The interferon (IFN)-stimulated gene Sp100 promoter contains an IFN-gamma activation site and an imperfect IFN-stimulated response element which mediate type I IFN inducibility.

Expression of the nuclear domain-associated proteins Sp100, PML, and NDP52, is enhanced by interferons (IFNs) on the mRNA and protein level. Increase both of Sp100 and PML mRNA is due to enhanced transcription of the corresponding genes which occurs independently of cellular protein synthesis immediately upon IFN-β addition. Here, we describe the molecular cloning and functional analysis of the Sp100 promoter. DNA sequence analysis revealed potential binding sites for several constitutive and IFN-inducible transcription factors. Consistent with the absence of a TATA box and an initiator element, several transcription initiation sites were found. Transient expression studies identified an imperfect IFN-stimulated response element within the first 100 nucleotides upstream of the major transcription start site. This element rendered a heterologous promoter IFN-β-inducible and bound IFN-stimulated gene factor 2 strongly but IFN-stimulated gene factor 3 only weakly. An IFN-γ activation site approximately 500 base pairs upstream of the IFN-stimulated response element was found to bind three IFN-α/β activation factors upon IFN-β induction and conferred both type I and type II IFN inducibility upon a heterologous promoter. These data demonstrate a novel arrangement of a nonoverlapping IFN-γ activation site and an IFN-stimulated response element mediating type I IFN inducibility, previously not reported for other IFN-stimulable promoters.

Expression of the nuclear domain-associated proteins Sp100, PML, and NDP52, is enhanced by interferons (IFNs) on the mRNA and protein level. Increase both of Sp100 and PML mRNA is due to enhanced transcription of the corresponding genes which occurs independently of cellular protein synthesis immediately upon IFN-␤ addition. Here, we describe the molecular cloning and functional analysis of the Sp100 promoter. DNA sequence analysis revealed potential binding sites for several constitutive and IFN-inducible transcription factors. Consistent with the absence of a TATA box and an initiator element, several transcription initiation sites were found. Transient expression studies identified an imperfect IFN-stimulated response element within the first 100 nucleotides upstream of the major transcription start site. This element rendered a heterologous promoter IFN-␤-inducible and bound IFN-stimulated gene factor 2 strongly but IFN-stimulated gene factor 3 only weakly. An IFN-␥ activation site approximately 500 base pairs upstream of the IFN-stimulated response element was found to bind three IFN-␣/␤ activation factors upon IFN-␤ induction and conferred both type I and type II IFN inducibility upon a heterologous promoter. These data demonstrate a novel arrangement of a nonoverlapping IFN-␥ activation site and an IFN-stimulated response element mediating type I IFN inducibility, previously not reported for other IFN-stimulable promoters.
Sp100 is a protein located in novel discrete nuclear domains ("nuclear dots," NDs) 1 and an autoantigen predominantly in patients with primary biliary cirrhosis (1)(2)(3)(4)(5). Experimental evidence suggests a transcriptional transactivating function for this protein (2,6). 2 Recently, the murine Sp100 gene (mSp100) was identified and found to be highly repetitive and variable in copy number (7,8). The expression of the Sp100 and mSp100 genes as well as of the genes of two additional ND-associated proteins, PML and NDP52 (9 -12) is greatly enhanced by type I (␣/␤) and type II (␥) interferons (IFNs) (7,(12)(13)(14)(15). Similar to Sp100, PML is also autoantigenic in primary biliary cirrhosis (4) and appears to be a transcriptional transactivator (16). It was discovered originally in patients with acute promyelocytic leukemia (APL), a hematopoietic malignancy in which the normal development of myeloid precursors is blocked at the promyelocytic stage. In APL cells the PML protein is fused to retinoic acid receptor-␣ sequences (17,18), and its expression leads to a redistribution of both the PML and Sp100 proteins. Since retinoic acid treatment of APL cells leads to normalization of the ND pattern and releases the differentiation block, it was speculated that disruption of the NDs causes APL (9 -11). The cell growth and tumor suppressing activities of PML support this speculation (19 -22). NDs and their associated proteins appear to play a role not only in tumorigenesis and autoimmunity but also in virus infection as changes in cellular localization are induced by various viral transcription factors (13,(23)(24)(25)(26). The autoimmunogenic character and putatively similar function of Sp100 and PML, the modulation of the NDs in tumorigenesis and virus infection, and the IFN-enhanced expression of all known ND proteins prompted us to investigate in more detail the mechanism of the transcriptional regulation of these genes. For Sp100 and PML, we could recently demonstrate that IFN-enhanced expression is mainly due to an increased transcription rate of the corresponding genes (14). However, from these and other studies (15), it has not become clear which factors and mechanisms are involved in IFN inducibility.
