Virus-specific Activation of a Novel Interferon Regulatory Factor, IRF-5, Results in the Induction of Distinct Interferon α Genes*

Interferon regulatory factor (IRF) genes encode DNA-binding proteins that are involved in the innate immune response to infection. Two of these proteins, IRF-3 and IRF-7, serve as direct transducers of virus-mediated signaling and play critical roles in the induction of type I interferon genes. We have now shown that another factor, IRF-5, participates in the induction of interferon A (IFNA) and IFNB genes and can replace the requirement for IRF-7 in the induction of IFNA genes. We demonstrate that, despite the functional similarity, IRF-5 possesses unique characteristics and does not have a redundant role. Thus, 1) activation of IRF-5 by phosphorylation is virus-specific, and its in vivo association with the IFNA promoter can be detected only in cells infected with NDV, not Sendai virus, while both viruses activate IRF-3 and IRF-7, and 2) NDV infection of IRF-5-overexpressing cells preferentially induced the IFNA8 subtype, while IFNA1 was primarily induced in IRF-7 expressing cells. These data indicate that multiple signaling pathways induced by infection may be differentially recognized by members of the IRF family and modulate transcription of individual IFNA genes in a virus and cell type-specific manner.

Interferon regulatory factor (IRF) genes encode DNAbinding proteins that are involved in the innate immune response to infection. Two of these proteins, IRF-3 and IRF-7, serve as direct transducers of virus-mediated signaling and play critical roles in the induction of type I interferon genes. We have now shown that another factor, IRF-5, participates in the induction of interferon A (IFNA) and IFNB genes and can replace the requirement for IRF-7 in the induction of IFNA genes. We demonstrate that, despite the functional similarity, IRF-5 possesses unique characteristics and does not have a redundant role. Thus, 1

) activation of IRF-5 by phosphorylation is virus-specific, and its in vivo association with the IFNA promoter can be detected only in cells infected with NDV, not Sendai virus, while both viruses activate IRF-3 and IRF-7, and 2) NDV infection of IRF-5overexpressing cells preferentially induced the IFNA8 subtype, while IFNA1 was primarily induced in IRF-7 expressing cells. These data indicate that multiple signaling pathways induced by infection may be differentially recognized by members of the IRF family and modulate transcription of individual IFNA genes in a virus and cell type-specific manner.
As an early response to viral infection, cells produce a spectrum of early inflammatory proteins that control virus growth and limit both lytic and nonlytic viral infections (1). Among these, IFNs 1 play a unique role because they can both activate immune cells and directly inhibit viral replication. In addition to their role in innate resistance against a large number of viruses, human type I IFNs (IFN ␣ and IFN ␤) promote differentiation of dendritic cells (2) and induce polarization of a peripheral CD4ϩ subset of T cells to T helper type 1 terminally differentiated effector cells producing high levels of IFN␥. Type I IFNs also protect CD8ϩ cells from antigen-induced cell death (3). Thus, activation of type I IFNs represents a defense mech-anism that bridges innate and T cell-mediated immunity.
The type I IFNs are encoded by a single IFNB gene and 13 closely related IFNA genes that are all clustered on human chromosome 9p22 (4). Although most cells do not constitutively produce type I IFNs, many DNA and RNA viruses have been found to induce IFN ␣ and ␤ production (5), although the molecular mechanism that leads to the activation of these IFN genes has not yet been clarified. However, it has been demonstrated that the transcription factors of the IRF family play a critical role in the activation of IFN genes (6 -10). The importance of IRFs in the induction of IFN genes is further supported by the observation that several viruses utilize a mechanism that prevents the IFN-induced antiviral response by targeting the activity of IRF transcription factors (11)(12)(13)(14)(15)(16).
To date, nine human IRFs have been identified, and all but two (IRF-5 and -6) have been characterized. All cellular IRFs share a region of homology in the amino terminus encompassing a highly conserved DNA binding motif characterized by five tryptophan repeats. Three of these repeats contact DNA by recognizing the GAAA sequence (17) found in ISRE in the promoters of IFN-stimulated genes or IRF-E present in the VRE in the promoters of IFNA and B genes. Two cellular IRFs (IRF-3 and -7) were shown to be direct transducers of virusmediated signaling and to play a critical role in the induction of IFNA and B genes in infected cells (12, 18 -21) as well as in the induction of some chemokines such as RANTES (regulated on activation normal T cell expressed and secreted) (22). IRF-3 was identified as a constitutively expressed protein present in a variety of cell lines and tissues that is strongly synergistic with the virus-induced expression of IFNA and B genes. In uninfected cells, IRF-3 is found predominantly in the cytoplasm, whereas in infected cells it is phosphorylated, binds to CBP/p300, and accumulates in the nucleus (19,21,(23)(24)(25). Overexpression of IRF-3 induces an antiviral state and expression of IFNB in uninfected cells, suggesting that IRF-3 can act as a primary inducer (7). In human cells, however, expression of IRF-3 is insufficient for the virus-mediated induction of IFNA genes, which also requires IRF-7 (9). Human IRF-7 is expressed in cells of lymphoid origin, and its expression is stimulated by type I IFNs (12,26,27). Silencing of IRF-7 by methylation of CpG clusters in the promoter was observed in several tumor cell lines (28), and these cells were unable to express IFNA when infected. Viral infection stimulated expression of IFNA genes only after reconstitution of IRF-7 by ectopic expression (9). Similarly, infected murine fibroblasts from DKO IRF9 Ϫ/Ϫ mice that do not respond to type I IFN treatment do not express IRF-7; nor does viral infection stimulate the expression of IFNA genes (29).
