Mesenchymal Stem Cells Promote Hepatocarcinogenesis via lncRNA–MUF Interaction with ANXA2 and miR-34a

Hepatocellular carcinoma (HCC) is the fifth most common cancer and the third leading cause of cancer-related mortality worldwide (1). HCC has been previously thought as a largely cell-autonomous process involving genetically transformed hepatic parenchymal cells or progenitor cells caused by hepatitis B virus/hepatitis C virus infection, food contamination, and nonalcoholic fatty liver disease (2). Recently, interactions between the tumor microenvironment and HCC cells have been increasingly appreciated as pivotal contributors to tumorigenesis and metastasis (3–5). Mesenchymal stem cells (MSC), first identified in the bone marrow, are a heterogeneous group of progenitor cells for tissue maintenance in physiologic conditions. MSCs have been increasingly appreciated as important parts of the tumor microenvironment in the last 10 years, and they can be recruited from the bone marrow or peripheral blood to participate in stromal desmoplastic reactions at tumor sites (4, 6). However, the number of MSCs in the peripheral blood is very low, even in mobilized peripheral blood (7), thus, these recruited MSCs cannot likely to be the main source of MSCs in the tumor microenvironment.

Recently, da Silva and colleagues (8) hypothesized that mouse MSCs reside in the perivascular zone. In addition, Crisan and colleagues (9) further identified human MSCs in the perivascular regions of multiple human organs. On the basis of these findings as well as the high vascular density and strong desmoplastic reaction in HCC tissues, we speculated that there may be numerous of MSCs localized in the perivascular region of HCC sites. Thereafter, we have provided the first evidence that HCC-associated MSCs (HCC-MSC) exist in primary HCC tissues (3). A thorough investigation of the interaction between MSCs and HCC cells might strengthen our understanding of the tumor microenvironment in HCC pathogenesis.

Recently, whole transcriptome sequencing has identified a large number of long noncoding RNAs (lncRNA), larger than 200 nucleotides and lack coding potential (10). Increasing evidence has indicated that lncRNAs can participate in many physiologic or pathologic processes through diverse mechanisms including RNA–protein (11–14) or RNA–RNA interactions (15). Dysregulated lncRNAs have been reported to regulate HCC proliferation (16), metastasis (17), and recurrence (18). However, the role of lncRNAs in MSC-mediated hepatocarcinogenesis remains unknown.

In this study, we investigated the contribution of lncRNA in HCC-MSCs promoting HCC progression. Using lncRNA microarray analysis, we found that a novel lncRNA named lncRNA–MUF (MSC-upregulated factor) was highly induced in HCC cells by HCC-MSCs. Moreover, high lncRNA–MUF expression was associated with poor prognosis of HCC patients. Loss- or gain-of-function analysis indicated that lncRNA–MUF promoted tumorigenesis and epithelial–ss mesenchymal transition (EMT) associated traits. Mechanistically, we found that lncRNA–MUF associated with Annexin A2 (ANXA2) to activate the Wnt/β-catenin signaling, consequently accelerating the EMT program. Furthermore, RNA immunoprecipitation (RIP) and RNA pulldown indicated that lncRNA–MUF can act as competing endogenous RNA (ceRNA) of miR-34a, thus allowing upregulation of Snail1 and subsequently activation of the EMT process. Collectively, our findings present the first evidence that lncRNA can play a pivotal role in MSC-mediated HCC progression.

293T, Hep3B, and PLC were obtained from ATCC. Huh7 from Health Science Research Resources Bank. HepG2 from the National Platform for Experimental Cell Resources. MHCC-97L, HCC-LM3, and SMMC-7721 from the Cell Bank of Shanghai Institute of Cell Biology, Chinese Academy of Sciences. Tumor cells were maintained in high glucose DMEM supplemented with 10% FBS, 100 μg/mL penicillin and 100 U/mL streptomycin. Cell lines were verified by PCR (SV40gp6 for 293T cells; HBVgp2 for PLC cells; AFP, ALB, HBVgp2 and A2M for Hep3B; AFP for Huh7 cells) and were not contaminated by mycoplasma.

