation analysis was performed to explore the association between each dominant bacteria and each dominant fungus. The Procrustes analysis was conducted to explore the correlation between bacteriome and mycobiome in HCC tumors20.

Construction of prognostic signatures

First, TCGA-HCC samples were divided into two groups based on the median of relative abundance of genus Malassezia, and then differentially expressed genes (DEGs) were identified between these two groups with the “limma” package (fold change > 2 and adjusted P-value < 0.05). Second, Malassezia-related genes were identified by Spearman correlation analysis with adjusted P-value < 0.05. Subsequently, an intersection was conducted on the two genesets, the final genes obtained were defined as Malassezia.Sig. Gene set variation analysis (GSVA) was used to calculate the pathway score of Malassezia.Sig with “GSVA” package. Functional enrichment analyses of Malassezia.Sig were based on Reactome database21 and Kyoto Encyclopedia of Genes and Genomes (KEGG).

Development of prognostic model

Univariate Cox regression and multivariate Cox regression analyses were conducted on the basis of the expression of Malassezia.Sig and overall survival (OS). Kaplan-Meier (K-M) curve was utilized to evaluate the differences of OS between two groups. The immunohistochemistry (IHC) staining images for the key genes were sourced from the human protein atlas database (HPA; https://www.proteinatlas.org/), a highly valuable resource encompassing IHC-based expression data for the 20 most prevalent types of cancer22. Dataset GSE76427 was used to independently validate the performance our prognostic model.

Statistical analysis

All statistical analyses in this study were carried out using R software (version 4.2.1; http://www.R-project.org). The statistical analysis of continuous variables was conducted using a two-tailed Wilcoxon rank-sum test. Nonmetric multidimensional scaling (NMDS) based on Bray-Curtis distance matrices was implemented in the R project, and the visualization of the results was achieved using the “ggplot2” package in R. Additionally, a permutational multivariate analysis of variance (PERMANOVA) test was executed utilizing the “vegan” package. Unless indicated otherwise, a P-value of below 0.05 was considered statistically significant.

Results

The intratumoral microbiota profile in HCC tumor and normal adjacent to tumor (NAT)

Intratumoral microbiota was correlated with HCC progression. First, we detected a significant difference in the intratumoral microbiota composition between tumor and NAT (Fig. 1a; PERMANOVA test, P = 0.001). The richness of intratumoral microbial community in tumors was significantly higher than that in NAT (Fig. 1b; P < 0.0001), while no significant differences were detected in the Shannon and Simpson indices (Fig. 1c and d, P > 0.05). Next, we depicted the microbial composition at the genus level, and found that tumor and NAT harbored similar dominant microbes (Fig. 1e and f). For example, in both tumor and NAT, Malassezia was the genus with the highest relative abundance. Besides, Bifidobacterium, Acinetobacter, Alcanivorax, and Arcobacter were enriched in tumor and NAT. Given the high abundance of Malassezia, an opportunistic fungus, we further investigated the species belonging to Malassezia. Three species was detected in tumors and Malassezia globosa accounted for a significant proportion of the readings (Fig. 1g).

Fig. 1
figure 1

Microbial community composition in tumor and NAT of HCC. (a) NMDS plot showing the difference in the intratumoral microbiome composition between tumor and NAT. The p value was generated by PERMANOVA test. Boxplot showing the difference in the (b) richness, (c) Shannon, and (d) Simpson index between tumor and NAT. Wilcoxon test, **** indicate p < 0.0001. ns, not significant. Microbial community composition at the genus level in (e) liver tumor and (f) NAT. (g) Species belonging to the genus Malassezia. The top of the figure represents the three species, the bottom represents the genus Malassezia, and the width of the lines connecting the top and bottom represents the number of sequences aligned to the species. (h) Procrustes analysis and Mantel analysis showing the correlation between bacteriome and mycobiome in liver tumors. (i) Heatmap showing the Spearman correlation coefficient between individual dominant bacteria and individual fungus. The column annotation indicates the relative abundance of bacteria. * indicate p < 0.05, ** indicate p < 0.01, and *** indicate p < 0.001.

Full size image

Multiple studies suggest bacteriome-mycobiome interactions in tumorigenesis2,23. First, from the perspective of community, Procrustes analysis showed a significant correlation between bacteriome and mycobiome in tumors (Fig. 1h; Correlation = 0.70, P = 0.001). Mantel test also verified this result (r = 0.67, P = 0.001). In addition, from the perspective of individual microbe, we also found significantly positive correlations between almost all dominant bacteria and fungi in HCC tumor (Fig. 1i). Consequently, these results demonstrate that HCC tumors create a microenvironment conducive to a bacteria-fungus synergistic relationship.