One of the remaining questions is what determines the profile of individual IFNA genes expressed upon viral infection. Is the cell type-specific expression profile of IFNA subtypes dependent on the relative levels of IRFs in the cell? Expression of IFNA and B genes in infected primary mouse fibroblasts with homozygous deletion of the IRF-3 gene was inhibited but not completely abolished; however, the subtype profile was altered (8). Also, the profile of IFNA genes expressed in infected human 2fTGH/IRF-7 cells was modulated when IRF-3 levels were down-regulated by the IRF-3-specific ribozyme (30). These data suggest that the ratio between relative levels of IRF-3 and IRF-7 affects the expression profile of IFNA subtypes. The contribution of other cellular IRFs to the differential expression of IFNA genes is unknown. Analysis of the relative levels of IRFs expressed in precursor dendritic cells, pDC2, that express high levels of IFN ␣ after virus stimulation (31) has shown that in addition to IRF-3 and IRF-7, other IRFs, such as IRF-5, may be important (32). Since IRF-5 has not been characterized, we have cloned IRF-5 cDNA from an EST cDNA library of dendritic cells and functionally characterized this new protein. Results presented here indicate that IRF-5 can synergize with the virus-mediated induction of IFNA genes and induce expression of these genes in the absence of IRF-7. Furthermore, IRF-5 is the first member of the IRF family that shows differential response to virus, since activation of IRF-5 by phosphorylation and induction of IFNA genes was only detected in NDV-infected and not in Sendai virus-infected cells. The expression profile of IFNA genes in 2fTGH NDV-infected cells expressing IRF-5 or IRF-7 showed distinct differences; in IRF-7-expressing cells, virus preferentially induced IFNA1, but IRF-5-overexpressing cells induced mainly IFNA8.

EXPERIMENTAL PROCEDURES
Cells and Virus-MDBK, HeLa, and Daudi cells were obtained from ATCC, and 2fTGH cells, obtained from G. Stark (Cleveland Clinic Foundation, Cleveland, OH), were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Sendai virus was purchased from Specific Pathogen Free Avian Supply (Preston, CT), and NDV was purchased from ATCC (VR-699). Infections were conducted with 640 hemagglutinin units/100-mm plate (80% confluency) for a given time period. 2fTGH cells constitutively expressing IRF-5 (2fTGH/IRF-5) were generated by co-transfection of 2fTGH cells with N-terminal FLAG-tagged IRF-5-expressing plasmid in which IRF-5 cDNA was under the control of the cytomegalovirus promoter (pCMV-Tag2B.IRF-5) and pSV 2 neo (ratio 10:1, respectively), and transfected cells were selected by growth in G418. Single clones were screened for the expression of IRF-5 by Western blot hybridization with monoclonal anti-FLAG antibody.
Transfections and CAT and SAP Assays-In the transient transfection assay, 2 ϫ 10 6 2fTGH cells were transfected with a constant amount of DNA (5 g/60-mm plate, 8 g/100-mm) by using Superfect transfection reagent (Qiagen). For the CAT and SAP assays, equal amounts (2.5 g) of reporter plasmid and IRF-expressing plasmid were transfected with the ␤-galactosidase expression plasmid (50 -100 ng). The transfected cells were divided 14 h later and infected with Sendai virus or NDV for 16 h. The CAT assay and the SAP assay were per-formed as previously described (35,36). Each experiment was repeated three times. The ␤-galactosidase expression levels were used to normalize the difference in transfection efficiency for both CAT and SAP assays.
Oligonucleotide Pull-down Assay and Western Blot-Doublestranded oligomers corresponding to the IFNA1 and IFNA14 VRE region (base pairs Ϫ110 to Ϫ53) were synthesized and biotin-labeled at the 5Ј-end of the antisense strand. Two micrograms of the doublestranded oligomers were incubated with streptavidin magnetic beads (200 l) (Dynal Inc.) for 1 h in TEN buffer (20 mM Tris, pH 8.0, 1 mM EDTA, 0.1 M NaCl), and the unbound DNA was removed by extensive washing with the same buffer. IRFs from the nuclear and cytoplasmic extracts were bound to DNA on magnetic beads as described previously (28). Beads were then washed, and the bound proteins were analyzed on SDS-polyacrylamide gel electrophoresis. IRFs were identified by Western blot with the respective antibodies as described (9).
Subcellular Localization of GFP-IRF-5 Proteins-2fTGH cells were transfected with GFP-IRF-5-expressing plasmid (5 g) and left uninfected or infected with Sendai virus or NDV. Fourteen hours after transfection, infected cells were divided and seeded in chambered coverglass slides (Nunc, Naperville, IL). Four hours after infection, cells were examined under fluorescence microscope at a wavelength of 507 nm. All pictures presented were recorded under the same magnification (ϫ 20) and exposure time.
Metabolic Labeling of Cellular Proteins with 32 P i -The 2fTGH and 2fTGH/IRF-5 cells (3 ϫ 10 6 ) were grown for 2 h in phosphate-free Dulbecco's modified Eagle's medium supplemented with 2% dialyzed fetal bovine serum. Cells were then left uninfected or were infected with Sendai virus or NDV and labeled with 0.5 mCi/ml of 32 P i (Amersham Pharmacia Biotech) for 2, 4, and 6 h. Samples were collected in lysis buffer (10 mM Nonidet P-40, 0.15 M NaCl, 0.01 M sodium phosphate, pH 7.2, 2 mM EDTA, 50 mM sodium fluoride, 0.2 mM sodium vanadate, protease inhibitor mixture) and left on ice for 30 min. The whole cell extracts (250 g) were used for immunoprecipitation with M2 anti-FLAG antibody, and precipitated proteins were resolved on 7% SDSpolyacrylamide gel electrophoresis (10).
