Inhibitor of growth protein 3 epigenetically silences endogenous retroviral elements and prevents innate immune activation.

Endogenous retroviruses (ERVs) are subject to transcriptional repression in adult tissues, in part to prevent autoimmune responses. However, little is known about the epigenetic silencing of ERV expression. Here, we describe a new role for inhibitor of growth family member 3 (ING3), to add to an emerging group of ERV transcriptional regulators. Our results show that ING3 binds to several ERV promoters (for instance MER21C) and establishes an EZH2-mediated H3K27 trimethylation modification. Loss of ING3 leads to decreases of H3K27 trimethylation enrichment at ERVs, induction of MDA5-MAVS-interferon signaling, and functional inhibition of several virus infections. These data demonstrate an important new function of ING3 in ERV silencing and contributing to innate immune regulation in somatic cells.

INTRODUCTION

Transposable elements (TEs) consist of repetitive nucleic acid sequences that make up as much as 40–60% of the mammalian genome (1,2). TEs are composed of DNA transposons and retrotransposons, which can be further categorized into elements with long terminal repeats (LTRs) such as endogenous retroviruses (ERVs) and TEs without LTRs such as long interspersed nuclear elements (LINEs) or short interspersed nuclear elements (SINEs) (3).

ERVs share substantial similarities with exogenous retroviral proviruses and are implicated in embryonic development, autoimmune diseases, and neurological disorders (4–6). Therefore, ERV transcription needs to be tightly regulated to provide genomic stability and avoid aberrant expression of neighboring genes and oncogenic transformation (7). However, only a handful of ERV regulators are known, including zinc-finger protein Zfp809, Trim28, Setdb1, Snerv and the HUSH complex, which epigenetically silence mouse ERV activation via H3K9 trimethylation (8–14). Even fewer epigenetic silencers are known for human ERVs: TRIM28 (11), TIP60 (14) and more recently KDM5B (15) and FBXO44 (16). Low expression of KDM5B or FBXO44 has been associated with ERV reactivation and an accelerated immune clearance of tumors (15,16). The suppressive mechanisms of most human ERVs in the genome remain largely unknown.

Inhibitor of Growth Family Member 3 (encoded by ING3) is a member of the ING tumor suppressor family (17) and encodes a plant homeodomain (PHD) finger, a motif common to chromatin-regulatory proteins. ING3 is found to be recruited to double-strand DNA breaks and contributes to DNA repair (18). ING3 is a key subunit of the human NuA4/Tip60 histone acetyltransferases complex (17). Recently, it was shown that TIP60 represses activation of ERV elements, which turns on STING-IRF7 mediated inflammatory responses (14). This suggested to us that ING3 may also participate in regulating ERV expression. However, there is little, if any, information on how ING3 may potentially impact virus infection and host innate immunity. In the current study, through a series of genetic, biochemical and functional experiments, we found that in cells lacking ING3, multiple ERV expression was re-activated and MDA5-MAVS signaling pathway was induced, which altogether identified ING3 to be a novel transcriptional repressor of ERVs.

MATERIALS AND METHODS

Cells, plasmids, and reagents

H1-Hela cells (CRL-1958) were obtained from American Type Culture Collection (ATCC) and cultured in complete Dulbecco's modified Eagle's medium (DMEM) medium. HT-29 cells (HTB38) were obtained from ATCC and cultured in complete advanced DMEM/F12 medium. ING3 KO HT-29 cells were transduced with pLenti-C-Myc-DDK-P2A-Puro empty vector, pLenti-C-Myc-DDK-P2A-ING3, pLenti-C-mGFP empty vector, or pLenti-C-mGFP-ING3 and cultured under puromycin selection (1 μg/ml). pSpCas9(BB)-2A-GFP (PX458) and lenti-CRISPR_v2 were purchased from Addgene (48138 and 52961, respectively). Ruxolitinib (Selleckchem, S1378) was reconstituted at 10 mM stock solution in dimethyl sulfoxide (DMSO) and used at 100 nM in cell culture. Recombinant human IFN-β was purchased from R&D Systems (285-IF-100). Poly(I:C) high molecular weight (InvivoGen, tlrl-piclv), LyoVec (InvivoGen, lyec-12), GSK126 (Selleckchem, S7061), JIB-04 (Tocris, 4972), and GSK-J4 (Tocris, 4594) were purchased from respective vendors.

