HP1γ promotes lung adenocarcinoma by downregulating the transcription-repressive regulators NCOR2 and ZBTB7A

Abstract

Lung adenocarcinoma (LUAD) is a major form of lung cancer, which is the leading cause of cancer death. Histone methylation reader proteins mediate the effect of histone methylation, a hallmark of epigenetic and transcriptional regulation of gene expression. However, their roles in LUAD are poorly understood. Here our bioinformatic screening and analysis in search of an LUAD-promoting histone methylation reader protein show that heterochromatin protein 1γ (HP1γ; also called CBX3) is among the most frequently overexpressed and amplified histone reader proteins in human LUAD, and that high HP1γ mRNA levels are associated with poor prognosis in LUAD patients. In vivo depletion of HP1γ reduced K-Ras^G12D-driven LUAD and lengthened survival of mice bearing K-Ras^G12D-induced LUAD. HP1γ and its binding activity to methylated histone H3 lysine 9 were required for the proliferation, colony formation, and migration of LUAD cells. HP1γ directly repressed expression of the transcription-repressive regulators NCOR2 and ZBTB7A. Knockdown of NCOR2 or ZBTB7A significantly restored defects in proliferation, colony formation, and migration in HP1γ-depleted LUAD cells. Low NCOR2 or ZBTB7A mRNA levels were associated with poor prognosis in LUAD patients and correlated with high HP1γ mRNA levels in LUAD samples. NCOR2 and ZBTB7A downregulated expression of tumor-promoting factors such as ELK1 and AXL, respectively. These findings highlight the importance of HP1γ and its reader activity in LUAD tumorigenesis and reveal a unique LUAD-promoting mechanism in which HP1γ downregulates NCOR2 and ZBTB7A to enhance expression of pro-tumorigenic genes.

Introduction

Lung cancer is the leading cause of global cancer-related death in both men and women. The overall 5-year survival rate for lung cancer patients is low (about 18.1%). Lung adenocarcinoma (LUAD), the most prevalent histological subtype of lung cancer, accounts for approximately 40% of lung cancer cases. The molecular etiology of LUAD is diverse. For example, activating mutations and gene amplification of oncogenic kinases (e.g. K-Ras, EGFR, MET, and ERBB2) and inactivating alterations in tumor suppressor genes (e.g. TP53 and LKB1/STK11) frequently occur in LUAD (1, 2). For the treatment of LUAD patients, much research has focused on kinase signaling. Kinase-targeted therapies (e.g., the EGFR inhibitor erlotinib) have been developed but had a limited success because of tumor recurrence. There is still a lack of well-defined molecular targets for treating LUAD. Therefore, there is a great need for a better mechanistic understanding of LUAD tumorigenesis.

Histone methylation is a key mark of epigenetic and transcriptional gene regulation and occurs at both lysine and arginine residues in histones. This modification plays an important role in regulating various biological processes, including cellular differentiation, stem cell maintenance, and cancer (3). Histone methylation is associated with either gene activation or silencing, depending on the modified sites, at the genome-wide levels. For example, methylation at histone H3 lysine 9 (H3K9) is generally coupled with gene silencing, whereas methylation at H3K4 is linked to gene activation (3). The levels of histone methylation can be dynamically regulated by histone methyltransferases and demethylases.

Histone methylation provides specific binding sites for ‘readers’ (also called binding modules). The reader-containing proteins mediate the effect of histone methylation. Such readers include chromodomain, Tudor, PHD (plant homeodomain), PWWP (Pro-Trp-Trp-Pro), bromo-adjacent homology (BAH), MBT (malignant brain tumor), WD40, ankyrin, and the zinc finger CW (zf-CW) (4). These readers are present in multiple types of proteins, including transcription factors, DNA-modifying enzymes, and histone-modifying enzymes. It has been known that deregulated histone methylation reader proteins contribute to cellular transformation and tumorigenesis. However, the pathogenic roles of histone methylation reader proteins in LUAD tumorigenesis are largely unknown.

Recently, multiple tumor-sequencing studies have compiled a list of mutations and genomic rearrangements in LUAD, providing great resources to screen potential candidate genes implicated in lung tumorigenesis. In search of an oncogenic histone methylation reader protein in LUAD, we performed in silico screening with use of several databases (e.g., The Cancer Genome Atlas [TCGA]) to determine which proteins of 124 histone methylation reader proteins undergo alterations in expression, DNA sequences, and copy numbers in LUAD. This screening led us to identify heterochromatin protein 1γ (HP1γ), an di- and trimethylated H3K9 (H3K9me2/3) reader protein (5–7), as one of the most frequently amplified and overexpressed histone methylation reader proteins in human LUAD. Because the in vivo tumor-promoting role of HP1γ and its mechanism of action in LUAD were unclear, we chose to study the role of HP1γ in LUAD. Our analysis showed that high HP1γ protein and mRNA levels correlated with poor prognosis in lung cancer patients. In vivo HP1γ knockdown in the K-Ras^G12D LUAD mouse model inhibited K-Ras^G12D-induced tumorigenicity and prolonged mouse survival. Our results from RNAi and rescue experiments demonstrated that HP1γ and its H3K9me2/3-binding activity were required for the proliferation, anchorage-independent growth, and migration of LUAD cells. Our mechanistic results provided evidence that HP1γ downregulates expression of the transcription-repressive regulators NCOR2 and ZBTB7A to upregulate expression of several tumor-promoting factors, such as AXL, PVT1, and ELK1. These findings provide a previously unknown mechanistic insight into the pathogenesis of LUAD and also suggest a rationale for targeting the binding activity of the prognostic biomarker HP1γ in LUAD patients with few therapeutic options.

Materials and Methods

Samples, reagents, cell lines, and antibodies.

For immunohistochemical (IHC) experiments, human lung normal and tumor tissue microarrays (TMAs) were purchased from Biomax (LC951 and LC1291) and Imgenex (IMH-305). IHC analysis was performed on these TMAs as previously described (8). All lung cancer cell lines were procured from ATCC (Rockville, MD, USA), which verifies cell lines using short tandem repeat analysis, and were cultured within 15 times of passages in 2–3 months. Mycoplasma testing was conducted for cell lines using the mycoplasma detection kit PlasmoTest™ (InvivoGen). Cell culture reagents and other chemicals were purchased from Gibco, Hyclone, Corning, Sigma-Aldrich and Fisher Bioreagents. The antibodies used for this study are listed in Supplementary Table S1.