Reverse Transcription PCR Analysis and Antiviral Assay-One microgram of total RNA isolated by the cesium-chloride method was reverse-transcribed to cDNA with oligo(dT) primers in a 30-l reaction. From this mixture of cDNAs, IFNA, IRF-7, IFNB, and ␤-actin cDNA were amplified by PCR as previously described (9). IRF-5 was amplified with the following primers: sense, 5Ј-GCCTTGTTATTGCATGCCAGC; antisense, 5Ј-AGACCAAGCTTTTCAGCCTGG. Individual IFNA subtypes were identified by cloning and subsequent sequencing of the PCR-amplified fragments as described recently (9). A one-way analysis of variance (from the VassarStats Web site) was used to analyze the significance of observed differences in IFNA subtype expression. Levels of biologically active IFN ␣ in the medium were determined by the antiviral assay (5).
Chromatin Immunoprecipitation Assay-2fTGH and 2fTGH/IRF-5 (2 ϫ 10 6 ) cells were transfected with 1 g of IFNA1SAP or IFNA14SAP. At 16 h post-transfection, cells were infected with Sendai virus or NDV for 6 h, and the proteins bound to DNA were cross-linked by the addition of 11% formaldehyde (0.1 M NaCl, 1 mM EDTA, 50 mM HEPES, pH 8.0) to a final concentration of 1% for 30 min at 37°C (20,37). The reaction was stopped by the addition of 0.125 M glycine. The cell pellets were washed with 5 ml of wash buffer (5 mM PIPES, pH 8.0, 85 mM KCl, 0.5% Nonidet P-40, 0.2 mM phenylmethylsulfonyl fluoride), resuspended in sonication buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.0) on ice, and lysed by sonication for 10 s. Samples were diluted 10-fold with dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl, pH 8.0, 167 mM NaCl) and cleared with protein A-Sepharose. Equal amounts of proteins were immunoprecipitated with 1 g of IRF-3 polyclonal or anti-FLAG monoclonal antibody for 4 h at 4°C. Immunocomplexes were extensively washed and treated with RNase A (50 g/ml), 0.5% SDS, and proteinase K (500 g/ml). The cross-linked DNA-protein complexes were reverted by heating at 65°C for 6 h, and the DNA was recovered by phenol-chloroform extraction. DNA purified by precipitation with 2 M ammonium acetate/ethanol was used as a template for PCR amplification with IFNA-SAP-specific primers (sense, 5Ј-AAGTTGGGTAACGCCAGGGT; antisense, 5Ј-ACAGTG-GCGAGGCAGGAATTGATCT) that could amplify both the IFNA1 and IFNA14 VRE.

RESULTS
Isolation of IRF-5 cDNA from Dendritic and B Cells: Expression in Lymphoid Tissues-The sequence of IRF-5 cDNA has been deposited in the human NCBI GenBank TM (U51127), but neither the gene nor the protein has been characterized. In order to determine the function of this protein and its possible role in the induction of early inflammatory genes, specifically type I interferon genes, we screened the Human Genome Sciences expressed sequence tag cDNA libraries for homology to IRF-5. The IRF-5 cDNA was identified in a dendritic cell library. By comparing the IRF-5 cDNA sequence obtained from the dendritic cell library with the sequence of human IRF-5 cDNA deposited in GenBank TM , we found that the dendritic IRF-5 cDNA had an internal deletion of 48 nucleotides (positions 583-631). To determine whether the deletion was unique to this cDNA, we also isolated IRF-5 cDNA from B cells (Daudi) and found that the isolated IRF-5 cDNAs contained the identical deletion and encoded a predicted protein of 488 amino acids. Similar to other IRF family members, IRF-5 contains five conserved tryptophan residues in the amino terminus. Examination of the predicted amino acid sequence of IRF-5 revealed several interesting features. The open reading frame encodes a polypeptide with a predicted molecular mass of about 60 kDa that contains two putative nuclear localization signals, one in the N-terminal domain and one in the C-terminal domain. These two nuclear localization signals, unique to IRF-5 and IRF-6, are not present in the other IRFs. In addition, 3Ј of the DNA-binding domain, IRF-5 contains a string of eight glutamic acid residues of yet unknown function. The full-length IRF-5 sequence shows close homology to IRF-6 (78%), IRF-7 (59%), IRF-4, IRF-8, and IRF-9 (58%).
Analysis of a data base of sequence-tagged sites revealed two independent tags, SHGC-63648 and sWSS3578 (38) that are overlapped by the IRF-5 coding region. Both tags map between chromosomal markers D7S504 and D7S530, a region corresponding cytogenetically to 7q32. No other IRF family members have been mapped to this genomic region, which is reported to contain a cluster of imprinted genes (39). Whether IRF-5 is imprinted or is associated with any disorders associated with 7q32 remains to be established.
