Development and validation of a hepatocellular carcinoma classifier based on macrophage-related gene set for rapid evaluation of patient response types and prognosis

doi:10.21203/rs.3.rs-3120720/v1

Download PDF

Research Article

Development and validation of a hepatocellular carcinoma classifier based on macrophage-related gene set for rapid evaluation of patient response types and prognosis

https://doi.org/10.21203/rs.3.rs-3120720/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Tumor-associated macrophages (TAMs) participate in and shape the tumor microenvironment of hepatocellular carcinoma (HCC), which is closely related to the formation of tumor heterogeneity. The aim of this study is to distinguish different subtypes of patients according to the activity level of macrophage functional gene set in HCC. We collected 1203 tissue samples from TCGA, ICGC and GEO databases. Using macrophage-associated gene set (MRRGS) from GSEA database, the score of MRRGS was calculated based on gene set variation analysis (GSVA). The key MRRGS was screened by univariate COX regression analysis and LASSO regression. Finally, non-negative matrix factorization (NMF) was used to classify HCC subtypes. Six immune cell infiltration algorithms, immune checkpoint expression differences, tumor immunity and rejection (TIDE) analysis, mutation data analysis, stem cell index based on mRNA expression (mRNAsi) were used to evaluate and reveal the differences of immunity, mutation and tumor cell malignancy among different HCC subtypes. Weighted gene coexpression network (WGCNA) is used to analyze the functional mechanism involved in MRRGS. CAMP and drug sensitivity analysis are used to explore drugs for different HCC subtypes. Two machine learning algorithms assist in screening characteristic genes among subtypes to facilitate subtype discrimination. Our study divides patients into two subtype (C1 and C2) by defining 12 MRRGS, which are similar to hot and cold tumors mentioned in previous studies. The stability of the macrophage functional classifier was validated in two independent HCC cohorts and this classifier can well predict the ability of patients to respond to immunotherapy, TACE treatment and various drug. Based on the above results, we built a bioinformatics tool to help users quickly distinguish patient subtypes and prognosis. In addition, immune signals (such as PD1-PDL1 signals), mutations, metabolic abnormalities, viral infection and chemical erosion in the environment are important upstream foundations of HCC heterogeneity caused by macrophages. This provides insights into the clinical treatment and management of HCC.

Hepatocellular carcinoma

Tumor-associated macrophages

Bioinformatics

Tumor heterogeneity

Liver cancer usually refers to primary liver cancer (PLC) originating from hepatic diseases, and 80%-90% of HCC is hepatocellular carcinoma (HCC). It is a major public health problem in the world with poor prognosis and high incidence [1, 2]. Statistics from the International Agency for Research on Cancer (IARC) show that about 830,000 people worldwide died of liver cancer, rank only second to lung cancer in 2020 [3]. Due to the lack of early symptoms, PLC patients are mostly diagnosed in the late stage, which resulting the low survival rate of liver cancer.

Tumor heterogeneity is the common result of multiple molecular mechanisms, which is related to the development of various cancers, poor therapeutic efficacy and drug resistance [4]. Tumor microenvironment (TME) is considered as the main cause of Tumor heterogeneity. Physical and chemical conditions and cytokines in TME can promote changes in the characterization of tumor cells, which are important basis for the plasticity of tumor cells [5]. At present, therapeutic efficacy of the existing therapeutic methods are not universal and complete enough. For example, surgical resection is only applicable to early patients, long-term use of chemotherapy drugs such as sorafenib will gradually develop drug resistance, accompanied by adverse reactions [6]. The most popular PD1/PDL1 (programmed death receptor 1/ programmed death ligand 1) inhibitors in immunotherapy also reflect drug resistance in some patients [7]. Therefore, it is important to identify patients with different tumor types for designing individualized treatment.

Tumor-associated macrophages (TAM) are the most abundant type of immune cells in TME, mostly derived from monocytes in blood [8]. Tumor cells and stromal cells in microenvironment can recruit the macrophage to tumors by secreting chemokines. Recruited macrophages are gradually domesticated by tumor cells and subsequently make phenotypic changes. They perform a series of tumor-promoting functions, such as promoting angiogenesis, tumor metastasis and anti-immune response, which are of great significance to the formation of TME. A great deal of studies have confirmed that high TAM infiltration level is an adverse factor affecting the prognosis of HCC patients [9, 10]. The traditional classification method divides macrophages into M1 macrophages featuring pro-inflammatory along with anti-tumor and M2 macrophages with the features of pro-tumor as well as anti-inflammatory. Although it has been proved that TAM has strong plasticity, that M1 and M2 subtype are only two extreme cases in TAM trajectory, it’s still great reference for many other studies. Researchers have found[11] that TAM usually presents M2 characterization and plays a crucial role in the formation of TME by secreting related cytokines, growth factors, chemokines and extracellular matrix components (ECM) [12, 13]. The changing TAM infiltration will present different TME levels, which is the basis for the occurrence of tumor heterogeneity. Other studies have shown that TAM is related to drug resistance of cancer patients, generation of immune checkpoints, and anti-tumor response mediated by T cells and natural killer cells [14–16].

In this study, we aimed to identify patients with different immune types by the level of macrophage-associated gene set(MRRGS). On this basis, we will explore differences in immune responses and mutational patterns between subtypes, potential downstream impact mechanisms and screen potential therapeutic drugs. It will provide new insights and assistance for the advancement of precision medicine in clinical HCC

2.1 Data sources and processing

HCC transcriptome and single nucleotide variant (SNV) data is from TCGA-LIHC project of TCGA database (https://portal.gdc.cancer.gov/)[17], the clinical data has been extracted and merged. Single-cell transcriptome data is from GEO database (https://www.ncbi.nlm.nih.gov/geo/)[18]. The samples were filtered, quality controlled and normalized according to the number of mitochondrial and ribosomal gene expression. Then PCA, t-SNE and UMAP are used for preliminary dimension reduction to determine whether there was batch effect on samples for elimination by CCA algorithm.The MRRGS were obtained by using the term of “macrophage” in MSigDB database (http://www.gsea-msigdb.org/gsea/index.jsp)[19]. The LINC-JP project under the ICGC database (https://dcc.icgc.org/) and the GSE54236 series of the GEO database will be used for independent external validation. Transcatheter arterial chemoembolization (TACE) is the standard clinical treatment strategy for stage B liver cancer. We used the cohort GSE104580 with or without response after TACE treatment to evaluate the ability of subtypes to predict the response to TACE treatment (see Fig. 1 for the research process of this study).