Malassezia abundance is related with the TME signatures

The TME impacts the initiation and progression of a variety of tumors24. Next, we investigated the potential role of intratumoral microbiota in the TME of HCC. First, we compared the TME signatures between tumor and NAT. The stromal score and immune score in tumor were significantly lower than those in NAT (Fig. 2a and b; P < 0.0001). Conversely, tumors harbored a significantly higher tumor purity than NAT (Fig. 2c; P < 0.0001). These results suggest a close correlation between the TME and HCC tumorigenesis.

Fig. 2
figure 2

Malassezia plays a regulatory role in the TME of HCC patients. Boxplot showing the differences in (a) stromal score, (b) immune score, and (c) tumor purity between tumor and NAT. (d) Boxplot showing the differences in the Malassezia load between tumor and NAT. Wilcoxon test, *** indicate p < 0.001 and **** indicate p < 0.0001. Linear regression analysis showing the correlation between the relative abundance of Malassezia and the TME profiles, including (e) stromal score, (f) immune score, and (g) tumor purity. (h) Boxplot showing the difference in the expression of gene SELE between tumor and NAT. (i-l) Linear regression analysis showing the correlation between the expression of gene SELE and four HCC-related key genes, including ALDH2, AXIN1, PRMT3, and BRAF. (m) Linear regression analysis showing the correlation between the expression of gene SELE and the relative abundance of Malassezia.

Full size image

Next, we found that the abundance of Malassezia in tumor was significantly higher than that in NAT (Fig. 2d; P < 0.001). Given the above results, we hypothesized that Malassezia might be closely related to the TME signatures. We correlated the relative abundance of Malassezia with stromal score, immune score, and tumor purity. Intriguingly, the relative abundance of Malassezia was significantly negatively correlated with stromal score and immune score, as well as positively correlated with tumor purity, although not reaching statistical significance (Fig. 2e and g).

Intratumoral microbes can induce a pro-tumor inflammatory environment to promote tumorigenesis25. We found that the expression of the SELE gene, a known cell adhesion molecule with a pivotal role in inflammation, was significantly higher in NAT than in liver tumors (Fig. 2h; P < 2.2e-16). Furthermore, the expression of SELE gene was significantly correlated with some key genes in HCC. For instance, ALDH2 expression is closely related to the risk, pathogenesis and prognosis of liver cancer, and can be used as a potential therapeutic target, which is significantly positively correlated with SELE gene (Fig. 2i; P = 0.001). SELE gene was significantly negatively correlated with AXIN1 (Fig. 2j; P = 0.002) and BRAF gene (Fig. 2l; P = 0.003), two common mutated genes in HCC. In addition, recent studies have identified PRMT3as a potential biomarker and therapeutic target to overcome immunotherapy resistance to HCC26, and we found that its expression was also significantly negatively associated with the SELE gene (Fig. 2k; P = 0.002). What’s more, we observed that the relative abundance of Malassezia was significantly negatively correlated with the SELE expression (Fig. 2m; P = 0.031), suggesting a potential relationship between intratumoral Malassezia and local inflammation.

Intratumoral Malassezia plays a role in bile acids synthesis

Next, we sought to explore the underlying biological characteristics of HCC tumors with distinct abundance of Malassezia. Here, we divided all samples into two groups based on the median of relative abundance of Malassezia, and identified the DEGs between the two groups. Functional enrichment results showed that the DEGs were significantly enriched in bile secretion, cAMP signaling pathway, and transmembrane transport (Fig. 3a and b). Consequently, we hypothesized that intratumoral Malasseziamay be involved in bile acid synthesis in the liver. Under normal conditions, the classical pathway accounts for at least 75% of bile acid production, triggered by CYP7 A1-catalyzed cholesterol 7α-hydroxylation27. CYP7 A1 is a rate-limiting enzyme and determines the amount of bile acid produced. The alternative pathway is initiated by CYP27 A1 (Fig. 3c). Intriguingly, we found that the relative abundance of Malassezia was significantly negatively correlated with the expression of gene CYP7 A1 (Fig. 3d; P = 0.031) and CYP27 A1 (Fig. 3e; P = 0.049). In addition, the relative abundance of Malassezia was significantly positively correlated with the MSTN gene (Fig. 3f; P = 0.027). One study showed that MSTNgene mutation up-regulates hepatic bile acid synthesis and regulates bile acid metabolism27. Collectively, our results demonstrate that intratumoral Malassezia may be involved in bile acid synthesis while the underlying mechanism remains to be explored.

