Aberrantly Expressed MicroRNAs in Cancer-Associated Fibroblasts and Their Target Oncogenic Signatures in Hepatocellular Carcinoma

Cancer-associated fibroblasts (CAFs) contribute to tumor progression, and microRNAs (miRs) play an important role in regulating the tumor-promoting properties of CAFs. The objectives of this study were to clarify the specific miR expression profile in CAFs of hepatocellular carcinoma (HCC) and identify its target gene signatures. Small-RNA-sequencing data were generated from nine pairs of CAFs and para-cancer fibroblasts isolated from human HCC and para-tumor tissues, respectively. Bioinformatic analyses were performed to identify the HCC-CAF-specific miR expression profile and the target gene signatures of the deregulated miRs in CAFs. Clinical and immunological implications of the target gene signatures were evaluated in The Cancer Genome Atlas Liver Hepatocellular Carcinoma (TCGA_LIHC) database using Cox regression and TIMER analysis. The expressions of hsa-miR-101-3p and hsa-miR-490-3p were significantly downregulated in HCC-CAFs. Their expression in HCC tissue gradually decreased as HCC stage progressed in the clinical staging analysis. Bioinformatic network analysis using miRWalks, miRDB, and miRTarBase databases pointed to TGFBR1 as a common target gene of hsa-miR-101-3p and hsa-miR-490-3p. TGFBR1 expression was negatively correlated with miR-101-3p and miR-490-3p expression in HCC tissues and was also decreased by ectopic miR-101-3p and miR-490-3p expression. HCC patients with TGFBR1 overexpression and downregulated hsa-miR-101-3p and hsa-miR-490-3p demonstrated a significantly poorer prognosis in TCGA_LIHC. TGFBR1 expression was positively correlated with the infiltration of myeloid-derived suppressor cells, regulatory T cells, and M2 macrophages in a TIMER analysis. In conclusion, hsa-miR-101-3p and hsa-miR-490-3p were substantially downregulated miRs in CAFs of HCC, and their common target gene was TGFBR1. The downregulation of hsa-miR-101-3p and hsa-miR-490-3p, as well as high TGFBR1 expression, was associated with poor clinical outcome in HCC patients. In addition, TGFBR1 expression was correlated with the infiltration of immunosuppressive immune cells.


Introduction
Hepatocellular carcinoma (HCC) is the fifth most common cancer and fourth leading cause of cancer-related mortality worldwide [1]. Although significant developments in therapeutic strategies have been made in the last 20 years, the long-term survival of HCC patients remains unsatisfactory. The development of novel therapeutic strategies based on an in-depth understanding of the molecular features of HCC is required to improve the prognosis of HCC patients.
The tumor microenvironment (TME) is a highly complex and dynamic ecosystem consisting of tumor cells, cancer-associated fibroblasts (CAFs), and a variety of immune cells [2]. In recent years, most studies examining the TME have focused on better understanding the role of TME components to improve immunotherapy efficacy [3]. CAFs make up a major cell type in tumor stroma that produce an extracellular matrix [4]. CAFs contribute to tumor growth, angiogenesis, invasiveness, and metastasis, not only by directly regulating the aggressiveness of malignant cells, but also by indirectly promoting an immunosuppressive TME [5].
MicroRNAs (miRs) are small, non-coding RNAs (usually~22 nucleotides) that play a key role in RNA silencing and regulating target gene expression [6]. In participating in tumor cell proliferation, differentiation, and metastasis, miRs can act as tumor suppressors or oncogenes by negatively regulating the expression of target mRNAs in nearly all cancer types, including HCC [7,8]. Accumulating evidence suggests that miRs are key players in regulating the tumor-promoting properties of CAFs; however, the role of miRs in CAFs of HCC (HCC-CAFs) remains poorly elucidated [9,10].
In the present study, to better understand the molecular mechanisms of HCC-CAFs, aberrantly expressed miR signatures in HCC-CAFs were evaluated using miR-sequencing data from primary cultured HCC-CAFs, para-cancer fibroblasts (PAFs), and normal fibroblasts (NFs). In addition, target gene signatures of the aberrantly expressed miRs in HCC-CAFs, as well as the clinical and immunological implications of these target genes, were evaluated using bioinformatic analyses.

