Proc. Natl. Acad. Sci. USAVol. 92, pp. 11657-11661, December 1995Biochemistry

Identification of a member of the interferon regulatory factorfamily that binds to the interferon-stimulated response elementand activates expression of interferon-induced genesWEI-CHUN AU*t, PAUL A. MOOREtt, WILLIAM LOWTHER*, YUANG-TAUNG JUANG*, AND PAULA M. PITHA*§¶*Oncology Center and §Department of Molecular Biology and Genetics, School of Medicine, The Johns Hopkins University, Baltimore, MD 21231; and tHumanGenome Sciences, Inc., Rockville, MD 20850

Communicated by Hamilton 0. Smith, The Johns Hopkins University, Baltimore, MD, August 17, 1995 (received for review June 21, 1995)

ABSTRACT A family of interferon (IFN) regulatory fac-tors (IRFs) have been shown to play a role in transcription ofIFN genes as well as IFN-stimulated genes. We report theidentification of a member of the IRF family which we havenamed IRF-3. The IRF-3 gene is present in a single copy inhuman genomic DNA. It is expressed constitutively in a varietyof tissues and no increase in the relative steady-state levels ofIRF-3 mRNA was observed in virus-infected or IFN-treatedcells. The IRF-3 gene encodes a 50-kDa protein that bindsspecifically to the IFN-stimulated response element (ISRE)but not to the IRF-1 binding site PRD-I. Overexpression ofIRF-3 stimulates expression of the IFN-stimulated gene 15(ISG15) promoter, an ISRE-containing promoter. The murineIFNA4 promoter, which can be induced by IRF-1 or viralinfection, is not induced by IRF-3. Expression of IRF-3 as aGal4 fusion protein does not activate expression of a chlor-amphenicol acetyltransferase reporter gene containing re-peats of.the Gal4 binding sites, indicating that this proteindoes not contain the transcription transactivation domain.The high amino acid homology between IRF-3 and ISG factor3 'y polypeptide (ISGF3y) and their similar binding propertiesindicate that, like ISGF3'y, IRF-3 may activate transcriptionby complex formation with other transcriptional factors,possibly members of the Stat family. Identification of thisISRE-binding protein may help us to understand the speci-ficity in the various Stat pathways.

The virus-induced expression of interferon (IFN) genes ininfected cells involves interplay of several constitutively ex-pressed and virus-activated transcriptional factors (1). Two ofthese factors binding to the virus-inducible element of theIFN-3 gene have been proposed to play a crucial role in theregulation of expression of IFN-a and IFN-,3 genes. Interferonregulatory factor 1 (IRF-1) was shown to act as an activator,and the closely related IRF-2 was a repressor (2). It wasproposed that induction of the IFN-,3 gene was the result of theremoval of repressor IRF-2 and the subsequent binding of theactivator IRF-1 (3). Several observations supported thismodel. Expression of IRF-1 was found to be upregulated invirus-infected cells. The IRF-1 binding sites were identified inthe IFN-3 gene promoter region and reporter plasmids withmultiple repeats of the AAGTGA hexanucleotides (which arethe strong IRF-1 binding site) were inducible by overexpres-sion of IRF-1. This transactivation could be repressed by IRF-2(2). In embryonal carcinoma cells, overexpression of IRF-1induced both the transfected and the endogenous IFN-a and-,B genes (4). Moreover, a decrease in IFN-,B induction wasobserved in cells expressing the IRF-1 antisense mRNA (5).

In contrast, studies of IFN-a promoters did not support therole of IRF-1 as the limiting factor in the virus-mediatedinduction of these genes. In mouse cells, overexpression of

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

IRF-1 was sufficient to induce transcription from the IFN-apromoters: however, it induced expression of both virus-inducible and -uninducible IFN-a promoters (6). Further, thebinding of IRF-1 to the respective IRF-1 binding sites inmurine and human IFN-a promoters was very weak, suggest-ing that if IRF-1 plays a role in induction of IFN-a genes, itcooperates with another binding protein(s) (7). The strongestargument against the limiting role of IRF-1 and IRF-2 is thatthe virus-mediated induction of IFN-a or -3 genes was notaltered in IRF-1 null mice or in cells with the IRF-1 genedeleted (8, 9), suggesting that IRF-1 function can be replacedby other transactivators, which may bind to the same domain.

IRF-1 was also shown to bind to the interferon-stimulatedresponse element (ISRE) present in promoters of genesactivated by IFN (10) and to have a direct role in regulation ofexpression of several IFN-induced genes (11-13). In addition,two other proteins [IFN consensus sequence-binding protein(ICSBP) and IFN-stimulated gene factor 3 'y polypeptide(ISGF3'y)] show similarity to IRF-1 and IRF-2 at the aminoacid level and bind to the ISRE, indicating the existence of afamily of IRF-1-like transcription factors (14, 15). All of thesefactors share a high degree of homology in the N-terminalDNA-binding domain but have diversity in the C-terminalregion, where IRF-1 contains the transcription activationdomain. ICSBP, which, unlike IRF-1 and -2, is expressed onlyin cells of lymphoid origin, was shown to bind the ISRE ofmany ISGs and repress the IFN-mediated activation of IFN-inducible promoters (16). The ISGF3'y (p48) present in thecytoplasm is assembled, in cells treated with IFN-a, togetherwith the Statl and Stat2 proteins into the ISGF3 complex andmediates binding of this complex to the ISRE (14, 17). c-Mybalso belongs to the family of IRF-1-like proteins, although itsrelationship to the IFN system is unclear (14).The aim of this study was to find a transcriptional factor that

can replace the function of IRF-1 and stimulate transcriptionof the IFN-a promoter. We have identified and characterizeda member of the IRF-1 family, IRF-3, that may function as aregulatory component in virus-infected cells.11