Activation of transcription by IFNs involves binding of the ligands to their receptors followed by tyrosine phosphorylation of the receptors, of specific tyrosine kinases, and of latent cytoplasmic transcription factor subunits called STATs (signal transducers and activators of transcription) (reviewed in Refs. [27][28][29]. The STATs are then assembled to form functional transcription factors which are translocated into the nucleus. Within their promoter regions IFN-␣/␤-inducible genes contain IFN-stimulated response elements (ISREs) (28) that interact specifically with three major complexes, the IFN-stimulated gene factors ISGF1, ISGF2, and ISGF3. ISGF3, the major primary transcriptional activator (30,31), consists of three ␣-subunits, p91 (STAT1␣), p84 (STAT1␤), and p113 (STAT2) that assemble in the cytoplasm upon phosphorylation (see above) and, together with the ISGF3␥ subunit p48, form transcriptionally active ISGF3. ISGF2, identical to IFN regulatory * This work was supported by a grant from the Bundesministerium fü r Forschung und Technologie. The Heinrich-Pette-Institut is supported by the Bundesministerium fü r Gesundheit and the Freie und Hansestadt Hamburg. 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.
The factor-1 (IRF-1) (32,33) also binds to the ISRE, but more slowly after IFN addition. Unlike ISGF3, ISGF2 binding requires ongoing protein synthesis (34). Currently, it is believed that ISGF2 functions as a positive transcriptional regulator which maintains IFN-enhanced transcription initiated by ISGF3 (35). ISGF1 is identical or at least contains IFN regulatory factor-2 (IRF-2) (36, 37) and appears to be a constitutive ISRE-binding factor that acts negatively on ISGF2 (36). In contrast to ISGF3, both ISGF1 and ISGF2 need only the ISRE core sequence 5Ј-TTCNNTTT-3Ј for binding (28). In contrast to type I IFNs, receptor-binding of IFN-␥ triggers tyrosine phosphorylation only of STAT1, which leads to homodimerization and subsequent translocation of the active transcription factor into the nucleus (reviewed in Refs. 27 and 29). This factor, IFN-␥ activation factor (GAF), and a cognate IFN-␥ activation site (GAS) were discovered originally through studies on the IFN-␣/␤and IFN-␥-induced transcription of the GBP gene (38 -40). Later on, GAS binding was also observed for an IFN-␣-inducible factor which was therefore termed IFN-␣ activation factor (AAF) (41). However, the discovery of the STAT proteins and the elucidation of the mechanisms of their activation strongly suggested that GAF and AAF are identical (42) and that a GAS may render a promoter responsible for both types of IFNs.
Here we describe the cloning and functional characterization of the Sp100 promoter region and the identification of sequence elements to which IFN-induced transcription factors bind. The results demonstrate that IFN-inducible Sp100 gene transcription is similar but not identical to the very well studied IFNinducible GBP gene and, therefore, represents a unique example of an IFN-regulated promoter.

EXPERIMENTAL PROCEDURES
Isolation of Genomic Sp100 DNA-For isolation of Sp100 promoter DNA fragments, a genomic library of the Burkitt's lymphoma cell line BL64 in vector EMBL3A (43) was screened with a full-length 32 Plabeled Sp100 cDNA probe (2). Positive plaques were further screened with 32 P-labeled oligonucleotides from the 5Ј-end of the Sp100 cDNA (Sp32, 5Ј-ATGGCAGGTGGGGGCGG-3Ј; Sp347, 5Ј-CCACTCTCTGTA-CAGGG-3Ј) according to standard protocols (44). The DNA of one of the positive clones was double-digested with SalI and XhoI, and all fragments were subcloned into SalI-digested pUC19. A plasmid containing the appropriate Sp100 promoter fragment was then identified by hybridization of the bacterial colonies with a fluorescein-11-dUTPlabeled oligonucleotide from the 5Ј-end of the Sp100 coding region (Sp61, 5Ј-GGTGCTCAGGTCGCCGCC-3Ј) and by enhanced chemiluminescence detection according to the manufacturer's protocol (Amersham Corp.). The nomenclature of the oligonucleotides corresponds to their 5Ј-positions in the Sp100-cDNA (2).
