Chromatin remodeling: demethylating H3K4me3 of type I IFNs gene by Rbp2 through interacting with Piasy for transcriptional attenuation

Abstract

Type I IFNs (IFNIs) are involved in the course of antiviral and antimicrobial activities; however, robust inductions of these can lead to host immunopathology. We have reported that the Pias (protein inhibitor of activated signal transducer and activator of transcription) family member, Piasy, possesses the ability to suppress IFNI transcriptions in mouse embryonic fibroblasts (MEFs), yet the specific molecular mechanism by which it acts remains elusive. Here, we identify that the H3K4me3 levels, one activation mark of genes, in MEFs that were stimulated by poly(I:C) were impaired by Piasy in the IFN-β gene. Piasy bound to the promoter region of the IFN-β gene in MEFs. Meanwhile, retinoblastoma binding protein 2 (Rbp2) was proven to be the only known and novel H3K4me3 demethylase that interacted with Piasy. Overexpression of Rbp2, but not its enzymatically inactive mutant Rbp2^H483G/E485Q, retarded the transcription activities of IFNI, whereas small interfering RNA–mediated or short hairpin RNA–mediated knockdown of Rbp2 enhanced IFNI promoter responses. Above all, coexpression of Piasy and Rbp2 led to statistically less IFNI induction than overexpression of either Piasy or Rbp2 alone. Mechanistically, Piasy bound to the Jmjc domain (451–503 aa) of Rbp2 via its PINIT domain (101–218 aa), which is consistent with the domain required for their attenuation of transcription and H3K4me3 levels of IFNI genes. Our study demonstrates that Piasy may prevent exaggerated transcription of IFNI by Rbp2-mediated demethylation of H3K4me3 of IFNI, avoiding excessive immune responses.—Yu, X., Chen, H., Zuo, C., Jin, X., Yin, Y., Wang, H., Jin, M., Ozato, K., Xu, S. Chromatin remodeling: demethylating H3K4me3 of type I IFNs gene by Rbp2 through interacting with Piasy for transcriptional attenuation.

Type I IFNs (IFNIs), including IFN-β (3 subtypes) and IFN-α (14 subtypes), are composed of a series of pleiotropic glycoproteins that are secreted by cells responding to virus or microorganism infection (1, 2); however, the substantial generation of IFNI can result in autoimmune disease, cytotoxicity, or inflammation (3, 4). To maintain cellular homeostasis, negative feedback must be activated by the host. Our preliminary study has shown that the Pias (protein inhibitor of activated signal transducer and activator of transcription) family member, Piasy, plays a crucial role in diminishing IFNI gene transcription mediated by TLRs and RIG-I–like receptor signaling (5); however, the exact molecular mechanism of TLR-mediated Piasy inhibition of IFNI gene transcription is not fully defined. TLRs critically participate in the innate immune responses to several kinds of pathogens, such as viruses. Mouse embryonic fibroblasts (MEFs) express high levels of TLR1–9, which are highly responsive to known TLR ligands, such as poly(I:C) activation, both in terms of mRNA elaboration and protein expression (6). Moreover, MEFs can produce IFNI, IL-6, TNF-α, etc., which play a significant role in the course of antivirus and inflammation, making these universal cells with which to investigate immunology (5, 7).

Pias family proteins, including Pias1, Piasx, Pias3, and Piasy 4, contain 4 conserved domains: a DNA-binding domain, SAP, which contains an LXXLL motif that is necessary for IFN-stimulated gene expression; a PINIT domain for nuclear localization; a RING finger-like domain for its small ubiquitin-like modifier (SUMO) E3 ligase; and a SIM motif for SUMO binding (5, 8, 9). In addition, Pias proteins are involved in the regulation of the expression of many transcription factors or regulators, which play essential roles in immune responses and cell motility (5, 10, 11). As Pias proteins possess the activity of SUMO E3 ligase, they can regulate transcription via SUMO-dependent mechanisms (5, 10, 11). Moreover, additional functions, such as decreasing DNA contact of transcription factors, recruiting transcriptional corepressors, cooperating with other Pias members, and translocation of transcription factors, have also been adopted by Piasy to adjust transcription (12^,–18).

Beginning in the 1990s, the epigenetic mechanism of histone modification has become a field of interest for the study of transcriptional regulation (19^,–21). To date, numerous histone methylases and demethylases have been discovered to control transcription by affecting histone modification (22^,–24). Retinoblastoma binding protein 2 (Rbp2) belongs to the Jarid1 (Jumonji/ARID domain-containing protein 1) family of histone demethylases, which can specifically impair dimethylation and trimethylation of histone (25, 26). Jarid1 family members contain a Jmjc domain, which allows them to possess demethylase activity (27). Rbp2 can regulate the expression of a substantial number of genes with a specially recognizing DNA sequence CCGCCC via K152 in its ARID domain (28). Furthermore, Rbp2 is reported to interact with P50 and the Socs promoter, which demethylates H3K4me3 of the Socs promoter and leads to increased IFN-γ production in NK cells (29). Rbp2 is also proposed to be recruited to the E-cadherin promoter to inhibit its expression (30). It is noted that immune cells have high Rbp2 expression. In addition, histone methyltransferase Ash1l-mediated H3K4 methylation is involved in immune response and inflammatory autoimmune disease (31). In conclusion, Rbp2 and histone methylation play critical roles in transcriptional regulation and immunoregulation. Whether Rbp2 takes part in inhibitory IFNI activity mediated by Piasy remains to be revealed; however, its characteristics suggest that it is a significant candidate that contributes to transcriptional modulation.

In our study, we found that Rbp2 interplays with Piasy, and we describe their collaborative role in attenuating IFNI transcription. Our data indicate that Piasy binds to the promoter region of the IFN-β gene and is likely to recruit Rbp2, a histone demethylase, to IFNI genes to demethylate H3K4me3 and reduce the transcription of IFNI genes, which protects the host against excessive immune response. This Piasy-Rbp2-H3K4me3-IFNI signaling pathway may be a novel mechanism for the inhibition of excessive immune reactions by IFNI, and targeting Rbp2 may have a treatment advantage for the prevention of autoimmune diseases.

MATERIALS AND METHODS

Reagents and Abs

Different reagents and Abs were obtained from various companies: poly(I:C) was obtained from InvivoGen (San Diego, CA, USA), polybrene from Millipore (Billerica, MA, USA), Lipofectamine 2000 from Thermo Fisher Scientific (Waltham, MA, USA), Fugene 6 from Roche (Basel, Switzerland), puromycin from Cayman Chemical (Ann Arbor, MI, USA), glycogen from Thermo Fisher Scientific, phenol solution from Sigma-Aldrich (St. Louis, MO, USA), small interfering RNA (siRNA) from RiboBio (Guangzhou, China), protease inhibitor cocktails from Roche, anti-V5 from Comwin (Beijing, China), anti-glyceraldehyde 3-phosphate dehydrogenase from Cell Signaling Technology (Danvers, MA, USA), anti-H3K36me3 from Cell Signaling Technology, anti-H3K27me3 from Cell Signaling Technology, anti-H3K4me3 from Millipore, anti-Flag from Sigma-Aldrich, anti–β-tubulin from Comwin, ECL kit from Pierce (Rockford, IL, USA), Trizol from Thermo Fisher Scientific, anti-Flag M2 affinity gel from Sigma-Aldrich, anti-HA affinity matrix from Roche, anti-HA from Roche, anti-Rbp2 from Abcam (Cambridge, United Kingdom), PrimeScript One Step RT-PCR Kit from Takara (Dalian, China), SYBR Premix from Takara, and dual-luciferase reporter assay kit from Promega (Madison, WI, USA).

