Advertisement
Cancer "-omics"

Identification of De Novo Enhancers Activated by TGFβ to Drive Expression of CDKN2A and B in HeLa Cells

Yen-Ting Liu, Lin Xu, Lynda Bennett, Jared C. Hooks, Jing Liu, Qinbo Zhou, Priscilla Liem, Yanbin Zheng and Stephen X. Skapek
DOI: 10.1158/1541-7786.MCR-19-0289 Published September 2019

Abstract

Disruption of the CDKN2A (INK4A/ARF) and B (INK4B) genes, which encode three function-independent tumor suppressors, is one of the most common events in human cancer. Because their relative importance in tumor prevention appears to be species- and context-specific, studying their regulation can shed light on mechanisms by which they are bypassed in malignant transformation. We previously unveiled a new pathway in which TGFβ selectively induces Arf at mouse Cdkn2a in eye development and cultured fibroblasts. As TGFβ signaling is often derailed in cancer development or progression, we investigated its control of CDKN2A/B in human cancer. Computational analyses of sequencing and array data from nearly 11,000 patients with cancer in TCGA showed discordant expression of ARF and INK4A in most cancer subtypes, with gene copy-number loss and promoter methylation involved in only a subset. Using HeLa cells as a model, we found that exogenous TGFβ induced ARF mRNA and protein, and ARF knockdown limited TGFβ-mediated growth suppression. TGFβ-mediated ARF mRNA induction required SMAD2/3, p38MAPK, and SP1, and ARF mRNA was induced without added RNAPII recruitment. Chromatin immunoprecipitation unveiled a remote enhancer element engaged by TGFβ by a mechanism that partially depended on p38MAPK. CRISPR-based editing of this enhancer limited induction of ARF and INK4B by TGFβ, but not by oncogenic RAS.

Implications: Our findings reveal new molecular mechanisms by which CDKN2A/B regulation is coupled to external cues, and those findings represent entry points to further explore pharmacologic strategies to restore their expression in cancer.

Introduction

Mammalian cells utilize a number of mechanisms to prevent cancerous transformation. By the time a cancer is clinically evident, genetic modifications or changes in gene expression provide a record of the barriers that were overcome during the neoplastic transformation. CDKN2A and B genes, evolutionarily conserved at human chromosome 9p21, encode three tumor suppressors (p14ARF, p16INK4A, and p15INK4B) that are commonly disrupted in cancer (1–3). CDKN2A generates p16INK4A and p14ARF (p19Arf in the mouse), which more or less directly activate what many argue are the two most important tumor suppressor effectors: RB and p53, respectively. CDKN2B encodes p15INK4B, also a positive regulator of RB.

The potential economy of colocalizing several tumor suppressors at this locus can have its pitfalls. CDKN2A is often deleted in many different types of cancer, and that deletion can include CDKN2B (4). When not deleted, expression of the CDKN2A/B genes is often silenced. Much effort has gone into understanding mechanisms underlying the inactivation of these genes. Established molecular mechanisms mainly address regulation at INK4A/ARF promoters, which include hypermethylation mediated by DNA methyltransferases (5, 6), and chromatin modification by Polycomb Group repressors, EZH2 (7), Bim-1 (8), and CBX7 (9). Emerging evidence suggests that a gene desert area that lies upstream of CDKN2A/B genes on human chromosome 9p21 is involved in their regulation, too. Whether that DNA segment controls CDKN2A/B gene expression by a trans-acting factor, like the long noncoding RNA, ANRIL (10), or by cis-acting enhancers (11), is not yet clear.

Beyond the global regulation of this locus, ARF and INK4A/B expression can be controlled as separate transcriptional units. Some of this work stems from insight gleaned from mouse models. For example, a number of genetically engineered mouse models show that Arf is required for normal eye development. Germline deletion of Cdkn2a exon 1β, unique to p19Arf (12), or exon 2, which encodes both p19Arf and p16Ink4a, or genetic changes silencing Arf expression (13, 14), results in hyperplastic expansion of cells in the developing primary vitreous, rendering the animals blind (12, 15). Pursuing the molecular basis for this developmental eye disease revealed a previously unrecognized signaling pathway that selectively controls Arf transcription in which extracellular cues from Tgfβ2 are essential for Arf expression and proper mouse eye development (13). Disrupting signaling by loss of this ligand or loss of Tgfβ receptor II recapitulates the eye disease found in Arf−/− mice (13). Furthermore, analyses of cultured mouse embryo fibroblasts (MEF) show that Tgfβ signaling leads to (i) rapid Smad2/3 phosphorylation and binding 5′ to Arf exon 1β; (ii) delayed recruitment of Sp1 and RNA polymerase II (RNAPII) to the Arf promoter; and (iii) p38Mapk activation. Moreover, Smad2/3 and p38Mapk activation and Sp1 binding are essential for induction of Arf mRNA and p19Arf protein (16, 17), and this signaling pathway has little effect on Ink4a expression in cultured MEFs (13). The ability to induce Arf in MEFs is shared by all three canonical TGFβ proteins (16).

Beyond this and other roles for TGFβ family members in normal mouse development (18), these proteins are well-known to influence human cancer formation and progression. Intriguingly, members of this family can contribute to either tumor suppression or promotion according to the stage of cancer progression but also cancer subtype (19). Nevertheless, efforts to elucidate the mechanisms of TGFβ signaling in tumorigenesis, especially on impeding tumor growth (20–22) or mediating prometastatic switches (23–25), further highlight the potential strategies in cancer treatment. Several proposed effectors of TGFβ-driven cancer suppression include cell-cycle regulators: p15INK4B (26), p21Cip1 (27), and p57Kip2 (28). The capacity for Tgfβ2 to induce Arf in the mouse raises the question of whether p14ARF might be also be a component of tumor suppression guided by TGFβ in human cells. We sought to evaluate this and to further explore whether TGFβ controlled these genes via a distant, cis-regulatory element.

Materials and Methods

Cell culture and treatment

Cancer cell lines PC3, HCT116, HeLa, SaoS2, U2OS, U87, MCF7, and MDA-MB-231 were all derived from ATCC. Cervical cancer cell lines siHa, C33A, and CaSki were provided by Dr. K. Alexander (University of Chicago, Chicago, IL). Rhabdomyosarcoma cells lines Rh18 and Rh30 were provided by P. Houghton (St. Jude Children's Research Hospital, Memphis, TN). Lung cancer cell lines H290 and H1666 were provided by Dr. J. Minna (UT Southwestern Medical Center, Dallas, TX). Cells were cultivated in RPMI or DMEM with 10% FBS supplemented with penicillin/streptomycin, or in medium recommended by ATCC. For TGFβ induction experiment, TGFβ1, 2, or 3 (5 ng/mL; R&D Systems) or an equivalent volume of vehicle (4 mmol/L HCl containing 1 mg/mL BSA) was added into the medium for the indicated time before harvesting the cells. When not otherwise specified in the text, TGFβ indicates TGFβ1. In some cases, a chemical inhibitor such as Actinomycin D (0.05 μg/mL; Sigma), SB431542 (10 μmol/L; Tocris Bioscience), or SB203580 (20 μmol/L; EMD chemicals) was added to the medium before or together with TGFβ. For stable ectopic gene expression, MSCV-based retrovirus vectors encoding ARF, H-RASV12, and GFP were produced and virus production and transduction were carried out as described previously (29).

The Cancer Genome Atlas genomic data analyses

We obtained The Cancer Genome Atlas (TCGA) somatic mutation, copy-number variation (CNV), DNA methylation, and RNA-seq data from Firehose (2016_01_28) at the Broad Institute (https://confluence.broadinstitute.org/display/GDAC/Home/). Custom PERL scripts were generated to compile and calculate the frequency of protein-altering somatic mutation, copy-number loss, and promoter methylation events on ARF gene directly from Firehose data. RNA-seq analysis and the following graph production were conducted using R programming packages. Gene expression between different cancer subtypes was normalized by ComBat (https://rdrr.io/bioc/sva/man/ComBat.html). The generation of heatmap and box and whisker plot were carried out by “pheatmap” package and “boxplot” function, respectively.

