Silencer Activity in the Interferon-A Gene Promoters*

Interferon-A (IFN-A) differential gene expression is modulated by a complex interplay between cis-acting DNA elements and the corresponding specific trans-regulating factors. Substitutions in the proximal virus-responsive element of the interferon-A (IFN-A) promoters contribute to their differential gene expression. The 5′ distal silencing region in the weakly virus-inducible murine IFN-A11 gene has been previously delimited. DNase I footprinting experiments and transient gene expression assays demonstrate identical silencing activity in equivalent regions of the genes for IFN-A11 and IFN-A4 promoters. A minimal 20-mer distal negative regulatory element (DNRE) in both promoters is necessary and sufficient for the silencing and a region in the highly inducible IFN-A4 promoter located between the silencer and the virus-responsive element overrides the silencer activity. Mutations in the central region of the DNRE, causing derepression, also altered the formation of one of the two major DNA-protein complexes. One of these contains a protein related to or identical to the high mobility group I(Y) proteins, while the other complex contains a major protein present in uninduced and virus-induced cells with a molecular mass of 38 kDa, which may be related to the silencer activity. Similar DNREs are present in other virus-uninducible IFN-A promoters, and these data suggest that a common silencer may mediate the transcriptional repression in different genes of this family.

The type I interferon (IFN-A and -B) 1 genes have been shown to be a model to examine the positive and negative transcriptional mechanisms controlling virus-inducible differential and cell type-specific gene expression (1)(2)(3)(4). The regulatory sequences that control the transcription of the IFN-A and -B genes are generally located within a 110-base pair region immediately upstream from the transcription site. The cis-ele-ments present in the IFN promoters and their trans-regulator proteins have been extensively studied within the human IFN-B promoter. Analysis of this region has revealed a complex organization of several synergistic overlapping positive (PRDI to -IV) and negative regulatory domains (NRDI and NRDII) (5)(6)(7). PRDI and PRDIII contain the core sequence of the interferon regulatory factor (IRF) element, named IRF-Es (8). The members of the IRF family including the activators IRF-1/ interferon-stimulated gene factor-2 (ISGF2) or ISGF3␥ and the repressor IRF-2/ISGF1 have been shown to bind to IRF-Es. All share in their DNA-binding domains a conserved amino-terminal region (9 -12). The IRF-1 (Ϫ/Ϫ) and IRF-2 (Ϫ/Ϫ) knockout studies have indicated that, for IFN virus induction, an IRF-1-independent pathway exists, whereas IRF-2 plays a role in the transcriptional shut-off of IFN-B and IFN-A gene transcription (13,14). Recently, the involvement of ISGF3␥ in the transcriptional regulation of the IFN-B gene has been suggested (15,16). NF-B, activated by virus infection, binds to PRDII motif and participates in the transcriptional activation of the IFN-B gene (17)(18)(19). Moreover, it has been demonstrated that high mobility group protein (HMG) I(Y) binds to the minor groove of an AT-rich region within PRDII and to two sites flanking the activating transcription factor-2/c-AMP regulatory element-binding protein binding site within PRDIV. HMG I(Y) interaction with DNA is constitutive; it bends the DNA and increases the binding activity of both NF-B and activating transcription factor-2/c-AMP regulatory element-binding protein (19 -21). The NRDI contains a negative regulatory element, which partially overlaps PRDII and is able to repress PRDII-mediated gene activity (22). Dorsal switch protein-1, a member of the HMG-1 protein family from Drosophila, binds to negative regulatory element and represses the activation due to NFB (23).
The IFN-B and -A genes are transcriptionally activated and repressed through different mechanisms, and the differences between the expression of the individual IFN-A genes reflect the transcriptional activity of the corresponding promoter regions (2, 4, 24 -28). In the virus-responsive element, Ϫ109/Ϫ64 of the IFN-A promoters, several proteins bind to this sequence and may be specific IFN-A regulators. An uncharacterized TG protein binds to the virus-responsive element of the human IFN-A1 (2). Analysis of the murine IFN-A4 gene promoter has delimited the inducible element (IE), Ϫ109/Ϫ75, that confers both virus inducibility and cell type-specific expression (29). IRF-2 binds to the IE because it contains an IRF-Es (27,29). The ␣F1/B complex binds to the IE (30) and was suggested to cooperate with IRF-1 for efficient activation. Furthermore, the virus-induced factor (VIF), which also binds to the IE, participates in transcriptional activation of the murine IFN-A4 gene after virus induction (27). However, none of the specific IFN-A factors has yet been isolated.
This study concerns the differential expression of two murine * This work was supported by the Centre National de la Recherche Scientifique, The Université René Descartes Paris V, and grants from Association pour la Recherche sur le Cancer (Contrat 1042) and Ligue Nationale Française contre le Cancer. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
IFN-A genes. The IFN-A11 was found to be poorly expressed upon Newcastle disease virus (NDV) induction in L929 cells (31), whereas the IFN-A4 gene is highly induced under the same conditions (25). The repression of the IFN-A11 gene is due in part to a single point substitution in the IE. This substitution inhibits the binding of VIF (27). On the other hand, negative regulatory sequences were found in the promoter of the IFN-A11 gene surrounding the IE that suppressed in part its expression (4). These data suggest that both substitutions in the IE and the presence of negative regulatory regions contribute to the differential expression of IFN-A genes. In our previous studies, we have demonstrated the presence of a silencer E1E2 (Ϫ241/Ϫ203) region in the IFN-A11 promoter (28).