MATERIALS AND METHODSIsolation of a cDNA Showing Homology to the IRF Family

Members. An expressed sequence tag (EST) cDNA database

Abbreviations: IFN, interferon; ICSBP, IFN consensus sequence-binding protein; IRF, IFN regulatory factor; ISG, IFN-stimulatedgene; ISGF3, ISG factor 3; ISRE, IFN-stimulated response element;CAT, chloramphenicol acetyltransferase; EST, expressed sequencetag; GST, glutathione S-transferase; NDV, Newcastle disease virus;PRD, positive regulatory domain.tW.-C.A. and P.A.M. contributed equally to this project.ITo whom reprint requests should be addressed at: Oncology Center,The Johns Hopkins University, 418 North Bond Street, Baltimore,MD 21231-1001.lThe sequence reported in this paper has been deposited in theGenBank database (accession no. Z56281).

11657

Dow

nloa

ded

by g

uest

on

Janu

ary

20, 2

020

Proc. Natl. Acad. Sci. USA 92 (1995)

was screened for homology by using the BLAST network serviceprovided by the National Center for Biotechnology Informa-tion. Several overlapping ESTs showing homology to IRF-1and IRF-2 were identified, and one that appeared to befull-length was sequenced and designated IRF-3. The firstdesignated methionine codon in the IRF-3 clone is likely to bethe start codon because its context (ACC-AUG-G) fits theKozak consensus sequence for translation (RCC-AUG-G).The IRF-3 clone was used for subsequent studies.

Plasmid Constructs. The IRF-3 expression plasmid wasprepared by cloning the BamHI-Xho I fragment containingthe IRF-3 cDNA from the pSKIRF-3 plasmid behind thecytomegalovirus promoter in the pCDNAINEO vector (In-vitrogen). The IRF-1 expression plasmid has been described(18). The Gal4-IRF fusion plasmids were constructed by firstamplifying full-length IRF-1, IRF-2, and IRF-3 cDNAs byPCR. After digestion of the amplified fragments with Xho IandXba I (for IRF-1 and IRF-3) or Sal I andXba I (for IRF-2),these fragments were cloned in-frame with the GAL4 sequencecorresponding to aa 1-147 of the Gal4 DNA binding domainin pSG424 (19). The indicator plasmids containing the chlor-amphenicol acetyltransferase (CAT) gene inserted behind thevirus-inducible 452-nt IFNA4 promoter region (IFNA4/CAT)or the IFN-inducible promoter of ISG15 (ISG15/CAT) orthree copies of ISRE (from ISG15) inserted in front of the-119 human immunodeficiency virus promoter-CAT plasmidhave been described (7). For construction of glutathioneS-transferase (GST)-IRF and GST-ISGF3,y fusion plasmids,the EcoRI-Xho I fragment of the IRF-3 cDNA or the PCR-amplified ISGF3,y cDNA were cloned into pGEX-4T-2 vector(Pharmacia Biotech) digested with the same enzymes. Forconstruction of the GST-IRF-1 fusion plasmid, murine IRF-1cDNA isolated as a Pst I fragment from pIRF-1AS (18) wascloned into pGEX-4T-2 digested with Sal I.

In Vitro Translation of IRF-3 mRNA and Expression of GSTFusion Protein. IRF-3 mRNA was translated in a rabbitreticulocyte lysate (Promega). The GST fusion proteins were

purified from bacterial lysates by affinity chromatography ona glutathione-agarose column (Sigma). To remove the GSTportion, fusion protein bound to glutathione-agarose beadswas treated with thrombin (20%, wt/wt) for 2 hr and recom-binant IRF-3 was eluted with phosphate-buffered saline.

Transfection and CAT Assay. Transfections were done bycalcium phosphate precipitation (6, 7). Treatment with type IIFN (100 units/ml) or infection with Newcastle disease virus(NDV) (multiplicity of infection, 5) was done 16 hr aftertransfection for 8 hr. IFN or virus inoculum was then removed,and cells were washed and incubated in medium for 16 hrbefore harvest for the CAT assay.

Gel Mobility-Shift Assay. The purified GST-IRF fusionproteins (100 ng) were incubated with labeled probes (1-10 pg)in the presence of nonspecific competitor poly(dIdC) (1 ,ug)as described (17) and the protein-DNA complexes wereresolved in a nondenaturing 4% polyacrylamide gel. Thefollowing oligonucleotides were used for DNA-binding studiesor as competitors: ISRE, 5'-CAGTTTCGGTTTCCCTTT-3';positive regulatory domain I (PRD-I), 5'-GAGAAGTGAA-AGTGGGAACCCTCTCCTT-3' (the underlined sequencewas used for annealing of primer 5'-AAGGAGAGGG-3' andsynthesis of a complementary strand).

