Multiple proteins interact with the nuclear inhibitory protein repressor element in the human interleukin-3 promoter.

T cell expression of interleukin 3 (IL-3) is directed by positive and negative cis-acting DNA elements clustered within 300 base pairs of the transcriptional start site. A strong repressor element, termed nuclear inhibitory protein (NIP), was previously mapped to a segment of the IL-3 promoter between nucleotides −271 and −250. Functional characterization of this element demonstrates that it can mediate repression when linked in cis to a heterologous promoter. DNA binding experiments were carried out to characterize the repressor activity. Using varying conditions, three distinct complexes were shown to interact specifically with the NIP region, although only one correlates with repressor activity. Complex 1 results from binding of a ubiquitous polypeptide that recognizes the 3′ portion of this sequence and is not required for repression. Complex 2 corresponds to binding of transcription factor (upstream stimulatory factor) to an E-box motif in the 5′ portion of the NIP region. DNA binding specificity of complex 3 overlaps with that of upstream stimulatory factor but is clearly distinct. To determine which of the latter two complexes represents NIP activity, we incorporated small alterations into the NIP site of an IL-3 promoter-linked reporter construct and examined their effects on NIP-mediated repression. Functional specificity for repression matches the DNA binding specificity of complex 3; both repressor activity and complex 3 binding require the consensus sequence CTCACNTNC.

Human interleukin 3 (IL-3) 1 is a potent growth factor, which supports early hematopoietic progenitor cells and potentiates lineage-specific effects of later acting growth factors (1,2). IL-3 expression is restricted to activated T cells and NK cells (3)(4)(5)(6) and is regulated primarily at the level of transcription (7)(8)(9). cis-acting promoter elements governing tissue-specific and activation-dependent expression are found within 300 base pairs (bp) of the transcriptional start site (8 -13). Promoter deletion experiments have identified two activating sites referred to as ACT-1 (NFIL-3) and AP-1/Elf-1 (8 -10, 12). In addition, there is a powerful repressor site, which binds an activity termed nuclear inhibitory protein (NIP) (9). Functional experiments in transfected T cells localized the NIP element between nucleotides Ϫ271 and Ϫ250 upstream of the transcriptional start (9). These experiments showed that, in the absence of AP-1 and Elf-1 sites, the NIP element blocks activation mediated by the downstream ACT-1 and CBF sites, thus silencing IL-3 expression in activated T cells (9).
The nature of the NIP repressor has remained elusive subsequent to this initial characterization. Studies with the human interleukin 2 (IL-2) promoter identified a repressor site, NRE-A, which interacts with a zinc finger DNA binding protein (14). Since the promoters of the IL-2 and IL-3 genes have common characteristics and function exclusively in activated T cells (5), it was possible that NIP and NRE-A sites might be related. A second possible identity for NIP was suggested by sequence analysis, which revealed that the NIP site contains a consensus E-box sequence (15,16) that might be recognized by a helix-loop-helix transcription factor.
To better define how NIP interacts with the IL-3 promoter and to characterize the protein mediating repressor activity, we have determined the nucleotide requirements for NIP function. In parallel we have investigated three proteins that bind to the repressor site to determine which correlates with the repressor activity of NIP. Taken together, our results show that NIP and NRE-A are distinct elements and that the repressor protein NIP does not belong to the E-box binding class of transcriptional regulators.

MATERIALS AND METHODS
Cell Lines-The gibbon T cell line MLA 144 (17) was maintained in 5% CO 2 at 37°C in RPMI medium supplemented with 10% fetal calf serum and antibiotics (50 units/ml penicillin; 50 g/ml streptomycin). The human Jurkat T cell line, the Epstein-Barr virus-transformed Raji cell line, the K562 erythroleukemia cell line, and a murine Abelsonmurine leukemia virus-transformed pre-B cell line were cultured in RPMI medium supplemented with 10% fetal calf serum and antibiotics. Mouse erythroleukemia cells were cultured in Dulbecco's modified Eagle's medium supplemented with 2.5% fetal calf serum, 7.5% calf serum, and antibiotics.
