DNA binding and transcription activation by chicken interferon regulatory factor-3 (chIRF-3)

Abstract

Interferon regulatory factors (IRFs) are a family of transcription factors involved in the cellular response to interferons and viral infection. Previously we isolated an IRF from a chicken embryonic liver cDNA library. Using a PCR-based binding site selection assay, we have characterised the binding specificity of chIRF-3. The optimal binding site (OBS) fits within the consensus interferon-stimulated response element (ISRE) but the specificity of chIRF-3 binding allows less variation in nucleotides outside the core IRF-binding sequence. A comparison of IRF-1 and chIRF-3 binding to ISREs in electrophoretic mobility shift assays confirmed that the binding specificity of chIRF-3 was clearly distinguishable from IRF-1. The selection assay also showed that chIRF-3 is capable of binding an inverted repeat of two half OBSs separated by 10–13 nt. ChIRF-3 appears to bind both the OBS and inverted repeat sites as a dimer with the protein–protein interaction requiring a domain between amino acids 117 and 311. In transfection experiments expression of chIRF-3 strongly activated a promoter containing the OBS. The activation domain was mapped to between amino acids 138 and 221 and a domain inhibitory to activation was also mapped to the C-terminal portion of chIRF-3.

INTRODUCTION

Interferon regulatory factors (IRFs) are a family of transcription factors that were originally isolated as positive and negative regulators of interferon (IFN) and IFN-responsive genes. More recently, members of the family have been found to have additional functions (reviewed in 1–3). IRF proteins appear to be involved in the regulation of cell growth, haematopoietic cell development and, under some conditions, apoptosis. To date, nine mammalian IRF proteins have been identified: IRF-1, IRF-2, IRF-3, IRF-4/Pip/ISCAT/LSIRF, IRF-5, IRF-6, IRF-7, ISGF3γ/p48 and ICSBP (4–12). Four viral IRFs have also been found in the genome of the human Herpes virus 8/Karposi sarcoma Herpes virus (13). In addition, we have cloned a novel chicken IRF, chIRF-3 (formerly termed cIRF-3) (14).

Members of the IRF family are characterised by a conserved N-terminal 115 amino acid DNA-binding domain (DBD). The signature motif of the DBD is a tryphophan repeat consisting of five residues spaced at 10–18 amino acid intervals (1). The crystal structures of the DBDs of IRF-1 and IRF-2 bound to IRF-binding elements have been determined and show that the IRF DBD forms a winged helix–turn–helix motif (15,16).

All the IRF proteins bind similar DNA sequences. Using a PCR-based binding site selection assay, Tanaka et al. identified a consensus IRF-1- and IRF-2-binding site, termed the IRF element (IRF-E), (G)AAA^G/_C^T/_CGAAA^G/_C^T/_C (17). The interferon-stimulated response element (ISRE), ^A/_GNGAAANNGAAACT, was derived from elements identified in the promoters of IFN-α-stimulated genes (18). However, a recent survey of genes containing IRF-binding sites shows that, although most sites fit the above consensus, some vary in both the number of GAAA repeats present (from one to three) and in the number of nucleotides between the repeats (from one to three) (16). Consequently, the manner in which IRF proteins recognise and contact specific DNA sequences is of growing interest and is potentially important in understanding the specific functions of the individual IRF family members.

When initially isolated, chIRF-3 was the first non-mammalian IRF protein to be identified. The chIRF-3 cDNA sequence encodes the N-terminal DBD characteristic of IRF proteins. In addition, a C-proximal region shares significant similarities with an ∼180 amino acid domain, termed the IRF association domain (IAD), that is involved in IRF protein–protein associations between IRF family members and with other transcription factors (14). More recently, the chicken orthologues of IRF-1, IRF-2 and ICSBP have been cloned and shown to be highly conserved between chicken and humans (19–21). They are 89, 97 and 97% identical within their DBDs and 59, 84 and 73% identical overall, respectively. ChIRF-3 was cloned prior to identification of the mammalian IRF, now designated IRF-3 (6). Human IRF-3 and chIRF-3 are 46% identical in their DBDs and 33% identical overall. The most closely related mammalian member of the family is IRF-7 (10). IRF-7 and chIRF-3 are 52% identical in their DBDs and 42% identical overall. Thus, it remains unclear whether a more highly conserved orthologue of chIRF-3 exists.

Previously we showed that the chIRF-3 mRNA is rapidly and transiently induced to high levels by dsRNA in the chicken hepatoma cell line LMH2A and, using EMSAs, we demonstrated that a λgt11 fusion protein, which contained the chIRF-3 DBD, was capable of binding a known ISRE from the promoter of the chicken Mx gene (14). More recently, we have shown that chIRF-3 expression is also rapidly and transiently induced by both type I and type II chicken IFNs (22).

In this report we have used a PCR-based binding site selection assay to identify sequences to which the chIRF-3 protein prefers to bind. We also demonstrate that chIRF-3 is a strong activator of transcription. We have localised the activation domain to a region between amino acids 138 and 221 and have detected an inhibitory domain near the C-terminal end of the protein. However, intact chIRF-3 is a strong activator of transcription even when this domain is present.