Fig. 3
figure 3

Malassezia is related to bile acid synthesis. (a) KEGG48 and (b) GO enrichment results of the DEGs between samples with high abundance of Malassezia and low abundance. Linear regression analysis showing the correlation between the relative abundance of Malassezia and bile acid synthesis-related genes, including (c) MSTN, (d) CYP7 A1, and (e) CYP27 A1. (f) Two main pathways of bile acid synthesis. The red ovals represent the key enzymes that work.

Full size image

Construction of Malassezia.Sig and its relationship with the TME

Given the strong association between the intratumoral Malassezia and the TME in HCC, to improve clinical utility, we sought to identify a Malassezia-related geneset to characterize tumor-resident Malassezia. First, we obtained 897 DEGs between samples with high and low relative abundance of Malassezia (Fig. 4a). Second, Spearman correlation analysis was performed to identify genes significantly correlated with the relative abundance of Malassezia, and 1,546 genes were obtained. Finally, we obtained 281 genes by intersection of these two genesets, called Malassezia.Sig.

Fig. 4
figure 4

Malassezia-related genes are strongly correlated with the TME profiles. (a) Pipeline of the construction of Malassezia.sig. Venn plot showing the intersection of DEGs and genes significantly correlated with Malassezia. (b) Reactome, (c) KEGG, and (d) GO enrichment results of Malassezia.sig. Linear regression showing the correlation between the GSVA score of Malassezia.sig and (e) stromal score, (f) immune score, (g) tumor purity, and (h) relative abundance of Malassezia. (i) Heatmap showing the normalized abundance of various immune cell populations in groups of high- and low GSVA score. Row annotation represents six approaches used to calculated the abundance of cells. Column annotation represents the two groups, high GSVA score and low GSVA score. (j) Boxplot showing the difference in the abundance of M1 macrophage, M2 macrophage, and CD4+ T cell between high GSVA score and low GSVA score.

Full size image

Next, we explored the biological significance of the Malassezia.Sig. Enrichment analyses showed that the Malassezia-related genes were significantly enriched in cilium assembly, cAMP signaling pathway, and histone demethylation (Fig. 4b and d). Then, we calculated the GSVA score of Malassezia.Sig and correlated it with the TME signatures. As we expected, the stromal score was significantly negatively correlated with the GSVA score of Malassezia.Sig (Fig. 4e; P = 0.035). Although not statistically significant, we also observed a negative relationship between the immune score and the GSVA score of Malassezia.Sig (Fig. 4f; P = 0.109). Conversely, tumor purity was significantly positively correlated with the GSVA score of Malassezia.Sig (Fig. 4g; P = 0.047). Meanwhile, the significant correlation between the relative abundance of Malassezia and the GSVA score of Malassezia.Sig (Fig. 4h; P = 7.5e-06) indicated that we successfully built a geneset to characterize the intratumoral Malassezia.

Considering the essential role of immune cell populations in the TME, we next divided all samples into two groups based on the median of the GSVA score of Malassezia.Sig. Six approaches were conducted to estimate the abundance of a variety of immune cells based on the gene expression. Consistent with above results, we observed that compared with the high group, the low group harbored much higher abundance of most of immune cells (Fig. 4i). Specifically, the abundance of M1/M2 macrophages was significantly higher in the low group compared with the high group, while the CD4+ T cell was opposite (Fig. 4j). Collectively, the Malasseiza.Sig is closely related with the TME in HCC.