Oligonucleotides and IL-3 Promoter Constructs-Sequences of one strand of the double-stranded oligonucleotides used in electrophoretic mobility shift assays (EMSAs) are indicated in Fig. 2. All IL-3 promoter segments are cloned upstream of a reporter gene construct consisting of a modified human IL-3 gene (IL3 x ) (9). The reporter gene Ϫ173/IL3 x has been described (9). For (NIP)-173/IL3 x , a double-stranded oligonucleotide spanning nt Ϫ267 to Ϫ244 of the IL-3 promoter (18) was ligated upstream of Ϫ173/IL3 x . For Ϫ267/IL3 x , polymerase chain reaction (PCR) was used to amplify a segment of the IL-3 promoter between nt Ϫ267 and Ϫ1. The 5Ј primer used in the reaction incorporated a HindIII restriction site for cloning purposes. The PCR fragment was digested with HindIII and SmaI (position Ϫ61 in the IL-3 promoter) and sub-* This work has been funded in part by National Institutes of Health Grant RO1 DK41758 (to B. M.-P.). 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.
§  1 The abbreviations used are: IL, interleukin; bp, base pair(s); NIP, nuclear inhibitory protein; nt, nucleotide(s); EMSA, electrophoretic mobility shift assay; GM-CSF, granulocyte-macrophage colony-stimulating factor. cloned into the unique HindIII and SmaI sites of the IL-3 promoter fragment in Ϫ173/IL3 x . This resulted in an IL3 x reporter gene linked to an IL-3 promoter extending to nt Ϫ267. Upstream primers that incorporated mutations in the NIP site listed in Fig. 2 were used on the Ϫ267/IL3 x DNA template to generate (by PCR) M1, M2, M5, M6, M7, M8, M9, and M10 IL3 x reporter gene mutants. Sequence integrity for all new constructs was confirmed by DNA sequencing of the promoter region between the HindIII and SmaI sites. All oligonucleotides (for PCR and EMSA) were obtained from the Oligonucleotide Core Facility at the Dana-Farber Cancer Institute.
Granulocyte-Macrophage Colony-stimulating Factor (GM-CSF) Promoter Constructs-For Ϫ215 GM /IL3 x , a fragment of the human GM-CSF promoter from nt Ϫ215 to Ϫ17 was subcloned immediately upstream of the presumed start site in the IL3 x reporter gene at a BanII site. For (NIP)-215 GM /IL3 x , the same double-stranded NIP oligonucleotide present in (NIP)-173/IL3 x was subcloned upstream of the Ϫ215 GM /IL3 x reporter gene. The area of interest was sequenced to confirm the integrity of the construct.
DNA Electroporation-4 ϫ 10 7 MLA 144 cells were resuspended in 3.5 ml of RPMI medium lacking fetal calf serum. For each electroporation, 50 g of plasmid DNA was added to the cell suspension and incubated on ice for 5 min. The suspension was split into four aliquots and transferred to disposable cuvettes for electroporation in ProGenetor II (Hoefer Scientific Instruments) set at 275 V, 960 F. Cells were then resuspended in a total of 50 ml of RPMI medium complete with fetal calf serum. After 36 -40 h cells were stimulated for 9 h with phorbol 12-myristate 13-acetate (10 ng/ml final concentration). Cells were harvested, and RNA was isolated as described (5).
RNA Isolation and RNase Protection Assay-Total RNA was isolated, and expression of the reporter genes was analyzed by an RNase protection assay (9). The probe used is complementary to the first exon of the IL-3 reporter gene, including an insertion of a 12-bp XhoI linker at a HincII site. After ribonuclease A and T1 digestion, transcripts corresponding to the IL-3 reporter genes protect a 226-nt fragment, whereas endogenous MLA 144 IL-3 transcripts (lacking the 12-bp linker) give rise to two protected fragments of 151 and 63 nt, respectively (9).