Southern and Northern Hybridizations and Si NucleaseAnalysis. Southern hybridization used 32P-labeled IRF-3cDNA as a probe. Tissue distribution of IRF-3 mRNA wasdetermined with a multiple-tissue Northern blot of poly(A)+RNA (Stratagene) probed with 32P-labeled IRF-3 cDNA bymodified hybridization (20). Total RNA isolated from NDV-infected HeLa cells was analyzed by hybridization with 32p-labeled IRF-3 cDNA probe as described (7). For S1 nucleaseanalysis, total RNA was hybridized with 32P-labeled IRF-3 and,3-actin RNA probes.

RESULTS

Identification and Characterization of the IRF-3 Clone. Adatabase of ESTs (21) generated from multiple human tissuesat Human Genome Sciences, Inc., and the Institute ofGenomic Research was searched for homologs of IRF-1 andIRF-2 by using the BLAST algorithm (22). A cDNA with anopen reading frame of 427 aa was identified (Fig. 1), desig-nated IRF-3, characterized, and used for further studies.

In addition to IRF-1 and IRF-2, IRF-3 is also homologousto ICSBP and ISGF3,y (Fig. 2). The homology to IRF-1 andIRF-2 is restricted to the N-terminal 110 aa containing thecharacteristic tryptophan repeats, whereas the homology toICSBP and ISGF3,y extends through the whole coding region.In addition, IRF-3, ISGF-y, and ICSBP contain an identicaldomain of 7 aa (positions 34-40) from which 6 aa are alsopreserved in IRF-1 and IRF-2. In the N-terminal region, IRF-3is 34% and 37% identical with IRF-1 and IRF-2 respectively,whereas the values for ICSBP and ISGF3,y are 39.8% and35.2%, respectively. In the C-terminal region, IRF-3 is 25.3%identical to ICSBP and 18.6% identical to ISGF3-y.

Expression of IRF-3 Gene in Various Tissues and Cells.Southern hybridization of human genomic DNA (HeLa cells)showed a single copy of the IRF-3 gene (Fig. 3A). Northernhybridization detected a 1.6-kb band in all tissues examined,indicating that this gene is expressed constitutively (Fig. 3B).To determine whether IRF-3 gene expression is further stim-ulated in virus-infected or IFN-treated cells, total RNA iso-lated from either NDV-infected or IFN-treated cells at varioustimes postinduction was analyzed by Northern hybridizationand S1 nuclease mapping. From 1 to 24 hr after infection, therelative levels of IRF-3 mRNA in NDV-infected HeLa cellsdid not change significantly, nor were there changes in the

M G T P KGGTTCCAGCTGCCCGCACGCCCCGACCTTCCATCGTAGGCCGGACCATGGGAACCCCAAAP R I L P W L V S Q L D L G Q L E G V A

GCCACGGATCCTGCCCTGGCTGGTGTCGCAGCTGGACCTGGGGCAACTGGAGGGCGTGGCW V N K S R T R F R I P W K H G L R Q D

CTGGGTGAACAAGAGCCGCACGCGCTTCCGCATCCCTTGGAAGCACGGCCTACGGCAGGAA Q Q E D F G I F Q A W A E A T G A Y VTGCACAGCAGGAGGATTTCGGAATCTTCCAGGCCTGGGCCGAGGCCACTGGTGCATATGTP G P D K P D L P T W K R N F R S A L N

TCCCGGGAGGGATAAGCCAGACCTGCCAACCTGGAAGAGGAATTTCCGCTCTGCCCTC'AR K E G L R L A E D R S K D P H D P H K

CCGCAAAGAAGGGTTGCGTTTAGCAGAGGACCGGAGCAAGGACCCTCACGACCCACATAAI Y E F V N S G V G D F S Q P D T S P D

AATCTACGAGTTTGTGAACTCAGGAGTTGGGGACTTTTCCCAGCCAGACACCTCTCCGGAT N G G G S T S D T Q E D I L D E L L G

CACCAATGGTGGAGGCAGTACTTCTGATACCCAGGAAGACATTCTGGATGAGTTACTGGGN M V L A P L P D P G P P S L A V A P E

TAACATGGTGTTGGCCCCACTCCCAGATCCGGGACCCCCAAGCCTGGCTGTAGCCCCTGAP C P Q P L R S P S L D N P T P F P N L

GCCCTGCCCTCAGCCCCTGCGGAGCCCCAGCTTGGACAATCCCACTCCCTTCCCAAACCTG P S E N P L K R L L V P G E E W E F E

GGGGCCCTCTGAGAACCCACTGAAGCGGCTGTTGGTGCCGGGGGAAGAGTGGGAGTTCGAV T A F Y R G R Q V F Q Q T I S C P E GGGTGACAGCCTTCTACCGGGGCCGCCAAGTCTTCCAGCAGACCATCTCCTGCCCGGAGGGL R L V G S E V G D R T L P G W P V T L

