IFR-9/STAT2 Functional Interaction Drives Retinoic Acid–Induced Gene G Expression Independently of STAT1

The retinoic acid–induced gene G (RIG-G), also known as IFN-inducible gene 60, was originally identified in all-trans retinoic acid (ATRA)–treated NB4 acute promyelocytic leukemia (APL) cells (1) and plays an important role in cell growth inhibition through up-regulation of p21 and p27 (2). RIG-G gene expression can be induced not only by ATRA along with the differentiation of NB4 cells but also by IFNα in a series of hematopoietic cell lines as well as various types of solid carcinoma cells (1, 3), which pointed to a possible role of RIG-G in signal cross-talk between ATRA and IFNα. However, at present, little is known about the regulation control of RIG-G gene expression. We previously reported two well-conserved IFN-stimulated response elements (ISRE) in the promoter region of RIG-G gene, these two ISREs being likely involved in RIG-G transcription (2). Besides, some other potential cis-acting elements for transcription factors [such as CAAT/enhancer binding protein (C/EBP) and PU.1] have been also noted in the RIG-G gene promoter. Therefore, the question of what are the transcription factors essential for RIG-G gene induction remains open.

The first identified multimeric complex to recognize the ISRE consensus sequence is the IFN-stimulated gene (ISG) factor 3 (ISGF3), which is composed of the phosphorylated forms of signal transducer and activator of transcription (STAT) 1 and STAT2, together with IFN regulatory factor (IRF)-9 (4). The IRF-9 protein is a member of the IRF family characterized by a well-conserved helix-turn-helix DNA-binding motif, which can recognize the ISRE. The mammalian IRF family contains nine members to date, IRF-1 to IRF-9, which have diverse roles in host defense, cell cycle regulation, apoptosis, oncogenesis, and immune cell development (5). STAT1 and STAT2 represent two of the seven mammalian STAT proteins that share several structurally and functionally conserved domains, including an NH₂-terminal coiled-coil region for interactions with other transcription factors and chaperone proteins, a phosphotyrosine-binding Src homology 2 domain required for receptor binding and STAT dimerization, and a COOH-terminal transcription activation domain (TAD) within which a highly conserved tyrosine residue is present. Importantly, the phosphorylation of this tyrosine allows a functional interaction between activated STAT proteins (6). The mature ISGF3 complex serves as a transcriptional regulator in type I IFN signaling pathways by translocating into the nucleus and then binding to ISRE of certain ISGs. Noteworthy, it is generally admitted that within the ISGF3 complex, STAT1 and IRF-9 mediate the DNA binding, whereas STAT2 does not seem to interact directly with DNA but provides a potent TAD (7).

In addition to the ISGF3, other types of STAT complexes with IRF-9 can also be formed to bind the ISRE. It was reported that a complex containing IRF-9 and the phosphorylated STAT1, but lacking STAT2, could be directed to ISRE-containing genes in response to IFNγ (8, 9). In mature B cells, Gupta and colleagues (10) found that IFNα was able to activate the formation of an ISGF3-like complex comprising STAT2, STAT6, and IRF-9. Whereas IRF-9 functions as an adaptor for tethering the different STAT dimers to ISRE, other members of IRF family have also been shown to be involved in the regulation of ISRE-containing genes. It was shown that IRF-1 could regulate expression of ISGs by directly binding to ISRE on its own or by dimerizing with other proteins, including IRF-3, IRF-7, IRF-8, or STAT1 (11–13). Recently, accumulating observations support the notion that STAT proteins can drive gene expression without tyrosine phosphorylation (14–19). It strongly suggests that some not yet identified mechanisms could exist in gene transcription regulation in response to these cytokines.

In this work, we intensively explored the mechanisms underlying the expression regulation of RIG-G gene and found that STAT2, together with IRF-9, was sufficient for RIG-G expression, even without the phosphorylation of STAT2, and provided a novel mechanism modulating the expression of ISGs that accounts for the signal cross-talk between IFNα and ATRA.