To determine in which cells IRF-5 is expressed, we examined the tissue distribution of IRF-5 mRNA and found that constitutive expression was primarily detected in lymphoid tissue and peripheral blood lymphocytes. Low levels were also detected in skeletal muscle and prostate; but no expression of IRF-5 was detected in thymus. In all of the tissues that expressed IRF-5, two transcripts of size 2.4 and 3.4 kilobases were detected (Fig. 1A). By Northern blot, a single IRF-5 transcript was detected in SW480 colorectal cancer, A549 lung carcinoma, and G361 melanoma cell lines (Fig. 1B). When RT-PCR was used for the detection of IRF-5 transcripts, most cell lines examined revealed low levels of IRF-5 expression that were enhanced by viral infection and IFN␣ treatment. 2

IRF-5 Contains a Transactivation Domain and Activates Promoters Containing IFNA1, IFNB, and ISRE in a Transient
Expression Assay-In order to determine the transcription potential of IRF-5, plasmid expressing GAL4/IRF-5 was cotransfected together with GAL4CAT reporter plasmid into 2fTGH cells. As a positive control, GAL4/IRF-1 was co-transfected or GAL4/IRF-2 was used as a negative control, with reporter plasmids. As shown in Fig. 2A, the transactivation potential of IRF-5 was similar to that of IRF-1 in this assay, while the increase by GAL4/IRF-2 was small (3-fold increase).
The role of IRF family members in the activation of type I IFN genes has been clearly established. We therefore examined the ability of IRF-5 to transactivate the ISRE-containing promoter as well as type I IFN promoters in a transient expression assay. As shown in Fig. 2B, overexpression of IRF-5 activated (4 -5-fold enhancement) the promoter containing four copies of ISRE in infected cells, and the activation was slightly higher in cells infected with NDV than with Sendai virus. This promoter was also activated efficiently by IRF-3 and to a lesser degree by IRF-1 and IRF-7, but activation by these IRFs was slightly more effective in cells infected with Sendai virus than with NDV. Overexpression of IRF-5 also activated a reporter plasmid containing SAP under the regulation of IFNA1 VRE or IFNB VRE (Fig. 2, C and D, respectively). Interestingly, significant activation of the IFNA1 promoter (2-3-fold enhancement) by IRF-5 was only observed in NDV infected cells while activation in Sendai virus-infected cells was negligible (less than 2-fold). This difference was not due to the inability of Sendai virus to induce this promoter because, in cells transfected with IRF-1, IRF-3, or IRF-7, activity of the IFNA1 promoter was stimulated both by Sendai virus and NDV. Thus, the higher inducing ability of NDV compared with Sendai virus was seen only in cells transfected with IRF-5. The IFNB VRE was stimulated by IRF-5 to about the same level as by IRF-1 in infected cells ( IRF-7 were more effective stimulators (5-6-fold enhancement). Although the IFNB promoter was stimulated by IRF-5 more effectively in NDV-than Sendai virus-infected cells, this difference was too small to be significant (Fig. 2D).
Taken together, these results concur with the previous observation that in a transient transfection assay both NDV and Sendai virus infection synergize with IRF-3 and IRF-7 in the activation of IFNA1 and IFNB promoters. However, the activation of the IFNA1 promoter by IRF-5 seems to be more efficient in cells infected with NDV than with Sendai virus.
NDV Infection Induces Nuclear Accumulation of IRF-5 and Phosphorylation-We have previously shown that IRF-3 stimulates transcription of only the IFNA1 promoter in infected cells, whereas IRF-7 also enhances expression of IFNA2, A4, and A14 VRE (9). To determine whether the IRF-5-mediated enhancement is also limited to specific IFNA promoters, we cotransfected the IRF-5 expression plasmid with IFNA1, IFNA2, IFNA4, and IFNA14 SAP reporter plasmids into 2fTGH cells. As shown in Fig. 3A, IRF-5 enhanced the expression of IFNA1, IFNA2, and IFNA14, but not IFNA4 in infected cells. In all of these transfections, IRF-5 stimulated the expression of IFNA1, A2, and A14 VRE also in the absence of viral infection. Yet, viral infection further enhanced the transactivation ability of IRF-5, and this activation was slightly more effective in NDV-infected cells than in Sendai virus-infected cells. The ability of IRF-5 to activate these promoters in the absence of viral infection to significantly higher levels than IRF-3 or IRF-7 is a unique property of this transcription factor.
The selective stimulation of IFNA VRE by IRF-3 and IRF-7 could be related to their ability to bind to the respective VREs (9). Therefore, we used the oligonucleotide pull-down assay to examine whether IRF-5 in infected cells could also bind to the IFNA VRE. For this analysis, the biotin-labeled IFNA1 or IFNA14 VRE coupled to magnetic beads was incubated with nuclear extract from uninfected and Sendai virus-or NDVinfected 2fTGH/IRF-5 cells. The specifically bound proteins were eluted, separated on SDS gels, and identified by Western blotting. As shown in Fig. 3B, IRF-5 was detected in nuclear extracts of both uninfected and NDV-infected cells. However, although the relative levels of IRF-5 in nuclear extracts of uninfected cells were slightly higher than in Sendai virusinfected cells, NDV infection significantly increased the levels of IRF-5 in the nucleus. When nuclear extracts from uninfected or Sendai virus-infected cells were incubated with IFNA1 or A14 VRE, only low levels of IRF-5 binding were detected. In contrast, much higher levels of IRF-5 were bound to these two VREs from the nuclear extracts of NDV-infected cells.