In this report, we study a corresponding E1E2 region in the highly inducible IFN-A4 promoter, which is protected from nuclease digestion by nuclear extracts. Transient transfection assays in L929 and HeLa cells with both IFN-A11 and -A4 E1E2 elements revealed identical negative regulatory activity within promoter constructs and the presence of a region in the IFN-A4 native promoter that overcame the silencer activity. A minimal distal negative regulatory element (DNRE) within the E1E2 regions in both gene promoters has been delimited by transfection assays. Electrophoretic mobility shift analysis (EMSA) and UV cross-linking, with the DNRE containing an AT-rich region, characterized one of the binding nuclear proteins as being related to, or identical with, the HMG I(Y). Since HMG I(Y) proteins are not transcriptional regulators themselves and mutated DNRE elements able to prevent the binding of HMG I(Y) do not abrogate silencer function, our results show that the DNRE, which has no demonstrable homology with known binding sites for transcription repressors, is specifically the binding site for a constitutive nuclear factor related to the silencer activity.

EXPERIMENTAL PROCEDURES
DNase I Footprinting Assays-The Mu IFN-A11-(Ϫ260/Ϫ192) and Mu IFN-A4-(Ϫ274/Ϫ206) gene promoter fragments (28) cloned into the SalI-blunted site of pBLCAT2 (32) were 32 P-3Ј-end-labeled by Klenow filling either at the BamHI site for the coding strand or at the HindIII site for the noncoding strand. For mutated gene promoters, fragments DM2, DM4, and DM6 were used. The DNase I footprinting experiments were carried out as described in the SureTrack protocol (Pharmacia Biotech Inc.).
Plasmid Constructions-Deletions and mutations in the native promoters of the IFN-A11 and -A4 were made by either the simple or nested polymerase chain reaction method using plasmids already described (4,28,31) or previous polymerase chain reaction products as the templates. Fragments were inserted (either PstI blunt/BamHI or PstI/ BamHI) into the corresponding sites of pBLCAT3 (32). All constructions were checked by nucleotide sequencing on double-stranded DNA template. For all constructions, the BamHI oligonucleotide (5Ј-CCCA-GATCTGGATCCTCT-3Ј) was used as 3Ј primer. For the 5Ј deletions, additions of IFN-A11 or IFN-A4 promoter fragments in IFN-A4 or IFN-A11 promoters and internal deletions in the IFN-A11 and -A4 promoters, which were made by polymerase chain reaction (for positions, see Figs. 2, and 3, A and B), different oligonucleotide primers from Eurogentec were used (data not shown). For the internal mutations in the IFN-A11 promoter, different mutated oligonucleotides were used (for mutated nucleotides see Fig. 3, C and D).
DNA Transfection, Viral Induction, and CAT Assays-L929 and HeLa S3 cells (six-well tissue culture plates), seeded in Dulbecco's modified Eagle's medium supplemented with 10% horse serum, were transfected at 50% confluence by the calcium phosphate precipitation method (33) with 1.25 g of reporter plasmid in the presence of 250 ng of pEF-LacZ plasmid (28), which was followed 4 h later (in the case of L929 cells) by a 1-min glycerol (10%) shock. NDV induction was carried out 48 h later as described (4). The mock-induced cells were set up as above except that no NDV was added. Cells were harvested 24 h postinduction, and cytoplasmic extracts were prepared by cell lysis with a buffer containing 0.65% Nonidet P-40, 10 mM Tris-HCl, pH 8.0, 1 mM EDTA, 150 mM NaCl, 0.5 mM phenylmethylsulfonyl fluoride. A CAT assay was carried out as described previously (34). Transfection effi-ciency was determined by the ␤-galactosidase activity assay with the Galactolight TM (Tropix, Inc.) chemiluminescent kit. In each experiment, a given construction was transfected in duplicate, and two different clones of each construction were tested.
Nuclear Extract Preparation and Gel Retardation Assays-After NDV induction for 1 h, the cells were treated for 5 h in the presence of serum with cycloheximide (CHX; 50 g/ml), and the medium was replaced with fresh medium. The mock-induced cells were set up as above except that no NDV was added. At different time points, nuclear extracts from NDV plus CHX-as well as CHX-induced or uninduced L929 and HeLa cells were prepared as described previously (35) in the presence of a mixture of protease inhibitors (pepstatin, leupeptin, antipain, aprotinin, benzamidine, chymostatin, and phenylmethylsulfonyl fluoride). The partially enriched 0.4 M KCl fraction of uninduced L929 nuclear extracts was eluted from heparin-Sepharose chromatography.
