The bromodomain inhibitor JQ1 up-regulates the long non-coding RNA MALAT1 in cultured human hepatic carcinoma cells

The epigenetic reader, bromodomain-containing 4 (BRD4), is overexpressed in hepatocellular carcinoma (HCC), and BRD4 inhibition is considered as a new therapeutic approach. The BRD inhibitor JQ1 is known to inhibit the enrichment of BRD4 at enhancer sites. Gene network analyses have implicated long non-coding RNAs (lncRNAs) in the effects of JQ1, but the precise molecular events remain unexplored. Here, we report that in HepG2 cells, JQ1 significantly reduced various proliferation-related lncRNAs, but up-regulated the known liver tumor marker, MALAT1. Using ChIP-sequencing data, ChIP-qPCR, luciferase reporter assays, and chromatin conformation capture (3C), we characterized the MALAT1 gene locus. We found that JQ1 elicited a rearrangement of its chromatin looping conformation, which involved the putative enhancers E1, E2, E3, the gene body, and the promoter. We further found that the forkhead box protein A2 (FOXA2) binds to E2 and the promoter; suppression of FOXA2 expression resulted in MALAT1 up-regulation and increased cell proliferation. These results suggest that the inhibition of MALAT1 may improve the effect of BET inhibitors as an anti-cancer therapy and that FOXA2 would be a suitable target for that approach.


JQ1 treatment leads to the upregulation of MALAT1.
To study the role of BRD4 in the HepG2 cells, we treated them with JQ1. This led to a significantly reduced proliferation within 24 h, and the effect increased further until at least 72 h ( Fig. 1A; Supplementary Fig. S1A). We used the 24 h-time point for RNA-seq analysis. Of a total of 856 differentially expressed lncRNAs (DElncRNAs), 333 were up-regulated and 523 down-regulated by JQ1 (Fig. 1B). Heatmaps of the top 40 up-and down-regulated DElncRNAs are shown in Fig. 1C (numerical values are listed in Supplementary Table S1). www.nature.com/scientificreports/ At least some of the downregulated lncRNAs (Fig. 1D, E) were previously found to be highly expressed in liver cancer and to promote proliferation and metastasis (AOC4P, PVT1, DANCR, DBH-AS1, HOXD-AS1, HNF1A-AS1, ANRIL) [20][21][22][23][24] . These lncRNAs probably are also important for HepG2 cells: When we randomly subjected one of them (DANCR) to RNA interference ( Supplementary Fig. S2A), this resulted in a markedly decreased number and proportion of EdU-positive HepG2 cells ( Supplementary Fig. S2B,C), in line with the known oncogenic role of DANCR.
In contrast, we could not make an obvious physiological link for an up-regulated lncRNA (Fig. 1D, E; MALAT1 and TUC338). Interestingly, one of them was MALAT1 ( Fig. 2A, B), which appeared paradoxical because MALAT1 is known to be highly expressed in liver cancer 25 , in line with our own bioinformatics analysis using The Atlas of non-coding RNA in Cancer (TANRIC; https:// ibl. mdand erson. org/ tanric/ design/ basic/ main. html) 26 (Supplementary Fig. S3). However, the stimulation of MALAT1 expression was observed not only with JQ1 but also with other BET inhibitors (OTX015 and ABBV-075) ( Fig. 2A, B, obtained by RNA-seq and qRT-PCR, respectively). Furthermore, an antisense oligonucleotide (ASO) directed against MALAT1 ( Supplementary  Fig. S4) increased the anti-proliferative effect of JQ1, although the oligo alone did not affect cell proliferation (Fig. 2C). This result indicated that the up-regulation of MALAT1 dampened the anti-proliferative effect of JQ1 (Fig. 1A). We, therefore, decided to take a closer look at the MALAT1 gene regulation in JQ1-treated HepG2 cells.
