A Novel Cancer-associated Fibroblasts-Related Gene Signature Based on Adjuvant Sorafenib Efficacy for Predicting Recurrence and Immunotherapy Response in Hepatocellular Carcinoma After Hepatectomy

doi:10.21203/rs.3.rs-2745988/v1

Download PDF

Research Article

A Novel Cancer-associated Fibroblasts-Related Gene Signature Based on Adjuvant Sorafenib Efficacy for Predicting Recurrence and Immunotherapy Response in Hepatocellular Carcinoma After Hepatectomy

https://doi.org/10.21203/rs.3.rs-2745988/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

Due to its high recurrence rate, hepatocellular carcinoma (HCC) has a poor prognosis after hepatectomy. An effective model to predict postoperative recurrence and identify high-risk patients is essential. Recent studies have revealed the important role of cancer-associated fibroblasts (CAFs) in predicting HCC prognosis. However, the prognostic value of CAFs-related gene signature in HCC recurrence remains unknown. According to the BIOSTORM study, adjuvant sorafenib efficacy data may help to predict the recurrence in HCC. Therefore, we aimed to create a novel CAFs-related gene signature based on adjuvant sorafenib efficacy to predict HCC recurrence.

Methods

The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases were used to obtain the transcriptomic gene expression profiles and corresponding clinical data of HCC patients. The CAFs-related genes based on adjuvant sorafenib efficacy were identified using EPIC and weighted gene co-expression network analysis (WGCNA) algorithm. Univariate and the least absolute shrinkage and selection operator (LASSO) Cox regression analyses were used to establish a novel risk model. Univariate and multivariate COX analyses were used to identify independent prognostic factors for disease-free survival (DFS), and a nomogram was developed. The CIBERSORT and ESTIMATE algorithms were used to assess the tumor microenvironment components. Tumor immune dysfunction and exclusion (TIDE) score was used to predict immunotherapy response.

Results

A novel risk model was created using ten CAFs-related genes based on adjuvant sorafenib efficacy (DCLRE1C, DDX11, MAP4K2, SHCBP1, ADAM12, PAQR4, BEND3, ADAMTSL2, NUP93 and MPP2). Survival analyses revealed that high-risk patients had worse DFS, and the risk model was found as an independent prognostic factor for DFS in both the training and validation groups. A novel nomogram combined with pathologic stage and risk score status was developed. In the high-risk group, the stromal and immune cell content was found significantly lower while the tumor purity was significantly higher. In addition, immune checkpoints genes were highly expressed in the high-risk group and a higher risk score may predict a better response to immunotherapy.

Conclusions

The novel risk model comprised of ten CAFs-related genes based on adjuvant sorafenib efficacy may accurately predict recurrence and immunotherapy response in HCC patients after hepatectomy.

Bioinformatics analysis

Hepatocellular carcinoma

Cancer‐associated fibroblasts

Adjuvant sorafenib efficacy

Recurrence

Immunotherapy

Hepatocellular carcinoma (HCC) is the leading cause of cancer-related global death (1). Although the proportion of HCC patients getting surgical resection has increased as diagnosis has improved, those who have received radical resection still have a 70% recurrence rate within 5 years (2). In order to enhance long-term outcomes, an efficient model that can identify patients at a high risk of postoperative recurrence and provide individualized adjuvant treatment is crucial. The traditional recurrence predictive model incorporates tumor stage, tumor size, microvascular invasion, tumor differentiation, and other pertinent clinical characteristics, as well as a single serum alpha-fetoprotein expression. However, its specificity and sensitivity were insufficient to differentiate individuals with heterogeneity.

Cancer-associated fibroblasts (CAFs), the predominant cell type within the stroma, orchestrate interaction with cancer cells and express a variety of prognostic markers (3). CAFs-related prognostic models have been described for colorectal cancer (4), breast cancer (5) and clear cell renal cell carcinoma (6). These findings suggest that CAFs may play a significant role in cancer prognosis and represent a prospective cancer therapeutic target. Multiple studies have demonstrated the prognostic value of CAFs in predicting HCC recurrence and metastasis after radical therapy (7–9). Recent reports have indicated that CAFs-related gene signatures can be used to predict HCC prognosis (10, 11). However, these signatures were primarily focused on overall survival (OS) prediction thus the prognostic significance of CAFs-related genes signature in HCC recurrence remains unknown.

There is currently no standard adjuvant therapy to prevent HCC recurrence after resection or ablation. The primary objective of the phase 3 STORM trial comparing sorafenib to placebo as adjuvant therapy for HCC recurrence-free survival (RFS) was not met (12). In the subsequent BIOSTORM study, however, the adjuvant sorafenib efficacy data have been used to generate a gene signature to identify HCC patients who can benefit from adjuvant sorafenib and experience an extension in recurrence-free survival (RFS) (13). The gene signature can identify 30% of patients as "sorafenib RFS responders," who benefited from sorafenib in terms of RFS irrespective of clinical-pathological variables. Therefore, we believe that adjuvant sorafenib efficacy has the potential to predict recurrence and may be beneficial in developing a recurrence-related gene signature for HCC.

In the present study, differentially expressed genes (DEGs) were first extracted from the adjuvant “sorafenib RFS responders” and “non-responders”, then intersections between CAFs and adjuvant sorafenib efficacy genes were exhaustively investigated. On this basis, the critical CAFs-related genes with prognostic significance based on adjuvant sorafenib efficacy were identified in order to construct a novel risk score model. Furthermore, prognostic analyses were performed to assess its ability to predict HCC recurrence in both the training and validation cohorts, and a prognostic nomogram was developed. Finally, the correlations between the risk score model and immune infiltration, immune checkpoint expression and drug susceptibility were evaluated further. This study may therefore provide new insights into the interaction between CAFs and the efficacy of adjuvant sorafenib, as well as a novel prognostic model for HCC recurrence.

Abundance of CAFs was a poor prognostic factor associated with the progression of HCC

Using EPIC (Estimating the Proportions of Immune and Cancer cells, a tool to estimate the proportions of different cell type from bulk gene expression data) (14), we first accessed the abundance of CAFs in the Cancer Genome Atlas (TCGA) cohort. CAFs were more abundant in HCC samples compared to normal samples (p = 0.0085, Fig. 2A). Moreover, high levels of CAFs in tumor tissue were significantly associated with poorer overall survival (p = 0.02, Fig. 2B) and disease-free survival (p = 0.036, Fig. 2C) in HCC patients. In addition, we investigate the differences in CAF distribution between samples at various phases of disease progression. Notably, there was a significant difference in CAFs abundance between different histologic grade (G1 vs. G2, p = 0.005; G1 vs. G3, p = 0.016, Fig. 2D), pathologic T classification (T1 vs. T3, p = 0.033, Fig. 2E), and pathologic stage (stage 1 vs. stage 3, p = 0.012, Fig. 2F).