These data indicate that NDV but not Sendai virus infection increases the nuclear localization of IRF-5 and consequently binding of IRF-5 to the VRE oligodeoxynucleotide. The relative levels of endogenous IRF-3 in the nuclear extracts of these cells were significantly increased by both Sendai virus and NDV infection. Furthermore, binding of IRF-3 from nuclear extracts of both Sendai virus-and NDV-infected cells to the IFNA1 VRE was greater than from nuclear extracts of uninfected cells, whereas no significant binding to IFNA14 VRE was observed. Finally, we determined the binding of IRF-7 to these two VREs, since IRF-7 was shown to activate both of these VREs in a transient transfection assay. Although the relative levels of IRF-7 in nuclear extracts of IRF-7-transfected 2fTGH cells were quite low, IRF-7 was detected in the nucleus of both Sendai virus-and NDV-infected cells. Accordingly, there was no detectable binding of IRF-7 from nuclear extracts of uninfected cells to A1 or A14 VRE, yet equivalent levels of IRF-7 were bound to these two VREs from nuclear extracts of Sendai virus-and NDV-infected cells. These results show that NDV but not Sendai virus induces the nuclear accumulation of IRF-5 and that IRF-5 binds to both A1 and A14 VRE. The cellular localization of IRF-5 in infected and uninfected cells was also examined by transient transfection with plasmid encoding the GFP-IRF-5 fusion protein. In the transfected cells, GFP-IRF-5 was detected only in the cytoplasm, and infection with Sendai virus did not result in nuclear accumulation. In contrast, the distribution of GFP-IRF-5 in NDV-infected cells was altered and IRF-5 was detected in both the nucleus and cytoplasm (Fig. 4A). These results correlate well with Western blot analyses of nuclear and cytoplasmic lysates from infected and uninfected 2fTGH cells (Fig. 4B).
It was previously shown that the virus-mediated phosphorylation of IRF-3 on serines 385, 386, 396, and 398 is critical for its nuclear accumulation and interaction with the transcription coactivator CBP/p300 (19,21,(23)(24)(25). In a similar manner, mutations of serines 483 and 484 of IRF-7H abolished virusmediated phosphorylation and nuclear translocation (10,40). IRF-5 does not contain the same conserved cluster of serine residues in this region as IRF-3 and IRF-7 (Fig. 5A); however, the increased accumulation of IRF-5 in the nucleus of NDVinfected cells suggested that this protein is also post-transcriptionally modified. To determine whether IRF-5 is phosphorylated in infected cells, 2fTGH cells were transfected with FLAGtagged IRF-5 and left uninfected or infected with Sendai virus or NDV for 2, 4, and 6 h. At the onset of infection, transfected cells were labeled with 32 P i for the appropriate time periods, and then the labeled IRF-5 was precipitated from cell lysates with anti-FLAG antibodies. As shown in Fig. 5B, low levels of 32 P-labeled IRF-5 were detected in uninfected cells, but NDV infection significantly enhanced the levels of phosphorylated IRF-5 within 2 h postinfection. In contrast, much lower levels of phosphorylation of IRF-5 by Sendai virus were observed and were not detected until 6 h postinfection. These data indicate that the kinetics of IRF-5 phosphorylation by NDV and Sendai virus are not identical. The relative levels of IRF-5 present in the cell lysate were not affected by viral infection during this time period as shown by Western blotting. Also, 32 P-labeled IRF-5 was not detected in untransfected cells, indicating that the precipitation with FLAG antibody was specific for IRF-5. Under the same conditions, IRF-3 and IRF-7 were effectively phosphorylated in Sendai virus-infected cells, 2 as demonstrated previously (10,19,21,40).

IRF-5 Stimulates Expression of Endogenous IFNA Genes and Binds to IFNA Promoters in Vivo in NDV-infected Cells-Since
transient expression assays showed that IRF-5 is an effective activator of IFNA promoters in infected cells, we examined whether overexpression of IRF-5 in 2fTGH cells that express IRF-3, but not IRF-7, would stimulate expression of type I IFN genes after virus infection. For these experiments, we generated a 2fTGH cell line that constitutively overexpressed active N-terminal FLAG-tagged IRF-5 (2fTGH/IRF-5). As shown in Fig. 6A, IRF-5 was expressed in transfected cells both on the RNA and protein level. In 2fTGH cells, infection with NDV resulted only in the expression of the IFNB gene, and no IFNA transcripts or biologically active IFN␣ were detected. In contrast, in 2fTGH/IRF-5 cells infected with NDV, we detected both higher levels of IFNB transcripts and the presence of IFNA transcripts. These cells also produced high levels of biologically active IFN␣ (1500 units/ml). Notably, when this stable 2fTGH/IRF-5 cell line was infected with Sendai virus, no IFNA gene expression or biologically active IFN␣ was detected.
These data suggest that IRF-5 plays a role similar to IRF-7 in its ability to restore the expression of IFNA genes in infected cells. Although IRF-7 mediated the induction of IFNA genes both in Sendai virus and NDV-infected cells, activation of endogenous IFNA genes by IRF-5 seemed to be limited to NDVinfected cells. Furthermore, to determine whether IRF-5 was a primary inducer of IFNA genes in 2fTGH cells, we examined whether IRF-5 up-regulated the IRF-7 promoter activity and/or induced expression of the IRF-7 gene. However, we could not detect the IRF-7 transcript in 2fTGH cells overexpressing IRF-5 either before or after infection, and transcription activity of the IRF-7 promoter was not stimulated by IRF-5. 2 To determine whether activation of individual IFNA subtypes is a distinct feature of NDV-infected 2fTGH cells expressing IRF-5 or IRF-7, we examined the IFNA subtypes expressed in these cells and the relative levels of their expression. Table  I shows that the IFNA subtypes expressed in NDV-infected IRF-5-or IRF-7-expressing cells are not identical. In NDVinfected, 2fTGH/IRF-5 cells, nine IFNA subtypes were expressed; IFNA8 was the major one (45%), and the other IFNA subtypes were expressed at significantly lower frequency (2-13%). In contrast, in 2fTGH/IRF-7 cells, NDV infection resulted in the induction of six IFNA subtypes, and IFNA1 was expressed at the highest level (40%). Although we did not detect expression of IFNA8, these cells efficiently expressed the IFNA17 (17%) subtype that was not identified in the IRF-5overexpressing cells. These data indicate that there is little redundancy in the function of IRF-5 and IRF-7 and that these factors not only respond to different viruses but also induce a unique set of IFNA genes.