Identification of putative MALAT1 enhancers. We examined ENCODE ChIP-seq and global runon sequencing (GRO-seq) data to localize the potential MALAT1 enhancers (Fig. 2D). Using the GRO-seq peaks (GSE92375), H3K27ac ChIP-seq peaks (GSE29611), and p300 ChIP-seq peaks (GSE32465) at the UCSC Genome browser, we analyzed the upstream regions of MALAT1. In region (chr11: 65,487,241-65,488,714), we found enrichment for H3K27ac that co-localized with the lncRNA gene LINC02736, whose expression was decreased by JQ1 ( Supplementary Fig. S5). In addition, we identified three putative enhancer loci (E1, E2, and E3) further upstream ( Fig. 2D; Table 1). We observed an approximately 20-fold increased luciferase reporter gene expression by the E2 region but not the E1 or E3 regions (Fig. 2E). From these results, we hypothesized that the increased MALAT1 expression in JQ1-treated HepG2 cells might be regulated by enhancer E2. www.nature.com/scientificreports/ FOXA2, but not FOS, is involved in MALAT1 expression and HepG2 cell proliferation. Next, we searched for potential regulators, especially transcription factors (TFs), that might be involved in the JQ1-caused MALAT1 gene upregulation. Using RNA-seq, we found that 274 mRNAs were up-regulated and 737 downregulated by JQ1 ( Supplementary Fig. S6A). The heatmaps of the top 40 up-and down-regulated differentially expressed mRNAs (DEmRNAs) are shown in Fig. 3A (numerical values are listed in Supplementary Table S2). The DEmRNAs were associated with cancer, hepatic system disease, cell death and survival, and cellular growth and proliferation (Fig. 3B). Many down-regulated genes were related to angiogenesis and negative regulation of apoptosis ( Supplementary Fig. S6B). IPA network analysis highlighted known tumor cell apoptosis-related genes (Fig. 3C), some of which we validated by qRT-PCR (Fig. 3D). More to the point, we found that several TFs were also altered, including the apoptosis-associated genes of Fig. 3C (Fig. 3E). Of these, we validated four up-regulated (FOS, EGR1, ZFP36, ID2, JUND) and two down-regulated (FOSL1 and FOXA2) TFs by qRT-PCR (Fig. 3F). Bioinformatics analysis (by IPA) suggested that one of the upregulated TFs, FOS, regulates MALAT1 (Supplementary Fig. S7A), in line with DNA sequence analysis that revealed the co-localization of FOS binding sites and putative MALAT1 enhancers ( Supplementary Fig. S7B). However, both in the absence and presence of JQ1, the levels of MALAT1 were not significantly changed by a FOS siRNA, neither was the JQ1-caused increment of MALAT1 expression ( Fig. 4A; Supplementary Fig. S7C). Furthermore, JQ1 did not increase the binding of FOS to the promoter and putative enhancer regions of MALAT1 ( Supplementary Fig. S7D). These results indicate that contrary to expectation, FOS is not involved in the regulation of MALAT1 in the HepG2 cells.
Next, we focused on the down-regulated TF, FOXA2 (Fig. 3E, F). Bioinformatics analysis of published HepG2 ChIP-seq data indicates that FOXA2 binds to the putative enhancer E2 (X-3) and the promoter (X-P) regions of the MALAT1 gene (Fig. 4B), as validated by our ChIP-qPCR analysis. These bindings were significantly reduced by JQ1 (Fig. 4C). In contrast, the binding of FOXA2 to X-1 and X-2 did not co-localize with E1 or E3 (Fig. 4B), and the JQ1 treatment did not elicit a statistically significant change of FOXA2 binding to X-1 and X-2 (Fig. 4C). These results suggest that the direct binding of FOXA2 to the MALAT1 promoter and enhancer E2, but not E1 or E3, interferes with the transcription of MALAT1, thus mirroring the effect of E2, but not E1 or E3, on luciferase reporter gene expression (compare with Fig. 2E).
The reduction of FOXA2 mRNA ( Supplementary Fig. S7E) and protein ( Supplementary Fig. S7F) by RNA interference led to a significant increase of MALAT1 expression (Fig. 4D) and an increase in the proliferation of the HepG2 cells (Fig. 4E). Of note, we observed the same reciprocal relationship between FOXA2 and MALAT1 in Huh7 cells, another human HCC line ( Supplementary Fig. S8). This result suggests that MALAT1 expression stimulates cell proliferation under negative control by FOXA2.

JQ1 treatment reconfigures the MALAT1 locus.
To better understand the mechanism of how JQ1 affects MALAT1 expression, we performed a 3C assay. Using the promoter region as the anchor (P), we assessed the relative positions of E1, E2, and E3 in the absence and presence of JQ1. Figure 5A shows that in the absence of JQ1, E2 (amplicon C2-P3) and the gene body M (amplicon M-P2), but neither E1 (amplicon C1-P3) nor E3 (amplicon C3-P1), associated with the promoter. Upon the addition of JQ1, all three putative enhancers became associated with the promoter, while the gene body was no longer associated (Fig. 5A). These interactions were confirmed by sequencing the agarose gel bands (Fig. 5B).