CRISPR/Cas9 knockout cells

Single clonal knockout HT-29 cells were obtained using the PX458 vector that expresses Cas9 and single-guide RNA (sgRNA) against ING3 (Supplementary Table S1). Green fluorescent protein (GFP)-positive single cells were sorted at 48 h post-transfection using BD Aria II into 96-well plates and screened for knockout based on western blot and Sanger sequencing using primers listed in Supplementary Table S1. Pooled knockout H1-Hela cells were obtained by lentiviral transduction with the lenti-CRISPR_v2 vector that expresses Cas9 and sgRNA against CHD8, COG7, ING3, KDSR, MGAT1, MMGT1, MYC, PHIP, or RALGAPB for 14 days under puromycin selection (10 μg/ml). Pooled knockout HT-29 cells were obtained by lentiviral transduction of ING3 KO HT-29 cells with the lenti-CRISPR_v2 vector that expresses Cas9 and sgRNA against RIG-I (DDX58), MDA5 (IFIH1), MAVS, cGAS (MB21DB1), STING (TMEM173), TBK1, IRF3, or IRF7 for 14 days under puromycin selection (10 μg/ml). All sgRNA targeting sequences were provided in Supplementary Table S1.

Viruses and virus infections

All human and animal RV strains used in this study were propagated in MA104 cells and RV infection was performed as previously described (19). Recombinant VSV (strain Indiana) expressing GFP was a kind gift from Dr. Jack Rose (Yale University) and propagated and titrated in BHK cells. Coxsackievirus B3 strain was propagated and titrated in Hela cells. Lentiviruses used in this study include pLenti-C-Myc-DDK-P2A-ING3, pLenti-C-mGFP-ING3, and lenti-CRISPR_v2 vector encoding Cas9 and sgRNA against CHD8, COG7, ING3, KDSR, MGAT1, MMGT1, MYC, PHIP, RALGAPB, RIG-I (DDX58), MDA5 (IFIH1), MAVS, cGAS (MB21DB1), STING (TMEM173), TBK1, IRF3 or IRF7. All of them were packaged in HEK293T cells by co-transfection with psPAX2 and pMD2.G as previously described (20). Supernatants were collected at 48 and 72 h post transfection, passed through a 450 nm filter, and added to target cells in the presence of polybrene (8 μg/ml).

Protein analysis

Western blot was performed using primary antibodies against cGAS (Cell Signaling Technology (CST), 15102S, 1:1000), EZH2 (CST, 5246S, 1:1000), GAPDH (CST, 5174S, 1:1000), H3K27me3 (CST, 9733S, 1:1000), Histone H3 (CST, 14269S, 1:1000), ING3 (Sigma-Aldrich, HPA067575, 1:500), IRF3 (CST, 11904S, 1:1000), IRF7 (CST, 72073S, 1:1000), MAVS (CST, 3993S, 1:1000), MDA5 (CST, 5321S,1:1000), RIG-I (CST, 3743S, 1:1000), STING (CST, 13647S, 1:1000) and TBK1 (CST, 3504S, 1:1000). IFN-λ secretion was quantified by the DuoSet ELISA for Human IL-29/IL-28B (IFNL-1/3) kit (R&D Systems).

Reverse transcription and quantitative PCR

Total RNA was extracted from cells using RNeasy Mini kit (Qiagen) and reverse transcription was performed using High Capacity RT kit and random hexamers as previously described (21). Quantitative PCR was performed on Stratagene Mx3005P (Agilent) with a 25 μl reaction, composed of 50 ng of cDNA, 12.5 μl of Power SYBR Green master mix (Applied Biosystems), and 200 nM both forward and reverse primers. All SYBR Green primers used in this study (Supplementary Tables S2 and S3) have been validated with both dissociation curves and electrophoresis of the correct amplicon size. Taqman primers used in this study were also provided in Supplementary Table S2.

RNA sequencing

Total RNA from WT and ING3 KO HT-29 cells was extracted using the RNeasy Mini Kit (Qiagen). RNA sample quality was examined by NanoDrop spectrophotometer (Thermo Fisher) and Bioanalyzer 2100 (Agilent). Total RNA sequencing libraries were constructed and sequenced on BGISEQ-500 platforms. Clean reads were generated by SOAPnuke (v.1.5.6) with the following parameters: -n 0.001 -l 20 -q 0.4 -A 0.25. The SE reads (50 bp) were aligned to the hg38 build using Bowtie2 (v.4.8.2 20140120) to map clean reads to reference gene and using HISAT2 (v.4.8.2) to reference genome (Homo_sapiens GCF_000001405.38_GRCh38.p12) with the following parameters: –phred64 –sensitive -I 1 -X 1000. KEGG (v.93.0) was used for gene annotation. Fragments per kilobase of transcript per million (FKPM) reads were counted using Subread (v.1.4.6), and differential gene expression analysis was performed using DESeq2 (v.1.16.1) based on the non-parametric NOIseq method (22). The FKPM information was provided as Supplementary Dataset 1.