Identification of Differentially Expressed miRs in HCC-CAFs
CAFs were isolated from HCC tissues, and PAFs were isolated from paired non-tumor tissues adjacent to the HCC (Figure 1a, middle and right panel). NFs were isolated and cultured from a normal liver tissue, acquired from a patient who did not have any chronic liver disease but had undergone surgical resection for a gradually growing benign tumor ( Figure 1a, left panel). A flowchart of this study's protocol is available in Figure 1b. The differential expression patterns of miRs in NFs, PAFs, and CAFs are presented in Figure 1c. While certain miRs were upregulated in CAFs compared to NFs and PAFs (Figure 1c, right), other miRs were relatively downregulated in CAFs (Figure 1c, left). Figure 1d shows a heatmap of 31 miRs that were significantly differently expressed in HCC-CAFs compared to NFs and PAFs, including 17 downregulated miRs (left panel) and 14 upregulated miRs (right panel). Integrative analyses were performed to identify the aberrantly expressed miRs showing clinical significance in HCC patients. Figure 1e displays a Venn diagram of differently expressed miRs between HCC-CAFs and The Cancer Genome Atlas Liver Hepatocellular Carcinoma (TCGA_LIHC) dataset. Among the 31 differently expressed miRs in HCC-CAFs, hsa-miR-101-3p and hsa-miR-490-3p were significantly downregulated in CAFs compared to PAFs, and their expression was also downregulated in tumor tissue compared to non-tumor tissue in TCGA_LIHC (Figure 1f). Furthermore, hsa-miR-95-3p was upregulated in CAFs, as well as in HCC tissue in TCGA_LIHC (Figure 1f). To evaluate the clinical significance of these three miRs in HCC progression, a clinical-stage analysis was performed using gene expression data from Catholic_LIHC and Tsinghua_LIHC. In this analysis, expression of hsa-miR-101-3p and hsa-miR-490-3p decreased significantly as HCC stage progressed in Catholic_LIHC, but hsa-miR-95-3p expression was not significant (Figure 1g). This expression pattern was also observed in the Tsinghua_LIHC dataset. The expression of hsa-miR-101-3p and hsa-miR-490-3p was lower in tumor and portal vein tumor thrombosis compared to normal liver tissue, while there was no significant difference in the expression of hsa-miR-95-3p (Figure 1h). In addition, the qRT-PCR analysis results also revealed that the paired CAFs exhibited the lowest levels of expression when compared to non-tumor tissues adjacent to HCC and tumor tissues from the same patient (Supplementary Figure S1). Thus, hsa-miR-101-3p and hsa-miR-490-3p were selected for further analysis, as their expression was associated with tumor progression, suggesting that they might have a central oncogenic role in HCC-CAFs.

Functional Analysis of Target Gene Signatures Regulated by miR-101-3P and miR-490-3p
The target genes of hsa-miR-101-3p and hsa-miR-490-3p were screened using the EN-CORI tool (https://starbase.sysu.edu.cn/ accessed on 12 January 2022) and CLIP-seq data. The candidate target genes were selected only when (1) they were identified as targets of the miRs by at least three of six ENCORI prediction tools and (2) they had at least one binding site for the miRs. As a result, a total of 1235 genes were selected as targets of hsa-miR-101-3p, while 352 genes were selected as targets of hsa-miR-490-3p (Figure 2a). To verify the association between the selected target gene signatures and miRs, miRTarBase, a representative target prediction tool, was used. The target gene signatures selected by the ENCORI tool and CLIP-seq data showed the closest correlation with hsa-miR-101-3p and hsa-miR-490-3p, respectively (Figure 2b).   Gene Ontology analysis considering biological processes (BP; left), molecular functions (MF; middle), and cellular components (CC; right) was performed to elucidate the functional role of the identified target gene signatures (Figure 2c,d). The target genes of hsa-miR-101-3p were enriched in "coronary vasculature morphogenesis" for BP, "phosphatidylinositol monophosphate phosphatase activity" for MF, and "ISWI-type complex" for CC (Figure 2c). The target genes of hsa-miR-490-3p were enriched in "positive regulation of protein acetylation" for BP, "cAMP-dependent protein kinase activity" for MF, and "NSL complex" for CC ( Figure 2d). Next, pathway enrichment analysis using KEGG 2021 Human and MSigDB Hallmark 2020 was performed. In analyses using KEGG 2021, the target genes of hsa-miR-101-3p were enriched in proteoglycans in cancer, while UV response was downregulated in MSigDB analyses ( Figure 2e). Meanwhile, target genes of hsa-miR-490-3p were enriched in ferroptosis in KEGG, and in myc target V1 in MSigDB ( Figure 2f). Common pathways included hedgehog signaling, TGF-beta signaling, and hypoxia, which are closely associated with hepatocarcinogenesis.