2.2 Identification of macrophage immune response subtypes in different HCC patients

The ImmuneScore of HCC patients was calculated by the Estimate algorithm. Gene set variation analysis (GSVA) was used to obtain the Enrichment score (ES) of MRRGS, which was used to evaluate the activity of each MRRGS in the samples. Difference analysis (|logFC | >= 1, FDR = < 0.05) and single factor COX regression analysis would screen the MRRGS which is different between normal samples and HCC patients and has a significant impact on survival. Subsequently, based on the survival information, Lasso regression and cross-validation are used to further converge the survival related MRRGS. Lasso regression is a commonly used algorithm. By defining the unknown parameter lambda and using cross-validation to select the optimal lambda, it can quickly and effectively extract important variables when there are many variables. At present, this analysis can be carried out in R language through glmnet program package. At the same time, we use the coefficient of each MRRGS obtained by lasso regression method to calculate the sample Risk score, and the formula is as follows:

Risk Score=∑Coefficents × Expression Levels.

Then, according to the median value of Risk Score, we divided the patients into high-risk group and low-risk group. The above analysis was carried out on the training set TCGA to screen out the MRRGS for the patient classification model. The HCC queues of ICGC and GEO will be used as completely independent external data sets to verify the results of the classification and subsequent results. The verification results of the two queues are given in the supplementary file. According to the median grouping of ES, the survival curve of MRRGS was drawn, and the difference of prognosis among groups was evaluated by log-rank test. And the ES of MRRGS is mathematically transformed to match the input of NMF clustering:

$${N}{E}{S}={{e}}^{{E}{S}}$$

(e is Euler's number which approximately equal to 2.71828)

Then, using the nmf function in the NMF program package of R language, all samples are clustered by non-negative matrix factorization (NMF) according to the NES of 12 MRRGS, and heatmap were drawn to observe the differences in clinical traits and the estimate immune scores between subtypes. P-value < 0.05 was considered statistically significant. In order to facilitate the clinical application of MRRGS classification, we use two machine learning classification algorithms: random forest (RF) and support vector machine (SVM-REF) to screen feature genes between subtypes. The area under (AUC) the receiver operating characteristic curve (ROC) greater than 0.7 indicates that the gene has a good effect on distinguishing different subtypes.

2.3 The survival, mRNAsi levels, immunity, and mutational landscape of subtypes

Survival curves were drawn according to the survival information of different subtypes, and the differences in prognosis between subtypes was evaluated by log-rank test. According to the immune function scores obtained by using TIMER[20], CIBERSORT[21], QuantTIseq[22], xCell[23], MCPcounte[24], EPIC[25] and gene expression levels of each immune checkpoint, Wil-coxon ranksum test was used to analyze the immune infiltration and immune checkpoint dif ferences between subtypes. At the same time, subtypes were used as phenotypes, and the immune gene set h.all.v7.5.1.symbols. gmt in GSEA software was used as the reference gene set for GSEA enrichment analysis. TIDE database (http://tide.dfci.harvard.edu/) [26] will be used to predict the patient's cancer immune response to treatment. And then, the differences of gene mutations in subtypes were counted and survival curves of single gene mutation groups were drawn subsequently. OCLR machine learning algorithm was used to calculate the dryness index (mRNAsi) based on mRNA expression of each sample and visualize the relationship with clinical traits. P-value < 0.05 is considered statistically significant.

2.4 Functional analysis of macrophage gene set

Extract the genes from 12 MRRGS (GFS). The weighted gene co-expression network was then constructed for GFS, and the scale-free network was constructed by determining the optimal soft threshold to conform to the biological model. The modules were clustered, and similar modules were combined at a shear height of 0.25, and calculated the correlation between modules and traits.At the same time, the gene of each module would be made enrichment analysis respectively by using KEGG database (https://www.kegg.jp/kegg/pathway.html) [27] annotation information.

2.5 Screening and discovery of the drug

With normal tissues as control, the different HCC types distinguished by NMF algorithm were used as experimental groups for differences analysis. Filtering and clearing up in accordance with the | logFC | > = 1, FDR = < 0.05, and digging out the subtypes of potential drugs by using CMAP database (https://portals.broadinstitute.org/cmap/). pRRophetic was used to predict the half inhibitory dose (IC50) of commonly used chemotherapeutic drugs in the samples to evaluate the differences in drug sensitivity between subtypes.

2.6 Macrophage interaction at the cellular level

The first 11 principal components (PC) of the processed data were selected. Principal component analysis (PCA) was used to reduce the dimension of feature space. K nearest neighbor classification algorithm was used to determine the optimal resolution of cells according to FindCluster function, and unified manifold approximation and projection (UMAP) method was used to visualize the clustering after reduced dimension. The classification results were annotated preliminarily, meanwhile the annotation was manually modified in combination with the literature (Fig S1). In order to furtherly explore the importance of each cell population in cell interaction network and time series, cell communication analysis and quasitime series analysis were conducted based on annotation results.

2.7 Statistical analysis and software environment used in visualization

The analysis was based on Linux or Windows system. The clusterProfiler[28], GGplot2[29],Seurat[30], SingleR[31], CCA[32], Cowplot[33], GSEABase[34], GSVA[35], pheatmap[36], limma[37], survivalROC[38], survival[39], survminer[40], WGCNA[41], glmnet[42], estimate[43], Maftools [44] ,NM[45], Monocle[46] program package and builtin functions had been used. In addition, CellPhoneDB[48] was used for the calculation of intercellular communicationhether there was batch effect on samples for elimination by CCA algorithm.

3.1 The sgnificant heterogeneity and survival differences between the subtypes differentiated by MRRGS

A total of 287 MRRGS were obtained from the GSEA database. Difference analysis showed that there was a significant difference in the scores of 176 MRRGS between normal and cancer tissues. Univariate COX regression analysis showed that a total of 43 MRRGS were significantly correlated with prognosis. After intersecting the above two results, 27 MRRGS were obtained. The activities of these 27 MRRGS were significantly different between normal and cancer tissues and correlated with prognosis, so they were used for follow-up analysis (Fig. 2A; Table S1). We further screened through Lasso and cross-validation, and obtained 12 MRRGS that are critical to patient survival, they respectively are: GSE5099_MONOCYTE_VS_CLASSICAL_M1_MACROPHAGE_DN, GSE25088_WT_VS_STAT6_KO_MACROPHAGE_DN, GSE11864_UNTREATED_VS_CSF1_PAM3CYS_IN_MAC_DN, GSE25088_WT_VS_STAT6_KO_MACROPHAGE_ROSIGLITAZONE_AND_IL4_STIM_UP, GSE14769_UNSTIM_VS_360MIN_LPS_BMDM_UP, GSE30971_WBP7_HET_VS_KO_MACROPHAGE_4H_LPS_STIM_DN, MANN_RESPONSE_TO_AMIFOSTINE_DN, GSE30971_WBP7_HET_VS_KO_MACROPHAGE_4H_LPS_STIM_UP, GSE11864_CSF1_PAM3CYS_VS_CSF1_IFNG_PAM3CYS_IN_MAC_DN, GSE14769_20MIN_VS_360MIN_LPS_BMDM_UP, BIOCAR-TA_TCAPOPTOSIS_PATHWAY, NEMETH_INFLAMMATORY_RESPONSE_LPS_DN (Fig. 2B-C). The survival curve showed that they were all factors affecting the survival of patients (Fig. 2D; Fig S2; Fig S3). In addition, the prognostic model based on 12 MRRGS has a good effect in predicting the prognosis of HCC, which has been verified in three independent cohorts. The results showed that the prognosis of the low-risk group was significantly better than that of the high-risk group, and the area under the ROC curve was more than 0.7 (Fig. 3). After NMF selection, the samples were divided into two subtypes (Fig. 4A-B; Table S2). PCA dimension reduction and contour coefficient results showed that the clustering effect was credible (Fig. 4C-E). The survival curve showed that cluster 2 (C2) had a better survival than cluster 1 (C1), with a five-year survival rate of more than 60% (Fig. 4D; Fig S4: Fig S5). At the same time, the results showed that most of the normal samples were clustered in C2, indicating that C2 had a MRRGS score level similar to that of the normal samples (Fig. 4F; Fig S6: Fig S7). These results are verified in independent datasets of GEO and ICGC.