Cell culture

HEK293T, NIH3T3, and MEF cells from Piasy^+/+, Piasy^−/−;Pias1^+/+, or Pias1^−/− mice, respectively, were maintained in DMEM/high glucose that was supplemented with 10% fetal bovine serum. Platinum E cells were cultured in DMEM/ high glucose that contained 10% fetal bovine serum, 1% penicillin/streptomycin, and a final 1 μg/ml puromycin. All cells were cultured in a 95% air/5% CO₂ incubator at 37°C.

Newcastle disease virus transduction and retrovirus infection

Adherent Piasy^+/+ or Piasy^−/− MEFs that were approximately 70% confluent were infected with Newcastle disease virus (NDV; Heartz strain) at a multiplicity of infection of 2. Cell RNA was harvested by using Trizol at different time points. (Experiments were conducted in K.O.’s laboratory.) Retrovirus-expressing shRbp2, HA-Rbp2, Flag, Flag-Piasy, or C-terminal truncated Piasy were packaged by transfecting platinum E cells with 8 μg of corresponding retrovirus expression vector using 20 μl of lipofectamine 2000. Each retrovirus was used to infect adherent Piasy^+/+, Piasy^−/−, Pias1^+/+, and Pias1^−/− MEFs that were approximately 70% confluent with equal volume of culture medium using 8 μg/ml polybrene. Plates were then centrifuged at 2000 rpm for 1 h at room temperature after equilibrating to 37°C. Cells were subsequently incubated in 2 μg/ml puromycin (Piasy^+/+ MEFs, NIH3T3, Pias1^+/+ MEFs, and Pias1^−/− MEFs) and 1.5 μg/ml puromycin (Piasy^−/− MEFs) for 2 wk.

Real-time quantitative PCR

Total RNA was extracted by using Trizol according to manufacturer protocol. Reverse transcriptase reactions were conducted with the PrimeScript One Step RT-PCR Kit. Real-time quantitative PCR (qPCR) reactions were performed on a Roche Light Cycler 480 system. Primers that were used for qRT-PCR reactions were obtained from Thermo Fisher Scientific and are given in Table 1. Gene mRNA levels were normalized to hypoxanthine-guanine phosphoribosyl transferase and analyzed by using the 2^−ΔCt method. All experiments were conducted in triplicate.

TABLE 1.

Primer sequences

Primer name	Primer sequence, 5′–3′
	Forward	Reverse
Mouse IFN-β	GCTCCTGGAGCAGCTGAATG	CGTCATCTCCATAGGGATCTTGA
Mouse IFNα4	TGATGAGCTACTACTGGTCAGC	GATCTCTTAGCACAAGGATGGC
Mouse IFN-αs	CCTGAGARAGAAGAAACACAGCC	GGCTCTCCAGAYTTCTGCTCTG
Mouse HPRT	GCTCGAGATGTCATGAAGGAGAT	AAAGAACTTATAGCCCCCCTTGA
Human IFN-β	AAACTCATGAGCAGTCTGCA	AGGAGATCTTCAGTTTCGGAGG
Human IFN-αs	GTGAGGAAATACTTCCAAAGAATCAC	TCTCATGATTTCTGCTCTGACAA
IFN-β-2.4 kb	CTCCAACCCACCCACTCC	CAACAATATGAACTAACCAG
IFN-β-500bp	GCAGTTTCTCCAGGATGTGG	TCCTGAAAGGATGGCACAGC
IFN-β-120 bp	GGCTGTCTCCTTTCTGTT	TCTCCCTTTCAGTTTTCC
IFN-β+177 bp	GCTCCAAGAAAGGACGAAC	TCTCATTCCACCCAGTGCT
IFN-β+588 bp	CTGGAGGGTGCAAAGGTAC	CTACCAGTCCCAGAGTCCG
IFN-β+2.3 kb	TGGGAGTCAAGTACAAGTT	TTTCACTTATTCTCCACTG
shRbp2	GAAGTTAGCTAAAGAAGAA
Rbp2 K152E	AGAGACCCAGTTCCTTCTCCTGGCAGATATCCC	GGGATATCTGCCAGGAGAAGGAACTGGGTCTCT
Rbp2 H483G/E485Q	GGAATAACTCCAGTGATCCTGAATGCCCCAGCAAAAAGAAGAGAAGC	GCTTCTCTTCTTTTTGCTGGGGCATTCAGGATCACTGGAGTTATTCC
Rbp2 K152A	AAGAGACCCAGTTCCTGCTCCTGGCAGATATCCCAAGC	GCTTGGGATATCTGCCAGGAGCAGGAACTGGGTCTCTT
Piasy C335A	CCGTGCAGAGACCGCCGCACACCTGCAG	CTGCAGGTGTGCGGCGGTCTCTGCACGG
Piasy W356A	GAGAAGAAGCCCACCGCGATGTGCCCTGTGTG	CACACAGGGCACATCGCGGTGGGCTTCTTCTC
Piasy C340A	CTGCGCACACCTGCAGGCCTTTGATGCTGTGTTC	GAACACAGCATCAAAGGCCTGCAGGTGTGCGCAG
Piasy C330A	CCTCTCGGTGCCCGCCCGTGCAGAGACC	GGTCTCTGCACGGGCGGGCACCGAGAGG
Pias1 C351S	CTGAAGGTGGGAGCTGGTAAGTGCCCGAC	GTCGGGCACTTACCAGCTCCCACCTTCAG
IRF7 K406R	GGCTCCAGCCTCACCAGGATCAGGGTC	GACCCTGATCCTGGTGAGGCTGGAGCC

HPRT, hypoxanthine-guanine phosphoribosyl transferase

Plasmid construct and transfection

Rbp2 and Piasy fragments were inserted into pcDNA3.1(+) HA expression vector and p3× Flag CMV7.1 expression vector, respectively. HA-Rbp2 mutants (H483G/E485Q, K152E, and K152A), Pias1 C351S, Piasy C330A, Piasy C340A, Piasy C335A, Piasy W356A, Piasy C335A/W356A, and IFN regulatory factor 7 (IRF7) K406R mutants were generated by using PCR using site-directed mutagenesis primers (https://www.genomics.agilent.com/primerDesignProgram; Table 1). Each transient expression plasmid was confirmed by sequencing and validated as functional by transfecting HEK293T cells followed by immunoblotting (IB). Retroviral pMSCV puro expression vectors for HA-Rbp2, Flag, Flag-Piasy full-length DNA, and Piasy C-terminal truncated parts were constructed by using standard cloning procedures. Subsequently, these retroviral vectors were sequenced and verified in retrovirus-infected target cells by IB. Oligo sequences that targeted Rbp2 were cloned into pSuper Retro vector. Plasmids (1 μg/ml) were transfected into HEK293T cells or platinum E cells by using lipofectamine 2000. siRNA that targeted Rbp2 were transfected into Piasy^+/+ MEFs by using lipofectamine 2000. Poly(I:C) (2.5 μg/ml) was used to transfect into Piasy^+/+ MEFs, Piasy^−/− MEFs, and Piasy mutants or Rbp2 stably overexpressed MEFs using lipofectamine 2000 according to manufacturer instructions.