MATERIALS AND METHODS

Binding site selection

Binding site selection was carried out essentially as described (23). Briefly, the probe consisted of a random 24 base sequence which was flanked by PCR primer sequences incorporating on one side EcoRI, NdeI and BamHI restriction sites and BglII, XhoI and HindIII sites on the other side. Each 25 µl binding reaction contained 2 µl of in vitro translated chIRF-3 in binding buffer [20 mM Tris, pH 7.5, 50 mM NaCl, 1 mM EDTA, 5% glycerol, 5 mM dithiothreitol (DTT), 1 µg bovine serum albumin (BSA) and 100 ng poly(dI:dC)-poly(dI:dC)]. After 1 h at room temperature, 6 µl of protein A affinity-purified antisera was added and the incubation continued, on ice, for a further 1.5 h. The immune complexes were then incubated for 2 h at 4°C with protein A–Sepharose (Amersham Pharmacia Biotech) in binding buffer. The protein A–Sepharose was washed twice with ice-cold binding buffer and DNA eluted by heating at 45°C for 5 min in elution buffer (50 mM Tris, pH 7.5, 0.1 M sodium acetate, 5 mM EDTA, 0.5% w/v SDS). After phenol/chloroform extraction and precipitation, the selected DNA was used to prime a PCR reaction with the primers described above. A total of seven selection/PCR cycles were performed with the amount of poly(dI:dC)-poly(dI:dC) increasing in each cycle up to a maximum of 1 µg in the final binding mix. The selected oligonucleotides were cloned into a pBluescript vector (Stratagene) using the HindIII and EcoRI restriction sites and sequenced using Sequenase version 2.0 (US Biochemicals).

chIRF-3 fusion proteins and antibodies

Rabbit polyclonal antibodies were raised against a fusion protein consisting of maltose-binding protein (vector pmal-c2; NEB) and chIRF-3 cDNA clone 12A (encoding amino acids 9–243) (14). Antibodies were affinity purified on a protein A–Sepharose column (Amersham Pharmacia Biotech) (24).

Recombinant full-length chIRF-3 protein (amino acids 1–491) was expressed in Escherichia coli as an intein fusion protein with a chitin-binding tag (vector pCYB1, IMPACT System; NEB). Recombinant protein without the chitin-binding tag was isolated according to the manufacturer’s directions, dialysed against cleavage buffer containing 10% glycerol and 1 mM DTT and concentrated on a Centricon-30 spin column (Amicon). After determining the protein concentration, aliquots of the protein were stored at –70°C.

Electrophoretic mobility shift assay (EMSA)

Double-stranded oligonucleotide probes for EMSAs were ³²P-labelled by fill-in reaction with Klenow fragment. Where a more stringent requirement for equivalent specific activities between probes was necessary, as in the comparison of the mutated binding sites, single-stranded oligonucleotides were annealled to a short complementary oligonucleotide previously ³²P-labelled with T4 polynucleotide kinase. The remainder of the second strand was filled-in with Klenow fragment and the double-stranded probes isolated from a 15% polyacrylamide gel (19:1) in 0.5× TBE.

Binding reactions (25 µl) contained binding buffer, 1 µg BSA, 100–500 ng poly(dI:dC)-poly(dI:dC), 1 or 2 µl of reticulocyte lysate-generated chIRF-3 or 20 ng recombinant chIRF-3 protein and 3–5 fmol (∼20 000 c.p.m.) of radiolabelled binding site. After incubating the binding reaction for 1 h at room temperature, the DNA–protein complexes were resolved on a 6% polyacrylamide gel (29:1) in 0.25× TBE. The unfixed gels were dried prior to autoradiography. Radioactivity in the complexes was quantified using an InstantImager (Packard Instrument Co.).

Construction of expression vectors

For chimeric GAL4/chIRF-3 expression vectors the yeast GAL4 DBD from pBD-GAL4 Cam (Stratagene) was cloned into the mammalian expression vector pcDNA1/Amp (Invitrogen) to generate clone pGAL4. Three regions of chIRF-3 encompassing amino acids 141–491, 141–311 and 295–491 were then PCR amplified. EcoRI sites were added to the ends of the fragments to facilitate the in-frame cloning of the chIRF-3 downstream of the yeast GAL4 DBD. All PCR fragments were sequenced to confirm fidelity. Additional chimeric expression vectors were generated using existing restriction enzyme sites in the chIRF-3 cDNA. These regions corresponded to chIRF-3 amino acids 138–221, 221–317, 138–190 and 175–221.

The chIRF-3 expression constructs were cloned into the expression vector pcDNA1/Amp. A haemaglutinin (HA) epitope was added to the C-terminus of each expression cassette.

Short-term transfection assays

The oligonucleotides to make the binding sites to be tested in the transfection assays were designed to concatenate in only one orientation. Two binding sites were tested; the chIRF-3 optimal binding site (OBS) (5′-TGGTCCGAAACCG AAAGTG-3′) and site B from Figure 3 (5′-TGGTCCGAAAG CGAAACTG-3′). For the OBS, clones were isolated with one, three and four copies and four copies in the reverse orientation (1F, 3F, 4F and 4R, respectively) flanking the SV40 minimal promoter in the firefly luciferase reporter vector, pGL3p (Promega). For site B, the 1F and 3F constructs were isolated. The reporter constructs were co-transfected with a control plasmid to assess the transfection efficiency together with a chIRF-3 expression vector and/or the empty expression vector. The control plasmid encoded sea pansy luciferase (Promega). Transfections were performed in 24-well dishes in triplicate using the chicken hepatoma cell line LMH/2A. For each determination 1.5 µg of reporter DNA, 75–300 ng luciferase control DNA and a total of 50 or 150 ng chIRF-3 expression vector and/or empty expression vector were mixed with 100 µl of Waymouth’s medium without serum and 6 µl of FuGENE 6 transfection reagent according to the manufacturer’s directions (Boehringer Mannheim). After 15 min the sample was divided equally among three wells of cells. Luciferase activities were determined 48 h after transfection using the Dual-Luciferase Reporter Assay System (Promega) and a microplate luminometer (EG+G Berthold MicroLumat Plus, model LB96V).

Figure 3.

Characterisation of a chIRF-3-binding site. (A) EMSAs were performed using 20 ng chIRF-3 protein made in E.coli (chIRF-3) and 1 µl of in vitro translated hIRF-1 (hIRF-1). Control binding reactions contained either no protein (0 Protein) or reticulocyte lysate primed with vector DNA alone (0 RNA). The binding sites were labelled to identical specific activities as described in Materials and Methods and consisted of point mutations (A–H) of the binding site consensus shown in Figure 2B. The amount of binding to each site was expressed relative to the binding obtained for site A, the site equivalent to Mx shown in Figure 1A. (B) EMSAs performed using 2 µl of in vitro translated full-length chIRF-3 (chIRF-3_1–491) or chIRF-3 truncated after amino acid 117 (chIRF-3_1–117) or amino acid 311 (chIRF-3_1–311). Each expression construct contained an HA epitope tag at the C-terminus. Binding sites were: a 40 bp oligonucleotide sequence containing the OBS (OBS) or a 54 bp oligonucleotide sequence from the PCR-based binding site selection that was excised from the cloning vector. Rat anti-HA antibody was used in the supershifts. The asterisk (*) indicates a non-specific complex formed with protein present in the lysate. This complex is present in EMSAs using uncharged lysate and does not supershift with the anti-HA antibody.