EMSAs-Crude nuclear extracts were prepared from different cell lines by the Dignam procedure (19) or a modified procedure by Andrews and Faller (20). T cell extracts (unstimulated and stimulated) were prepared from the Jurkat and MLA 144 cell lines. Cells were stimulated with 10 ng/ml phorbol 12-myristate 13-acetate and 0.5 M ionomycin for 6 h prior to nuclear extract preparation. Protease inhibitors (from Boehringer Mannheim and Sigma) were included in all extract preparation steps: 0.2 mM phenylmethylsulfonyl fluoride, 1 g/ml pepstatin A, 0.1 g/ml leupeptin, and 0.4 g/ml aprotinin. Nuclear extracts from mouse myotube cells (C2DM) were the kind gift of Dr. S. Skapek, Harvard Medical School. Purified upstream stimulatory factor (USF) from HeLa cells was generously provided by Dr. D. Fisher (Dana-Farber Cancer Institute). Oligonucleotide duplexes were labeled with [␣-32 P]dCTP by filling in sticky ends with Klenow polymerase. Labeled oligonucleotides were purified by electrophoresis through nondenaturing polyacrylamide gels. Bands were excised, and oligonucleotides were eluted, precipitated, and resuspended in H 2 O. Oligonucleotide probes were incubated with nuclear proteins at a concentration of approximately 100 fmol/15 l assay. Two different buffers were used for the binding reaction: buffer A (10 mM Hepes/NaOH, pH 7.8, 50 mM potassium glutamate, 5 mM MgCl 2 , 1 mM dithiothreitol, 5% (v/v) glycerol, 1 mM EDTA, and 1 g/15 l poly(dI-dC)) and buffer B (10 mM Tris/HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA, 4% (v/v) glycerol, 5 mM dithiothreitol, 20 M ZnSO 4 , and 1 g/15 l poly(dA)⅐poly(dT)). Nuclear extracts were incubated at a concentration of about 5 g of total protein/assay with radiolabeled oligonucleotide probes for 20 min at 22°C. Competition experiments were performed by adding excess unlabeled oligonucleotides to the binding reaction (at 40 ng/15 l, a 50 -100-fold excess), 15 min prior to adding the probe. The unrelated oligonucleotide (nonspecific) used in competition was AGCTTACGTCTGTGGATC. Samples were electrophoresed at 22°C through 5% nondenaturing polyacrylamide gels in 0.5 ϫ TBE. Gels were dried on Whatman 3MM paper and used to expose Kodak x-ray film with an intensifying screen.
Immunologic Assays of Basic Helix-loop-helix Proteins-Anti-E12 antiserum was generously provided by Dr. C. Murre (UCSD). Monoclonal antibody preparations recognizing E2-2, E2-2/E12, and E12/E47 were provided by PharMingen Inc. (San Diego, CA). Polyclonal rabbit IgG recognizing a 20-amino acid C-terminal peptide of USF and a sample of that peptide (at 0.2 mg/ml) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-USF antibody (0.05 g) was incubated at 22°C in 15 l of binding buffer plus nuclear extract for 1 h prior to addition of the radiolabeled oligonucleotide probe. Antibody specificity was confirmed by preincubation with 1 l of the 20-amino acid Cterminal USF peptide for 5 h at 22°C before proceeding with the binding reaction as described above.

RESULTS
The NIP Element Is a General Repressor Site-Deletion analysis of the IL-3 promoter revealed the existence of a silencing element, NIP, between bp Ϫ267 and Ϫ244 (8,9). To determine whether this region exhibits a repressor function outside of its usual context, it was placed upstream of a truncated version of the IL-3 and GM-CSF promoters (Fig. 1, A and B). The activity of these constructs was measured by RNase protection after transient transfection into the gibbon T cell line MLA 144, as described previously (9). Expression directed by the truncated Ϫ173 IL-3 promoter (Ϫ173/IL-3 x reporter gene) was completely silenced when the NIP element was placed just upstream of it (Fig. 1A, lanes a and b). Consistent with the IL-3 promoter results, expression of the reporter gene was drasti-

FIG. 1. The NIP repressor site functions in a different position in the IL-3 promoter and in a heterologous promoter.