Gel retardation assays were performed as described (36) with some modifications. Nuclear extracts (2 g) were preincubated with 1 g of poly(dI-dC)⅐poly(dI-dC) or poly(dG-dC)⅐poly(dG-dC) in the presence of 100 ng of sonicated salmon sperm DNA (Pharmacia) for 10 min on ice. The mixture was then added to the binding buffer containing 10 fmol of 32 P-end-labeled probe (50,000 cpm, 0.1 ng) either with or without specific competitor oligonucleotide in a final volume of 20 l, and the incubation was carried out for a further 30 min at room temperature.  The shaded and white boxes correspond, respectively, to the IFN-A4 and IFN-A11 promoters. Since the data are pooled from several experiments, they are presented in arbitrary units of CAT activity. CAT activity values represent CAT/␤-galactosidase activity ratios relative to the induced activity Ϫ119A4wt, which was arbitrarily set at 100%. CAT activities for each plasmid are the means Ϯ S.E. for at least five separate transfections with at least two separate plasmids. Error bars indicate S.E. values. Inducibility (FOLD) is the ratio of the NDV-induced activity over the mock-induced activity. B, E1E2 and 4E1E2 repression of IFN-A11 promoter. CAT expression in cells transfected was relative to the induced activity of Ϫ119A11wt, which was set at 100%. Assay conditions are as described in the legend to Fig. 2A. C, E1E2 and 4E1E2 repression of IFN-A4 promoter in the presence of 4D substituted by 11D. CAT expression in transfected cells was relative to the induced activity of Ϫ119A4wt, which was set at 100%. Assay conditions are as described in the legend to Fig. 2A. D, E1E2 and 4E1E2 derepression of IFN-A11 promoter in the presence of 11D substituted by 4D. CAT expression in transfected cells was relative to the induced activity of Ϫ119A11wt, which was set at 100%. Assay conditions are as described in the legend to Fig. 2A. E, nucleotide sequences of the 4D and 11D regions. The nucleotide substitutions are underlined.
Gel shift assays with recombinant human HMG I protein (rhuHMG I) and purified polyclonal rabbit IgG anti-rhuHMG I antibodies (37) were performed under the same conditions, except that the antiserum and preimmune IgG were preincubated with the proteins for 30 min at room temperature prior to the addition of the probe. Complexes were resolved by electrophoresis on prerun 5% native polyacrylamide gels. After drying, the gels were autoradiographed overnight.
Heat-treated extracts were prepared by incubating the nuclear extracts for 5 min at 40 -80°C and for 15 min at 80°C. Competition with distamycin (Sigma) was achieved by preincubating the extracts with 1 l of the drug at 1 nM, 100 nM, 500 nM, and 1 M for 10 min and then by adding extracts or rhuHMG I and EMSA as described above.
UV Cross-linking Experiments-UV cross-linking of nuclear proteins of uninduced L929 nuclear extracts to the DNRE DNA was performed as described (38). DNRE is obtained by annealing the aaaactgcagATT-TAAGTCTAATTTAAAGT coding and the acgACTTTAAATTAGACT-TAAATctg noncoding strands, thus generating 5Ј and 3Ј overhangs. The duplex molecules were blunt-ended by Klenow filling, using bromodeoxyuridine and labeled [␣-32 P]dCTP and [␣-32 P]dGTP nucleotides.

Characterization of a Similar Distal Silencer Region in IFN-A11 and IFN-A4 Promoters-
The E1E2 (Ϫ241/Ϫ203) region was described to act as a silencer in the murine IFN-A11 promoter in L929 cells (28). Among other IFN genes examined, the murine IFN-A4 promoter contains an E1E2 element with only three nucleotide substitutions as compared with the IFN-A11 E1E2 sequence (Fig. 1B). The effect of these substitutions was studied by DNase I footprinting experiments and transient transfection assays in L929 and HeLa cells.
DNase I footprinting analysis of IFN-A11 and -A4 promoters exhibited similar protection (Ϫ241/Ϫ203 for IFN-A11 and Ϫ255/Ϫ218 for IFN-A4) with nuclear extracts from uninduced and NDV-induced L929 cells on both the coding and noncoding strands ( Fig. 1, A and B). We have designated the protected region in the IFN-A4 promoter as 4E1E2.
To determine if the 4E1E2 and E1E2 sequences had a comparable repressing effect, different constructions in the native IFN-A4 promoter were generated and tested in transient transfection assays in L929 and HeLa cells. In the experiments depicted in Fig. 2A (constructions 1-5), 5Ј deletions of the native IFN-A4 promoter, including a deletion the 4E1E2 element ( Fig. 2A, compare constructions 2 and 3), did not significantly modify the virus-induced transcriptional activity and inducibility. As described previously (25), the deletion of a region containing a G-rich cluster (Ϫ159 to Ϫ136) slightly decreased the inducibility of the promoter (Fig. 2A, compare constructions 4 and 5). However, 4E1E2 (as did E1E2) when independently inserted upstream of the proximal promoter of the IFN-A4 Ϫ119A4wt, showed a similar repression ( Fig. 2A,  compare constructions 5-7). These results suggest that the three point substitutions between 4E1E2 and E1E2 did not modify the repression activity, whereas regions surrounding 4E1E2 in the native promoter may contain regions able of overcoming the silencing effect of 4E1E2 in the IFN-A4 gene. In defining more precisely those regions that appear to prevent 4E1E2 from negatively regulating the IFN-A4 promoter, various internal deletions that modified neither the G-rich cluster nor the virus-responsive proximal region were carried out. As shown by the 5Ј deletion of the 4E1E2 element ( Fig. 2A, compare constructions 2 and 3), an internal deletion 4E1E2 did not modify the transcriptional activity of the promoter ( Fig. 2A,  construction 8). The 5Ј deletion of the upstream region of the 4E1E2 element (Ϫ470 to Ϫ258; Fig. 2A, compare constructions 1 and 2) was shown to have no effect, while the internal deletion of a downstream region of 4E1E2 element was able to modify the activity of the promoter. Indeed, when the region named 4D (located between Ϫ212 and Ϫ161; see also Fig. 2E) was deleted, a repressing effect was observed ( Fig. 2A, construction 9). It appears that the 4D region does not function as an activator but rather as an inhibitor of the silencer element 4E1E2 in IFN-A4 gene expression. The 5Ј deletion experiments and results described previously (25) suggest that 4D does not contain activator element(s) by itself. No significant modification in activity was observed after its deletion in the absence of the silencer 4E1E2 element (Fig. 2A, compare constructions 1,  3, and 4). In the absence of 4D, 4E1E2 appears to mediate the repressing activity, since subsequent internal deletion of 4E1E2 restores the full activity of the promoter ( Fig. 2A, compare constructions 1, 9, and 10).