Expression data and survival analysis.

TCGA, NCI Director’s Challenge Consortium, and lung cancer transcriptomic datasets (SE37745 and GSE29013) were used for expression data and survival analysis (see Supplementary methods for details).

IHC experiments, IHC scoring, and immunofluorescence.

IHC experiments were performed as described previously (8). TMA samples (n = 73) with clear IHC staining and survival follow-up data of patients were scored on a scale of 0–5 based on the percentage of tumor cells stained: 0, negative; 1, 1%–10%; 2, 11%–25%; 3, 26%–50%; 4, 51%–70%; 5, >71%. The samples were further divided into low (score 0–3) and high (≥4) score categories. For immunofluorescence, Alexa 488-conjugated anti-mouse IgG and Alexa 568-conjugated anti-rabbit IgG secondary antibodies were used for detection, and images were captured using a laser confocal microscope.

Mouse strains and in vivo lung tumorigenesis study.

Briefly, shRNAs against mouse HP1γ (mHP1γ) were designed and cloned into U6-shRNA pgkCre vector (Addgene plasmid # 24971; provided by Tyler Jacks) (Supplementary Table S1). To induce tumors in the lungs of 6‒8 week old K-Ras^LSL-G12D mice (Strain number 01XJ6, NCI mouse repository), each mouse was infected with 1×10⁵ lentivirus particles that expressed either Cre alone (a control) or both Cre and shHP1γ (shmHP1γ−1-Cre) via an intratracheal intubation method. The mice were monitored for tumor growth for 12 to 14 months using micro-computed tomography (micro-CT). Mouse survival was compared between the control group and shmHP1γ−1-Cre group. Humane end-points were used in the mouse experiments. The lungs were collected at necropsy and lung tumors were analyzed microscopically (see Supplementary methods for details).

Stable knockdown

For knockdown experiments, lentivirus-based, puromycin-resistant shRNAs were purchased from Sigma and Open Biosystem (Supplementary Table S1). The shRNA-infected cells were selected in puromycin-containing medium (1 μg/ml). shLuciferase (shLuc)-infected cells were used as a control.

Quantitative RT-PCR, Western blot, and chromatin immunoprecipitation (ChIP) assays

The quantitative RT-PCR, Western blot, and ChIP assays were performed as previously described (8, 9). The primers and antibodies used for quantitative RT-PCR and ChIP assays are listed in Supplementary Table S1.

Cell proliferation, cell migration, and soft agar assays

Cell proliferation, cell migration, and soft agar assays were performed as previously described (8, 10).

Clonogenic cell survival assay

Cells (1×10³) were plated onto six-well plates in triplicate. After 10‒14 days passed, the cells were fixed with 4% paraformaldehyde, followed by staining with 0.5% crystal violet for 1 hour at room temperature. Stained cells were then washed with 1 x PBS, and images were captured.

Rescue and overexpression experiments.

For rescue experiments, human HP1γ cDNA was cloned into the lentivirus vector pLenti6.3/V5-DEST (Thermo Fisher Scientific) using standard cloning methodology. To generate an H3K9me2/3-binding mutant of HP1γ (mtHP1γ), HP1γ cDNA cloned in pLenti6.3/V5-DEST was mutated by site-directed mutagenesis. Then, shRNA-resistant HP1γ and mtHP1γ constructs were generated by introducing silent mutations into the target site of shHP1γ−16. The primer sequences used for mutagenesis are listed in Supplementary Table S1. Cells infected by pLenti6.3 were selected in blasticidin-containing medium (5 μg/ml). For HP1γ overexpression experiments, cells infected with pLenti6.3-GFP and pLenti6.3 vector were used as controls.

Microarrays.

The whole transcriptomic analysis was performed using Affymetrix GeneChip™ Human Genome U133 Plus 2.0 Arrays. Transcriptomic profile data of HP1γ-depleted cells were compared with those of shLuc-infected cells. Gene ontology analysis was performed using DAVID functional annotation tools (http://david.abcc.ncifcrf.gov). The accession number for transcriptomic profile data reported in this paper is GEO: GSE111321.

Statistical analysis.

Student’s t-test was used to determine the statistical significance of two groups of data. For correlation analysis, the chi-squared test was performed to calculate the level of significance. The log-rank method was used to test the statistical significance of survival data. Data are presented as means ± standard error of the mean (SEM; error bars). P-values less than 0.05 were considered statistically significant. *, P <0.05; **, P <0.01; and ***, P <0.001 indicate statistically significant differences.

Study approval.

The care and use of all mice was approved by the Institutional Animal Care and Use Committee of The University of Texas MD Anderson Cancer Center.

Results

HP1γ is frequently amplified and overexpressed in LUAD, and its high levels are linked to poor prognosis in LUAD patients

To identify a histone methylation reader protein with oncogenic function in LUAD, we first analyzed whether histone methylation reader proteins, including those in the WERAM database (http://weram.biocuckoo.org/) (11), are altered in TCGA human LUAD samples (Fig. 1A). For this analysis, we considered the following three types of alterations: copy number variations, mutations, and changes in mRNA expression. Of 124 histone methylation reader proteins that we analyzed using TCGA database, 58 proteins showed ≥8% alterations (Fig. 1B; the top 15 proteins are detailed in Supplementary Fig. S1A and B). We then determined whether these 58 proteins are upregulated in LUAD tumors compared with adjacent normal lung tissues. Of 58 reader proteins, 14 showed ≥1.5-fold upregulation in their mRNA levels (Fig. 1C; Supplementary Fig. S2; Supplementary Table S2). These top 14 hits were further subjected to Kaplan-Meier survival analysis using a publicly available microarray dataset of 443 lung adenocarcinoma patients (stages I, II, and III) from the NCI Director’s Challenge Consortium for the Molecular Classification of Lung Adenocarcinoma. Of the 14 hits, data for 12 reader proteins, all except PYGO2 and CDYL2, were available, and high HP1γ and MSH6 mRNA levels showed statistically significant shorter survival in LUAD patients (Fig. 1D; Supplementary Fig. S3). Median survival time of HP1γ-high LUAD patients (21.2 ± 3.42 months) was significantly shorter than that of MSH6-high LUAD patients (30.1 ± 13.0 months). In addition, results from TCGA database analysis demonstrated that alterations in HP1γ but not MSH6 correlated with shorter overall and disease-free survival in LUAD patients (Supplementary Fig. S4A–D). These results led us to choose HP1γ as a putative oncogenic histone methylation reader protein for subsequent analyses.