Transfection assays testing the chimeric GAL4/chIRF-3 expression vectors were performed essentially as described above except that pFR-Luc (Stratagene) was used as the reporter vector. This reporter has five copies of the GAL4 DNA-binding site upstream of the firefly luciferase gene.

Western blot analysis

To compare the level of expression of chIRF-3 in transfection experiments an identical experiment was run in parallel and the cells from each triplicate determination pooled for total protein extracts. Protein concentrations were determined by Bradford assay (Bio-Rad). Protein samples (10 µg) were resolved on a 10% SDS–PAGE gel which was then blotted to Immobilon P membrane (Millipore). The chIRF-3 protein was detected using a rat anti-HA primary antibody (clone 3F10; Boehringer Mannheim), a chemiluminescence reagent (NEN Life Science) and X-Omat Blue XB-1 film (Kodak). An InstantImager (Packard Instrument Co.) was used for densitometry of the film.

RESULTS

Identification of a chIRF-3-binding motif

Although the IRF family is defined by its conserved DBD, members differ widely in their ability to bind DNA as assessed by in vitro binding assays (1). To characterise the DNA binding properties of chIRF-3, we performed EMSAs using in vitro translated full-length protein and as probes IFN response elements from the promoter regions of two chicken genes, Mx and a MHC class I gene (Mx and chICS sites, respectively) (25,26). In addition, a previously characterised synthetic IRF-binding site, C13, was also used (27). As shown in Figure 1A, these sites all conform to IRF-E or ISRE consensus sequences (17,18). As expected, the chIRF-3 protein bound very well to the Mx site. However, binding to the C13 site was markedly lower and no binding to the chICS site could be detected (Fig. 1B). This binding preference for the Mx site over very similar sequences found in the C13 and chICS sites was consistent despite different probe and lysate preparations. In addition, similar binding preferences were obtained when a recombinant full-length chIRF-3 protein was used in EMSAs instead of the lysate (data not shown). For comparison, we performed a similar EMSA using in vitro translated human IRF-1 (hIRF-1). In contrast to chIRF-3, hIRF-1 exhibited strong binding to both the Mx and C13 sites (Fig. 1B). However, like chIRF-3, virtually no DNA–protein complex was detectable with the chICS probe. Thus, chIRF-3, like some IRF proteins such as IRF-1, is capable of binding DNA in isolation. In addition, chIRF-3 and hIRF-1 demonstrated clear and distinct binding site preferences.

Figure 1.

chIRF-3 displays binding site preferences in EMSAs. (A) Sequences of IFN response elements found in two chicken genes, Mx (Mx ISRE) and a MHC class I gene (chICS), and a synthetic IRF-binding site (C13) (27). Also shown are the consensus IRF1- and IRF-2-binding sites, IRF-E (17) and ISRE (18). (B) EMSA showing binding of in vitro translated chIRF-3 and hIRF-1. Binding assays were performed with either 2 µl of chIRF-3 or 1 µl of hIRF-1 reticulocyte lysate and ³²P-labelled Mx, C13 and chICS double-stranded oligonucleotides as probes as described in Materials and Methods.

To further define binding sites preferred by chIRF-3, a PCR-based binding site selection assay was performed that was similar to that used to obtain the IRF-E consensus binding site for IRF-1 and IRF-2 (17). After seven rounds of enrichment the selected sequences were cloned, sequenced and visually aligned. The consensus sequences were calculated according to the frequency of each nucleotide at a given position. In 31 of the 42 sequences analysed putative binding sites were found. These were assigned to two categories, 18 sequences in the first and 13 in the second (Fig. 2A and B).

Figure 2.

Sequence analysis of the oligonucleotides cloned from the binding site selection assay. Sequences were aligned and the number of clones containing each nucleotide at each position of the binding site is shown. Of the clones sequenced, 18 contained the binding site shown in (A) and 13 contained the binding site shown in (B). The consensus binding sites derived from the cloned sequences are shown below the figures.

The first consensus sequence consisted of a 7 bp inverted repeat, 5′-CGAAA^G/_C^T/_CN_10–13^G/_A^C/_GTTTCG-3′ (Fig. 2A). The repeat was separated by 10 and 11 random nucleotides, in six and 10 of the selected sequences, respectively. In two of the selected sequences the repeat was separated by 12 and 13 nt. Based on the crystal structures of IRF-1 and IRF-2 bound to IRF-binding sites, the IRF DBDs bind GAAA and AANNGAAA core motifs, respectively (15,16). It is probable, therefore, that a chIRF-3 molecule binds to each of the GAAA motifs in the inverted repeat. The presence of 10–13 random nucleotides between the inverted repeats places the two chIRF-3 molecules on the same side of the α-helix, which suggests that the two molecules could interact as a homodimer on the DNA.

