Regulation of MYO18B mRNA by a network of C19MC miRNA-520G, IFN-γ, CEBPB, p53 and bFGF in hepatocellular carcinoma

MYO18B has been proposed to contribute to the progression of hepatocellular carcinoma (HCC). However, the signals that govern MYO18B transcription are not known. Here we show that, a network of C19MC miRNA-520G, IFN-γ, CEBPB and p53 transcriptional-defects promote MYO18B mRNA expression in HCCs. IFN-γ by itself suppresses MYO18B transcription, but promotes it when miRNA-520G is stably overexpressed. Similarly, CEBPB-liver-enriched activator protein (LAP) isoform overexpression suppresses MYO18B transcription but promotes transcription when the cells are treated with IFN-γ. Furthermore, miR-520G together with mutant-p53 promotes MYO18B transcription. Conversely, bFGF suppresses MYO18B mRNA irrespective of CEBPB, miR-520G overexpression or IFN-γ treatment. Finally high MYO18B expression reflects poor prognosis while high MYL5 or MYO1B expression reflects better survival of HCC patients. Thus, we identified a network of positive and negative regulators of MYO18B mRNA expression which reflects the survival of HCC patients.

www.nature.com/scientificreports/ Using human hepatocellular carcinoma (HCC) patient data here we show that, C19MC overexpression is tightly linked to MYO18B mRNA expression in patients who harbor transcription incompetent p53. In p53 defective Hep3B cells, the expression of MYO18B is suppressed by interferon-γ (IFN-γ) and that the presence of C19MC miRNA-520G reverses this suppressive effect to promote the expression of MYO18B mRNA. Stable overexpression of CEBPB mimics the effect of miR-520G in promoting MYO18B mRNA expression. Furthermore, wild-type and mutant p53s promote the expression of MYO18B mRNA in the presence of miR-520G. On the other hand, basic Fibroblast Growth Factor (bFGF) suppresses MYO18B mRNA expression irrespective of IFN-γ treatment, CEBPB overexpression or miR-520G expression. Thus our study significantly expose the transcriptional regulatory network of MYO18B, which in future will help to study the role of these signaling pathways in myopathy, cirrhosis of the liver, and the development of hepatocellular carcinoma.

Results
High MYO18B mRNA expression is correlated with C19MC overexpression and poor survival in hepatocellular carcinoma. Deregulated expression of Myosin-18B is linked to HCC progression, stress fiber formation, cirrhosis of liver, and cardiac dysfunction, through defects in myosin-II z-stack formation of muscle fibers (Fig. 1A). In hepatocellular carcinoma patients, high MYO18B mRNA expression is significantly associated with poor survival (Fig. 1B). We next examined whether MYO18B mRNA expression is associated with any of the integrated molecular classification clusters (iClusters) using TCGA iCluster dataset 27 . MYO18B mRNA was significantly enriched in iCluster-3 (Fig. 1C), a cluster known to harbor most p53 defects in HCCs 27 . Furthermore, examination of integrated RNA-seq and miRNA-seq data revealed that, MYO18B mRNA is significantly expressed in tumors with high C19MC miRNA expression (Fig. 1D). Taken together, these data demonstrate that, high MYO18B mRNA expression is correlated with C19MC overexpression and poor survival in iCluster-3 of HCCs.
Genomic structure of MYO18B gene enhancer reveals multiple CEBPB binding sites. To understand the transcriptional cause for the high expression of MYO18B mRNA in iCluster-3, we examined the 5′-regulatory region of MYO18B gene in UCSC genome browser. MYO18B gene is located on Chr22q12.1 and has a very strong enhancer marked by H3K27Ac (chr22: 26,137,162,170: hg19) (Fig. 2). Transcription factor ChIP-Seq data from UCSC genome browser (ENCODE) revealed numerous transcription factor binding sites within this enhancer which includes CEBPB, p53, Myc, Max, GATA-2 and others (data not shown), but we focused our attention on CEBPB because of the following reasons: (i) CEBPB has the capability to regulate enhancers in liver environment 28 (ii) CEBPB is tightly linked to obesity 5 and (iii) CEBPB sites were also present at C19MC region on chromosome-19 17 . MYO18B has 3 CEBPB binding sites within its enhancer region and a fourth CEBPB binding site located upstream to the enhancer region (Fig. 2). CEBPB is capable of binding to these regions as evaluated by examining ENCODE ChIP-seq data (Fig. 2). Notably, CEBPB binds to a fourth site close to transcriptional start site (TSS) upon forskolin induction (Fig. 2). While the CEBPB sites may regulate different isoforms of MYO18B mRNAs, we chose exon-3 of the longest isoform for expression analysis by RT-PCR because it is shared by multiple isoforms of MYO18B mRNAs (Fig. 2). Taken together these data reveal that, the MYO18B gene harbors a strong enhancer with four CEBPB binding sites.