TGFBR1 Was Identified as a Common Target Gene of miR-101-3p and miR-490-3p
The possible target gene network of hsa-miR-101-3p and hsa-miR-490-3p was predicted by using the miRWalks database (http://mirwalk.umm.uni-heidelberg.de/ accessed on 8 April 2022), miRDB, and miRTarBase. Genes were selected when they (1) exceeded a 0.95 score and bound to 3'UTR in miRWalks and (2) were predicted as bounding genes of hsa-miR-101-3p and hsa-miR-490-3p in miRDB and miRTarBase. As a result, GEN1, CLCC1, and SMARCD1 were predicted as target genes of hsa-miR-490-3p, and TRIB1, PSPC1, ACVR2B, GRSF1, MCL1, LMNB1, FBN2, LIFR, DCBLD2, and RAP1B were determined to be possible target genes of hsa-miR-101-3p. TGFBR1 was predicted as a common target gene of both hsa-miR-101-3p and hsa-miR-490-3p ( Figure 3a). Expression of the 14 target gene signatures was significantly associated with the prognosis of HCC patients in the TCGA_LIHC database. Specifically, patients with higher expression of the target gene signatures had a significantly poorer prognosis (Figure 3b, left panel). Subgroup analysis was performed according to iCluster classification. iCluster involves integrative clusters classified by genomic, expression, and epigenetic data [11]. Interestingly, only in iCluster 1, which is known as an immune-low cluster, patients with higher expression of the target gene signatures had a significantly poorer prognosis, while there was no difference in iCluster 2/3 (Figure 3b, middle and right panel). In the enrichment analysis, the 14 target genes were highly enriched in TGF-beta signaling in both KEGG 2021 and MSigDB ( Figure 3c). In CBioPortal analyses, alterations of the target gene signatures were related to the TGFB-SMAD and Activin-SMAD pathways, which were associated with cancer cell proliferation and stem/progenitor phenotypes ( Figure 3d). These results suggest that the target genes of the CAF-related miRs contribute to HCC progression by activating the TGF-beta/SMAD pathway.

TGFBR1 Expression Was Negatively Regulated by hsa-miR-101-3p and hsa-miR-490-3p
We performed further analyses to evaluate whether the miRs regulated TGFBR1 expression. First, to determine the specific binding sites of hsa-miR-101-3p and hsa-miR-490-3p in 3 -untranslated regions (3 -UTR) of TGFBR1 mRNA, we analyzed the binding sites and context ++ scores with the TargetScan algorithm. This analysis result showed that both miRNAs bind to the specific sites of TGFBR1 3 -UTR and the absolute binding context scores indicates a high probability of the direct regulation of TGFBR1 by miRs (Figure 4a and Supplementary Table S1). Next, TGFBR1 expression in CAFs and PAFs were evaluated (Figure 4b). TGFBR1 was significantly upregulated in CAFs compared to PAFs. In correlation analyses, TGFBR1 expression was significantly inversely correlated with expression of hsa-miR-101-3p and hsa-miR-490-3p (Figure 4c). In the TCGA_LIHC database, TGFBR1 was upregulated in HCC tissue compared to non-tumor tissue (Figure 4d, left panel: comparison across entire TCGA_LIHC cohort; right panel: paired comparison of tumor and non-tumor tissue). TGFBR1 expression levels were also evaluated in several other studies in the Gene Expression Omnibus (GEO) and International Cancer Genomic Consortium (ICGC) databases. The expression of TGFBR1 in HCC tissues was generally upregulated compared with that in adjacent non-tumor tissues (Figure 4e). We next obtained the expression of TGFBR1 in HCC tissues from the Human Protein Atlas (HPA) database. TGFBR1 was mainly expressed in the cytoplasm/membrane and showed a positive expression of 54.5% in liver cancer tissues (Figure 4f). Next, correlation of TGFBR1 and expression of the miRs were evaluated in twenty pairs of surgically resected HCC tissues and corresponding non-tumor tissues from the Ajou University Hospital (Suwon, South Korea). TGFBR1 was found to be significantly upregulated in tumor tissue compared to non-tumor tissue in 18 of 20 patients (Figure 4g). In addition, TGFBR1 expression was inversely correlated with expression of hsa-miR-101-3p (p = 0.01), while there was no significant correlation between the expression of TGFBR1 and hsa-miR-490-3p (Figure 4h).     To validate the regulatory effect of hsa-miR-101-3p on TGFBR1 expression, an hsa-miR-101-3p mimic was transfected into Huh-7 cells and TGFBR1 expression was measured using quantitative reverse-transcription polymerase chain reaction (qRT-PCR). Interestingly, when the hsa-miR-101-3p mimic was transfected, TGFBR1 expression was markedly lower. Similar results were also observed with the hsa-miR-490-3p mimic transfection; however, the negative regulatory effect on TGFBR1 expression was more potent when hsa-miR-101-3p was transfected (Figure 4i). The Western blot analysis results confirmed the decrease in TGFBR1 protein expression in HCC cells upon treatment with CAF-conditioned medium (CAF-CM) when the miR mimics were transfected. Additionally, the phosphorylation of Smad2, a downstream molecule of the TGF-beta signaling pathway, was significantly reduced in HCC cells when they were treated with CAF-CM and transfected with the miR mimics ( Figure 4j).