Since the induction of endogenous IFNA in the IRF-5-expressing cells occurred only in NDV-infected and not in Sendai virus-infected cells, we would expect that assembly of IRF-5 black, NDV-infected cells. The SAP levels were normalized to the constant level of ␤-galactosidase expressed. B, binding of nuclear IRFs from infected and uninfected 2fTGH cells to IFNA1 and IFNA14 VRE. 2fTGH cells were transfected with FLAG-tagged IRF-5 or untagged IRF-7, and nuclear extracts were prepared from uninfected or infected cells. The relative levels of transfected IRF-5, IRF-7 and endogenous IRF-3 in the nuclear extracts (25 g) were estimated by Western blot hybridization. The same extracts (250 g) were used for the oligonucleotide pull-down assay as described under "Experimental Procedures," and the levels of IRF-3, IRF-5, and IRF-7 bound to the respective VREs were detected by immunoblotting.
would take place preferentially on the endogenous IFNA promoter in NDV-infected cells. To examine whether this assumption is correct, we performed the chromatin immunoprecipitation assay to determine the binding of IRF-5 to IFNA1 and A14 promoters in vivo in response to NDV and Sendai virus infection. As a control, we determined whether IRF-3 could bind to either of these two promoters because IRF-3 can be activated by phosphorylation in both NDV-and Sendai virus-infected cells. The 2fTGH/IRF-5-expressing cells were transfected either with IFNA1-or IFNA14 SAP-expressing plasmid and left uninfected or infected with NDV or Sendai virus for 6 h. Pro-teins were then cross-linked to DNA, and the protein-DNA complexes were precipitated with either anti-IRF-3 polyclonal or M2 anti-FLAG antibody (20,37). The DNA present in the precipitates was then amplified by PCR with primers detecting the IFNA1 and A14 VRE.
As shown in Fig. 6B, the fragment corresponding to the IFNA1 VRE was amplified from DNA immunoprecipitated by IRF-3 antibodies both from Sendai virus-and NDV-infected cells. No amplification was seen in precipitates from uninfected cells. Amplification of the IFNA14 promoter from these precipitates was extremely low. These results correlate with the  and underlined). B, specific phosphorylation of IRF-5 by NDV infection. 2fTGH/IRF-5 cells were left uninfected or were infected with Sendai virus or NDV for 2, 4, and 6 h (lanes 4 -12). 32 P labeling was done for the entire time of infection, and 32 P-labeled IRF-5 was precipitated from cell lysates with anti-FLAG antibody. The proteins were separated by electrophoresis in 7% SDS gel, dried, and then exposed to a PhosphorImager screen (Molecular Dynamics, Inc., Sunnyvale, CA). The top panel shows levels of radiolabeled IRF-5, and the bottom panel shows the levels of IRF-5 protein in cell lysates detected by immunoblotting with anti-FLAG antibody. As a control, 2fTGH cells transfected with untagged IRF-5 were left uninfected or infected with Sendai virus or NDV for 2 h (lanes 1-3).
previous observation that IRF-3 does not bind effectively to the VRE of IFNA14 (9). When the precipitation was done with anti-FLAG antibodies to detect IRF-5, low levels of amplification of the IFNA1 promoter were observed in Sendai virusinfected cells, but a band of greater intensity was amplified from immunoprecipitates of NDV-infected cells. No fragment was amplified from the immunoprecipitates of uninfected cells. The IFNA14 promoter was amplified from FLAG precipitates of both uninfected and Sendai virus-infected cells at low levels, but a much stronger amplification was observed from NDVinfected cells. To determine whether there was some nonspecific amplification from the FLAG immunoprecipitates, we performed chromatin immunoprecipitation on lysates from 2fTGH cells transfected with untagged IRF-5 and the respective IF-NASAP plasmids. However, we were unable to amplify any IFNA promoter-specific bands from FLAG immunoprecipitates either before or after infection. We conclude from these data that IRF-5 associates with the VRE of the IFNA1 promoter much more effectively in NDV-than Sendai virus-infected cells, while IRF-3 associates with this VRE effectively both in Sendai virus-and NDV-infected cells. The low level of amplification of IFNA14 VRE from FLAG precipitates of uninfected cells was surprising and indicates that low levels of IRF-5 can bind to this promoter even before viral infection. Nevertheless, the levels of amplification were substantially increased after NDV infection. Of note, in transient transfection assays, IRF-5 was also able to activate the IFNA14 reporter plasmid in uninfected cells. Since 2fTGH cells do not express IRF-7, we could not determine in these experiments whether binding of IRF-5 and IRF-7 is mutually exclusive or cooperative. These experiments are being performed.