CCTGCGGCTGGTGGGGTCCGAAGTGGGAGACAGGACGCTGCCTGGATGGCCAGTCACACTP D P G M S L T D R G V M S Y V R H V L

GCCAGACCCTGGCATGTCCCTGACAGACAGGGGAGTGATGAGCTACGTGAGGCATGTGCTS C L G G G L A L W R A G Q W L W A Q R

GAGCTGCCTGGGTGGGGGACTGGCTCTCTGGCGGGCCGGGCAGTGGCTCTGGGCCCAGCGL G H C H T Y W A V S E E L L P N S G H

GCTGGGGCACTGCCACACATACTGGGCAGTGAGCGAGGAGCTGCTCCCCAACAGCGGGCAG P D G E V P K D K E G G V F D L G P FTGGGCCTGATGGCGAGGTCCCCAAGGACAAGGAAGGAGGCGTGTTTGACCTGGGGCCCTTI V D L I T F T E G S G R S P R Y A L W

CATTGTAGATCTGATTACCTTCACGGAAGGAAGCGGACGCTCACCACGCTATGCCCTCTGF C V G E S W P Q D Q P W T K R L V M V

GTTCTGTGTGGGGGAGTCATGgCCCCAGGACCAGCCGTGGACCAAGAGGCTCGTGATGGTK V V P T C L R A L V E M A R V G G A S

CAAGGTTGTGCCCACGTGCCTCAGGGCCTTGGTAGAAATGGCCCGGGTAGGGGGTGCCTCS L E N T V D L H I S N S H P L S L T S

CTCCCTGGAGAATACTGTGGACCTGCACATTTCCAACAGCCACCCACTCTCCCTCACCTCD Q Y K A Y L Q D L V E G M D F Q G P G

CGACCAGTACAAGGCCTACCTGCAGGACTTGGTGGAGGGCATGGATTTCCAGGGCCCTGGE S *

GGAGAGCTGAGCCCTCGCTCCTCATGGTGTGCCTCCAACCCCCCTGTTCCCCACCACCTCAACCAATAAACTGGTTCCTGCTATGAAAAAAAAAAAAAAAAAAAAA

FIG. 1. Nucleotide and predicted amino acid sequence of the IRF-3cDNA. Asterisk marks termination codon.

11658 Biochemistry: Au et aL

Dow

nloa

ded

by g

uest

on

Janu

ary

20, 2

020

Proc. Natl. Acad. Sci. USA 92 (1995) 11659

A

MGTPKPRILPWLVSQIDLGQLEGVAlVNIKSRTREMIPAGLRQDAQQE-DF)GI}OANAEATGAYVPGRDXCPDLPTM5N^RSANRKEGLRLAEDRSK-DPHDPHKIY 108MCDRNGGGRLRQI@LIEQIDSSMYPGL VENEEKSHMFR PAGKQDYNQEVDASIFKAVFKGKFKEG-DXAEPATWIICTRRCALNKS PDFEEVTDRSQLD ISEPYKVY 111

MASGRARCTRKIRNIvVEQVESGQFPGVCWV DTAKTQFMPG TAGKQDFRED EDAQFDKAUFKGKYKErG-DTGGPAVNKTRLACALNKSSEFKEVPERGRMDVAEPYKVY 113MPITREMMQPEMINSNQI PGLIlfINltEEMIFQI" AKHGWDINKDACLFRSlEI HTGRYKAGEKEPDPKllMMCAMNSLPDIEEVKDQSRNKGSSAVRVY 110MPVERMERMEEQINSNTIPPGLK2MEKKIFQIPW2IAARHGWDVZKDAPLFRNRAIHTGKHQPGVDDPKTOKANlCAMNSLPDIEEVKDKSIKKGNNAFRVY 110

210 YRGRVFISCPZG--IRSVGSEGDTL-PGWPVTLPDPGMSLTDRGVMSYVRHVLSCGGGLLWRAGaIJ WAQRLGHCHTYWAVSEELLPNSGH 305211 YGGKL GCRLSLSQPGLPGTKLYGP30LELVRFPPADTI6PSERQRQVFGHGV-LHSSR GRVF-CSGNAVCK--- 306226 YNGRVVGEAQVQSLDC--RLVAEPSGSESSM ---Z--QVLFPKP------- GPLEPTQRLLQLZGILVASNPRGP7VgRLPIPISWNAPQ ---APPGP 309

306 GPDGEVPKDKEGGVFDLGPFIVDLITFTEGSRSPRYALWFCVGSWPQDQPWTKRLVMVKVVPTCLRALV-EMARVGGASSL 387307 GRPNKLERDZVVQV!DTSQFFEELQQTYNSQGRLPDGRVVLCIYZFPDMAPLRSKLILVQIEQLYVRQLAEZAGKSCGAGSV 389310 GPH-LLPSNZCVELFRTAYFCRDLVRYFQGLGPPPKFQVTLNFWZSHGSSHTPQNLITVKISQAFARYLL-EQTPEQQAAIL 390

FIG. 2. (A) N-terminal amino acid homology among members of the IRF family. Conserved amino acids are shown in bold type. (B) C-terminalamino acid homology among IRF-3, ISGF3-y, and ICSBP.

levels of IRF-3 mRNA in these cells after treatment withIFN-a (data not shown). To determine whether IRF-3 isinduced by IFN in cells of lymphoid origin (as observed forICSBP), we compared by Si analysis the levels of IRF-3mRNA with those of the constitutively expressed ,B-actinmessage in IFN-a-treated Jurkat cells (T cells) and Namalwacells (B cells). All samples showed a single 268-nt protectedband, the intensity of which did not change (after normaliza-tion to the actin mRNA levels) during 16 hr of IFN treatment(Fig. 3C). These data indicate that expression of the IRF-3gene is not stimulated in infected or IFN-treated cells.