Cell culture and reagents. The NB4 cells (gift from Dr. Lanotte, INSERM U685, Paris, France) were cultured in RPMI 1640 (Life Technologies) supplemented with 10% fetal bovine serum, 2 mmol/L l-glutamine, and antibiotics. The U3A, U2A, and HT1080 cells were kindly provided by G.R. Stark (Lerner Research Institute, Cleveland, OH; refs. 20, 21) and cultured in DMEM supplemented with 10% calf serum, 2 mmol/L l-glutamine, and antibiotics. ATRA and cycloheximide (Sigma) were dissolved in ethanol as stock solution at 10⁻² mol/L and 10 mg/mL, respectively, and stored at −20°C. Human IFNα was purchased from Schering-Plough Co. and stored at 4°C.

Plasmids. The expression constructs for wild-type (wt) STAT1 and mutant STAT1-Y701F (converting tyrosine 701 to phenylalanine) were kindly provided by Dr. J.E. Darnell, Jr. (Rockefeller University, New York, NY). STAT2 full-length cDNA expression plasmid was obtained by subcloning the insert of pBlueScript-STAT2 (from Dr. J.E. Darnell, Jr.) to the vector pcDNA3.1. The mutant STAT2-Y690F (converting tyrosine 690 to phenylalanine) was prepared by Mutagenesis kit (Stratagene). The luciferase-containing reporter plasmids pXP2 (−310) and pXP2 (−87) were constructed as previously described (2). The mutant reporter constructs pXP2-mut-1, pXP2-mut-2, and pXP2-mut-3 were generated from pXP2 (−310) by Mutagenesis kit. All the constructs were verified by DNA sequencing.

Reverse transcription-PCR. Total RNA was extracted with Trizol reagent (Invitrogen). After reverse transcription, cDNA was amplified under the following conditions: 94°C for 5 min; 25 cycles of 94°C for 30 s, 58°C for 45 s, and 70°C for 50 s; and 70°C for 10 min. The primers 5′-ACCTCGAGACACAGAGGGCAGTC-3′ and 5′-ACGGATCCGCCTTGTAGCAGCACCCAATC-3′ were used to detect RIG-G transcripts. The amplification of β-actin gene was used as an internal control for cDNA loading with the primers 5′-CATCCTCACCCTGAAGTACCCC-3′ and 5′-AGCCTGGATGCAACGTACATG-3′.

Western blot analysis. The performance of Western blot and the generation of rabbit polyclonal anti-RIG-G sera were described in our previous work (2). Other primary antibodies were from Santa Cruz Biotechnology (STAT1, STAT2, IRF-9, and IRF-1), Sigma (β-actin), and Cell Signaling (pTyr⁷⁰¹-STAT1 and pTyr⁶⁹⁰-STAT2).

Transfection and luciferase reporter assay. The cells were cotransfected with the reporter gene constructs and related expression plasmids as indicated by using SuperFect (Qiagen) followed by transcriptional activity assay using Dual-Luciferase Assay System (Promega) according to the manufacturer's instruction. Ten nanograms pRL-TK vector (Promega) were used as an internal control for normalizing the transfection efficiency. Data were shown as means ± SD of three independent experiments.

RNA interference. To make small interfering RNA (siRNA) against IRF-1, 19-bp sequence (5′-CTTCCAGGTGTCACCCATG-3′) within the coding region of IRF-1 gene was selected and cloned into the pTER⁺ vector (22). The plasmid pTER-si-IRF-1 was then transfected into U3A cells by using SuperFect.

Coimmunoprecipitation. The expression plasmid encoding green fluorescent protein (GFP)-IRF-9 hybrid protein was transfected with the indicated STAT2 constructs into U3A cells. Twelve hours after transfection, cells were treated with or without IFNα for another 24 h. The protein extracts were prepared in the buffer containing 150 mmol/L NaCl, 50 mmol/L Tris-HCl (pH 8.0), and 0.5% NP40 and then mixed with protein A-agarose and rabbit anti–IRF-9 antibody (Santa Cruz Biotechnology) at 4°C overnight with rotation. The precipitated proteins were eluted by boiling beads in SDS-loading buffer and analyzed by Western blotting.