Survival prediction with the Malassezia.Sig

Considering the pivotal role of the TME in prognosis of patients with HCC28,29, we next sought to correlate the Malassezia.Sig with patients’ overall survival (OS). First, univariate Cox regression analysis was performed to explore the association between the individual gene and OS. Results showed that among the 281 genes, 95 genes were significantly correlated with OS (P < 0.05), accounting for 33.81% (Fig. 5a). Subsequently, based on these prognostic genes, multivariate Cox regression analysis was conducted. Finally, we obtained a prognostic model consisting of five genes, including ACYP1, GPD2, ENTPD2, SUPT7L, and COG1 (Fig. 5b). Survival analysis suggested that all these five genes were significantly correlated with patient OS, and high expression of them indicated adverse survival (Fig. 5c and g). In addition, our prognostic model achieved excellent performance in predicting the OS in both TCGA (Fig. 5h; P < 0.0001) and an independent HCC cohort (Fig. 5i; P = 0.021). Next, immunohistochemical staining was used to verify the expression of key genes in the prognostic model, and the results showed that the expression of these five prognostic genes was higher in liver tumors than those in NAT (Fig. 5j), which was also consistent with previous conclusions.

Fig. 5
figure 5

Malassezia.sig is closely related with prognosis in HCC patients. (a) Pie chart showing the proportion of prognostic genes in Malassezia.sig. The prognostic genes were obtained by univariate Cox regression with a threshold of p < 0.05. (b) Forest plot showing the result of multivariate Cox regression. (c-g) Survival analysis of five genes in Fig. 5b. Samples were divided into two groups (high and low expression) based on the median expression of each gene, The p values were generated by log-rank test. Performance of our prognostic model in (h) TCGA and (i) GSE76427. (j) Representative immunohistochemical staining images of five genes in tumor and NAT.

Full size image

Next, we correlated these five genes with the TME signatures. In TCGA, we observed significant correlations between COG1 gene and all TME signatures, including stromal and immune score, and tumor purity (Fig. 6a, e and g). In GSE16757, SUPT7L gene was significantly correlated with the TME signatures (Fig. 6b). Besides, GPD2 gene was strongly correlated with all the TME signatures in GSE20017 (Fig. 6c and l), while COG1 gene was closely related with tumor purity and stromal score in GSE56140 (Fig. 6d and k). Given the substantial correlation between COG1 gene and the TME in multiple datasets (Fig. 6e, g, i and k), we then calculated the TIDE score of TCGA samples and observed a significantly positive correlation between COG1 expression and the TIDE score (Fig. 6h; P = 3.54e-05), indicating a potential relationship between high expression of COG1 and poor response to immunotherapy.

Fig. 6
figure 6

Our prognostic genes are significantly correlated with the TME signatures. Heatmap showing the Spearman correlation coefficient between the five prognostic genes and the TME profiles in (a) TCGA, (b) GSE16757, (c) GSE20017, and (d) GSE56140. (e-l) Linear regression plot showing the correlation between specific prognostic genes and the TME profiles.

Full size image

Discussion

Accumulating evidence provided insights into the association between a wide range of cancers and intratumoral mycobiome2,30,31, however, the role of intratumoral fungal composition in the development of HCC remains unclear. Herein, we found that Malassezia, an opportunistic pathogenic fungus, was highly enriched in HCC tumors. Moreover, the relative abundance of Malasseziawas strongly correlated with a depleted stroma and immune components in the TME. In addition to its normal physiological function, bile acids are also closely related to the occurrence and development of HCC32,33. We found that intratumoral Malassezia could inhibit bile acid synthesis by down-regulating the expression of CYP7 A1 and CYP27 A1. Finally, we developed a Malassezia-related gene signatures and prognostic model was built based on these signatures to predict the survival. Although more in-depth mechanistic studies are needed to clarify the intrinsic link between Malassezia and the development of HCC, our study provides new insights into the association of intratumoral mycobiome with HCC.

The relationship between Malassezia and cancers has been previously reported, particularly in pancreatic cancer. In pancreatic cancer patients, the number of fungi in the pancreas can increase to 3,000 times the normal state, with Malasseziaaccounting for a significant proportion, about 20% of the total number of fungi34. The specific mechanism by which Malasseziapromotes pancreatic cancer has not been fully elucidated, but the researchers suggest a possible pathway, namely the activation of mannose-binding lectins (MBL) and complement C334. In addition, in breast cancer, a recent study reported that sphingosine kinase 1 (Sphk1) is upregulated to promote breast tumor growth after Malasseziacolonizes tumors35. Our study reveals for the first time the potential role of intratumoral Malassezia in HCC. Unlike the previous two studies, we did not find a direct relationship between Malassezia and HCC tumor growth, possibly due to tumor heterogeneity. Interestingly, we observed a prominent correlation between the relative abundance of Malasseziaand the TME characteristics, including stromal and immune cell populations. A large number of studies have shown that the TME plays a key role in HCC initiation, development and anti-tumor therapy response, especially the efficacy of immunotherapy is closely related to the TME28,36,37,38. Consequently, intratumoral Malassezia may indirectly affect the development and prognosis of HCC by regulating the TME.

