The SANT Domain of Human MI-ER1 Interacts with Sp1 to Interfere with GC Box Recognition and Repress Transcription from Its Own Promoter*

ld;&.2qTo gain insight into the regulation of hmi-er1 expression, we cloned a human genomic DNA fragment containing one of the two hmi-er1 promoters and consisting of 1460 bp upstream of the translation initiation codon of hMI-ER1. Computer-assisted sequence analysis revealed that the hmi-er1 promoter region contains a CpG island but lacks an identifiable TATA element, initiator sequence and downstream promoter element. This genomic DNA was able to direct transcription of a luciferase reporter gene in a variety of human cell lines, and the minimal promoter was shown to be located within–68/+144 bp. Several putative Sp1 binding sites were identified, and we show that Sp1 can bind to the hmi-er1 minimal promoter and increase transcription, suggesting that the level of hmi-er1 expression may depend on the availability of Sp1 protein. Functional analysis revealed that hMI-ER1 represses Sp1-activated transcription from the minimal promoter by a histone deacetylase-independent mechanism. Chromatin immunoprecipitation analysis demonstrated that both Sp1 and hMI-ER1 are associated with the chromatin of the hmi-er1 promoter and that overexpression of hMI-ER1 in cell lines that allow Tet-On-inducible expression resulted in loss of detectable Sp1 from the endogenous hmi-er1 promoter. The mechanism by which this occurs does not involve binding of hMI-ER1 to cis-acting elements. Instead, we show that hMI-ER1 physically associates with Sp1 and that endogenous complexes containing the two proteins could be detected in vivo. Furthermore, hMI-ER1 specifically interferes with binding of Sp1 to the hmi-er1 minimal promoter as well as to an Sp1 consensus oligonucleotide. Deletion analysis revealed that this interaction occurs through a region containing the SANT domain of hMI-ER1. Together, these data reveal a functional role for the SANT domain in the action of co-repressor regulatory factors and suggest that the association of hMI-ER1 with Sp1 represents a novel mechanism for the negative regulation of Sp1 target promoters.

hmi-er1 (human mesoderm induction-early response 1) is a growth factor-induced immediate early gene encoding a novel transcriptional regulator (1,2) that is differentially expressed in breast carcinoma cell lines and tumors (3).
hmi-er1 is transcribed from two distinct promoters P1 and P2 (4), and transcripts undergo alternative splicing to produce six protein isoforms differing in their N and C termini (4). Two of the N-terminal variants differ in the sequence of their 5Ј-UTR, 1 whereas the third contains an additional exon that encodes 25 amino acids (aa). The C-terminal variants, hMI-ER1␣ and -␤, differ both in the size and sequence of their C-terminal domains: the ␣ C terminus consists of 23 aa and includes an LXXLL motif, a domain known to be important for interaction with nuclear hormone receptors (5). In this regard, the ␣ isoform mRNA is only detectable in endocrine tissues (4). The ␤ C terminus contains 102 aa and includes the only functional nuclear localization signal (4,6). Whereas the divergent Cterminal amino acid sequences would suggest that these two isoforms have distinct functions, so far no difference in function has been determined. Instead, both isoforms can act as transcriptional repressors, and this repression was shown to involve recruitment of histone deacetylase 1 (HDAC1) (2).
The common internal sequence of hMI-ER1 contains conserved domains found in a number of transcriptional regulators, including an acid activation domain (1), an ELM2 domain (7), and a signature SANT domain (8). In a recent report, we showed that the ELM2 domain functions in the recruitment of HDAC1 and transcriptional repression (2). The SANT domain is located immediately downstream of the ELM2 domain and, in other proteins, has been implicated in DNA binding as well as in protein-protein interactions (8), including interactions with histone deacetylase 3 (HDAC3)-and histone acetyltransferase-containing complexes (9 -11), as well as nonacetylated histones (12). Recently, the SANT domain of the nuclear receptor corepressor SMRT has been implicated in the ability of this protein to target specific promoters through interpretation of the so-called histone code (12). To date, no function has been ascribed to the SANT domain of hMI-ER1.