Identification of a 50-kDa IRF-3 Peptide and DNA-BindingProperties of IRF-3 Synthesized in Bacteria. Translation ofIRF-3 mRNA in the rabbit reticulocyte system yielded a

50-kDa polypeptide (data not shown), indicating that IRF-3 isvery similar in size to ISGFy (48 kDa) and IRF-1 (45 kDa) (2,14). IRF-3 expressed as a GST fusion protein in bacteria waspurified on a glutathione-agarose beads column and used inDNA-binding assays. The GST protein alone did not bind toany of the probes used. In a mobility-shift assay, the GST-

B

A

IRF-3 protein formed with the ISRE probe one major, fast-moving complex (Fig. 4A, lane 2) and two minor, slowlymoving complexes (seen after a prolonged exposure; Fig. 4A,lane 4). Addition of antibodies to the GST fusion proteincompletely eliminated the complex formation, indicating thatthe multiple complexes formed represented binding of theGST-IRF-3 fusion protein and were not the result of itsdegradation (data not shown). Removal of GST fusion proteinfrom IRF-3 by cleavage with thrombin did not change thebinding pattern (data not shown). Complex formation wasspecific; it was inhibited by a 50-fold excess of nonradioactiveISRE oligonucleotide but not by a nonspecific polynucleotide(Fig. 4B). The binding of GST-IRF-1 fusion protein to ISREshowed only one DNA-protein complex (Fig. 4A, lanes 1 and3), suggesting that the binding properties of IRF-1 and IRF-3are not identical. Since IRF-1 binds very effectively to thePRD-I domain present in the IFN-,B gene promoter region, wecompared the binding of these two proteins to the PRD-Iprobe. IRF-1 bound very strongly to this probe and formed a

single DNA-protein complex (Fig. 4A, lane 5), but no complexformation was detected between IRF-3 and PRD-I (lane 6).Thus, IRF-3 is a DNA-binding protein that can bind specifi-cally to ISRE, but not to the IRF-1 binding site. Since ISGF3,ybinds only to the ISRE, but not to PRD-I (17), we comparedthe binding of ISGF3,y and IRF-3 GST fusion proteins with the

1 2 3 4 5 6 7 8

C

A-I I?- 7 W (

-.....--.

4.

BISRE poly dl.dC

0 1 10 50 100 1 10 50 100

_~~~~

1 2 1 2 3 4 5

FIG. 3. (A) Southern blot hybridization of human (HeLa cell)DNA digested with HindlIl (lane 1) or EcoRI (lane 2) and probed withIRF-3 cDNA. (B) Northern blot analysis of poly(A)+ RNA fromhuman tissues. Lanes: 1, spleen; 2, thymus; 3, prostate; 4, testis; 5,ovary; 6, small intestine; 7, colon; 8, peripheral blood leukocytes. (C)S1 analysis of total RNA (5 jig) isolated from HeLa cells treated withIFN-a and IFN--y for various times. Lane 1, untreated cells; lanes 2 and3, cells treated with IFN-a (500 units/ml) for 5 and 16 hr, respectively;lanes 4 and 5, cells treated with IFN--y (100 units/ml) for 5 and 16 hr,respectively. The lower band (145 nt) represents ,B-actin mRNA; theupper band (268 nt) represents IRF-3 mRNA.

1 2 3 4 5 6 7 8

FIG. 4. Gel mobility-shift assay. (A) An ISRE oligodeoxynucle-otide was used as probe for binding of the recombinant GST-IRF-1(lanes 1 and 3), GST-IRF-3 (lanes 2,4, and 8), and GST-ISGF3'y (lane7) fusion proteins. Equal amounts (100 ng) of fusion proteins wereused in all binding reactions. Lanes 1 and 2 show a 4-hr exposure; lanes3 and 4 show a 16-hr exposure; and lanes 7 and 8 show a 24-hr exposure,respectively. A PRD-I oligodeoxynucleotide was used as probe forbinding of IRF-1 and IRF-3 GST fusion proteins (lanes 5 and 6,respectively; 24-hr exposure). (B) Specificity of GST-IRF-3 fusionprotein binding to the ISRE. Binding reaction mixtures containedunlabeled ISRE oligodeoxynucleotide (specific competitor) or

poly(dIdC) (nonspecific competitor) at the indicated molar ratio(0-100) relative to the radioactive ISRE probe.

1m3ICSBPISGF3yIRF1IRF2

B

w3ICSBPISGF3y

Biochemistry: Au et al.

k.A, -Mo.

I"pw0

Dow

nloa

ded

by g

uest

on

Janu

ary

20, 2

020

Proc. Natl. Acad. Sci. USA 92 (1995)

ISRE probe. The gel mobility-shift assay showed weak bindingof GST-ISGF3-y fusion protein and only one DNA-proteincomplex was detected (Fig. 4A, lane 7), indicating that recom-binant IRF-3 has a much stronger affinity for the ISRE thandoes recombinant ISGF3y.

IRF-3 Activates the ISG15 Promoter Region and MinimalPromoter Containing Multiple Copies of the ISRE in aTransient Expression Assay. To determine whether IRF-3 canactivate promoters containing ISREs or promoters of IFNgenes containing the virus-inducible element, the IRF-3cDNA was placed under the control of the cytomegaloviruspromoter.

