Tumor suppressor Hypermethylated in Cancer 1 represses expression of cell cycle regulator E2F7 in human primary cells

Hypermethylated in Cancer 1 (HIC1) is an established tumor suppressor, which is frequently inactivated in various cancers. In colorectal carcinoma (CRC), silencing of HIC1 has been recognized as one of the important events in malignant tumor progression. Strikingly, CRC patients with high HIC1 expression have a worse prognosis than patients with relatively low HIC1 mRNA levels. To analyze the function of HIC1, we performed expression profiling of human primary fibroblasts after downregulation of HIC1 by RNA interference. We show that HIC1 deficiency triggers a p53-dependent response and that disruption of the HIC1 gene in human colon cells delays cell cycle progression under serum deficiency conditions. Moreover, treatment with etoposide, a DNA-damaging agent, significantly impairs the proliferation rate and dynamics of damaged DNA repair in HIC1-deficient compared with wild-type cells. One of the genes upregulated in HIC1-depleted cells encodes cell cycle regulator E2F7. E2F7 is an atypical member of the E2F family, which functions primarily as a transcriptional repressor, and its downregulation is essential for proper cell cycle progression and expression of genes involved in DNA repair. We demonstrated that E2F7 is indeed the target of transcriptional repression mediated by HIC1. Moreover, our results suggest that the phenotypic manifestations associated with loss of the HIC1 gene, in particular the changes in cell cycle progression and slowed repair of damaged DNA, are caused by dysregulation of E2F7 expression. Finally, we observed an inverse relationship between HIC1 and E2F7 in a panel of CRC. Importantly, CRC patients who express relatively high levels of E2F7 have a remarkably better prognosis than patients with intermediate or low levels of E2F7 expression.


Introduction
The Hypermethylated in Cancer 1 (HIC1) gene is frequently lost or silenced in many human cancers, suggesting its role as a tumor suppressor (1). Functional evidence for this conclusion has been provided by gene targeting studies in mice. While mice with homozygous loss of Hic1 exhibited multiple developmental defects leading to embryonic and perinatal lethality (2), mice with a functional Hic1 allele were viable and fertile. However, Hic1 +/animals spontaneously developed multiple tumors in various tissues later in life (3). The HIC1 protein has been characterized as a transcriptional repressor containing the N-terminal Broad complex, Tramtrack, and Bric à brac/POx viruses and zinc finger (BTB/POZ) protein-protein interacting domain responsible for HIC1 multimerization and binding with multiple partners, and the C-terminal portion containing five Krüppel-like C2H2 zinc fingers. The latter portion provides affinity for the HIC1-responsive element (HiRE) present in the regulatory regions of genes repressed by HIC1 (1). Histone deacetylase sirtuin 1 (SIRT1) was one of the proteins associated with HIC1 via the BTB/POZ domain. Interestingly, the HIC1-SIRT1 complex binds directly to the SIRT1 promoter and represses transcription of the SIRT1 gene. Inactivation of HIC1 attenuates SIRT1 production, leading to deacetylation of p53, followed by suppression of the proapototic response induced by DNA damage (4).
Based on the presence of HiRE in regulatory regions, several other HIC1 target genes have been identified. The genes encode various proteins involved in proliferation, differentiation and apoptosis (5)(6)(7)(8). Another important function of HIC1 is to inhibit the transcriptional complexes that mediate STAT3 (9) and canonical Wnt signaling (10). In many clinical studies, inactivation of HIC1 correlates with a more aggressive phenotype and poorer survival in various tumor types (reviewed in (11)). Our analysis of gene expression profiles of colorectal neoplasia samples revealed that HIC1 expression was indeed decreased in precancerous stages. However, patients with tumors with relatively high HIC1 mRNA expression in more advanced lesions, i.e., colorectal carcinomas (CRCs), had a lower survival rate than patients with CRCs with low HIC1 expression (12).
To uncover the molecular basis of the tumor suppressive role of HIC1, we expression profiled human primary fibroblasts after HIC1 mRNA silencing by RNA interference (RNAi). In addition, we used the clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 system to disrupt the HIC1 gene in human colon (HC) cells. Gene set enrichment analysis (GSEA) revealed that small inhibitory (si) RNA-mediated silencing of HIC1 primarily triggers a p53dependent transcriptional response. The cell morphology and proliferation rate of HIC1-deficient HC cells were normal, but serum starvation resulted in a decreased proliferation rate of these cells.
Cell cycle analysis revealed a statistically significant accumulation of HIC1-deficient cells in G2/M phase, suggesting delayed progression through the cell cycle. In addition, treatment with etoposide, an agent that inhibits DNA topoisomerase II and leads to double-strand breaks in genomic DNA, significantly impaired the proliferation rate of HIC1 KO HC cell compared with HIC1 wild-type (WT) cells. Subsequent analysis revealed that DNA repair is impaired in the absence of HIC1.
One of the genes upregulated after HIC1 knockdown or HIC1 gene disruption encoded transcription factor E2F7. The function of the E2F family of transcription factors is to regulate cyclin-dependent kinases (CDK) and retinoblastoma protein (RB), which control cell cycle progression. Importantly, dysregulation of E2F7 (and E2F8) leads to delayed cell cycle transition (13). We confirmed that the E2F7 gene is indeed repressed by HIC1. Our additional results suggest that the cell cycle defects and lower DNA damage repair dynamics observed in HIC1-deficient cells are due to the dysregulation of E2F7 expression. We also analyzed the mRNA levels of HIC1 and E2F7 in tumor samples obtained from colorectal tumors. We observed an inverse relationship between the expression levels of HIC1 and E2F7 in CRC samples. Importantly, a negative correlation between the expression of these genes was observed, and, in addition, patients with high E2F7 expression had longer survival than individuals with tumors producing increased levels of HIC1 mRNA.