3.2 Differences in immune response states and malignant characteristics between C1 and C2

The clinical heatmap showed that C2 had a lower tumor grade but higher immune Estimate score (Fig. 4F; Fig S6: Fig S7) than C1. GSEA enrichment analysis showed that DNA repair, downstream function of E2F transcription factor, G2M checkpoint, mitotic spindle assembly and downstream function of MYC transcription factor were active in C1. While apoptosis pathway, coagulation system related genes, complement system related genes, IL2 signal, IL6 signal, inflammatory response, interferon response, KRAS signal and tumor necrosis factor response were active in C2 (Fig. 4G). These signals also appear in the GSEA results of the external verification set (Fig S8).

Subsequent immune infiltration analysis showed that 59 of the 119 types of immune cells and related function scores were significantly different between C1 and C2. Datasets from ICGC have the same trend (Table S3; Fig S9). Among them in TCGA, 8 kinds of CD8 + T cells function level, 10 kinds of CD4 + T cells function level, 4 kinds of natural killer (NK) cells function level, 4 kinds of dendritic cells function level, 3 kinds of monocytes function level, 7 kinds of macrophage function level, 3 kinds of cancer fibroblast function level, 4 kinds of B cells function level and a kind of neutrophils function level existed significant differences (Fig. 5). Difference analysis showed that there were significant differences in the expression of 42 genes encoding immune checkpoints in C1 and C2 (Fig. 6A). Among them, IDO1, IDO2, PD1, BTLA, CTLA4, LAG3 and TIGIT were upregulated in C2 patients. These are also verified in the ICGC dataset (Fig S10A).

TIDE analysis was consistent with the results of immune infiltration (Fig. 6B-E; Fig S10B-E). C1 patients had higher immune exclusion score and M2 TAM infiltration levels, while C2 patients had higher T cell dysfunction and TIDE scores. The higher TIDE score means the better response to immune checkpoint inhibitors (ICI). In addition, C1 patients had a higher mRNAsi level, which means that these patients had more actively proliferating tumor cell (Fig. 6F; Fig S10F; Table S4). Not only that, C2 is also more responsive to TACE. 73% of C2 patients responded to TACE treatment, while 46% of C1 patients (Fig. 11, P = 0.003).

3.3 TP53 mutation difference is an important factor in tumor heterogeneity

Mutation data showed that TP53 mutations were common in patients with liver cancer, and 1/4 of the samples had mutations(Fig. 7A). Single nucleotide polymorphisms (SNP) are the most common type of mutation in HCC(Fig. 7C). The tumor mutation burden (TMB) of HCC was higher than that of other cancer species, and C1 was higher than C2 (Fig. 7A,B). We further analyzed the difference of gene mutation frequency between C1 and C2 and drew the survival curve of the top 10 genes with the highest mutation frequency. We found that there was a significant difference in the mutation frequency of TP53 between the subtypes of TCGA datasets. 35% C1 patients carry TP53 mutation and only 8% C2 patients carry TP53 mutation (Fig. 7D). The results of ICGC also showed that there was difference in the frequency of TP53 mutation between C1 and C2. 12% C1 patients carried TP53 mutations and only 4% C2 patients carried TP53 mutations (Fig S11D). At the same time, the mutation of TP53 has an impact on the survival of patients in both TCGA and ICGC data sets (Fig. 7E; Fig S11E)

3.4 12 MRRGS affects the heterogeneity of patients by mediating all kinds of HCC-related signals

A total of 1657 genes were extracted from 12 MRRGS, and 1439 genes were obtained after filtering according to the average expression level of 0.5, which could distinguish normal and HCC patients well. And then, the scale-free network was constructed where the optimal β value is 4, and five modules were obtained by merging similar modules (Fig. 8A-C; Table S6). Four of these modules were significantly associated with the occurrence of liver cancer, and five were significantly associated with the patient's Estimate immune score (Fig. 8D). The gray module is an unclustered set of genes, containing only 6 genes and not being significantly associated with the occurrence of HCC or the estimated immune score.

Further functional analysis of genes was made for each of the five modules respectively, and the results showed that the blue module involved the purine metabolism, amino acid biosynthesis, B cell receptor pathway, carbon metabolism, cancer of the PD1 checkpoints, tyrosine metabolism, retinol metabolism, endoplasmic reticulum protein processing, ctype lectin receptors way and T cell receptors pathway; Cyan module involves DNA replication, cell cycle, base excision and repair, carbon metabolism, mismatch repair, P53 signaling pathway, proteasome, cell senescence, pyrimidine metabolism, glutathione metabolism, nucleotide excision and repair, amino acid synthesis and degradation, citric acid cycle. The yellow module covers antigen processing and presentation, natural killer cell mediated cytotoxicity, helper T cells (CD4 + T cells) differentiation, apoptosis, primary immune deficiency, PDL1 expression, PD1 checkpoint in cancer, NF-κB signaling pathway, T cell receptor pathway, TNF signaling pathway, transcription disorders in cancer. The brown module contains with Ras signaling pathway, PDL1 expression, PD1 checkpoint and B cell receptor pathway in cancer. The green module involved the TNF signaling pathway, complement and clotting cascades (Fig. 8E; Table S7).

3.5 Macrophages had the highest intensity of intercellular communication

A total of 1263 cells were obtained after quality control, and the first 11 PC were selected for dimensionality reduction according to standard deviation and P-value of PC. Then, 884 T cells, 230 hepatocytes, 81 macrophages, and 68 B cells were annotated respectively (Fig. 9A). Cell communication analysis showed that the communication intensity between macrophages and other cells was the highest, followed by T cells (Fig. 9B-C). In addition, Pseudotime analysis showed that macrophages and T cells appeared at all stages of the trajectory, which means they had a great influence on HCC patients (Fig. 9D).