DNA Sequencing and Polymerase Chain Reaction (PCR)-Doublestranded DNA was sequenced using the Sequenase Version 2.0 kit (U. S. Biochemical Corp.) and 35 S-dATP, or by PCR cycle-sequencing with fluorescence dyes and use of a Li-Cor 4000 DNA sequencer. For PCR amplifications, buffers, reaction mixes, and cycling conditions were chosen as described elsewhere (45).
Preparation of Total and Poly(A) ϩ RNA-Total cellular RNA was prepared by the urea/LiCl method (46). Poly(A) ϩ RNA was purified from total RNA using Oligotex-dT spin columns (Qiagen).
Cell Culture, Cytokines, and Transient Transfection Assays-Hela S3, HEp2, and HuH7 (47) cells were maintained as monolayers in Dulbecco's modified eagle medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum. Human IFN-␤ and -␥ were used at concentrations of 1000 and 100 units/ml, respectively. Plasmid DNAs were introduced into cells by the calcium phosphate procedure (44). One picomole of Sp100 promoter CAT construct and 1 g of a CMV-lacZ construct, adjusted to 10 g of DNA with pUC19, were precipitated per 6-cm dish. For transfection of thymidine kinase/ISRE-and thymidine kinase/GAS-CAT constructs, 3 pmol of plasmid were used. Calcium phosphate-DNA co-precipitates were left on the cells overnight and removed by exchange of the medium, and cells were then incubated for a further 24 h in Dulbecco's modified Eagle's medium with or without IFN.
CAT ELISA-The amount of CAT protein was determined immunologically with a commercially available CAT ELISA kit (Boehringer Mannheim) according to the manufacturer's instructions.
Protein Extracts and Gel Mobility Shift Assays-Nuclear protein extracts were prepared essentially as described previously (48), except that 10 g/ml aprotinin, 5 g/ml leupeptin, and 2.5 g/ml pepstatin A were added to both hypotonic and high salt buffer, and additionally 0.1% Nonidet P-40 to the hypotonic buffer. After cell lysis on ice for 10 min, nuclei were collected by centrifugation in a microcentrifuge for 1 min at 4°C (13,000 rpm), resuspended in high salt buffer, and incubated on ice for 20 min. After centrifugation for 10 min at 4°C (13,000 rpm) the supernatant containing the soluble nuclear proteins was recovered. For gel mobility shift assays (49, 50) 10 g of nuclear proteins were preincubated for 5 min at room temperature with 4 g of poly[d(I-C)] and 100 ng of unrelated single-stranded oligonucleotide in 25 l of 20 mM HEPES (pH 7.9), 50 mM NaCl, 1 mM MgCl 2 , 0.1 mM EGTA, 0.5 mM dithiothreitol, 0.1% Nonidet P-40, and 10% glycerol. After addition of 32 P-labeled double-stranded oligonucleotide (50,000 cpm, corresponding to about 5 fmol) and incubation for 20 min at room temperature, DNA-protein complexes were analyzed on 6% acrylamide (80:1) gels run in 0.25 ϫ TBE. For competition experiments unlabeled double-stranded oligonucleotides were used in an 100-fold molar excess.