Figure 1: HP1γ levels are upregulated in lung adenocarcinoma and high HP1γ levels correlate with shorter survival in LUAD patients.

(A) An in silico screening strategy for the search of putative oncogenic histone methylation reader proteins. LUAD, Lung adenocarcinoma. (B) Bar graph showing alterations in the 58 histone methylation reader proteins in LUAD samples in TCGA database. The cutoff value for alterations was 8%. (C) Fourteen histone methylation reader proteins with >1.5-fold upregulation in LUAD tumors (n = 357) compared with adjacent normal tissue samples (n = 54) in TCGA dataset. P-value for the 14 reader proteins, <0.05. T, tumor; N, normal. (D) Kaplan-Meier survival rate analysis on the basis of HP1γ mRNA levels (probe set, 200037_s_at) using NCI LUAD dataset. Tumor samples (n = 443) were divided into HP1γ-high or HP1γ-low mRNA groups by using a cutoff value of the mean plus 2 standard deviations. (E) Representative images of IHC staining of HP1γ in normal lung and lung tumors. Scale bars, 50 μm. (F) Survival analysis of lung cancer patients on the basis of HP1γ protein levels (see also Supplementary Table S3 and S4). HP1γ protein levels in lung cancer samples in TMAs were determined by IHC staining (n = 73).

To further determine the lung cancer relevance of HP1γ, we assessed the association of HP1γ levels with survival in lung cancer patients by analyzing several other lung cancer datasets. Our analysis demonstrated that high HP1γ mRNA levels, but not high MSH6 mRNA levels, were consistently associated with shorter survival in lung cancer patients. (Supplementary Fig. S5A–D). Interestingly, HP1γ mRNA levels were also increased in lung squamous cell carcinomas (Supplementary Fig. S5E). In addition, IHC analysis showed that in contrast to only 12.5% (1/8) of normal lung samples, 32.9% (24/73) of lung cancer samples displayed high HP1γ protein levels (Fig. 1E; Supplementary Fig. S5F; Supplementary Table S3). Survival analysis of these lung cancer patients (n = 73) demonstrated that patients with high HP1γ protein levels survived shorter than did those with low HP1γ protein levels (Fig. 1F; Supplementary Table S4). Together, these results indicate that high HP1γ levels correlate with worse survival in lung cancer patients.

In vivo knockdown of HP1γ in mouse lung impedes K-Ras^G12D–driven LUAD tumorigenicity and increases survival of mice bearing K-Ras^G12D–driven LUADs.

To determine the importance of HP1γ in in vivo LUAD tumorigenesis, we examined the effect of HP1γ knockdown on K-Ras^G12D-driven mouse lung tumorigenesis. The K-Ras^G12D mouse model was used because it is a well-established LUAD mouse model (12) and HP1γ overexpression partially overlapped with K-Ras mutations in LUAD samples (Supplementary Fig. S6A). To achieve HP1γ knockdown in vivo, we employed the lentivirus-based U6-shRNA-pgkCre system (Supplementary Fig. S6B) (13). This lentiviral vector system not only expresses Cre recombinase under the pgk promoter to induce K-Ras^G12D expression by Cre-mediated deletion of the LSL (loxP-STOP-loxP) cassette but also produces shRNA under the U6 promoter to deplete a specific protein. We cloned four different shRNAs against mouse HP1γ (shmHP1γ−1, −2, −3 and −4) in the U6-shRNA-pgkCre vector (see Supplementary Table S1). Our analysis of knockdown efficiency of U6-shHP1γ-pgkCre constructs in mouse lung 393P cells showed that shmHP1γ−1 was the most effective shRNA against mouse HP1γ (Supplementary Fig. S6C). Using HEK-293 cells containing a Cre-Reporter, we also determine whether Cre expressed from this vector is active. The Cre-Reporter HEK-293 cells were used because they normally express a green fluorescence protein (GFP) signal but, upon Cre-mediated deletion of GFP-Stop in loxP-GFP-Stop-loxP-RFP in the Cre-reporter, express red fluorescence protein (RFP). Our results showed that GFP signals switched to RFP signals after infection with lentiviruses containing U6-shHP1γ-pgkCre, indicating that all constructs expressed active Cre (Supplementary Fig. S6D).

We next infected lungs of 6- to 8-week-old K-Ras^LSL-G12D mice with shmHP1γ−1-Cre or control (only Cre expression) viruses using an intratracheal intubation method and monitored tumor growth in mouse lungs at 3 and 9 months post-infection using micro-CT (Fig. 2A). Tumor sizes were significantly smaller in the shmHP1γ−1-Cre group of mice than in the control group (Fig. 2B and C). At 12 months post-infection, the average lung tumor burden in the shmHP1γ−1-Cre group was about one-half of that in the control group (Fig. 2D and E; Supplementary Fig. S7A). Moreover, the percentage of the lung effaced by tumors was analyzed and microscopically scored. As shown in Fig. 2F, the shmHP1γ−1-Cre group of mice had a lower percentage of their lung affected (grades I and II) than did the control group (grades II and III). Consistent with slower growth of HP1γ-depleted tumors, fewer Ki-67-expressing cells were present in HP1γ-depleted tumors than in control tumors (Fig. 2G and H). IHC analysis for the well-known LUAD marker TTF1/NKX2–1 showed that lung tumors from both groups expressed TTF1, confirming that the K-Ras^G12D model develops LUAD (Supplementary Fig. S7B). In addition, HP1γ expression in K-Ras^G12D mice was higher in tumor cells than in adjacent normal lung cells (Supplementary Fig. S7C). Notably, shmHP1γ−1-Cre mice had significantly longer survival times than did control mice. Median survival times of shmHP1γ−1-Cre and control mice were 354 and 286 days, respectively (Fig. 2I). These results indicate that in vivo depletion of HP1γ inhibits LUAD tumorigenicity in the K-Ras^G12D mouse model.

Figure 2: HP1γ is required for in vivo growth of K-Ras^G12D-driven LUAD tumors.