Identification Of 140 Cafs Related Degs Based On Adjuvant Sorafenib Efficacy In Hcc

A total of 1,297 genes were differentially expressed between the HCC tumors and normal controls in the TCGA cohort (Fig. 3A), whereas 2,315 genes were differentially expressed between sorafenib RFS responder and non-responder in the GSE109211 dataset (Fig. 3B). Therefore, the overlapped 399 DEGs in the TCGA and GSE109211 were screened out as adjuvant sorafenib efficacy related DEGs in HCC (Fig. 3C). Next, weighted gene co-expression network (WCGNA) was used to identify modules that had a significant association with elevated CAF levels in the TCGA cohort. On the basis of scale independence and average connectivity, 4 was selected as the soft threshold (scale-free R² = 0.9) (Figs. 3D). Four modules (brown, yellow, green and blue) were positively correlated with CAFs with a correlation coefficient higher than 0.30 (Fig. 3E, 3F). Finally, 140 genes in these four modules were chosen for further analyses as CAFs-related genes based on adjuvant sorafenib efficacy.

Protein-protein Interaction And Function Enrichment Analysis

Then, the potential interactions between these CAFs-related genes based on adjuvant sorafenib efficacy were analyzed by the protein-protein interaction (PPI) network, and the results revealed significant interactions among most of the CAFs-related genes (Fig. 4A). The PTPRC, IL1B, MKI67, DNMT3B, NOTCH3, IL10, MCM8, UBE2C, CSF1R and CTLA4 genes were identified as the top ten core genes in the network (Table 1). With the gene ontology (GO) functional enrichment analyses, the CAFs-related genes were found enriched in the T cell activation, cell-cell adhesion, T cell differentiation, extracellular matrix structural constituent and transmembrane receptor protein kinase activity (Fig. 4B). Kyoto encyclopedia of genes and genomes (KEGG) pathway analysis showed that MAPK signaling pathway, T cell receptor signaling pathway, alcoholic liver disease, TNF signaling pathway, Th17 cell differentiation and Th1 and Th2 cell differentiation were mainly enriched (Fig. 4C). These findings indicated that these CAFs-related genes based on adjuvant sorafenib efficacy were involved in the regulation of HCC tumor microenvironment.

Table 1

A list of 10 core genes among 140 CAFs-related genes in PPI networks
Gene Symbol	Average Shortest Path Length	Betweenness Centrality	Closeness Centrality	Degree
PTPRC	1.9	1.53258911	0.52631579	37
IL1B	1.90566038	1.35914212	0.52475248	35
MKI67	1.72972973	1.46805843	0.578125	30
DNMT3B	1.78873239	0.68738682	0.55905512	28
NOTCH3	1.9375	1.6066114	0.51612903	27
IL10	2.05555556	0.43948887	0.48648649	27
MCM8	2.3	0.8922638	0.43478261	26
UBE2C	1.25	0.2473715	0.8	26
CSF1R	2	0.47599578	0.5	26
CTLA4	1.94520548	0.60220174	0.51408451	26
CAFs, cancer-associated fibroblasts; PPI, protein-protein interaction.

Development And Validation Of The Novel Prognostic Risk Score Model Based On Crucial Cafs-related Genes

Sixty-eight CAFs-related genes were found to be significantly associated with disease-free survival (DFS) prognosis by univariate Cox analysis (all p < 0.05, Additional file 1, Table S1), and sixteen genes were found to be independent prognostic factors for DFS by multivariate Cox analysis (all p < 0.05, Table 2). After screening for optimal gene combinations using the least absolute shrinkage and selection operator (LASSO) algorithm, ten crucial CAFs-related genes were identified (DCLRE1C, DDX11, MAP4K2, SHCBP1, ADAM12, PAQR4, BEND3, ADAMTSL2, NUP93 and MPP2, Fig. 5A-B).

Table 2

Sixteen CAFs-related genes found to be independent prognostic factors for DFS by multivariate COX analysis
Gene Symbol	P value	HR	HR lower 95%	HR upper 95%
DCLRE1C	2.53E-02	2.478	1.187	6.154
DDX11	7.61E-03	1.865	1.127	3.086
MAP4K2	2.32E-04	2.795	1.572	4.968
SHCBP1	2.24E-03	3.217	1.437	7.201
ADAM12	2.31E-06	3.176	1.937	5.207
PAQR4	1.64E-02	1.478	1.033	2.116
BEND3	1.66E-03	4.247	1.617	11.154
ADAMTSL2	6.85E-04	0.728	0.600	0.884
NUP93	1.18E-02	0.312	0.114	0.855
MPP2	2.95E-02	0.254	0.051	0.825
HECTD2	4.57E-02	0.350	0.096	0.882
LIMK1	1.53E-02	0.566	0.337	0.948
RIBC2	4.22E-02	0.573	0.293	0.979
KRBA1	2.49E-02	0.604	0.365	0.998
ATAD3B	3.11E-02	1.707	1.162	2.994
CEP290	3.31E-02	2.405	1.833	6.137
CAFs, cancer-associated fibroblasts; DFS, disease-free survival; HR, Hazard Ratio.

Then, the expression and coefficients of these ten genes were utilized to develop a risk score model. The risk score = (-0.0631) * Exp _MPP2 + (-0.0577) * Exp _NUP93 + (-0.0311) * Exp _ADAMTSL2 + (0.0139) * Exp _BEND3 + (0.023) * Exp _PAQR4 + (0.0252) * Exp _ADAM12 + (0.0285) * Exp _SHCBP1 + (0.0467) * Exp _MAP4K2 + (0.0586) * Exp _DDX11 + (0.114) * Exp _DCLRE1C. As a result, each patient in the TCGA and GSE76427 cohorts was given a risk score for risk evaluation of DFS. The median risk scores were used to divide the patients into high- and low-risk subgroups in the TCGA and validation groups. Survival analyses showed that the DFS of high-risk subgroups in the TCGA cohort (p < 0.001, Fig. 5C-5D) were significantly worse than the DFS of low-risk subgroups. Similarly, Survival analyses showed that the DFS of high-risk subgroups in the GSE76427 cohort (p = 0.034, Fig. 5F-5G) were significantly worse than the DFS of low-risk subgroups. The ROCs curves were plotted further. In the TCGA cohort, the area under the curve (AUC) for DFS was 0.784, while the specificity and sensitivity were 0.737 and 0.744, respectively (Fig. 5E). In the GSE76427 cohort, the AUC for DFS was 0.724, and the specificity and sensitivity were 0.683 and 0.729, respectively (Fig. 5H).