The functional features of the 4E1E2 and 4D sequences on promoter activity were further investigated by addressing the question of whether the corresponding elements could have the same effect in the native IFN-A11 promoter. Several lines of evidence seem to indicate that the IFN-A11 and IFN-A4 promoter context do not display the same effect. Indeed, in contrast to 5Ј deletions of the native IFN-A4 promoter, the experiments depicted in Fig. 2B show that 5Ј deletions of the E1E2containing fragment cause a derepression of the virus-induced transcriptional level and inducibility in native IFN-A11 promoter (Fig. 2B, compare constructions 2 and 3). The constructions containing E1E2 or 4E1E2 directly inserted upstream of the proximal promoter of the IFN-A11 Ϫ119A11wt showed a similar repression (Fig. 2B, compare constructions 5-7). The individual E1E2 and 4E1E2 elements are able to repress both the IFN-A11 and IFN-A4 virus-induced proximal promoters. In contrast to the internal deletion of 4E1E2 in the IFN-A4 promoter, internal deletion of E1E2 in the context of the IFN-A11 promoter derepresses the native promoter (Fig. 2B, compare constructions 1 and 8). The region corresponding to 4D in the IFN-A11 promoter 11D (located between Ϫ198 and Ϫ147; see also Fig. 2E) does not seem to have any effect either by itself or on the repressive effect of E1E2. The 5Ј-deleted construction (Fig. 2B, compare constructions 3 and 4) as well as the construction with internal deletion of 11D (Fig. 2B, compare constructions 1 and 9) did not modify the activity of the promoter of the IFN-A11 gene. The 11D region in the IFN-A11 promoter, in contrast to the 4D region in the IFN-A4 promoter, does not overcome the silencing effect of E1E2.
The individual contributions of 4E1E2, 4D, E1E2, and 11D to IFN-A4 and IFN-A11 promoter activities were assessed by performing single and double substitutions between these regions in the two promoters. No significant improvement was observed by a single substitution that replaced 4E1E2 by E1E2 in the native promoter of the IFN-A4 gene (Fig. 2C, construction 2), whereas, when the 4D region was substituted by 11D, the repressing effect was released (Fig. 2C, construction 3). Similar results were obtained when both 4E1E2 and 4D were substituted by E1E2 and 11D (Fig. 2C, construction 4). Several 5Ј deletions of different substituted IFN-A4 promoters were also performed and gave results consistent with those from previous deletion analyses of native promoters (data not shown). These findings confirm that 4E1E2 and E1E2 negatively regulate promoter activity of IFN-A4 but only in the absence of the 4D region or in the presence of the 4D region  FIG. 3. Repression of IFN-A11 and -A4 promoters by a minimal DNRE. A, determination of DNRE by deletions within the E1E2. For experimental conditions, see the legend to Fig. 2A. Transfections were relative to the induced activity of Ϫ119A11wt, which was set at 100%. B, effect of DNRE on transcription of native proximal IFN-A4 promoters. For experimental conditions, see the legend to Fig. 2A. Transfections were substituted by 11D. On the contrary, when the 11D region was substituted by 4D in the presence of E1E2 or 4E1E2 in the promoter of the IFN-A11, the repressive activity was abolished (Fig. 2D, constructions 3 and 4). Single substitution between E1E2 and 4E1E2 did not modify the repressing activity (Fig.  2D, construction 2).
All of above experiments suggest identical silencing activity of the E1E2 and the 4E1E2 elements of both promoters and indicate that the 4D region overcomes the silencer activity in the natural context of the IFN-A4 gene. To understand the mechanism of this silencing we first tried to define more precisely the repressor-binding site of the silencer.
Identification of a Minimal DNRE within the IFN-A11 and -A4 Promoters-To delineate a minimal negative regulatory element within the E1E2, different 5Ј and internally deleted fragments of the native IFN-A11 promoter were generated and tested in transient transfection experiments for reporter gene expression in L929 cells (Fig. 3A). The results of the transfection assays (Fig. 3A, constructions 1-9) indicate that the IFN-A11 promoter domain corresponding to the 20-mer core sequence located from Ϫ228 to Ϫ209 within E1E2 (DNRE) is responsible for the distal negative regulatory activity.
The functional properties of the DNRE observed within the IFN-A11 promoter were studied in the corresponding element 4DNRE (Ϫ242 to Ϫ223) in the IFN-A4 promoter. The short 20-base pair elements DNRE and 4DNRE were inserted directly upstream of the Ϫ119/ϩ19 promoter of both the IFN-A11 and IFN-A4 genes. In the experiments depicted in Fig. 3A (compare constructions 6, 10, and 11) and in Fig. 3B, the DNRE and 4DNRE showed a repression of the virus-induced transcriptional level and inducibility of both promoters. These results demonstrate that DNRE and 4DNRE can repress transcription of the two proximal IFN-A promoters (Ϫ119 to ϩ19) to the same extent. Similar results were obtained in HeLa cells (data not shown). In the absence of the surrounding sequences in the IFN-A4 promoter, these DNRE and 4DNRE cis-elements behave similarly.