Western blotting

Western blotting was performed as described previously (29) with primary antibodies directed against p14ARF (sc8613, Santa Cruz Biotechnology; cs2407, Cell Signaling Technology), SMAD2/3 (cs3102, Cell Signaling Technology), phospho-SMAD2 (cs3101, Cell Signaling Technology), p38 (cs9212, Cell Signaling Technology), phospho-p38 (cs9211, Cell Signaling Technology), HSC70 (sc1059/sc7298, Santa Cruz Biotechnology), or HRAS (ab55391, Abcam). Chemiluminescence detection of proteins was visualized either by exposure to film or through ChemiDoc XRS+ imaging system (Bio-Rad).

Quantitative RT-PCR

Total RNA extraction and cDNA preparation were carried out as described previously (14). Quantitative mRNA analysis was performed using KAPA SYBR FAST qPCR Master Mix (Kapa Biosystems) and the CFX96 Touch System (Bio-Rad) for real-time PCR detection. Sequence for qRT-PCR primers is listed in Supplementary Table S1. Results are pooled from at least three individual samples.

siRNA

siRNAs used for gene knockdown in this article are listed in Supplementary Table S1. The knockdown experiments were either assayed at 48 hours after the initial transfection or combined with treatment of vehicle or TGFβ after 24 hours of the initial transfection. Cells then harvested for RT-PCR, Western blotting analysis, or used for cell proliferation assay.

Cell proliferation assay

Cell proliferation assay was carried out using either Cell Proliferation Kit I (MTT; Sigma-Aldrich) or CellTiter-Glo Assay Kit (Promega) according to the manufacturer's instructions. Data are presented as pooled data from six replicates and plotted as fold change in cell number from plating to harvest.

ChIP-qPCR and ChIP-seq

Chromatin immunoprecipitation (ChIP) was carried out as described previously (16). For immunoprecipitation, antibodies against H3K27Ac (ab4729, Abcam), SMAD3 (ab28379, Abcam), or RNAPII (sc899, Santa Cruz Biotechnology) were used in this article. Quantitative analysis of the precipitated and input DNA through real-time PCR was described above. The primer sets for ChIP-qPCR are listed in Supplementary Table S1. Results are pooled from at least three individual samples.

For ChIP-seq, the precipitated DNA was followed library generation (TruSeq ChIP Library Preparation Kit, Illumina) and sequencing (NextSeq500 system, Illumina) performed by DNA Link, Inc. The generated reads were aligned to the human reference genome (hg19) using MACS2 (30). BAM files for ChIP-seq data of RNAPII in multiple cell lines were downloaded from ENCODE (https://www.encodeproject.org/).

GRO-seq analysis

Sequence Read Archive (SRA) files for GRO-seq data in HeLa cells were downloaded from SRA repository with accession number SRP050563. FASTQ files were converted from SRA files by SRA Toolkit for CentOS Linux 64 bit architecture (https://www.ncbi.nlm.nih.gov/sra/docs/toolkitsoft/). BAM files were generated by aligned reads from FASTQ files through Bowtie2 with default settings (31). Figure was plotted by IGV (32).

Generation of genomic deletions through CRISPR/Cas9

CRISPR-based genomic deletion in HeLa cells was carried out and modified according to the protocol described in Bauer and colleagues (33). Briefly, pairs of sgRNAs targeting the boundaries of desired deletion region were designed through the online prediction tool (34). Lentiviral expression vectors containing both the Cas9 and the sgRNA (targeted sgRNA or control sgRNA) were prepared and used as described previously (35). After transduction of a pair of lentiviruses with similar titers and puromycin selection for at least 7 days, the deletion-bearing bulk cells were subjected to two consecutive, 96-well format screen with 10 cells/well in the first round and 1 cell/well in the second round, in order to obtain homozygous deletion clones. PCR validation of deletion-bearing bulk cells and the homozygous deletion clones was performed with primer sets either flanking or internal to the targeted deletion region. The deletion clones were further validated through Sanger sequencing before functional assays. The sequences for sgRNA cloning and PCR deletion screen are listed in Supplementary Table S1.

Statistical analysis

Quantitative data are mainly presented as column bar graph with either mean ± SD or mean ± SEM from at least three representative experiments, except box and whisker plots in Fig. 1B and C. Statistical significance (P < 0.05) between two populations was determined by Student t test.

Figure 1.
Figure 1.

CDKN2A regulation across human cancer and cultivated cell lines. A, Heatmap showing the differential expression of ARF and INK4A in a subset of cancers in which the percent of cases with CDKN2A copy-number, mutation, or promoter methylation is ranging from high to low. The correlation of ARF and INK4A expression (r) is calculated and listed at side. B, Box and whisker diagrams show that promoter methylation (excluding cases with copy-number loss and mutation) significantly correlates with ARF expression in STAD, but not in LUAD and COAD. P < 0.05 was considered significant. C, The box and whisker diagram showing the ARF expression without cases regulated by copy-number loss, nonsynonymous somatic mutation, and promoter methylation in representative cancer subtypes. D, Cell proliferation assays of indicated cell lines transduced with GFP- or ARF-expressing retrovirus. E, Schematic diagram showing CDKN2A locus with overlapping ARF and INK4A gene products. F, Representative Western blot showing basal levels of p14ARF in a collection of human cancer cell lines. The PCR image showing the genome integrity of ARF locus of respective cancer cell lines are listed below. The primer sets (red arrows) for detecting genome integrity are located within the ARF exons and indicated in the schematic diagram in E. CN, copy number; SKCM, skin cutaneous melanoma; GBM, glioblastoma; ESCA, esophageal carcinoma; STAD, stomach adenocarcinoma; LUAD, lung adenocarcinoma; COAD, colon adenocarcinoma; PCPG, pheochromocytoma and paraganglioma; THCA, thyroid carcinoma; LAML, acute myeloid leukemia. Whisker bars, min to max; error bars, ±SD (*, P < 0.05, ARF vs. GFP).

Results

CDKN2A regulation across human cancer

To begin to study the role of the TGFβ–ARF pathway in human tumor suppression, we took advantage of publicly available data to describe how ARF expression or p14ARF function may be impaired across a wide spectrum of human cancers studied in TCGA (see Materials and Methods). We found that gene copy-number losses are relatively more common than nonsynonymous somatic mutation or ARF promoter methylation across different histologic subtypes of cancers (Supplementary Fig. S1A), and at least one of these three variables was evident in approximately 70% of cases of cutaneous melanoma (SKCM), glioblastoma multiforme (GBM), and esophageal cancer (ESCA). In contrast, less than 5% of cases of acute myelogenous leukemia (LAML), thyroid cancer (THCA), or pheochromocytoma and paraganglioma (PCPG) harbored these molecular changes (Supplementary Fig. S1B). In certain histologic subtypes, such as GBM, ESCA, bladder cancer (BLCA), lung adenocarcinoma (LUAD), cholangiocarcinoma (CHOL), and diffuse, large B-cell lymphoma (DLBC), over 10% of cases harbored both copy-number losses and promoter methylation.

We selected a subset of cancer types to cover a spectrum with relatively high (e.g., SKCM, GBM, and ESCA), moderate [e.g., stomach adenocarcinoma (STAD), LUAD, and colon adenocarcinoma (COAD)], or low (e.g., PCPG, THCA, and LAML) fractions of cases documented to have genomic alterations of CDKN2A for further analysis. We found that ARF expression did not strongly correlate with presence of gene copy-number losses or promoter methylation (Fig. 1A). For example, cancer subtypes with the lowest frequency of these modifications (LAML, THCA, and PCPG) also had a higher percentage of cases with lower ARF expression (Fig. 1A, blue color) when compared with SKCM, GBM, and ESCA–subtypes with the highest frequency of copy-number loss and/or promoter methylation. To further understand the impact of promoter methylation, we focused only on cases without detectable gene deletion or mutation. In certain groups, such as stomach cancer (STAD), ARF expression was significantly lower in cases with promoter methylation, but in LUAD and COAD, ARF expression was not significantly associated with the presence or absence of methylation (Fig. 1B). The inconsistent relationship between ARF expression and gene copy-number loss and promoter methylation implied that expression of this particular tumor suppressor was governed by other mechanisms. Indeed, when considering ARF expression across different cancer subtypes and excluding cases with deletion/mutation/promoter methylation, we still observed over a 10-fold variation in ARF expression between cancer subtypes with relatively high (e.g., GBM, ESCA and SKCM) and low expression (e.g., THCA and LAML; Fig. 1C). This indicated the existence of mechanisms controlling ARF regulation beyond gene copy-number and promoter methylation. We also investigated the coregulation of ARF with either INK4A and INK4B across these histologic subtypes (Fig. 1A; Supplementary Fig. S1C). GBM showed the strongest correlation between expression of ARF and INK4A (r = 0.9) and, in COAD, the regulation of the two genes was completely uncoupled (r = 0.06). In all other cancer subtypes, the expression of these two genes was loosely correlated (r = 0.5–0.6). The analysis of ARF and INK4B showed similar levels of coregulation in these cancer subtypes.