Preparation of dsRNA

Cells from two biological replicates were lysed in TRIzol reagent (Invitrogen) and kept at −80°C until use. Total cellular RNA from wild-type or ING3 KO HT-29 cells was extracted using Direct-zol RNA Miniprep Kit (Zymo Research). Extracted RNA was treated enzymatically with RNase-free DNase I (Ambion) 37°C for 1 hour, and recovered using RNA Clean & Concentrator-25 (Zymo Research). RNA quantification was done using a Qubit RNA Broad Range Assay kit and Qubit fluorometer according to the manufacturer's instructions. DsRNA was prepared by treating 100 mg of total cellular RNA in 0.3 M NaCl buffer with 200 units of S1 nuclease (Thermo Fisher Scientific) at 37°C for 45 min (23), and size of dsRNA longer than 200 nucleotides was purified and used in the transfection experiments.

Chromatin immunoprecipitation

ChIP was performed following the protocol of Pierce™ Magnetic ChIP Kit (Thermo Fisher Scientific, 26157). Briefly, WT and ING3 KO HT-29 cells complemented with empty vector or pLenti-C-Myc-DDK-P2A-ING3 were cross-linked with 1% formaldehyde for 10 min at room temperature. The cells were treated using Membrane Extraction Buffer containing protease/phosphatase inhibitors and incubated on ice for 10 min. After centrifuge, cells were resuspended with MNase Digestion Buffer Working Solution and MNase in 37°C for 15 min. The lysate was then sonicated for 10 cycles at 30% amplitude (30 s ON and 30 s OFF). The sonicated samples were then diluted in ChIP dilution buffer and used for the immunoprecipitation with anti-Myc (Thermo Fisher Scientific, MA116637). The EZH2, H3K27me3, and H3K9me3 ChIP was performed using the following antibodies: EZH2 (CST, 5246S), H3K27me3 (CST, 9733S), H3K9me3 (Abcam, ab8898), Histone H3 (CST, 4620S). After an over-night incubation with antibody, ChIP Grade Protein A/G Magnetic Beads were added to each IP and incubate at 4°C overnight with mixing. The bound DNA was washed sequentially with IP Elution Buffer and reverse cross-linked using NaCl for 1.5 hr at 65°C. The eluted DNA was purified and used for qPCR reactions. QPCR was performed as described above using primers described in Supplementary Table S4.

Immunoprecipitation

Immunoprecipitation was performed as described previously (24). Briefly, cells were washed by ice-cold PBS for 3 times, then lysed by NP40 cell lysis buffer (Invitrogen, FNN0021) supplemented with protease inhibitor cocktail (Thermo Scientific, 78430) and PMSF (Santa Cruz, sc-482875) on ice for 30 min. Cell lysates were clarified by centrifugation at 14 000 g for 10 min at 4°C. The lysates were incubated with 2 μg of anti-EZH2 antibody (CST, 5246S), or normal rabbit IgG (Millipore, 12-370) at 4°C overnight. In the meantime, Dynabeads (Invitrogen, 10004D) were incubated with 5% BSA diluted in PBS to reduce the non-specific background at 4°C overnight. Next, the antibody-lysates mixture was added to the dynabeads at 4°C for 4 h. The complex was washed with ice-cold NP40 lysis buffer for three times before the complex was eluted in 2× Laemmli sample buffer (Bio-Rad, 1610737) at 95°C for 5 min.

Statistical analysis

All bar graphs were displayed as means ± SEM. Statistical significance of data in Figures 1D, 2B, E, 4A–D, Supplementary Figures S2F, and S3D was calculated by Student's t test using Prism 8 (GraphPad). Statistical significance of data in Figures 1A–C, 2C–D, 3B–D, Supplementary Figures S1E, and S3C was calculated by pairwise ANOVA using Prism 8. All data were presented as asterisks (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001). All experiments other than Figure 2A and Supplementary Figure S1C have been repeated at least twice. The transcriptome analysis of wild-type and ING3 KO HT-29 cells in Figure 2A was performed once in duplicate. The raw data is included in Dataset S1.