RESULTS
Cloning and Sequence Analysis of the Sp100 Promoter Region-A genomic DNA fragment containing promoter sequences of the Sp100 gene was isolated from a genomic phage library of the Burkitt's lymphoma cell line BL64 by hybridization with a 32 P-labeled full-length Sp100 cDNA (2). An approximately 10-kilobase pair long restriction fragment of the insert was subcloned into pUC19. The promoter region of this insert, comprising 1114 bp upstream of the translational start codon of the Sp100 gene and 146 bp downstream thereof extending into intron I, was sequenced using synthetic oligonucleotides as primers (Fig. 1). Inspection of the sequence revealed neither a consensus CCAAT or TATA box nor a consensus initiator element that can functionally substitute a TATA box (51). However, we found the sequence element ACTTCN 18 GCCA at the transcription initiation sites (see below and Fig. 1) which diverges by only one nucleotide from the housekeeping initiator protein 1 (HIP1) binding site (ATTTCN 1-19 GCCA) described for several other TATA-less promoters (52). As HIP1 was shown to be sufficient for mediating transcription initiation at the ATTTC and/or GCCA site when combined with at least one binding site for transcription factor Sp1 (52,53) and as the Sp100 promoter also has a closely spaced Sp1 binding site (Fig. 1), these sequences are probably important for Sp100 promoter activity and transcription initiation. Similarly, potential binding sites for the transcription factor Ap1 (54) and for NF-B (55) could also play a role in promoter activity (Fig. 1). Three sequence motifs, two consensus GAS located at positions Ϫ655 to Ϫ647 and Ϫ811 to Ϫ803, and one with similarity to an ISRE at position Ϫ162 to Ϫ149, which may confer IFN inducibility upon the Sp100 gene, were also identified (Fig. 1). Both GAS motifs exactly match the consensus sequence (5Ј-TTCNNNGAA-3Ј) (56), whereas the ISRE diverges from the consensus sequence (5Ј-AGTTTC-NNTTTCNY-3Ј) (28) at three nucleotide positions (Ϫ161, Ϫ152, and Ϫ151), one of them (Ϫ161) being highly conserved. Taken together, computer analysis of the Sp100 promoter sequence revealed several features typical for promoters of housekeeping genes and three potential elements which may be relevant for IFN-enhanced transcription of the Sp100 gene.
Mapping of the Transcription Initiation Sites-A prerequisite for functional analysis of promoters is the mapping of the start sites of transcription. Therefore, primer extension analysis was carried out by reverse transcription of RNA from IFN-␤-treated or untreated HEp2 cells, using a 32 P-labeled oligonucleotide specific for Sp100 or, as a control, for ␤-actin as primers. Two distinct size classes of reverse transcription products were observed with RNA from IFN-treated cells (three separate experiments, one example shown in Fig. 2, lane 2) with the Sp100 oligonucleotide indicating several transcription start sites at the potential ACTTC-and GCCA-HIP1 binding region (Fig. 1). In untreated cells, essentially the same type of signals were seen but only after a much longer exposure (Fig. 2,  lanes 1 and 1*). Although the sizes of the products were the same in IFN-treated and untreated cells, the major transcription initiation site corresponding to one of the fast migrating fragments was predominant only in IFN-treated cells (Fig. 2,  compare lanes 1* and 2). This indicates an IFN-dependent quantitative change in transcription initiation site usage. The correct mapping of the 5Ј-end of the ␤-actin mRNA in the control experiment with RNA from IFN-␤-treated or untreated HEp2 cells, the lack of reverse transcripts when Escherichia coli tRNA was used as negative control template (Fig. 2, lanes  3, 4, and 5, respectively), and the much stronger Sp100 signals in IFN-treated compared to untreated cells demonstrate that the Sp100 primer extension products reflect true transcription initiation sites. These results were confirmed by RNase protection experiments (data not shown).