Taken together, we conclude that ARF expression or p14ARF function can be impaired by gene copy-number loss, promoter methylation, and protein-altering somatic mutations to varying extents across human cancer. However, gene copy-number loss and ARF promoter methylation does not completely explain silencing of either of the two transcripts encoded by CDKN2A. Moreover, the expression of these two transcripts, together with INK4B, is independently controlled in many different human cancers.

TGFβ increases ARF mRNA and protein in HeLa cells

We next sought to identify cancer-derived cell lines in which ARF expression could be enhanced by TGFβ, one of two extracellular signals previously implicated in its regulation (13, 36), as that could serve as an entry point to pharmacologic strategies to restore its expression. The functional importance of such a manipulation was evidenced by the fact that ectopic expression of p14ARF in human lung cancer (H290, H1666) and rhabdomyosarcoma (RH18) cell lines decreased cell accumulation (Fig. 1D). We initially interrogated a number of human cancer cell lines to find a subset in which exons 1β, 2, and 3 were intact, but displayed little to no detectable expression of p14ARF (Fig. 1E and F). Interestingly, TGFβ failed to enhance ARF expression in those lines (negative data not shown), and this prompted a larger search for responsive human cancer cell lines.

We next leveraged Encyclopedia of DNA Elements (ENCODE) data to find cell lines with evidence for basal transcriptional activity at CDKN2A based on chromatin immunoprecipitation-sequencing (ChIP-seq) for RNAPII binding to CDKN2A promoters. Our analysis of 10 cell lines showed that HeLa cells had the greatest RNAPII binding near exon 1β, the first coding region for p14ARF (Fig. 2A). Indeed, HeLa cells represented a good candidate as CDKN2A/B coding sequence is intact and there is little DNA methylation in CpG islands at promoter regions of CDKN2A/B transcripts (ref. 37; Supplementary Fig. S2). HeLa cells also show a high level of aneuploidy with an average of approximately 3 copies of CDKN2A/B locus (37). In agreement with baseline activity and presumed accessibility for further induction by TGFβ, we observed increased p14ARF in those cells following a 24-hour exposure to TGFβ1 (Fig. 2B), and that TGFβ1, 2, and 3 shared this capacity, consistent with previous findings in cultured MEFs (ref. 16; Supplementary Fig. S3A). In contrast, HCT116, which showed little RNAPII binding around ARF exon 1β (Fig. 2A), had no detectable p14ARF increase following TGFβ treatment at 48 hours, even though the gene was intact (Fig. 2B; Supplementary Fig. S3B). Given HeLa cell derivation, we investigated several other cervical cancer cell lines, but none showed p14ARF induction, even though p14ARF was detected at baseline (Supplementary Fig. S3C). The TGFβ-driven ARF induction in HeLa cells does not appear to be a direct effect of virus transformation (38) as it was not affected by the knockdown of HPV18 E7 gene (Supplementary Fig. S3D).

Figure 2.
Figure 2.

TGFβ induces ARF expression in HeLa cells. A, Analysis of RNAPII ChIP-seq data suggests cell lines with active gene activity around ARF locus. B, Representative Western blot analysis for the indicated proteins using lysate from HeLa and HCT116 cells treated with vehicle or TGFβ for 24 or 48 hours, as indicated. C, Cell proliferation assay showing TGFβ treatment decreases HeLa cell accumulation over 10 days of TGFβ treatment and such a decrease was partially reversed by ARF knockdown. D, Quantitative analysis of mRNA expression of the indicated genes in HeLa cells treated with TGFβ for 24 or 48 hours, as indicated. mRNA expression, measured by qRT-PCR and normalized to GAPDH, is presented as fold change over expression in vehicle-treated cells at 24 hours. qRT-PCR analysis ARF mRNA using total RNA isolated from HeLa cells treated with vehicle or TGFβ for 48 hours after 24-hour transfection with either siRNA control (Scramble), or siRNA targeting SMAD4, p38MAPK (E), or SP1 (F) as indicated. Right, the knockdown efficiency of corresponding targeted siRNAs is shown. V, Vehicle; T, TGFβ. Error bars, ±SD (*, P < 0.05, T vs. V; #, P < 0.05 siRNA targets vs. siRNA control).

We assessed the functional consequences of TGFβ induced p14ARF in vitro. Exposure of HeLa cells to TGFβ repressed cell accumulation by approximately 40%, and concurrent ARF knockdown significantly diminished that effect (Fig. 2C). We conclude that the capacity for TGFβ to induce ARF correlated with RNAPII binding at baseline to p14ARF in HeLa cells and imposes an ARF-dependent slowing of cell accumulation; however, baseline expression levels and a cervical cancer lineage was not sufficient to explain the TGFβ responsiveness.

We explored mechanisms of ARF regulation by TGFβ, as mouse p19Arf is principally controlled by transcriptional induction (13, 39, 40). Following exposure to TGFβ, ARF mRNA increased at 24 and 48 hours (Fig. 2D), and the increased ARF mRNA was detected as early as 6 hours after TGFβ treatment, at which time p14ARF was not measurably elevated (Supplementary Fig. S4A and S4B). After exposure to actinomycin D, which impairs mRNA synthesis (41), TGFβ failed to induce ARF mRNA, even though there was no overt evidence of toxicity to the cells (Supplementary Fig. S4C and S4D). We also noted that, as in cultured MEFs, TGFβ differentially regulated expression of other genes at this locus: INK4A mRNA was not significantly changed by TGFβ, whereas the expression of INK4B was induced, consistent with a previous report (ref. 26; Fig. 2D). Hence, TGFβ-mediated induction of p14ARF depends on increased ARF transcription and, while ARF induction parallels INK4B, INK4A expression is not controlled by TGFβ in this context.

Signaling pathways and RNA polymerase II binding dynamics influenced by TGFβ in HeLa cells

TGFβ activates both SMAD-dependent and independent pathways, the latter of which includes p38MAPK (42). We previously established that both pathways are needed for TGFβ-dependent induction of mouse Arf (16). In HeLa cells, ARF induction by TGFβ also correlated with phosphorylation of SMAD2, and knockdown of SMAD4, an obligate heterodimeric partner for SMAD2 and SMAD3 (43), or chemical inhibition of the SMAD2 regulator, TGFβ receptor I, impaired ARF induction (Fig. 2B and E; Supplementary Fig. S5A–S5C). Similarly, p38MAPK knockdown or chemical inhibition with SB203580 (SB20) limited ARF and p14ARF induction (Fig. 2E; Supplementary Fig. S5A–S5C). Interestingly, phospho-p38MAPK was readily detected at baseline, treatment of TGFβ not only showed no effect on its levels at 24 hours but even showed diminished activity at 48 hours (Fig. 2B), suggesting that levels of this phosphoprotein were sufficient at baseline, and the activation of the pathway may only be required in the early stages of ARF induction. The transcription factor SP1 was previously implicated in MEFs as a positive regulator of TGFβ-driven Arf induction (17). In HeLa cells, ARF mRNA and p14ARF induction by TGFβ was significantly decreased when transiently knocking down SP1 (Fig. 2F; Supplementary Fig. S5D and S5E).