We also found that intratumoral Malasseziamay affect the bile acid synthesis pathway. The liver is the main site of bile acid synthesis. Multiple studies have shown significant abnormalities in bile acid levels in liver tissue and plasma in HCC patients39,40. The levels of a variety of bile acids in HCC patients were significantly higher than those in healthy people40. Besides, excess bile acids can induce sustained damage, apoptosis and inflammation of liver cells, thereby increasing the risk of HCC41,42. We found that the relative abundance of Malassezia was negatively correlated with the expression of both CYP7 A1 and CYP27 A1, indicating the inhibitory effect of intratumoral Malassezia on bile acid synthesis. Future studies should focus on the functional mechanisms by which Malassezia inhibits bile acid synthesis, thereby contributing to the development of tests and treatments related to bile acid synthesis factors. For example, the development of more sensitive and effective markers for early diagnosis and prognostic prediction of HCC, microbiome preparations for remodeling intratumoral microbiota and specific bile acid receptor agonists have important clinical application value and need to be further studied.

Due to the low content of microorganisms in the tumor, the high proportion of host DNA, and the extremely difficult to control contamination, the accurate detection of intratumoral microbes becomes a major obstacle. Therefore, in order to improve clinical practicability, we established a geneset to characterize intratumoral Malassezia. Intriguingly, the biological function of this geneset was enriched in the TME-related pathways. Based on five Malassezia-derived genes, including ACYP1, GPD2, ENTPD2, SUPT7L, and COG1, the multivariate Cox proportional hazards regression model successfully predicted the OS in HCC patients. Consistent with previous studies, Cui et al. developed a model based on five metabolic genes, including POLA1, UCK2, ACYP1, ENTPD2, and TXNRD1, and successfully predicted the prognosis of HCC patients43. Besides, Wei et al. established a prognostic model based on 19 lipid metabolism-related genes, including GPD2, to stratify the prognosis of gastric cancer patients44.

Specifically, studies have shown that ACYP1 can promote the proliferation, invasion and migration of HCC cells both in vitro and in vivo45. Zheng et al. reported that GPD2 can regulate hepatic glucose homeostasis and control nonalcoholic fatty liver, while the elimination of GPD2 will increase the accumulation of triglycerides induced by liver diet and promote lipogenesis to induce fatty liver46. In the human body, liver cancer cells produce a protease called ENTPD2, which indirectly promotes the growth of liver tumors and weakens the effectiveness of immunotherapy47. Inhibition of ENTPD2 is a promising strategy to enhance the efficacy of ICIs targeting PD-1/CTLA-4. We found that SUPT7L and COG1 genes were risk factors for survival in HCC patients, but their underlying roles in HCC remains unclear and needs to be further explored.

The limitations of our study encompass a comparatively modest sample size and a cross-sectional study design, which, while enabling the identification of associations, falls short of establishing causal relationships. Given the constraints of the sample size, we were unable to comprehensively adjust the data for all potential confounding variables. Future endeavors necessitate the conduction of studies with expanded sample sizes and employing longitudinal designs to further elaborate upon our present findings. Additionally, the biological functions of tumor-resident mycobiome in HCC, especially the relationship between Malassezia and the TME of HCC, require further exploration. Conducting further experimental studies is imperative to elucidate these mechanisms, which could significantly contribute to our understanding of HCC pathogenesis and facilitate the identification of potential therapeutic targets. Finally, the etiology of HCC has far-reaching implications for gene expression based analyses, such as metabolic dysfunction-associated steatotic liver disease (MASLD)-associated liver cancer and hepatitis B. Due to the lack of relevant etiological data, we were not able to stratify patients and perform comparative analyses to provide insight into the heterogeneity of HCC.

In conclusion, we showed that Malassezia is enriched in HCC tumor tissues and, notably, that the levels of Malassezia are correlated with the TME and bile acid synthesis. In the future, elucidations of the underlying mechanisms by which intratumoral Malassezia regulates HCC behavior will contribute to the development of new prognostic and predictive biomarkers for HCC. These relationships in the hepatic microenvironment, similarly to those observed in the gut, suggest that the intratumoral mycobiome may emerge as a novel therapeutic target as well as a promising area for biomarker discovery.