The second consensus sequence was found in 13 of the 31 sequences obtained from the binding site enrichment procedure (Fig. 2B). This 11 bp sequence, 5′-AAAN_0–1CGAAAGT-3′, is very similar to the consensus binding site, 5′-G(A)AAA^G/_C^T/_CGAAA^G/_C ^T/_C-3′ (IRF-E), for IRF-1 and IRF-2. However, there are two potentially important differences between the two consensus sites. First, IRF-E contains two of the GAAA core binding motifs while the chIRF-3 site contains only one. In the chIRF-3 binding sites, the 5′ ‘triplet A’ motif is preceded by a G in only six of the 13 sequences examined whereas in the IRF-1 and IRF-2 selected sequences there is a G in this position 75 and 90% of the time, respectively (Fig. 2B; 17). The second notable difference between the chIRF-3 site and IRF-E is the lack of variation among the sequences selected by the chIRF-3 protein compared to IRF-1 and IRF-2. For example, in the chIRF-3 site the nucleotide preceding the GAAA motif is C in 11 of the 13 sequences, whereas in the IRF-1- and IRF-2-selected sequences T and C are present at this position almost equally (Fig. 2B) (17). In addition, the dinucleotides following the GAAA core motif are GT in 12 of the 13 sequences selected by chIRF-3 and CA in only one sequence. In IRF-E these nucleotides are highly variable (^G/_T^T/_C). In the binding sites used in Figure 1 these nucleotides are CT, GT and GG in the Mx, C13 and chICS sites, respectively. Since a GG dinucleotide is not part of the consensus binding site for either IRF-1 or chIRF-3, its presence in the chICS site may partly explain why neither protein bound this site in EMSAs (Fig. 1B). In contrast, GT was found almost exclusively at this location in the chIRF-3-selected sequences. Therefore, its presence in the C13 binding site does not explain the minimal binding of chIRF-3 to this site.

Analysis of the chIRF-3-binding motif

To further define the chIRF-3 binding site, we used EMSAs to assess the relative binding of a recombinant chIRF-3 protein to mutated forms of the second chIRF-3 consensus binding site. As in the Mx and C13 sites, two core GAAA motifs were present in each of the eight mutated binding sites tested and the effect of varying the number of bases (one or two) and the base content (^G/_C^C/_T) between the two cores was examined. In addition, the base composition of the dinucleotide following the second core GAAA motif was varied (CT, GT or CA). Each of the mutated binding sites was labelled to the same specific activity and the binding of chIRF-3 to each site was assessed by counting the radioactivity in the bound complex.

The first binding site (A of Fig. 3) was equivalent to the chicken Mx site, 5′-GAAA_CGAAACT-3′, where _ represents the absence of a corresponding nucleotide at this position. In the B binding site, the corresponding nucleotide is a G. In the selected sequences no base or a G at this position occurs with equal frequency (four of 13 sequences each) (Fig. 2B). The presence of a G at this position decreased the relative binding of chIRF-3 to site B by 30–40% of that observed for the Mx site (site A) (Fig. 3). The GAAA core motif was followed by GT in 12 of the 13 sequences selected (Fig. 2B). Therefore, we tested whether this change affected the need for one or two bases between the two GAAA cores in the mutant sites. Binding to the C probe, 5′-GAAA_CGAAAGT-3′, was 30–50% greater than that obtained for the A site (Fig. 3, compare A and C). Therefore, changing the dinucleotide following the second GAAA core motif to GT from CT increased the binding of chIRF-3. However, inserting a G residue between the cores in site D, 5′-GAAAGCGAAAGT-3′, decreased the binding to levels that were roughly equivalent to those obtained for site A and ∼70–80% of those obtained for site C (Fig. 3, compare A and D and C and D).

Although a _G or GC was found most often preceding the core motif in the selected sequences, CC, AC and TC were also found (data not shown). When GC was exchanged for CC in site E, a marked increase in binding of chIRF-3 was noted; binding to E was over twice that noted for A, B and D and 50–70% greater than that observed for site C (Fig. 3). Consequently, binding of chIRF-3 was decreased if two bases, in this case GC, replaced the single base C between the GAAA core motifs (compare A and B and C and D). However, when the G was replaced with a second pyrimidine residue, GC in site D to CC in site E, binding of chIRF-3 was from 1.5- to 4-fold greater than obtained with any of the other sites (compare E with A, B, C or D). Presumably, the presence of two pyrimidines instead of a purine and a pyrimidine at this position provides additional stabilising contacts between the DNA and protein.

As mentioned previously, in the sites selected by chIRF-3 the nucleotide immediately preceding the GAAA core motif was a C in 11 of the 13 sequences (Fig. 2B). To test whether a pyrimidine at this location was sufficient for binding, GC was changed to GT in site F and to _T in site G. As expected from the results shown in Figure 1, binding to the F probe (which is equivalent to the C13 site), 5′-GAAAGTGAAAGT-3′, was ∼30% of that obtained for site A (equivalent to Mx). When the G residue preceding the T was deleted in site G, 5′-GAAA_TGAAAGT-3′, no binding was detected (Fig. 3). Therefore, a _T at this position allowed less binding of chIRF-3 than _C, GC or CC (compare F with B, D with E and G with A and C).

As mentioned above, in the chIRF-3-selected sequences the dinucleotide following the GAAA core motif was GT in 12 sequences and CA in one (data not shown). To test whether a pyrimidine was also necessary at the 3′-end of the binding site, the dinucleotide in site H was changed to GA. As shown in Figure 3, chIRF-3 did not bind to site H, 5′-GAAA_CGAAAGA-3′ (compare H with A and C). Therefore, the relative binding of chIRF-3 to the binding sites mutated in the dinucleotide following the core is GT > CT >>> GA.

For comparison, the identical binding assays were done with in vitro translated hIRF-1. Binding of hIRF-1 to the tested sites was diminished ∼2-fold whenever spacing between the core GAAA motifs was decreased from two to one nucleotides, 5′-GAAANNGAAA-3′ to 5′-GAAA_NGAAA-3′ (Fig. 3, compare A, C and G with B, D, E and F). Changing the two nucleotides between the core GAAA motifs from GC to CC or GT did not noticeably affect the degree of binding (compare D, E and F). In addition, a CT or GT dinucleotide immediately following the second GAAA motif had no effect on binding of hIRF-1 (compare B and D). Thus, our results agree with the broad consensus binding site, 5′-G(A)AAA^G/_C^T/_CGAAA^G/_C^T/_C-3′ (IRF-E), previously obtained for IRF-1 and IRF-2 (17).