(A) A scheme for monitoring the effect of in vivo knockdown of HP1γ on K-Ras^G12D-driven lung tumorigenesis. The lungs of 8-week-old mice were infected by an intratracheal intubation of shmHP1γ−1-Cre lentiviruses, which express both Cre recombinase and shRNA targeting mouse HP1γ. Lentiviruses with an empty vector expressing only Cre were used as a control. (B) Micro-CT-scan images of mouse lungs infected with shmHP1γ−1-Cre (n = 3) and control (n = 3) viruses at 3 months and 9 months post-infection. (C‒H) The effect of in vivo HP1γ knockdown on K-Ras^G12D-driven lung tumorigenesis. Lung tumors in mice infected with control viruses were compared with those in mice infected with shmHP1γ−1-Cre viruses. Representative images of massive lung tumors in the control group and of tiny lung tumors in the shmHP1γ−1-Cre group are shown; tumors are encircled by dotted green lines (C). Representative images of H&E-stained lung lobes effaced by lung tumors in the control and shmHP1γ−1-Cre groups are shown (D). The percentages of tumor area per lobe in control (n = 10) and shmHP1γ−1-Cre (n = 10) groups were quantified (E). Tumor grades, based on tumor size, in control (n = 4) and shmHP1γ−1-Cre (n = 4) groups were scored (F). IHC staining data for HP1γ and the cell proliferation marker Ki-67 in lung tumors from control and shmHP1γ−1-Cre groups are shown (G). Ki-67-positive cells in eight random fields of three different tumors of control and shmHP1γ−1-Cre groups were quantified (H). (I) Kaplan-Meier survival analysis of the control group of mice (n=17) and the shmHP1γ−1-Cre group of mice (n=14). Scale bars, 7 mm (D) and 100 μm (G).

HP1γ and its H3K9me2/3-binding ability are necessary for the proliferation, colony formation, and migration of LUAD cells.

Oncogenic proteins commonly play a critical role in the proliferation, anchorage-independent colony formation, and migration of cancer cells. To investigate whether HP1γ is required for these cellular properties of LUAD cells, we examined the effect of HP1γ knockdown on such cellular properties of H1792 cells (an LUAD cell line with relatively high HP1γ levels) using the two most effective shRNAs (shHP1γ−16 and shHP1γ−17) of five HP1γ shRNAs tested (Supplementary Fig. S8A–C). HP1γ knockdown in H1792 cells reduced cell proliferation and cell cycle S phase but increased subG1 phase (Supplementary Fig. S8D and E). In addition, HP1γ depletion highly inhibited colony formation in the clonogenic cell survival assay, anchorage-independent colony formation in soft agar, and cell migration in Boyden’s chamber assay (Supplementary Fig. S8F–H). We confirmed the inhibitory effect of HP1γ knockdown on the proliferation, colony formation, and migration of LUAD cells using another HP1γ-high LUAD cell line, H23 (Supplementary Fig. S8A and S9A–E). These results indicate the importance of high HP1γ levels for the proliferation, anchorage-independent growth and migration in LUAD cells. In line with these results, HP1γ overexpression using a lentiviral system increased the sizes of anchorage-independent colonies of H460 cells (a large cell lung cancer cell line with a relatively low HP1 level) in soft agar (Supplementary Fig. S9F–H).

To examine whether the H3K9me2/3-binding activity of HP1γ is necessary for the proliferation, colony formation, and migration of LUAD cells, we generated an shHP1γ-resistant HP1γ lentivirus construct and introduced a valine-to-methionine mutation at position 32 (V32M) into the HP1γ construct using site-directed mutagenesis. It should be noted that the V32M mutation ablates H3K9me2/3-binding ability of HP1γ (7). We then infected HP1γ-depleted cells with lentiviruses expressing shHP1γ-resistant wild type HP1γ or its H3K9me2/3-binding mutant (mtHP1γ). Protein levels of HP1γ and mtHP1γ ectopically expressed in HP1γ-depleted cells were similar to endogenous levels of HP1γ in shLuc-treated cells, although their ectopically expressed mRNA levels were much higher than endogenous HP1γ mRNA levels (Fig. 3A and B). Interestingly, H3K9me3 levels were not changed by HP1γ knockdown or exogenous expression of HP1γ and mtHP1γ (Fig. 3B). Importantly, HP1γ but not mtHP1γ rescued defects in the proliferation, colony formation in clonogenic cell survival assay, migration, and anchorage-independent colony formation of HP1γ-depleted cells (Fig. 3C–F). These results suggest that the tumor-promoting function of HP1γ is dependent largely on its H3K9me2/3-binding activity.

Figure 3: HP1γ and its binding activity are essential for the proliferation, colony formation and migration of LUAD cells.

(A and B) Analysis of mRNA levels of ectopically expressed HP1γ and mtHP1γ and endogenous HP1γ by quantitative RT-PCR (A) and Western blot analysis of HP1γ, mtHP1γ, H3K9me3, and H3 levels (B). In HP1γ-depleted H1792 cells, HP1γ and mtHP1γ were ectopically expressed using a lentivirus system. (C) The effect of ectopic expression of HP1γ and mtHP1γ on the proliferation of HP1γ-depleted H1792 cells. (D) The effect of ectopic expression of HP1γ and mtHP1γ on the colony formation ability of HP1γ-depleted H1792 cells in a clonogenic cell survival assay. (E and F) The effect of ectopic expression of HP1γ and mtHP1γ on the migration (E) and anchorage-independent colony growth (F) of HP1γ-depleted H1792 cells. Black scale bars: 200 μm; white scale bars, 400 μm.

HP1γ enhances the proliferation, colony formation, and migration of LUAD cells by directly repressing NCOR2 and ZBTB7A expression.

To delineate the molecular mechanisms by which HP1γ may regulate cell proliferation, colony formation, and cell migration, we compared the whole transcriptomic profiles between HP1γ-depleted (shHP1γ−16 and shHP1γ−17) and control H1792 cells (shLuc) using Affymetrix microarray. Our analysis showed that genes upregulated and downregulated by shHP1γ−16 significantly overlapped with those upregulated and downregulated by shHP1γ−17, respectively (Fig. 4A). Gene ontology analysis showed that several gene expression programs, including apoptosis and negative regulation of cell proliferation, were upregulated by HP1γ knockdown. In contrast, other gene expression programs, such as cell migration and positive regulation of cell proliferation, were downregulated by HP1γ knockdown (Fig. 4B). These results corroborated the inhibitory effects of HP1γ knockdown on the proliferation, colony formation, and migration of LUAD cells.