Furthermore, univariate and multivariate Cox analyses were performed to assess the independent prognostic values of the risk score model in the training and validation cohort. In the TCGA cohort, the pathologic stage and risk score were found to be independently significant in both univariate [stage, p < 0.001, HR = 1.749 (1.468–2.084); risk score status, p < 0.001, HR = 2.041 (1.504–2.770)] and multivariate [stage, p < 0.001, HR = 1.574 (1.259–1.968); risk score status, p < 0.001, HR = 2.012 (1.398–2.897)] Cox analyses (Table 3). Similarly, in the GSE76427 cohorts, risk score was found to be independently significant in both univariate [risk score status, p < 0.05, HR = 1.956 (1.054–3.629)] and multivariate [risk score status, p < 0.05, HR = 1.975 (1.039–3.753)] Cox analyses (Table 4).

Table 3

Univariate and multivariate Cox analysis in the TCGA cohort
Clinical characteristics	Uni-variable Cox		Multi-variable Cox
Clinical characteristics	HR (95% CI)	P value	HR (95% CI)	P value
Age (years, mean ± SD)	0.997 [0.985–1.009]	6.252E-01	-	-
Gender (Male/Female)	0.866 [0.632–1.187]	3.715E-01	-	-
Pathologic M (M0/M1)	4.834 [1.515–15.43]	3.229E-01	-	-
Pathologic N (N0/N1)	1.411 [0.348–5.722]	6.477E-01	-	-
Pathologic T (T1/T2/T3/T4)	1.688 [1.441–1.976]	2.439E-01	-	-
Pathologic stage (I / II / III / IV)	1.749 [1.468–2.084]	1.168E-10	1.574 [1.259–1.968]	6.700E-05
Histologic grade (G1/G2/G3/G4)	1.113 [0.914–1.354]	2.866E-01	-	-
Vascular invasion (No/Yes)	1.990 [1.407–2.814]	7.259E-05	1.425 [0.960–2.116]	7.862E-02
RS status (Low/High)	2.041 [1.504–2.770]	2.917E-06	2.012 [1.398–2.897]	1.680E-04
SD, standard deviation; TCGA, The Cancer Genome Atlas; HR, Hazard Ratio; RS, risk model.

Table 4

Univariate and multivariate Cox analysis in the GSE76427 cohort
Clinical characteristics	Uni-variable Cox		Multi-variable Cox
Clinical characteristics	HR (95% CI)	P value	HR (95% CI)	P value
Age (years)	1.005 [0.983–1.028]	6.359E-01	-	-
Gender (Male/Female)	1.673 [0.773–3.621]	1.867E-01	-	-
Pathologic stage (I/II/III/IV)	1.270 [0.891–1.810]	1.853E-01	1.172 [0.814–1.688]	3.941E-01
RS status (Low/High)	1.956 [1.054–3.629]	3.341E-02	1.975 [1.039–3.753]	4.443E-02
RS, risk score

Prognostic Nomogram Construction

On the basis of the results of multivariate Cox analyses for DFS in the TCGA, we devised a novel prognostic nomogram for DFS prediction with the pathologic stage and our risk score model (Fig. 6A), which could provide a simple method for predicting the 1-, 3-, and 5-year DFS of HCC. The nomogram was effective in predicting DFS with a high C-index of 0.7835 (95% CI, 0.7518–0.8152, p < 0.001). The C-indexes for predicting 1-, 3-, and 5-year DFS were 0.711 [95% CI, 0.679–0.743], 0.753 [95% CI, 0.721–0.785] and 0.789 [95% CI, 0.757–0.821], respectively. The calibration curves for 1-, 3- and 5-year DFS rates were also largely overlapped with the standard lines for the TCGA cohorts (Fig. 6B). DCA curves demonstrated that the combined model had the best net predictive value for DFS (Fig. 6C). In the TCGA cohort, the AUCs for predicting 1-, 3-, and 5-year DFS with the nomogram were 0.824 [95% CI, 0.792–0.856], 0.795 [95% CI, 0.763–0.827] and 0.790 [95% CI, 0.785–0.823], respectively (Fig. 6D). Taking together, these results indicated that the nomogram combining pathologic stage and our risk score model was effective for predicting HCC recurrence with a high discriminative capacity, which may facilitate the clinical management.

Gene Set Enrichment Analyses Between Different Risk Groups

To investigate the underlying molecular mechanisms of the risk model, we compared the high-risk and the low-risk groups of HCC patients in the TCGA cohort using gene set enrichment analyses (GSEA) analysis. Totally 25 KEGG signaling pathways were significantly enriched. In the high-risk group, oncological signatures dominated the enriched KEGG pathways (including cell cycle, DNA replication, p53 signaling, mismatch repair and so on, Fig. 7A-I). In the low-risk group, KEGG pathways were enriched primarily for metabolism process (including arachidonic acid, fatty acid, linoleic acid, retinol, tyrosine and so on, Fig. 7J-T). In addition, several biosynthesis and degradation pathways were substantially enriched in the low- and high-risk groups (Fig. 7U-Y).

Immune Related Analyses Between Different Risk Score Groups

Since the KEGG pathway enrichment analysis revealed that oncological pathways were enriched in the high-risk group, we questioned the association between risk score and tumor microenvironment. Thus, we calculated the ESTIMATE score, immune score, stromal score and tumor purity using ESTIMATE. We found that risk score was negatively correlated with stromal score (p < 0.001, Fig. 8A), immune score (p < 0.05, Fig. 8B), and ESTIMATE score (p < 0.001, Fig. 8C), and was positively correlated with tumor purity (p < 0.05, Fig. 8D), indicating that the stromal and immune cell content was significantly lower and the tumor purity was significantly higher in the high risk group, resulting in a poorer prognosis for the HCC patients. In addition, we assessed the proportional distribution of immune cells in the TCGA samples using CIBERSORT. In the high-risk group, we observed significantly higher proportions of B cell plasma, activated memory CD4 T cell, follicular helper T cell, M0 Macrophage and lower resting memory CD4 T cell (all p < 0.05) (Fig. 8E). These findings suggested that the infiltration of these immune cell subtypes may have a significant influence on the prognosis of HCC patients.