Taken together, our data suggest that there is a definite hierarchy in the nucleotides acceptable around the GAAA core motif that affects the binding potential of chIRF-3. For example, binding is maximised if two pyrimidines, specifically CC, precede the core motif. In the absence of CC, as in _C or GC, binding by chIRF-3 requires that a T, or perhaps a C since this was not tested, be present at the end of the site, AAA_GAAA^C/_GT. Such fine discrimination in the binding sequence was not found with the IRF-1 protein. Therefore, the OBS for in vitro binding of chIRF-3 is 5′-GAAACCGAAAGT-3′.

chIRF-3 binds as a homodimer

In EMSAs, mutation of the 5′ AAA motif in the OBS abrogates all binding by chIRF-3 (data not shown). This suggests that part of the 5′ A triplet is essential for binding. Thus, it is possible that chIRF-3 binds as a dimer where each molecule may make contact with an AAA motif or as a monomer that makes contacts with both the 5′ and 3′ AAA motifs. To determine if chIRF-3 is capable of forming homodimers, we used a co-immunoprecipitation assay in which ³⁵S-labelled in vitro translated chIRF-3 proteins, with and without an HA epitope tag, were incubated together, followed by immunoprecipitation with an anti-HA antibody (Boehringer Mannheim). The two chIRF-3 proteins have a different apparent M_r in SDS–PAGE which allowed analysis by fluorography. We found no evidence that the untagged chIRF-3 protein was co-precipitated with the tagged protein, even when the two proteins were translated in the same mix (data not shown). It is possible that contact with a DNA-binding site stabilises the interactions necessary to form a homodimer, in which case we would not have detected the interaction in this assay.

To determine if chIRF-3 binds DNA as a dimer, we performed EMSAs using both full-length and truncated chIRF-3 proteins. As shown in Figure 3B, full-length chIRF-3 (chIRF-3_1–491) and proteins truncated after amino acids 117 (chIRF3_1–117) and 311 (chIRF-3_1–311) bind to the OBS. The results revealed an unexpectedly large difference between the mobility of the complex formed with chIRF3_1–117 and the complexes formed with either full-length protein or chIRF-3_1–311. The mobility of the complex formed with chIRF-3_1–311 was only ∼20–25% faster than that of the complex involving full-length chIRF-3 despite the 180 amino acid difference in polypeptide length. In contrast, the mobility of the complex with chIRF3_1–117 was ∼3-fold faster than that of the complex formed with chIRF-3_1–311. Similarly, supershift assays with an anti-HA tag antibody resulted in complexes that failed to enter the gel with both chIRF-3_1–311 as well as full-length protein, while the complex involving the DBD alone (chIRF3_1–117) migrated with a mobility that was ∼50% that of the chIRF-3_1–311 complex in the absence of antibody. These observations strongly suggest that chIRF3_1–117 may bind as a monomer while chIRF-3_1–311 and full-length protein bind as dimers and that stabilisation of the dimer requires a domain between amino acids 117 and 311.

We also performed similar EMSAs using binding sites that contained the selected inverted repeat sequences as probes. One of these assays is shown in Figure 3B. In this case the relative mobilities of the complexes formed with chIRF-3_1–491, chIRF-3_1–311 and chIRF3_1–117 were identical to those obtained using the OBS as probe, suggesting that the stoichiometry of chIRF-3 molecules binding to the two sites is the same. In addition, the complexes formed with the longer inverted repeat probe migrate only marginally slower than the equivalent complexes formed with the OBS probe. Therefore, we suggest that chIRF-1_1–117 also binds the inverted repeat probe as a monomer and chIRF-3_1–491 and chIRF-3_1–311 bind as single homodimers.

The two half-sites of the consensus inverted repeat sequence are identical and spaced at least 10 nt apart. This suggests that either the sites have to be on the same side of the DNA helix or that steric factors dictate the minimum distance between them. The fact that chIRF-3_1–117 also appears to bind to this site as a monomer suggests that protein–protein interactions stabilise binding of the dimer of chIRF-3_1–311 and full-length protein. The lack of detectable binding of two molecules of chIRF-3_1–117 to the inverted repeat may be attributable to the relatively weak binding observed with the DBD alone and the relatively large excess of unbound oligonucleotide.

chIRF-3 is a transcriptional activator

The function of chIRF-3 in transcriptional regulation was tested in short-term transfection assays in LMH/2A cells. The OBS was concatenated and cloned in the luciferase reporter vector, pGL3p. As shown in Figure 4, introduction of the OBS decreased the basal level of transcription from the reporter. The relative luciferase activity was decreased 20-fold compared to the vector control when one copy of the OBS was present, but only 3- to 4-fold when three or four copies were introduced. Since IRF-2 has been shown to repress transcription from promoters containing IRF-binding sites (4,28), it is possible that the decrease in basal level of transcription could be mediated by binding endogenous IRF-2. We have shown previously that untreated LMH/2A cells contain high levels of IRF-2 mRNA, suggesting the presence of significant amounts of IRF-2 protein (22). Why one copy of the OBS should produce markedly more repression than three copies is, at present, unknown.

Figure 4.

chIRF-3 acts as a transcriptional activator in short-term transfection assays. LMH/2A cells were transfected with pGL3p in which the beetle luciferase reporter gene is under the control of a minimal SV40 promoter (vector) or pGL3p in which one, three or four copies of an IRF-binding site in the forward orientation (1F, 3F and 4F, respectively) or four copies in the reverse orientation (4R) were cloned upstream of the SV40 promoter. The IRF-binding sites used were the optimum chIRF-3-binding site (site E, Fig. 3), designated OBS, and a less than optimal chIRF-3-binding site (site B, Fig. 3). Cells were co-transfected with a sea pansy luciferase expression vector, pRL-TK, as an internal control for transfection efficiency and 0, 8, 16, 32 or 50 ng chIRF-3 expression vector and/or the empty expression vector, pcDNA1/Amp, to a total of 50 ng. Activity of each transfection is expressed as beetle luciferase activity/sea pansy luciferase activity. Shown are the results of a typical experiment. The error bars represent ± SD of triplicate determinations.