Chromatin immunoprecipitation assay. Chromatin immunoprecipitation (ChIP) assay was performed using EZ ChIP kit (Millipore) according to the instruction of the manufacturer. The primers specific to the promoter region of RIG-G gene were as follows: 5′-TGGTGAGTATGGCCTTAGAATT-3′ and 5′-TTCCTGTCTGCCTCAAGTAAAT-3′. Normal rabbit IgG (Millipore) was used to control the nonspecific immunoprecipitation of chromatin by immunoglobulins.

Quantitative measurement of human IFNα in medium. The cell culture supernatant was collected and concentrated if necessary, and then 100 μL aliquots were used for detection with ELISA kit (PBL Biomedical Laboratories) following the protocol of the manufacturer. The ratio unit/pg for human IFNα is approximately 3 to 5. Data were shown as means ± SD of three independent experiments.

RIG-G is not a primary target of IFNα-activated ISGF3 complex. By Western blot analysis, we first examined whether the Janus-activated kinase (JAK)-STAT pathway–activated ISGF3 complex was involved in IFNα-induced RIG-G up-regulation. We found that in NB4 cells the total amounts of STAT1 and STAT2 continued to increase throughout the IFNα treatment in parallel with the RIG-G induction. In contrast, the level of tyrosine-phosphorylated STAT1 increased rapidly, peaking at 10 minutes to 1 hour after IFNα treatment and then returning to the levels characteristic of untreated cells, whereas the level of tyrosine-phosphorylated STAT2 followed a similar time course but with a much slower decline (Fig. 1A). The differential expression kinetics between RIG-G and the activated ISGF3 hardly exhibited a possibility that RIG-G could be a direct target of ISGF3. In addition, no tyrosine-phosphorylated STAT1 was detected in NB4 cells treated either by IFNα for 6 hours or by ATRA for 72 hours (Fig. 1A), further suggesting that the RIG-G gene expression was most likely activated through a novel mechanism.

Using the protein synthesis inhibitor cycloheximide, we questioned whether RIG-G was an IFNα primary responsive gene. In comparison with the direct transcriptional response of IRF-1 to IFNα, the RIG-G mRNA induction was entirely blocked by cycloheximide (Fig. 1B), indicating that the RIG-G was not a primary responsive gene in IFNα-induced JAK-STAT pathway.

Both IRF-9 and STAT2 are key factors required for RIG-G expression. We then conducted a survey of transcription factors essential to the expression of RIG-G. Several IFN signaling pathway–related key regulators, including STAT1, STAT2, IRF-1, and IRF-9, were transfected individually or in combination into STAT1-deficient U3A cells, together with a luciferase reporter gene containing RIG-G promoter. Consequently, STAT1, STAT2, or IRF-9 alone exhibited no obvious effect on the reporter gene expression, whereas the combination of IRF-9 and STAT2 could cooperatively cause an up to 7-fold induction in luciferase activity, as the IRF-1 alone (Fig. 1C). Some important factors implicated in myelopoiesis, such as C/EBPs and PU.1, were also evaluated for their abilities to transactivate the reporter gene. No significant change was observed when these expression plasmids were transfected into U3A cells, except that C/EBPα had a potent repressive activity (data not shown).

We found that enforced expression of STAT1 in U3A cells could rescue the response of RIG-G gene to IFNα, but mutant STAT1-Y701F had no such an effect (Fig. 1C), suggesting that phosphorylation of STAT1 was required for RIG-G expression in IFNα-treated U3A cells. However, when we coexpressed mutant STAT2-Y690F with wt STAT1 in U3A cells, IFNα-induced luciferase activity was effectively inhibited by mutant STAT2-Y690F in a dose-dependent manner (Fig. 1C). Furthermore, we observed no induction of luciferase activity in IRF-9–deficient U2A cells after IFNα treatment, unless the cells were reconstituted by IRF-9 (Fig. 1D). Taken together, these results provided an indication that IRF-9 and STAT2, rather than STAT1, were key factors necessary for RIG-G expression.