Materials and methods
Cell culture, CRISPR/Cas9 editing of the HIC1 locus, and chromosome analysis Human WI38, HFF-1, and HEK293 cells were purchased from ATCC (cat. no. CCL-75, SCRL-1041, CRL-1573, respectively). Human HC cells immortalized by human telomerase reverse transcriptase (hTERT) expression were purchased from Applied Biological Materials (cat. no. T0570). The cell lines used were not further authenticated. All cell lines were regularly tested for mycoplasma using the MycoStrip TM detection system (InvivoGen, Toulouse, France); all cell lines were maintained in Dulbecco's Modified Eagle's Medium (DMEM) with GlutaMAX (Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum (Thermo Fisher Scientific) and 100 U/mL penicillin and 100 μg/mL streptomycin (both from Thermo Fisher Scientific) in a humidified atmosphere containing 5% CO2 at 37°C. Knockout HIC1 HC cell lines (HIC1 KO) were generated using the lentiCRISPRv2 vector (Addgene, #52961); HIC1-specific single guide RNA (sgRNA) was designed using the CRISPR design tool available at crispr.mit.edu; see supplementary Table S1 for the list of gRNA sequences. Cells were co-transfected with the lentiCRISPRv2 vector and pARv-RFP reporter plasmid (14); the plasmids contained the corresponding DNA sequence that drives production or is recognized by the sgRNA, respectively. Red fluorescent protein (RFP)-positive cells were sorted into 96-well plates as single cells 72 hours after transfection. HIC1 knockout was validated by sequencing DNA fragments amplified from genomic DNA derived from multiple clones. Primers are listed in supplementary   Table S1. For chromosome analysis, cells were fixed in suspension and maintained in Carnoy's solution. Twenty metaphases were evaluated for each cell clone; HIC1-deficient and control HC cell clones were analyzed approximately 15 generations after HIC1 gene/"mock" targeting.
Chromosome banding was performed using the standard G-banding procedure. The chromosome number and structure were evaluated using an Olympus BX60 microscope (Olympus Czech Group, Prague, Czech Republic) and Ikaros software (MetaSystems, Altussheim, Germany).
Karyotypes were assembled according to the International System for Human Cytogenomic Nomenclature (15). Bioconductor (16). Datasets obtained with DNA microarrays were analyzed in the R environment using the Linear models for microarray data analysis (LIMMA) package (17).