IB

Cells were lysed in lysis buffer (1% Triton X-100, 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA) that contained freshly added protease inhibitor cocktails. Equal amounts of protein were boiled for 10 min after concentration quantification by the BCA protein assay kit (Comwin), then separated by SDS-PAGE and transferred to PVDF membranes. Membranes were then blocked by 5% (w/v) nonfat milk, dissolved in Tris-buffered saline/Tween 20 at room temperature for 1 h and probed with anti-glyceraldehyde 3-phosphate dehydrogenase, anti-HA, anti-Flag, anti–β-tubulin, anti-Rbp2, or anti-V5 at 4°C overnight. Thereafter, membranes were incubated with corresponding horseradish peroxidase–conjugated secondary Abs for 1 h at room temperature. Immunoreactive bands were imaged by using ECL. All experiments were conducted at least in triplicate.

Immunoprecipitation assay

Immunoprecipitation (IP) was conducted by using the anti-Flag M2 affinity gel according to manufacturer instructions. Whole-cell lysates were obtained in IP lysis buffer (1% Triton X-100, 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA) with protease inhibitor cocktails. Approximately 10% of whole-cell lysates boiled at 95°C for 10 min with 5× SDS loading buffer were used as input, and the remaining protein (400–500 μg) was incubated with 40 μl anti-Flag M2 affinity gel or 60 μl anti-HA affinity matrix for IP at 4°C overnight. Immunoprecipitated proteins were washed 5 or 7 times with wash buffer (0.1% SDS, 1% Triton X-100, 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA) and eluted with 5× SDS loading buffer at 95°C for 10 min, then underwent IB with input samples.

In vivo SUMOylation assay

Flag-IRF7 and V5-SUMO3 were coexpressed with Piasy or Pias1 in HEK293T cells by lipofectamine 2000. Flag-IRF7, T7-SUMO1, or V5-SUMO3 were coexpressed in Piasy^+/+ or Piasy^−/− MEFs cells by lipofectamine 2000, with or without NDV infection, for indicated time points. Cells were lysed with 2% SDS at 24 h post-transfection or after NDV infection for indicated time points. IP was completed by using the previous protocol with the anti-Flag M2 affinity gel. IB was finished as previously described.

Luciferase reporter assay

HEK293T cells were transfected with pIFN–β-luciferase, internal control plasmid (pRL-CMV), and an indicated variety of expression plasmids or control (pcDNA3.1) plasmids by lipofectamine 2000. At 24 h post-transfection, cells were transfected with 2.5 μg/ml poly(I:C). Luciferase activity was measured at 12 h after poly(I:C) transfection by using a dual-luciferase reporter assay system according to manufacturer instructions. The fold induction of IFN-β luciferase activity normalized to internal control Renilla activity was evaluated.

Chromatin immunoprecipitation

Chromatin immunoprecipitation (ChIP) assay was carried out on the basis of the quick and quantitative ChIP (Q²-ChIP) protocol using H3K36me3, H3K27me3, and H3K4me3 Abs and anti-FLAG affinity gel. In brief, approximately 7 × 10⁵ MEF cells were seeded in one 10-cm dish and transfected with poly(I:C) (2.5 μg/ml) 24 h later. Total protein was then cross-linked for 10 min with chromosomal DNA by 1% formaldehyde at the indicated time post-transfection, followed by 0.13 M glycine treatment for 5 min. After washing with PBS twice on ice, cells were harvested with 11 ml PBS by cell scraper. Cells were then spun down and resuspended in 200 μl lysis buffer (50 mM Tris-HCl, pH 8, 10 mM EDTA, 1% SDS) that was supplemented with protease inhibitor cocktail. DNA was then fragmented into approximately 0.5 kb by using a sonicator. After centrifugation at maximum speed for 10 min, supernatants were diluted 10-fold in RIPA buffer (10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.5 mM EGTA, 0.1% SDS, 1% Triton X-100, 0.1% Na-deoxycholate, 140 mM NaCl) that contained protease inhibitor cocktail. ChIP assays were performed with an anti-H3K36me3, anti-H3K27me3, and anti-H3K4me3 Abs and anti-Flag affinity gel, with mouse IgG as negative control. Real-time qPCR analyses were conducted to calculate the bound DNA in triplicate by using SYBR Premix. Sequences of the primers are provided in Table 1. The relative amount of immunoprecipitated DNA was normalized to input DNA.

Statistical analysis

Data are shown as means ± sem. A Student’s t test was used to analyze differences between groups. Differences were considered statistically significant with values of P ≤ 0.05.

RESULTS

SUMO E3 ligase activity of Pias1 is required for the inhibition of IRF7

We previously reported that Piasy negatively regulated IFNI expression upon vesicular stomatitis virus, encephalomyocarditis virus, or Sendai virus infection (5). To investigate its detailed mechanism, we first confirmed the role of Piasy in IFNI expression. After verifying the transcript level of Piasy in Piasy^+/+ and Piasy^−/− MEFs (Supplemental Fig. 1A), we tested the mRNA expression levels of IFN-αs and IFN-β at indicated times during the first 10 h postinfection of NDV. As shown in Supplemental Fig. 1B, mRNA levels of IFN-αs and IFN-β were markedly higher in Piasy^−/− MEFs than in Piasy^+/+ MEFs at 9 or 10 h postinfection. Given that several TLRs are activated during NDV infection, it seems that a ligand that is linked to a specific TLR may be a better choice with which to study related pathways (32); therefore, we transfected poly(I:C), a TLR3 agonist, into Piasy^+/+ and Piasy^−/− MEFs. In contrast to that in Piasy^+/+ MEFs, both IFN-αs (Fig. 1A, bottom) and IFN-β (Fig. 1A, top) mRNA were relatively higher in Piasy^−/− MEFs, especially at 8 and 12 h post-transfection (Fig. 1A). Ectopic expression of Flag-tagged Piasy in Piasy^+/+ and Piasy^−/− MEFs (Supplemental Fig. 1C) down-regulated poly(I:C)-induced IFN-β production compared with those transfected with empty vector (Supplemental Fig. 1D). These data verify that Piasy is involved in the suppression of IFNI transcription.

Figure 1.