Prognostic Implication of TGFBR1 in HCC Patients
In the survival analysis, patients exhibiting higher expression of TGFBR1 had significantly poorer overall survival (OS) and progression free survival (PFS) in TCGA_LIHC (Figure 5a). In an analysis of hsa-miR-101-3p and hsa-miR-490-3p expression, patients with higher expression of both miRs demonstrated a significantly better prognosis in OS, disease free survival (DFS), and PFS ( Figure 5b). Interestingly, combinations of TGFBR1 expression and expression of the two miRs demonstrated more potent prognostic implications on OS, DFS, and PFS than any one gene's expression (Figure 5c). Patients with high expression of TGFBR1 and low expression of the two miRs demonstrated significantly poorer OS (Hazard ratio (HR) = 2.08, p = 0.022), DFS (HR = 2.63, p = 0.003), and PFS (HR = 2.44, p = 0.002) than patients with low TGFBR1 and high miR expression.

Immunological Implication of TGFBR1 in HCC Patients
The immunological implication of TGFBR1 expression was evaluated in TCGA datasets using a TIMER analysis (Figure 5d). As expected, TGFBR1 expression was highly correlated with CAF infiltration. TGFBR1 expression was positively correlated with the infiltration of M2 macrophages, regulatory T cells, and myeloid-derived suppressor cells (MDSCs), and negatively correlated with the infiltration of CD8+ T cells. Further, we also evaluated the correlation between TGFBR1 expression and regulatory T-cell markers (ENTPD1 and CCR8), M2 macrophage markers (PPARD and STAT3), and MDSC markers (LOX and CD83). Interestingly, TGFBR1 expression was significantly positively correlated with Treg, M2, and MDSC markers (Figure 5e).