Functional Characterization of the Sp100 Promoter-Mapping of the transcription initiation sites provided the basis for the identification of sequence elements regulating the constitutive and IFN-inducible Sp100 promoter activity by transient transfection experiments. Several Sp100 promoter fragments were generated by PCR (Fig. 3A) and cloned blunt end into SmaI-digested vector pGCAT-C (58) upstream of the CAT gene. All constructs, except pGCAT-1091 and Ϫ266, contain the authentic Sp100 translation start codon for expression of CAT fusion proteins (Fig. 3A). Three transient transfection experiments with HuH7 cells and two different DNA preparations (Fig. 3B, transfections 1, 2, and 3) of each CAT construct containing promoter fragments in sense orientation (Fig. 3A) resulted in similar, although slightly fluctuating, amounts of CAT enzyme at levels comparable to that achieved with a reference herpes simplex virus (HSV) thymidine kinase promoter construct (pBLCAT2) (59) (Fig. 3B). Consistent with our previous computer predictions, these experiments indicate that the sequence elements required for constitutive Sp100 promoter activity are located between positions Ϫ143 and Ϫ14. When the same experiments were performed with cells treated after transfection with IFN-␤, all fragments in sense orientation with 5Ј-ends between positions Ϫ1104 and Ϫ192 and 3Ј-ends either at position ϩ12 or Ϫ14 exhibited much higher promoter activity than in untreated cells (Fig. 3B). Progressive 5Ј-deletions up to nucleotide position Ϫ193 (constructs pGCAT-1116 to Ϫ204) gradually reduced IFN-␤ inducibility from 13.0-to 7.4-fold but did not completely abolish IFN-␤ responsiveness. The fragment with the 5Ј-end at position Ϫ143 and the 3Ј-end at position ϩ12 retained basal promoter activity but lacked IFN inducibility completely. As expected, the promoter fragment tested in reverse orientation (Fig. 3, pGCAT-1116(Ϫ)) as well as the vector without insert had neither constitutive nor IFN-inducible promoter activity. These data locate an IFN-␤-responsive element within region Ϫ192 to Ϫ144 of the Sp100 promoter. This is consistent with our previous identification of an ISRE at position Ϫ162 to Ϫ149 (Fig. 1) presumed to play a role in IFN inducibility. However, the gradual decrease in IFN inducibility associated with the progressive 5Ј-deletions of the promoter fragments up to the position of the ISRE argues for additional sequence elements upstream of the ISRE required for full IFN-inducibility of the Sp100 promoter.
The Sp100-ISRE Confers IFN-␤ Inducibility-Transient transfection experiments suggested but did not prove that the ISRE within region Ϫ192 to Ϫ144 contributes to IFN-␤-enhanced Sp100 promoter activity. In order to corroborate this suggestion, HuH7 cells were transfected with CAT plasmids containing in both orientations either a synthetic Sp100 oligonucleotide spanning region Ϫ168 to Ϫ143 and including the ISRE (5Ј-TTAAATACTTTCACTTCTCTTTTCCC-3Ј) or an oligonucleotide containing the ISRE from the IFN-inducible gene ISG15 (60) (5Ј-GGGAAAGGGAAACCGAAACTGAAGCC-3Ј) upstream of the HSV thymidine kinase promoter of plasmid pBLCAT2. When the transfected cells were treated with IFN-␤, the promoter activities of all four constructs increased from 3.6to 9.2-fold (data not shown), demonstrating that irrespective of the orientation both oligonucleotides convey IFN inducibility.
Identification of Proteins Binding to the Sp100-ISRE-IFN-␤-enhanced Sp100 gene transcription occurs rapidly and is independent of protein synthesis (14). Therefore, we tested in gel mobility shift assays whether ISGF3, the primary transcriptional activator of IFN-␣/␤-stimulated genes, and/or other proteins bind to the Sp100-ISRE. For these experiments, we used the oligonucleotides containing either the Sp100-ISRE or, as a control, the ISG15-ISRE (see above). The nuclear protein extracts were prepared from untreated Hela S3 cells or from cells treated for 16 h with IFN-␥ and IFN-␤ during the last 30 min to obtain maximal levels of ISGF3 (61). With the Sp100-ISRE, one major DNA-protein complex was evident in untreated cells, its amount strongly increasing upon IFN treatment (Fig. 4, A and B, lanes 1 and 2). Upon longer exposure of the gel, additional DNA-protein complexes became apparent (Fig. 4B, lanes 2 and 4 -6). Competition with unlabeled Sp100-ISRE (Fig. 4, A and B, lanes 3) but not with an unrelated oligonucleotide (Sp1-oligo, Fig. 4, A and B, lanes 4) strongly inhibited formation of the major and four minor DNA-protein complexes (one of them can be seen only after much longer exposure; Fig. 4B, lanes 2 and 4), demonstrating that they are due to sequence specific protein binding. When antibodies against ISGF2/IRF1 were added to the binding reactions, the amount of the major DNA-protein complexes drastically decreased, and supershifted ones appeared (Fig. 4, A and B, lanes  6). This demonstrates that ISGF2/IRF1 binds to the Sp100-ISRE and is a component of the major DNA-protein complex. A similar experiment performed with or without antibodies against p84/p91 (both proteins are components of ISGF3) revealed an IFN treatment-dependent DNA-protein complex containing ISGF3 as it supershifted after antibody addition (visible only after long exposure; Fig. 4B, lanes 2 and 5). The oligonucleotide containing the ISG15-ISRE showed higher levels of the same ISGF3-DNA complex in untreated cells, which increased even further after IFN-␤ treatment (Fig. 4, A and B,  lanes 7 and 8). Taken together, these experiments revealed very similar IFN-enhanced complex formation of ISGF2/IRF1 with the Sp100-ISRE and the ISG15-ISRE, much weaker binding of ISGF3 to the Sp100-ISRE, and similar complexes formed by both oligonucleotides (compare Fig. 4A, lane 8, and Fig. 4B,  lane 2).