Figure 1.

ING3 deficient cells are resistant to RV infection. (A) WT and ING3 KO (clones #1, #2, and #3) HT-29 cells were infected with bovine RV UK strain (MOI = 1) for 24 h. RV NSP5 level was measured by RT-qPCR and normalized to that of GAPDH. (B) WT and ING3 KO HT-29 cells transduced with lentiviruses either in empty vector (ev) or encoding mGFP-ING3 were infected with bovine RV UK strain (MOI = 1) for 24 h. RV NSP5 level was measured by RT-qPCR and normalized to that of GAPDH. (C) Same as (B) except that at 24 h post infection, infectious RV titers were determined by a focus-forming unit assay. (D) WT and ING3 KO HT-29 cells were infected with a panel of different human and animal RV strains (MOI = 1) for 24 h. RV NSP5 level was measured by RT-qPCR and normalized to that of GAPDH. For all panels, experiments were repeated at least three times with similar results. Data are represented as mean ± SEM. Statistical significance is from pooled data of the multiple independent experiments (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001).

Figure 2.

Loss of ING3 induces robust IFN and ISG expression. (A) Volcano plots of RNA-sequencing analysis of WT and ING3 KO (clones #1 and #2) HT-29 cells. Genes that were up-regulated in ING3 KO cells are highlighted in red, and the highly induced ISGs were marked with more prominent red dots (Padj: adjusted p-value). (B) IFI44L and IFITM1 transcript levels were measured by RT-qPCR and normalized to that of GAPDH in WT and ING3 KO HT-29 cells. (C) IFNL3 transcript level was measured by RT-qPCR and normalized to that of GAPDH in WT and ING3 KO HT-29 cells transduced with lentiviruses in empty vector (ev), encoding mGFP-ING3, or encoding Myc-ING3. (D) Same as (C) except that cell culture supernatants were collected and secreted IFN-λ levels were measured by ELISA. (E) WT and ING3 KO HT-29 cells were treated with vehicle control DMSO or JAK inhibitor ruxolitinib (100 nM) for 24 hr and infected with RV (MOI = 1) for another 24 h. RV NSP5 level was measured by RT-qPCR and normalized to that of GAPDH. For all panels except A, experiments were repeated at least three times with similar results. Figure A was performed once with technical duplicates. Raw data is listed in Dataset S1. Data are represented as mean ± SEM. Statistical significance is from pooled data of the multiple independent experiments (**P ≤ 0.01; ***P ≤ 0.001).

Figure 4.

ING3 recruits EZH2 and H3K27 trimethylation to the ERV promoters. (A) ING3 KO HT-29 cells transduced with lentiviruses either in empty vector (ev) or encoding Myc-ING3 were subjected to chromatin immunoprecipitation (ChIP) using an anti-Myc antibody and RT-qPCR assay to examine the forward (FWD) and reverse (REV) LTR regions of indicated ERVs. (B) WT and ING3 KO HT-29 cells were cultured with vehicle control DMSO or EZH2 inhibitor GSK126 (10 μM) for 6 days. MER21C, MER4D and ERVL transcript levels were measured by RT-qPCR and normalized to that of GAPDH. (C) WT and ING3 KO HT-29 cells were subjected to ChIP using anti-EZH2 or Histone H3 antibodies and RT-qPCR assay to examine the forward (FWD) and reverse (REV) LTR regions of MER21C. (D) Same as (C) except that H3K27 trimethylation (H3K27me3) or H3K9 trimethylation (H3K9me3) antibodies were used in ChIP instead of EZH2. For all panels, experiments were repeated at least three times with similar results. Data are represented as mean ± SEM. Statistical significance is from pooled data of the multiple independent experiments (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001).

Figure 3.