Discussion
Accumulating evidence indicates that miRs are involved in carcinogenic transformation in the TME [12]. In particular, miRs have been shown to further the ability of CAFs to promote tumor progression [13,14]. However, the oncogenic role of miRs in HCC-CAFs remains poorly evaluated. In this study, hsa-miR-101-3p and hsa-miR-490-3p were identified as major downregulated miRs in HCC-CAFs, and lower expression of hsa-miR-101-3p and hsa-miR-490-3p was associated with a poor prognosis in HCC patients. Bioinformatic analyses revealed that TGFBR1 was a common target gene, and a validation study demonstrated that the expression of TGFBR1 was directly regulated by hsa-miR-101-3p and hsa-miR-490-3p in HCC.
Both hsa-miR-101-3p and miR-490-3p are known as tumor suppressors in many cancers. Aberrant expression of miR-101-3p in CAFs has been reported in lung cancer and breast cancer [15,16]. Guo et al. [15] demonstrated that CAFs promote migration and invasion of cancer cells via miR-101-3p-mediated VEGFA secretion in non-small cell lung cancer. In a study of HCC, Yang et al. [17] reported that CAF-derived TGF-β and SDF1 promote vascular mimicry formation, which was reversed by miR-101. Several prior studies reported that miR-490-3p inhibited migration, invasion, and epithelial-mesenchymal transition of cancer cells by suppressing TGFβR1 expression in colorectal and ovarian cancer [18,19]. In the present study, expression of hsa-miR-101-3p and hsa-miR-490-3p was consistently and significantly downregulated in HCC-CAFs and significantly associated with poor OS. In addition, overexpression of TGFBR1, which was identified as a common target gene of hsa-miR-101-3p and hsa-miR-490-3p, was associated with a poor prognosis in HCC patients. In the same context, previous studies have demonstrated that TGFBR1 acts as a potent modifier of cancer risk, and TGFBR1 overexpression has been associated with cancer cell aggressiveness and poor clinical outcomes in many malignancies [20][21][22][23][24]. TGFBR1 is associated with the TGF-β /SMAD pathway, as demonstrated in Figure 3d [25]. The TGF-β signaling pathway contributes to HCC progression and is known as one of the major oncogenic pathways of CAFs [26,27].
iCluster performs HCC subtyping based on multi-omics technology, including evaluations of DNA copy number and methylation, as well as mRNA, microRNA, and protein arrays, proposed by the TCGA research network [28]. iCluster 1, known as the immune-low cluster, is characterized by a high tumor grade and the presence of macrovascular invasion with significantly worse prognosis [11,29]. The target gene signatures of hsa-miR-101-3p and hsa-miR-490-3p showed significant prognostic implications in iCluster 1. It suggests that aberrant expression of hsa-miR-101-3p and hsa-miR-490-3p in CAFs may play a specific role in iCluster 1 by regulating target gene expression. Further, this role may relate to creating immune-suppressive TMEs. Thus, we evaluated the immunological implication of TGFBR1, which is the common target gene of the two miRs. TGF-β signaling plays a central role in enabling tumor immune evasion, and recent studies have reported that it is associated with poor responses to cancer immunotherapy. The present study revealed that TGFBR1 had a consistent, positive correlation with the infiltration of MDSCs, Treg cells, and M2 macrophages, which are known as key players in promoting an immune-suppressive TME [30]. We also showed that TGFBR1 expression was negatively correlated with CD8+ T-cell infiltration.
This study has several limitations. First, although CAF-specific dysregulated miR signatures were identified in this study, it is difficult to say that these findings are representative of all HCC-CAFs, as the number of included CAF and PAF pairs is only nine. Second, several recent studies have revealed the heterogeneity of the CAF population through single-cell RNA-sequencing (scRNA-seq) [31,32], but this study was based on bulk RNA-sequencing data and did not reflect the heterogeneity of HCC-CAFs. Third, although this study revealed the dysregulated profile of miRs and their target gene signatures in HCC-CAFs through bioinformatic analysis, with attempts to demonstrate its clinical and immunological implications, these results are inferences based on analytical methods. Additional in vitro and in vivo study is required to validate these results. Fourth, only one biological sample of NF was included in this study. Acquiring normal liver tissue for NF primary culture was very difficult, because most of patients with benign liver tumor followed up without surgical resection. Fifth, although we demonstrated significant downregulation of hsa-miR-101-3p and hsa-miR-490-3p in CAFs compared to their paired tumor and non-tumor tissue (Supporting Figure 1), selective downregulation of these miRs in CAFs compared to the other cells in the tumor microenvironment could not be evaluated in the present study. To accurately demonstrate the selective downregulation of these miRs in CAFs, the use of scRNA-seq would be ideal. However, there are currently no studies that have analyzed miRNA expression in HCC tissue using scRNA-seq.