Study of the DNRE-The sequence of the DNRE reveals the presence of partially overlapping direct (R2, 8/9 nucleotide homology) and inverted (R2X, 5/5 nucleotide homology) tandem repeats and palindromic sequences from position Ϫ228 to Ϫ209 (Fig. 3D). This element does not show any homology with already known binding sites for transcription repressors, but it contains two AT-rich sequences that may be binding sites for HMG I(Y). A related but not identical consensus recognition DNA sequence (IRF-Es) for IRF-1 and IRF-2, G(A)AAA(G/C)(T/ C)GAAA(G/C)(T/C) (8) is present in E1E2 (TAAAGTGAAACC) and overlaps part of the DNRE. Furthermore, a variant hepatocyte nuclear factor-1-binding site (38) is present 5Ј to the DNRE (GTGGTTAATGA). These potential DNA-binding sites were tested within the Ϫ244/ϩ19 promoter by mutation. Mutations were chosen for their inability to create new binding sites for already described factors. These mutations included the following (Fig. 3D): the related IRF-Es, which was mutated as described previously (39,40) without affecting the DNRE (IM1, IM2, and IM3); the three bases C, T, and A, which separate the palindromic half-sites (DM1); either one (DM3 and DM4) or both (DM2) palindromic half-sites (these mutations also affect the tandem repeats); either one (DM6 and DM4) or both (DM5) tandem repeats (these mutations also affect the palindrome; either one (DM7 and DM9) or both (DM8) AT-rich sequences; and, as a control, the variant hepatocyte nuclear factor-1-binding site (HM). The results after transfection in L929 cells are shown Fig. 3C. The mutations affecting the IRF-Es (IM1-IM3) are not able to release the DNRE effect. The integrity of the palindromic structure (and also that of the tandem repeats) does not seem to be required, since some mutations affecting these structures are not able to abolish the DNRE silencing effect (DM3 and DM6). When the AT-rich sequences were perturbed, no effect on the repression was observed (DM7-DM9). Nevertheless, the transcriptional activity of the Ϫ244/ϩ19 promoter was released by some of the mutations that affected nucleotides present in the core sequence of the DNRE e.g. DM2, DM4, and DM5. Similar results were obtained in HeLa cells (data not shown). Together, these results suggest that the nucleotides involved in the repressive effect may define a specific DNA binding sequence.
Identification of Nuclear Factors Binding to DNRE-The negative effect observed on promoter activity is due to the binding of trans-repressor(s) to the E1E2 as demonstrated by previous ex vivo transfection-competition analysis (28). Factors binding to DNRE and 4DNRE were identified by EMSAs with nuclear extracts from uninduced and NDV plus CHX-as well as CHX-induced L929 and HeLa cells. The major complexes I and II (complex II actually contains two closely migrating complexes) were observed under different conditions using DNRE as probe. When poly(dI-dC)⅐poly(dI-dC) and sonicated salmon sperm DNA were used as nonspecific competitors, only complex I was observed (Fig. 4A, lanes 1-7). In the presence of poly(dG-dC)⅐poly(dG-dC) and sonicated salmon sperm DNA, complex I and the faster mobility complex II are detected (Fig. 4A, lanes  8 -14), whereas the formation of complexes I and II is inhibited by the use of poly(dA-dT)⅐poly(dA-dT) (data not shown). The partially enriched 0.4 M KCl fraction of uninduced L929 nuclear extracts eluted from heparin-Sepharose chromatography gave similar patterns of complexes (data not shown). Complexes I and II are constitutively present before virus induction and were still maintained even following induction by NDV plus CHX or by CHX. The specificity of the formation of complexes I and II was demonstrated using unlabeled DNRE element in the presence of poly(dG-dC)⅐poly(dG-dC) (Fig. 4A, lanes 15-18).
In the presence of poly(dI-dC)⅐poly(dI-dC), similar results concerning the specificity of the formation of complexes I were obtained (data not shown). On the other hand, EMSAs were also carried out using uninduced and NDV-induced HeLa cells nuclear extracts, and similar patterns were observed (Fig. 4B). The specificity of both complexes using HeLa cell extracts was also demonstrated (data not shown). The same specific complexes were detected using 4DNRE as probe (data not shown).
To correlate the DNRE binding of nuclear proteins with the ability of DNRE to repress transcription, we analyzed the binding properties of the DNRE mutants DM1-DM9, using them as probes in EMSA. The mutated DNRE fragments DM2, DM4, and DM5 (which are able to release the native promoter activity in transfection assays) when used as probes prevent the formation of the complexes I and II (Fig. 5). The other mutated DNRE fragments (which are unable to release the promoter activity) show differences in EMSA. Whereas mutants DM1, DM3, and DM9 showed a binding pattern identical or similar to that of wild-type DNRE, mutants DM6 -DM8 allowed the formation of complex I but not of complex II. When mutant DM6 was used, a faint band at the faster mobility complex level was relative to the induced activity of Ϫ119A4wt, which was set at 100%. C, effect of mutations within DNRE. For experimental conditions, see the legend to Fig 1-7) or 1 g of poly(dG-dC)⅐poly(dG-dC) (lanes 8 -14) in the presence of 125 ng of sonicated salmon sperm DNA and nuclear extracts. Extracts were prepared from L929 cells (lanes 1-14) uninduced or induced with NDV plus CHX as well as with CHX for the indicated periods. Lanes 15-18, specific formation of complexes I and II. Uninduced L929 nuclear extracts were incubated with a 0, 10, 50, or 100-fold molar excess of unlabeled wild-type DNRE. B, extracts were prepared from HeLa cells uninduced or induced with NDV plus CHX as well as with CHX for the indicated periods.