Figure 4: HP1γ directly represses NCOR2 and ZBTB7A expression while upregulating expression of oncogenes, such as AXL, PVT1, and ELK1.

(A) Venn diagrams and heat maps for genes upregulated (>1.5-fold) or downregulated (<0.5-fold) by HP1γ knockdown. RNA was isolated from HP1γ-depleted H1792 cells (shHP1γ−16 and shHP1γ−17) and control (shLuc) H1792 cells. The mRNA levels were assessed by Affymetrix Human Genome U133 plus 2.0 Array in duplicate. (B) Ontology analysis of genes upregulated or downregulated by HP1γ knockdown. The functional annotation tool DAVID was used. (C) Analysis of the effect of HP1γ knockdown on EHF, NCOR2, ZBTB7A, AXL, PVT1, ARF1 and ELK1 mRNA levels using quantitative RT-PCR. (D) Western blot analysis of NCOR2, ZBTB7A and EHF protein levels in HP1γ-depleted (shHP1γ−16 and shHP1γ−17) H1792 cells. (E‒G) Chromatin levels of HP1γ at the NCOR2 (E), ZBTB7A (F), and EHF (G) genes. Schematic representations of individual genes are shown (top panels). Quantitative ChIP assay was performed using shLuc-infected cells (control) and HP1γ-depleted cells. Arrows indicate the primer sites for PCR amplification of ChIP-enriched DNAs. TSS, transcription start site.

Interestingly, AXL, PVT1, ELK1, and ADP-ribosylation factor 1 (ARF1) were the top four tumor-promoting factors downregulated by HP1γ knockdown (Fig. 4A and C). The receptor tyrosine kinase AXL is overexpressed in multiple types of cancer, and its inhibition blocks tumor growth in a mouse model (14). The long non-coding RNA PVT1 has been found to be upregulated in multiple types of cancer and to promote tumorigenesis (15). ELK1 is an ETS-domain-containing transcription factor that promotes tumor progression in bladder cancer (16). In addition, AXL and PVT1 are associated with enhanced proliferation and migration of lung cancer cells (17, 18). ARF1 is a member of the Ras superfamily of small GTPases that may enhance cell proliferation and cancer progression (19).

Although it has been reported that its levels in the gene body may be positively associated with gene transcription (20), HP1γ often acts as a transcriptional corepressor (21, 22). Therefore, we reasoned that HP1γ indirectly enhances expression of tumor-promoting factors by repressing tumor-suppressive transcriptional repressors (or corepressors). We thus searched for transcription-repressive regulators that may act as tumor suppressors among the first 200 genes that are upregulated by HP1γ knockdown. Our search showed that nuclear receptor corepressor 2 (NCOR2; also called SMRT), zinc finger and BTB domain containing 7A (ZBTB7A; also called LRF, Pokemon, and FBI-1), and ETS Homologous Factor (EHF; also called ESE-3) were the top three transcription-repressive regulators with putative tumor-suppressive function (Fig. 4A and C), although their functions in LUAD tumorigenesis are not clear. NCOR2 acts as a transcriptional corepressor that interacts with the histone deacetylase 3 (HDAC3) and regulates chromatin structure and genome stability (23). ZBTB7A is a transcriptional repressor that belongs to the POZ/BTB and Krüppel (POK) transcription factor family (24, 25). It has been reported that NCOR2 and ZBTB7A may have pro-tumorigenic or tumor-suppressive functions in a tissue-dependent manner (see Discussion). EHF is a transcriptional repressor for a set of genes containing ETS/AP-1-binding sites and can have a tumor-suppressive function (26, 27).

Our quantitative RT-PCR data and Western blot analysis confirmed that NCOR2, ZBTB7A, and EHF levels in H1792 cells were upregulated by HP1γ depletion (Fig. 4C and D). Similar to this, NCOR2, ZBTB7A, and EHF mRNA levels in H23 cells were increased by HP1γ knockdown (Supplementary Fig. S10A–C). We also performed chromatin immunoprecipitation (ChIP) experiments to examine whether NCOR2, ZBTB7A, and EHF are occupied by HP1γ. Our ChIP results showed that HP1γ was significantly enriched at the proximal promoters of NCOR2, ZBTB7A and EHF genes (Fig. 4E–G), indicating that NCOR2, ZBTB7A, and EHF genes are directly repressed by HP1γ in LUAD cells. However, HP1γ knockdown did not have any significant effect on H3K9me3 levels at these genes (Supplementary Fig. S10D–F).

To determine whether the tumor-promoting function of HP1γ in LUAD cells is dependent on HP1γ-mediated repression of NCOR2, ZBTB7A, and EHF expression, we first selected the most effective shRNAs against NCOR2, ZBTB7A and EHF (shNCOR2–1, shZBTB7A-1, and shEHF-1) in H1792 cells (Supplementary Fig. S11A). We then examined the effects of NCOR2, ZBTB7A and EHF knockdown on the proliferation, colony formation, and migration of HP1γ-depleted H1792 cells. For this, we compared these cellular characteristics among H1792 cells treated with five groups of shRNAs: 1) shLuc, 2) shHP1γ−16, 3) shHP1γ−16 + shNCOR2–1, 4) shHP1γ−16 + shZBTB7A-1, and 5) shHP1γ−16 + shEHF-1. Knockdown efficiencies of HP1γ, NCOR2, ZBTB7A, and EHF in these five groups of cells were assessed by quantitative RT-PCR and Western blot analysis (Fig. 5A and B). As shown in Fig. 5C–E, knockdown of NCOR2 or ZBTB7A substantially restored defects in the proliferation, colony formation in clonogenic cell survival assay, and migration of HP1γ-depleted cells, whereas knockdown of EHF had no significant effect on such cellular characteristics of HP1γ-depleted cells. These results indicate that HP1γ-mediated transcriptional repression of NCOR2 and ZBTB7A is important for the cellular function of HP1γ that enhances the proliferation, colony formation, and migration of LUAD cells.