Isolation of CAFs, PAFs, and NFs from Surgically Resected Liver Tissue
The Biobank of Ajou University Hospital, a member of the Korea Biobank Network, provided HCC tissues, corresponding adjacent para-tumor tissues, and a normal liver tissue used in this study. All experiments were performed according to the Declaration of Helsinki and the study protocol was approved by the Institutional Review Board of Ajou University Hospital (approval no. AJIRB-BMR-SMP-17-188; 28 July 2017). HCC tissues and paired para-tumor tissues were collected from HCC patients who underwent surgical resection at Ajou University Hospital (Suwon, South Korea) between 2017 and 2019. Fresh liver tissues were washed with phosphate-buffered saline (GenDEPOT, Barker, TX, USA) and finely minced into small fragments (<1 mm 3 ). Then, the tissue fragments were placed in a culture dish and incubated in fresh culture medium with a cover slip to promote fibroblast attachment. Isolated fibroblasts were maintained in Dulbecco's modified Eagle's medium (DMEM, GenDEPOT) containing 10% fetal bovine serum (FBS; Invitrogen, Waltham, MA, USA) and 100 units/mL penicillin-streptomycin (GenDEPOT) and kept at 37 • C in a humidified incubator with 5% CO 2 .

Small-RNA-Sequencing
Small RNA libraries were constructed from total RNA using the Illumina HiSeq 2000 system (Illumina Inc., San Diego, CA, USA). After small-RNA-sequencing, the cutadapt program was used to remove adapters and low-quality sequences, trimming reads to 18~26 bp in length considering the length of mature miR. Then, the trimmed reads were collapsed to remove duplicates and estimate sequence abundance and annotated using BLAST with miRBase. To enable comparisons between samples, counts of each sample were normalized in units of transcripts per million.

Gene Expression Profiling Using Public Omics Databases
To assess the expression level of miRs and candidate target genes in HCC patients, RNA-sequencing data were obtained from TCGA_LIHC, ICGC, and the GEO databases from the National Center for Biotechnology Information (NCBI) projects: GSE114564; Catholic_LIHC, GSE76903; Tsinghua_LIHC, GSE22058, GSE14520, GSE54236, GSE64041, and GSE76427.

Gene Ontology Analysis and Molecular Pathway Mining
To identify the biological functions and molecular pathways related to the target candidates of hsa-miR-101-3p and hsa-miR-490-3p, Gene Ontology (GO), KEGG 2021 Human, and MSidDB Hallmark 2020 databases were used in Enrichr (https://maayanlab.cloud/ Enrichr/ accessed on 5 April 2022). A p-value < 0.05 was defined as significant in both GO and pathway enrichment analyses. 4.6. Network Analysis between the Hub-Target Genes and CAF-Related miRs miRWalk (version 3.0, http://mirwalk.umm.uni-heidelberg.de/ accessed on 8 April 2022) was used to predict the network between hub-target genes and two CAF-related miRs. In the miRWalk platform, for each miR, we considered all experimentally validated targets reported by the miRTarBase tool and predicted targets identified by both TargetScan and miRDB tools.

Survival Analyses
The relationship between the expression of 14 signatures and LIHC prognosis was analyzed through the Gene Expression Profiling and Interactive Analysis (GEPIA2) database. The GEPIA2 survival analysis tool was used to evaluate this relationship based on gene expression levels, and the Log rank test was applied for hypothesis testing. The expression of the 14 signatures was divided into high-and low-expression cohorts, with the median value of 50% used as the threshold in GEPIA2. Patients with expression levels above 50% were categorized as the high-expression cohort (high 14 signatures), while those with expression levels below 50% were categorized as the low-expression cohort (low 14 signatures). The Cox proportional hazard ratio and 95% confidence interval were included in the survival plots. To evaluate OS, DFS, and PFS for two miRNAs and TGFBR1, clinical data of liver hepatocellular carcinoma (TCGA, PanCancer Atlas) from cBioPortal (https://www.cbioportal.org/ accessed on 5 September 2022) were downloaded and analyzed. The levels of TFGBR1 were analyzed using the median value as the threshold and the combination of two miRNAs was categorized as low/high based on the median value of each individual miRNA. If both miRNAs were found to be high, they were grouped as "2miR_High," and if both miRNAs were found to be low, they were grouped as "2miR_Low" and analyzed. In cases where the TGFBR1 and two miRNAs showed contrasting results, patients with high TGFBR1 and low 2 miRNAs were designated as "TGFBR1_High & 2 miR_Low," and patients with low TGFBR1 and high 2 miRNAs were designated as "TGFBR1_Low & 2 miR_High" and analyzed accordingly.