We previously reported an approximately 24-hour delay between immediate Smad2/3 activation by TGFβ at 1.5 hours and measurable induction of Arf mRNA at 24 hours in MEFs, and we found a temporal correlation between Arf mRNA increase and RNAPII binding to the promoter (16). This suggested that RNAPII recruitment was limiting in the induction of mouse Arf by TGFβ. Using MEFs, we confirmed our ability to detect RNAPII recruitment by showing higher RNAPII binding in p53−/− MEFs than in wild-type controls, consistent with higher basal Arf expression in absence of p53 (44); and we detected a further increase in RNAPII binding following exposure of p53−/− MEFs to TGFβ (Fig. 3A and data not shown). In contrast, we could not detect a change in RNAPII binding to the human ARF promoter following TGFβ stimulation (Fig. 3B). Our finding that additional RNAPII recruitment was not correlated with increased human ARF mRNA suggested that the polymerase might be paused at baseline in HeLa cells. However, reanalysis of published ChIP-seq and GRO-seq data (45) was not consistent with RNAPII release from a paused state at baseline (ref. 46; Fig. 3C). We conclude that signaling pathways by which TGFβ induces ARF expression are conserved in MEFs and HeLa cells, but the dynamics of RNAPII recruitment and kinetics of the transcriptional response differs in the two models. In HeLa cells, the increased ARF expression in response to TGFβ does not represent “release” from a paused state, but likely represents increased RNAPII processivity that was evidenced as a trend toward increased RNAPII detected within intron 1 (Fig. 3B).

Figure 3.
Figure 3.

TGFβ fails to further recruit RNAPII to the ARF locus in HeLa cells. Quantitative analysis of representative ChIP assays by using antibodies specific to RNA polymerase II (RNAPII). ChIP was performed using either p53-null MEFs exposed to vehicle or TGFβ for 0, 48 hours (A) or HeLa cells exposed to vehicle or TGFβ for 24 hours (B). The results from A and B suggested that inactivation of p53 in HeLa cells does not account for differential RNAPII dynamics in two different models. Immunoprecipitated DNA and input DNA were amplified with primers for genomic regions indicated in the respective schematic diagrams above. C, Reanalysis of GRO-seq and RNAPII ChIP-seq around CDKN2A/B region in HeLa cells suggests an active but not paused gene activity followed the classification described previously. Representative diagrams of different types of gene activities and their corresponding pattern of GRO-seq are shown above. Error bars, ±SEM (*, P < 0.05, T vs. V).

TGFβ activates de novo distant enhancers upstream of CDKN2A/B genes

Given the importance of transcriptional induction of ARF by TGFβ in human cancer cells, we sought to identify enhancer elements that might be engaged by TGFβ. We focused on a large gene desert at chromosome 9p21 previously identified by genome-wide association studies as a coronary artery disease (CAD) risk interval (47). Deletion of a 70-kb orthologous region on mouse chromosome 4 (Chr4Δ70kb) severely depressed the Cdkn2a/b expression in the heart and other tissues, leading to increased mortality on a high-fat, high-cholesterol diet and to cancer predisposition (11). The animals also exhibited persistent hyperplastic primary vitreous, an ocular development defect also found in Arf−/− mice (12, 15). Indeed, the effect of the 70-kb deletion is not limited to Cdkn2a/b regulation: our RNA-seq analysis of cultured Chr4Δ70kb/Δ70kb MEFs demonstrated the dampened basal expression of seven genes, regardless of their transcriptional orientation, from the targeted region through approximately 600 kb toward the centromere (Supplementary Fig. S6), suggesting that the deleted DNA segment included cis-acting regulatory element with far-reaching effects.

Because TGFβ-mediated induction of Arf and Ink4b is also blunted in the Chr4Δ70kb/Δ70kb MEFs (14), we used ChIP-seq for acetylation at H3K27 to search for molecular evidence of enhancers that were altered by TGFβ (48). Comparing TGFβ-treated to untreated samples showed no visible differences in H3K27Ac modifications within the CDKN2A/B genes (Fig. 4A). However, approximately 110 kb distal to INK4B, at least three peaks emerged following TGFβ stimulation: one as a 10-fold enrichment from a previous peak and two de novo peaks (denoted as peaks 1, 2, and 3; Fig. 4A). We noted that the peaks did not overlap with the long noncoding RNA, ANRIL, a reported trans-acting factor with diverse gene regulation functions (10). The enrichment of H3K27Ac at these peaks, but not in adjacent DNA, was confirmed by H3K27Ac ChIP-qPCR (Fig. 4B, blue vs. red solid lines). Given that p38MAPK knockdown or chemical inhibition blunted ARF induction by TGFβ, we tested whether this kinase acted proximal to these apparent enhancers. H3K27Ac ChIP-qPCR for the three peaks showed diminished, but still significant, remodeling by TGFβ, even after pretreatment with SB20 for 16 hours (Fig. 4B, blue vs. red dashed lines). Importantly, SB20 did not generally decrease the baseline H3K27Ac signal at this DNA segment (Fig. 4B, blue dashed vs. solid lines). We also tested whether SMAD3 was recruited to these peaks. Indeed, 24 hours following TGFβ induction, ChIP-qPCR revealed increased SMAD3 binding at all three peaks, but especially at peak 3, and decreased SMAD3 binding between the peaks (Fig. 4C). Overall, our results revealed a previously unrecognized capacity for TGFβ to recruit SMAD3 to a preexisting and de novo candidate cis-enhancer elements distal to CDKN2A/B, in a manner that partially depended on p38MAPK activity.

Figure 4.
Figure 4.

The TGFβ-driven, de novo enhancer regions at chromosome 9p21 in HeLa cells. A, H3K27Ac ChIP-seq signal profiles from HeLa cells treated with two individual vehicle or TGFβ for 24 hours. The normalized ChIP-seq read coverage across the chromosome 9p21 region is presented as wiggle plots, and the corresponded CDKN2A/B locus together with the long noncoding RNA ANRIL are schemed below. The positions of the de novo peaks following TGFβ stimulation (denoted as P1–3) and from here the so-called suspected enhancer regions (highlighted with green line) are indicated in diagram or plots from A to C. B, H3K27Ac ChIP-qPCR data were obtained with primer sets designed in 1-kb intervals across the suspected enhancer peaks and nearby regions. The ChIP assays were performed in HeLa cells treated with vehicle or TGFβ for 24 hours following overnight preincubation with either p38 inhibitor (SB203580, 10 μmol/L; referred as SB20) or DMSO. Pull-down signal is presented as relative pull down normalized against signal obtained at coordinate: 22121 k in “V(DMSO+)” group. C, SMAD3 ChIP-qPCR across the suspected enhancer peaks in HeLa cells treated with vehicle or TGFβ for 24 hours. Pull-down signal is presented as relative pull down normalized against signal obtained at coordinate: 22121 k in vehicle-treated cells. V, Vehicle; T, TGFβ. Error bars, ±SEM [*, P < 0.05, T (DMSO+) vs. V (DMSO+); #, P < 0.05, T (DMSO+) greater than T (SB20+); §, P < 0.05, T vs. V].

Deletion of candidate distal enhancers blunts the ability of TGFβ to induce ARF/INK4B expression

We sought to confirm that segment of DNA contained TGFβ-induced enhancers by using CRISPR/Cas9-mediated DNA editing (33, 35). Recognizing that additional copies of chromosome 9 in HeLa cells (37) could enhance the difficulty of gaining complete deletion, we designed pairs of single guide RNA (sgRNA) targeting the boundaries of the 20-kb region involving peaks 1–3 (ΔChr9:22121-22141K, Δ20 kb) and cloned them into the lentiCRISPR v2 lentiviral plasmid vector also encoding Cas9 (35). The pair of sgRNAs was cointroduced into HeLa cells, and separate puromycin-resistant clones were verified by PCR to harbor complete or nearly complete deletion of the region (Fig. 5A; Supplementary Fig. S7A).

Figure 5.
Figure 5.

Verification of the TGFβ-mediated enhancer region through HeLaΔ20kb cells. A, HeLaΔ20kb cell clone was generated through CRISPR/Cas9-mediated homologous recombination. The predicted digestion sites according to the design of sgRNAs are highlighted in the schematic diagram. To validate the clones with desired deletion, PCR was performed with primer sets either flanking (red arrows) or internal (blue arrows) to the 20-kb deletion region as indicated in the schematic diagram, and is shown as representative gel. Compared with the control strains, the deletion clones can only amplify flanking but not internal signal, suggesting the homozygous phenotype on genome deletion. B, Quantitative analysis of mRNA expression of the indicated genes in HeLaΔ20kb cells (denoted as “Δ20 kb”) treated with TGFβ (T) for 24 hours, in comparison with one from control strains (denoted as “NO sgRNA” and “Gal4 sgRNA”). mRNA expression is presented as fold change overexpression in vehicle (V)-treated cells at 24 hours. Reanalysis of Hi-C data from HeLa cells to address the interaction between Δ20 kb region and ARF promoter (C), INK4B promoter (D), and INK4A promoter (E). As the schematic diagram indicated above, the reference regions chosen for analysis, including Δ20 kb region, are around 10 kb for each and 45 kb away from each other. The average interaction strength from each region was further normalized to one from Region #1. I, Internal; F, Flanking. Error bars, ±SEM [*, P < 0.05, “Δ20 kb” vs. “No sgRNA” and “Gal4 sgRNA”].