IRF-3 Associates with IRF-5 and Enhances Recruitment of IRF-5 to the IFNA Promoter in NDV-infected Cells-It has been shown that IRF-3 associates with IRF-7 and that the IRF-3/ IRF-7 heterodimers are biologically active (10,41). We therefore asked whether IRF-5 could also interact with IRF-3 and whether the decrease of IRF-3 levels by using an IRF-3-targeted ribozyme would affect IFNA and IFNB expression in NDV-infected 2fTGH/IRF-5 cells. We have recently shown that the hammerhead ribozyme targeted to IRF-3 mRNA specifically down-modulated IRF-3 levels in transfected cells, although it did not affect levels of IRF-2 or IRF-7 (30). As shown in Fig. 7, the expression of IRF-3 was significantly down-modulated in cells transfected with the IRF-3 ribozyme. Also, the levels of IFNB and IFNA transcripts induced by Sendai virus or NDV in 2fTGH cells were significantly lower in cells cotransfected with IRF-5 in the presence of IRF-3 ribozyme than in its absence. These data indicate that both IRF-3 and IRF-5 are required for the effective induction of IFNA gene transcription in infected cells.
When the direct association between IRF-5 and IRF-3 was analyzed by the glutathione S-transferase pull-down assay, we found that only low levels of IRF-5 from cellular lysates of uninfected or Sendai virus-infected cells bound specifically to glutathione S-transferase-IRF-3. However, IRF-5 binding to glutathione S-transferase-IRF-3 was greatly enhanced in lysates from NDV-infected cells. 2 These data suggest that IRF-5 can form heterodimers with IRF-3 and that formation of these heterodimers is enhanced when IRF-5 is phosphorylated.

DISCUSSION
This study brings up several novel features regarding the molecular mechanism of IFNA gene induction in infected cells.
Although it was previously shown that IRF-7 plays a critical role in the induction of IFNA genes in both mouse and human cells (9,12,29,42), we have now shown that human IRF-5 can reconstitute the virus-mediated induction of IFNA genes in the absence of IRF-7 and that this activation is virus-specific. Furthermore, the functions of IRF-5 and IRF-7 do not appear to be redundant, since they each result in the induction of distinct IFNA subtypes.
The human IRF-5 cDNAs identified encode a protein of 488 amino acids. Of note, the IRF-5 cDNAs that we have identified in both dendritic cells and human B cells (Daudi) contain a 48-nucleotide internal deletion in comparison with the IRF-5 cDNA deposited in GenBank TM . The expression of IRF-5 is limited to lymphoid organs, but no IRF-5 transcripts were detected in thymus. Thus, the expression of IRF-5 is more restricted than the expression of IRF-7, which is also expressed in thymus and appendix (12), and IRF-3, expressed in all human tissues examined (18). Type I IFN, which stimulates transcription of IRF-7 (12,28), also increases levels of IRF-5 transcripts, 2 but it does not modulate expression of IRF-3 (18).
The presence of IRF-7 is required for induction of IFNA genes. Human fibrosarcoma cells (2fTGH) express IRF-3 but do not express IRF-7, and Sendai virus infection leads only to the induction of IFNB and not IFNA genes. Nevertheless, reconstitution of IRF-7 synthesis in these cells rescued expression of IFNA upon infection with either Sendai virus or NDV, and the predominant subtype induced by both of these viruses was The levels of IRF-5 in cell lysates were determined by immunoblotting with anti-FLAG antibody, and the levels of biologically active IFN␣ synthesized in these cells determined by antiviral assay are shown at the bottom. B, in vivo binding of IRF-5 to IFNA1 and IFNA14 subtypes analyzed by chromatin immunoprecipitation. 2fTGH cells were cotransfected with FLAG-tagged IRF-5 and either IFNA1SAP or IFNA14SAP plasmids. Cellular DNA and proteins were cross-linked and subjected to the chromatin immunoprecipitation assay as described under "Experimental Procedures." The immunoprecipitations were performed with either IRF-3 monoclonal antibody or anti-FLAG antibody to precipitate FLAG-tagged IRF-5. DNA recovered from chromatin immunoprecipitation by heating was amplified using primers specific for IFNA1 and A14 VRE. Lanes 1-3 depict amplification of IFNA1 VRE, and lanes 4 -6 show amplification of IFNA14 VRE. IFNA1 (9). The constitutive expression of IRF-5 in 2fTGH cells is very low and IRF-5 transcripts can only be detected by RT-PCR analyses. However, overexpression of IRF-5 from an ectopic IRF-5-expressing plasmid rescued induction of IFNA genes, but only in NDV-infected and not in Sendai virus-infected cells. Notably, IRF-5 neither stimulates the transcription activity of the IRF-7 promoter nor induces the expression of IRF-7 in these cells in a transient expression assay. 2 Furthermore, although IFNA1 was the major subtype induced by NDV in IRF-7-expressing cells, NDV infection of IRF-5-expressing cells led to the expression of IFNA8 as the major subtype, further confirming that the induction of IFNA genes was not mediated by IRF-7. It has recently been shown that IFNA8 has higher antiviral activity than the other IFNA subtypes, which may account for the high levels of IFN␣ detected in the 2fTGH/IRF-5 cells (43,44). It was also observed that in mouse fibroblasts from DKO STAT1 Ϫ/Ϫ or DKO IRF-9 Ϫ/Ϫ mice defective in type I IFN signaling, induction of IFNA genes was impaired due to the absence (low levels) of IRF-7 expression (29,42). Whether these mice are also unable to express IRF-5 remains to be determined. However, we were not able to detect IRF-5 expression in mutant 2fTGH (U2A) cells that are unable to respond to type I IFN due to a defect in IRF-9, while we could detect very low levels of IRF-5 transcripts in 2fTGH cells. 