A schematic diagram of the wild-type IL-3 promoter is shown at the top. Reporter genes indicated on the right were transfected into MLA 144 cells, and expression of these constructs was monitored by RNase protection using a modified IL-3 first exon probe (9). The filled arrowhead indicates a 226-nt protected fragment resulting from expression of the reporter gene (9). The prominent 151-nt fragment (in the lower part of the gels) results from protection of the probe by gibbon endogenous IL-3 mRNA and serves as internal control for loading (9). A, constructs a and b contain a truncated portion of the IL-3 promoter (Ϫ173), containing the known regulatory sites ACT-1 (10), CBF (11), and DB1 (13) and other potential sites (CK1, CK2, and CACC) (9,11). In construct b, the NIP segment was fused to the Ϫ173 promoter, thus altering the position of the NIP site relative to the wild-type promoter. B, in constructs c and d, the IL-3 promoter was replaced by a GM-CSF promoter fragment (Ϫ215) known to direct regulated expression of GM-CSF in activated T cells (28). Construct d contains the NIP site upstream of the Ϫ215 cally reduced in activated T cells, when the NIP element was placed immediately upstream of the Ϫ215 GM-CSF promoter (Fig. 1B, lanes c and d). These results indicate that the NIP site can function as a silencer element in a heterologous position and in a heterologous promoter.
The NIP Site interacts with Three DNA-binding Complexes-To characterize the protein(s) responsible for the silencing function of the NIP site, in vitro binding experiments were carried out. A double-stranded oligonucleotide spanning the region between Ϫ270 and Ϫ241 of the IL-3 promoter (see Fig. 2) was used as a probe in an EMSA. In agreement with our previous findings (9), a dominant protein-DNA complex (complex 1) was detected with this probe (Fig. 3A). Complex 1 is specific, since it is competed by excess unlabeled NIP probe and not by excess nonspecific competitor (Fig. 3A). NIP-3Ј, an oligonucleotide consisting of the 3Ј half of the NIP probe (Fig. 2) also competed complex 1 as well as the faster migrating bands, which represent degradation products of complex 1. In addition, a fainter complex (complex 2) forms over the NIP probe. Because of its relatively weak binding over the NIP probe, complex 2 had not been recognized in past experiments (8,9). Complex 2 binding was competed by excess NIP and NIP-5Ј oligonucleotides but not by NIP-3Ј oligonucleotides. Last, there was a slower migrating doublet at the top of the gel, which was competed by NIP but not by NIP-5Ј or NIP-3Ј oligonucleotides. The nature of this doublet was not pursued since its detection failed to correlate with repressor activity (data not shown).
To analyze formation of complexes 1 and 2 over the NIP region, NIP-3Ј and NIP-5Ј probes (Fig. 2) were used in EMSA (Fig. 3, B, C, and D). The NIP-3Ј probe specifically bound complex 1 (Fig. 3B), while the NIP-5Ј probe failed to do so (Fig.  3C). Instead, NIP-5Ј bound complex 2. Complex 2 is specific and often gives rise to two bands, one major band and a slightly faster migrating minor band. The two bands exhibit identical DNA binding specificity in EMSA competition experiments ( Fig. 5 and data not shown).
To maximize detection of all proteins binding to the NIP region, we carried out EMSAs under a variety of different conditions. There was a striking difference in the pattern of complexes detected when poly(dA)⅐poly(dT) was used as nonspecific carrier DNA in place of poly(dI-dC). As shown in Fig.  3D, an additional complex (complex 3) was detected using the 5Ј portion of the NIP sequence, which migrated differently from complex 2 described above. The assignment of the slower migrating band as complex 2 is based on competition experiments using a variety of mutated probes (Figs. 2 and 8 and data not shown). The faster migrating band in Fig. 3D is not consistently seen in our different nuclear extracts (see Fig. 8), and it cannot be fully competed by excess unlabeled oligonucleotide. Thus, three distinct, specific complexes (complexes 1, 2, and 3) are reproducibly detected over the NIP site.