RNA isolation and reverse-transcription quantitative polymerase chain reaction (RT-qPCR)
For RT-qPCR analysis, total RNA from WI38 and HFF-1 cells was transcribed using  Table S1.

Proliferation assay and cell cycle analysis
The cell number and viability were determined using CellTiter-Blue reagent (Promega

Luciferase reporter assay
The assay was performed using the Dual-Glo Luciferase Assay System and a GloMax® 20/20 Luminometer (all from Promega). To test the activity of the E2F7 promoter, a 3kb region of the promoter containing HIC1 binding sites was amplified by PCR from human genomic DNA and cloned into the pGl4.26-luc vector (Promega); primers are listed in supplementary Table S1.

Western blotting, immunohistochemistry, and antibodies
purchased from Thermo Fisher Scientific) staining before measurement of nuclear γH2AX and 53BP1 staining, and foci were identified using the Spot detection module. Image analysis was performed with at least 400 cells per given condition.  (12)). Datasets GSE39582 and GSE3958 were retrieved from the Gene Expression Omnibus (20). Expression data based on RNA sequencing (RNA-seq) were retrieved from the colorectal adenocarcinoma dataset (TCGA, PanCancer Atlas) available at cBioportal (21). The data were analyzed using RNA-seq by the expectation maximization (RSEM) software package (22).

Statistical analysis and availability of expression data
Statistical tests were performed in the R environment. Expression profile data were deposited in the ArrayExpress database (https://www.ebi.ac.uk/arrayexpress/); accession number: E-MTAB-10678.

Expression profiling of primary human cells after HIC1 silencing
We used primary human lung fibroblasts WI38 treated with HIC1-specific siRNAs to investigate changes in the expression profile caused by decreased HIC1 expression. Cells were transfected with two different HIC1 siRNAs (labeled 'Amb HIC1' or 'Dhar HIC1') or with a nonsilencing control siRNA and harvested at 24, 48 and 72 hours post-transfection. Quantitative RT-PCR analysis of cells transfected with HIC1-specific siRNAs showed a strong decrease in HIC1 mRNA levels compared to cells transfected with the non-silencing siRNA (Fig. 1A). In addition, gradually decreasing levels of the HIC1 protein were observed at these time points ( Since HIC1 functions as a transcriptional repressor, we analyzed the genes whose levels increased after HIC1 silencing. We performed GSEA of 80 and 49 genes that were upregulated 48 and 72 hours after HIC1 silencing, respectively. We used BioPlanet and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway datasets included in the Enrichr web tool (23); the datasets catalog all biological pathways, their interactions and changes in various healthy or diseased conditions. The analysis revealed that the pathway most affected by the decreased expression of HIC1 was the p53 pathway (Fig. 1D). Note that the significance criterion (adjusted p-value < 0.05) for the p53-pathway dataset was reached only for the 72-hour siRNA treatment time interval. Subsequent analysis of potential binding sites of transcription factors (TFs) using the Encyclopedia of DNA Elements (ENCODE) project web tool (24) and the ChIP Enrichment Analysis (ChEA) database derived from the integration of genome-wide experiments using ChIP-Chip, ChIP-seq, ChIP-PET, and DamID (these methods were referred to as ChIP-X (25)) revealed that eight and nine genes that were upregulated 48 and 72 hours after HIC1 downregulation, respectively, were possibly regulated by p53. In addition, comparison of the same gene list with gene expression signatures from RNA-seq studies obtained after application of drugs or specific gene manipulations/disruptions (the corresponding datasets were obtained from the Gene Expression Omnibus (GEO) repository) showed the most significant match (13 genes) with the group of genes obtained after activation (stabilization) of p53 in MCF-7 breast cancer cells treated with small molecule Nutlin-3a (26) (Fig. 1D and not shown).