In this report, we investigate mechanisms of transcriptional regulation of the hmi-er1 P2 promoter and demonstrate that promoter activity is regulated by Sp1 protein and that this activation is repressed in a dose-dependent manner by hMI-ER1 protein. This autorepression does not depend upon HDAC activity but involves interference of Sp1 binding to the chromatin surrounding the hmi-er1 P2 promoter and from the cognate binding sites by physical association with a region containing the hMI-ER1 SANT domain. This represents a novel mechanism for negative regulation of Sp1 target promoters and a functional role for the SANT domain in the activity of co-repressor regulatory factors.

EXPERIMENTAL PROCEDURES
Cell Lines-All cell lines were obtained from the American Tissue Culture Collection and cultured at 37°C in 5% CO 2 in Dulbecco's modified Eagle's medium containing 10% fetal calf serum.
Plasmids and Constructs-A 1460-bp sequence, containing 144 bp of hmi-er1 5Ј-UTR and 1316 bp of 5Ј-flanking genomic sequence, was generated by PCR from human genomic DNA, using the primer pairs listed in Table I. The PCR product (Ϫ1316) was cloned into the pCR2.1 vector (pCR(Ϫ1316)), using the TOPO-TA cloning kit (Invitrogen) and sequenced on both strands. The sequence was verified by comparison to previously reported hmi-er1 sequences (4) and to the human genome sequence data (Sanger Centre). The pGL3(Ϫ1316) plasmid was generated by digestion of pCR(Ϫ1316) and subcloning into the XhoI/HindIII sites of the promoterless pGL3-Basic vector (Promega Corp.). The pGL3(AS) plasmid was generated by digestion of pCR(Ϫ1316) and subcloning into the KpnI/XhoI sites of pGL3-Basic.
To obtain GST-hmi-er1␣ or ␤ fusion constructs, cDNA representing the appropriate isoform was subcloned in-frame into the pGEX-4T-1 vector (Amersham Biosciences). A series of hMI-ER1 deletion mutations was generated by first amplifying fragments encoding the appropriate amino acid residues of hMI-ER1␣ or -␤, using the primer pairs listed in Table I. PCR products were cloned into pCR3.1, and EcoRI fragments were then inserted into the complementary sites of the pGEX-4T-1 plasmid. Deletion constructs were named according to the encoded amino acid residues of the hMI-ER1␣ or -␤ proteins. The GST-Sp1 fusion was constructed by subcloning an Sp1 EcoRI fragment from Sp1-pCR3.1 in-frame into pGEX-4T-2.
Computer Analysis of the hmi-er1 Promoter Region-Computer-assisted analysis was performed using the following programs: 1) for promoter prediction: Promoter Scan (PROSCAN Transfection and Reporter Assays-All transfections and TSA treatments were performed as previously described (2), in duplicate in 6-well plates, using the indicated amount of plasmid DNA, and cells were harvested after 48 h in culture. Luciferase assays were performed on cell lysates using a Monolight 2010 Luminometer (Analytical Luminescence Laboratory) and a luciferase assay reagent (Promega), according to the manufacturer's directions. The values obtained, in relative luciferase units (RLU), were normalized to the amount of cellular protein in each sample and plotted either as a -fold increase over that obtained with the control vector or as a percentage of control activity.