Figure 5: Knockdown of NCOR2 or ZBTB7A substantially rescues the proliferation, colony formation, and migration of HP1γ-depleted LUAD cells.

(A and B) Analysis of relative HP1γ, NCOR2, ZBTB7A, and EHF mRNA (A) and protein (B) levels in H1792 cells that were treated with the following five groups of shRNA-containing viruses: 1) shLuc, 2) shHP1γ−16, 3) shHP1γ−16 and shNCOR2–1, 4) shHP1γ−16 and shZBTB7A-1, and 5) shHP1γ−16 and shEHF-1. (C‒E) The effect of NCOR2 or ZBTB7A knockdown on the colony formation (C), proliferation (D), and migration (E) of HP1γ-depleted H1792 cells. shLuc-infected cells were used as controls. Scale bars, 200 μm.

Interestingly, ectopic expression of HP1γ but not mtHP1γ in HP1γ-depleted H1792 cells substantially downregulated NCOR2 and ZBTB7A mRNA levels while restoring AXL and PVT1 mRNA levels (Supplementary Fig. S11B and C). These results indicate that HP1γ’s binding activity is required for HP1γ-mediated regulation of these genes.

Low NCOR2 or ZBTB7A levels are associated with increased cell proliferation and migration, high HP1γ levels in LUAD tumors, and poor prognosis in LUAD patients.

To determine whether HP1γ has a negative effect on NCOR2 and ZBTB7A expression in our mouse model, we compared NCOR2 and ZBTB7A levels between K-Ras^G12D lung tumors and mHP1γ-depleted K-Ras^G12D tumors using IHC analysis. NCOR2 and ZBTB7A levels were increased by in vivo HP1γ knockdown (Fig. 6A). To assess whether there is any anti-correlation between HP1γ and either NCOR2 or ZBTB7A expression in human tumor samples, we analyzed the LUAD dataset (n = 357) in TCGA database. This analysis showed that HP1γ mRNA levels inversely correlated with NCOR2 and ZBTB7A levels (Fig. 6B and C). Similar results were obtained from our analysis using the NCI Director’s Challenge Consortium dataset (Supplementary Fig. S11D and E). These results support the notion that transcriptional repression of NCOR2 and ZBTB7A by HP1γ contributes to the tumor-promoting function of HP1γ in LUAD.

Figure 6: Low NCOR2 or ZBTB7A levels are linked to enhanced cell proliferation and migration, high HP1γ levels in LUAD tumors, and poor prognosis in LUAD patients.

(A) IHC levels of NCOR2 and ZBTB7A in K-Ras^G12D lung tumors and mHP1γ-depleted K-Ras^G12D tumors. (B and C) Scatter plots showing an inverse correlation between HP1γ mRNA levels and either NCOR2 (B) or ZBTB7A (C) mRNA levels in TCGA LUAD sample dataset. (D and E) The effect of NCOR2 knockdown on colony formation (D) and migration (E) of H1792 cells. (F and G) The effect of ZBTB7A knockdown on colony formation (F) and migration (G) of H1792 cells. (H) Box plots showing downregulation of NCOR2 (left panel) and ZBTB7A (right panel) mRNA levels in LUAD tumor samples (n = 357) compared with their adjacent normal samples (n = 54) in TCGA dataset. T, tumor; N, normal. (I and J) The Kaplan-Meier survival analysis showing the correlation of low NCOR2 (I) and ZBTB7A (J) mRNA levels with shorter survival in LUAD patients. The auto cutoff was used to divide samples into low and high groups in the KM Plotter database (http://kmplot.com/analysis). NCOR2 cutoff, 2870 in the range between 578 and 10323; ZBTB7A cutoff, 788 in the range between 259 and 20494; NCOR2 probe set, 207760_s_at; ZBTB7A probe set, 226554_at; black scale bars, 100 μm (A) and 200 μm (E and G).

To assess the roles of NCOR2 and ZBTB7A in regulating tumor-relevant characteristics of LUAD cells, we examined the effects of their knockdown on colony formation and migration of LUAD cells. Knockdown of NCOR2 or ZBTB7A significantly increased colony formation and migration of H1792 cells (Fig. 6D–G). Interestingly, NCOR2 and ZBTB7A mRNA levels were significantly downregulated in TCGA LUAD samples (n = 357) compared with adjacent normal samples (n = 54) (Fig. 6H), and low NCOR2 and ZBTB7A mRNA levels correlated with worse survival in the LUAD patients (Fig. 6I and J; Supplementary Fig. S11F and G). Analysis of TCGA database showed that NCOR2 and ZBTB7A genes in LUAD had approximately 12% and approximately 6% mutations, respectively, which contained loss-of-function mutations (Supplementary Fig. S12A and B). These results suggest that NCOR2 and ZBTB7A have tumor-suppressive functions in LUAD.

Expression of the tumor-promoting factors AXL, PVT1, and ELK1 in LUAD cells requires HP1γ-mediated downregulation of NCOR2 and ZBTB7A.

As mentioned above, AXL, PVT1, ELK1, and ARF1 have tumor-promoting functions. In line with this, their high mRNA levels correlated with worse survival in LUAD patients (Fig. 7A–C; Supplementary Fig. S12C). In our effort to understand how HP1γ upregulates expression of the oncogenes AXL, PVT1, ELK1, and ARF1, we determined whether the transcription-repressive regulators NCOR2 and ZBTB7A whose genes are repressed by HP1γ downregulate expression of these oncogenes. Specifically, we examined the effect of NCOR2 or ZBTB7A knockdown on expression of these oncogenes in H1792 cells. NCOR2 knockdown increased ELK1 mRNA levels (Fig. 7D) and ZBTB7A knockdown upregulated AXL and PVT1 expression (Fig. 7E). In contrast, knockdown of EHF did not show any significant effect on AXL, PVT1, ELK1 and ARF1 mRNA levels (Fig. 7F), consistent with the above results (Fig. 5C–E). Interestingly, our ChIP results showed that ZBTB7A was recruited to the promoter regions of AXL and PVT1 while NCOR2 was located to the ELK1 promoter region in H1792 cells (Fig. 7G–I). These results indicate that expression levels of AXL, PVT1 and ELK1 but not ARF1 are directly downregulated by NCOR2 or ZBTB7A.

Figure 7: HP1γ indirectly upregulates expression of the oncogenes AXL, PVT1, and ELK1 by downregulating NCOR2 and ZBTB7A expression.