Co-transfection of the chIRF-3 expression vector alleviated the repression due to the presence of the OBS, but had little or no effect on the pGL3p vector control. Expression of chIRF-3 increased the relative luciferase activity obtained with the vector containing a single copy of the OBS ∼20-fold, to a level comparable to that produced by the pGL3p vector control. The luciferase activity of the reporter vector containing three copies of the OBS was increased ∼46-fold over the level of activity found when the cells were co-transfected with the empty expression vector and 12-fold over the level produced from the pGL3 vector control. The fold induction obtained with four copies of the OBS was the same as that obtained with three copies regardless of the orientation. Thus, chIRF-3 is a strong transactivator and its activity is unaffected by the orientation of its binding site, at least in the context of this simplified promoter.

We also assessed whether chIRF-3 could discriminate between strong and weak binding sites in the transfection assays. In the EMSAs, binding of chIRF-3 to site B was ∼4-fold lower than binding to the OBS (site E) (Fig. 3). Consequently, we placed either one or three copies of site B flanking the SV40 promoter and compared the relative luciferase activities of these reporters to those containing the OBS (Fig. 4). The presence of site B also repressed activity from the promoter to a similar extent as that obtained from the OBS-containing reporters. Furthermore, co-transfection of the chIRF-3 expression vector induced luciferase activities that approximated those obtained with the OBS contructs whether one or three copies of site B were present. Thus, chIRF-3 did not discriminate between the OBS site and site B in transfection assays even though binding to these in vitro was clearly different (Fig. 3). It is possible that, in the artificial conditions of the transfection assay, the large amounts of chIRF-3 and binding target present in the transfected cells obscures such subtle differences in binding affinity.

Mapping the activation domain of chIRF-3

To define the region of chIRF-3 responsible for its ability to function as a transactivator, we designed several expression vectors encoding fusion proteins composed of the DBD of the yeast transactivator GAL4 located N-proximal to different portions of chIRF-3. These expression vectors were tested for their ability to transactivate a promoter containing five copies of the GAL4 DNA-binding element. Figure 5A is a schematic of the chimeric expression vectors showing which portions of chIRF-3 were fused to the DBD of yeast GAL4. Construct pGAL4/ch 141–491, which contains most of chIRF-3 except its DBD, activated the reporter 27-fold, confirming the presence of a strong activation domain (Fig. 5B). When the C-terminal 180 amino acids of pGAL4/ch 141–491 were removed to generate pGAL4/ch 141–311, the level of transactivation increased from 27- to 44-fold, suggesting that we had removed an inhibitory domain between amino acids 312 and the end of the protein. Construct pGAL4/ch 305–491, which contains the entire C-terminal region, was essentially inactive. pGAL4/ch 138–221 strongly transactivated the reporter 53-fold, while the N-terminal portion of this construct (pGAL4/ch 138–190) only transactivated 19-fold. The two remaining constructs (pGAL4ch 175–221 and pGAL4/ch 221–317) were incapable of transactivating the reporter. Taken together, the data indicated that the chIRF-3 protein between amino acids 138 and 221 is sufficient for transactivation.

Figure 5.

Mapping the transactivation domain of chIRF-3. (A) A schematic diagram of the chimeric expression vectors used in the transfection experiments compared to the structure of full-length chIRF-3 protein (amino acids 1–491). The chIRF-3 DBD and the IAD are shown as grey and black boxes, respectively. The numbers in the chimeric expression vector names correspond to chIRF-3 amino acids. Non-chIRF-3 amino acids that were added to the C-terminus during cloning are as follows: pGAL4/ch 141–491 (none), pGAL4/ch 141–311 (EFARASSPVRL), pGAL4/ch 305–491 (none), pGAL4/ch 138–221 (GGSTSSRAAATAVEGSTLEGPIL), pGAL4/ch 221–317 (DPAATAVEGSTLEGPFL), pGAL4/ch 138–190 (RGRL) and pGAL4/ch 175–221 (GGSTSSRAAATAVEGSTLEGPIL). (B) LMH/2A cells were co-transfected with the pFR-Luc reporter construct, pRL-TK and 50 ng of either a chimeric or control expression vector. Fold induction was calculated by dividing the normalised luciferase activity of the chimeric transfectants by the normalised luciferase activity of the control transfectants. The error bars represent ± SD of three independent experiments.

To confirm the location of the activation domain, the co-transfection experiments were repeated in LMH2A cells using a reporter construct containing three copies of the chIRF-3 OBS (OBS3F) and chIRF-3 expression constructs containing the DBD, amino acids 1–117 (chIRF-3_1–117), amino acids 1–311 (chIRF-3_1–311) and the full-length protein, amino acids 1–491 (chIRF-3_1–491). A C-terminal HA epitope tag was added to both the truncated and full-length proteins and used to assess their relative expression levels following transfection. Approximately equivalent levels of chIRF-3_1–491 were expressed in the pGL3p control and OBS3F samples (Fig. 6B). The level of chIRF-3_1–311 protein was ∼15 and 22% of the level of the full-length protein in the control and OBS3F samples, respectively. Control and OBS3F samples co-transfected with chIRF-3_1–117 contained ∼140 and 70% of the level of samples expressing the full-length chIRF-3 protein, respectively. As shown in Figure 6A, expression of chIRF-3 has little or no effect on the luciferase control vector, pGL3p. Interestingly, co-transfection of the reporter OBS3F and the chIRF-3_1–117 expression vector resulted in a 2-fold decrease in luciferase activity compared to the background activity of the OBS3F reporter alone, suggesting that binding of the DBD by itself inhibits the basal level of transcription from the OBS3F promoter. As expected, co-transfection with full-length chIRF-3_1–491 resulted in a 37-fold increase in luciferase activity. However, when chIRF-3 was truncated to include only the DBD and the region containing the putative activation domain, chIRF-3_1–311, luciferase activity was more than double that obtained for the full-length protein (83- versus 37-fold), despite the fact that there is ∼5-fold less of the truncated protein than the full-length protein present in the samples. These observations support the results obtained with GAL4/chIRF-3 hybrid proteins and strongly suggest the presence of an inhibitory domain in the C-terminal region of the protein and a strong activation domain between amino acids 117 and 311.