Hsa-miR-520G-3p remodels IFN-γ but not bFGF signaling to regulate MYO18B transcription. Although MYO18B has multiple transcription factor binding sites in addition to CEBPB and a strong  www.nature.com/scientificreports/ enhancer, additional signals are likely needed to activate transcription from MYO18B gene. Therefore we screened the mRNA expression of cytokines and chemokines for correlation to MYO18B mRNA expression in HCCs. We chose the cytokines bFGF, IFN-γ, EGF and IL-6 that are related to myopathy and or cirrhosis [29][30][31][32] , and found that, bFGF is negatively correlated and the remaining three cytokines are positively correlated to MYO18B mRNA expression (Fig. 3A). To understand the effect of C19MC miRNA expression on MYO18B mRNA expression we chose miR-520G which is known to promote drug resistance in cancer cells 33,34 . Analysis of miR-520G in HCCs revealed that this miRNA is expressed more in iCluster-3 (Fig. 3B), a cluster also expresses more MYO18B mRNA (Fig. 1C). Treatment of Hep3B cells with 1 nM each of IFN-γ, IL-6, EGF and bFGF promoted CEBPB mRNA expression while IFN-γ, EGF and bFGF suppressed MYO18B mRNA expression (Fig. 3C). However, in miR-520G stably transfected cells, IFN-γ promoted both CEBPB and MYO18B mRNAs whereas EGF and bFGF treatment downregulated both CEBPB and MYO18B mRNAs (Fig. 3C). The effect of CEBPB promotion was stronger in IFN-γ treated conditions compared to the other cytokines tested (Fig. 3C). The effect of MYO18B mRNA suppression was stronger in bFGF treated conditions (Fig. 3C), which stand in line with the negative correlation of FGF2 with MYO18B in HCC patients (Fig. 3A). Although IL-6 could promote CEBPB in Hep3B untransfected cells, it could not promote CEBPB mRNA in miR-520G stable cells (Fig. 3C). Taken together these results demonstrated that, miR-520G remodels IFN-γ signaling to promote MYO18B transcription and that bFGF negatively regulate MYO18B mRNA expression.
CEBPB mimics the effect of miR-520G in MYO18B mRNA expression but bFGF counteracts it. We noted a striking correlative upregulation or downregulation of CEBPB with MYO18B mRNA levels in response to IFN-γ or bFGF respectively (Fig. 3C) raising the question that, the CEBPB expression level could mimic the effect of these cytokines or miR-520G (Fig. 4A). Of note, miR-520G overexpression does not alter the mRNA expression of MYO18B/IFNG/bFGF/CEBPB/cytokines receptors compared to control pMIR transfected cells (Supplementary figure 1A). Complete lack of IFNG mRNA expression prompted us to examine whether IFNG gene is deleted in Hep3B cells. However, IFNG was not deleted in Hep3B cells as per copy number data (Supplementary figure 1B). To test whether CEBPB expression level could mimic the effect of the cytokines or miR0520G, we stably overexpressed the LAP-isoform of CEBPB (which is known to promote transcription compared to its short isoform: liver-enriched inhibitor protein (LIP) 35 ) in Hep3B cells. Overexpression of CEBPB-LAP isoform itself suppressed MYO18B transcription compared to control empty vector transfected cells (Fig. 4B). Importantly, IFN-γ treatment tremendously promoted MYO18B transcription in CEBPB-LAP overexpressed cells compared to empty vector transfected cells (Fig. 4B). However, bFGF treatment suppressed both CEBPB and MYO18B mRNAs even when co-treated with IFN-γ (Fig. 4B). The reduction of CEBPB mRNA ) in UCSC Genome Browser track (blue peaks) was focused to show CEBPB binding at this region using ENCODE CEBPB ChIP-seq HepG2 data (red and black peaks). Note the three CEBPB binding sites (indicated by red peaks) within the enhancer region (indicated by red box) in uninduced HepG2 cells and a notable fourth binding site close to transcription start site (TSS) in forskolin induced HepG2 cells. One another CEBPB site was far upstream to enhancer region (indicated by green shaded box). The exon chosen for RT-PCR analysis is indicated at exon-3 of long isoform. www.nature.com/scientificreports/ in CEBPB-LAP overexpressed condition suggests that, bFGF promotes the degradation of CEBPB mRNA rather than suppressing transcription from CEBPB promoter because the overexpression vector employs a different promoter (CMV). To understand whether CEBPB binding site is involved in the promotion of the transcription of IFN-γ target genes, we chose IFI27 gene which is known to get transcribed in response to IFN-γ 36 but lack CEBPB binding sites or binding within its enhancer region ( Fig. 4C and Supplementary figure-2). We examined the same cDNA set that was used for Fig. 4B (CEBPB-LAP overexpressed and its control) and found that, IFI27 mRNA was not promoted by mere overexpression of CEBPB or when the CEBPB-LAP overexpressed cells were treated with IFN-γ (Fig. 4D). However, bFGF abolished the expression of IFI27 mRNA or impeded the IFN-γ induced IFI27 mRNA expression. These data demonstrate that, CEBPB binding is required for IFN-γ to promote transcription as IFN-γ could not promote IFI27 mRNA in CEBPB-LAP overexpressed cells.