ERVs activate MDA5-MAVS signaling in ING3 KO cells. (A) WT and ING3 KO HT-29 cells were transduced with lentiviruses encoding Cas9 and single-guide RNAs (sgRNAs) against the indicated genes and selected under puromycin for 10 days. Cell lysates were harvested and the indicated proteins levels were examined by western blot. (B) Same as (A) except that the cellular RNA was harvested and IFNL3 level was measured by RT-qPCR and normalized to that of GAPDH. (C) Same as (A) except that cells were infected with RV (MOI = 1) for 24 hr. RV NSP5 level was measured by RT-qPCR and normalized to that of GAPDH. (D) Same as (C) except that at 24 hr post infection, infectious RV titers were determined by a focus-forming unit assay. (E) WT HT-29 cells were transfected with LyoVec in complex with dsRNA longer than 200 nucleotides from WT HT-29 cells (10 ng), dsRNA from ING3 KO HT-29 cells (10 ng), or high-molecular weight poly(I:C) (100 ng) for 24 hr. IFNL3 transcript level was measured by RT-qPCR and normalized to that of GAPDH. (F) Indicated ERV transcript levels were measured by RT-qPCR and normalized to that of GAPDH in WT and ING3 KO HT-29 cells. For all panels, experiments were repeated at least three times with similar results. Data are represented as mean ± SEM. Statistical significance is from pooled data of the multiple independent experiments (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001).

RESULTS

In a genome-wide CRISPR/Cas9 loss-of-function screen for rotavirus (RV) dependence factors, we identified a number of host genes with novel pro-viral functions (25). Here, we performed a focused validation screen using a lentivirus based pooled CRISPR/Cas9 knockout (KO) approach. Genetic deletion of ING3 in H1-Hela cells decreased RV gene NSP5 mRNA levels and viral titers as compared to mock transduced cells (Supplementary Figure S1A, B). We next generated single clonal ING3 KO HT-29 cells, a human colonic epithelial cell line commonly used for RV research (26). RV replication was significantly reduced (>100-fold) in three independent ING3 KO clones (Figure 1A). These cells did not exhibit noticeable defects in survival or proliferation (data not shown). Complete ING3 deletion was confirmed by Sanger sequencing of the genomic locus targeted by Cas9 and western blot (Supplementary Figure S1C, D). Compared to the wild-type HT-29 cells, ING3 KO cells had substantially lower levels (∼500-fold) of bovine RV UK strain replication, as measured by RV RNA levels and virus titers (Figure 1B, C). ING3 supported the replication of a variety of human and animal RV strains (Figure 1D). Importantly, transduction of ING3 KO HT-29 cells with lentiviruses encoding monomeric GFP (mGFP)-tagged ING3 partially restored RV replication (Figure 1B, C and Supplementary Figure S1D). Functional complementation of genetic deletion of ING3 was also verified by a lentiviral vector that encodes Myc-tagged ING3 (Supplementary Figure S1E).

In addition to RV, which is a double-stranded (ds) RNA virus, ING3 KO cells were also highly resistant to infections by single-stranded (ss)RNA (+) and ssRNA (-) viruses such as coxsackievirus B3 strain and vesicular stomatitis virus, respectively (Supplementary Figure S2A), suggesting that ING3 may regulate a cellular pathway common to different RNA virus families. To shed light on the molecular mechanisms of the broad pro-viral effect of ING3, we profiled the cellular transcriptomes of wild-type and two independent clones of ING3 KO HT-29 cells by RNA-sequencing. Both gene ontology and KEGG pathway analysis revealed up-regulation of pathways mediating innate immune activation towards viruses and endogenous stimuli in ING3 KO cells (Supplementary Figure S2B-C). The expression of canonical interferon (IFN)-stimulated genes (ISGs) such as IFI44L and IFITM1, was up-regulated by over 100-fold in ING3 KO cells compared to wild-type HT-29 cells (Figure 2A). This ISG signature was confirmed by reverse transcriptase-quantitative polymerase chain reaction (RT-qPCR) (Figure 2B and Supplementary Figure S2D). Consistent with our previous observation with primary intestinal epithelial cells and cell lines (19), type III IFN expression was preferentially induced in ING3 KO cells compared to wild-type HT-29 cells (Figure 2C and Supplementary Figure S2E). In the absence of virus infection, we detected robust IFN-λ secretion in the supernatants of ING3 KO HT-29 cells by enzyme-linked immunosorbent assay (ELISA) (Figure 2D). IFN-λ mRNA and secreted cytokine levels were partially restored to homeostatic status upon re-expression of mGFP-tagged or Myc-tagged ING3 (Figure 2C, D).

We reasoned that the heightened IFN and ISG responses could mediate broad resistance to virus infections in ING3 KO cells. Indeed, treatment of cells with a JAK kinase inhibitor ruxolitinib, which effectively blocks STAT1 phosphorylation (27), led to a dramatic reduction in IFN and MX1 expression (Supplementary Figure S2F) and partially restored the susceptibility to RV infection in ING3 KO HT-29 cells (Figure 2E). Taken together, these data establish ING3 as a previously unknown regulator of IFN signaling.