IFN-␤-inducible Factors Bind to the Sp100-GAS-The very weak complex formation of the Sp100-ISRE with ISGF3 suggested the participation of additional transcription factors in IFN-enhanced Sp100 promoter activity. This prompted us to investigate whether AAF/STAT1 may be such a factor by testing its binding to the canonical GAS sequences previously identified in the Sp100 promoter region (Fig. 1). In gel mobility shift assays with the Sp100-GAS at position Ϫ655 to Ϫ647 (Fig.  1), three complexes were detected almost exclusively with extracts from Hela S3 cells treated for 30 min with IFN-␤ (Fig. 5,  lanes 1 and 2). Binding of all three complexes proved to be specific since competition with unlabeled Sp100-GAS but not with Sp1-oligo prevented their formation (Fig. 5, lanes 3 and 5,  respectively). As a positive control, an oligonucleotide containing the ␤-casein-GAS known to bind not only STAT5 but also IFN-induced STAT1 homodimers (63) was used. This oligonucleotide formed two protein complexes corresponding in mobility to the two fast migrating Sp100-GAS complexes and both also formed only when extracts from IFN-treated cells were used (Fig. 5, lanes 9 and 10). The co-migration of the ␤-casein-GAS and the two fast migrating Sp100-GAS complexes as well  1 and 3) and from cells treated with IFN-␤ (10 h) (lanes 2 and 4) were reverse transcribed, using only a Sp100-specific primer (indicated in Fig. 1) or the same primer together with a ␤-actin-specific oligonucleotide (5Ј-CGTCGC-CCGCGAAGCCGGCC-3Ј, resulting in a 157-bp long fragment) (57) (lanes 1 and 2, and 3-5, respectively). Lane 5, negative control with 10 g of E. coli tRNA as a template. Lanes 1* and 3*, longer exposures of lanes 1 and 3. Lanes C, T, A, and G, Sp100 promoter sequence obtained with the Sp100-specific primer. Sp100-and ␤-actin-specific signals are marked on the right. Nucleotides on the left corresponding to transcription start sites are displayed in boldface letters. The arrow indicates the major initiation site.
as the complete disappearance of the latter by competition with unlabeled ␤-casein-GAS strongly indicate that they contain common components. These are STAT1␣ and/or STAT1␤ as deduced from the disappearance of both Sp100-GAS complexes and the emergence of two supershifted bands after addition of ␣-p84/p91 antibodies to the binding reaction (Fig. 5, lane 6). Currently, we cannot determine whether the slow migrating Sp100-GAS complex also contains these proteins as the band at the corresponding position in this blot may correspond to a supershifted complex of one of the fast migrating ones or to the third Sp100-GAS complex which did not supershift. The lack of supershifted complexes in the gel mobility shift experiments with antibodies against ISGF3␥/p48 or ISGF2/IRF1 (Fig. 5,  lanes 7 and 8, respectively) indicates that these proteins are not components of any of the Sp100-GAS complexes. Interestingly, with an oligonucleotide containing the second potential Sp100-GAS (nucleotide position Ϫ811 to Ϫ803, see Fig. 1) no complex formation could be observed, irrespective of whether protein extracts from untreated or IFN-␤-treated cells were used (data not shown). Thus, only one of the two GAS elements appears to contribute to IFN inducibility of the Sp100 promoter.