We used clones with complete (HeLaΔ20kb) or partial deletion (HeLaΔ20kb-PD) and two controls, the parental HeLa cells (no sgRNA) and pooled HeLa cells transduced with sgRNA targeting Gal4 (Gal4 sgRNA), in assays measuring CDKN2A/B expression in response to TGFβ. These experiments demonstrated that induction of both ARF and INK4B mRNA was blunted in HeLaΔ20kb and HeLaΔ20kb-PD cells, whereas INK4A expression was not affected by the deletion. Furthermore, SERPINE1 was induced approximately 6- to 10-fold in all tested HeLa lines, which confirmed that the transcriptional response to TGFβ was not generally corrupted in the experimental clones (ref. 49; Fig. 5B; Supplementary Fig. S7B). The physical interaction between the 20-kb region and the promoters for CDKN2A/B transcripts was supported by reanalysis of published data from chromosome conformation capture with high-throughput sequencing (Hi-C; ref. 50). That data showed enrichment for interactions between the 20-kb region and the promoter regions of all three transcripts in culture independently of exogenous TGFβ (Fig. 5C–E). We note that the interaction between the 20-kb region and the INK4A promoter, which is high at baseline, may be driven by FOXA1, which has a binding element within the 20-kb region and was previously reported to control INK4A expression (51). While confirming that the 20-kb DNA segment contains enhancer elements that can directly interact with CDKN2A/B genes, these data also reveal that the relative level of interaction with individual promoters at baseline does not necessarily correlate with relative induction of those transcripts by TGFβ.

The distant TGFβ-dependent enhancer does not control RAS-mediated induction of CDKN2A/B

The CDKN2A/B locus is also well known to be regulated by oncogenic signals from RAS (52). To test whether the 20-kb enhancer influenced CDKN2A/B induction by oncogenic RAS, we transduced HeLa cells with a retrovirus expressing H-RASV12 (Fig. 6A). After oncogenic RAS expression, we detected elevated ARF, INK4B, and INK4A with 1.5- to 2-fold changes compared with control (Fig. 6B). However, this induction was not impaired in the absence of the remote enhancer; indeed, induction of INK4B was even higher (Fig. 6B). This finding was consistent with H3K27Ac profiles following H-RASV12 expression: H3K27Ac ChIP-qPCR revealed no obvious difference in H3K27Ac signal within the 20-kb enhancer region in HeLa cells expressing either GFP or H-RASV12 (Fig. 6C).

Figure 6.
Figure 6.

Oncogenic stimulation on CDKN2A/B regulation is independent of TGFβ-driven enhancer region in HeLa cells. A, Representative Western blots showing HRAS expression using lysate from HeLa cells after transduction using GFP- or HRASV12-expressing retrovirus for 8 days. B, Quantitative analysis of mRNA expression of the indicated genes in HeLa and HeLaΔ20kb cells with ectopic expression of GFP or HRAS through retrovirus transduction, in comparison with parental cells (no vector). C, H3K27Ac ChIP-qPCR across the TGFβ-driven enhancers (highlighted with green line as indicated) and nearby regions in HeLa cells with ectopic expression of GFP or HRAS through retrovirus transduction. Pull-down signal is presented as relative pull down normalized against signal obtained at coordinate: 22121 k in GFP-expressing cells. Error bars, ±SEM (*, P < 0.05, HRAS vs. GFP or no vector).

Discussion

Our work significantly extends our understanding of how transcription of CDKN2A and CDKN2B genes is controlled (Fig. 7). Focusing first on TGFβ, which guides mouse Arf expression in a process that is required for proper mouse eye development and vision (12–14), we demonstrate that the signaling pathway, generally conserved between mouse and human cells (43), includes SMAD2/3 and p38MAPK activation as well as SP1 in both species. However, the kinetics of mouse and human ARF mRNA induction and RNAPII binding differ in early-passage MEFs and HeLa cells: Arf is essentially silenced in mouse embryos and early-passage MEFs (13, 53) and its induction requires recruitment of RNAPII to the promoter (16), whereas we show here that ongoing ARF transcription in HeLa cells positions them to respond to TGFβ without that step. Our computational and functional analyses identity a 20-kb DNA segment, lying approximately 110 kb and approximately 125 kb upstream of INK4B and ARF transcription start sites, respectively, and that element represents a cis enhancer engaged de novo in response to TGFβ. The magnitude of H3K27 acetylation at this site is dampened by p38MAPK inhibition, correlating with diminished ARF induction. Our chromatin conformation capture analyses reveal direct associations between this new enhancer and promoters of ARF, INK4A, and INK4B. We also show that the expression of ARF and INK4A is largely uncoupled in most cancer types, while ARF and INK4B share transcriptional induction by TGFβ in HeLa cells; hence, the three transcripts encoded by CDKN2A/B are controlled individually. We also demonstrate that the TGFβ-activated enhancer element does not measurably influence induction of these genes by oncogenic HRAS expression. Together, our findings add to our understanding of how an extracellular signaling protein controls the expression of individual genes at this important tumor suppressor locus, while also underlining gaps that still need to be closed.

Figure 7.
Figure 7.

A model for regulation of CDKN2A/B expression in cancers. For majority of cancers whose CDKN2A and B locus are not deleted, the expression of CDKN2A/B (either repressed or derepressed) is mainly controlled at the transcriptional level and focus on their promoter regions. For some cancers (HeLa cell model in our study), the limited expression of CDKN2A/B can further be activated by interacting with remote enhancers, probably due to chromatin remodeling at upstream 9p21 region. TAs, transcription activators; TRs, transcription repressors.

We recognize that the contributions that TGFβ makes to cancer is complex in that it can either foster or block tumor formation, and it acts on cancer cells as well as elements of the tumor stroma (54). Focusing just on the tumor cell, understanding how TGFβ induces CDKN2A/B expression represents an entry point for novel cancer therapeutics aimed at restoring expression of these genes. These genes are often deleted en bloc, but when not, restoring expression of either p14ARF or p16/p15INK4A/B could trigger apoptosis through p53 (52) or slow or arrest proliferation by mechanisms that may be independent of p53 (for p14ARF; ref. 55) or the RB1 gene (for INK4 proteins; ref. 56). The relative frequency by which CDKN2A/B is corrupted by gene copy-number loss or promoter methylation varies across different histologic subtypes, but virtually all human cancer subtypes contain a relatively large fraction of cases in which low expression is not fully explained by these two mechanisms. Defining a new enhancer element near these genes could help unravel silencing mechanisms that, in principle, might be reversible. Our findings highlight that expression of CDKN2A/B genes is governed by discrete elements, depending on whether the stimulus is cell-intrinsic (such as from oncogene expression) or extrinsic (TGFβ signaling). We define a distinct enhancer element required for the latter, but not the former. Interestingly, the 70-kb CAD orthologous region adjacent to mouse Cdkn2a/b genes, which partially overlaps the 20-kb region studied here, is similarly required for TGFβ-driven Arf induction in MEFs, but not for induction of this transcript by oncogenic RAS (14). Conceptually, ChIP-seq for enhancer-associated marks in presence or absence of oncogenic RAS could be utilized to identify other regulatory elements activated de novo by oncogenic signals in human cells. As we showed in HeLa cells, being able to chemically/pharmacologically restore ARF (and INK4A/B) expression may represent a novel therapeutic strategy for a variety of cancers.