Binding of Complex 1 Is Not Necessary for Repression-The most prominent complex detected with the NIP probe is complex 1. To examine further the DNA binding specificity and function of complex 1, a mutant oligonucleotide was prepared in which four bp were deleted from the 3Ј portion of the NIP sequence (mutant NIP-⌬; Fig. 2). As shown in Fig. 4A, this oligonucleotide did not compete for complex 1 binding when added in excess and failed to bind complex 1 when radiolabeled and used as a probe. Binding of complex 2 to NIP-⌬ was unaffected, but detection required a much longer exposure than that shown for this experiment (data not shown).
The NIP-⌬ mutation was incorporated into the Ϫ267/IL3 x reporter construct to determine whether loss of complex 1 binding correlated with a change in repressor activity at the NIP site. Its activity was compared with Ϫ267/IL3 x (wildtype), which shows repression of the reporter gene, and a construct in which the NIP site has been deleted, Ϫ250/IL3 x (NIP minus), which shows no repression. As displayed in Fig. 4B, the 4-bp NIP-⌬ deletion had no effect on repression function. Thus, complex 1 does not correlate with NIP activity, and its binding is not required for repression to take place.
Complex 2 Is an E-box Complex Immunologically Related to USF-To characterize the DNA binding specificity exhibited by complex 2, several mutated oligonucleotides (listed in Fig. 2) were used to compete binding of complex 2 to the NIP-5Ј probe (Fig. 5). This experiment demonstrates that the integrity of the E-box consensus sequence in the NIP-5Ј probe is essential for complex 2 binding. Indeed, a E2 site (21), which has only the E-box consensus sequence in common with the NIP-5Ј probe, completely competes binding of complex 2. A compilation of the results observed with the several mutants reveals that only nucleotides forming the E-box consensus, i.e. CACNTG, are necessary. Nucleotides flanking the E-box core sequence do not influence binding of complex 2 in vitro. Also, the second of the two cytidines in the middle of the E-box can be mutated without affecting binding of complex 2 (mutant M6). Since complex 2 can be detected in all cells tested and requires the E-box consensus for binding, we investigated the possibility that complex 2 forms from binding of known ubiquitous basic helix-loophelix proteins. Antisera raised against proteins belonging to the E2-2 family of transcription factors failed to interfere with the formation of complex 2 on the NIP-5Ј oligonucleotide (data not shown). In contrast, formation of complex 2 on the NIP-5Ј probe was blocked by the addition of an antiserum recognizing the 20 C-terminal amino acids of USF (22) (Fig. 6A). This

FIG. 2. Sequence of wild-type and NIP oligonucleotides and other probes used in EMSA. Only the top strand is indicated. Wildtype nucleotides are shown in capital letters, mutations in boldface.
Nucleotides added for the purpose of labeling duplexes using Klenow polymerase are shown in lowercase. Numbering of the nucleotides is relative to the start site of the IL-3 gene (18). Sequences used for non-NIP oligonucleotides were obtained from published reports: E2 (21), USF site in the adenovirus major late promoter (29). The region corresponding to the E-box in the various oligonucleotides is shaded.
interaction appears to be specific, since preincubation of the antiserum with excess peptide against which it was raised largely restores binding of complex 2 with the NIP-5Ј probe (Fig. 6A). The E-box in the NIP-5Ј probe differs from the previously described USF consensus sequence (22). We therefore used a probe containing a bona fide USF binding site (from the adenovirus major late promoter) to confirm our initial observation. Using the USF probe under identical conditions, a complex with the same mobility as complex 2 (Fig. 6, A and B) can be detected. However, the affinity of the binding to the USF

FIG. 4. Complex 1 is not involved in repression.