Discussion
This study found that when HepG2 cells were treated with JQ1, the long non-coding RNA MALAT1, which has been positively correlated with malignancy, was up-regulated. Our data suggest the down-regulation of the transcription factor FOXA2 and a reconfiguration of the associated chromatin complex as an underlying mechanism.
The JQ1-caused up-regulation of MALAT1 appears paradoxical because BET inhibitors are being considered as anti-cancer agents. However, MALAT1 is highly expressed in various cancers, including liver, lung, and breast cancer, and plays a role in cancer progression 27 . In addition, the MALAT1 expression level is negatively correlated with the survival rate in cancer patients 28 . MALAT1 induces cell proliferation and metastasis via the MAPK/ERK and PI3K/AKT signaling pathways in retinoblastoma and ovarian cancer, respectively 29,30 , and it is known to enable the high expression of the key oncogene MYC in thymic epithelial tumors 31 . Interestingly, in HepG2 cells, MALAT1 was also found in mitochondria, and its knockdown limited ATP synthesis and tumor cell invasion 32 . In addition, MALAT1 causes chemotherapy resistance by regulating miR-216b in HCC 23 . Taken together, literature strongly suggests that MALAT1 expression should be considered as an undesired feature of HCC and other tumors.
We have recently shown that JQ1 down-regulates MYC in HCC cells 33 , which is in line with the anti-cancer effects of JQ1 in other tumors. Similarly, JQ1 reduced the expression of pro-apoptotic BCL2L11 in HCC 9 . However, in prostate cancer, JQ1 inhibited the transcriptional repressor FOXA1, thereby increasing the expression of invasion genes 34 or even activating the DNA damage response 35 . The increased expression of MALAT1 after www.nature.com/scientificreports/ the JQ1 treatment that we described here may also contribute to the unwanted effects of JQ1. These findings collectively emphasize the need to learn more about the mechanisms of BET inhibitors as potential anti-cancer agents. Hence, investigating the mechanisms regulating the overexpression of MALAT1 by JQ1 treatment may contribute to understanding the unwanted side effects of the BET inhibitors. In our study, contrary to expectations 36 , the general TF FOS did not regulate MALAT1. Instead, we identified the lineage-specific TF FOXA2 as a candidate for the modulation of MALAT1 expression in HepG2 cells. The forkhead box (FOX) proteins are transcription factors related to cancer development and progression. FOXA1 is a well-studied regulator of estrogen receptor (ER) and androgen receptor (AR) activity in breast and prostate cancer 37 . In this context, FOX proteins play a crucial role in the rearrangement and reprogramming of super-enhancers 38,39 . FOXA1 and FOXA2 regulate the transcription of liver-specific genes and are known to complement each other 40 . In addition, the importance of FOXA2, particularly concerning liver disease, has been demonstrated 41 . Interestingly, FOXA1 and FOXA2 play dual roles as tumor suppressors and oncogenes 42 . FOXA1 is a transcriptional repressor and reduces the viability and motility in liver cancer cells 43 , while FOXA2 inhibits www.nature.com/scientificreports/ EMT in HCC, breast cancer, and lung cancer 37,44,45 . Hence, our data suggest that a focus on FOXA2 in HCC may help address the problem of JQ1's and potentially other BET inhibitors' detrimental effects in anti-cancer therapy.
In the present study, we associated the JQ1-promoted MALAT1 expression with decreased binding of FOXA2 to the promoter and E2 along with the formation of an (E1, E2, E3)-promoter complex, where E1, E2, and E3 are putative enhancers that we identified. JQ1 has been shown to directly bind to FOXA1, which neutralizes the repressor function of that TF 34 . Our finding that JQ1 reduced the binding of FOXA2 to the MALAT1 gene locus, along with an increase of MALAT1 expression, points to a similar mechanism.