The Sp100-GAS Renders a Heterologous Promoter IFN-␤and IFN-␥-inducible-The gel mobility shift assays demonstrated that the Sp100-GAS at position Ϫ655 to Ϫ647 is able to form complexes with three AAFs. To study the functional significance of this finding we cloned the Sp100-GAS oligonucleotide (see legend to Fig. 5) as a monomer and as a trimer upstream of the HSV thymidine kinase promoter of plasmid pBLCAT2 and investigated the IFN inducibility of the corresponding plasmids. Transient transfection experiments of two different preparations of each plasmid revealed that the Sp100-GAS monomer construct was only poorly inducible by IFN-␤ (1.3-fold), whereas it was strongly inducible by IFN-␥ (10.6fold) (data not shown). In contrast to the monomer construct, however, the Sp100-GAS trimer construct was equally well inducible by IFN-␤ and -␥ (10.4-and 9.5-fold, respectively) (data not shown). These data demonstrate that the Sp100-GAS is able to confer IFN-␤ and IFN-␥ responsiveness to a heterologous promoter. The very poor IFN-␤ inducibility of the Sp100-GAS monomer construct is in good agreement with the transient transfection experiments shown in Fig. 3, where deletion of the region containing the Sp100-GAS did not result in a reduction of the IFN-␤ inducibility of the corresponding Sp100 promoter construct. Furthermore, our data are consistent with results obtained with other GAS elements in transient transfection experiments (see, for example, Ref. 42). The weak type I IFN responsiveness of promoters containing only one GAS element is believed to result from a shorter lived interaction of AAFs with GAS elements compared to IFN-␥-activated GAFs.

DISCUSSION
In this report we have described the isolation and functional characterization of the promoter region of the IFN-inducible Sp100 gene. Heterogeneous transcription start sites differentially used in IFN-treated and untreated cells as well as sequence elements and transcription factors involved in constitutive and IFN-inducible promoter activity were identified. Unlike all other IFN-inducible genes known so far, IFN inducibility of the Sp100 promoter appears to be controlled by the combination of an imperfect ISRE and a GAS separated by more than 400 nucleotides and probably by additional sequence elements.
We found that transcription from the Sp100 promoter is initiated at various sites as described for example for the IFNinducible GBP and 9 -27 genes (64,65). Indirect evidence has been provided that Sp100 transcription initiation is controlled by the transcription factors HIP1 and Sp1, similar as described for promoters of several housekeeping genes, the ISGF2/IRF1 gene, and for the SV40 late promoter (52). The start sites of transcription for the Sp100 promoter are identical in IFNtreated and untreated cells. However, similar to the situation described for the IFN-inducible murine 202 gene (66), we found preferential initiation at a specific site in IFN-treated cells, suggesting that IFN can also modulate the efficiency of transcription initiation at specific sites. As circumstantial evidence suggests that the type of splicing of Sp100 mRNAs may also be affected by IFN (14), it appears that IFN modulates Sp100 gene transcription on the level of transcriptional initiation as well as posttranscriptionally.
With the exception of the GBP gene, all genes known to respond to type I IFNs contain either an ISRE or a GAS but not both in their promoter regions (see Refs. 28 and 67 and references therein). The GBP gene is the only example where an ISRE and an overlapping GAS are both necessary for transcriptional induction by IFN-␣/␤ (40,41). The IFN inducibility of the Sp100 and GBP promoters appears similar in some but not all aspects. The ISREs of the Sp100 and GBP promoters both diverge from the consensus sequence (a G/C transversion at the second, very highly conserved, position) and, therefore, have a very low affinity for ISGF3 (our data) (40). Unlike the ISRE and GAS elements of the GBP promoter, those in the Sp100 promoter are separated by several hundred nucleotides. More-over, the GAS motifs of both promoters are different both in sequence and function as reflected by the different DNA-protein complexes formed in gel mobility shift assays (our data) (67). These differences may be in part responsible for the biphasic and monophasic kinetics of Sp100 and GBP gene transcription upon type I IFN treatment, respectively, and the less pronounced induction of Sp100 gene transcription upon IFN-␥ treatment (14,39). In addition, differences in sequences flanking both elements of the two promoters could also play a role as they can alter the specificity and the extent of the IFN response as shown for the ISRE of the 6-16 and (2Ј-5Ј)oligoadenylate synthetase genes, respectively (65,68).