contains AT-rich domains that may be putative sites for the binding of the nonhistone chromatin-associated proteins HMG I(Y), we compared some of the properties of complexes I and II with properties of these proteins. The high electrophoretic mobility of complex II suggested also that the proteins present in complex II may be related to HMG I(Y) proteins. Furthermore, in the case of HMG I(Y) implicated in complex II, we wanted to test more precisely if complex II formation modifies the formation of complex I or if HMG I(Y) can also be detected as a part of complex I.
HMG proteins are heat-stable and bind DNA in the minor groove. The relative thermostability was examined by heating the nuclear extracts for 5 min at various temperatures or for 15 min at 80°C. Under these conditions, complex II was still maintained after heating at 80°C for 5 min and to a lesser extent after 15 min (data not shown). Complex I was undetectable after heating at these temperatures. Since distamycin, a minor groove-binding drug, blocks DNA binding of HMG proteins, we tested the effect of distamycin on the DNRE binding activity. rhuHMG I was used to analyze its ability to bind to DNRE. The promoter of the lymphotoxin gene containing an HMG I(Y)-binding site (LT) was used as a control (Fig. 6A, lane 1) as described previously (37). The DNRE probe (Fig. 6A, lane  3) is recognized by rhuHMG I, and the use of distamycin blocks both the LT and the DNRE binding activities (Fig. 6A, lanes 2  and 4). With nuclear extracts from uninduced L929 cells using the DNRE as probe, the formation of complex II was inhibited by distamycin (Fig. 6A, lanes 11 and 12), and similar results were obtained with the major complex formed using LT as probe (Fig. 6A, lane 7). At a similar concentration, the drug had no inhibitory effect on the formation of complex I (Fig. 6A, lanes  11 and 12). In contrast, the distamycin seemed to weakly increase the formation of complex I, which may be explained by the use of the drug by itself (Fig. 6A, compare lanes 9 and 10).
In conclusion, the formation of complex I was neither directly inhibited by distamycin nor dependent on the presence of complex II.
These results suggest that complex II may contain factors related to the HMGs. A purified polyclonal rabbit IgG anti-HMG I(Y) antibody was used to determine whether the complexes were immunologically related to HMG I(Y). In EMSA with nuclear extracts from uninduced L929 cells using the DNRE as probe, the anti-HMG I(Y) antibody (Fig. 6B, lanes  11 and 13), but not the preimmune serum (Fig. 6B, lanes 12  and 14), was able to inhibit or supershift the formation of complex II. When LT was used as a probe, similar results were obtained with the major complex formed (Fig. 6B, lane  8) and rhuHMG I (Fig. 6B, lanes 1-6). The blocking of complex II formation by the antibody suggests an antigenic relation between complex II and HMG I(Y). The rhuHMG I protein (Fig. 6A, lane 3, and 6B, lane 4) yielded a complex with DNRE as a probe that did not comigrate exactly with complex II (Fig. 6, A, lane 8, and B, lane 10). Probably, these slight differences in electrophoretic mobilities are caused by secondary biochemical modifications of the HMG I(Y) proteins that occur in vivo in mammalian cells (e.g., phosphorylations) but that are absent from the bacterially produced recombinant proteins. It seems likely that the doublet seen in the complex II from nuclear extracts is the result of binding to the slightly differently sized HMG I and HMG Y isoforms of the protein found in mammalian cells. It should be noted that, under the same conditions of EMSA with several dilutions of the antiserum, complex I was not displaced by the antibody (Fig. 6B, lanes 11 and 13, and data not shown). Complex I does not appear to be immunologically related to the HMG I(Y) proteins. The DNA binding affinities of both complexes were also measured, and we have observed that complex II protein(s) bind to DNRE with a low affinity, whereas those of complex I bind to DNRE with a high affinity (data not shown).
The DNA binding ability of rhuHMG I was measured using DNRE and the DNRE mutants DM1-DM9 as probes and compared with that of the DNA-protein complex formation corresponding to complex II. The pattern of binding of HPLC-purified rhuHMG I to these probes (Fig. 6C) is similar or identical to the pattern of formation of complex II previously shown (Fig.  5). The presence or absence of the DNA-protein complex II formation and the binding of rhuHMG I using DNRE and mutants DM1-DM9 as probes may be explained by the properties of interactions of HMG I(Y) proteins with DNA. The HMG I(Y) proteins contain three DNA binding motifs known as AT hooks (41), which have been shown to bind to two or three appropriately spaced AT tracts in DNA (42). DNRE includes two AT tracts, and mutations that disrupt one of them such as in DM4, DM6, and DM7 are sufficient to inhibit totally the binding of HMG I(Y). One of the mutations within the 3Ј AT tract of DNRE (DM9) poorly affects the affinity of binding of rhuHMG I to the probe. This result is in agreement with previous work showing that five AT nucleotides within two AT tracts are sufficient for the binding of HMG I(Y) (42).