A, nuclear extracts from unstimulated Jurkat cells were used in an EMSA with probes shown at the bottom of the gels (NIP probe and mutant NIP-⌬ probe (see Fig. 2). The presence or absence (No) and the type of competitor oligonucleotides are indicated above each lane. Identical results were obtained with extracts from stimulated Jurkat cells or extracts from MLA 144 cells before and after T-cell activation (data not shown). B, the functional consequence of the 4-bp deletion (NIP-⌬) engineered in the NIP site was tested in MLA 144 cells as described in Fig. 1. Expression of the IL-3 reporter gene (9) mediated by a wild-type IL-3 promoter fragment containing the NIP site (Ϫ267/IL3 x ) and by the same fragment containing the NIP-⌬ mutation was compared with that obtained with a promoter fragment lacking a full NIP site (Ϫ250/IL3 x ). The level of expression of this latter construct is equivalent to that of Ϫ173/IL3 x (data not shown). The protected fragment corresponding to the reporter gene is indicated by the arrowhead. As in Fig. 1, the prominent 151-nt fragment at the bottom of the gel (resulting from protection by endogenous IL-3 mRNA) indicates equal loading for RNA. probe is about 2 orders of magnitude higher than it is for the NIP-5Ј probe. Anti-USF antiserum interferes with binding of this complex (Fig. 6B). In addition, a faint supershift is apparent for this probe. While a similar supershift is not readily observed in Fig. 6A, it is likely that detection of such a supershift requires a stronger signal for complex 2 than what is generally observed with the NIP-5Ј probe. Alternatively, complex 2 may only be related to USF and react less strongly with the antiserum. To confirm our observation that complex 2 is USF or USF-related, purified USF from HeLa cells was incubated with NIP-5Ј and USF probes (Fig. 6C). A single complex with identical mobility is detected with both probes. Again, purified USF binds with much higher affinity to the viral USF site than it does to the NIP-5Ј probe. Competition experiments using USF, NIP-5Ј, and unrelated oligonucleotides confirmed the specificity of these bindings (data not shown). Taken together, these experiments suggest that the protein component of complex 2 and USF are the same or highly related polypeptides.
Repression by NIP Does Not Require an Intact E-box-As shown in Fig. 3D, complex 3 also forms on the NIP-5Ј probe. To test whether binding of either complex 2 or complex 3 correlated with repressor function, we performed a detailed mutational analysis. Nucleotides in and around the E-box consensus sequence were altered in the context of the Ϫ267/IL3 x reporter gene. The effect of these mutations on reporter gene expression was assayed (Fig. 7). The majority of the mutations interfered with repression and restored reporter gene expression to a level comparable with that observed with the Ϫ173 promoter (Fig. 7,  compare lanes M1, M2, M5, M7, M9, and M10 with lane NIP minus). In contrast, mutations present in the M6 and M8 constructs do not abrogate repression. These observations, together with results obtained from three additional mutations in the area of the E-box sequence, define the functional consensus for NIP repression to be CTCACNTNC ( Fig. 7 and data not shown). This consensus indicates that complex 2/USF cannot represent the repressor, since the invariant G in the E-box can be mutated to a T (mutant M8) without affecting repression, even though it abrogates binding of this factor in competition experiments (Fig. 5, lane M8).
Complex 3 Is the NIP Repressor-Having excluded complex 2/USF as the repressor, we turned our attention to complex 3. The NIP-5Ј oligonucleotide duplex and four informative mutant probes were used in EMSA to identify a protein complex that showed a binding pattern matching the NIP functional consensus. The NIP-5Ј wild-type probe as well as mutant M6 and M8 probes should bind the repressor complex, while the mutant M1 and M9 probes should not. In addition, probe M8 can help discriminate between E-box binding proteins (which cannot be the repressor) and the NIP repressor complex itself. Only complex 3 fulfills the above requirements (Fig. 8A). Taken together, these results indicate that complex 3 is the NIP repressor complex. We have designated this activity NIP-C3, for NIPcomplex 3.