Tumor Immune Dysfunction And Exclusion Analysis

Given the clinical significance of therapeutic strategies based on immune checkpoint blockade in HCC, we investigate the relationship between the risk score, immune checkpoints and HLA family genes. As shown in Fig. 9A, the gene expression of CD274, CD40, CD70, CD86, CTLA-4, HAVCR2, ICOS, PDCD1, and TIGIT was substantially increased in the high-risk group. In the high-risk group, we also observed a significant decrease in the expression of B2M, HLA.B, and HLA.DOB (Fig. 9B). These findings suggested that the risk score model may affect the immunotherapy response of HCC patients. To investigate further whether our prognostic model can be used to evaluate the efficacy of immunotherapy in HCC patients, we analyzed the immunotherapy response of TCGA-LIHC patients using the tumor immune dysfunction and exclusion (TIDE) database. The results demonstrated that risk score was positively correlated with immunotherapy response (p = 0.0082, Fig. 9C), the high-risk group had a higher ratio of immunotherapy responders (p = 0.03157, Fig. 9D) and the responder group in HCC patients had a better DFS than the non-responder group (p = 0.037, Fig. 9E). It indicated that the HCC high-risk group had a greater immune escape potential and a more effective immunotherapy.

Correlations Between Risk Score Model And Drug Susceptibility

Since we observed that the oncological and metabolism-related pathways were enriched in according to the KEGG pathway enrichment analysis, and the oncological and metabolic changes in tumor may serve as potential targets for chemotherapeutic and targeted drugs. Therefore, we compared the IC50s of a number of chemotherapeutics between different risk groups. Patients in the high-risk group had lower IC50s of cisplatin, camptothecin, gemcitabine, paclitaxel and vinorelbine than those in the low-risk group, suggesting that patients may derive greater benefit from chemotherapy (Fig. 10). In addition, we assessed the sensitivity of patients in different risk subgroups to a number of targeted inhibitors. Patients in the low-risk group had substantially lower IC50s to multiple targeted drugs including lapatinib, erlotinib, gefitinib, dasatinib and lapatinib (Fig. 10). These results indicated the prospective benefits of specific targeted therapy for patients in the low-risk group.

Recently, the concept of tumor microenvironment (TME) was proposed and widely acknowledged, and the effect of TME on HCC recurrence has been intensively researched (15, 16). The majority of tumor-associated stroma cells in the TME of HCC are CAFs, which are distinct from normal fibroblasts and present in abnormally high numbers. CAFs serve a crucial role in tumor growth, invasion, and metastasis by providing a major source of secreted cytokines that promote tumor growth (17). Several studies have demonstrated the utility of CAFs in predicting prognosis of HCC after radical therapies (7–9), and CAFs-related gene signatures have been reported to predict HCC prognosis (10, 11). However, the existed signatures were mainly focused on OS prediction, the prognostic value of CAFs-related genes in HCC recurrence has yet to be thoroughly studied.

In this study, we developed a novel prognostic risk model comprised of ten CAFs-related genes (DCLRE1C, DDX11, MAP4K2, SHCBP1, ADAM12, PAQR4, BEND3, ADAMTSL2, NUP93 and MPP2) based on the efficacy of adjuvant sorafenib to predict HCC recurrence after hepatectomy. According to the risk score, patients can be divided into high- and low-risk groups and the high-risk patients had worse DFS. The risk model was an independent predictive factor for DFS in HCC in both training and validation cohorts, and a novel DFS nomogram was subsequently created. The BCLC staging system is a popular predictive model for HCC recurrence; however, it cannot effectively differentiate heterogeneous patients using only clinicopathologic characteristics (DFS prediction C-index = 0.62) (18). In our research, we incorporated the prognostic nomogram with the CAFs-related risk model and the American Joint Committee on Cancer (AJCC) pathologic stage. Therefore, our nomogram compensated for the deficiencies of the conventional model and attained a greater capacity for discrimination (DFS prediction C-index = 0.78).

Multiple studies suggest that the biological characteristics of the tumor may be the main cause of recurrence and metastasis. Recently, it has been reported that the underlying tumor biological characteristics are associated with treatment response and subsequent prognosis (19–21). Treatment response and biologic markers have been recommended to be incorporated into the AJCC staging system in breast cancer (21). There is currently no standard adjuvant therapy to prevent HCC recurrence after resection. The phase 3 STORM study demonstrated the ineffectiveness of adjuvant sorafenib for HCC after resection or ablation (12). In the subsequent BIOSTORM study, however, a multigene signature associated with enhanced recurrence-free survival on sorafenib was generated (13), indicating that adjuvant sorafenib efficacy may be useful to predict recurrence. According to the gene signature, 30% of patients were “sorafenib RFS responders”, who captured benefit to sorafenib in terms of RFS regardless of whatever clinical-pathological variable. Therefore, the adjuvant sorafenib efficacy data from the BIOSTORM study may be beneficial for developing recurrence-related gene signature in HCC, and DEGs were extracted from “sorafenib RFS responders” and “non-responders” to develop our recurrence-related gene signature.

Previous studies have also evaluated the effect of CAFs-related gene signatures on the prognosis of HCC (10), bladder cancer (22), renal cell carcinoma (6) and colon cancer (23). However, neither the source of CAFs nor CAFs-related genes have been precisely defined. Consequently, the CAFs-related genes utilized to develop the signatures in existing studies varied. Some studies obtained CAFs-related genes from mesenchyme stem cells (MSC) grown in a tumor conditioned medium through Molecular Signature Database, as well as by directly copying the CAFs-related genes from another study (23), which may greatly reduce the specificity of the subsequent gene signature associated with HCC. To increase the specificity of the gene signature, we first obtained DEGs from adjuvant sorafenib efficacy data from BIOSTORM study, then we used EPIC and WGCNA algorithm to cluster transcriptomes and correlate with CAFs (6, 22), and finally we used LASSO algorithm to identify the crucial CAFs-related genes. Using data from a large-scale clinical trial exploring adjuvant sorafenib efficacy, combining the EPIC, WGCNA and LASSO algorithm to identify CAFs-related genes may make our risk model more reliable.