Figure 6.

The region between amino acids 117 and 311 of chIRF-3 activates transcription in the context of the chIRF-3 DBD. (A) LMH/2A cells were co-transfected with a reporter (pGL3p/OBS3F), pRL-TK and a chIRF-3 expression construct or control vector. The chIRF-3 expression vectors each contained the chIRF-3 DBD and various portions of the C-terminal end of the protein; chIRF-3_1–117 contains only the DBD, chIRF-3_1–311 contains the putative activation domain and chIRF-3_1–491 contains the entire protein coding sequence. Each determination was performed in triplicate and the experiment was repeated several times with similar results. Shown is a typical experiment. Fold induction was calculated as discussed in Figure 5. (B) Western blot showing the relative expression levels of the HA-tagged chIRF-3 proteins expressed in the transfection experiment shown in (A).

DISCUSSION

As more becomes known about individual members of the IRF family of transcription factors, the problem of how specific genes are regulated by these highly related factors is being elucidated. On one level, some specificity is imparted by the pattern of expression of the factors themselves, whether they are constitutively expressed or inducible and whether their expression is cell-type specific or ubiquitous, long- or short-lived (reviewed in 1–3). Promoter context and cell type-specific expression of other transcription factors also provides an important additional level of control. Although some IRF factors have been shown to bind as homodimers or monomers, others can heterodimerise with other IRF proteins or other transcription factors. Both Tyr and Ser/Thr phosphorylations have been shown to be important for protein–protein interactions, DNA–protein interactions and also for correct localisation to the nucleus in some cases (7,29–38). Finally, the IRF-binding sites themselves are likely important. A recent survey of IRF-regulated promoters showed that although all of the IRF-binding sites contained the core GAAA motif, the number of these motifs and their spacing was variable (16).

The manner in which the DBDs of IRF proteins bind to DNA is just beginning to be studied (15,16,39). Recent co-crystallisation studies of IRF DBDs with oligomers containing IRF-binding sites revealed that the core binding sequences for IRF-1 and IRF-2 are GAAA and AANNGAAA, respectively. The diverse functions of IRF proteins and variations in IRF-binding sites strongly suggest that the simple core recognition sequence is probably not sufficient to target these proteins to specific promoters. In addition, IRF proteins can demonstrate different affinities for IRF-binding sites in in vitro assays despite the similarities in their DBDs. Thus, it is probable that other primary sequence information within the DBD is important for binding specific DNA sequences. Since the crystal structures were determined for the DBDs alone, the full-length protein may adopt conformations that increase or decrease the binding site specificity of a particular IRF protein.

In vitro analysis of binding site preferences of the most closely related members of the IRF family, IRF-1 and IRF-2, produced an identical consensus for both proteins, the IRF-E (17). Although the IRF-E contains the core binding motif, IRF-1 and IRF-2 proteins showed decided preferences in nucleotides at a number of positions outside the core. We have confirmed these preferences for IRF-1. As shown in Figure 3, whenever the mutant binding sites departed from the consensus IRF-E, binding of IRF-1 was reduced by at least half. The most critical element appeared to be the spacing between the GAAA core motifs, with IRF-1 demonstrating a clear preference for two nucleotides rather than one. Of the 36 sites selected by IRF-1 and IRF-2 by Tanaka et al., >80% also contained two nucleotides rather than one separating the cores (17).

Using a similar PCR-based binding site preference assay we have shown that chIRF-3 is more selective in its site preferences than either IRF-1 or IRF-2 (Fig. 2; 17). In 13 of the 31 sequences selected by chIRF-3 the consensus binding site was 5′-AAAN_0–1CGAAAGT-3′. Although this binding motif fits within both IRF-E and ISRE consensus sequences, it is longer than the core binding motifs found in the structural studies of IRF-1 and IRF-2 (15,16). In addition, the heterogeneity found in the IRF-E and ISRE sites outside the core GAAA motifs is considerably lower in the sites selected by chIRF-3. The C before the GAAA core motif was present in 11 of 13 sequences selected and G was present in the remaining two. The dinucleotide GT followed the core in 12 sequences with CA in the other. Subsequent testing in EMSAs confirmed these preferences and also showed that although chIRF-3 would bind when N was either no nucleotide or G or C, binding was approximately doubled when a C was present at this location. The IRF-1 protein bound avidly to all these oligonucleotides.

These results explain our initial observations of the binding site preferences of chIRF-3 (Fig. 1). Although the ICS site, 5′-GAAAGCGAAAGG-3′, contains an acceptable GC preceeding the GAAA core binding motif, the GG dinucleotide following the core abrogates binding of chIRF-3. Similarly, the GT dinucleotide following the core in the C13 site, 5′-GAAAGTGAAAGT-3′, fits the consensus chIRF-3-binding site but the GT preceeding the core results in reduced binding (compare site F with sites C–E, Fig. 3). In contrast, the _C preceeding and CT following the core binding motif in the Mx site, 5′-GAAACGAAACT-3′, although not optimal for chIRF-3 binding, fall within the acceptable variations in the chIRF-3-binding site.

The remaining 18 selected sequences contained a consensus inverted repeat consisting of 5′-CGAAA^G/_C^C/_T-3′, separated by 10–13 nt. Interestingly, we did not isolate any binding sites that contained a single copy of this motif. In binding the ISRE-like sequence at least part of the 5′ AAA motif appears essential, since binding is abrogated when it is removed by mutation. Whether or not this is true for the binding site containing the inverted repeat sequence has not been determined. However, it is possible that other sequence information, perhaps within the non-random portions of the selection probes, may be involved in stabilising the binding. However, since two consensus half-ISRE-like sites were found in each selected site, at least two chIRF-3 DBDs contact the DNA. Although we were unable to find evidence that chIRF-3 was capable of homodimerisation using a co-immunoprecipitation assay, data from EMSAs suggest that chIRF-3 binds both the ISRE-like and inverted repeat sites as a homodimer. EMSAs also suggest that protein–protein interactions are important for binding both half-sites of the inverted repeat probe since the chIRF-3 DBD alone binds as a monomer.