3.6 Subtype responses to drugs are significantly different

CMAP analysis helped to screen out treatments for C1 and C2 respectively (Table S8). The results of TCGA dataset showed that histone deacetylase (HDACs) inhibitor MS-275 had the highest therapeutic potential for C1(P = 0), while cyclosynthase inhibitor (COX) NS-398 had the highest therapeutic potential for C2(P = 0.00312). They also appear in the results of cmap analysis of ICGC datasets. Drug sensitivity analysis showed that the half-maximal inhibitory concentration (IC50) of 81 drugs including MS-275 had significant differences between C1 and C2 (Fig. 9E-F; Fig S12; Fig S13; Table S9). In addition, SASH3 ranks second in the results of the two machine learning algorithms. The three cohorts of TCGA, ICGC and GSE54236 showed excellent performance in distinguishing subtypes, with AUC of 0.901, 0.801 and 0.829, respectively (Fig. 10).

3.7 Precision medical tools based on shiny

Finally, based on the shiny(https://shiny.rstudio.com/), we built a bioinformatics tool for 12 MRRGS. Users only need to input the txt format file of the expression matrix to get the subtype and Risk Score of each sample, which will hopefully become a part of HCC precision medicine(Fig. 12). This tool can be accessed through the following address: http://117.62.203.72:3838/MNMF/.

Biological functions are regulated by the set of upstream coding genes of proteins. The activity of the biological function in a certain sample could be determined by evaluating the overall expression level of the gene sets corresponding to each biological function. Based on this, we calculated the macrophage-related function score. TAM plays an important role in shaping TME. In this study, the samples were divided into C1 and C2 by evaluating functional differences related to macrophages. Lasso regression screening of 12 MRRGS has a significant effect on the survival, indicating that macrophages play an important role in the prognosis of HCC. Notably, Almost all the paracancerous tissues contained in TCGA and normal tissues in ICGC are clustered in C2, which means that C2 has a level of macrophage response similar to that of normal tissue. Therefore, we can use normal tissue as a reference to distinguish C2 subtypes in HCC, which can effectively simplify the clinical application. Meanwhile, C2 also had a better prognosis, which means that normal macrophage function is important for patient survival. In addition, the results showed that C2 had higher immune levels,, which may be related to the shaping of TME by macrophage function. Previous studies have also shown that TAM in HCC inhibits the overall immune level of TME by expressing cell surface proteins or releasing soluble factors related to immunosuppressive function, such as arginase 1, IL-10, TGF-β [49–51]. TAM could recruit regulatory T cells to mediate immune depletion of CD8 + T cells in TME [49].

Functional analysis showed that the main functions of C1 were mainly concentrated on cell cycle, cell proliferation and protein genesis, which were considered to be more likely to promote the occurrence and deterioration of HCC. Recent studies have shown[52] that high levels of DNA repair are associated with proliferation of HCC cell, poor prognosis, heterogeneity and mutation. E2F is the encoding gene of a set of transcription factors family, related to cell cycle and DNA synthesis, and has been confirmed to be involved in regulating the proliferation and progression of HCC in a number of researches[53–56]. G2M checkpoint ensures adequate repair of DNA during G2M transformation and plays a key role in the proliferation, apoptosis and invasion of HCC [57]. Myc is a transcription factor encoded by protooncogene, which is closely related to the occurrence and deterioration of liver cancer and drug resistance[58]. UPR and Myc have a synergistic effect on cancer and could jointly promote tumor progression[59]. These evidence suggest that abnormal cell proliferation and metabolism are important factors leading to poor prognosis of C1. On the other hand, C2, has relatively stronger level of immunity and apoptosis, such as tumor necrosis factor (TNF) signal and interferon (INF) signal, which have been confirmed to be related to the formation of anti-HCC tumor immunity in literature studies[60–62]. It is worth noting that C2 is accompanied by activation of HCC signaling and immune negative feedback related functions, including IL2/STAT5 signaling, cell proliferation, angiogenesis and tumor migration. Among them, IL2/STAT5 is an important maintenance signal of regulatory T cells (Tregs). Cytokines participate in the depletion of T cells in TME and promote the immune escape of tumor cells [63]. This suggests that the reason why C2 has a better survival than C1 is its strong immune level. As a diagnosed HCC, it also has some malignant characteristics of cancer, and immunosuppression caused by high-level immune response may be an important factor in the occurrence of this subtypes.Therefore, for C1 and C2 patients, using different therapies respectively will have a better effect. The former is more suitable for targeting TAM reprogramming and inhibiting cell proliferation, while the latter may be more suitable for immune checkpoint therapy and anti-proliferative metastasis drugs.

Immune infiltration analysis showed that C2 has stronger infiltration levels of CD8 + T cells, CD4 + T cells, NK cells and M2 macrophages compared with C1, which is consistent with our previous analysis results and more strongly proved that TAM had a strong shaping ability to the overall immunity level of TME. At the same time, the gene expression of 42 immune checkpoints were significantly different between C1 and C2. The different expression of these immune checkpoints will have a great impact on the therapeutic efficacy of patients, which is one of the important reasons for tumor heterogeneity and drug resistance. Interestingly, the inhibitory checkpoint molecules such as IDO1, IDO2, PD1, BTLA, CTLA4, LAG3 and TIGIT are upregulated in C2 patients, which are all markers of T cell exhaustion. The highly expressed immunosuppressive components suggest that C2 may have a good response to immune checkpoint inhibitors. Previous studies [64] have shown that some cancer patients have high levels of CD8 + T cell infiltration, but these infiltrating cytotoxic T cells (CTL) do not seem to give full play to their anticancer effects and are in a states of dysfunction. At the same time, high infiltration level of CD8 + T cells can promote the activity of immunosuppressive components such as PD-L1, IDO and Tregs in TME [65], which is strikingly consistent with the phenotype of C2 patients. TIDE analysis confirmed our conclusion that C2 patients had a higher T cell dysfunction score and ICI response, while C1 patients had a higher level of TAM infiltration and tumor immune exclusion score. The higher TAM infiltration level is an important shaping factor of immune rejection in C1 patients, which leads to lower TME immune cell infiltration level. Therefore, the design of personalized treatment for HCC patients with different subtypes of macrophage function will help to save drug resources and improve the therapeutic effect. For the C2 patients with high PD1 expression levels, the use of ICI such as PD1/PDL1 seems to reactivate the saturated functional impaired CTL in their TME and improve the patient's condition. For C1 patients with low PD1 expression level, ICI seems to be difficult to achieve good efficacy. It may be a better choice to use targeted TAM to regulate immune exclusion and inhibit tumor cell proliferation. Our research also showed that two subtypes with different macrophages response have appeared in the HCC patients. The first subtype has similar macrophage function states to normal and often behave a higher immune cell infiltration levels, but with a certain degree of immunosuppression, while the other subtype displays higher tumor proliferation characteristics and lower immune cell infiltration levels. The former subtype is pretty similar to the "hot tumor" proposed in previous studies[66] and is sensitive to ICI treatment. The latter subtype seems to be similar to the "cold tumor". At the same time, these results suggest that constructing macrophage classification model has great clinical value in evaluating the response potential of patients receiving immunotherapy. Moreover, the classifier constructed by macrophages can also be used to evaluate the response of HCC to TACE therapy. Due to the large number of genes involved in the 12 MRRGS, in order to facilitate clinical use, machine learning method has mined the characteristic gene SASH3 between C1 and C2. Evaluated by ROC in three cohorts, it has a very reliable diagnostic performance. At present, there are few studies on this gene, and a small number of studies have shown that SASH3 is very important for the survival and proliferation of T cells, which is related to immune deficiency and disorder in patients. These evidences show that SASH3 has further research and application value.