The data from Figs. 3 and 4 together demonstrates that CEBPB is sufficient to mimic the effect of miR-520G in IFN-γ induced alterations of MYO18B transcription but bFGF suppresses MYO18B mRNA levels irrespective of IFN-γ or CEBPB overexpression.
Transcription defective p53, increased miR-520G and MYO18B expression reflect a lethal phenotype with cellular transformation in HCCs. We next examined the possible reasons why patients with high MYO18B exhibited poor overall survival. In general, p53 defective tumors are the indication for poor survival and p53 defects can be of one or many of the different types (such as copy number loss, transcriptional repression, degradation at protein level, or gain-of-function due to mutations). Therefore we classified HCCs into p53-transcription competent (p53TC) or p53-transcription incompetent (p53TI) groups using a p53-target gene transcription signature that consists of 30 genes 27 and integrated this dataset to miRNA-seq data. Interestingly, miR-520G, MYO18B and IFNG RNAs were significantly expressed more in p53-transcription incompetent tumors than in p53-transcription competent tumors (Fig. 5A). On the other hand, bFGF (FGF2) mRNA was significantly downregulated in p53-transcription incompetent tumors than in p53-transcription competent tumors (Fig. 5A).
Considering the fact that Hep3B cells harbor p53 defects 37,38 , MYO18B is constitutively expressed at mRNA level in this cell line (Fig. 5B). Transient overexpression of wild-type (WT) or gain-of-function mutant p53s (R175H and R273H) did not promote MYO18B transcription in Hep3B cells but promoted MYO18B transcription in miR-520G stably overexpressed cells (Fig. 5B). Furthermore, sphere formation represents cancer cells with aggressive and transformed phenotype [39][40][41] , therefore, we examined whether MYO18B transcription is altered in monolayer versus sphere forming Hep3B cells. MYO18B mRNA is expressed more in sphere forming cells than monolayer cells (Fig. 5C). We further tested the miR-520G expression in miR-520G stably transfected Hep3B monolayer cells versus spheres at 48 h and found that, the spheres accumulate 3.76 (SEM = ± 0.014) fold higher amount of miR-520G-3p compared to monolayer (Fig. 5C). Survival analysis of p53TC and p53TI tumors revealed that the p53TI patients had significantly poor prognosis than the p53TC patients (Fig. 5D). www.nature.com/scientificreports/ Taken together these data demonstrate that transcription defective p53, increased C19MC miRNA-520G expression in patients and transformed state of cells and increased MYO18B transcription reflects a lethal phenotype with cellular transformation in HCCs.
MYO18B is negatively correlated to MYL5 and MYO1B expression to reflect survival outcome. MYO18B is part of a large family of myosin genes which constitutes both myosin heavy chains and light chains to provide structural organization of cells and tissues such as liver. Cirrhotic liver often show abrupt texture of liver and therefore more myosins may have redundant roles along with or against MYO18B. Therefore we next examined whether other family members of myosin-18B positively or negatively correlate with MYO18B expression. For this purpose we subjected the 52 myosin family member genes from the p53TC/p53TI RNA-seq expression dataset to correlation analysis and found that, many myosins were positively or negatively correlated to MYO18B (Fig. 6A). We focused on two myosins, MYO1B and MYL5 that were negatively correlated to MYO18B expression in HCCs and were expressed significantly lower quantities in p53TI tumors compared to p53TC tumors (Fig. 6A,B). This result suggested that, higher expression of these myosins may reflect better survival and p53-transcriptional competence. In line with this, overall survival analysis based on MYL5 or MYO1B revealed that, higher expression of MYL5 or MYO1B is significantly associated with better survival in HCCs (Fig. 6C), which is in contrast to high MYO18B expression (Fig. 1B).