Electrophoretic Mobility Shift Assays-Electrophoretic mobility shift assays (EMSAs) were performed as in (13). Briefly, the double-stranded consensus Sp1 oligonucleotides (Promega) or the minimal functional promoter fragment (Ϫ68 to ϩ144) was labeled with [␥-32 P]ATP and T4 polynucleotide kinase and purified on NucTrap probe purification columns (Stratagene, Inc.). The labeled double-stranded oligonucleotides were incubated with 2 l of HeLa nuclear extract (Promega) or GST fusion protein at room temperature for 20 min in 20 l of reaction buffer containing 5% glycerol, 5 mM MgCl 2 , 1 mM dithiothreitol, 50 mM KCl, 10 M ZnSO 4 , 85 g/ml bovine serum albumin, and 50 mM HEPES (pH 7.5). Poly(dI-dC) was used as a heterologous competitor in the reaction (2 g/reaction). Where indicated, a 20-fold molar excess of unlabeled probe was included in the binding reaction. For antibody supershift assays, the extract was incubated for 30 min at room temperature with Sp1-specific antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Bound and free probes were resolved by nondenaturing electrophoresis on 4% polyacrylamide gels and analyzed by autoradiography.
GST Fusion Protein Production-GST fusion proteins were expressed in Escherichia coli BL21 and purified according to the instructions supplied with the pGEX-4T-1 and pGEX-4T-2 vectors. GST fusion protein level and purity were determined by SDS-PAGE.
Co-immunoprecipitation (Co-IP), GST Pull-down Assays and Western Blot Analysis-Coupled transcription-translation (TNT; Promega) reactions and in vitro co-IP assays were performed as in Ding et al. (2); the antibody used was either an anti-Sp1 antibody (Santa Cruz Biotechnology) or the anti-Myc monoclonal antibody, 9E10 (a kind gift from Dr. K. Kao, Memorial University). Pull-down assays with GST fusion proteins were performed as described in Ref. 14, using 1 g of GST fusion protein, 50 l of glutathione-Sepharose beads (Amersham Biosciences) and 50,000 cpm of 35 S-labeled TNT product. Bound proteins were analyzed by SDS-PAGE and autoradiography. For all assays, one-twentieth volume of the indicated TNT was loaded into the input lanes.
For the in vivo co-IP assays, either nontransfected HeLa cells or cells transfected with pMyc, pMyc-␣, or pMyc-␤ were used. Immunoprecipitation using anti-Sp1 or anti-pan-hMI-ER1 (4) antibodies and Western blot analysis using anti-Myc or anti-Sp1 were performed as described (2). Input lanes contained one-third of the volume used for immunoprecipitation.
Establishment of Stable hMI-ER1␣ and -␤ Tet-On HeLa Cell Lines and Doxycycline Induction-hMI-ER1␣ and ␤ Tet-On HeLa cell lines were generated using the hmi-er1␣ or -␤ coding region inserted into the a Deletion constructs were named according to the encoded aa residues of the hMI-ER1␣ or -␤ protein hMI-ER1 SANT Domain Regulates Sp1 Activity pTRE2 vector and a Tet-On gene expression system (Clontech), according to the manufacturer's protocol. Control cell lines were generated by transfection with the pTRE2 empty vector. Stable clones were induced to express hMI-ER1␣ or -␤ using 2 g/ml doxycycline (dox), and expression was verified by Western blot analysis of whole cell extracts, using an anti-hMI-ER1 antibody (4). Quantitative Real Time PCR and Semiquantitative PCR Analysis-RNA was extracted from uninduced and dox-induced Tet-On cell lines and reverse transcribed as in Ref. 1. Quantitative real time PCR was performed using the Syber Green PCR Master Mix and the ABI Prism 7000 SDS (Applied Biosystems), according to the manufacturer's protocol. The 5Ј-UTR of hmi-er1 was amplified using 5Ј-AGTGGCGGCG-GGAGCGGCAGAGA-3Ј (forward) and 5Ј-CGTACTGCCGGGTCACAT-CTCC-3Ј (reverse), whereas ␤-actin was amplified using 5Ј-ATCTGGC-ACCACACCTTCTACAATGAGCTGCG-3Ј (F) and 5Ј-ATGGCTGGGGT-GTTGAAGGTCTC-3Ј (R). Data analysis and the ratio of expression in induced cells relative to uninduced cells (relative expression ratio) was calculated as described in Ref. 15. Semiquantitative PCR was performed as in Ref. 4 using the primers listed above to amplify ␤-actin and the 5Ј-UTR of hmi-er1, as well as 5Ј-CAAGGGCTGAAGGCCTATGG-3Ј (forward) and 5Ј-CCAAATCGTGTTTGCTGAGC-3Ј (reverse) to amplify the coding region of hmi-er1.