(A‒C) Kaplan-Meier survival analysis of AXL (A), PVT1 (B), and ELK1 (C) levels in LUAD datasets. The auto cutoff was used to divide samples into low and high groups in the KM Plotter database (http://kmplot.com/analysis). AXL cutoff, 136 in the range between 5 and 677; PVT1 cutoff, 454 in the range between 12 and 15881; ELK1 cutoff, 13 in the range between 2 and 524; AXL probe set, 202685_s_at; PVT1 probe set, 1558290_a_at; ELK1, 220802_at. (D‒F) The effect of NCOR2 (D), ZBTB7A (E), or EHF (F) knockdown on AXL, PVT1, ELK1 and ARF1 mRNA levels in H1792 cells. (G‒I) Chromatin levels of ZBTB7A at the AXL (G) and PVT1 (H) genes and of NCOR2 at the ELK1 gene (I). Schematic representations of individual genes are shown (top panels). Arrows indicate the primer sites for PCR amplification of ChIP-enriched DNAs. TSS, transcription start site. (J‒L) The effect of NCOR2 or ZBTB7A knockdown on AXL (J), PVT1 (K), and ELK1 (L) mRNA levels in HP1γ-depleted H1792 cells. H1792 cells were treated with the following four groups of shRNA-containing viruses: 1) shLuc, 2) shHP1γ−16, 3) shHP1γ−16 and shNCOR2–1, and 4) shHP1γ−16 and shZBTB7A-1. shLuc-treated cells were used as controls. For analysis of mRNA levels in J‒L, quantitative RT-PCR was used. (M) A hypothetical model for the molecular mechanism underlying the tumor-promoting function of HP1γ in LUAD. HP1γ interacts with H3K9me3 at the NCOR2 and ZBTB7A promoters and represses NCOR2 and ZBTB7A expression. NCOR2 downregulates ELK1 expression, and ZBTB7A represses expression of AXL and PVT1. HP1γ-mediated repression of NCOR2 and ZBTB7A increases expression levels of tumor-promoting factors (e.g., AXL, PVT1, and ELK1) to enhance the proliferation, migration, and tumorigenic growth of LUAD cells.

To determine whether expression of AXL, PVT1 and ELK1 is dependent on HP1γ-mediated repression of NCOR2 and ZBTB7A expression, we examined whether NCOR2 or ZBTB7A knockdown increases expression of these oncogenes in HP1γ-depleted cells. Basically, we compared AXL, PVT1, ELK1, and ARF1 expression among H1792 cells treated with the following four groups of shRNAs: 1) shLuc, 2) shHP1γ−16, 3) shHP1γ−16 + shNCOR2–1, and 4) shHP1γ−16 + shZBTB7A-1. In HP1γ-depleted cells, NCOR2 knockdown restored ELK1 mRNA levels, and ZBTB7A depletion rescued AXL and PVT1 mRNA levels (Fig. 7J–L). In contrast, NCOR2 and ZBTB7A knockdown did not affect ARF1 expression (Supplementary Fig. S12D). These results suggest that HP1γ positively and indirectly regulates expression of oncogenes, such as AXL, PVT1, and ELK1, by directly downregulating expression of NCOR2 and ZBTB7A.

Discussion

In the current study, our bioinformatic screening and analysis of more than 120 histone methylation reader proteins with use of several databases identified HP1γ as one of the most frequently overexpressed and amplified histone methylation reader proteins in human LUAD samples. Our analysis of a cohort of lung tumor samples and several databases demonstrated that high HP1γ protein and mRNA levels were associated with worse survival in lung cancer patients, indicating that HP1γ may be a prognostic marker for LUAD patients. Our results also indicate that the proliferative, colony-forming, and migratory abilities of LUAD cells are dependent on HP1γ and its H3K9me2/3-binding activity. Furthermore, in vivo knockdown of HP1γ in the K-Ras^G12D LUAD mouse model decreased K-Ras^G12D-driven lung tumorigenesis while increasing mouse survival. Thus, our findings indicate that HP1γ, along with its reader activity, is necessary for LUAD tumorigenesis.

Three members in the HP1 family (HP1α, HP1β, and HP1γ) are conserved from Drosophila to mammals and play a critical role in diverse biological processes, including gene regulation, chromosome segregation, heterochromatin formation and maintenance, and DNA repair (28, 29). HP1α and HP1β are present largely in heterochromatin but HP1γ localizes in both heterochromatin and euchromatin (29, 30). Although their roles in LUAD have not been well characterized, the HP1 family of proteins has been linked to tumorigenesis of other cancer types. For example, HP1α is upregulated in several cancer types (e.g., breast, prostate, pancreatic and uterine cancer), and high HP1α levels are required for cell proliferation (31). HP1β promotes tumor growth by enhancing androgen receptor’s transcriptional activity in prostate cancer (32). On the contrary, low HP1α and HP1β levels correlate with cancer metastatic phenotype (33, 34). Therefore, it is possible that HP1α and HP1β may have pro-tumorigenic or tumor-suppressive functions depending on cancer stage. HP1γ has been shown to be upregulated in different types of tumors, including prostate, colorectal, esophageal, cervical, breast and lung cancer (35–37). It has been reported that HP1γ represses the cyclin-dependent kinase inhibitor p21 in colon cancer cells and enhances prostate cancer progression by repressing expression of miR-451a (36, 38). HP1γ knockout mice study has reported that HP1γ may be not required for normal somatic cell proliferation although it positively regulates cell cycle of primordial germ cells (39). However, the in vivo tumorigenic role of HP1γ and its mode of action in cancer cells, especially lung cancer cells, have been poorly documented. Using a genetically engineered mouse model (i.e., the well-established K-Ras^G12D LUAD mouse), our current study demonstrated the in vivo requirement of HP1γ in LUAD tumorigenesis. In addition, our results revealed that HP1γ directly downregulated expression of the transcription-repressive regulators NCOR2 and ZBTB7A to indirectly upregulate expression of tumor-promoting factors (e.g., AXL, PVT1, and ELK1). We furthermore showed that NCOR2 repressed ELK1 expression while ZBTB7A downregulated expression of AXL and PVT1 (Fig. 7D–I). Thus, our findings, distinct from other HP1γ studies, provide unique epigenetic and mechanistic insights into LUAD pathogenesis (Fig. 7M).