In extending studies of mouse Arf regulation by TGFβ to human cancer cells, we were initially struck by the paucity of human cancer cells in which TGFβ enhanced ARF or INK4 gene expression. To date, in screening 15 human cancer cell lines, we found only HeLa cells to have robust ARF induction by TGFβ. This finding underscores the biological importance of tumor suppression exerted by TGFβ. Deleterious mutations of individual members of the TGFβ pathway, including TGFBR2, TGFBR1, and SMAD4, are evident in many cancers including colorectal and prostate cancer (57). Given the capacity for ARF and INK4 gene expression to hinder or arrest proliferation, disrupting CDKN2A/B could be a relatively early step in tumorigenesis. When this does not occur by gene deletion, the signaling pathway to CDKN2A/B genes could be disrupted. We have already highlighted a number of nodal points in that cascade, all the way to a 20-kb segment of noncoding DNA, abnormalities of which could underlie failed response to TGFβ in some cancers. As an HPV-generated human cancer (38), HeLa cells have an alternative mechanism to corrupt both RB and p53, leaving little selective pressure to otherwise derail TGFβ-ARF/INK4 signaling. Interrogating the numerous “nonresponsive” cell lines could reveal ways to restore their responsiveness. For example, initial reversal of DNA or histone methylation, long recognized as a silencing mechanism at CDKN2A/B, may enable TGFβ to induce ARF or INK4A/B in cancers where the other elements of the signaling pathway are intact.

The paucity of TGFβ “responsive” cell lines could also highlight an under-appreciated cell lineage dependency with respect to ARF expression. This is best evidenced in mouse development. Even though all three canonical TGFβ proteins induce mouse Arf in MEFs, and all three drive diverse facets of mouse embryo development (58), Arf is only detectable in perivascular cells enveloping the hyaloid vasculature in the primary vitreous and the internal umbilical arteries during development (12, 13, 29). Hence, there must be lineage-specific controls, such as cooperating transcriptional activators or histone or DNA-modifying enzymes that may limit accessibility of cis enhancers to TGFβ-driven activators. ChIP-seq analyses for enhancer marks in different tissues in the developing mouse may suggest a basis for responsiveness or nonresponsiveness of these genes to TGFβ during development. Those differences may also underlie the tissue-specific capacity for certain oncogenes to induce CDKN2A/B expression as an oncogene “sensor” (59). This concept can also help explain the cancer histology–specific mechanisms to bypass this tumor suppressor: selective pressure to delete CDKN2A/B should be lost in cell lineages where the expression is silenced by other means.

While our findings add to our understanding of how distant regulatory elements control ARF and INK4B expression in human cancer cells (Fig. 7), there are remaining questions. First, the identified remote enhancer elements are specific to TGFβ signaling but not for intrinsic oncogenic signaling, such as by oncogenic RAS. Second, the role for sequence-specific transcriptional regulators remains to be further elucidated, especially for individual CDKN2A/B transcripts. For example, SP1 is relevant for both ARF (ref. 17 and this study) and INK4B (60) induction by TGFβ in mouse and human cells. It is likely that those classes of transcriptional regulators can help to explain the induction of specific CDKNA/B transcripts even though our chromosome conformation analyses reveal broader contacts between the distant enhancer elements and those genes. While we focused on induction of these genes, deletion of the so-called CAD risk interval in MEFs (Chr4Δ70kb) showed lowered basal expression of Arf and Ink4a (11, 14), and we noticed lower basal ARF expression in HeLa cells lacking the 20-kb enhancer element (Y.-T. Liu and S.X. Skapek, data not shown). Finally, beyond cancer, it will be quite interesting to explore how yet-to-be defined elements might contribute to the CAD risk associated with SNPs in this region (47) and also the enhanced mortality in mice fed a high-calorie/carbohydrate diet (11).

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Authors' Contributions

Conception and design: Y.-T. Liu, L. Xu, Y. Zheng, S.X. Skapek

Development of methodology: Y.-T. Liu, J.C. Hooks, Y. Zheng, S.X. Skapek

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y.-T. Liu, L. Bennett, J.C. Hooks, J. Liu, P. Liem, Y. Zheng

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y.-T. Liu, L. Xu, L. Bennett, J.C. Hooks, J. Liu, Q. Zhou, Y. Zheng, S.X. Skapek

Writing, review, and/or revision of the manuscript: Y.-T. Liu, L. Xu, Y. Zheng, S.X. Skapek

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Liu, Y. Zheng

Study supervision: S.X. Skapek

Acknowledgments

The authors gratefully acknowledge technical assistance provided by J. Singh and R. Wilson, and many helpful comments by other members of the Skapek laboratory. We also acknowledge support from the NIH National Eye Institute (R01 EY019942), from the Cancer Prevention and Research Institute of Texas (RP120685-P2, RP180319), and from the UTSW Harold C. Simmons Comprehensive Cancer Center and from the National Cancer Institute (CA142543).

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.

Footnotes

  • Note: Supplementary data for this article are available at Molecular Cancer Research Online (http://mcr.aacrjournals.org/).

  • Mol Cancer Res 2019;17:1854–66

  • Received March 15, 2019.
  • Revision received May 1, 2019.
  • Accepted June 7, 2019.
  • Published first June 12, 2019.