2 In infected cells, IRF-3 and IRF-7 are activated by phosphorylation on the carboxyl-terminal serines 385/386 and 483/484, respectively, and then are retained in the nucleus, where IRF-3 binds to the transcription coactivator CBP/p300 (10, 19, 21, 24). Both of these factors were found to be phosphorylated by a variety of viruses including Sendai virus and NDV (10,40,45,46) and IRF-3 is also phosphorylated by double-stranded RNA (23). Therefore, the observation that only NDV and not Sendai virus infection induces phosphorylation and activation of IRF-5 is rather surprising and novel. The kinase(s) that phosphorylates IRF-3 and IRF-7 in infected cells has not been identified yet. However, two observations suggest that a distinct signaling pathway may activate IRF-5 and IRF3 or IRF-7 in NDVinfected cells. 1) The configuration of the two serine clusters found in IRF-3 and IRF-7 is not preserved in the carboxylterminal region of IRF-5; and 2) IRF-5 phosphorylation can be detected within 2 h postinfection, while the phosphorylation of IRF-3 and IRF-7 and nuclear translocation are barely detectable before 6 h postinfection. Additional studies are necessary to determine which other viruses can induce the phosphorylation and activation of IRF-5. The phosphorylation of IRF-3 and IRF-7 by UV and stress-inducing agents was also recently observed (47,48). Whether the same agents induce phosphorylation of IRF-5 will be of interest.
Genetic evidence suggests that a quantitative response of type I IFN induction in mice is modulated by several autosomal, virus-specific loci. These were designated as If loci and have high and low producer alleles. It was shown that the response to NDV and Sendai virus is governed by two independently segregated loci, If-1 and If-3/4 (49). Whether there is any relationship between these loci and the IRFs is not known, but it is of interest that the human IRF-1, IRF-3, IRF-7, and IRF-5 are also not closely linked and are located on different chromosomes, 5q23, 19q13, 11p15, 3 and 7q32, respectively (50,51). It is interesting to note that both IRF-7 and IRF-5 are present in the chromosomal region that contains imprinted genes, and we have recently demonstrated that in some cancer cell lines IRF-7 expression is silenced by methylation (28).
It is likely that all of the transcription factors that bind to the VRE of IFNA genes have not yet been identified (33), although binding of activated IRF-3 and IRF-7 from the lysates of infected cells to IFNA promoters has been demonstrated by the DNA pull-down assay (10). Both IRF-3 and IRF-7 were also shown to form homodimers and heterodimers (10,40,41). The observation that the virus-mediated activation of IFNA gene expression in IRF-7-expressing cells was substantially inhibited in cells that had decreased levels of IRF-3 (30) or in DKO IRF-3 Ϫ/Ϫ fibroblasts suggests that the IRF-3/IRF-7 heterodimer is functional. Interestingly, the down-regulation of IRF-3 levels also resulted in an alteration of the expressed 3 P. A. Moore and P. M. Pitha, unpublished results. a Values not significantly different between IRF-5 and IRF-7 based on analysis of variance statistical analysis. F values from these analyses are presented. The null hypothesis (indicating no significant difference in the two groups) is true if the F value is less than or equal to 1. b Values in boldface type illustrate the major subtype induced by NDV infection in IRF-5-or IRF-7-expressing 2fTGH cells. profile of IFNA subtypes, indicating either that the homodimers have different affinities for the individual IFNA promoters or that binding of the homodimers and heterodimers alters the chromatin structure around the IFN loci. The fact that distinct major IFNA subtypes are expressed in IRF-5-and IRF-7-expressing cells indicates that IRF-3 targets different promoters when in complex with IRF-5 or IRF-7. The analyses of the relative levels of IRF factors in cell lines producing high levels of IFN␣, such as B cells (Namalva) or precursors of dendritic cells (pDC2), indicate that these cells express most of the IRFs constitutively at relatively high levels (32). Thus, the two-step expression of IFNA genes (42) that requires the autocrine and paracrine loop and transcriptional up-regulation of IRF-7 may not be necessary in all cells.
The observation that viral infection results in the activation of multiple signaling pathways and transcription factors was first established with the IFNB promoter, activation of which is mediated by at least two virus-induced signaling pathways. One leads to the activation of NF-B (52), and the other, not yet well characterized, leads to the activation of IRF-3 and IRF-7 (24,45). There is also a requirement for cooperative binding of multiple transcription factors forming a multicomponent complex-enhanceosome for the initiation of IFNB transcription (53). It seems that the cell-specific activation of IFNA promoters requires binding of multiple transcription factors of the IRF family and possibly other transcription factors (54). As suggested in the present study, some of these factors may be activated by separate terminal virus-induced signaling pathways. The observed cell-specific variation of the expressed individual IFNA subtypes may then be related to the relative levels of IRF expressed in different cell types as well as their distinct virus-specific activation. Whether the cell type-restricted expression of IRF-5 together with its selective activation by a specific virus(es) and the resulting induction of an interferon subtype with high antiviral activity reflects the need to protect these cells against the inducing virus remains to be established. The change in the ratio of activated IRFs in infected cells, which affects their self-interaction as well as interaction with other activated IRFs, could then be a major determinant of the tissue-and virus-specific expression patterns of IFNA genes.