Cotransfection of IRF-3 expression plasmid with a reporterplasmid containing 2000 nt of the ISG15 promoter regioninserted 5' of the CAT gene resulted in a dose-dependentincrease in CAT activity (Fig. 5A). This promoter region isactivated by IFN-a (17, 23, 24). When IFN induction was doneunder suboptimal conditions (8-hr treatment) leading to onlya small increase in CAT activity, transfection with IRF-3,followed by 8 hr of IFN treatment, showed a highly synergisticactivation (25-fold increase, compared with the 6-fold increase

80 A

r 60.2

(a0

C> 40U

- 20

...,.- ...

z

U-

co

ztL

CN C(N

z z

U-

0 r

trm

ISG 15/CAT:IRF-3

c

0.)

0

16 B

14121086420

0

C')

cc

i, -r

z z

LLcm

0

zCL)

cc

FIG. 5. IRF-3 induces expression from the ISG15 promoter but notthe IFNA4 promoter. (A) Hybrid plasmid (5 ,g) containing 2000 nt ofthe ISG15 promoter region inserted in front of the CAT gene was

transfected into murine L929 cells either alone or with increasingamounts of IRF-3 expression plasmid. CAT activity was assayed 24 hrafter transfection. All transfections were done in the presence of a

reference plasmid, pCH1 10 (1 jig), encoding ,B-galactosidase. Whereindicated, transfected cells were treated with murine IFN (100 units/ml) for 8 hr or 24 hr. (B) Hybrid plasmid (5 ,ug) containing 452 nt ofthe murine IFNA4 promoter region inserted in front of the CAT gene(6) was cotransfected with either IRF-I or IRF-3 expression plasmid(5 ,ug) and pCHI 10 (1 ,ug). Where indicated, the transfected cells wereinfected with NDV. Percent chloramphenicol conversion was calcu-lated by dividing the radioactivity (cpm) present in 3-acetylchloram-phenicol and 1-acetylchloramphenicol fractions by the sum of radio-activity in unconverted chloramphenicol and these two fractions. Thelevels of CAT activity were normalized to the constant level of/3-galactosidase. The maximal difference in 3-galactosidase activity ina single experiment was 2-fold. An average value from three inde-pendent experiments is given. There was 15% variability betweenindividual experiments.

10

80.6

0

U4,

0~~C)40--

2

o6 > zLL IL

4-J

(9

0zU-cc4

(9

z

....

U-

4

!-J

(9

M E:a.a m

(N M > z

cc cc z

e C'X LLJ -j uL cc

< < A

( ( 4<

< (9

(9

FIG. 6. Gal4-IRF-3 fusion protein does not activate transcriptionwhen bound to a Gal4 DNA-binding domain upstream from thetranscription start site. Gal4-IRF-1, -2, or -3 expression plasmid (5 jig)was transfected into L929 cells together with an indicator plasmid thatcontained four Gal4 DNA binding sites cloned 5' of the minimalthymidine kinase promoter-CAT plasmid. CAT activity was assayed48 hr after the transfection. Where indicated, transfected cells weretreated with murine IFN (500 units/ml) or infected with NDV(multiplicity of infection, 5) for 8 hr.

by IRF-3 and 3-fold increase by IFN). While treatment withIFN for 24 hr led to a significant increase in CAT activity, thesynergy between IFN and IRF-3 was less obvious (20-foldincrease by IFN and IRF-3 and 10-fold increase by IFN alone).

In contrast, cotransfection of IRF-3 expression plasmid withthe reporter plasmid containing the IFNA4 promoter region infront of the CAT gene did not lead to increased CAT activity,indicating that the IFNA4 promoter is not activated by IRF-3(Fig. SB). This promoter is, however, activated by cotransfec-tion with the IRF-1 expression plasmid or by viral infection(NDV). Although IRF-3 did not affect IRF-1-mediated stim-ulation (data not shown), infection of IRF-3-transfected cellsled to an -3-fold increase in virus-mediated induction of theIFNA4 promoter. Thus, whereas overexpression of IRF-3alone was unable to induce this promoter, it could enhance theNDV-mediated induction of the IFNA4 promoter.

Gal4-IRF-1, but Not Gal4-IRF-3, Fusion Protein CanActivate Transcription. To determine whether IRF-3 containsan activation region similar to that identified in IRF-1 (25), wetested the ability of IRF-3 to stimulate transcription and fusedIRF-3 cDNA in-frame to the coding sequence for the DNA-binding domain of Gal4. The reporter contained five Gal4binding sites inserted upstream of a minimal thymidine kinasepromoter. Gal4-IRF-1, but not Gal4-IRF-3 or Gal4-IRF-2,stimulated the transcription of the CAT gene (Fig. 6). It hasbeen shown by others that IRF-2, in contrast to IRF-1, does notcontain the transactivation domain (J. Hiscott, personal com-munication). In cotransfection studies, we have observedenhancement of IRF-3-mediated transactivation in IFN-treated or virus-infected cells. Therefore, we tested whethervirus or IFN can also modulate the activity of the Gal4-IRF-3fusion protein. Neither IFN treatment nor viral infectionactivated the inducing potential of Gal4-IRF-3 or Gal4-IRF-1(Fig. 6). These results indicate that IRF-3 does not contain atransactivation domain and that the observed activation of theISRE may be mediated by coassembly of IRF-3 with anothertranscriptional activator(s).