Next, we attempted to explored the biological and prognostic significance of the ten crucial CAFs-related genes included in our risk model. DCLRE1C is a nuclear protein that is involved in DNA repair (24). Recent research indicated that developing a peptide that inhibits DNA repair by blocking the binding of DCLRE1C and DNA ligase IV can increase tumor radio-sensitivity (25). DDX11 is DNA-dependent ATPase and ATP-dependent DNA helicase that participates in various functions in genomic stability, including DNA replication and DNA repair (26, 27). A recent study demonstrated that DDX11 was an oncogene in HCC via a E2F1/DDX11/EZH2 positive feedback loop (28). MAP4K2 is a member of the serine/threonine protein kinase family, which acts as an essential component of the famous tumor-related MAPK signal transduction pathway(29). SHCBP1 is a proto-oncogene which mediates cell proliferation and carcinogenesis in human carcinomas and overexpression of SHCBP1 could significantly promote HCC cell proliferation, survival and colony formation in HCC cell lines (30). ADAM12 is an ADAM family that modulates cell-cell and cell-matrix interactions (31). In liver cancer, ADAM12 was significantly highly expressed and an independent predictor of poor prognosis (32). In addition, ADAM12 was up-regulated by transforming growth factor beta (TGF-β) in human activated hepatic stellate cells (33, 34). PAQR4 has been reported to be related to the development of multiple cancers (35). Gang Zhao et al. demonstrated that PAQR4 was highly expressed in HCC samples and negatively correlated with the overall survival of HCC patients (36). BEND3 is a transcriptional repressor which associates with the nucleolar remodeling complex and plays a key role in repressing rDNA transcription (37). ADMATSL2 is a secreted glycoprotein that binds the cell surface and extracellular matrix. Recent research suggested ADAMTSL2 was a negative regulator of TGF-β in human cardiac fibroblasts, inhibiting myofibroblast differentiation and pro-fibrotic properties (38). Nucleoporin 93 (NUP93) mediates b-catenin nuclear import to promote hepatocellular carcinoma progression and metastasis (39). MPP2 is a member of a family of membrane-associated proteins termed MAGUKs, which interact with the cytoskeleton and regulate cell proliferation (40). A new study found that liver cancer tissues had lower MPP2 expression than normal tissues, and overexpression of MPP2 increased apoptosis and decreased liver cancer cell proliferation, invasion, and migration (41). According to our GSEA analysis, some of these ten CAFs-related genes (such as DCLRE1C, DDX11, MAP4K2, SHCBP1, BEND3 and NUP93) may play crucial roles in the oncological signaling pathways like cell cycle, DNA replication, p53 signaling, and DNA repair. Furthermore, according to our GO and KEGG analyses, some of them (such as PAQR4, ADAM12, ADAMTSL2 and MPP2) may be involved in the tumor microenvironment remodeling, such as cell-cell adhesion, T cell differentiation, extracellular matrix structural components and transmembrane receptor protein kinase activity. Taking together, these ten CAFs-related genes based on adjuvant sorafenib efficacy may play significant roles in HCC recurrence thus were of great value in predicting recurrence in HCC.

With a better understanding of tumor microenvironment, we further investigated immune infiltration in various groups to obtain more insight into the relationship between risk scores and immune components. In the high-risk group, we observed substantially higher proportions of B cell plasma, activated memory CD4 T cell, follicular helper T cell, M0 Macrophage and lower resting memory CD4 T cell, indicating that immune regulation was disturbed in TME of the high-risk HCC patients. In addition, correlation analyses suggested that low-risk HCC patients had a higher stromal and immune cell content and reduced tumor purity. Taken together, the high-risk group had higher tumor purity and lower infiltration of immune cells, resulting in more aggressive tumors and poorer prognosis in HCC patients, which is consistent with the previous study (42, 43).

Although immunotherapy offers new hope to patients with HCC, validated biological markers for predicting the therapy's efficacy and guiding clinical decision-making are still lacking (44). By analyzing immunotherapy-related markers, we discovered that our CAFs-related prognostic model may accurately predicted the efficacy of immunotherapy in HCC. The high-risk group has a higher immune escape potential and was more likely to benefit from immunotherapy. In addition, we discovered that the risk model could predict the efficacy of a number of common anticancer medications, including cisplatin, camptothecin, gemcitabine, paclitaxel, vinorelbine, lapatinib, erlotinib, gefitinib, dasatinib, lapatinib and so on. Together, the CAFs-related risk score model and other indicators may be used to evaluate the efficacy of immunotherapy, chemotherapy and targeted therapy, which may aid in resolving difficulties in the adjuvant tailored treatment of HCC.

Our study has limitations as well. First, the CAFs-based risk signature was generated using retrospective data from public databases. In the future, it should be validated in additional prospective and multi-center HCC cohorts. Second, we only investigated the prospective prognostic value of predicting recurrence using the CAFs-related risk model; therefore, additional research is necessary to investigate the underlying mechanisms of the signature in the development of HCC recurrence. Third, there were no experimental data available for the identified signature. Therefore, further research is needed to elucidate the relationship between the CAFs-related signature and the prognosis of HCC patients.

In conclusion, our study constructed a novel prognostic model comprised of ten CAFs-related genes based on adjuvant sorafenib efficacy. The prognostic model exhibited stable predictive power in predicting recurrence and immunotherapy response, providing a possible direction for the development of personalized therapy for HCC patients after hepatectomy.

Data source and processing

The transcriptome data of 373 HCC tumor samples and 50 normal controls were obtained from TCGA database (TCGA-LIHC, https://portal.gdc.cancer.gov, 2022-03-10). The clinical and pathological information (age, gender, tumor grade, stage, follow-up time, and survival status) was downloaded from UCSC Xeno (https://xena.ucsc.edu/ public) on 2022-03-10. Disease-free survival time was defined as the time between surgery and disease recurrence or the last follow-up. Ultimately, 315 HCC patients with valid DFS information were designated as a training cohort for the development of a risk model. In addition, the gene expression microarray data and clinical characteristics of validation cohort GSE76427 were downloaded from the Gene Expression Omnibus (GEO) database. Finally, 115 HCC patients with complete recurrence information after liver resection were enrolled.

In addition, the gene expression microarray data of GSE109211, which included 67 HCC patients treated with sorafenib, were extracted from the GEO database in order to identify genes correlated with the efficacy of adjuvant sorafenib (GSE109211, authors presented results of the biomarker companion study BIOSTORM in which they defined predictors of recurrence prevention with sorafenib as an adjuvant therapy in patients with HCC after surgical resection or local ablation, https://www.ncbi.nlm.nih.gov/geo/, platform GPL13938). Among these 67 samples, 21 patients benefited from sorafenib adjuvant therapy in terms of extended recurrence-free survival (so-called "sorafenib RFS responders"), while 46 patients had no effect ("sorafenib RFS non-responders"). The flowchart of the present study design is shown in Fig. 1. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013).

Cafs Assessment

EPIC (https://gfellerlab.shinyapps.io/EPIC 1–1/) was used to compute the CAFs score of patients recruited from the TCGA database in order to estimate the proportion of CAFs in matrix components of tumor tissues. The EPIC application is designed to estimate the proportion of immune cells and cancer cells from bulk tumor gene expression data (45). According to the median, tumor samples were classified into “low” and “high” CAFs groups. Then, correlation analyses between different CAFs groups and clinical variables were conducted. Using Kaplan–Meier survival curves and the log-rank test, the difference in OS and DFS between the high and low CAFs groups was determined.