The mutation data showed that the TMB of C1 was higher than that of C2, which further showed a significant heterogeneity between the two subtypes of patients. Studies have shown that somatic mutations are important factors driving tumorigenesis[67]. This study found that the mutation level of TP53 in C1 was significantly increased, and had a significant impact on the prognosis of patients. Previous studies have shown that TP53 mutation exists in a large number of HCC, and the p53 protein encoded downstream is a key molecule in the p53 signal pathway, which plays an important role in regulating cell proliferation and immune level, and is an important molecular marker and therapeutic target. The higher TP53 mutation level of C1 than C2 is an important upstream basis affecting the difference of macrophage function and a key factor leading to different immune types of patients. At the same time, it shows that C1 patients with high TP53 mutation are more suitable to be treated with related molecular therapy. In addition, C1 patients generally had high mRNAsi scores, which confirmed that C1 patients had a more active states of cell proliferation.

Most of the genes extracted from MRRGS were confirmed to be related to the occurrence of tumors and the level of immunity through WGCNA analysis. Subsequent KEGG enrichment analysis of five gene modules revealed that these genes involved PD1-PDL1 signal, NF-κB signal, Ras signal pathway, TNF signal pathway, IL17 signal, viral infection pathway, reaction to chemical substances, immune cell-associated receptor pathway, cell cycle and proliferation, amino acid synthesis and decomposition, protein processing, DNA shearing and repair, substance metabolism, complement system, and coagulation cascade reactions and C-type lectin receptor pathways etc. are mutually confirmed with our previous analysis, and most of them have been mentioned in the literatures related to the progression of liver cancer[68–72]. These function differences between C1 and C2 are important components of tumor heterogeneity. Drug design by targeting these pathways and their core molecule will have a better chance to improve patient survival. Single-cell analysis further revealed that the recruited TAM plays a crucial role in the shaping of TME. The identification of patients with TAM response states will be of great benefit to the choice of drugs and prognosis of HCC.

Finally, we identified potential therapeutic for each subtype based on the patient expression profile. And we found that HDACs inhibitors seem to be more suitable for C1 patients, while COX inhibitors may be more effective for C2 patients. The result will help us better understand the therapeutic mechanism of HCC and be used for subsequent drug combination and exploration. HDACs inhibitors in a number of clinical studies were used for the treatment of various cancers [73–76]. Studies have shown that inhibition of HDACS leads to increased acetylation of relevant tumor suppressive genomic proteins and non-histones regulated by chromatin modification, resulting in transcriptional activation and recovery of downstream cell proliferation inhibition and apoptosis [77, 78]. At the same time, HDACs inhibitors have been shown to target to reshape the epigenetic state of TAM in TME, which may have a good therapeutic effect on C1 patients with high TAM infiltration [79, 80]. COX inhibitors can not only inhibit the growth, invasion and migration of tumor cells [81], but also effectively inhibit the expression of IDO and the accumulation of Tregs in TME [82]. COX inhibitors can be used in cancer-related inflammation with overactive IDO1 to help regulate the immune response in TME [82]. A study shows that the combination of HADC inhibitors and COX inhibitors has no obvious side effects on patients, but has synergistic anticancer effect [83]. These results suggest that a combination of these drug may have a therapeutic effect on most patients. At the same time, drug sensitivity analysis showed that there were significant differences in the response of HCC with different subtypes to most chemotherapeutic drugs, which further emphasized the importance of accurate medical treatment.

TAM plays an important role in the shaping of TME and the formation of tumor heterogeneity. Our research shows that HCC can be divided into two subtypes according to the functional response states of macrophages. Among them, patients with high levels of immune cell infiltration are usually accompanied with lower TMB, less M2 TAM infiltration and stronger immunosuppression, and the potential negative immune feedback mechanism is the primary driving force of tumor deterioration. However, patients with low levels of immune cell infiltration are usually accompanied with higher TMB, more M2 TAM immune infiltration and stronger cell proliferation, and immune exclusion and extensive mutations are the primary driving forces of tumor occurrence and deterioration in this subtype. For the former, it seems to be necessary to curb the aggravation of negative immune regulation, and targeted suppression of high expression PD1, IDO1, IDO2, BTLA, CTLA4, LAG3 and TIGIT immune checkpoints is a better choice. For the latter, it seems to be necessary to improve the proportion of immune cell infiltration and inhibit the active proliferation of tumor cells. Immune-related pathways (e.g., PD1-PDL1 signaling), mutations, cell proliferation, viral infection, various metabolic responses, chemicals (e.g., alcohol) are important upstream bases for the formation of patient heterogeneity. Through the classification of MRRGS, we can well predict the response of patients to immunotherapy, TACE and various drugs, which will save resources more effectively and help to achieve accurate treatment. It is noteworthy that our study further emphasizes that the TME shaped by macrophages is an important factor in the formation of tumor heterogeneity and an important research breakthrough for tumor immunotherapy in the future. In addition, two potential therapeutics like COX inhibitors and HDACs inhibitors are worthy of further clinical trials.

Nonetheless, this research still has some limitations. First, although our results used more than 1000 HCC samples and were validated in two independent cohorts, this needs to be further incorporated into a larger clinical cohort for study. Secondly, other epigenetic methods such as non-coding RNA may have an impact on the production of subtypes, which is worthy of further analysis. Finally, the mechanism of 12 MRRGS and SASH3 on tumor heterogeneity is also worthy of further exploration. We believe that macrophages play an important role in the formation of HCC tumor heterogeneity, and the evaluation of HCC immunity through macrophage function will provide great help for the realization of accurate medicine.