HCC-MSCs were extracted as reported previously (3). HCC-MSCs characterization was verified by differentiation assays and surface marker analysis (19). Passages 3–10 of HCC-MSCs were used.

Confluent HCC-MSCs were pretreated with 10 μg/mL mitomycin C for 1.5 hours. Then HCC-MSCs were enzymatically dissociated and 5 × 10⁵ HCC-MSCs were placed on the top inserts (6.5-mm diameter with polycarbonate membrane filters containing 0.4-μm pores; Corning Inc.). After culturing 24 hours, the inserts were placed into the fresh wells of ultra-low attachment plates filled with tumorsphere medium and 2,000 HCC cells. Tumor spheres were countered under stereomicroscope after 7–9 days. Sphere medium contained DMEM/F-12 supplemented with B27 (Life Technologies; 1:50), N2 (Life Technologies,1:100), 20 ng/mL EGF, 10 ng/mL bFGF, 100 U/mL penicillin, and 100 ng/mL streptomycin (19).

GFP-positive Huh7, SMMC-7721, and HepG2 cells were cocultured with HCC-MSCs: The initial plating ratio of HCC cells to HCC-MSCs was 1:3. After coculturing for 7 days, GFP-positive HCC cells were sorted with BD FACSAria II cell sorter (BD Biosciences). Total RNA was extracted by TRIzol and hybridized to the lncRNA V3.0 microarray (Agilent). The array data were subject to quality control and background subtraction by the GeneSpring software. Quantile normalization was modulated by limma package. Probes were selected to be differential expressed when fold change was greater than 2 between experimental and control group.

The data were download from The Cancer Genome Atlas (TCGA) website and DESeq2 package was used to detect differential expressed genes. Fold change ≥2 and P value ≤0.05 were set as the threshold for significantly differential expression. Hierarchical cluster analysis of differentially expression lncRNAs was performed to explore the expression pattern. The results showed that the differentially lncRNAs could distinguish between HCC samples and adjacent tissue samples.

We labeled TCGA samples as “high” or “low” according to whether the expression of lncRNA-MUF was higher or lower than the median value among all samples. The log-rank test was used to measure whether the survival time was significantly different between the “high” and “low” expressed groups. The Kaplan–Meier plots were made by the R packages.

RNA pulldown assays were performed as described below (14). Briefly, Biotin-labeled RNAs were transcribed with Biotin RNA Labeling Mix and T7 RNA polymerase, purified with the RNeasy Mini Kit. Total RNA was heated and annealed to form secondary structure, mixed with cytoplasm extract in RIP buffer at room temperature for 1 hour. Streptavidin agarose beads were incubated at room temperature for 1 hour. Beads were extracted by TRIzol reagent for quantitative PCR (qPCR) analysis.

A total of 2 × 10⁶ tumor cells were injected subcutaneously into nude mouse. After 4 weeks, when subcutaneous tumors reached about 1 cm in diameter, mice were sacrificed, tumors were harvested, weighed, and stained with hematoxylin and eosin. The animal studies have been conducted in accordance with an Institutional Animal Care and Use Committee of Institute of Biophysics, Chinese Academy of Sciences.

Values are shown as the mean ± SD and all experiments were conducted at least three times. Significance was determined using the two tailed Student t test: *, P < 0.05; **, P < 0.01. Other methods are detailed in the Supplementary Materials and Methods.