Our data show that JQ1 affects MALAT1 expression by two mechanisms. The first mechanism is indirect and is mediated by the reduced expression of FOXA2, probably caused by the interference of JQ1 with the activity of BRD4 at the FOXA2 locus. This mechanism would be similar to the typical effects of JQ1 on other genes. It leads to the increased expression of MALAT1, as supported by our findings that FOXA2 binds to E2 and that a knockdown of FOXA2 increased the expression of MALAT1. These data reveal that FOXA2 is a repressor of the MALAT1 gene in the HepG2 cells. The second mechanism directly affects MALAT1 expression, as indicated by our finding (by ChIP-qPCR) of a reduced association of BRD4 with the MALAT1 promoter region upon JQ1 treatment. However, the outcome (stimulation versus inhibition of MALAT1 expression) is not yet certain. In general, one might expect that the reduced BRD4 availability reduces the expression of MALAT1 just like it reduces the expression of FOXA2 and other genes. Such a mechanism would counteract the indirect, FOXA2mediated effect. However, our 3C analysis of the MALAT1 promoter and upstream region points to the opposite possibility. We found that JQ1 treatment, which implies a reduced BRD4 level, led to a re-organization of the enhancer-containing chromatin loops associated with the MALAT1 promoter. We note that even reduced levels of BRD4/mediators by BET inhibitors are sufficient to maintain enhancer-promoter interaction 46 . In addition to E2 (now free of its repressor), the putative enhancers E1 and E3 became directly associated with the promoter, suggesting the possibility of a stimulatory effect on MALAT1 gene expression. Future experiments will need to determine the direct effect of JQ1 on MALAT1 gene expression and the relative contributions of the indirect vs. direct mechanisms. It is worth mentioning that we observed the reciprocal relationship between the FOXA2 and MALAT1 also in the independently derived Huh7 human HCC cell line ( Supplementary Fig. S8), indicating that the mechanistic relationships that we studied in the HepG2 cells are not a cell line-specific artifact.
In conclusion, our study suggests a regulatory model for the up-regulation of the lncRNA MALAT1 due to JQ1 treatment (Fig. 5C). The model predicts that manipulating MALAT1 expression could improve the therapeutic effect of BET inhibitors in HCC. Firstly, JQ1 inhibits the binding of FOXA2, a repressor of MALAT1 expression, to the MALAT1 enhancer E2 and the promoter. Secondly, alteration of chromatin looping recruits the enhancers  www.nature.com/scientificreports/ USA). The medium was replaced every 3-4 days. The cells were cultured in a humidified incubator at 37 °C with a 5% CO 2 atmosphere. JQ1 was purchased from MedChemExpress (Monmouth Junction, NJ, USA). JQ1 was present at a concentration of 5 μM for 24 h.
Total RNA sequencing. RNA sequencing (RNA-seq) was performed as previously described 47 . Total RNA was extracted from HCC cells using RNAiso Plus (Takara, Shiga, Japan) and a Qiagen RNeasy Mini kit (Qiagen, Hilden, Germany). RiboMinus Eukaryote kit (Invitrogen, Carlsbad, CA, USA) was used for Ribosomal RNA (rRNA) depletion. An RNA library was created by a NEBNext Ultra directional RNA library preparation kit from Illumina (New England BioLabs, Ipswich, MA, USA). RNA library sequencing was performed on the Illumina HiSeq2500 platform (Macrogen, Seoul, Korea). Transcriptome sequencing was performed on independent RNA samples from DMSO-treated (3 samples) and JQ1-treated (3 samples) HepG2 cells in biological triplicate.
Differentially expressed genes analysis using RNA-seq data. For mRNA analysis, FASTQ files from RNA-seq were clipped and trimmed of adapters, and low-quality reads were removed using Trimmomatic 48 . These FASTQ files were aligned using STAR (version 2.7.8) aligner software with a UCSC hg38 reference 49 . Differentially expressed mRNAs (DEmRNAs) were analyzed using DESeq2 with the default parameters 50 . For lncRNA analysis, the raw data were trimmed with Trimmomatic (version 0.36) 48 and processed using Bow-tie2 (version 2.3.5) 51 or STAR (version 2.7.8) 49 aligner software with a GenCode GRCh38 reference (https:// www. genco degen es. org/ human/) or an LNCipedia reference (https:// lncip edia. org/; version 5.2) 52 . RNAs that exhibited an absolute log 2 -fold change larger than 1.5 or smaller than − 1.5 (log 2 -fold change ≥ 1.5 and log 2 -fold change ≤ −1.5, p-adjusted < 0.05) were designated as DEmRNAs or DElncRNAs. The dataset accession number GSE158552 was deposited in the Gene Expression Omnibus database 53 .