DISCUSSIONWe have isolated and expressed a cDNA clone designatedIRF-3 that shows a high degree of similarity to members of theIRF family. The IRF-3 gene is present as a single copy ingenomic DNA and is expressed constitutively at the mRNA

11660 Biochemistry: Au et al.

3.,3p

Dow

nloa

ded

by g

uest

on

Janu

ary

20, 2

020

Proc. Natl. Acad. Sci. USA 92 (1995) 11661

level in various human tissues. Its expression is not furtherstimulated by virus infection or IFN treatment.The predicted amino acid sequence in the N-terminal part

of IRF-3 shows 34-40% identity with other members of theIRF family, including five tryptophan residues preserved in thefirst 90 aa of all IRF-like proteins. The DNA-binding domainof IRF-1, IRF-2, and ISGF3,y is localized in the N-terminal120-180 aa (refs. 1 and 14; J. Hiscott, personal communica-tion). Thus, although the binding domain of IRF-3 has notbeen determined, we assume that it is also localized in theN-terminal region.

IRF-3 shows DNA-binding specificity similar to that ofISGF3y. IRF-3 binds the ISRE sequence but fails to bind thePRD-I oligodeoxynucleotide that serves as a strong bindingsite for IRF-1. However, in a mobility-shift assay, the com-plexes formed between IRF-3 or ISGF3y and the ISRE did notshow identical mobility. Whereas the binding of ISGF3y to theISRE is rather weak and yields a single complex, IRF-3 showeda stronger affinity for the ISRE and complexes with threedifferent mobilities were detected. Since purified IRF-3 pro-tein was used, these data suggest that IRF-3 may be dimerized.Interestingly, the central part of IRF-3 is proline-rich andcontains one cluster (aa 150-197) with prolines making up34% of the amino acids. Mutational analysis is needed todetermine whether this proline-rich region is involved in theinteraction of IRF-3 with other proteins or its dimerization. InISGF3y, the region required for interaction with the Statl andStat2 proteins lies between aa 271 and 377 (26). IRF-3 andISGF3y share some homology in this region; however, a proline-rich cluster (aa 285-309) is found only in ISGF3,y, and not inIRF-3 or the other IRF family members. It is possible that thisregion is involved in the interaction of ISGF3,y with the Statproteins. After this work was completed, another member of theIRF-I family, Pip (27), was described that is closely related toICSBP. IRF-3 and Pip are not identical proteins.

Several observations indicate that IRF-3 may act throughan association with other proteins. (i) IRF-3 fused to theDNA-binding domain of Gal4 is unable to activate transcrip-tion of a reporter gene containing the Gal4 binding site. Incontrast, IRF-1, which contains a transactivation domain in theC-terminal part of the molecule, is a very efficient transacti-vator in the same system. (ii) IRF-3-mediated transactivationof the ISG15 promoter is significantly more effective in cellstreated with IFN for a short time (that alone does not resultin an effective transactivation), indicating that enhancement isthe result of an IFN-induced modification of a cellular pro-tein(s) that interacts with IRF-3. This enhancement of IRF-3activity by IFN can be demonstrated only when IRF-3 binds tothe ISRE and not when IRF-3 binds to DNA through the Gal4binding site. This result indicates that interaction with otherproteins may depend on the conformation of IRF-3 and/orthat the protein interacting with IRF-3 also binds to the ISREor the adjacent sequences. Similarly, in the ISGF3 complex,not only ISGF3y but also Statl and Stat2 interact with theISRE (28). (iii) Although IRF-3 does not activate the IFNA4promoter, it enhances the virus-mediated induction of thatpromoter. This activation may result from the association ofIRF-3 with virus-modified cellular factors and a consequentincrease in the IRF-3 binding to the virus-inducible element inthe IFNA4 promoter. Modification of the DNA-binding do-main through interaction of the DNA-binding protein with aprotein unable to bind DNA was reported for the Oct-Iprotein, which recognized a G+A-rich binding site uponinteraction with the viral protein VP16 (29, 30).

In conclusion, our data suggest the existence of a DNA-binding protein, IRF-3, that by association with cellular pro-teins activated by viral infection or by IFN, increases tran-scriptional activity of targeted promoters. The proteins thatinteract with IRF-3 and the target gene that these complexesactivate are unknown. Recently, a novel nuclear factor binding

to the ISRE was identified in cells infected with vesicularstomatitis virus (31). Whether this protein is IRF-3 is un-known. However, the existence of another DNA-binding pro-tein of the IRF family that can function as a regulatorycomponent in infected or IFN-treated cells could providefurther complexity and specificity to the regulatory networkmediating the responses to viral infection.

The contribution of the Human Genome Sciences and the Institutefor Genomic Research sequencing facilities in sequencing the initialESTs displaying homology to IRF-1 and IRF-2 is appreciated. Thisstudy was supported by an American Foundation for AIDS Researchscholarship (P.A.M.) and by Grant A119713 from the National Insti-tutes of Health (P.M.P.). W.L. is a student in the predoctoral trainingprogram in Human Genetics (T32GM07814).