SUMO E3 ligase activity of Pias1, but not Piasy, is involved in negatively regulating IFN-β transcription. A) Piasy^+/+ and Piasy^−/− MEFs were transfected with poly(I:C). IFN-β and IFN-αs were measured by real-time qPCR at indicated time points post-transfection, and results were normalized by control gene HPRT (hypoxanthine-guanine phosphoribosyl transferase). A representative result of 3 independent experiments was shown. B) Sumoylation of IRF7 in Piasy^+/+ and Piasy^−/− MEFs. Piasy^+/+ and Piasy^−/− MEFs were coexpressed with indicated plasmids, and 24 h later, cells were lysed with lysis buffer. IP was accomplished by anti-Flag M2 gel, then analyzed by IB with indicated Abs. C) In vivo sumoylation assay of IRF7 by Piasy or Pias1. Flag-IRF7 and V5-SUMO3 were coexpressed with Piasy or Pias1 in HEK293T cells by lipofectamine 2000. Whole-cell extracts were collected by 2% SDS, then IP and IB were conducted as before. D) IP between HA-Piasy and Flag-Pias1 in HEK293T cells. HEK293T cells were cotransfected with indicated plasmids by lipofectamine 2000 for 24 h and collected by lysis buffer. IP was conducted by anti-Flag M2 gel, then analyzed by IB with the indicated Abs. E) Luciferase reporter assay of IFN-β. Indicated plasmids were transfected into HEK293T cells by Fugene 6. These cells were transfected by 2.5 μg/ml poly(I:C) 24 h later. Luciferase activities were analyzed 12 h after poly(I:C) transfection. F) IB analysis of the overexpression of Flag-tagged Piasy in Pias1^+/+ and Pias1^−/− MEFs. G) Fluorescent real-time qPCR assessment of IFN-β and IFN-α4 mRNA levels of the indicated groups after 2.5 μg/ml poly(I:C) transfection for indicated time points. H) Fluorescent real-time qPCR analysis of IFN-β and IFN-α4 mRNA levels in Pias1^+/+ and Pias1^−/− MEFs. Cells that stably overexpress Flag or Flag-Piasy were transfected with wild-type IRF7 or IRF7 K406R, with 2.5 μg/ml poly(I:C) transfection for 12 h. ns, not significant. Data are expressed as means ± sem. *P < 0.05, **P < 0.01, *** P < 0.001.

We explored whether Piasy inhibits IFNI promoter activity mediated by IRF7, which is a transcription factor (TF) that is required for positive feedback of IFNI transcription (5). Recent studies have indicated that Piasy can increase or reduce the activities of transcription factors by binding with them (16, 33). Piasy has been proposed to interact with IRF7 (34), an interaction that we verified (Supplemental Fig. 2A). As is well established, Piasy possesses SUMO E3 ligase activity, and the inhibitory activity of Piasy on IRF7 is independent of its SUMO E3 ligase activity, but dependent on sumoylation (5). This result was confirmed by our sumoylation experiments using IRF7 during NDV infection in vivo. In Piasy^+/+ MEFs, sumoylation of IRF7—mediated by either SUMO-1 (Supplemental Fig. 2B, left) or SUMO-3 (Supplemental Fig. 2B, right)—was enhanced at the indicated time points after NDV infection, especially at 12 h postinfection. In addition, in Piasy^+/+ MEFs, sumoylation of IRF7 was much more apparent than in Piasy^−/− MEFs (Fig. 1B). Another study has shown that VP-35 can negatively regulate the activity of IRF7 via Pias1-mediated sumoylation of IRF7, thereby repressing IFNI transcription (11). In our study, IRF7 could be sumoylated by both Piasy and Pias1 in HEK293T cells as shown in Fig. 1C. To determine the inhibition activity of Piasy and Pias1 on IRF7, we performed luciferase reporter assays. Whereas the SUMO E3 ligase–deficient mutants Piasy C335A, Piasy W356A, or Piasy C335A/W356A retained their inhibitory activities on IRF7, the enzymatically deficient mutant Pias1 C351S lost its inhibitory function on IRF7 (Fig. 1E) (35, 36), which suggests that the SUMO E3 ligase of Pias1, but not Piasy, was indispensable for the inhibition of IRF7. Of interest, coexpression of Piasy and Pias1 reduced the promoter response of IRF7 dramatically, more evidently than that of the overexpression of either (Fig. 1E, bottom). With regard to the effect of Pias1 or Piasy mutant on IRF7 sumoylation, we found that Pias1 C351S, Piasy C335A, Piasy W356A, and Piasy C335A/W356A lost sumoylation activity on IRF7 (Supplemental Fig. 2C), whereas other Piasy mutants (shown in Supplemental Fig. 2C) retained their sumoylation activity. In addition, it has been previously reported that Piasy can enhance or reduce transcription by interacting with Pias1 (12). As shown in Fig. 1D, Piasy could bind to Pias1 to form a complex. Moreover, stably aberrant expression of Flag-tagged Piasy in Pias1^−/− MEFs (Fig. 1F) led to no statistical change in IFNI transcription upon poly(I:C) transfection compared with control (Fig. 1G), which indicates that Pias1 may play a key role in mediating the inhibition of IFNI promoter response. To investigate the role of IRF7 sumoylation in IFNI transcription attenuation by Piasy, we overexpressed Flag-Piasy or Flag in Pias1^+/+ or Pias1^−/− MEFs (Fig. 1F). Thereafter, we transfected these cells with vectors that encoded wild-type IRF7 or IRF7 K406R, followed by poly(I:C) transfection. IRF7 K406 was identified as 1 SUMO conjugation site and IRF7 K406R was an IRF7 mutant that was defective in sumoylation. IFN-β and IFN-α4 promoter responses were then detected with real-time qPCR. In Pias1^−/− MEFs, there were no significant difference among the 4 indicated groups (Fig. 1H, top); however, in Pias1^+/+ MEFs, in contrast to that in Flag-overexpressed cells, IFN-β and IFN-α4 mRNA levels were lower in the Flag-tagged Piasy overexpression group (Fig. 1H, bottom). Moreover, wild-type IRF7 led to less IFN-β and IFN-α4 mRNA induction than did IRF7 K406R (Fig. 1H, bottom), which indicates that IRF7 sumoylation plays a critical role in IFNI transcription inhibition. Taken together, we propose that Piasy attenuates IFNI transcription via Pias1-mediated sumoylation of IRF7.