In summary, this study identified two immune subtypes of hepatocellular carcinoma patients by defining macrophages activity classification indicators. 12 MRRGS can help predict the ability of patients to respond to various types of treatment, which will be beneficial to the accurate treatment and management of clinical patients with liver cancer. More importantly, through the bioinformatics tools built in this study, researchers can quickly use classifiers based on 12 MRRGS, which greatly reduces the threshold for use. On this basis, we explored the functional signal of TAM promoting the formation of TME heterogeneity, which deepened the role of TAM in the carcinogenesis and progression of hepatocellular carcinoma.

Data availability

The data used to support the results are available at the TCGA (https://portal.gdc.cancer.gov/), ICGC(https://dcc.icgc.org/), GEO (https://www.ncbi.nlm.nih.gov/geo/). Meanwhile, you can also get it on Github: https://github.com/mumdark/Macrophage-cluster-R-script , which also contains the code of macrophage functional gene set classification. The bioinformatics tools constructed in this study can be accessed through the following address: http://117.62.203.72:3838/MNMF/.

Ethics approval and consent to participate

None Declared.

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Author Contributions

The Conception and design: Defu Liu. Collection and assembly of data: Zhengjun Wang, Xiaonan Huang and Zhanfeng Dai. Data analysis: Defu Liu, Zhanfeng Dai, Jing Yang. Construction of bioinformatics tool: Defu Liu, Jing Yang. Reviewing and editing: Defu Liu, Jing Yang, Caoyu Xie. Project administration: Defu Liu. Commented and approved the text: All the authors. All authors made substantial contributions to the conception or design of the study, or to the collection, analysis, or interpretation of study data.

Acknowledgments

This study was funded by the Innovation and Entrepreneurship Training Program of GZUCM (202110572005) to Defu Liu.