Chromatin Immunoprecipitation Assays-Chromatin from HeLa cells or dox-induced Tet-On cell lines was cross-linked with formaldehyde, sheared, and then subjected to immunoprecipitation as described by Weinmann et al. (16), with the following modifications. Anti-Sp1 (Santa Cruz) or anti-pan-hMI-ER1 (4) antibodies were used for immunoprecipitation, and 50 l of protein A-Sepharose (Amersham Biosciences) was used instead of Staph A cells. Immunoprecipitated chromatin was eluted, treated with RNase A, and then treated with proteinase K as in Ref. 16. PCR was performed as in Ref. 1 using 5 l of resuspended DNA and 30 cycles, with 5Ј-TTTCCTCTGCTGTGTCA-ATGC-3Ј and 5Ј-GAGATGTGACCCGGCAGTAC-3Ј as forward and reverse primers, respectively. For experiments using 3,3Ј-dithiobispropionimidate-2HCl (DTBP), cells were treated for 30 min with 5 mM DTBP, as described in Ref. 17, prior to cross-linking with formaldehyde.

RESULTS
Cloning of hmi-er1 5Ј Regulatory Sequence and Identification of the Minimal Functional Promoter-A genomic DNA fragment encompassing the promoter region, P2 (4), and containing 1460 bp upstream of the ATG translational start codon of hmi-er1 was isolated by PCR from human genomic DNA. This sequence contains 144 bp of 5Ј-UTR (ϩ1 to ϩ144) and 1316 bp (Ϫ1 to Ϫ1316) upstream of the start of transcription (4).
Computer-assisted analysis of the 1460-bp fragment to identify potential cis-regulatory elements predicted a CpG island located between nucleotide position Ϫ389 and the start of translation (Fig. 1A). Further analysis revealed that, like many other GC-rich promoters, the upstream sequence does not contain a TATA box; nor does it contain an initiator (18) or a downstream promoter element (19). A number of potential transcription factor binding sites, including multiple Sp1 binding sites, were identified using TFSEARCH, PROSCAN, and TESS (Fig. 1A).
The 1460-bp fragment was subcloned into the promoterless luciferase reporter vector, pGL3-Basic, in sense and antisense orientations, and its promoter activity was analyzed in HeLa cervical carcinoma cells. As shown in Fig. 1B, luciferase activity was 8 -10-fold higher when expression was driven by the 1460-bp fragment, compared with the pGL3-Basic vector and was markedly lower when the 1460-bp fragment was in the antisense orientation. We tested the cellular specificity of the hmi-er1 promoter in six different cell lines: C33A, HEK293, BT-20, MCF-7, SK-OV-3, and U87. Luciferase activity was high in all transfected cell lines and ranged from 10-to 34-fold higher than control levels, demonstrating that the hmi-er1 promoter can function in a variety of cell types (data not shown).