References

  1. 1.↵
    1. Foulkes WD,
    2. Flanders TY,
    3. Pollock PM,
    4. Hayward NK
    . The CDKN2A (p16) gene and human cancer. Mol Med 1997;3:5–20.
    OpenUrlPubMed
  2. 2.↵
    1. Shintani S,
    2. Nakahara Y,
    3. Mihara M,
    4. Ueyama Y,
    5. Matsumura T
    . Inactivation of the p14ARF, p15INK4B and p16INK4A genes is a frequent event in human oral squamous cell carcinomas. Oral Oncol 2001;37:498–504.
    OpenUrlCrossRefPubMed
  3. 3.↵
    1. Simon M,
    2. Köster G,
    3. Menon AG,
    4. Schramm J
    . Functional evidence for a role of combined CDKN2A (p16-p14ARF)/CDKN2B (p15) gene inactivation in malignant gliomas. Acta Neuropathol 1999;98:444–52.
    OpenUrlCrossRefPubMed
  4. 4.↵
    1. Kim WY,
    2. Sharpless NE
    . The regulation of INK4/ARF in cancer and aging. Cell 2006;127:265–75.
    OpenUrlCrossRefPubMed
  5. 5.↵
    1. Kusy S,
    2. Larsen C-J,
    3. Roche J
    . p14ARF, p15INK4b and p16INK4a methylation status in chronic myelogenous leukemia. Leuk Lymphoma 2004;45:1989–94.
    OpenUrlCrossRefPubMed
  6. 6.↵
    1. Esteller M
    . CpG island hypermethylation and tumor suppressor genes: a booming present, a brighter future. Oncogene 2002;21:5427–40.
    OpenUrlCrossRefPubMed
  7. 7.↵
    1. Chen H,
    2. Gu X,
    3. Su IH,
    4. Bottino R,
    5. Contreras JL,
    6. Tarakhovsky A,
    7. et al.
    Polycomb protein Ezh2 regulates pancreatic β-cell Ink4a/Arf expression and regeneration in diabetes mellitus. Genes Dev 2009;23:975–85.
    OpenUrlAbstract/FREE Full Text
  8. 8.↵
    1. Jacobs JJ,
    2. Kieboom K,
    3. Marino S,
    4. DePinho RA,
    5. van Lohuizen M
    . The oncogene and Polycomb-group gene bmi-1 regulates cell proliferation and senescence through the ink4a locus. Nature 1999;397:164–68.
    OpenUrlCrossRefPubMed
  9. 9.↵
    1. Bernard D,
    2. Martinez-Leal JF,
    3. Rizzo S,
    4. Martinez D,
    5. Hudson D,
    6. Visakorpi T,
    7. et al.
    CBX7 controls the growth of normal and tumor-derived prostate cells by repressing the Ink4a/Arf locus. Oncogene 2005;24:5543–51.
    OpenUrlCrossRefPubMed
  10. 10.↵
    1. Pasmant E,
    2. Laurendeau I,
    3. Héron D,
    4. Vidaud M,
    5. Vidaud D,
    6. Bièche I
    . Characterization of a germ-line deletion, including the entire INK4/ARF locus, in a melanoma-neural system tumor family: identification of ANRIL, an antisense noncoding RNA whose expression coclusters with ARF. Cancer Res 2007;67:3963–9.
    OpenUrlAbstract/FREE Full Text
  11. 11.↵
    1. Visel A,
    2. Zhu Y,
    3. May D,
    4. Afzal V,
    5. Gong E,
    6. Attanasio C,
    7. et al.
    Targeted deletion of the 9p21 non-coding coronary artery disease risk interval in mice. Nature 2010;464:409–12.
    OpenUrlCrossRefPubMed
  12. 12.↵
    1. McKeller RN,
    2. Fowler JL,
    3. Cunningham JJ,
    4. Warner N,
    5. Smeyne RJ,
    6. Zindy F,
    7. et al.
    The Arf tumor suppressor gene promotes hyaloid vascular regression during mouse eye development. Proc Natl Acad Sci U S A 2002;99:3848–53.
    OpenUrlAbstract/FREE Full Text
  13. 13.↵
    1. Freeman-Anderson NE,
    2. Zheng Y,
    3. McCalla-Martin AC,
    4. Treanor LM,
    5. Zhao YD,
    6. Garfin PM,
    7. et al.
    Expression of the Arf tumor suppressor gene is controlled by Tgfβ2 during development. Development 2009;136:2081–9.
    OpenUrlAbstract/FREE Full Text
  14. 14.↵
    1. Zheng Y,
    2. Devitt C,
    3. Liu J,
    4. Mei J,
    5. Skapek SX
    . A distant, cis-acting enhancer drives induction of Arf by Tgfβ in the developing eye. Dev Biol 2013;380:49–57.
    OpenUrlCrossRefPubMed
  15. 15.↵
    1. Martin AC,
    2. Thornton JD,
    3. Liu J,
    4. Wang X,
    5. Zuo J,
    6. Jablonski MM,
    7. et al.
    Pathogenesis of persistent hyperplastic primary vitreous in mice lacking the Arf tumor suppressor gene. Invest Ophthalmol Vis Sci 2004;45:3387–96.
    OpenUrlAbstract/FREE Full Text
  16. 16.↵
    1. Zheng Y,
    2. Zhao YD,
    3. Gibbons M,
    4. Abramova T,
    5. Chu PY,
    6. Ash JD,
    7. et al.
    Tgfβ signaling directly induces Arf promoter remodeling by a mechanism involving Smads 2/3 and p38 MAPK. J Biol Chem 2010;285:35654–64.
    OpenUrlAbstract/FREE Full Text
  17. 17.↵
    1. Zheng Y,
    2. Devitt C,
    3. Liu J,
    4. Iqbal N,
    5. Skapek SX
    . Arf induction by Tgfβ is influenced by Sp1 and C/ebpβ in opposing directions. PLoS One 2013;8:e70371.
    OpenUrlCrossRefPubMed
  18. 18.↵
    1. Wu MY,
    2. Hill CS
    . TGF-β superfamily signaling in embryonic development and homeostasis. Dev Cell 2009;16:329–43.
    OpenUrlCrossRefPubMed
  19. 19.↵
    1. Massagué J
    . TGFβ in cancer. Cell 2008;134:215–30.
    OpenUrlCrossRefPubMed
  20. 20.↵
    1. Markowitz S,
    2. Wang J,
    3. Myeroff L,
    4. Parsons R,
    5. Sun L,
    6. Lutterbaugh J,
    7. et al.
    Inactivation of the type II TGF-β receptor in colon cancer cells with microsatellite instability. Science 1995;268:1336–8.
    OpenUrlAbstract/FREE Full Text
  21. 21.↵
    1. Becker C,
    2. Fantini MC,
    3. Schramm C,
    4. Lehr HA,
    5. Wirtz S,
    6. Nikolaev A,
    7. et al.
    TGF-β suppresses tumor progression in colon cancer by inhibition of IL-6 trans-signaling. Immunity 2004;21:491–501.
    OpenUrlCrossRefPubMed
  22. 22.↵
    1. David CJ,
    2. Huang YH,
    3. Chen M,
    4. Su J,
    5. Zou Y,
    6. Bardeesy N,
    7. et al.
    TGF-β tumor suppression through a lethal EMT. Cell 2016;164:1015–30.
    OpenUrlCrossRefPubMed
  23. 23.↵
    1. Roberts AB,
    2. Tian F,
    3. Byfield SD,
    4. Stuelten C,
    5. Ooshima A,
    6. Saika S,
    7. et al.
    Smad3 is key to TGF-β-mediated epithelial-to-mesenchymal transition, fibrosis, tumor suppression and metastasis. Cytokine Growth Factor Rev 2006;17:19–27.
    OpenUrlCrossRefPubMed
  24. 24.↵
    1. Tarasewicz E,
    2. Rivas L,
    3. Hamdan R,
    4. Dokic D,
    5. Parimi V,
    6. Bernabe BP,
    7. et al.
    Inhibition of CDK-mediated phosphorylation of Smad3 results in decreased oncogenesis in triple negative breast cancer cells. Cell Cycle 2014;13:3191–201.
    OpenUrl
  25. 25.↵
    1. Yeh HW,
    2. Hsu EC,
    3. Lee SS,
    4. Lang YD,
    5. Lin YC,
    6. Chang CY,
    7. et al.
    PSPC1 mediates TGF-β1 autocrine signalling and Smad2/3 target switching to promote EMT, stemness and metastasis. Nat Cell Biol 2018;20:479–91.
    OpenUrl
  26. 26.↵
    1. Hannon GJ,
    2. Beach D
    . p15INK4B is a potential effector of TGF-β-induced cell cycle arrest. Nature 1994;371:257–61.
    OpenUrlCrossRefPubMed
  27. 27.↵
    1. Datto MB,
    2. Li Y,
    3. Panus JF,
    4. Howe DJ,
    5. Xiong Y,
    6. Wang XF
    . Transforming growth factor β induces the cyclin-dependent kinase inhibitor p21 through a p53-independent mechanism. Proc Natl Acad Sci U S A 1995;92:5545–9.
    OpenUrlAbstract/FREE Full Text
  28. 28.↵
    1. Scandura JM,
    2. Boccuni P,
    3. Massagué J,
    4. Nimer SD
    . Transforming growth factor β-induced cell cycle arrest of human hematopoietic cells requires p57KIP2 up-regulation. Proc Natl Acad Sci U S A 2004;101:15231–6.
    OpenUrlAbstract/FREE Full Text
  29. 29.↵
    1. Silva RL,
    2. Thornton JD,
    3. Martin AC,
    4. Rehg JE,
    5. Bertwistle D,
    6. Zindy F,
    7. et al.
    Arf-dependent regulation of Pdgf signaling in perivascular cells in the developing mouse eye. EMBO J 2005;24:2803–14.
    OpenUrlAbstract/FREE Full Text
  30. 30.↵
    1. Zhang Y,
    2. Liu T,
    3. Meyer CA,
    4. Eeckhoute J,
    5. Johnson DS,