Although the Sp100-ISRE diverges from the consensus sequence, it binds the same proteins as the "consensus" ISRE from the ISG15 gene promoter. Five comparable specific DNAprotein complexes, one of them containing ISGF2/IRF1 and one identified as ISGF3, formed with each of the two oligonucleotides. However, as concluded from the intensities of the signals the proteins of these complexes appeared to have higher affinities for the ISG15-ISRE than for the Sp100-ISRE. This is probably due to the T/C transition in the second T-triplet of the Sp100-ISRE affecting both ISRE and ISRE core binding.
ISGF3 and ISGF2/IRF1 were unequivocally identified as components of two of the Sp100-ISRE-protein complexes. The complex migrating below the very faint Sp100-ISRE-ISGF3 Gel mobility shift assays with oligonucleotides containing either the Sp100-GAS (5Ј-CTGATTTCCAGGAAGGCTAA-3Ј, nucleotide position Ϫ660 to Ϫ641) or the GAS of the ␤-casein promoter (5Ј-AGATTTCTAG-GAATTCAAATC-3Ј) (62) as indicated at the bottom of the gel. The type of treatment of Hela S3 cells and the components added to the reaction mixture are indicated above the gel. The retarded complexes (AAF, IFN-␣ activation factor) are marked on the left. complex supershifted when extracts were incubated with ␣-ISGF2/IRF1 antibodies, suggesting that this complex is identical to a slower migrating form of ISGF2/IRF1 which has been described for extracts containing large amounts of ISGF2/IRF1 (33). As we focussed on the identification of positively acting ISRE-binding factors, the additional complexes were not studied.
As the Sp100-ISRE-protein complexes, those formed with the Sp100-GAS were also characterized. Three and two IFN-␤inducible complexes were observed with one of the Sp100-GAS and a functional GAS from the ␤-casein gene, respectively. This Sp100-GAS represents a functional GAS novel in sequence and, therefore, increases the growing number of GAS elements (67). Yet, it differs only in one nucleotide position from the ␤-casein-GAS (see legend to Fig. 5). The different complex pattern seen with the two GAS elements is consistent with and extends recent reports which have demonstrated that the type of GAS complexes formed is not only dependent on the inducing agent and cell type but also on the type of GAS sequence used (63,67,69,70). The two fast migrating complexes seen with the Sp100-GAS apparently both contain STAT1␣, whereas the large complex appears to consist of different protein(s). Unlike the GBP-GAS-protein complex (the only one described thus far to contain ISGF3␥/p48) (67), none of the Sp100-GAS complexes contain ISGF3␥/p48, corroborating our previous assumption of functional differences between both GAS elements. As STAT3 is also inducible in some cell types by type I IFN and was shown to bind to some GAS sequences as homodimer and as heterodimer with STAT1␣ (69,70), it is conceivable that the largest AAF complex represents a STAT3 homodimer, the second largest a STAT1␣/STAT3 heterodimer, and the smallest a STAT1␣ homodimer. A gel mobility shift experiment with nuclear protein extracts from IFN-␥-induced cells identified indeed the smallest complex as a STAT1␣ homodimer (data not shown).
Taken together, these data suggest that binding of the STAT1␣ homodimer and eventually the other AAFs to the Sp100-GAS mediates primary transcriptional activation, followed by ISGF2/IRF1 binding to the Sp100-ISRE at later stages of activation. This model is consistent with the demonstration that ISGF2/IRF1 can stimulate expression of a reporter gene under the control of an ISRE-core (36). Moreover, ISGF2/IRF1 cannot account for the immediate and protein synthesis-independent rise of gene transcription rates upon IFN induction shown previously for Sp100 (14), whereas AAFs are immediately activated upon IFN induction. Mutational analysis of Sp100 promoter sequences combined with expression studies in stable cell lines are required to further support this model.