We next dissected the cellular pathway required for IFN induction triggered by ING3 deficiency. Since ING3 mediates the repair of double-strand DNA breaks (18), our data are reminiscent of findings with STAG2 (25) and Banf1 (28), of which genetic ablation leads to STING activation and IFN induction. Hence we posited that the cytoplasmic DNA sensing pathway might be involved in ING3’s innate immune regulatory activity. We used CRISPR/Cas9 editing to knock out major components in the IFN induction signaling pathway, including RIG-I, MDA5, MAVS, cGAS, STING, TBK1, IRF3 and IRF7. The levels of these proteins were efficiently depleted in ING3 KO cells (Figure 3A). Surprisingly, deletion of the dsRNA sensor MDA5 or downstream factors of MDA5, including MAVS, TBK1, and IRF3, led to a restoration of the baseline IFN mRNA levels (Figure 3B). Consistently, CRISPR-mediated inhibition of the MDA5-MAVS-TBK1-IRF3 signaling enhanced intracellular RV RNA levels and virus titers in the ING3 KO cells (Figure 3C, D). In contrast, knocking out cytosolic DNA sensor cGAS or its adaptor protein STING had a relatively modest effect on IFN expression and virus infection (Figure 3B–D).

To further probe the role of dsRNAs in IFN responses in WT and ING3 KO cells, we used S1 nuclease to generate dsRNAs from total cellular RNAs (23). When transfected into cells, dsRNA from ING3 KO HT-29 cells induced higher (2.9-fold) IFN mRNA levels than dsRNA of wild-type HT-29 cells (Figure 3E). We next sought to determine the nature of cellular dsRNA ligands that activate MDA5 signaling in ING3 KO cells. A number of potential sources for dsRNAs are known, including circular RNA, mitochondrial RNA, and retrotransposon elements such as ALU and ERVs (29–31). To assess whether any known TEs were regulated upon ING3 depletion, we quantified the levels of 24 individual ERVs through RT-qPCR, spanning human ERV classes I and III (32). Most ERVs examined were up-regulated by 8–16-fold in ING3 KO cells compared to wild-type HT-29 cells (Figure 3F). Augmented expression of multiple ERVs was verified in a separate ING3 KO clone (Supplementary Figure S3A). In contrast, transcripts from other TEs such as ALU and LINE-1 were not affected (Supplementary Figure S3B), and complementation of ING3 KO cells with exogenously introduced ING3 restored the expression of ERVs to homeostatic levels (Supplementary Figure S3C), highlighting the specific ING3 targeting of ERVs. To exclude the possible secondary effect of IFN induction on ERV transcription, we measured the mRNA levels of ERVs, which were not induced by IFN-β treatment as canonical ISGs OAS3 and MX1 (Supplementary Figure S3D).

Next, we tested whether ING3 is present at the ERV promoters by chromatin immunoprecipitation (ChIP). With a weak antibody to detect endogenous ING3 (Supplementary Figure S1D), we turned to the ING3 KO cells complemented with Myc-tagged ING3 (Supplementary Figure S1E). Anti-Myc ChIP in rescued cells revealed enriched occupancy of ING3 at the LTR regions of several ERVs, including MER21C, MER4D and MLT2B4 (Figure 4A), all of which were up-regulated upon ING3 deletion (Figure 3F). In contrast, anti-Myc ChIP assay in rescued cells showed that ING3 did not bind to the promotor regions of LINE-1 and ALU elements (Supplementary Figure S4A), both of which were not activated in ING3 KO HT-29 cells (Figure 3F). In addition, anti-Myc ChIP assay also specifically detected ING3 binding at the RPSA promoter as previously described (33), and this was highly specific to Myc-ING3 expressing cells and not the control ING3 KO cells (Supplementary Figure S4B). ING3 was reported to interact with multiple epigenetic regulators, including KAT5/TIP60 complex, DNA methyltransferase 1 associated protein 1, and polycomb-group proteins (17,34). To define the mechanism of how ING3 potentially precludes ERV activation at steady state, we utilized a set of small-molecule inhibitors that target respective histone methyltransferases and demethylases. Treatment of wild-type HT-29 cells with GSK126, a highly selective inhibitor of enhancer of zeste homolog 2 (EZH2), an H3K27 methyltransferase (35), resulted in an appreciable increase of MER21C, MER4D, and ERVL expression (Figure 4B), without inducing a global decrease of H3K27 trimethylation (Supplementary Figure S5A). GSK126 treatment did not alter the expression levels of LINE-1 or ALU elements (Supplementary Figure S5B). MER21C expression was not reactivated by histone demethylase inhibitors JIB-04 and GSK-J4 (Supplementary Figure S5C). The enhancing effect of GSK126 treatment on ERV levels was limited in ING3 KO cells (Figure 4B). Using ING3 KO HT-29 cells complemented with Myc-ING3, we found that EZH2 co-precipitated with ING3 (Supplementary Figure S5D). We then detected by ChIP-qPCR that the degree of EZH2 binding reduced in ING3 KO HT-29 cells compared to wild-type cells (Figure 4C). These data suggest that ING3 likely recruits EZH2 to the ERV sites for transcriptional suppression.