Cancer-associated MSCs are becoming increasingly appreciated as an important part of the tumor microenvironment (20), but the role of HCC-MSCs in HCC progression has been less reported. Here, we found that HCC-MSCs significantly enhanced tumor sphere formation of HCC cells via Transwell coculture system (Fig. 1A). We investigated the effects of the interaction between HCC-MSCs and HCC cells (GFP). Flow cytometry analysis showed that cancer stem cell (CSC) markers such as CD90 and CD13 were significantly induced by HCC-MSCs (Fig. 1B). To further evaluate the CSC-promoting effects of HCC-MSCs in vivo, limiting dilution tumorigenicity assays were performed by subcutaneously inoculating luciferase-labeled SMMC-7721 cells in nude mice, admixed with HCC-MSCs or not. We found that 100 or 1,000 SMMC-7721 cells alone cannot form tumor, and even 10⁴ SMMC-7721 cells alone only formed tumors in 1 of 5 mice. When coinoculated with HCC-MSCs, 100 and 1,000 SMMC-7721 cells resulted in tumor formation in 2 of 4 mice (Fig. 1C). When 1 × 10⁶ SMMC-7721 or HepG2 cells were subcutaneously inoculated, tumor weights in the HCC-MSC coinoculation group were significantly enhanced (Fig. 1D). Furthermore, both the bioluminescence intensity and metastatic nodules at liver sites indicated that HCC-MSCs significantly promoted metastasis in vivo (Fig. 1E and F). These results provide strong evidence that the HCC cells switch to a more aggressive phenotype when cocultured with HCC-MSCs.

To understand the mechanism of how HCC-MSCs promote HCC progression, we used microarray analysis to evaluate changes in lncRNAs and mRNAs in HCC cells cocultured with HCC-MSCs (Fig. 2A). Transcriptome profiling identified that 57 lncRNA transcripts (Fig. 2B; Supplementary Table S1) and 110 mRNAs (Supplementary Table S1; Supplementary Fig. S1A) were differentially expressed in three different HCC cell lines when cocultured with HCC-MSCs. The transcriptome data have been deposited in the Gene Expression Omnibus (GEO) database (GSE86160). Until now most lncRNAs still remain uncharacterized, we first investigated the altered mRNAs of HCC cells induced by HCC-MSCs. Gene Set Enrichment Analysis (GSEA) and Gene Ontology (GO) assays both indicated that the differentially expressed mRNAs enriched in extracellular matrix, which is closely associated with the EMT signaling (21) and metastasis (Supplementary Fig. S1B–S1D; ref. 22). In addition, qPCR and Western blotting assays confirmed that EMT-related genes were significantly upregulated by HCC-MSC conditioned medium (Supplementary Fig. S1E and S1F). TGF-β1, a well-characterized inducer of EMT, served as the positive control (Supplementary Fig. S1G). Our results suggest that HCC-MSCs might largely promote HCC malignancy through the EMT process.

To further identify which lncRNAs might play important roles in HCC-MSCs promoting HCC progression, candidate lncRNAs were screened using the following criteria: lncRNA expression was significantly altered by HCC-MSCs in lncRNA microarrays (Fig. 2B, Supplementary Table S1) and TCGA database, indicating differential expression of the lncRNAs in HCC specimens compared with adjacent nontumor tissues (Fig. 2C). Among the five most differentially expressed lncRNAs, a novel lncRNA (LINC00941, Ensembl ID: ENSG00000235884, fold changes >5 and P < 0.01) particularly drew our attention (Fig. 2D). As this is an uncharacterized lncRNA, we termed it lncRNA–MUF (MSCs upregulated factor). Consistent with the microarray results, qPCR assays confirmed that lncRNA–MUF was significantly upregulated in HCC cells by HCC-MSCs (Supplementary Fig. S2A and S2B). In addition, it was highly expressed in HCC specimens compared with adjacent nontumor tissues (Fig. 2E). The TCGA data (Fig. 2F) and our previous lncRNA microarray (GSE70880, Supplementary Fig. S2C) confirmed these results (23, 24). Notably, Kaplan–Meier survival analysis indicated that high lncRNA-MUF expression was correlated with poor survival of HCC patients (Fig. 2F). Furthermore, lncRNA–MUF expression was significantly enhanced by HCC-MSC conditioned medium and TGF-β1 (Fig. 2G; Supplementary Fig. S2D). Moreover, lncRNA-MUF was highly expressed in tumor spheres and highly invasive HCC cells (Fig. 2G). Collectively, these findings indicate that lncRNA–MUF is significantly upregulated by HCC-MSCs.