Piasy reduces H3K4me3 levels of IFNI genes upon activation in MEFs

Just as widely recognized, transcription can mainly be regulated by modulating TF or changing the structure or function of chromatin (37, 38). From the point of TF, we concluded that Piasy suppressed IFNI transcription via Pias1-mediated sumoylation of IRF7. With respect to chromatin, histone methylation is the most stable among all histone modifications. Piasy was reported to regulate acetylation, thereby promoting autophagy (39). To determine whether Piasy can influence the structure or function of chromatin, such as altering the state of histone modifications, thereby affecting transcription of IFNI, we conducted ChIP assays to test H3K36me3, H3K27me3, or H3K4me3 levels. Whereas H3K36me3 and H3K4me3 are mainly associated with activated transcription, H3K27me3 is linked to repressed transcription. Our results demonstrated that H3K36me3 (Fig. 2A) and H3K27me3 (Fig. 2B) levels presented no increased or different signal around the transcription start site (TSS) of the IFN-β gene at indicated times post-transfection of poly(I:C) between Piasy^+/+ and Piasy^−/− MEFs. Of note, we observed that at 6 h post-transfection of poly(I:C), the H3K4me3 level of approximately 588 bp downstream of TSS of the IFN-β gene in Piasy^−/− MEFs was markedly increased and was much higher (∼3-fold) than that in Piasy^+/+ MEFs (Fig. 2C). H3K4me3 is a kind of histone modification that promotes the transcription of targeted genes and exists solely in the promoter zone (40); therefore, the H3K4me3 level is indicative of the mRNA expression level. This result coincided well with the data shown in Fig. 1A. To further confirm the role of Piasy in influencing the H3K4me3 level, we reintroduced Piasy into Piasy^−/− MEFs (Supplemental Fig. 1C). We subsequently carried out ChIP experiments to detect the H3K4me3 level for groups with or without poly(I:C) transfection at the indicated time points. As shown in Fig. 2D, to some extent, ectopic expression of Piasy in Piasy^−/− MEFs reversed the H3K4me3 level at 177–588 bp downstream of TSS of the IFN-β gene, which was in contrast to that in Piasy^−/− MEFs, especially at 12 h after poly(I:C) transfection. Moreover, the ChIP assay in the Flag-tagged Piasy overexpression Piasy^−/− MEFs using anti-Flag affinity gel indicated that Piasy bound to the promoter zone and increased at 500 bp upstream of TSS of the IFN-β gene at 14 h after poly(I:C) transfection (Fig. 2E). Taken together, these results indicate that Piasy can decrease the H3K4me3 level of the IFN-β gene upon poly(I:C) transfection.

Figure 2.

Piasy can demethylate the H3K4me3 of IFNI upon activation. A) H3K36me3 level in different zones near the TSS of IFN-β was analyzed by ChIP assay in Piasy^+/+ and Piasy^−/− MEFs at the indicated time points after 2.5 μg/ml poly(I:C) transfection. Mouse IgG signal is shown as control. B) H3K27me3 level in different zones near the TSS of IFN-β was analyzed by ChIP assay in Piasy^+/+ and Piasy^−/− MEFs at the indicated time points after 2.5 μg/ml poly(I:C) transfection. Mouse IgG control signal ias also shown. C) H3K4me3 level in different zones near the TSS of the IFN-β gene was analyzed by ChIP assay in Piasy^+/+ and Piasy^−/− MEFs at the indicated time points after 2.5 μg/ml poly(I:C) transfection. Mouse IgG signal is shown as control. D) H3K4me3 level in the indicated parts around the TSS of IFN-β was assessed by ChIP assay in Piasy^−/− MEFs and Piasy overexpressed Piasy^−/− MEFs at the indicated time points after 2.5 μg/ml poly(I:C) stimulation. Mouse IgG control signal is also presented. E) Flag-Piasy level in different zones near the TSS of IFN-β was analyzed by ChIP assay in Flag-Piasy overexpressed Piasy^−/− MEFs at the indicated time points after 2.5 μg/ml poly(I:C) transfection. Mouse IgG signal is shown as control. Data are expressed as means ± sem. *P < 0.05, ***P < 0.001.

Rbp2 directly binds to Piasy and inhibits IFNI promoter responses

Piasy, although not itself a demethylase, decreased the H3K4me3 level of the IFN-β gene upon poly(I:C) transfection. This caused us to speculate that Piasy may interact with one demethylase, which can demethylate H3K4me3. To verify this function, we performed co-IP between Piasy and the hitherto reported H3K4me3 demethylases, Lid, Rbp2, and Jarid1b/c (data not shown) (25, 41, 42). Among them, we noticed that Piasy could uniquely combine with Rbp2 (Fig. 3A). To determine whether the inhibitory effect of Piasy on IFNI was a result of the interaction between Piasy and Rbp2, a Rbp2-specific siRNA was used to effectively diminish the mRNA (Fig. 3B) and protein (Fig. 3C) expression of Rbp2 in Piasy^+/+ MEFs. Consequently, IFN-β and IFN-α4 mRNA levels were higher at 12 h after poly(I:C) transfection in cells that were transfected with Rbp2 siRNA compared with that of control cells that were transfected with scrambled siRNA (Fig. 3D). This finding was verified in NIH3T3 cells. Rbp2-targeted short hairpin RNA was introduced into NIH3T3 (Fig. 3E, F). Thereafter, control NIH3T3 cells and Rbp2 knockdown NIH3T3 cells were transfected with poly(I:C). At 24 h post-transfection of poly(I:C), IFN-β and IFN-α4 mRNA levels in Rbp2 knockdown NIH3T3 cells were much higher than those in control cells (Fig. 3G). Simultaneously, we introduced HA-tagged Rbp2 into Piasy^+/+ MEFs (Fig. 3H), then treated them with poly(I:C) and analyzed IFN-β and IFN-α4 mRNA expression levels. Of note, IFN-β and IFN-α4 mRNA levels were approximately 50% lower at 12 h after poly(I:C) transfection in cells aberrant expression of HA-Rbp2 than in control cells (Fig. 3I). In short, these data of Rbp2 loss- and gain-of-function experiments demonstrate that Piasy likely attenuates IFNI transcription via Rbp2. Furthermore, luciferase reporter assays were conducted in HEK293T cells to investigate whether Piasy really attenuates IFNI transcription via Rbp2. As shown in Fig. 3J, Rbp2 repressed the IFN-β promoter activity. Above all, Piasy and Rbp2 together inhibited the IFN-β promoter more than either Piasy or Rbp2 alone. These data suggest that Piasy and Rbp2 repress IFN-β transcription by inactivating the IFN-β promoter in a synergistic manner.

Figure 3.

Piasy binding to Rbp2 is essential for the inhibition of IFNI response. A) IP between Flag-Piasy and HA-Rbp2 in HEK293T cells. HEK293T cells were cotransfected with indicated plasmids by lipofectamine 2000 for 24 h and collected by lysis buffer. IP was completed by anti-Flag M2 gel (upper panel) or anti-HA affinity matrix (lower panel), then analyzed by IB with indicated Abs. B) Fluorescent real-time qPCR analysis of relative Rbp2 mRNA level in Rbp2 knockdown Piasy^+/+ MEFs. C) IB analysis of Rbp2 protein level in Rbp2 knockdown Piasy^+/+ MEFs and control cells. D) Transcript levels of IFN-β and IFN-α4 in Piasy^+/+ MEFs and Rbp2-deficient Piasy^+/+ MEFs with or without 2.5 μg/ml poly(I:C) transfection for indicated time points were determined by real-time qPCR. Data are normalized to HPRT (hypoxanthine-guanine phosphoribosyl transferase). E) Real-time qPCR analysis of Rbp2 mRNA expression level in Rbp2-deficient NIH3T3 cells. F) IB analysis of Rbp2 protein level in Rbp2 knockdown NIH3T3 and control cells. G) Transcript levels of IFN-β and IFN-α4 in NIH3T3 and Rbp2-deficient NIH3T3 cells with or without 2.5 μg/ml poly(I:C) transfection for indicated time points were determined by real-time qPCR. H) IB analysis of HA-Rbp2 stable overexpression in Piasy^+/+ MEFs. I) Fluorescent real-time qPCR analysis of IFN-β and IFN-α4 mRNA expression levels after 2.5 μg/ml poly(I:C) transfection for indicated time points. J) Fold-change of relative luciferase activities of IFN-β. Indicated plasmids were transfected into HEK293T cells by lipofectamine 2000. These cells were transfected with 2.5 μg/ml poly(I:C) 24 h later. Luciferase activities were analyzed 12 h after poly(I:C) transfection. Data are shown as means ± sem. *P < 0.05, **P < 0.01, *** P < 0.001.