In conclusion, complex II contains a protein related to or identical to HMG I(Y). Complex I appears to be necessary and sufficient for repressive activity and does not seem to contain any protein(s) related to the HMG I(Y). HMGI(Y) proteins seem to be dispensable for the repressive activity. However, the precise relationship between both complexes remains to be established by isolation of complex I. DNRE Mutants DM2 and DM4 Alter the DNase I Footprint Pattern of the Wild Type Promoter-Because some of the mutations appear to prevent the repression of the IFN-A11 promoter by the DNRE, it was important to determine if the absence of the DNRE-binding factors using EMSA assays had any influence on the binding pattern of the wild type promoter and the mutated promoters in footprinting experiments. The mutations in DM2 and DM4 (but not in DM6) significantly alter the DNase I footprint pattern of the core sequence of the DNRE (Fig. 7, A and B). However, the mutations in DM2 and DM4 did not completely abolish the footprint at E1E2, suggesting that region(s) surrounding the DNRE may be able to bind different factors from those binding the DNRE. Nevertheless, it is the DNRE element that acts on repression, and the additional proteins that bind E1E2 do not themselves seem to act as repressors.
A DNRE-binding Protein of 38 kDa Present in Complex I-The molecular sizes of the DNRE-binding proteins composing complex II and complex I were determined by UV crosslinking using nuclear extracts from uninduced L929 cells (Fig.  8, A and B). Only one protein of approximately 15 kDa in size was detected in complex II (Fig. 8B). This is similar in molecular size to the rhuHMG I protein used as a control. All of the experiments described above suggest that the complex II is related or identical to HMG I(Y).
Complex I appeared to contain a major 38-kDa nuclear protein (Fig. 8A). We suggest that this major protein could be related to the silencer activity. The isolation of this protein is now necessary to analyze its involvement in the distal repression of the IFN-A genes. DISCUSSION In the present study, we have within the IFN-A11 promoter, between Ϫ228 and Ϫ209, a silencer element (DNRE) that is necessary and sufficient for the distal repression of the IFN-A11 gene after NDV induction in both L929 and HeLa cell lines. The isolated DNRE of the IFN-A11 promoter or the similar element 4DNRE found in IFN-A4 is able to reduce the inducibility of both the native proximal (Ϫ119/ϩ19) promoters of the IFN-A11 and -A4 genes. These results suggest that DNRE may exert a more general modulation of the transcriptional strength of the promoters of the IFN-A genes. Among eukaryotic gene promoters examined (Eukaryotic Promoter Database), in the 20 best scores, the DNRE of the murine IFN-A11 gene sequence is highly homologous to a sequence within the promoters of two other IFN-A genes that are different from the murine IFN-A4: the human IFN-A6 (Ϫ387 to Ϫ368, ATTTAACTTTTAGTTAAATT, 75.0% identity in a 20nucleotide overlap in the coding strand; and Ϫ388 to Ϫ369, ATTTAACTAAAAGTTAAATT, 75.0% identity in a 20-nucleotide overlap in the noncoding strand) and the murine IFN-A7 (Ϫ275 to Ϫ258, ATCTAA-TCTAATATAAAG, 84.2% identity in a 19-nucleotide overlap in the noncoding strand) cloned in our laboratory (43). In these two cases, the genes are weakly inducible or uninducible after virus induction. Indeed, the human IFN-A6 has been described as an uninducible gene in several cell types and with different virus (44). Human IFN-A6 may be in fact a pseudogene with regard to its promoter, since the gene has a deletion of 12 nucleotides, from Ϫ73 to Ϫ61, within the Fig. 6. Relationship between complex II and HMG I(Y) proteins. A, a minor groove-binding drug was tested by preincubating the DNRE probe with 1 l of 1 nM, 100 nM, 500 nM, and 1 M of distamycin before binding. B, complex II contains factors immunologically related to HMG proteins. rhuHMG I (lanes 1-6) or nuclear extracts from uninduced L929 cells (lanes 7-14) were used in the presence of LT probe (lanes 1-3 and lanes 7-9) and DNRE probe (lanes 4 -6 and lanes 1O-14). rhuHMG I or nuclear extracts were preincubated with the amounts indicated (1:10 and 1:20 dilutions), or a 1:10 dilution if not indicated, of anti-HMG I(Y) antibody (lanes 2, 5, 8, 11, and 13). Preimmune serum was used as the negative control (lanes 3, 6, 9, 12, and 14). C, binding of HPLCpurified rhuHMG I to wild-type and point-mutated DNRE probes. HPLC-purified rhuHMG I was incubated with the DNRE and DNRE mutant (DM1-DM9) probes and 1 g of poly(dG-dC)⅐poly(dG-dC) in the presence of 125 ng of sonicated salmon sperm DNA. presumed virus-responsive element. In contrast, the murine IFN-A7 gene may be repressed by the existence of a silencer element such as DNRE. This gene is poorly expressed or not expressed after NDV induction in several cell lines, and transient transfection assays in L929 and HeLa cell lines have demonstrated that its promoter is weakly inducible by NDV (data not shown). The participation of the DNRE in the silencing of the transcription of the murine IFN-A7 gene in different cell types remains to be elucidated. It is interesting to note that other gene promoters have been shown to contain sequence homologies with the DNRE silencer element of the IFN-A11 and -A4 promoters. Therefore, homologies were found with the canavalin, the serum albumin, the adenovirus 7 E1b, the globin, and the interleukin-4 promoters. The repressive effect of DNRE-related factors, suggested by this IFN-A study, may participate in the transcriptional modulation of these different promoters.