LncRNA–MUF resides on chromosome 12p11.21 and is localized near active regulatory elements including the histone mark H3K27Ac marks and DNaseI hypersensitivity clusters (Fig. 3A). A 1884-base pair (bp) transcript was determined by 5′ and 3′ rapid amplification of cDNA ends assays (RACE) and displayed no coding potential, as determined by the CNCI, CPAT, and PhyloCSF score analysis (Supplementary Table S2). Northern blot assays confirmed that one transcript of about 1,800 bp was mainly found in HCC cells (Supplementary Fig. S2E). Cellular fractionation and RNA FISH (RNA-FISH) analysis indicated that lncRNA–MUF was mainly localized in the cytoplasm (Fig. 3B). Knockdown of lncRNA–MUF significantly attenuated the capacity of tumor sphere formation and invasion in vitro (Fig. 3C; Supplementary Fig. S2F). Moreover, lncRNA–MUF depletion significantly repressed the subcutaneous xenograft formation and metastatic capacity in vivo (Fig. 3D). In contrast, tumor sphere formation was increased in lncRNA–MUF–overexpressing cells (Fig. 3E). Given the fact that HCC-MSCs stimulate the EMT process, we examined the expression of EMT-related gene after lncRNA–MUF alteration. Both qPCR and Western blotting demonstrated that lncRNA–MUF depletion significantly repressed the expression of mesenchymal-related genes, whereas lncRNA–MUF overexpression had the opposite effect (Fig. 3F). Taken together, these data show that lncRNA–MUF can regulate the EMT program to promote HCC progression.

LncRNAs can exert their functions through RNA–protein interactions to modulate target genes (12, 25). Therefore, we used RNA pulldown assays followed by silver staining and mass spectrometry (MS) to identify lncRNA–MUF–protein interactions. ANXA2 was found to specifically bind to lncRNA–MUF (Fig. 4A; Supplementary Table S3). As shown in Fig. 4B, ANXA2 was detected in the biotin-labeled sense lncRNA–MUF group through RNA pulldown assays followed by Western blotting. Moreover, the interactions between lncRNA–MUF and ANXA2 were verified by RNA immunoprecipitation (RIP) assays (Fig. 4C). In addition, a series of truncated lncRNA–MUF were constructed to map the specific binding region between lncRNA–MUF and ANXA2, suggesting that nucleotides 800 to 1,600 of lncRNA-MUF could bind to ANXA2 (Fig. 4D). Competitive RNA pulldown assays further confirmed the interaction of ANXA2 with lncRNA–MUF (Fig. 4E). Moreover, RNA-FISH and immunofluorescence assays indicated that lncRNA–MUF colocalized with ANXA2 in the cytoplasm of HCC cells (Fig. 4F). The GEO data showed that ANXA2 was highly expressed in HCC tissues and metastatic samples (Fig. 4G). Depletion of ANXA2 significantly repressed tumor sphere formation and tumorigenicity (Fig. 4H). These data indicate that lncRNA–MUF–ANXA2 axis can modulate HCC progression.