Consistent with EZH2 binding, we measured hallmarks of chromatin repression and found more H3K27 trimethylation but not H3K9 trimethylation at the MER21C promoter in wild-type HT-29 cells and it was diminished significantly in the absence of ING3 (Figure 4D). Although DNA methyltransferase inhibitor treatment was previously shown to induce ERV activation and IFN production (36,37), the methylation status of the MER21C promoter was not altered in the presence or absence of ING3 (data not shown). Further, treatment of wild-type cells with 5-aza-2′-deoxycytidine, a DNA methyltransferase inhibitor, did not trigger ERV reactivation (data not shown), suggesting a different mechanism of ING3-mediated epigenetic silencing. In summary, our data demonstrate that ING3 is a new ‘molecular checkpoint’ of a broad range of ERVs, whose expression needs to be tightly regulated to prevent autonomous innate immune activation.

DISCUSSION

Here, using a genome-wide CRISPR/Cas9 screen, RNA sequencing, and ChIP, we determined that ING3 functions as a novel ‘gatekeeper’ of ERV expression and IFN homeostasis. ING3 deficiency leads to ERV de-suppression, activation of the MDA5-MAVS signaling, and excessive IFN production. Of note, both ING1 and ING5, members in the same family as ING3, were also among the top 200 hits of our CRISPR/Cas9 screen (25). Whether they play a similar or distinct role in ERV suppression remains to be determined.

Although TIP60 is known to associate with ING3 to regulate gene expression (17,34), TIP60 deficiency led to a loss of H3K9 trimethylation and activation of the ERV promoters (14). However, we found that in the case of ING3, H3K27 trimethylation via EZH2 but not H3K9 trimethylation is the predominant repressive marker (Figure 4B-D), suggesting a distinct silencing mechanism. Since ING3 encodes a highly conserved PHD domain (38), we speculate that unlike the other epigenetic regulators, ING3 is able to directly recognize and bind DNA sequences and motives. In future studies, it will be of interest to determine whether there are significant overlaps in the ERVs targeted by KAP1, TIP60, KDM5B, FBXO44, versus ING3.

The potential limitations of the study include the specificity of ING3 targeting ERVs and whether the IFN induction is directly mediated through ERVs. Specifically, not all repetitive elements seem to be regulated by ING3, the underlying mechanisms of which we will examine by ChIP-seq analysis in future studies. Also, we have not definitively shown that the dsRNA ligands directly originated from ERV sequences. To address this problem, we may be able to take advantage of S1 nuclease digestion to enrich cytoplasmic dsRNA molecules and couple with RNA-seq analysis.

In the human population, the occurrence of ING3 loss-of-function mutations is substantially under-represented (observed-expected ratio of loss-of-function variants in ING3 = 0.07; missense variants = 0.56) according to the Genome Aggregation Database (39). This suggests a highly detrimental outcome of pre-mature stop codons in ING3, especially because loss of Ing3 has been associated with embryonic lethality in mice (40). These alleles are potentially too rare to be directly linked to autoimmune diseases via genome-wide association studies, but our results and the regulatory mechanism described can influence MDA5 activation thresholds in humans and hence might affect the likelihood of some autoimmune disorders or cancer progression. Further in vivo studies will test whether selective inhibition of ING3 induces viral mimicry in cancer cells and promotes anti-tumor immunity.

DATA AVAILABILITY

Raw data is deposited and available at GEO (GSE166219).

Click here for additional data file.