E2F7 expression was suppressed by HIC1
One of the genes that was upregulated after knockdown of HIC1 in the 48-and 72-hour time interval was the gene encoding transcription factor E2F7 (supplementary Table S2). Using RT-qPCR analysis, we confirmed that the expression of E2F7 mRNA was increased after HIC1 knockdown with both HIC1-specific siRNAs not only in WI38 cells, but also in primary human foreskin fibroblasts HFF-1 ( Fig. 2A and data not shown). We then generated an E2F7-luciferase reporter (E2F7-luc) by cloning the E2F7 promoter region containing a putative HiRE upstream of the luciferase gene and tested its response to HIC1. The reporter activity was increased after HIC1 knockout (Fig. 2B, left graph). In contrast, co-transfection of the reporter with increasing amounts of the HIC1 expression construct resulted in decreased E2F7-luc reporter activity, confirming transcriptional repression of E2F7 by HIC1 (Fig. 2B, right diagram). Furthermore, we used the CRISPR/Cas9 system to disrupt the HIC1 gene in human HC cells (27). Alignment of the HIC1 locus was confirmed by sequencing the corresponding genomic regions, and three HIC1 KO HC cell clones, #13, #20, and #25 (supplementary Fig. S1), were used for further analysis. The control HIC1-proficient clones transfected with the Cas9-expressing vector lacking the sgRNA expression cassette were designated CTRL #1 and CTRL #2. We encountered a problem in detecting endogenous HIC1 protein in cell lysates obtained by Western blotting from HC cells. Nevertheless, RT-qPCR analysis revealed a marked decrease in HIC1 mRNA in HC cells transfected with HIC1specific gRNA (Fig. 2C). Because the region spanning the translation initiation region of HIC1 mRNA was targeted (supplementary Fig. S1), we suspected that HIC1 mRNA levels were reduced by a nonsense-mediated RNA decay mechanism (28) (Fig. 2C).

E2F7 displays elevated expression in human 'HIC-low' CRCs
We have previously shown that the expression of HIC1 varies in CRCs (12). Therefore, we sought to analyze a possible relationship between E2F7 and HIC1 in CRC samples. To determine changes in E2F7 expression during colorectal neoplastic progression, we analyzed samples from neoplastic progression stage (Fig. 3A). Moreover, RT-qPCR analysis of total RNA isolated from colorectal tumor samples showed that the increased E2F7 expression correlated with the decline in HIC1 mRNA (Fig. 3B). Next, we used the cBioPortal for Cancer Genomics Webtool (https://www.cbioportal.org/) to analyze the expression profile of 524 CRC samples. The RNA sequencing data showed an inverse correlation between E2F7 and HIC1 expression levels; this correlation was relatively weak, although significant (Fig. 3C). The inverse relationship between E2F7 and HIC1 expression was confirmed by statistical analysis of public datasets GSE37892 and GSE39582, which contain DNA microarray-based expression analyses of 130 and 443 colorectal cancer samples, respectively (29, 30) (supplementary Fig. S2). Next, we performed immunohistochemical detection of HIC1 and E2F7 in the progressive stages of colon carcinogenesis consisting of healthy mucosa, hyperplastic polyps, adenomas with low-or highgrade dysplasia, and invasive adenocarcinomas. In the healthy colonic mucosa, HIC1 expression showed an apicobasal gradient during the progression of colorectal cancer. In the healthy mucosa, nuclear and cytoplasmic E2F7 staining was observed in almost all epithelial cells with marked positivity in the superficial colonocytes; immunopositivity was maintained in the hyperplastic lesion. Starting with the low-grade adenoma, we observed heterogeneous nuclear and strong cytoplasmic E2F7 immunostaining (Fig. 3D). Finally, we investigated whether the E2F7 expression level was related to some specific clinicopathological features included in the GSE39582 dataset (30). First, we determined the relative E2F7 mRNA expression in each CRC.
Samples that fell in the first and last deciles according to E2F7 expression were defined as 'E2F7high' and 'E2F7-low', respectively. The rest of the samples were labeled as 'Other'; HIC1 expression levels were reported in the same way in each CRC. Interestingly, patients with high E2F7 expression had better survival than patients with low E2F7 expression. In contrast, HIC1high/low CRCs showed the opposite trend, in agreement with previously published results (12) (Fig. 3E).