To investigate the mechanism by which the lncRNA–MUF–ANXA2 axis promotes HCC progression, we examined the expression of EMT-related genes upon ANXA2 alteration. Mesenchymal-related genes and β-catenin were significantly altered by ANXA2 overexpression or knockdown (Fig. 5A). β-Catenin in different cellular compartments mediates distinct cellular function (26), thus we determined whether ANXA2 affected the subcellular distribution of β-catenin. As shown in Fig. 5B, Western blotting assays showed an increase of β-catenin in the cytosol and nucleus of ANXA2-overexpressing cells. In contrast, β-catenin in the nucleus decreased substantially in ANXA2-knockdown cells. Furthermore, several β-catenin target genes were upregulated in ANXA2-overexpressing cells and downregulated in ANXA2 knockdown cells (Supplementary Fig. S3A). These results suggest that ANXA2 can alter the subcellular localization of β-catenin to activate the Wnt cascade. Phosphorylation of β-catenin by glycogen synthase kinase 3β (GSK-3β) causes its degradation by the ubiquitin-proteasome system (27). Therefore, we examined whether the altered expression of β-catenin resulted from phosphorylation by GSK-3β. As shown in Fig. 5A, phosphorylation of β-catenin at S33/S37 and T41 residues was reduced in ANXA2-overexpressing cells and increased in ANXA2 knockdown cells. Coimmunoprecipitation (co-IP) assays showed that GSK-3β precipitated less endogenous β-catenin in ANXA2-overexpressing cells (Fig. 5C) and more β-catenin in ANXA2-knockdown cells (Fig. 5D). Likewise, endogenous GSK-3β precipitated by β-catenin were affected by ANXA2 alterations (Fig. 5C and D). Furthermore, co-IP and RNA-FISH assays showed that ANXA2 could bind with GSK-3β (Fig. 5E; Supplementary Fig. S3B) but not directly interact with β-catenin (Supplementary Fig. S3C). Taken together, these results show that ANXA2 can bind to GSK-3β and disrupt the formation of the GSK-3β/β-catenin complex. Moreover, we investigated the effect of lncRNA–MUF in the interaction between GSK-3β and ANXA2, and found that knockdown of lncRNA–MUF repressed the promoting effect of ANXA2 on β-catenin and EMT-related proteins (Fig. 5F). In addition, co-IP assays confirmed that lncRNA-MUF overexpression significantly enhanced the interactions between ANXA2 and GSK-3β, whereas lncRNA–MUF depletion repressed the binding effect (Fig. 5G). Analysis of lncRNA–MUF truncations indicated that GSK-3β but not β-catenin binds to nucleotides 800 to 1,200 of lncRNA–MUF (Supplementary Fig. S3D), showing that lncRNA–MUF interacts with ANXA2 and GSK-3β via distinct but partially overlapping domains. Taken together, the binding of lncRNA–MUF and ANXA2 plays important roles in the activation of Wnt/β-catenin signaling and the EMT process.

Emerging evidence suggests that cytoplasmic lncRNAs can act as ceRNAs to modulate the functions of miRNAs (15, 28, 29). To examine whether cytoplasmic localized lncRNA–MUF can bind endogenous miRNAs, we predicted the target miRNAs using PITA and RNAhybrid softwares. Only the miRNAs with highly ranking target sites and low expression in HCC tissues were selected for further analysis (Supplementary Fig. S4A). RIP and RNA pulldown assays showed that miR-34a and miR-133a were significantly enriched in the lncRNA–MUF group compared with the control (Fig. 6A and B). Our preliminary results showed that miR-34a repressed tumor sphere formation more significantly than miR-133a (data not shown). Thus, we mainly focused on the effect of lncRNA–MUF binding to miR-34a. Luciferase activities and AGO₂ RIP assays showed that lncRNA-MUF could bind with miR-34a (Fig. 6C). Furthermore, RNA-FISH analysis demonstrated that lncRNA–MUF was colocalized with miR-34a in the cytoplasm of HCC cells (Fig. 6D). Next, we found that miR-34a repressed tumor sphere formation and tumor growth of HCC cells (Supplementary Fig. S4B and S4C). Moreover, when lncRNA–MUF and miR-34a were cotransfected into HCC cells, lncRNA-MUF significantly rescued the inhibitory role of miR-34a on tumor sphere formation and invasion (Fig. 6E; Supplementary Fig. S4D). In addition, qPCR and Western blotting analysis showed that lncRNA–MUF could recover the inhibitory effect of miR-34a on the EMT process (Fig. 6F). These data indicate that lncRNA–MUF acts as a ceRNA to affect the function of miR-34a in HCC cells.