The co-crystallisation of the DBD of IRF-2 on a concatenated ISRE resulted in the cooperative binding of IRF-2 molecules on each AANNGAAA core where the 5′ AA forms part of the GAAA core motif on the opposite face of the DNA (16). The cooperative binding was not due to interaction of the IRF-2 molecules but resulted from bending of the DNA towards the first IRF-2 bound, which allowed the second to bind on the opposite face. As mentioned above, our data strongly suggest that the chIRF-3 DBD binds an ISRE-like element as a monomer and that the full-length or shorter truncations of chIRF-3 bind as homodimers. A single homodimer also binds the inverted repeat probes although, in this case, severe constraints are imposed on both the orientation of the half-ISRE-like binding sites and the spacing between them. The binding sites must be inverted and at least 10 nt apart, suggesting that chIRF-3 homodimers are arranged with their DBDs juxtaposed on the same side of the DNA α-helix. Consequently, an arrangement of chIRF-3 on the DNA like that observed for the IRF-2 DBD does not appear to be possible, probably due to the bulk of the C-terminal portions of the molecules.

It is possible that protein modifications could alter the shape of the C-terminus of chIRF-3 and allow a homodimer to bind over a shorter stretch of DNA. Mammalian IRF-3 in nuclear extracts does not bind DNA in the absence of virally induced Ser/Thr phosphorylation of a region near the C-terminus (34–36,40,41). Once phosphorylated, however, the protein homodimerises, is retained in the nucleus and binds an ISRE (34,40,42). It is presumed that the unphosphorylated protein is in a ‘closed’ conformation that acts as an inhibitory domain for DNA binding and protein associations. A similar region is phosphorylated in IRF-7 (37,38,43). The phosphorylated region is equally conserved between chIRF-3 and mammalian IRF-3 and IRF-7. At present, we do not know whether this region of chIRF-3 becomes phosphorylated in response to virus infection or IFN treatment. However, the in vitro translated proteins used here clearly bind DNA and homodimerise, suggesting that they are not in a completely ‘closed’ conformation. In fact, at least one interaction domain of chIRF-3 has been mapped to between amino acids 117 and 311.

It is also possible that chIRF-3 functions as a heterodimer or larger complex and that the interaction with another protein(s) alters the shape of the molecule to allow binding of the complex to a smaller DNA element. For example, the IRF protein ISGF3γ can bind to an ISRE independently. However, it exerts its effect on transcription exclusively through the ISGF3 complex, formed by association with phosphorylated STAT-1 and STAT-2. The trimeric complex binds the ISRE of IFN-inducible genes (11,44,45). In B cells IRF-4 is recruited to its DNA-binding site in the immunoglobulin light chain gene enhancers (a composite Ets/IRF element) by interaction with PU.1. PU.1, an Ets family transcription factor, must be phosphorylated on Ser148 to interact with IRF-4 (7,46). For both ISGF3 and IRF-4/PU.1 the interactions, at least in part for IRF-4, are formed through the IAD (29,47). This domain, based on sequence similarities, is shared with chIRF-3 (14,48).

Transfection experiments indicate that chIRF-3 can transactivate a minimal promoter containing an optimised chIRF-3-binding site. Furthermore, the level of transactivation correlated well with the amount of chIRF-3 expression vector used and, like other ISREs, transactivation through the OBS was independent of orientation. We confirmed that chIRF-3 has a strong activation domain by creating vectors that express yeast GAL4 DBD/chIRF-3 fusion proteins and assessing their ability to transactivate a GAL4-responsive reporter. These studies revealed that amino acids 138–221 of chIRF-3 are sufficient for strong activation in this system. In addition, the transactivation potential of this region was maintained even when the GAL4/chIRF-3 chimeric proteins were tested in yeast (data not shown), suggesting that chIRF-3 may interact directly with the basal transcriptional machinery. Experiments with truncated proteins containing the native chIRF-3 DBD supported results obtained with the GAL4 fusion proteins and delimited the activation domain to a region between amino acids 117 and 311.

The transactivation domain of chIRF-3, unlike many such domains, is not particularly rich in Pro, Glu or acidic amino acids (49). Furthermore, the domain contains no known protein–protein interaction motif and is in a region of limited sequence similarity among the IRFs. Although the activation domain of IRF-7 is in a similar location, the two proteins share only 25% sequence identity in this region (10,43). Alignment of IRF-3 and chIRF-3 indicates that the exceptionally long activation domain defined for the mammalian IRF overlaps the C-terminal portion of the chIRF-3 activation domain but, in the region of overlap, sequence identity with chIRF-3 is lower than with IRF-7 (42). A computer search of the translated GenBank database for protein sequences with similarities to amino acids 138–221 of chIRF-3 failed to identify any similar peptide motifs. Thus, it will be of interest to determine whether the mechanism of activation by chIRF-3 involves interaction with the co-activator p300/CBP as demonstrated for some mammalian IRFs, including IRF-1, IRF-3 and IRF-7.

Transfection experiments also revealed the presence of an inhibitory domain between amino acid 311 and the C-terminal end of the protein. Inhibitory domains at the C-proximal ends of mammalian IRF-3, IRF-4 and IRF-7 have been described (42,43,47,48). In IRF-3 and IRF-7 the inhibitory regions encompass amino acid residues that have been shown to be phosphorylated in response to viral infection, which has been hypothesised to result in the exposure of functional domains in these proteins and stabilisation of the interaction with the transcriptional co-activator p300/CBP (34,37,38,40,43,50,51). As mentioned above, chIRF-3 shares a short relatively conserved C-proximal region with IRF-3 and IRF-7 which includes motifs similar to those known to be phosphorylated in IRF-3 and IRF-7. These and previous observations suggest that regulation of chIRF-3 may occur at multiple levels which include not only transcriptional induction and rapid turnover of chIRF-3 mRNA, but also post-translational events that control the subcellular localisation and activity of the protein.