The IL-2 promoter, which also directs T cell-specific expression, contains a repressor site, NRE-A, which binds a zinc finger protein (14). Comparison of the complex(es) binding to the NRE-A and NIP sites showed them to have different mobilities in EMSA. Furthermore, NIP oligonucleotides did not compete specific binding to the NRE-A probe; nor did NRE-A oligonucleotides compete NIP-C3 binding to its cognate site (data not shown). This suggests that the IL-2 repressor and NIP-C3 are distinct from each other.
NIP-C3 Can Be Detected in Different Cell Types and Organisms-To further characterize NIP-C3, nuclear extracts from various primate and murine cell lines were tested with the same set of probes (Fig. 8, A and B). NIP-C3 was detected in all cases. An equivalent amount of this complex is present in nuclear extracts from unstimulated and stimulated human Jurkat T cells (Fig. 8, A and B). As already shown, NIP-C3 is detected in nuclear extracts from the gibbon T cell line MLA 144. The prominent complex that migrates ahead of NIP-C3 is unlikely to be a breakdown product of NIP-C3 since it binds to probe M9. Its significance, if any, is unknown. NIP-C3 is also detected in erythroleukemic K562 cells, mouse erythroleukemia cells, human B cells (Raji), mouse pre-B cells (2M3), and mouse myotube (C2DM) (Fig. 8B). Although the relative amount of NIP-C3 varies from cell line to cell line, NIP-C3 migration appears to be the same in all preparations, with the exception of the faster component in the NIP-C3 doublet in mouse pre-B cells. This component may have arisen from proteolytic degradation or may represent an altered form of NIP-C3. DISCUSSION The presence of a negative regulatory element within the IL-3 promoter was originally recognized when a truncated IL-3 promoter lacking the AP-1/Elf-1 sites was tested in T cell lines (8,9,12). These experiments localized this element between nt Ϫ244 and Ϫ270 of the IL-3 promoter and showed that it also functions as a repressor in primary T lymphocytes (23), B cells, and HeLa cells (12). We have shown that the repression function of the NIP region (Ϫ267 to Ϫ244) is preserved when placed in a different location within the IL-3 promoter or transferred in cis to a heterologous promoter (Fig. 1). Thus, repression through the NIP site is not cell-restricted and may play a more general role than the one involved in IL-3 regulation. The NIP region was used as a probe to identify proteins that might mediate its repression function. Using various assay conditions for DNA binding in vitro, we were able to detect three specific complexes (Fig. 3). The most prominent complex is complex 1. Partial characterization and purification of complex 1 indicate that it is a ubiquitous protein of with a molecular mass of 63 kDa. 2 It forms on the 3Ј half of the NIP site and is dispensable for repression function (Fig. 4). Its role in IL-3 regulation remains unknown. A complex with similar proper-ties has been described, which binds a site in the stromelysin gene promoter and appears to mediate phorbol ester activation in cooperation with a nearby AP-1 site (24).
The 5Ј half of the NIP site interacts with two distinct protein complexes to form complex 2 and complex 3. Both complexes are expressed ubiquitously ( Fig. 8 and data not shown) and appear to be conserved across different species. Since a consensus sequence for an E-box binding protein is present in the 5Ј half of the NIP site, we sought to determine whether either of the two complexes might be a member of the basic helix-loophelix family of transcription factors. Eleven mutant oligonucleotides of the NIP-5Ј probe were tested as competitors in EMSAs revealing that the binding specificity of complex 2 (CACNTG) matched the general consensus sequence of an Ebox site (CANNTG). The identity of complex 2 was explored using antibodies directed against ubiquitous E-box proteins. Complex 2 was disrupted by an antiserum recognizing the general transcriptional activator USF. Additional experiments established that complex 2 is identical or closely related to USF.
Complex 2 cannot be the NIP repressor, however, since its DNA-binding specificity differs from the functional repressor consensus sequence CTCACNTNC. Other E-box binding proteins are also unlikely candidates, since a mutation altering the invariant G in the NIP E-box core had no effect on repressor function. Since neither complex 1 nor complex 2 binding correlated with repressor activity, we focused on complex 3, the only remaining candidate for this function. The use of different informative mutant oligonucleotides as DNA probes in EMSA showed that, among the three complexes that form on the NIP region, only complex 3, designated NIP-C3, displayed the appropriate specificity for being the repressor.