To further elucidate the mechanism underlying the contribution of the lncRNA–MUF–miR-34a axis to HCC progression, TargetScan software predicted that Snail1 was one of the targets of miR-34a (Fig. 7A). And miR-34a significantly repressed the expression of Snail1 in HCC cells; luciferase reporter assays demonstrated that miR-34a could bind the 3′-untranslated region (UTR) of Snail1. Importantly, lncRNA–MUF attenuated the binding capacity of miR-34a to the 3′-UTR of Snail1 (Fig. 7A). Moreover, Western blotting analysis confirmed that lncRNA–MUF could reduce the inhibitory effect of miR-34a on Snail1 expression (Fig. 7B). In addition, Snail1 knockdown suppressed tumor sphere formation induced by lncRNA-MUF overexpression (Fig. 7C). qPCR and Western blotting analysis demonstrated that Snail1 depletion inhibited the promoting effect of lncRNA-MUF on the EMT progress (Fig. 7D and E). Taken together, these data show that lncRNA–MUF can function as a ceRNA of miR-34a to regulate the expression of Snail1 and enhance the EMT process.

A schematic representation of the cross-talk between HCC-MSCs and HCC cells is illustrated in Fig. 7F. The interaction between HCC-MSCs and HCC cells induced the upregulation of lncRNA–MUF, thereafter, lncRNA–MUF can associate with ANXA2 and miR-34a to regulate the EMT process and tumor progression.

In this study, we investigated the role of lncRNAs in HCC-MSCs promoting HCC progression. Interestingly, we found that the novel lncRNA–MUF was significantly increased in HCC cells by HCC-MSCs. Moreover, lncRNA–MUF significantly promoted tumorigenesis and EMT traits. Mechanistically, lncRNA–MUF-interacted with ANXA2 and miR-34a to participate in the EMT process and tumorigenesis. Collectively, the results of this study present a new mechanism by which HCC-MSCs promote HCC progression.

Recent studies have indicated that tumorigenesis relies heavily on the reciprocal interactions between tumor cells and the surrounding stroma (30–33). Identifying the critical pathway involved in this cross-talk could potentially improve the efficiency of treatment (34, 35). It has been increasingly recognized that cancer-associated MSCs act as important contributors to tumor progression (20, 36). However, most studies have focused on the role of growth factors or cytokines in the cross-talk between MSCs and tumor cells. Breast cancer–associated MSCs promoted sphere formation via the EGF/EGFR/Akt pathway (19). McLean and colleagues (36) demonstrated that ovarian cancer–associated MSCs expressed more bone morphogenetic proteins, thereby altering the stemness of ovarian tumors. Waghray and colleagues (20) indicated that pancreatic adenocarcinoma-associated MSCs can secret more granulocyte macrophage-colony stimulating factor to enhance tumor proliferation. Our previous results demonstrated that S100A4 secreted from HCC-MSCs could participate in HCC progression (3). Recently, accumulating evidence has suggested that dysregulated lncRNAs are involved in the proliferation (16, 37), metastasis (17), CSCs maintenance (38), and prognosis of HCC patients (39). However, it has remained unknown whether lncRNAs can play roles in the cross-talk between HCC-MSCs and HCC cells.

In this study, we used mRNA and lncRNA microarrays to investigate the mechanism of the interactions between HCC-MSCs and HCC cells. We firstly analyzed the altered mRNAs in HCC cells by HCC-MSCs, GSEA, and GO analyses indicated that HCC-MSCs conferred EMT characteristics to HCC cells. The EMT causes epithelial cells to acquire mesenchymal traits and is closely associated with CSC-like features (21, 22, 40). In accordance with the role of HCC-MSCs in liver cancer progression as well as inducing the EMT in HCC cells, we hypothesized that the differentially expressed lncRNAs induced by HCC-MSCs might play important roles in modulating the EMT program.