The Jmjc domain of Rbp2 and the PINIT domain of Piasy are prerequisites

To identify the domains that are responsible for the interaction between Rbp2 and Piasy, we constructed a series of C-terminal deletion mutants (Fig. 4A, D). We then conducted co-IP studies using these mutants. As shown in Fig 4B, full-length Piasy could precipitate Rbp2 1–735 aa and Rbp2 1–635 aa, but not other mutants (Fig 4B, left), which suggests that Piasy may interact with Rbp2 451–635 aa; therefore, we also performed a co-IP assay between Piasy and Rbp2 Δ451–635 aa. As expected, Piasy could not pull down Rbp2 Δ451–635 aa (Fig. 4B, right). To narrow down the interaction domain between Rbp2 and Piasy, we carried out more precise co-IP assays and found that Piasy could precipitate other Rbp2 mutants than Rbp2 1–450 aa (Fig. 4C), which indicates that the region around amino acid 450 of Rbp2 is essential for their interaction. Conversely, co-IP between Piasy N-terminal parts (Fig. 4D) and Rbp2 demonstrated that Piasy mutants, excluding Piasy 1–100 aa, coprecipitated Rbp2 (Fig. 4E, left). Furthermore, prolonged co-IP assay between Piasy 101 and 218 aa and Rbp2 demonstrated that Piasy 101–218 aa could precipitate Rbp2 (Fig. 4E, right). These data imply that 451–503 aa (Jmjc domain 470–586 aa) of Rbp2 and 101–218 aa (PINIT domain 125–270 aa) of Piasy is required for their interaction. To address the functionality of their interaction, HEK293T cells that had been transfected with Piasy or Rbp2 fragment constructs, or their indicated combinations, were stimulated with poly(I:C). Whereas Rbp2 Δ451–635 aa lost its effect on IFN-β and IFN-αs mRNA expression, Rbp2 451–635 aa and full-length Rbp2 attenuated their transcription significantly (approximately 45% of control cells; Fig. 5A). Moreover, when coexpressed with Piasy, the Rbp2 451–635 aa and full-length Rbp2 groups impaired IFNI transcription more obviously than did the Rbp2 Δ451–635 aa group (Fig. 5A), which suggests that Rbp2 451–635 aa is required for its inhibitory function. To narrow down its functional region, HEK293T cells were transfected with other mutants of Rbp2 (Fig. 5B), followed by poly(I:C) treatment. Rbp2 1–503 aa kept its function in IFN-β and IFN-αs, but Rbp2 1–450 aa lost this ability (Fig. 5C). Analogously, Piasy Δ101–218 aa was not capable of repressing IFN-β and IFN-αs expression (Fig. 5D), whereas Piasy 101–218 aa and full-length Piasy suppressed IFN-β and IFN-αs transcription remarkably (Fig. 5E). Of special note, among groups that were transfected with Piasy fragments and Rbp2, the Piasy 101–218 aa and Rbp2 coexpressed groups had the same inhibitory activity as the group that overexpressed Rbp2, with other coexpressed groups exerting more obvious inhibitory function (Fig. 5E). Hence, these data indicate that the binding of Rbp2 451–503 aa to Piasy 101–218 aa is a prerequisite for the negative modulation of IFNI responses.

Figure 4.

Screening for interactional domains between Rbp2 and Piasy. A) Schematic illustration of full-length and truncated mutants of HA-Rbp2 used in the binding domain test and functional domain assay. B, C) IP analysis of HA-Rbp2 truncated mutants with Flag-Piasy. HEK293T cells were coexpressed with indicated plasmids by lipofectamine 2000, then interaction was analyzed by IB with indicated Abs. D) Schematic diagram of full-length and deleted mutants of Flag-Piasy used in the interaction domain screening and functional domain analyses. E) IP analysis of Flag-Piasy truncated mutants with HA-Rbp2. HEK293T cells were coexpressed with indicated plasmids, then interaction was analyzed by IB with indicated Abs.

Figure 5.

Functional domain screening of Rbp2 and Piasy. A) Fluorescent real-time qPCR analysis of IFN-β and IFN-αs mRNA expression levels in HEK293T cells. Cells were transfected with indicated plasmids, with or without poly(I:C) transfection. B) IB analysis of HA-Rbp2 and its mutant expression in HEK293T cells with indicated Abs. C) Fluorescent real-time qPCR analysis of IFN-β and IFN-αs mRNA expression levels in HEK293T cells that were transfected with HA-Rbp2 fragments, with or without poly(I:C) transfection. D) IB analysis of the indicated protein expression in HEK293T cells with referential Abs. E) Fluorescent real-time qPCR analysis of IFN-β and IFN-αs mRNA expression levels in HEK293T cells that were transfected with indicated plasmids, with or without poly(I:C) transfection. ns, not significant. Error bars indicate means ± sem. *P < 0.05, **P < 0.01.

These findings, in part, were verified in Piasy^+/+ and Piasy^−/− MEFs. First, we introduced Flag-tagged Piasy or its truncated mutants into Piasy^+/+ and Piasy^−/− MEFs (Fig. 6A). Thereafter, we transfected these cells with poly(I:C) and observed that overexpression of full-length Piasy or its mutants, excluding Piasy 1–100 aa, down-regulated the level of IFN-β and IFN-α4 transcript to a certain degree (40–75% of control cells; Fig. 6B). Moreover, after introducing Piasy Δ101–218 aa into Piasy^−/− MEFs (Supplemental Fig. 3A), we tested its function on IFNI transcription. As shown in Supplemental Fig. 3B, compared with control, Piasy Δ101–218 aa was unable to exert its inhibitory effect. Taken together, Piasy 101–218 aa plays an essential role in reducing IFNI signaling.

Figure 6.

The PINIT domain of Piasy is necessary for repressing IFNI promoter response and decreasing H3K4me3 level in the IFN-β gene. A) IB analysis of Piasy or its truncated mutants that are stably ectopically expressed in Piasy^+/+ and Piasy^−/− MEFs. Right arrow indicates a nonspecific band. B) Transcript levels of IFN-β and IFN-α4 in cells with 2.5 μg/ml poly(I:C) transfection for 12 h were determined by real-time qPCR. Data are normalized to HPRT (hypoxanthine-guanine phosphoribosyl transferase). C) H3K4me3 levels in the indicated parts near the TSS of IFN-β were analyzed by ChIP assay in Piasy^−/− MEFs and Piasy truncates ectopically expressed Piasy^−/− MEFs at the indicated time points after 2.5 μg/ml poly(I:C) transfection. Mouse IgG signal is shown as control. NC, negative control; ns, not significant. Error bars indicate means ± sem. *P < 0.05, **P < 0.01, *** P < 0.001.