HIC1 knockout in HC cells disturbed cell cycle progression
Based on functional studies, E2Fs can be considered as transcriptional activators (E2F1-E2F3) or repressors (E2F4-E2F8). Production of E2F activators peaks in late G1 phase and promotes cell entry into S phase. In contrast, the level of E2F repressors reaches its maximum in S and early G2 phases (31,32). Yuan and colleagues have shown that downregulation of E2F7 is essential for proper cell cycle progression (13). Cell morphology and proliferation of HIC1 KO cells clones were normal compared to parental HC cells or CTRL #1 and CTRL #2 cell clones (Fig. 4A, left graph, and data not shown). Therefore, we decided to expose these cells to stress conditions. Interestingly, serum deficiency-induced stress resulted in decreased cell proliferation in HIC1 KO cell clones (Fig. 4A, right diagram) Therefore, we subsequently tested p53 protein levels in HIC1-deficient and control HC cells; immunoblotting was also performed with an antibody that recognizes transcriptionally active p53 protein phosphorylated at serine 15 (35). As shown in Fig. 5E, we did not observe stabilization of p53 protein in cells that had not been incubated with etoposide. In addition, HIC1-deficient HC cells exhibited increased levels of p53 protein as well as its form phosphorylated at serine 15 after 24 hours of exposure to etoposide. Surprisingly, we found no significant overlap between our gene sets and the gene signatures obtained by expression profiling of MEFs after deletion of the Hic1 allele, as previously published in two independent studies (33,36). This fact likely indicates the cell specificity of HIC1-regulated genes and/or the manner in which HIC1 expression was repressed, i.e., gene inactivation of the conditional Hic1 allele in mouse cells compared to siRNA-mediated knockdown in human cells. In addition, the HIC1 gene did not meet the significance criteria, although RT-qPCR analysis of total RNA prior to microarray-based gene expression profiling clearly showed robust downregulation of HIC1 mRNA. We suspected that the observed discrepancy might be caused by the very high GC content of HIC1 mRNA leading to the low signal on the DNA chip (37).