To investigate how IRF-9 and STAT2 synergistically regulate the expression of RIG-G gene, we detected whether these two proteins could interact. Unlike IFNα-activated ISGF3 complex in which tyrosine-phosphorylated STAT proteins are required, IRF-9 could successfully interact with either wt STAT2 or mutant STAT2-Y690F in the absence of IFNα, although the interaction between IRF-9 and wt STAT2 could be obviously enhanced by IFNα via tyrosine phosphorylation of STAT2 (Fig. 2A). In agreement with this, both wt STAT2 and mutant STAT2-Y690F displayed a similar effect on the luciferase reporter gene expression in U3A cells without IFNα. By contrast, in IFNα-treated U3A cells, the luciferase activity showed an additional 7-fold increase by the complex of IRF-9 with wt STAT2, but not in the case of mutant STAT2-Y690F (Fig. 2B). Noticeably, no obviously increased binding ability of IRF-9 to RIG-G promoter was observed in the presence of IFNα after normalizing the amounts of the detected samples according to input (Fig. 2C).

Dramatic induction of RIG-G by ATRA is related to the IFNα secretion in NB4 cells. The idea that both STAT2 and IRF-9 were basic components necessary for RIG-G expression was also supported by the fact that ATRA could not only induce the total amounts of STAT2 and IRF-9 proteins but also increase the tyrosine phosphorylation level of STAT2 in NB4 cells (Fig. 1A). Because ATRA could cause IFNα synthesis and secretion in NB4 cells by up-regulating IRF-1 (23), we thus quantitated the IFNα in culture supernatant of ATRA-treated NB4 cells. We noted that the IFNα level was undetectable until the NB4 cells were treated with ATRA for 72 hours when the IFNα titer was ∼7 pg/mL. After a 96-hour treatment, a further 9-fold increase in IFNα level could be measured (Fig. 3A).

To test whether the IFNα secreted from ATRA-treated NB4 cells was able to induce the STAT2 phosphorylation and RIG-G expression, we incubated the IRF-9/STAT2–cotransfected U3A cells in the supernatant of ATRA-treated NB4 cells. The results revealed that the supernatant of 72-hour ATRA-treated NB4 cells was sufficient to induce the tyrosine phosphorylation of STAT2 and the endogenous RIG-G level in U3A cells, in comparison with the relative consistent level of total STAT2 (Fig. 3B). These data suggested that the effective IFNα secretion might be one of the causes resulting in RIG-G induction in ATRA-treated NB4 cells.

IRF-1 plays an important role in the modulation of RIG-G expression. It was noteworthy that IRF-1 alone could significantly induce the expression of the reporter gene containing RIG-G promoter (Fig. 1C). To investigate the role of IRF-1 in RIG-G expression regulation, IRF-1 gene-specific siRNA treatments were included into the reporter gene assays. Consequently, we found that si-IRF-1 could significantly reduce IFNα-induced luciferase activity in STAT1-transfected U3A cells (Fig. 4A), showing that IRF-1 was indeed involved in transcriptional activation of RIG-G.

Thus, we tried to better understand the mechanisms by which IRF-1 induced RIG-G gene expression. An obvious up-regulation of both STAT2 and IRF-9 proteins was found in HT1080 cells transfected with IRF-1 (Fig. 4B), implying that the effect of IRF-1 on RIG-G expression might also be attributed to IRF-9 and STAT2. Moreover, we observed that the transactivity of IRF-9 complexed with wt STAT2, rather than mutant STAT2-Y690F, could be enhanced 4-fold more when IRF-1 coexpressed in U3A cells (Fig. 4C), suggesting a role of IRF-1 on STAT2 phosphorylation. Whether IRF-1 could also induce the IFNα synthesis as well as the phosphorylation of STAT2 was then determined. The IFNα level in the culture supernatant of IRF-1–transfected U3A cells was 3-fold elevated when compared with control cells (Fig. 3C). Similarly, increased level of STAT2 tyrosine phosphorylation was detected as well in IRF-1–transfected HT1080 cells (Fig. 4B). Together, these data indicated that IRF-1 could regulate RIG-G expression through an IRF-9/STAT2–dependent mechanism.