To evaluate whether gain of Piasy mutants can affect the H3K4me3 level of the IFN-β gene, we performed ChIP assays that measured the H3K4me3 level of the IFN-β gene of Piasy or its mutants overexpressed Piasy^−/− MEF cells and control cells, with or without poly(I:C) transfection. Consistent with real-time qPCR results for IFN-β and IFN-α4 mRNA, overexpression of Piasy mutants, excluding Piasy 1–100 aa, reduced the H3K4me3 level of the IFN-β gene to some extent (approximately 50% of control cells; Fig. 6C). Indeed, in parallel with Piasy and other Piasy mutants, Piasy 1–100 aa had no effect on decreasing IFNI transcription and the H3K4me3 level of the IFN-β gene (Fig. 6B, C), which implies that Piasy 101–218 aa is critical for its inhibitory activity.

Demethylase activity of Rbp2, but not DNA contact by K152, is indispensable

Our data show that Rbp2 plays a key role in attenuating IFNI transcription. Considering that Rbp2 modulates gene expression via DNA contact (28) and is a demethylase that adjusts IFNI expression via demethylating H3K4me3, we evaluated whether these two activities contribute to the regulatory function of Rbp2. For this purpose, we initially overexpressed HA-tagged wild-type Rbp2 or demethylase inactive Rbp2^H483G/E485Q (25) in HEK293T cells (Fig. 7A). As shown in Fig. 7B, Rbp2 ^H483G/E485Q, similar to the empty vector, was unable to impede IFN-β and IFN-αs expression compared with wild-type Rbp2. These results demonstrate that the enzymatic activity of Rbp2 is necessary for its action as a negative regulator. Meanwhile, DNA contact–defective mutant Rbp2^K152E (28) was also introduced into HEK293T cells (Fig. 7C), followed by poly(I:C) transfection. Of interest, compared with wild-type Rbp2 (65% of control cells), Rbp2^K152E (37% of control cells) exerts even a little more inhibition on IFN-β and IFN-αs expression at 12 h post-transfection (Fig. 7D). This caused us to think about potential causes. Rbp2 repressed IFNI transcription via demethylating H3K4me3, and both histone and lysine (K) carry positive charges, whereas glutamate (E) carries a negative charge. It is possible that the difference in charge carried led to more Rbp2 K152E binding to the IFN-β promoter than that of Rbp2 and IFNI mRNA production. Thus, we constructed Rbp2 K152A—alanine carries no charge—and performed the above assay with it. After a plasmid that expressed HA epitope-tagged Rbp2 or Rbp2 K152A was transfected into HEK293T cells for 24 h (Fig. 7E), cells were stimulated with poly(I:C). Results of real-time qPCR demonstrated that, just as we speculated, Rbp2 K152A had similar effects on the suppression of IFN-β and IFN-αs (Fig. 7F), which suggests that DNA binding activity affected the inhibitory function of Rbp2 to some degree. It has been reported that Rbp2 could specifically bind to the DNA sequence CCGCCC via K152; however, upon DNA sequence blasting, we could not find a conserved sequence CCGCCC in the IFN-β gene. More remarkable, Rbp2 S156, L157, and R112 also contribute to sequence-dependent DNA binding (28). In other words, in addition to CCGCCC, Rbp2 might recognize other DNA sequences. In brief, in our system, the demethylase activity of Rbp2, rather than DNA contact by Rbp2 K152, is indispensable for its modulatory function.

Figure 7.

Demethylase activity of Rbp2, but not DNA contact by Rbp2 K152, is indispensable for its inhibitory effect. A) IB analysis of empty vector (EV), wild-type Rbp2, or its enzymatically inactive mutant H483G/E485Q aberrant expression levels in HEK293T cells. B) Transcript levels of IFN-β and IFN-αs in cells with or without 2.5 μg/ml poly(I:C) transfection for indicated time points were determined by real-time qPCR. Data are normalized to HPRT (hypoxanthine-guanine phosphoribosyl transferase). C) IB analysis of EV, wild-type Rbp2, or Rbp2^K152E overexpression levels in HEK293T cells. D) Transcript levels of IFN-β and IFN-αs in cells with 2.5 μg/ml poly(I:C) transfection for 12 h were determined by real-time qPCR. Data are normalized to HPRT. E) IB analysis of EV, wild-type Rbp2, or Rbp2^K152A transient ectopic expression levels in HEK293T cells. F) Transcript levels of IFN-β and IFN-αs in cells with 2.5 μg/ml poly(I:C) transfection for 12 h were determined by real-time qPCR. Data are normalized to HPRT. G) Schematic illustration of IFNI transcription suppression by Piasy. A summary diagram of Piasy-regulated IFNI transcription via Rbp2. Piasy interacts with Rbp2 and recruits it to the promoter zone of IFNI genes, which demethylates the H3K4me3 of the IFNI gene via the demethylase activity of Rbp2. As the H3K4me3 level of the IFNI gene represents the transcription level of IFNI, IFNI transcriptions are inhibited. Piasy also interacts with IRF7 and Pias1. Furthermore, Piasy can diminish IFNI transcription by synergy with Pias1. Thus, it should be noted that Piasy has an additional mechanism to attenuate IFNI transcription, and Piasy may involve Pias1 to increase IRF7 sumoylation by interacting with Pias1. Ns, not significant. Error bars indicate means ± sem. *P < 0.05, **P < 0.01.

DISCUSSION

In this study, we explored the mechanisms by which Piasy attenuates IFNI transcription. As recent studies have indicated that transcription is mainly regulated by modulating TF or changing the structure or function of chromatin (37, 38), we conducted our study on the basis of these 2 aspects. From the point of TF regulation, we focused on IRF7, a crucial TF that participated in the positive feedback of IFNI production (5). Our data demonstrate that Piasy interacts with Pias1, and that Piasy diminishes IFNI transcription via Pias1-mediated sumoylation of IRF7, which is similar to the published result that VP-35 reduced IFNI transcription via Pias1-mediated sumoylation of IRF7 (11). It should be noted that we only mutated the SUMO conjugation site, K406, and our data suggest that it is an important SUMO conjugation site that is involved in the inhibition of IFN-β, but that it is not a main site in the attenuation of IFN-α4.

Histone modification is one of the factors that can alter the structure or function of chromatin. Commonly, methylation, acetylation, phosphorylation, ubiquitination, and sumoylation of histone are mediated by histone modification enzymes (43). Recent findings have demonstrated that abundant TFs modulate transcription by remodeling the chromatin structure of targeted genes by recruiting histone methylase or deacetylase to targeted promoters (44, 45). With the discovery of various histone demethylases, studies of transcription regulation mediated by histone methylation represent the cutting edge. Rbp2—as a newly identified histone demethylase that can specifically demethylate H3K4me2 and H3K4me3—is expressed in humans and mice (25). Meanwhile, Rbp2 has been observed to play a key role in cell proliferation, differentiation, development, and senescence by regulating the transcription of certain genes (46^,–48).