NIP-C3 is expressed in a wide array of tissues and species ( Fig. 8 and data not shown). Its binding in vitro is exquisitely sensitive to the type of nonspecific DNA present in the binding FIG. 6. Complex 2 is related to or identical to USF. A, nuclear extracts from unstimulated Jurkat cells were tested in EMSA with a NIP-5Ј probe in buffer A. The nuclear extract was either directly added to the binding reaction (first lane), incubated with an antibody raised against the 20 C-terminal amino acids of USF (lane ␣USF-Ab), or incubated with the same antibody blocked with excess peptide against which the antibody was raised (lane ␣USF-Ab ϩ peptide) prior to the addition of the probe. B, the same experiment was performed except that a USF binding site derived from the adenovirus major late promoter (see Fig. 2) was used as a probe. Experiments in panels A and B were run on the same gel. As apparent from the intensity corresponding to free probe, exposure time for panel A was longer than for panel B. C, binding of purified USF (from HeLa cells) to NIP-5Ј and USF probes. The two reactions were run on the same gel. The lane with NIP-5Ј as a probe was exposed longer.
FIG. 7. The functional consensus for NIP repression is different from the E-box consensus. Reporter gene constructs with different mutations in the Ϫ267 IL-3 promoter fragment (see Fig. 2 for sequences) were assayed as described in Fig. 1. Migration of the protected fragment derived from expression of the reporter gene is indicated by the arrowhead. Plasmid Ϫ173/IL3 x (lane labeled NIP minus), which lacks the NIP site, and plasmid Ϫ267/IL3 x (lane labeled wildtype) containing an intact NIP element served as controls. Equal loading for RNA was achieved as indicated by the prominent 151-nt fragment at the bottom of the gel. Results from the eight mutants shown here and three additional mutants (data not shown) define the consensus for NIP repressor function indicated at the bottom. reaction. We were unable to detect NIP-C3 when poly(dI-dC) was used as carrier. This behavior is reminiscent of the zinc finger transcription factor ␦EF1, which was originally identified as a transcriptional activator of the ␦-crystallin enhancer and was subsequently found to repress E2-box-mediated gene activation through competition of binding. Although NIP-C3 and ␦EF1 share similar binding specificities, they differ in recognition of at least one important position, where the presence of a specific nucleotide inactivates NIP repressor function while it favors ␦EF1 function (25). Although the two proteins may share similar mechanisms of repression, the above difference strongly argues that the two factors are different.
There is accumulating evidence that a growing number of proteins function by displacing helix-loop-helix transcription factors from their cognate E-box sequences. Indeed, another zinc finger protein, ZEB, was recently shown to be able to silence the IgH enhancer by displacing the E2A complex from its cognate site (26). Since binding sites for complex 2/USF and the repressor significantly overlap, a dynamic equilibrium of occupancy by complex2/USF and NIP-C3 over the 5Ј portion of the NIP site may dictate appropriate IL-3 expression in an analogous manner (27).
Why should the IL-3 promoter contain a strong repressor element? IL-3 is a potent growth factor, which plays multiple, complementary roles in normal hematopoiesis (1). Meticulous control of its production may be necessary to prevent unbridled proliferation of progenitor cells and to meet the body's fluctuating needs for differentiated cells. IL-3 may be an example of a protein that is so deleterious when overproduced that blocking its expression may be of greater importance than activating its expression. It has previously been shown that the proximal portion of the IL-3 promoter activates a basal level of transcription in a variety of cell types (12), yet IL-3 is produced only by activated T cells (5). The NIP repressor may serve as a clamp that normally prevents IL-3 transcription in all cells. Its effect, however, can be specifically abrogated when the upstream AP-1 and Elf-1 sequences engage proteins expressed only in activated T cells. Future studies must be aimed at understanding the mechanism by which factors acting at the AP-1 and Elf-1 sites relieve NIP repression.