Discussion
Cirrhosis of the liver is a major and classical risk factor for HCC 2,3 and obesity is thought to play a role in this context 4 . Therefore, the pathways that modulate cirrhosis and obesity may play a role in the prognosis of HCC patients. MYO18B has been shown to promote progression of HCCs through PI3K/Akt/mTOR pathway 14 . However, the regulation of MYO18B is not characterized in detail. Hereby, we show for the first time that, a complex network of IFN-γ, CEBPB (a transcription factor drives obesity through adipogenesis 5,7,8 ), miR-520G, and p53defects co-operatively regulate the expression of MYO18B mRNA which in turn reflects the poor survival of HCC patients (Fig. 7). On the other hand we show the interesting negative regulatory aspect of bFGF in counteracting MYO18B mRNA expression induced by IFN-γ/CEBPB network (Fig. 7). Presence of CEBPB binding sites is a crucial aspect in the promotion of MYO18B mRNA because another IFN-γ target gene IFI27 failed to get promoted and lacks CEBPB binding site (Fig. 4C,D and Supplementary figure-2). In fact, bFGF may promote the degradation of CEBPB mRNA to achieve the negative regulation of IFN-γ-induced MYO18B mRNA expression because, bFGF almost silenced CEBPB mRNA expression despite the fact that CEBPB was overexpressed using a CMV promoter (therefore it is not due to repression of original genomic CEBPB promoter alone) (Fig. 4B). Defects in p53 can result in aggressive phenotype involving cancer stem cell expansion through blebbishield emergency program-mediated cellular transformation/sphere formation 33,39,40,[42][43][44][45][46][47][48][49][50][51][52] . Therefore, MYO18B expression may not be the direct cause of poor survival in HCCs but it reflects the poor survival due to its association with p53 transcriptional incompetence and associated aggressive therapy resistance and stem-cell expansion phenotypes. The increased expression of MYO18B mRNA and miR-520G in spheres compared to monolayer cells supports this notion. High expression of MYL5 and MYO1B mRNAs indicates an opposite outcome compared to Myosin-18B may contribute to proliferation of cancer cells as targeting MYO18B expression is linked to skeletal muscle cell proliferation in rheumatoid arthritis 53 . In ovarian and colorectal cancers Myosin-18B is considered as a tumor suppressor 54,55 . However, the C19MC miRNAs and IFN-γ (analogous to inflammatory environment of the cirrhotic liver) in p53 defective background may render it as an oncogene in HCCs as per our data.
In summary, our study identified a complex network of IFN-γ, CEBPB, C19MC miR-520G and p53-transcriptional incompetence as positive regulators of MYO18B mRNA expression and bFGF as negative regulator of MYO18B mRNA expression to reflect the survival outcome of HCC patients.  www.nature.com/scientificreports/

Materials and methods
The cancer genome atlas (TCGA) and iCluster details. LIHC RNA-seq, miRNA-seq data were from TCGA (https:// gdac. broad insti tute. org/) and an integrated patient data sub-set was used which is based on the patient IDs of integrated cluster (iC1 + iC2 + iC3 = 183 samples). The integrated iCluster dataset was based on the expression of 528 signature genes (200 + 128 + 200 genes from iC1, iC2 and iC3 respectively) as described previously 27  Evaluation of CEBPB binding to MYO18B and IFI27 regulatory regions: ChIP-seq data analysis. The CEBPB ChIP-seq data were accessed from Encyclopedia of DNA Elements (ENCODE) 56 . CEBPB ChIP-seq data sets with or without forskolin induction in HepG2 cells [ENCODE: ENCSR000EEX file: ENCFF000XPP (fold change over control hg19) and ENCSR000BQI file: ENCFF321NDM (fold change over control hg19)] were examined for CEBPB binding at MYO18B enhancer region (Chr22:26,135,000-26,160,000, hg19) and visualized using Integrative Genomics Viewer (IGV: BROAD institute, version 2.4.10). The data range was kept constant (40) for both uninduced and forskolin induced tracks. For IFI27, same data sets were used with same settings but by focusing on IFI27 regulatory region (Chr14:94,576,511-94,577,956, hg19). The peaks are comparable qualitatively and quantitatively between MYO18B and IFI27 genes within same tracks however, the uninduced and forskolin