    6. Bernstein BE,
    7. et al.
    Model-based analysis of ChIP-Seq (MACS). Genome Biol 2008;9:R137.
    OpenUrlCrossRefPubMed
  31. 31.↵
    1. Langmead B,
    2. Salzberg SL
    . Fast gapped-read alignment with Bowtie 2. Nat Methods 2012;9:357–9.
    OpenUrlCrossRefPubMed
  32. 32.↵
    1. Thorvaldsdóttir H,
    2. Robinson JT,
    3. Mesirov JP
    . Integrative Genomics Viewer (IGV): high-performance genomics data visualization and exploration. Brief Bioinform 2013;14:178–92.
    OpenUrlCrossRefPubMed
  33. 33.↵
    1. Bauer DE,
    2. Canver MC,
    3. Orkin SH
    . Generation of genomic deletions in mammalian cell lines via CRISPR/Cas9. J Vis Exp 2015;95:e52118.
    OpenUrlCrossRefPubMed
  34. 34.↵
    1. Labun K,
    2. Montague TG,
    3. Gagnon JA,
    4. Thyme SB,
    5. Valen E
    . CHOPCHOP v2: a web tool for the next generation of CRISPR genome engineering. Nucleic Acids Res 2016;44:W272–6.
    OpenUrlCrossRefPubMed
  35. 35.↵
    1. Sanjana NE,
    2. Shalem O,
    3. Zhang F
    . Improved vectors and genome-wide libraries for CRISPR screening. Nat Methods 2014;11:783–4.
    OpenUrlCrossRefPubMed
  36. 36.↵
    1. Sandoval R,
    2. Xue J,
    3. Pilkinton M,
    4. Salvi D,
    5. Kiyokawa H,
    6. Colamonici OR
    . Different requirements for the cytostatic and apoptotic effects of type I interferons: induction of apoptosis requires ARF but not p53 in osteosarcoma cell lines. J Biol Chem 2004;279:32275–80.
    OpenUrlAbstract/FREE Full Text
  37. 37.↵
    1. Landry JJ,
    2. Pyl PT,
    3. Rausch T,
    4. Zichner T,
    5. Tekkedil MM,
    6. Stütz AM,
    7. et al.
    The genomic and transcriptomic landscape of a HeLa cell line. G3 (Bethesda) 2013;3:1213–24.
    OpenUrlCrossRef
  38. 38.↵
    1. Werness B,
    2. Levine A,
    3. Howley P
    . Association of human papillomavirus types 16 and 18 E6 proteins with p53. Science 1990;248:76–9.
    OpenUrlAbstract/FREE Full Text
  39. 39.↵
    1. Zindy F,
    2. Eischen CM,
    3. Randle DH,
    4. Kamijo T,
    5. Cleveland JL,
    6. Sherr CJ,
    7. et al.
    Myc signaling via the ARF tumor suppressor regulates p53-dependent apoptosis and immortalization. Genes Dev 1998;12:2424–33.
    OpenUrlAbstract/FREE Full Text
  40. 40.↵
    1. Palmero I,
    2. Pantoja C,
    3. Serrano M
    . p19ARF links the tumour suppressor p53 to Ras. Nature 1998;395:125–6.
    OpenUrlCrossRefPubMed
  41. 41.↵
    1. Sobell HM
    . Actinomycin and DNA transcription. Proc Natl Acad Sci U S A 1985;82:5328–31.
    OpenUrlAbstract/FREE Full Text
  42. 42.↵
    1. Derynck R,
    2. Zhang YE
    . Smad-dependent and Smad-independent pathways in TGF-β family signalling. Nature 2003;425:577–84.
    OpenUrlCrossRefPubMed
  43. 43.↵
    1. Moustakas A,
    2. Souchelnytskyi S,
    3. Heldin CH
    . Smad regulation in TGF-β signal transduction. J Cell Sci 2001;114:4359–69.
    OpenUrlAbstract/FREE Full Text
  44. 44.↵
    1. Zeng Y,
    2. Kotake Y,
    3. Pei XH,
    4. Smith MD,
    5. Xiong Y
    . p53 binds to and is required for the repression of Arf tumor suppressor by HDAC and polycomb. Cancer Res 2011;71:2781–92.
    OpenUrlAbstract/FREE Full Text
  45. 45.↵
    1. Duttke SHC,
    2. Lacadie SA,
    3. Ibrahim MM,
    4. Glass CK,
    5. Corcoran DL,
    6. Benner C,
    7. et al.
    Human promoters are intrinsically directional. Mol Cell 2015;57:674–84.
    OpenUrlCrossRefPubMed
  46. 46.↵
    1. Core LJ,
    2. Waterfall JJ,
    3. Lis JT
    . Nascent RNA sequencing reveals widespread pausing and divergent initiation at human promoters. Science 2008;322:1845–8.
    OpenUrlAbstract/FREE Full Text
  47. 47.↵
    1. McPherson R,
    2. Pertsemlidis A,
    3. Kavaslar N,
    4. Stewart A,
    5. Roberts R,
    6. Cox DR,
    7. et al.
    A common allele on chromosome 9 associated with coronary heart disease. Science 2007;316:1488–91.
    OpenUrlAbstract/FREE Full Text
  48. 48.↵
    1. Creyghton MP,
    2. Cheng AW,
    3. Welstead GG,
    4. Kooistra T,
    5. Carey BW,
    6. Steine EJ,
    7. et al.
    Histone H3K27ac separates active from poised enhancers and predicts developmental state. Proc Natl Acad Sci U S A 2010;107:21931–6.
    OpenUrlAbstract/FREE Full Text
  49. 49.↵
    1. Dennler S,
    2. Itoh S,
    3. Vivien D,
    4. ten Dijke P,
    5. Huet S,
    6. Gauthier JM
    . Direct binding of Smad3 and Smad4 to critical TGFβinducible elements in the promoter of human plasminogen activator inhibitortype 1 gene. EMBO J 1998;17:3091–100.
    OpenUrlAbstract/FREE Full Text
  50. 50.↵
    1. Rao SS,
    2. Huntley MH,
    3. Durand NC,
    4. Stamenova EK,
    5. Bochkov ID,
    6. Robinson JT,
    7. et al.
    A 3D map of the human genome at kilobase resolution reveals principles of chromatin looping. Cell 2014;159:1665–80.
    OpenUrlCrossRefPubMed
  51. 51.↵
    1. Zhang Y,
    2. Tong T
    . FOXA1 antagonizes EZH2-mediated CDKN2A repression in carcinogenesis. Biochem Biophys Res Commun 2014;453:172–8.
    OpenUrlCrossRefPubMed
  52. 52.↵
    1. Gil J,
    2. Peters G
    . Regulation of the INK4b–ARF–INK4a tumour suppressor locus: all for one or one for all. Nat Rev Mol Cell Biol 2006;7:667–77.
    OpenUrlCrossRefPubMed
  53. 53.↵
    1. Kamijo T,
    2. Zindy F,
    3. Roussel MF,
    4. Quelle DE,
    5. Downing JR,
    6. Ashmun RA,
    7. et al.
    Tumor suppression at the mouse INK4a locus mediated by the alternative reading frame product p19ARF. Cell 1997;91:649–59.
    OpenUrlCrossRefPubMed
  54. 54.↵
    1. Bachman KE,
    2. Park BH
    . Duel nature of TGF-β signaling: tumor suppressor vs. tumor promoter. Curr Opin Oncol 2005;17:49–54.
    OpenUrlCrossRefPubMed
  55. 55.↵
    1. Weber JD,
    2. Jeffers JR,
    3. Rehg JE,
    4. Randle DH,
    5. Lozano G,
    6. Roussel MF,
    7. et al.
    p53-independent functions of the p19ARF tumor suppressor. Genes Dev 2000;14:2358–65.
    OpenUrlAbstract/FREE Full Text
  56. 56.↵
    1. Dyson NJ
    . RB1: a prototype tumor suppressor and an enigma. Genes Dev 2016;30:1492–502.
    OpenUrlAbstract/FREE Full Text
  57. 57.↵
    1. Levy L,
    2. Hill CS
    . Alterations in components of the TGF-β superfamily signaling pathways in human cancer. Cytokine Growth Factor Rev 2006;17:41–58.
    OpenUrlCrossRefPubMed
  58. 58.↵
    1. Sanford LP,
    2. Ormsby I,
    3. Gittenberger-de Groot AC,
    4. Sariola H,
    5. Friedman R,
    6. Boivin GP,
    7. et al.
    TGFβ2 knockout mice have multiple developmental defects that are non-overlapping with other TGFβ knockout phenotypes. Development 1997;124:2659–70.
    OpenUrlAbstract
  59. 59.↵
    1. Sherr CJ
    . Tumor surveillance via the ARF–p53 pathway. Genes Dev 1998;12:2984–91.
    OpenUrlFREE Full Text
  60. 60.↵
    1. Seoane J,
    2. Pouponnot C,
    3. Staller P,
    4. Schader M,
    5. Eilers M,
    6. Massagué J
    . TGFβ influences Myc, Miz-1 and Smad to control the CDK inhibitor p15INK4b. Nat Cell Biol 2001;3:400–8.
    OpenUrlCrossRefPubMed
View Abstract
Back to top
View this article with LENS

Open full page PDF
Article Alerts
Email Article
Citation Tools
Identification of De Novo Enhancers Activated by TGFβ to Drive Expression of CDKN2A and B in HeLa Cells
Yen-Ting Liu, Lin Xu, Lynda Bennett, Jared C. Hooks, Jing Liu, Qinbo Zhou, Priscilla Liem, Yanbin Zheng and Stephen X. Skapek
Mol Cancer Res September 1 2019 (17) (9) 1854-1866; DOI: 10.1158/1541-7786.MCR-19-0289
del.icio.us logo Digg logo Reddit logo Twitter logo CiteULike logo Facebook logo Google logo Mendeley logo

Jump to section

Advertisement