Chon YE, Jeong SW, Jun DW (2021) Hepatocellular carcinoma statistics in South Korea. Clin Mol Hepatol 27:512
Beal EW, Tumin D, Kabir A, Moris D, Zhang X, Chakedis J, Washburn K, Black S, Schmidt CM, Pawlik TM (2017) Trends in the mortality of hepatocellular carcinoma in the United States. J Gastrointest Surg 21:2033–2038
Ferlay J, Colombet M, Soerjomataram I, Parkin DM, Piñeros M, Znaor A, Bray F (2021) Cancer statistics for the year 2020: An overview. Int J Cancer
Stanta G, Bonin S (2018) Overview on clinical relevance of intra-tumor heterogeneity. Front Med 5:85
Hass R, von der Ohe J, Ungefroren H (2020) Impact of the tumor microenvironment on tumor heterogeneity and consequences for cancer cell plasticity and stemness. Cancers 12:3716
Anwanwan D, Singh SK, Singh S, Saikam V, Singh R Challenges in liver cancer and possible treatment approaches. Biochimica et Biophysica Acta (BBA)-Reviews on Cancer 2020, 1873, 188314
Shergold AL, Millar R, Nibbs RJ (2019) Understanding and overcoming the resistance of cancer to PD-1/PD-L1 blockade. Pharmacol Res 145:104258
Biswas SK, Allavena P, Mantovani A (2013) Tumor-associated macrophages: functional diversity, clinical significance, and open questions. Semin Immunopathol 585–600
Chew V, Lai L, Pan L, Lim CJ, Li J, Ong R, Chua C, Leong JY, Lim KH, Toh HC (2017) Delineation of an immunosuppressive gradient in hepatocellular carcinoma using high-dimensional proteomic and transcriptomic analyses. Proceedings of the National Academy of Sciences 114, E5900-E5909
Thorsson V, Gibbs DL, Brown SD, Wolf D, Bortone DS, Yang TO, Porta-Pardo E, Gao GF, Plaisier CL, Eddy JA (2018) The immune landscape of cancer. Immunity 48:812–830
Huang Y, Ge W, Zhou J, Gao B, Qian X, Wang W (2021) The role of tumor associated macrophages in hepatocellular carcinoma. J Cancer 12:1284
Zhou K, Cheng T, Zhan J, Peng X, Zhang Y, Wen J, Chen X, Ying M (2020) Targeting tumor–associated macrophages in the tumor microenvironment. Oncol Lett 20:1
Chanmee T, Ontong P, Konno K, Itano N (2014) Tumor-associated macrophages as major players in the tumor microenvironment. Cancers 6:1670–1690
Cassetta L, Kitamura T (2018) Targeting tumor-associated macrophages as a potential strategy to enhance the response to immune checkpoint inhibitors. Front cell Dev biology 6:38
Binenbaum Y, Fridman E, Yaari Z, Milman N, Schroeder A, David GB, Shlomi T, Gil Z (2018) Transfer of miRNA in macrophage-derived exosomes induces drug resistance in pancreatic adenocarcinoma. Cancer Res 78:5287–5299
Dong D, Zhang G, Yang J, Zhao B, Wang S, Wang L, Zhang G, Shang P (2019) The role of iron metabolism in cancer therapy focusing on tumor-associated macrophages. J Cell Physiol 234:8028–8039
Tomczak K, Czerwińska P, Wiznerowicz M (2015) The Cancer Genome Atlas (TCGA): an immeasurable source of knowledge. Contemp Oncol 19:A68
Barrett T, Suzek TO, Troup DB, Wilhite SE, Ngau W, Ledoux P, Rudnev D, Lash AE, Fujibuchi W, Edgar R (2005) NCBI GEO: mining millions of expression profiles—database and tools. Nucleic Acids Res 33:D562–D566
Liberzon A, Subramanian A, Pinchback R, Thorvaldsdóttir H, Tamayo P, Mesirov JP (2011) Molecular signatures database (MSigDB) 3.0. Bioinformatics 27:1739–1740
Li B, Severson E, Pignon J, Zhao H, Li T, Novak J, Jiang P, Shen H, Aster JC, Rodig S et al (2016) Comprehensive analyses of tumor immunity: implications for cancer immunotherapy. Genome Biol 17:174. 10.1186/s13059-016-1028-7
Finotello F, Mayer C, Plattner C, Laschober G, Rieder D, Hackl H, Krogsdam A, Loncova Z, Posch W, Wilflingseder D et al (2019) Molecular and pharmacological modulators of the tumor immune contexture revealed by deconvolution of RNA-seq data. Genome Med 11:34. 10.1186/s13073-019-0638-6
Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, Hoang CD, Diehn M, Alizadeh AA (2015) Robust enumeration of cell subsets from tissue expression profiles. Nat Methods 12:453–457
Becht E, Giraldo NA, Lacroix L, Buttard B, Elarouci N, Petitprez F, Selves J, Laurent-Puig P, Sautès-Fridman C, Fridman WH et al (2016) Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression. Genome Biol 17:218. 10.1186/s13059-016-1070-5
Aran D, Hu Z, Butte AJ (2017) xCell: digitally portraying the tissue cellular heterogeneity landscape. Genome Biol 18:220. 10.1186/s13059-017-1349-1
Racle J, de Jonge K, Baumgaertner P, Speiser DE, Gfeller D (2017) Simultaneous enumeration of cancer and immune cell types from bulk tumor gene expression data. Elife 6:e26476
Jiang P, Gu S, Pan D, Fu J, Sahu A, Hu X, Li Z, Traugh N, Bu X, Li B (2018) Signatures of T cell dysfunction and exclusion predict cancer immunotherapy response. Nat Med 24:1550–1558
Kanehisa M, Furumichi M, Tanabe M, Sato Y, Morishima K (2017) KEGG: new perspectives on genomes, pathways, diseases and drugs. Nucleic Acids Res 45:D353–D361
Yu G (2018) clusterProfiler: universal enrichment tool for functional and comparative study. BioRxiv 256784
Wickham H (2011) ggplot2. Wiley Interdisciplinary Reviews: Computational Statistics 3, 180–185
Satija R, Farrell JA, Gennert D, Schier AF, Regev A (2015) Spatial reconstruction of single-cell gene expression data. Nat Biotechnol 33:495–502
Aran D, Looney AP, Liu L, Wu E, Fong V, Hsu A, Chak S, Naikawadi RP, Wolters PJ, Abate AR (2019) Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage. Nat Immunol 20:163–172
González I, Déjean S, Martin P, Baccini ACCA (2008) An R package to extend canonical correlation analysis. J Stat Softw 23:1–14
Wilke CO, Wickham H, Wilke MCO Package ‘cowplot’. Streamlined Plot Theme and Plot Annotations for ‘ggplot2 2019
Morgan M, Falcon S, Gentleman R, Maintainer MBP, AnnotationDbi I, biocViews GeneExpression G, GraphAndNetwork G (2013) Package ‘GSEABase’.
Hänzelmann S, Castelo R, Guinney J (2013) GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics 14:1–15
Kolde R, Kolde MR Package ‘pheatmap’. R package 2015, 1, 790
Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK (2015) limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43:e47
Heagerty PJ, Saha-Chaudhuri P, Saha-Chaudhuri MP (2013) Package ‘survivalROC’. GitHub, San Francisco
Therneau TM, Lumley T (2014) Package ‘survival’. Survival Anal Published CRAN 2:119
Kassambara A, Kosinski M, Biecek P, Fabian S (2017) Package ‘survminer’. Drawing Survival Curves using ‘ggplot2’(R package version 03 1)
Langfelder P, Horvath S (2008) WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 9:1–13
Friedman J, Hastie T, Tibshirani R, Narasimhan B, Tay K, Simon N, Qian J (2021) Package ‘glmnet’. CRAN R Repositary
Yoshihara K, Shahmoradgoli M, Martínez E, Vegesna R, Kim H, Torres-Garcia W, Treviño V, Shen H, Laird PW, Levine DA et al (2013) Inferring tumour purity and stromal and immune cell admixture from expression data. Nat Commun 4:2612. 10.1038/ncomms3612
Mayakonda A, Lin D, Assenov Y, Plass C, Koeffler HP (2018) Maftools: efficient and comprehensive analysis of somatic variants in cancer. Genome Res 28:1747–1756. 10.1101/gr.239244.118
Gaujoux R, Seoighe C (2010) A flexible R package for nonnegative matrix factorization. BMC Bioinformatics 11:367. 10.1186/1471-2105-11-367
Qiu X, Mao Q, Tang Y, Wang L, Chawla R, Pliner HA, Trapnell C (2017) Reversed graph embedding resolves complex single-cell trajectories. Nat Methods 14:979–982. 10.1038/nmeth.4402
Tang Y, Li M, Wang J, Pan Y, Wu F (2015) CytoNCA: a cytoscape plugin for centrality analysis and evaluation of protein interaction networks. Bio Syst 127:67–72. 10.1016/j.biosystems.2014.11.005
Efremova M, Vento-Tormo M, Teichmann SA, Vento-Tormo R (2020) CellPhoneDB: inferring cell-cell communication from combined expression of multi-subunit ligand-receptor complexes. Nat Protoc 15:1484–1506. 10.1038/s41596-020-0292-x