induced tracks are comparable only qualitatively but not quantitatively. Survival analyses and statistics. TCGA LIHC (HCC) survival data for MYO18B were obtained through Oncolnc (https:// www. oncol nc. org/) using 40% settings for high and low groups, and matched with RNAseq dataset sorted based on MYO18B expression values and selected high and low groups (n = 144 each). For MYO1B, survival data was obtained similarly using 35% settings, matched with RNA-seq dataset sorted based on MYO1B expression values and selected high and low groups (n = 126 each). For MYL5, survival data was obtained similarly using 35% settings, matched with RNA-seq dataset sorted based on MYL5 expression values and selected high and low groups (n = 39 each). For p53TC versus p53TI survival analysis the survival data were matched to p53TC (n = 50) and p53TI (n = 41) dataset where the patient number is one less for each group due to non-availability of data. The survival data were plotted using GraphPad Prism v.7.04 (La Jolla, CA, USA) and the log-rank (Mantel-Cox test) p-values were considered for level of significance. The p-values < 0.05 were considered significant and < 0.001 were considered robust significance.
The iCluster or myosin datasets from RSEM normalized LIHC (HCC) TCGA RNAseq were log transformed to the base of 10 before generating matrix table in R. The insignificant correlations were coded white and thus white indicates either correlation value = 0 or insignificant. The lentiviral expression cassettes were used as plasmids for transfection rather than as viruses or with accompanying plasmids to package viruses, because C19MC is a cluster that responds to viral infections. Wild-type p53 (#16434) and mutant p53s p53-R175H (#16436) and p53-R273H (#16439) plasmids under CMV promoter were a gift from Bert Vogelstein 60 . An empty CMV promoter containing plasmid was used as empty control. All plasmids were isolated using Qiagen MIDI prep kit (#12143).
Hep3B cells were stably transfected using plasmids (not viruses in the case of lentiviral plasmids) and Lipofectamine 2000 (Life Technologies # 11668019) and selected using 4 μg/ml puromycin (Invitrogen # A1113803) for 2 months while GFP/RFP positive clones were picked, expanded and frozen. For transient transfections, 1 μg plasmid DNA/10 cm dish was used with Lipofectamine for 12-14 h., in complete MEM, the media were washed off, and the cells were then collected at 48 h. duration (from the time of addition of DNA + Lipofectamine complex to cells). Sphere formation and microscopy. Hep3B cells were plated at high density (500,000 cells/ml) in regular tissue culture 10 cm dish (for monolayer) or in low attachment flasks (for spheres) in complete MEM and cultured for 48hrs with a media change at 24hrs. The spheres were stained with Hoechst-33342 and imaged at 48hrs using Zeiss Observer.Z1 microscope equipped with Axiocam 503 mono (Zeiss) camera. The individual channel images of Hoechst-33342 were pseudo-colored to red, merged with bright field and exported using ZEN 2.3 Pro software (Carl Zeiss Microscopy, GmbH, 2011, Blue edition). The final composite was done using Adobe Photoshop CS5 (Adobe Systems Inc., San Jose, CA, USA). Similar experiments were performed to collect monolayer cells and spheres for RNA isolation for RT-PCR/qRTPCR analysis using Hep3B parental cells or miR-520G stably transfected cells.

RNA-seq evaluation of genes of interest.
RNAs from stable miR-520G and pMIR control cells were isolated using miRNeasy Mini Kit (Qiagen #217004, Germantown, MD, USA), with an on-column RNAse free DNAse (Qiagen # 79254) digestion as per manufacturer's protocol. RNA-seq was then performed in quality control tested RNAs using the NuGen Ovation RNA-seq FFPE System (PN 7150-08) to prepare the libraries and were run on the Illumina NextSeq 500 with a 76-base paired-end read. The adapter reads were trimmed using Cutadapt (v1.8.1) and raw reads were then aligned to human genome (build: hg19) using STAR (v2.5.3a). Gene expression was evaluated as read count at gene level with HTSeq (v0.6.1) and Gencode gene model v28. Gene expression data were then normalized using DEseq2. The genes of interest were then visualized using Microsoft Excel (2010).