Identification Of Degs Base On Adjuvant Sorafenib Efficacy Data In Hcc

Using R package “limma” with the criteria of |log₂fold-change| > 1 and false discovery rate (FDR) < 0.05, we identified DEGs between HCC tumors and normal/adjacent controls in the TCGA cohort and between sorafenib RFS responders and non-responder in the GSE109211 dataset. The overlapped DEGs in the TCGA and GSE109211 were subsequently selected as DEGs based on the adjuvant sorafenib efficacy in HCC for the subsequent WCGNA analysis.

Wcgna For The Determination Of Key Cafs-related Genes Based On Adjuvant Sorafenib Efficacy In Hcc

The WGCNA R-package Version 1.61 was used for co-expression network analysis (https://cran.r-project.org/web/packages/WGCNA/index.html), with cut height detection threshold of 0.995. The modules that were significantly positively correlated with CAFs (correlation coefficient > 0.3) were reserved, and the genes within those modules were defined as CAFs-related genes based on adjuvant sorafenib efficacy.

Ppi Interaction And Function Enrichment Analysis

The PPI network of CAFs-related genes based on adjuvant sorafenib efficacy in HCC was analyzed at the STRING website (https://www.string-db.org/). Then, the network was further analyzed in the Cytoscape software (Version 3.9.0, https://cytoscape.org/) for identification of key genes with “CentiScaPe” (Version2.2, http://apps.cytoscape.org/apps/centiscape) (46). GO and KEGG pathway enrichment analyses were performed using R package “clusterProfiler” (Version 4.4.4) (47). The screening threshold was set at 0.05.

Construction and validation of the risk score model based on the prognostic CAFs-related genes based on adjuvant sorafenib efficacy

Univariate Cox regression analysis was used to further screen for CAFs-related genes based on adjuvant sorafenib efficacy that were significantly associated with DFS. These genes were utilized to conduct a multivariate cox regression analysis in order to identify the independent DFS-related genes. Finally, LASSO algorithm was applied to prevent overfitting of the prognostic model and to identify the optimal prognostic genes for constructing the prognostic risk score model (48). The risk score was calculated in both the TCGA and validation cohort using the following formula: riskscore (RS) = ∑Coef_genes × Exp _genes. Coef_genes represents the LASSO prognostic coefficient of the target gene, and Exp _genes represents the expression level of the target gene. Taking the median risk score as a cutoff, HCC patients were divided into high-risk group and low-risk group according to the risk value. DFS was further analyzed in the high- and low-risk groups using the “survival” R package.

Prognostic Nomogram Construction And Assessment

In both the training and validation cohorts, the univariate and multivariate Cox regression analyses were used to evaluate the independent prognostic significance of the risk score model. A nomogram for DFS prediction for HCC was developed using the R package “rms”. The prediction performance was evaluated using the concordance index (C-index), calibration curves, and time-dependent receiver operating characteristics (ROCs) through the R packages “survcomp”, “rmda” and “timeROC”.

Gene Set Enrichment Analysis

To elucidate the relationship between the risk scores and biological functions, the GSEA analysis was performed to evaluate the enriched pathways within different groups (49). This was performed using the GSEA software (http://www.gsea-msigdb.org/gsea/index.jsp) with the KEGG gene set (C2.cp.kegg.v7.2.symbols.gmt). An FDR < 0.25 and adjusted p < 0.05 were considered statistically significant.

Immune Infiltration And Tumor Microenvironment Analysis

CIBERSORT (https://cibersort.stanford.edu/index.php) was utilized to assess the proportion of immune cell types and compare the variability of distinct immune cell distributions in various risk groups (50). Then, we calculated the ESTIMATE scores, immune scores, stromal scores and tumor purity using the “estimate” R package (51).

Tumor Immune Dysfunction And Exclusion Analysis

To determine whether the risk model can be used to evaluate the efficacy of immunotherapy in patients, we analyzed the correlations between risk score model, the immune checkpoint genes and HLA-family genes. TCGA sample expression profiles were then transmitted to the tumor immune dysfunction and exclusion website (http://tide.dfci.harvard.edu/) for TIDE and dysfunction scores of each sample. Using the TIDE score, we were able to predict the response of individual HCC patients to immune checkpoint therapy.

Drug Susceptibility Analysis

To evaluate the potential significance of the risk score model in the treatment guidance for HCC, we compared the IC50s of regular chemotherapeutic agents and several targeted drugs between high- and low-risk groups using the R package “pRRophetic”.

Statistics Analysis

All statistical analyses were performed using R software (version 4.0.4; The R Foundation for Statistical Computing, Vienna, Austria). A two-sided p-value < 0.05 was considered statistically significant.

HCC, Hepatocellular carcinoma, CAFs, Cancer-associated fibroblasts, OS, Overall survival, RFS, Recurrence-free survival, DEGs, Differentially expressed genes, TCGA, The cancer genome atlas, PPI, Protein-protein interaction, GO, Gene ontology, KEGG, Kyoto encyclopedia of genes and genomes, DFS, Disease-free survival, WGCNA, Weighted gene co-expression network analysis, LASSO, Least absolute shrinkage and selection operator, AUC, Area under the curve, GSEA, Gene set enrichment analyses, TIDE, Tumor immune dysfunction and exclusion, GEO, Gene expression omnibus, TME, Tumor microenvironment, C-index, Concordance index, ROCs, receiver operating characteristics, RS, Risk score.

Ethics approval and consent to participate

The study involving human material were performed in accordance with the Declaration of Helsinki. Both TCGA and GEO data were obtained from public databases, thus, there was no need for additional ethics approval.

Consent for publication

Not applicable.

Availability of data and materials

TCGA-LIHC can be downloaded from UCSC Xeno (https://xena.ucsc.edu/public). GSE76427 and GSE109211 are accessible in Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo).

Competing interests

The authors declare that they have no competing interests.

Funding: This work was supported by the Natural Science Foundation of Guangdong Province (Grant No. 2022A1515010519), the Guangzhou Basic and Applied Basic Research Project (Grant No. 202201020439) and the Science and Technology Program of Guangzhou (Grant No. 202201020439).

Author Contributions: (I) Conception and design: RBL, GW, (II) Administrative support: GW, (III) Provision of study materials or patients: RBL, (IV) Collection and assembly of data: RBL, ZHL, JF, (V) Data analysis and interpretation: RBL, ZHL, JF, (VI) Manuscript writing: All authors, (VII) Final approval of manuscript: All authors.

Acknowledgments

The authors express their appreciation to TCGA and GEO databases for the availability of the data.

Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2018,68(6):394-424.
EASL Clinical Practice Guidelines: Management of hepatocellular carcinoma. J Hepatol. 2018,69(1):182-236.
Baglieri J, Brenner DA, Kisseleva T. The Role of Fibrosis and Liver-Associated Fibroblasts in the Pathogenesis of Hepatocellular Carcinoma. Int J Mol Sci. 2019,20(7).
Pan H, Pan J, Wu J. Development and validation of a cancer-associated fibroblast-derived lncRNA signature for predicting clinical outcomes in colorectal cancer. Front Immunol. 2022,13:934221.
Hu J, Jiang Y, Wei Q, Li B, Xu S, Wei G, et al. Development of a Cancer-Associated Fibroblast-Related Prognostic Model in Breast Cancer via Bulk and Single-Cell RNA Sequencing. Biomed Res Int. 2022,2022:2955359.
Liu B, Chen X, Zhan Y, Wu B, Pan S. Identification of a Gene Signature for Renal Cell Carcinoma-Associated Fibroblasts Mediating Cancer Progression and Affecting Prognosis. Front Cell Dev Biol. 2020,8:604627.
Fang M, Yuan J, Chen M, Sun Z, Liu L, Cheng G, et al. The heterogenic tumor microenvironment of hepatocellular carcinoma and prognostic analysis based on tumor neo-vessels, macrophages and α-SMA. Oncol Lett. 2018,15(4):4805-12.
Lau EY, Lo J, Cheng BY, Ma MK, Lee JM, Ng JK, et al. Cancer-Associated Fibroblasts Regulate Tumor-Initiating Cell Plasticity in Hepatocellular Carcinoma through c-Met/FRA1/HEY1 Signaling. Cell Rep. 2016,15(6):1175-89.
Takamura H, Nakanuma S, Hayashi H, Tajima H, Kakinoki K, Sakai S, et al. Evaluation of eligibility criteria in living donor liver transplantation for hepatocellular carcinoma by α-SMA-positive cancer-associated fibroblasts. Oncol Rep. 2013,30(4):1561-74.
Dong W, Xie Y, Huang H. Prognostic Value of Cancer-Associated Fibroblast-Related Gene Signatures in Hepatocellular Carcinoma. Front Endocrinol (Lausanne). 2022,13:884777.
Zou B, Liu X, Gong Y, Cai C, Li P, Xing S, et al. A novel 12-marker panel of cancer-associated fibroblasts involved in progression of hepatocellular carcinoma. Cancer Manag Res. 2018,10:5303-11.
Bruix J, Takayama T, Mazzaferro V, Chau GY, Yang J, Kudo M, et al. Adjuvant sorafenib for hepatocellular carcinoma after resection or ablation (STORM): a phase 3, randomised, double-blind, placebo-controlled trial. Lancet Oncol. 2015,16(13):1344-54.
Pinyol R, Montal R, Bassaganyas L, Sia D, Takayama T, Chau GY, et al. Molecular predictors of prevention of recurrence in HCC with sorafenib as adjuvant treatment and prognostic factors in the phase 3 STORM trial. Gut. 2019,68(6):1065-75.
Racle J, Gfeller D. EPIC: A Tool to Estimate the Proportions of Different Cell Types from Bulk Gene Expression Data. Methods Mol Biol. 2020,2120:233-48.
Sun Y, Wu L, Zhong Y, Zhou K, Hou Y, Wang Z, et al. Single-cell landscape of the ecosystem in early-relapse hepatocellular carcinoma. Cell. 2021,184(2):404-21.e16.
Sas Z, Cendrowicz E, Weinhäuser I, Rygiel TP. Tumor Microenvironment of Hepatocellular Carcinoma: Challenges and Opportunities for New Treatment Options. Int J Mol Sci. 2022,23(7).
Yin Z, Dong C, Jiang K, Xu Z, Li R, Guo K, et al. Heterogeneity of cancer-associated fibroblasts and roles in the progression, prognosis, and therapy of hepatocellular carcinoma. J Hematol Oncol. 2019,12(1):101.
Wang JC, Hou JY, Chen JC, Xiang CL, Mao XH, Yang B, et al. Development and validation of prognostic nomograms for single large and huge hepatocellular carcinoma after curative resection. Eur J Cancer. 2021,155:85-96.
Bear HD, Anderson S, Brown A, Smith R, Mamounas EP, Fisher B, et al. The effect on tumor response of adding sequential preoperative docetaxel to preoperative doxorubicin and cyclophosphamide: preliminary results from National Surgical Adjuvant Breast and Bowel Project Protocol B-27. J Clin Oncol. 2003,21(22):4165-74.
Cortazar P, Zhang L, Untch M, Mehta K, Costantino JP, Wolmark N, et al. Pathological complete response and long-term clinical benefit in breast cancer: the CTNeoBC pooled analysis. Lancet. 2014,384(9938):164-72.
Mittendorf EA, Vila J, Tucker SL, Chavez-MacGregor M, Smith BD, Symmans WF, et al. The Neo-Bioscore Update for Staging Breast Cancer Treated With Neoadjuvant Chemotherapy: Incorporation of Prognostic Biologic Factors Into Staging After Treatment. JAMA Oncol. 2016,2(7):929-36.
Du Y, Jiang X, Wang B, Cao J, Wang Y, Yu J, et al. The cancer-associated fibroblasts related gene CALD1 is a prognostic biomarker and correlated with immune infiltration in bladder cancer. Cancer Cell Int. 2021,21(1):283.
Herrera M, Berral-González A, López-Cade I, Galindo-Pumariño C, Bueno-Fortes S, Martín-Merino M, et al. Cancer-associated fibroblast-derived gene signatures determine prognosis in colon cancer patients. Mol Cancer. 2021,20(1):73.
Ma Y, Pannicke U, Schwarz K, Lieber MR. Hairpin opening and overhang processing by an Artemis/DNA-dependent protein kinase complex in nonhomologous end joining and V(D)J recombination. Cell. 2002,108(6):781-94.
Zhu C, Wang X, Li P, Zhu Y, Sun Y, Hu J, et al. Developing a Peptide That Inhibits DNA Repair by Blocking the Binding of Artemis and DNA Ligase IV to Enhance Tumor Radiosensitivity. Int J Radiat Oncol Biol Phys. 2021,111(2):515-27.
Hirota Y, Lahti JM. Characterization of the enzymatic activity of hChlR1, a novel human DNA helicase. Nucleic Acids Res. 2000,28(4):917-24.
Shah N, Inoue A, Woo Lee S, Beishline K, Lahti JM, Noguchi E. Roles of ChlR1 DNA helicase in replication recovery from DNA damage. Exp Cell Res. 2013,319(14):2244-53.
Su SG, Li QL, Zhang MF, Zhang PW, Shen H, Zhang CZ. An E2F1/DDX11/EZH2 Positive Feedback Loop Promotes Cell Proliferation in Hepatocellular Carcinoma. Front Oncol. 2020,10:593293.
Delire B, Stärkel P. The Ras/MAPK pathway and hepatocarcinoma: pathogenesis and therapeutic implications. Eur J Clin Invest. 2015,45(6):609-23.
Zou A, Wu A, Luo M, Zhou C, Lu Y, Yu X. SHCBP1 promotes cisplatin induced apoptosis resistance, migration and invasion through activating Wnt pathway. Life Sci. 2019,235:116798.
White JM. ADAMs: modulators of cell-cell and cell-matrix interactions. Curr Opin Cell Biol. 2003,15(5):598-606.
Du S, Sun L, Wang Y, Zhu W, Gao J, Pei W, et al. ADAM12 is an independent predictor of poor prognosis in liver cancer. Sci Rep. 2022,12(1):6634.
Le Pabic H, Bonnier D, Wewer UM, Coutand A, Musso O, Baffet G, et al. ADAM12 in human liver cancers: TGF-beta-regulated expression in stellate cells is associated with matrix remodeling. Hepatology. 2003,37(5):1056-66.
Le Pabic H, L'Helgoualc'h A, Coutant A, Wewer UM, Baffet G, Clément B, et al. Involvement of the serine/threonine p70S6 kinase in TGF-beta1-induced ADAM12 expression in cultured human hepatic stellate cells. J Hepatol. 2005,43(6):1038-44.
Wang K, Meng J, Wang X, Yan M, Liu S, Yang S, et al. Pan-cancer analysis of the prognostic and immunological role of PAQR4. Sci Rep. 2022,12(1):21268.
Zhao G, Shi X, Sun Z, Zhao P, Lu Z. PAQR4 promotes the development of hepatocellular carcinoma by activating PI3K/AKT pathway. Acta Biochim Biophys Sin (Shanghai). 2021,53(12):1602-13.
Khan A, Giri S, Wang Y, Chakraborty A, Ghosh AK, Anantharaman A, et al. BEND3 represses rDNA transcription by stabilizing a NoRC component via USP21 deubiquitinase. Proc Natl Acad Sci U S A. 2015,112(27):8338-43.
Rypdal KB, Erusappan PM, Melleby AO, Seifert DE, Palmero S, Strand ME, et al. The extracellular matrix glycoprotein ADAMTSL2 is increased in heart failure and inhibits TGFβ signalling in cardiac fibroblasts. Sci Rep. 2021,11(1):19757.
Lin CS, Liang Y, Su SG, Zheng YL, Yang X, Jiang N, et al. Nucleoporin 93 mediates β-catenin nuclear import to promote hepatocellular carcinoma progression and metastasis. Cancer Lett. 2022,526:236-47.
Ivanova S, Gregorc U, Vidergar N, Javier R, Bredt DS, Vandenabeele P, et al. MAGUKs, scaffolding proteins at cell junctions, are substrates of different proteases during apoptosis. Cell Death Dis. 2011,2(1):e116.
Li FY, Fan TY, Zhang H, Sun YM. Demethylation of miR-34a upregulates expression of membrane palmitoylated proteins and promotes the apoptosis of liver cancer cells. World J Gastroenterol. 2021,27(6):470-86.
Xue YN, Xue YN, Wang ZC, Mo YZ, Wang PY, Tan WQ. A Novel Signature of 23 Immunity-Related Gene Pairs Is Prognostic of Cutaneous Melanoma. Front Immunol. 2020,11:576914.
Wu M, Li X, Liu R, Yuan H, Liu W, Liu Z. Development and validation of a metastasis-related Gene Signature for predicting the Overall Survival in patients with Pancreatic Ductal Adenocarcinoma. J Cancer. 2020,11(21):6299-318.
Sangro B, Sarobe P, Hervás-Stubbs S, Melero I. Advances in immunotherapy for hepatocellular carcinoma. Nat Rev Gastroenterol Hepatol. 2021,18(8):525-43.
Racle J, de Jonge K, Baumgaertner P, Speiser DE, Gfeller D. Simultaneous enumeration of cancer and immune cell types from bulk tumor gene expression data. Elife. 2017,6.
Scardoni G, Petterlini M, Laudanna C. Analyzing biological network parameters with CentiScaPe. Bioinformatics. 2009,25(21):2857-9.
Yu G, Wang LG, Han Y, He QY. clusterProfiler: an R package for comparing biological themes among gene clusters. Omics. 2012,16(5):284-7.
Goeman JJ. L1 penalized estimation in the Cox proportional hazards model. Biom J. 2010,52(1):70-84.
Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005,102(43):15545-50.
Chen B, Khodadoust MS, Liu CL, Newman AM, Alizadeh AA. Profiling Tumor Infiltrating Immune Cells with CIBERSORT. Methods Mol Biol. 2018,1711:243-59.
Hu D, Zhou M, Zhu X. Deciphering Immune-Associated Genes to Predict Survival in Clear Cell Renal Cell Cancer. Biomed Res Int. 2019,2019:2506843.