However, because IRF-1 also has the ability to directly bind the ISRE elements (24), an IRF-9/STAT2–independent mechanism in which IRF-1 acted should not be ruled out. This notion was sustained by the fact that IRF-1 could activate the RIG-G gene promoter in IRF-9–deficient U2A cells and by our ChIP assay, confirming the ability of IRF-1 to bind the RIG-G gene promoter (Fig. 4D). Moreover, we found that IRF-9 could significantly inhibit the transactivity of IRF-1 in a dose-dependent manner (Fig. 5A), reflecting again their competition for direct binding with RIG-G promoter.

Two ISRE elements are molecular basis required for RIG-G expression. Because both IRF-1 and the complex IRF-9/STAT2 could bind the RIG-G gene, we therefore conducted a detailed functional analysis on RIG-G promoter to precise the molecular basis for RIG-G expression. Several luciferase reporters with wt or mutant RIG-G promoters were constructed (Fig. 5B and C): pXP2 (−87) was totally in lack of two ISRE elements; pXP2-mut-1 and pXP2-mut-2 contained mutations in the site of ISRE I and ISRE II, respectively; and pXP2-mut-3 possessed double mutations. Our results showed that both pXP2 (−87) and pXP2-mut-3 lost all activities in response to indicated transcription factors. In contrast, pXP2-mut-1 could significantly reduce the transactivities of both IRF-9/STAT2 complex and IRF-1, whereas pXP2-mut-2 had no difference with the wt construct pXP2 (−310) (Fig. 5D). The similar results were also obtained in the presence of IFNα. These observations clearly indicated that the two ISRE sequences in RIG-G promoter were the molecular basis for RIG-G gene expression. Notably, it seems that both IRF-9 and IRF-1 preferentially bind the ISRE I.

This work intended to identify the transcription factors essential for RIG-G gene expression. The transcription factors usually function as homodimers or heterodimers to regulate the gene expression in eukaryotic cells and their activities are dependent on distinct protein-protein interaction. In this regard, one of the best models is IFN-stimulated JAK-STAT pathways, in which both STAT1 and STAT2 are activated by tyrosine phosphorylation in response to IFNs (4). However, it has become apparent that the activation of classic JAK-STAT pathways alone cannot account for the pleiotropic biological effects of IFNs on target cells. Increasing number of studies show that other types of STAT complexes can be formed in response to IFNs (7). Several type I IFN-mediated STAT2-dependent, but STAT1-independent, mechanisms were reported (25–30). Because STAT2 is the only member of the STAT family without DNA-binding activity, it can act as a transcription factor only when complexed to other STATs or proteins (31). Two independent reports have pointed the existence of a complex composed of STAT2 homodimers and IRF-9 without STAT1 in IFNα-treated cells (32, 33), but this finding was tempered by the facts that their authors mentioned that this complex displayed only limited DNA-binding affinity for an ISRE sequence. Therefore, for a long time, the biological significance of IRF-9/STAT2 complex has been difficult to assess.

In this study, we provide the first evidence that in STAT1-deficient U3A cells, STAT2 forms a complex with IRF-9 on the ISRE regions of RIG-G promoter and effectively mediates the transcription of RIG-G gene, even without the tyrosine phosphorylation. Because IFNα failed to induce RIG-G expression in STAT1-deficient U3A cells unless the cells were transfected with exogenous STAT1, we previously concluded that STAT1 was a prerequisite to RIG-G expression. However, our conclusion no longer proved exact because of our following new findings. First, mutant STAT2-Y690F could exert a dominant-negative effect in STAT1-reconstituted U3A cells in the presence of IFNα (Fig. 1C). Second, in IRF-9–deficient U2A cells, IFNα-induced luciferase activity was detected only when enforcing IRF-9 expression in these cells (Fig. 1D). Third, ectopic overexpression of wt STAT1 in U3A cells was unable to induce the reporter gene expression without IFNα, whereas mutant STAT1-Y701F remained responseless even under IFNα treatment (Fig. 1C), indicating that STAT1 activity depended on its phosphorylation. Taken together, data are currently convincing that STAT1 alone is not sufficient for the expression of RIG-G gene. The primary function of exogenous STAT1 transfected into U3A cells might be to restore the JAK-STAT pathways through which the downstream genes, including IRF-1, IRF-9, STAT2, and RIG-G, can be directly or indirectly induced (Fig. 6).