Gene and lncRNA expression analysis using quantitative reverse transcription-PCR (qRT-PCR). Total RNA was extracted from HepG2 cells using RNAiso Plus (Takara, Shiga, Japan) according to the manufacturer's instructions. cDNA was synthesized by PrimeScript reverse transcriptase (Takara, Shiga, Japan) and amplified using gene-specific primers (Supplementary Table S4). The primers were designed by BLAST (https:// blast. ncbi. nlm. nih. gov/ Blast. cgi). qRT-PCR was performed with TBGreen Premix Ex Taq II (Takara, Shiga, Japan). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) or RNU6-1 (U6) were used as an internal control. After performing qRT-PCR, the results were analyzed using the critical threshold (△C T ) and the comparative critical threshold (△△C T ) methods in ABI 7500 (Applied Biosystems, Foster City, CA, USA) software with the NormFinder and geNorm PLUS algorithms. The data represent three independent experiments (n = 3).
Cell proliferation assay. Cell proliferation was assessed using a premixed water-soluble tetrazolium salt Chromatin immunoprecipitation quantitative PCR (ChIP-qPCR). The chromatin immunoprecipitation (ChIP) assay was performed as previously described 54 . Briefly, the HepG2 cell chromatin was incubated with antibodies against BRD4 (Bethyl; A301-985A50), FOS (SCBT; sc-166940x), FOXA2 (Abcam; ab256493) and then precipitated with Dynabeads Protein A beads (Invitrogen, CA, USA); normal rabbit IgG (CST; 2729) and normal mouse IgG (Santa Cruz; sc-2025) were used as controls. The immunoprecipitated DNA was ana- www.nature.com/scientificreports/ lyzed by qRT-PCR, and the expression levels were normalized to the amounts of input DNA. The data represent three independent experiments (n = 3). Primers used for ChIP-qPCR are listed in Supplementary Table S5.
Genomic data analysis. We re-analyzed public H3K27ac ChIP-sequencing (seq) data sets in Gene Expression Omnibus (GEO) (GSE29611) as described previously 55 and global run-on sequencing (GRO-seq) data sets in GEO (GSE92375). For the re-analysis, Trimmomatic (version 0.36) 48 was used to trim the raw data and processed using Bowtie2 (version 2.3.5) 51 or STAR (version 2.7.8) 49 aligner software with a UCSC hg 38 reference. The ChIP-seq and GRO-seq peaks identified were analyzed with Homer (version 4.11) 56 and visualized using UCSC Genome Browser (https:// www. genome. ucsc. edu).
Luciferase reporter assay. Putative enhancer regions (E1, E2, and E3) were amplified with LongAmp Taq 2X Master Mix (New England Biolabs, Ipswich, MA, USA), using forward and reverse primers that generated NheI and XhoI sites, respectively. These amplicons were cloned into the pGL4.26 construct (Promega, Madison, WI, USA). The primers used for cloning are listed in Supplementary Table S3. The cells were seeded into 24-well plates and transfected with Lipofectamine 3000 (Thermo Fisher Scientific, Waltham, MA, USA). Luciferase activity was measured using the Dual-Glo Luciferase Assay kit (Promega, Madison, WI, USA). PRL-TK (Renilla luciferase expression construct; Promega) was used as an internal control. Luciferase activity was normalized to Renilla luciferase and the control (empty vector).
Chromosome conformation capture assay. Chromosome conformation capture (3C) assay was performed as previously described, with minor modifications 57 . HepG2 cells were cross-linked with 1% formaldehyde, and nuclei were prepared from approximately 1-2 × 10 6 cells. Five hundred units of BamHI, BglII, and EcoRI were used to digest the DNA overnight, followed by ligation and purification. The 3C products were quantified by Qubit assay kits (Thermo, Q32851) and amplified by PCR using TB Green Premix Ex Taq (Takara, BR420). The ligation of fragments was analyzed using agarose gel electrophoresis. Sequences of primers are presented in Supplementary Table S6.
Statistical analysis. Data are presented as the mean ± standard deviation (SD) of the mean. All statistical analyses were performed using the IBM SPSS Statistics 26.0 program (IBM corp., Armonk, NY). We used a oneway analysis of variance followed by Tukey's honestly significant difference post hoc test. p-values < 0.05 were considered significant.

Data availability
The raw data of RNA-sequencing were deposited in the Gene Expression Omnibus (GEO) database with accession number GSE158552 51 .