Discussion
The loss of the HIC1 gene had a minimal effect on the cell "behavior" of HC cells.
However, growth in serum-free medium resulted in lower proliferation of these cells. Subsequent cell cycle analysis showed that cells lacking HIC1 have tendency to accumulate in the G2/M phase.
This observation was rather unexpected considering the tumor suppressive role of HIC1. On the other hand, Kumar described a similar effect of HIC1 depletion in glioma cells (38). In addition, we treated HC cells with etoposide, which induces double-strand breaks in replicating cells that trigger activation of ataxia telangiectasia mutated (ATM) kinase, leading to phosphorylation of multiple proteins such as H2AX and 53BP1 (39). Culture with etoposide caused a decreased proliferation rate in HIC1-deficient cells, which was probably related to lower dynamics of repair of damaged DNA. This effect of HIC1 loss is consistent with the observation of Dehennaut and colleagues, who found that HIC1 is involved in DNA repair (8). Szczepny and colleagues recently reported that inactivation of Hic1 in MEFs leads to chromosomal instability and G2/M arrest (33).
In relation to the above observation, we performed karyotype analysis of HC cells. HIC1-deficient cells indeed exhibit increased chromosomal aberrations. However, a clone of control HC cells was also chromosomally unstable. Therefore, it is not possible to say whether the chromosomal aberrations observed in HIC1-deficient HC cells are indeed directly related to HIC1 loss. Another remarkable effect of the HIC1 gene loss was an apparent activation of the DNA damage response (DDR) even in cells cultured under standard conditions, i.e., without etoposide.
One of the genes whose expression was increased in WI38 and HFF-1 cells treated with HIC1-specific siRNAs and in HIC1-deficient HC cells was E2F7. Transcription factor E2F7 is an atypical member of the E2F family. E2F7 acts primarily as a transcriptional repressor that antagonizes the action of the "classical" E2F proteins, e.g., E2F1 and E2F2 (31). Interestingly, the expression of genes involved in DDR and DNA repair is regulated during the cell cycle and shows the highest expression at the G1/S transition and then decreases. The decrease in expression of these genes then depends on the transcriptional repressive function of E2F7 (40). Another interesting fact is that E2F7 levels are regulated at the posttranslational level during the cell cycle by ubiquitination and subsequent degradation after interaction with cyclin F, a substraterecognizing component of the S phase kinase-associated protein 1 (SKP1)-cullin 1 (CUL1)-F-box protein complex (SCF). SCF is a multiprotein E3 ubiquitin ligase that controls the transition between G1/S and G2/M phases and regulates the cell cycle by targeting a number of key cell cycle regulators for proteasomal degradation (41). Knockdown of cyclin F or the E2F7 mutant, which cannot interact with cyclin F and therefore remains stable during G2 phase, results in a delay of the G2/M transition accompanied by decreased expression of genes involved in DNA repair (42). Based on our results, we conclude that the cell cycle perturbations and lower dynamics of damaged DNA repair observed in HIC1-deficient cells are possibly caused by dysregulated E2F7 expression. More specifically, E2F7 accumulation interferes with G2/M progression and/or affects the expression of genes involved in recovery after DNA damage. Why do p53-dependent genes dominate the expression signature of HIC1 siRNA-treated primary cells? Especially considering that the loss of HIC1 increases the level of SIRT1, which deacetylates (and inactivates) p53 (4). However, some recent results suggest (in agreement with our observation) that inhibition of HIC1 leads to an increase in total p53 levels, possibly accompanied by activation of the p53-dependent response (8,38). It should be noted that analysis of the transcriptional regulatory function of HIC1 is complicated by the fact that the genes regulated by HIC1 and p53 (partially) overlap (8). Therefore, without detailed analysis, it is not possible to confirm that a "gene of interest" is indeed the target of HIC1. It is also evident that p53 in the 'first line' blocks the transformation of HIC1-deficient cells. It should be emphasized that the human HIC1 and TP53 genes (the latter gene encodes p53) are located in the same chromosomal region. Since the region is rearranged or lost in many human cancers, the condition of loss of both genes is often met. In addition, the exact nature of the cellular stress that triggers the p53-dependent response in human cells with suppressed HIC1 gene expression remains to be determined. Finally, the HIC1 expression is upregulated by E2F1 (43), indicating a more complex interplay between HIC1 and E2F cell cycle regulators. Our data suggest that exploitation of the HIC1-E2F7 relationship may influence the sensitivity of cancer cells to treatment.

Availability of data and materials
All data generated or analyzed during this study are included in this published article (and in the supplementary files). Raw expression profile data were deposited in the ArrayExpress database (https://www.ebi.ac.uk/arrayexpress/); accession number: E-MTAB-10678.      E2F7 expression in cells treated with the solvent (DMSO) only were set to 1; immunoblotting with an anti-β-ACTIN antibody was used as a loading control. (E) Western blot analysis of p53 protein in HIC1-deficient and control HC cells. Cells were treated with etoposide (80 µM) for 24 hours or left untreated. Whole-cell lysates were immunoblotted with an antibody that recognizes the entire pool of p53 protein (p53) or its transcriptionally active form phosphorylated on serine 15 (P-p53); the asterisk indicates a nonspecific band; CtBP, loading control; 53BP1, TP53-binding protein 1;