In addition to H3K9 methylation readers, the H3K9 methyltransferases appear to regulate tumorigenesis. For example, G9a has been reported to repress the cell adhesion molecule Ep-CAM and to promote cell invasion and metastasis of lung cancer (40). Equally interesting, H3K9 methylation has been linked to silencing of tumor-suppressor genes, such as p16^INK4A (41). Therefore, it is possible that an H3K9 methylation-epigenetic system, including writers and readers, cooperates to promote lung tumorigenesis.

HP1γ has been associated with gene silencing of multiple genes, such as E2F- and Myc-responsive genes in G₀ cells and the HIV gene (21, 22). In contrast, HP1γ is enriched in the gene body, and its levels positively correlate with gene activity (20, 42). In the latter studies, HP1γ was linked to gene activation via transcriptional elongation and co-transcriptional RNA splicing. In the current study, we showed that HP1γ occupied and repressed expression of transcription-repressive regulator genes, such as NCOR2 and ZBTB7A, in LUAD cells (Fig. 4). Moreover, our results showed that NCOR2 and ZBTB7A mRNA levels inversely correlated with HP1γ mRNA levels in human LUAD datasets (Fig. 6B and C). Although we cannot exclude the possibility that HP1γ could directly activate the transcription of some tumor-promoting factors in LUAD cells, our results support the notion that HP1γ-mediated transcriptional repression of tumor-suppressor genes contributes to the progression of LUAD.

The transcriptional repressor ZBTB7A has a tumor-suppressive function that downregulates oncogenic glycolytic genes (24), and ZBTB7A undergoes mutations that disrupt the anti-proliferative function of ZBTB7A in leukemia (43). However, ZBTB7A has an oncogenic function for lymphoma and represses expression of the tumor-suppressor gene p14^ARF (25). Similar to ZBTB7A, the transcriptional corepressor NCOR2 appears to have pro-tumorigenic and anti-tumorigenic functions. NCOR2 is downregulated in multiple myeloma (44) and its loss is linked with the neoplastic transformation of non-Hodgkin’s lymphoma (45). However, increased NCOR2 levels are associated with faster recurrence of estrogen receptor α-positive breast tumors (46). It is possible that NCOR2 and ZBTB7A may function as tumor suppressors or oncogenic proteins in a tissue-dependent manner. In the current study, tumor-suppressive functions of NCOR2 and ZBTB7A in lung are supported by the following results. First, NCOR2 or ZBTB7A knockdown rescued, at least in part, defects in the proliferation, colony formation and migratory abilities of HP1γ-depleted LUAD cells (Fig. 5). Second, NCOR2 or ZBTB7A knockdown significantly increased colony formation and migration of LUAD cells (Fig. 6D–G). Third, our TCGA analysis showed that NCOR2 and ZBTB7A mRNA levels were downregulated in LUADs compared with their adjacent normal samples (Fig. 6H). Fourth, low NCOR2 and ZBTB7A mRNA levels correlated with shorter survival in human LUAD patients (Fig. 6I and J). Fifth, in vivo HP1γ knockdown upregulated NCOR2 and ZBTB7A levels in mouse K-Ras^G12D lung tumors (Fig. 6A) and NCOR2 and ZBTB7A mRNA levels were inversely associated with HP1γ mRNA levels in human LUAD datasets (Fig. 6B and C). Finally, our database analysis showed that NCOR2 and ZBTB7A underwent loss-of-function of mutations in LUAD samples (Supplementary Fig. S12A and B). For these reasons, it could be concluded that HP1γ promotes lung tumorigenesis, at least in part, by repressing expression of ZBTB7A and NCOR2.

Histone methylation reader proteins are frequently dysregulated in cancer. Such dysregulation includes recurrent mutations within the methylation reader domains, oncogenic fusion proteins containing readers, and overexpression of histone reader proteins. For instance, the H3K4me3/2-binding PHD in the inhibitor of growth 1 (ING1) is mutated in various tumors, including breast cancer and melanoma (47). Examples of oncogenic fusion proteins containing readers include fusion proteins between nucleoporin-98 and H3K4me3-binding PHD fingers in the PHD finger-containing proteins JARID1A and PHF23, which cause acute myeloid leukemia in mice (48). As reported here, HP1γ is an example of a reader protein that is overexpressed in LUAD and is a prognostic marker necessary for LUAD tumorigenesis. In addition to dysregulation in histone methylation reader proteins, alterations of their partner proteins affect their function and tumorigenesis. For example, a recent study from our laboratory has shown that ZYMND8 (also called RACK7) acts as a reader protein for the dual histone mark H3K4me1-H3K14ac via its PHD/bromodomain and suppresses metastasis-linked genes in cooperation with the H3K4 demethylase JARID1D (9). Because JARID1D is frequently deleted or downregulated in prostate tumors and their metastases (9, 10), the gene-repressive function of ZYMND8 at metastasis-linked genes may be impaired in these tumors. Similar to this, it is possible that the tumor-promoting activity of HP1γ is regulated by its partner proteins. With respect to HP1γ-interacting proteins, it has been shown that the chromatin remodeler CHD4 is isolated as an HP1γ-interacting protein (49) and has an oncogenic function by epigenetically suppressing multiple tumor-suppressor genes (50). Thus, it would be interesting to examine in the future whether CHD4 might cooperate with HP1γ for tumorigenesis.

The inhibition of the binding activities of oncogenic histone readers can be an attractive strategy to block their function. For example, the inhibition of the BET family of readers that recognize histone lysine acetylation has been shown to be a promising therapeutic strategy. Because our results indicate that HP1γ’s reader activity is required for LUAD tumorigenicity, abrogating the interaction of HP1γ with H3K9me2/3 using small molecule inhibitors may be relevant to therapeutic intervention for the treatment of LUAD. In summary, our findings provide new insights into how overexpression of a tumor-promoting histone reader protein represses expression of tumor suppressors to promote LUAD tumorigenicity and also highlight the clinical relevance of an HP1γ-regulated mode of mechanism to LUAD.

Significance.

Direct epigenetic repression of the transcription-repressive regulators NCOR2 and ZBTB7A by the histone reader protein HP1γ leads to activation of pro-tumorigenic genes in lung adenocarcinoma.