Through screening of our microarrays and the TCGA data, we found that lncRNA–MUF was one of the most highly expressed lncRNAs in HCC cells induced by HCC-MSCs and it played a critical role in EMT modulation. Through RNA pulldown and mass spectrometry analyses, ANXA2 was found to bind with lncRNA–MUF, and the binding led to upregulated β-catenin and the EMT process. Upregulation of β-catenin occurs in several cancers such as colorectal cancer (41), breast cancer (42), and liver cancer (43). Canonical Wnt signaling depends on β-catenin/GSK-3β complex conformation and phosphorylation by GSK-3β, leading to β-catenin degradation (44, 45). Our results showed that ANXA2 could bind to GSK-3β and disrupt the formation of GSK-3β/β-catenin complex. In addition, co-IP experiments showed that lncRNA–MUF could promote the interaction between GSK-3β and ANXA2. Our results demonstrate that lncRNA–MUF might play a critical role in Wnt signaling to promote HCC development.

It was recently shown that cytoplasmic-localized lncRNAs can act as ceRNAs to regulate miRNAs (29, 46). LncRNA-ATB activated by TGF-β1 promotes the HCC invasion-metastasis cascade by competitively binding miR-200s to regulate the EMT factors ZEB1 and ZEB2 (17). Exosome-transmitted lncARSR acts as a ceRNA of miR-449 to facilitate c-MET expression and promote sunitinib resistance (47). The pseudogene BRAF, which encodes a subclass of lncRNAs, functions as ceRNAs to induce lymphoma formation (15). In our study, we found that lncRNA–MUF functions as a ceRNA of miR-34a to regulate Snail1 activity and modulate HCC progression.

In this study, we mainly focused on the downstream mechanism by which lncRNA–MUF participates in HCC development. However, it still remains unknown how HCC-MSCs upregulate the expression of lncRNA–MUF. Because HCC-MSC–conditioned medium can enhance the expression of lncRNA–MUF, we speculated that growth factors or cytokines secreted from HCC-MSCs may participate in the regulation of lncRNA–MUF expression. However, we cannot rule out the possibility that HCC-MSCs can modulate the expression of lncRNA–MUF via direct physical contact. It will be of great interest to explore in future studies.

In summary, our study highlights the importance of lncRNA–MUF as a mediator of MSC niche modulating HCC progression. To the best of our knowledge, this is the first study to characterize the role of lncRNA in the cross-talk between MSCs and HCC cells. These results provide new insights into the mechanism underlying the interactions between the tumor microenvironment and HCC cells.

No potential conflicts of interest were disclosed.

Conception and design: X. Yan, D. Zhang, S. Wu, R. Chen, Z. Fan

Development of methodology: X. Yan, D. Zhang, S. Wu, J. Qian, J. Wu, Y. Huang, X. Liu, B. Liu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Yan, D. Zhang, J. Qian, F. Yan, G. Huang, X. Liu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Yan, D. Zhang, W. Wu, Y. Hao, F. Yan, P. Zhu, Y. Huang

Writing, review, and/or revision of the manuscript: X. Yan, D. Zhang, W. Wu, Z. Fan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Yan, D. Zhang, J. Luo, X. Chen, Y. Du

Study supervision: R. Chen, Z. Fan

We are grateful to Dr. Geir Skogerbø for manuscript reading and we thank LetPub (www.letpub.com) for its linguistic assistance of the manuscript. We thank Prof. Xiaofei Zheng (Beijing Institute of Radiation Medicine, Beijing, China) for providing MS2bs plasmids and Dr. Junying Jia for FACS analysis.

This work was supported by grants from National Natural Science Foundation of China (81472341 and 81772617 to X. Yan; 31401098 to D. Zhang; and 31520103905 to R. Chen); Beijing Natural Science Foundation (7152095 to X. Yan); Beijing Municipal Education Commission (KM201710005031 to X. Yan); 973 Program of the MOST of China (2015CB553705 to Z. Fan); National High Technology Research and Development Program of China (2015AA020108 to R. Chen).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.