The mutations in DNRE responsible for the derepression of the transcriptional activity of the IFN-A11 promoter also alter the formation of complexes using nuclear extracts from uninduced or virus-induced L929 and HeLa cells. These results indicate the presence of constitutive binding proteins involved in the silencing effect. One of the complexes corresponds to a protein related to or identical to HMG I(Y) but does not seem to modulate the binding to DNRE of a factor present in uninduced and virus-induced nuclear extracts with a molecular mass of 38 kDa. HMG I(Y) has been implicated in the regulation of a number of genes, including the IFN-B (19), E-selectin (45,46), interleukin-4 (47), interleukin-2 receptor ␣-chain (48), lymphotoxin (37), herpes simplex virus type 1, and papovavirus JC virus (49,50) genes. To date, HMG I(Y) proteins have not been shown to be implicated in the regulation of the expression of the IFN-A genes. In human IFN-B gene expression, HMG I(Y) proteins facilitate the binding of both NF-B and activating transcription factor-2 to mediate virus induction and play an essential role in the assembly of a higher order transcription enhancer complex named enhanceosome (51). In the case of DNRE, it seems that HMG I(Y) does not modify the binding of the 38-kDa protein. Indeed, complex I is also formed in the absence of complex II. Furthermore, the formation of complex II but not complex I was inhibited by anti-HMG I(Y) antibody and distamycin. In these experiments, it seems that the formation of complex I may not be dependent on the presence of complex II. Furthermore, we have shown a strict correlation between the activity of DNRE ex vivo and its capacity to form complex I but not complex II in vitro. Nevertheless, our observations cannot exclude completely the possibility that HMG I(Y) may account for part of the repression of the IFN-A genes and may interact with the 38-kDa protein. This hypothesis will be explored when this latter protein will be isolated.
Recently, it has been demonstrated in the interleukin-2 receptor ␣-chain promoter that both Elf-1 and HMG I(Y) can be detected in the same complex in EMSAs and that they can interact with each other in the absence of DNA (48). Our results from EMSAs in the presence of antibodies against HMG I and from cross-linking experiments suggest that HMG I(Y) is not a component of complex I. Another function of HMG I(Y) could be to alter the structure of the bound DNA, which may result in its bending, bringing distant regulator factors, already bound to the elements, into closer proximity, thus facil- itating interaction between these factors and therefore bringing regulating elements closer to the start site (52). Whereas an increase of positive transcriptional factors may result in stronger recruitment of the basal transcriptional machinery, the opposite effect may be observed when negative transcriptional factors are involved.
For most of the IFN-A promoters as well as the IFN-B promoter, positive activator proteins bind to the proximal virusresponsive element and can act to overcome the constitutive silencing of the genes and allow high levels of transcription initiation after virus induction. Here we show that novel elements located upstream from the IE of the IFN-A promoters may also modulate the transcription of the genes. Indeed, the same distal silencer elements DNRE and 4DNRE are present in both IFN-A4 and IFN-A11 promoters, and similar DNAbinding proteins may be implicated in the negative activity. However, the presence of a cis-acting region, 4D, located between the 4DNRE and the IE of the IFN-A4 promoter, overcomes the silencer activity and could be considered as an antisilencer. Antisilencer or antirepressor elements in different genes have been identified by their ability to override a negative effect and restore gene transcription without themselves showing a true positive activity, since they function only in the presence of the silencers (53)(54)(55)(56)(57). Multiple upstream silencer elements as well as a unique antisilencer element have been shown to be responsible for regulating the vimentin gene (53). During myogenesis, down-regulated vimentin gene expression is correlated with the increased binding activity of the silencer proteins and coordinated with the decreased binding activity of the antisilencer protein. Therefore, these upstream silencer and antisilencer elements correlate with the developmental expression pattern of vimentin.
In this report, we demonstrate that the presence of the 4D region overcame the silencer activity of the 4E1E2 element in the natural context of the IFN-A4 gene. Therefore, we suggest that the 4D region be considered as an antisilencer. The presence or absence of silencer and antisilencer elements may contribute to the differential expression of the IFN-A genes. We have searched for sequence homologies within the 4D antisilencer region that would be absent in the 11D region (see Fig.  2E), and we have found a perfect homology to the consensus recognition DNA sequence for IRF-1 and IRF-2, G(A)AAA(G/ C)(T/C)GAAA(G/C)(T/C) (IRF-Es), found within the promoters of IFN-A, IFN-B, and many IFN-inducible genes (8). The 4D antisilencer element, which contains the IRF-Es, GAAACT-GAAAGC (Ϫ187/Ϫ176), shares two nucleotide substitutions in the corresponding element within the nonfunctional 11D region, GAAAGAGAAAGC (Ϫ173/Ϫ162). Different genes of the IRF family have been cloned and characterized, but, to date, none of them has been described as an antisilencer (9, 12, 39, 58 -61). Experiments are in progress to elucidate the mechanism by which the 4D antisilencer element overcomes the silencer 4DNRE in the IFN-A4 gene promoter. The isolation of the 38-kDa DNRE-binding protein will be critical and could contribute in the precise characterization of the 4D cis-acting antisilencer element and the different trans-acting factors that may bind to the 4D and 11D regions. Indeed, protein-protein interactions between silencer and antisilencer as described for the carbamyl phosphate synthetase I promoter (55) may be suggested. Clearly, additional experiments are required for a more detailed understanding of the complex differential expression pattern of the IFN-A gene family, which involves a combination of the actions of different activator(s) and also silencer(s) and antisilencer(s).