In addition, the role of the methylation of H3K4 in modulating transcription has been confirmed by many studies (44, 49). As reported by a genomic experiment, the mono- and dimethylation of H3K4 exist in regions other than the promoter, whereas the trimethylation of H3K4 was found exclusively in the promoter zone (40); therefore, the modification level of H3K4me3 represents transcription activity to a certain degree. Our current study demonstrated that, among the 3 kinds of involved histone modifications, only the H3K4me3 level—especially at 6 h post-transfection of poly(I:C) —in approximately 588 bp downstream of the TSS of the IFN-β gene in Piasy^−/− MEFs is much higher than in Piasy^+/+ MEFs, which is consistent with the tendency of the IFN-β mRNA level with the same treatment. Furthermore, the ectopic expression of Piasy in Piasy^−/− MEFs could regain the inhibitory effect of Piasy in IFNI transcription, accompanied by a lower H3K4me3 level in approximately 588 bp downstream of the TSS of the IFN-β gene. Accordingly, data on both the loss and gain of Piasy strongly suggest that Piasy down-regulated the H3K4me3 level of the IFN-β gene. It should be noted that Piasy may influence other important events, including histone acetylation, nucleosome positioning, RNA Pol II status, etc. Whether Piasy can inhibit IFNI transcription by affecting these events should be investigated.

Using the co-IP assay to screen for H3K4me3 demethylase that can interact with Piasy, we observed that only Rbp2 could be pulled down by Piasy, and vice versa. Moreover, functional analyses indicated that Rbp2 knockdown led to more IFNI production in Piasy^+/+ MEFs and NIH3T3 cells, whereas the ectopic expression of Rbp2 in Piasy^+/+ MEFs resulted in less generation of IFNI. In addition, a luciferase reporter assay indicated that Piasy and Rbp2 attenuated the IFN-β promoter response in a synergistic manner, and Piasy bound to the promoter zone of the IFN-β gene and increased at 500 bp upstream of TSS of the IFN-β gene at 14 h after poly(I:C) transfection. These results support the essential role of Rbp2 in the Piasy-mediated repression of IFNI, but it should be noted that as a result of Abs, endogenous levels of Piasy and Rbp2, or lysis buffer, etc., we failed to induce the endogenous interaction of Piasy and Rbp2.

It has been reported that the structure of a protein is associated with its function (50). Of note, Piasy binds to Rbp2 via its 101–218 aa (Fig. 4E), and its 101–218 aa is required for Piasy to suppress IFNI transcription in HEK293T cells (Fig. 5E) and MEFs (Supplemental Fig. 3). We have previously reported that the LXXLL (20–24 aa) motif of Piasy is not implicated in the transcription regulation of IFNI (5). Data from the present study indicate that Piasy 1–100 aa has no inhibitory effect on IFNI transcription and that Piasy 101–218 aa itself is sufficient to reduce IFNI promoter responses, which correlates well with our previous findings. In addition, Piasy 101–218 aa overlaps the PINIT (125–270 aa) domain, which makes Piasy locate in the nucleus, where transcriptions take place. Of importance, the result that the overexpression of other Piasy mutants in Piasy^−/− MEFs, but not Piasy 1–100 aa, can demethylate the H3K4me3 of IFN-β is in agreement with the functional domain of Piasy. H3K4me3 modification is also associated with inactive genes, but we failed to find such genes in this study.

In our research, 451–635 aa of Rbp2 is indispensable for its inhibitory activity, yet it should be noted that the transfection of Piasy, combined with Rbp2 or Rbp2 451–635 aa, exerts only little more inhibitory effect on IFN-αs transcription than that of Piasy. This may the result of the fact that IFN-αs is composed of different subtypes. For some kinds of IFN-α, Piasy may have no inhibitory effect or even an enhanced effect. Our study indicated that some amino acids around 451–503 aa were necessary. Given that the demethylase activity of Rbp2 plays an important role in demethylating H3K4me3, we mutated well-established Rbp2 H483 and Rbp2 E485 2 enzymatic activity sites (25, 44). Overexpressing these enzymatically inactive mutant Rbp2^H483G/E485Q in HEK293T cells hardly attenuated IFNI transcription compared with wild-type Rbp2 (Fig. 7B). Thus, our study provides convincing evidence that the demethylase activity of Rbp2 is involved in the transcription regulation of IFNI.

Rbp2 can specifically bind to the DNA sequence, 5′-CCGCCC-3′, via K152 in its ARID (81–170 aa) domain, thereby regulating transcription (28); however, upon analyzing the DNA sequence of the IFN-β gene, we could not find this conserved sequence. In addition, the introduction of DNA contact–defective mutant Rbp2^K152E into HEK293T cells led to less IFNI production than that of wild-type Rbp2 (Fig. 7D); however, Rbp2^K152A negatively regulated IFNI transcription to the same degree as WT Rbp2 (Fig.7F). Judging from the difference in the charge that aa 152 carries, we found that this may result in different amounts of Rbp2 binding to IFNI genes, which, in turn, brings about diverse impacts on IFNI induction. Another study reported that Rbp2 could recognize DNA sequence other than 5′-CCGCCC-3′, which supports our hypothesis that Rbp2 binds to other DNA sequences, but not 5′-CCGCCC-3′ in our model (51). In contrast, we found that overexpression of Rbp2 in Piasy^−/− MEFs could repress the induction of IFNI by Poly(I:C) on a certain level (data not shown), which indicates that Rbp2 might play a multilevel role in regulating IFNI transcription—Piasy is only one of its partners.

In summary, we identified the H3K4me3 demethylase, Rbp2, as a novel member of the Piasy inhibitory complex for IFNI transcription. We observed that Piasy recruited Rbp2 to the promoter of targeted IFNI genes to remove the activation mark, H3K4me3, thereby impairing transcription (Fig. 7G). Thus, this Rbp2-mediated IFNI transcription control is an effective way to prevent excessive host innate immunity elicited by IFNI. Improved understanding of the Rbp2-mediated regulation of IFNI transcription may facilitate the development of therapies that are aimed at decreasing autoimmune diseases, cytotoxicity, or inflammation induced by exorbitant IFNI.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

ChIP

chromatin immunoprecipitation

IFNI

type I IFN

IB

immunoblotting

IP

immunoprecipitation

IRF

IFN regulatory factor

MEF

mouse embryonic fibroblast

NDV

Newcastle disease virus

Pias

protein inhibitor of activated signal transducer and activator of transcription

qPCR

quantitative PCR

Rbp2

retinoblastoma binding protein 2

siRNA

small interfering RNA

SUMO

small ubiquitin-like modifier

TF

transcription factor

TSS

transcription start site

AUTHOR CONTRIBUTIONS

S. Xu designed the experiments; K. Ozato provided plasmids and cell lines; X. Yu, H. Chen, C. Zuo, X. Jin, and S. Xu performed the research; X. Yu, C. Zuo, Y. Yin, H. Wang, M. Jin, K. Ozato, and S. Xu analyzed the data; X. Yu and S. Xu wrote the manuscript; and all authors read and approved the manuscript.