No competing interests reported.

Additionalfile1.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

A Novel Cancer-associated Fibroblasts-Related Gene Signature Based on Adjuvant Sorafenib Efficacy for Predicting Recurrence and Immunotherapy Response in Hepatocellular Carcinoma After Hepatectomy

Status:

Version 1

Abstract

Background

Methods

Results

Conclusions

Figures

Background

Results

Abundance of CAFs was a poor prognostic factor associated with the progression of HCC

Identification Of 140 Cafs Related Degs Based On Adjuvant Sorafenib Efficacy In Hcc

Protein-protein Interaction And Function Enrichment Analysis

Development And Validation Of The Novel Prognostic Risk Score Model Based On Crucial Cafs-related Genes

Prognostic Nomogram Construction

Gene Set Enrichment Analyses Between Different Risk Groups

Immune Related Analyses Between Different Risk Score Groups

Tumor Immune Dysfunction And Exclusion Analysis

Correlations Between Risk Score Model And Drug Susceptibility

Discussion

Conclusions

Methods

Data source and processing

Cafs Assessment

Identification Of Degs Base On Adjuvant Sorafenib Efficacy Data In Hcc

Wcgna For The Determination Of Key Cafs-related Genes Based On Adjuvant Sorafenib Efficacy In Hcc

Ppi Interaction And Function Enrichment Analysis

Prognostic Nomogram Construction And Assessment

Gene Set Enrichment Analysis

Immune Infiltration And Tumor Microenvironment Analysis

Tumor Immune Dysfunction And Exclusion Analysis

Drug Susceptibility Analysis

Statistics Analysis

Abbreviations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1