1. Tjian, R. & Maniatis, T. (1994) Cell 77, 5-8.2. Fujita, T., Kimura, Y., Miyamoto, M., Barsoumian, E. L. & Taniguchi,

T. (1989) Nature (London) 337, 270-272.3. Harada, H., Fujita, T., Miyamoto, M., Kimura, Y., Maruyama, M.,

Furia, A., Miyata, T. & Taniguchi, T. (1989) Cell 58, 729-739.4. Harada, H., Willison, K., Sakakibara, J., Miyamoto, M., Fujita, T. &

Taniguchi, T. (1990) Cell 63, 303-312.5. Reis, L. F., Harada, H., Wolchok, J. D., Taniguchi, T. & Vilcek, J.

(1992) EMBO J. 11, 185-193.6. Au, W.-C., Raj, N. B. K., Pine, R. & Pitha, P. M. (1992) Nucleic Acids

Res. 20, 2877-2884.7. Raj, N. B. K., Au, W.-C. & Pitha, P. M. (1991) J. Biol. Chem. 266,

11360-11365.8. Ruffner, H., Reis, L. H., Naf, D. & Weissmann, C. (1994) EMBO J.

13, 4798-4806.9. Matsuyama, T., Kimura, T., Kitagawa, M., Watanabe, N., Kundig, T.,

Amakawa, R., Kishihara, K., Wakeham, A., Potter, J., Furlonger, C.,Narendran, A., Suzuki, H., Ohashi, P., Paige, C., Taniguchi, T. & Mak,T. (1993) Cell 75, 83-97.

10. Pine, R., Decker, T., Kessler, D. S., Levy, D. E. & Darnell, J. E., Jr.(1990) Mol. Cell. Biol. 10, 2448-2457.

11. Briken, V., Ruffner, H., Schultz, U., Schwarz, A., Reis, L. F. L.,Strehlow, I., Decker, T. & Staeheli, P. (1995) Mol. Cell. Biol. 15,975-982.

12. Neish, A. S., Read, M. A., Thanos, D., Pine, R., Maniatis, T. &Collins, T. (1995) Mol. Cell. Biol. 15, 2558-2569.

13. Kamijo, R., Harada, H., Matsuyama, T., Bosland, M., Gerecitano, J.,Shapiro, D., Le, J., Koh, S. I., Kimura, T., Greene, S. J., Mak, T. W.,Taniguchi, T. & Vilcek, J. (1994) Science 263, 1612-1615.

14. Veals, S. A., Schindler, C., Leonard, D., Fu, X.-Y., Aebersold, R.,Darnell, J. E., Jr., & Levy, D. E. (1992) Mol. Cell. Biol. 12, 3315-3324.

15. Driggers, P. H., Ennist, D. L., Gleanson, S. L., Mak, W.-H., Marks,M. S., Levi, B.-Z., Flanagan, J. R., Appella, E. & Ozato, K. (1990)Proc. Natl. Acad. Sci. USA 87, 3743-3747.

16. Nelson, N., Marks, M. S., Driggers, P. H. & Ozato, K. (1993) Mol. Cell.Biol. 13, 588-599.

17. Kessler, D. S., Veals, S. A., Fu, X.-Y. & Levy, D. E. (1990) Genes Dev.4, 1754-1765.

18. Au, W.-C., Su, Y., Raj, N. B. K. & Pitha, P. M. (1993) J. Biol. Chem.268, 24032-24040.

19. Sadowski, I. & Ptashne, M. (1989) Nucleic Acids Res. 17, 7539.20. Church, G. M. & Gilbert, W. (1984) Proc. Natl. Acad. Sci. USA 81,

1991-1995.21. Adams, M. D., Dubnick, M., Kervlavage, A. R., Moreno, R., Kelley,

J. M., Utterback, T. R., Nagle, J. W., Fields, C. & Venter, J. C. (1991)Nature (London) 355, 632-634.

22. Altschul, S. F., Gish, W., Miller, W., Myers, E. W. & Lipman, D. J.(1990) J. Mol. Biol. 215, 403-410.

23. Stark, G. R. & Kerr, I. M. (1992) J. Interferon Res. 12, 147-151.24. Su, Y., Popik, W. & Pitha, P. M. (1995) J. Virol. 69, 110-121.25. Lin, R., Mustafa, H., Nguyen, N. & Hiscott, J. (1994) J. Biol. Chem.

269, 17542-17549.26. Veals, S. A., Santa Maria, T. & Levy, D. E. (1993) Mol. Cell. Biol. 13,

196-206.27. Eisenberg, C. S., Singh, H. & Storb, U. (1995) Genes Dev. 9, 1377-

1387.28. Qureshi, S. A., Salditt-Georgieff, M. & Darnell, J. E., Jr. (1995) Proc.

Natl. Acad. Sci. USA 92, 3829-3833.29. Kristie, T. M., LeBowitz, J. H. & Sharp, P. A. (1989) EMBO J. 8,

4429-4238.30. Stern, S. & Herr, W. (1991) Genes Dev. 5, 2555-2566.31. Bovolenta, C., Lou, J., Kanno, Y., Park, B.-K., Thornton, A. M.,

Coligan, J. E., Schubert, M. & Ozato, K. (1995) J. Virol. 69, 4173-4181.

Biochemistry: Au et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

20, 2

020