Wu Q, Zhou W, Yin S, Zhou Y, Chen T, Qian J, Su R, Hong L, Lu H, Zhang F (2019) Blocking triggering receptor expressed on myeloid cells-1‐positive tumor‐associated macrophages induced by hypoxia reverses immunosuppression and anti‐programmed cell death ligand 1 resistance in liver cancer. Hepatology 70:198–214
Yang J, Zhang J, Wang H, Wang G, Hu Q, Zheng Q (2012) Hepatocellular carcinoma and macrophage interaction induced tumor immunosuppression via Treg requires TLR4 signaling. World J Gastroenterology: WJG 18:2938
Xiang X, Wang J, Lu D, Xu X (2021) Targeting tumor-associated macrophages to synergize tumor immunotherapy. Signal Transduct Target Therapy 6:1–12
Oshi M, Kim TH, Tokumaru Y, Yan L, Matsuyama R, Endo I, Cherkassky L, Takabe K (2021) Enhanced DNA repair pathway is associated with cell proliferation and worse survival in hepatocellular carcinoma (HCC). Cancers 13:323
Ye W, Lu H, Li J, Chen G, He R, Wu H, Zhou X, Rong M, Yang L, He W (2021) Clinical Implication of E2F Transcription Factor 1 in Hepatocellular Carcinoma Tissues. Cancer Biotherapy & Radiopharmaceuticals
Lü Y, Zhang J, Li L, Li S, Yang Z (2020) Carcinogenesis effects of E2F transcription factor 8 (E2F8) in hepatocellular carcinoma outcomes: an integrated bioinformatic report. Bioscience Rep 40:R20193212
Shao-xin DXG, Li-rong YSH, Fang Z (2018) MicroRNA-539 inhibiting the occurrence and development of hepatocellular carcinoma by targeting E2F transcription factor 3. Acta Anat Sinica 49:469
Liu J, Xia L, Wang S, Cai X, Wu X, Zou C, Shan B, Luo M, Wang D (2021) E2F4 Promotes the Proliferation of Hepatocellular Carcinoma Cells through Upregulation of CDCA3. J Cancer 12:5173
Liu L, Zhu J, Yu X, Zhu H, Shi X, Bilegsaikhan E, Guo H, Wu J, Shen X (2019) UBE2T promotes proliferation via G2/M checkpoint in hepatocellular carcinoma. Cancer Manag Res 11:8359
Wang J, Ma L, Liang Y, Yang X, Wei S, Peng H, Qiu S, Lu X, Zhu Y, Wang B (2021) RCN1 induces sorafenib resistance and malignancy in hepatocellular carcinoma by activating c-MYC signaling via the IRE1α–XBP1s pathway. Cell Death Discovery 7:1–15
Zhang T, Li N, Sun C, Jin Y, Sheng X (2020) MYC and the unfolded protein response in cancer: synthetic lethal partners in crime? Embo Mol Med 12:e11845
Tan W, Luo X, Li W, Zhong J, Cao J, Zhu S, Chen X, Zhou R, Shang C, Chen Y (2019) TNF-α is a potential therapeutic target to overcome sorafenib resistance in hepatocellular carcinoma. Ebiomedicine 40:446–456
Yan Y, Zheng L, Du Q, Yazdani H, Dong K, Guo Y, Geller DA (2021) Interferon regulatory factor 1 (IRF-1) activates anti-tumor immunity via CXCL10/CXCR3 axis in hepatocellular carcinoma (HCC). Cancer Lett 506:95–106
Chen H, Tang C, Wu F, Lu L, Li S, Li Z, Lu W, Wu Z (2021) IL-2 Increases Tams-Derived Exosomal Mir-375 to Ameliorate Hepatocellular Carcinoma Development and Progression.
Fan MY, Low JS, Tanimine N, Finn KK, Priyadharshini B, Germana SK, Kaech SM, Turka LA (2018) Differential Roles of IL-2 Signaling in Developing versus Mature Tregs. Cell Rep 25:1204–1213. 10.1016/j.celrep.2018.10.002
Joyce JA, Fearon DT (2015) T cell exclusion, immune privilege, and the tumor microenvironment. Science 348:74–80
Spranger S, Spaapen RM, Zha Y, Williams J, Meng Y, Ha TT, Gajewski TF (2013) Up-regulation of PD-L1, IDO, and Tregs in the melanoma tumor microenvironment is driven by CD8 + T cells. Sci Transl Med 5:116r–200r
Chen DS, Mellman I (2017) Elements of cancer immunity and the cancer–immune set point. Nature 541:321–330
Ding X, Zhu Q, Zhang S, Guan L, Li T, Zhang L, Wang S, Ren W, Chen X, Zhao J (2017) Precision medicine for hepatocellular carcinoma: driver mutations and targeted therapy. Oncotarget 8:55715
Luo L, Chen L, Ke K, Zhao B, Wang L, Zhang C, Wang F, Liao N, Zheng X, Liu X (2020) High expression levels of CLEC4M indicate poor prognosis in patients with hepatocellular carcinoma. Oncol Lett 19:1711–1720
De Matteis S, Ragusa A, Marisi G, De Domenico S, Casadei Gardini A, Bonafè M, Giudetti AM (2018) Aberrant metabolism in hepatocellular carcinoma provides diagnostic and therapeutic opportunities. Oxid Med Cell Longev 2018
Li WT, Zou AE, Honda CO, Zheng H, Wang XQ, Kisseleva T, Chang EY, Ongkeko WM (2019) Etiology-specific analysis of hepatocellular carcinoma transcriptome reveals genetic dysregulation in pathways implicated in immunotherapy efficacy. Cancers 11:1273
Berndt N, Eckstein J, Heucke N, Gajowski R, Stockmann M, Meierhofer D, Holzhütter H (2019) Characterization of lipid and lipid droplet metabolism in human HCC. Cells-Basel 8:512
Yan H, Kamiya T, Suabjakyong P, Tsuji NM (2015) Targeting C-type lectin receptors for cancer immunity. Front Immunol 6:408
Baretti M, Durham JN, Walker R, Mitcheltree A, Christmas B, Cope L, Jaffee EM, Azad NS (2018) Entinostat in combination with nivolumab for patients with advanced cholangiocarcinoma and pancreatic adenocarcinoma.
Brown TJ, Karasic TB, Schneider CJ, Teitelbaum UR, Reiss KA, Mitchell TC, Massa RC, O'Hara MH, DiCicco L (2021) ; Garcia-Marcano, L. Phase I trial of regorafenib, hydroxychloroquine, and entinostat in metastatic colorectal cancer.
Trapani D, Esposito A, Criscitiello C, Mazzarella L, Locatelli M, Minchella I, Minucci S, Curigliano G (2017) Entinostat for the treatment of breast cancer. Expert Opin Inv Drug 26:965–971
Lin J, Elkon J, Ricart B, Palmer E, Zevallos Delgado C, Noonepalle S, Burgess B, Siegel R, Ma Y, Villagra A Phase I Study of Entinostat in Combination with Enzalutamide for Treatment of Patients with Metastatic Castration-Resistant Prostate Cancer. The Oncologist 2021.
Knipstein J, Gore L (2011) Entinostat for treatment of solid tumors and hematologic malignancies. Expert Opin Inv Drug 20:1455–1467
Atadja PW (2011) HDAC inhibitors and cancer therapy. Epigenetics and Disease 175–195
DeNardo DG, Ruffell B (2019) Macrophages as regulators of tumour immunity and immunotherapy. Nat Rev Immunol 19:369–382
Guerriero JL, Sotayo A, Ponichtera HE, Castrillon JA, Pourzia AL, Schad S, Johnson SF, Carrasco RD, Lazo S, Bronson RT (2017) Class IIa HDAC inhibition reduces breast tumours and metastases through anti-tumour macrophages. Nature 543:428–432
Yao M, Lam E, Kelly C, Zhou W, Wolfe M (2004) Cyclooxygenase-2 selective inhibition with NS-398 suppresses proliferation and invasiveness and delays liver metastasis in colorectal cancer. Brit J Cancer 90:712–719
Cesario A, Rocca B, Rutella S (2011) The interplay between indoleamine 2, 3-dioxygenase 1 (IDO1) and cyclooxygenase (COX)-2 in chronic inflammation and cancer. Curr Med Chem 18:2263–2271
Sonnemann J, Hüls I, Sigler M, Palani CD, Hong LTT, Völker U, Kroemer HK, Beck JF (2008) Histone deacety lase inhibitors and aspirin interact synergistically to induce cell death in ovarian cancer cells. Oncol Rep 20:219–224
Delmonte OM et al (2021) SASH3 variants cause a novel form of X-linked combined immunodeficiency with immune dysregulation." Blood. J Am Soc Hematol 138(12):1019–1033

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Development and validation of a hepatocellular carcinoma classifier based on macrophage-related gene set for rapid evaluation of patient response types and prognosis

Status:

Version 1

Abstract

Figures

Introduction

Materials and Methods

Result

3.1 The sgnificant heterogeneity and survival differences between the subtypes differentiated by MRRGS

3.2 Differences in immune response states and malignant characteristics between C1 and C2

3.3 TP53 mutation difference is an important factor in tumor heterogeneity

3.4 12 MRRGS affects the heterogeneity of patients by mediating all kinds of HCC-related signals

3.5 Macrophages had the highest intensity of intercellular communication

3.7 Precision medical tools based on shiny

Discussion

Conclusions

Declarations

Data availability

Ethics approval and consent to participate

Conflict of Interest Statement

Author Contributions

Acknowledgments

References

Additional Declarations

Supplementary Files

Status:

Version 1