TGFBR1 Protein Expression in Human HCC tissues
The Human Protein Atlas (https://www.proteinatlas.org accessed on 20 September 2022) was used to analyze the TGFBR1 protein expression level in human HCC tissues. The HPA is a Swedish program initiated in 2003 that aims to map all human proteins in cells, tissues, and organs by integrating various omics technologies, including antibody-based imaging. The representative immunohistochemistry pictures were downloaded from the Tissue Atlas and Pathology Atlas in the HPA.

Cell Culture, Collection of CAF-CM, and miR Mimic Transfection
CAF cells were derived from patient hepatocellular carcinoma tissues. CAF cells were cultured in DMEM with 10% FBS and Huh-7 cells (Korean Cell Line Bank, Seoul, South Korea) were cultured in RPMI-1640 (Sigma-Aldrich) containing 10%; both CAF and Huh-7 cells contained 100 units/mL penicillin-streptomycin (GenDEPOT), and were kept at 37 • C in a humidified incubator with 5% CO 2 .
To analyze the effects of secreted factors from CAFs on tumor cells, CAFs were cultured with the respective media for 48 h. Cell-free conditioned media was collected and stored at −70 • C until used. Synthetic miR mimics (Genolution, Seoul, South Korea) or miR NC mimics (Bioneer, Daejeon, South Korea) were transfected into Huh-7 cells using Lipofectamine 2000 (Invitrogen), according to the manufacturer's instructions. We transfected 3 × 105 Huh-7 cells treated with or without CAF-CM with the miR mimic and cultured for 48 h. Thereafter, the whole-cell extracts were prepared from Huh-7 cells treated with or without CAF-CM with the miR mimic. Cells were harvested, washed with ice-cold phosphate-buffered saline (PBS), and lysed in RIPA buffer (10 mM Tris (pH 7.2), 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 1.0% Triton X-100, and 5 mM EDTA) supplemented with protease inhibitors for 30 min on ice. In this study, the absence of mycoplasma in the cultures was confirmed (Supplementary Figure S2).

RNA Isolation and Quantitative Reverse-Transcription Polymerase Chain Reaction Analysis
QIAzol reagent (Qiagen, Hilden, Germany) was used to extract the total RNA from tissues and cells. cDNA was synthesized from 500 ng of RNA using the miScript RT II kit (Qiagen) or PrimeScript™ RT Master Mix (Takara Bio, Shiga, Japan), in accordance with the manufacturers' instructions. qRT-PCR was performed using the amfiSure qGreen Q-PCR Master Mix (GenDEPOT) and monitored in real time using a CFX Connect Real-Time PCR Detection System (Bio-Rad Laboratories, Hercules, CA, USA). The cycling conditions were as follows: 95 • C for 2 min, 40 cycles of 95 • C for 15 s, 58−62 • C for 34 s, and 72 • C for 30 s, followed by a dissociation stage at 95 • C for 10 s, 65 • C for 5 s, and 95 • C for 5 s. Relative expression levels were calculated using the 2−∆∆Cq method. Utilized primer sequences are listed in Table S2. All assays were performed in triplicate.

Statistical Analysis
All experiments were performed at least three times and all samples were analyzed in triplicate. Between-group differences were analyzed using a paired t-test, unpaired Welch's t-test, or two-way ANOVA with GraphPad Prism version 8.0 software (GraphPad Software Inc., San Diego, CA, USA). Differences were considered statistically significant when p < 0.05.

Conclusions
In conclusion, hsa-miR-101-3p and hsa-miR-490-3p were downregulated in HCC-CAFs, and their common target gene was identified as TGFBR1. The downregulated hsa-miR-101-3p and hsa-miR-490-3p and upregulated TGFBR1 was associated with a poor clinical outcome in HCC patients. TGFBR1 expression was correlated with immunosuppressive immune cell infiltration, involving MDSCs, M2 macrophages, and Treg cells. This is the first study to analyze the aberrant expression of miRs in HCC-CAFs and their target gene signatures through bioinformatic analysis. The results of this study enhance scientific understanding of the molecular signatures of HCC-CAFs and may support further study of HCC therapeutics and biomarkers.

Informed Consent Statement: Not applicable.
Data Availability Statement: RNA-sequencing datasets generated in this study and all other supporting data are available from the corresponding author upon reasonable request.