During the past decade, data on the putative roles of STAT proteins in mediating gene expression without tyrosine phosphorylation have been accumulating. For example, a complex of unphosphorylated STAT1 and IRF-1 was found to support the constitutive expression of low molecular mass polypeptide 2 gene coding for a subunit of 20S proteasome (15). Unphosphorylated STAT3 could drive overexpressions of certain oncogenes (such as M-RAS and MET), which contributed to cellular transformation (34). Similarly, unphosphorylated STAT6 was shown to be able to cooperate with p300 to enhance cyclooxygenase-2 gene expression in human non–small cell lung cancer (16). Here, we have shown for the first time that the unphosphorylated STAT2 could play an important role in RIG-G gene expression by interacting with IRF-9, further reinforcing the idea that STAT proteins could function as transcription factors in the absence of tyrosine phosphorylation. Interestingly, it was reported that unphosphorylated STAT2 was not static but dynamically shuttling between the nucleus and cytoplasm (31, 35, 36). The nuclear import of unphosphorylated STAT2 depended on its association with IRF-9, which contained a constitutive nuclear localization signal (35). These findings offer no doubt a prerequisite for the transactivation potential of unphosphorylated STAT2 on RIG-G expression.

Another interesting question addressed by our investigation is how ATRA can induce RIG-G expression in NB4 cells. In addition that ATRA greatly increases the protein level of both STAT2 and IRF-9, IFNα synthesis and secretion in NB4 cells after ATRA treatment for 72 hours and onward strongly suggests that ATRA-induced RIG-G may represent one of the key molecular nodes of cross-talk between ATRA and IFNα. This idea is supported by the finding that the lack of IFNα synthesis in ATRA-resistant NB4-R1 cells is exquisitely coincident with no induction of RIG-G by ATRA in these cells (data not shown). Recently, some other cytokines (such as tumor necrosis factor) or chemical drugs were also found to up-regulate several IFN-inducible gene family members (including RIG-G) through production of type I IFNs (37, 38), further indicating that the induction of RIG-G by ATRA in NB4 cells was tightly related with an IFNα autocrine pathway.

In summary, we propose here a novel mechanism for the transcriptional regulation of RIG-G, an IFNα-induced gene. As shown in Fig. 6, the binding of IFNα to its cognate receptors activates the classic JAK-STAT pathway, which is required to initiate the transcription of a set of primary target genes, such as IRF-1. The expression of unphosphorylated STAT2 and IRF-9 is greatly increased in response to IFNα probably through an indirect manner. STAT2 interacts with IRF-9 to form a transcription factor complex, which is both necessary and sufficient to transactivate the expression of RIG-G gene. On the other hand, IRF-1 can also induce RIG-G gene expression through an IRF-9/STAT2–dependent or IRF-9/STAT2–independent mechanism. In NB4 cells, ATRA can up-regulate IRF-1, which was followed by the production and secretion of IFNα. Accordingly, the induction of RIG-G by ATRA is also related to IFNα (Fig. 6). Our work deciphered the signal cross-talk between ATRA and IFNα in APL NB4 cells, but our finding should have broader implications in cancer. We furnished here an alternative pathway for the diversity of IFN signaling to mediate its multiple biological properties in normal and tumor cells.

No potential conflicts of interest were disclosed.

Grant support: Chinese National Key Program for Basic Research 973, Chinese National High Tech Program 863 (2006AA02Z19A), National Natural Science Foundation of China (30570778 and 30670882), “Shu Guang” Program of Shanghai Municipal Commission for Education (03SG37), and Samuel Waxman Cancer Research Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Lanotte for help in reviewing the manuscript and constructive discussion and suggestion and all members of the Shanghai Institute of Hematology for their support.