To identify the minimal functional region of the hmi-er1 promoter, a series of 5Ј deletions was constructed in pGL3 and transiently transfected into HeLa cells. Deletion of nucleotides Ϫ1316 to Ϫ133 resulted in little change in luciferase activity (Fig. 1B). Further deletion to nucleotide Ϫ68 reduced but did not abolish luciferase activity, whereas deletion to nucleotide ϩ28 completely abolished activity (Fig. 1B). These data dem-FIG. 1. The hmi-er1 minimal promoter is located within ؊68/ ؉144 bp and contains Sp1 binding sites. A, nucleotide sequence of the CpG island. The predicted promoter region (PROSCAN) is shaded gray, the transcriptional start site is indicated by ϩ1, the ATG translational start codon is shown in boldface type, and the putative Sp1 binding sites are shown in white letters with black highlighting. Additional transcription factor binding sites are underlined, and the positions of the 5Ј-end of the ϩ28, Ϫ68, and Ϫ133 constructs are indicated by the relevant number. B, identification of the hmi-er1 minimal promoter region. The schematic on the left shows a scaled representation of the full-length pGL3(Ϫ1316) and the various 5Ј deletion constructs; the transcriptional start site is indicated by ϩ1. HeLa cells were transiently transfected with the luciferase reporter vector pGL3, pGL3 containing the 1460-bp fragment in the sense (Ϫ1316) or antisense orientation, or one of the indicated deletion constructs. The RLU were determined as described under "Experimental Procedures." The RLU for each sample was normalized to the amount of cellular protein and plotted as a -fold increase over the value obtained with the empty pGL3 vector. Shown are the average values and S.D. for three independent experiments. C, Sp1 binds to the minimal promoter region. EMSAs were performed as described under "Experimental Procedures," using a HeLa cell nuclear extract and a 32 P-labeled probe representing the hmi-er1 minimal promoter (bp Ϫ68 to ϩ144; 32 P(Ϫ68)). In lane 1, the labeled probe was incubated without HeLa nuclear extract. In hMI-ER1 SANT Domain Regulates Sp1 Activity onstrate that the minimal functional promoter is located within the sequence Ϫ68/ϩ144.
The hmi-er1 minimal promoter is predicted to contain four Sp1 binding sites (Fig. 1A); therefore, we investigated whether Sp1 binds to this sequence. EMSAs were performed using a HeLa cell nuclear extract and a probe consisting of the minimal promoter sequence (Ϫ68/ϩ144). Two bands representing DNA-protein complexes appeared in samples containing nuclear extract (Fig.  1C, lane 2); however, only the larger complex was specific, as revealed by competition with excess unlabeled probe (Fig. 1C,  lane 3). Supershift assays, using an Sp1 antibody that does not cross-react with Sp2, Sp3, or Sp4, resulted in the appearance of a higher mobility band that disappeared in the presence of excess unlabeled probe (Fig. 1C, lanes 4 and 5), thus confirming the presence of Sp1 in the DNA-protein complexes.
hMI-ER1␣ and -␤ Repress Activation of the Minimal Promoter by Sp1-The ability of Sp1 to regulate transcription from the hmi-er1 promoter was investigated in vivo using luciferase reporter assays. HeLa cells were co-transfected with a luciferase reporter construct containing the minimal promoter sequence Ϫ68/ϩ144 (pGL3(Ϫ68)), along with an Sp1 expression vector (pCR-Sp1) or control empty vector (pCR). Co-transfection with Sp1 resulted in a dose-dependent increase in luciferase activity to 4-fold that of control ( Fig. 2A), demonstrating that Sp1 can activate transcription from the hmi-er1 minimal promoter, either directly or indirectly.
We have shown previously that hMI-ER1 functions as a HDAC-dependent transcriptional repressor (2); therefore, we investigated whether hMI-ER1 could regulate transcription from its own promoter. HeLa cells were co-transfected with pGL3(Ϫ68) along with a plasmid expressing Myc tag alone (pMyc) or fused to hMI-ER␣ (pMyc-␣) or hMI-ER1␤ (pMyc-␤). Myc-␣ and Myc-␤ repressed the activity of the hmi-er1 minimal promoter to 40 and 33% of control, respectively (Fig. 2B). The HDAC dependence of this repression was determined by assay-ing in the presence of TSA. Fig. 2C revealed that repression was not relieved by TSA, demonstrating that hMI-ER1 represses its own promoter by a HDAC-independent mechanism. HeLa cells were co-transfected with pGL3(Ϫ68) and pCR-Sp1; the amount of plasmid (g) used for transfection is indicated below each bar. In each case, the amount of pCR empty vector was adjusted so that the total amount of DNA used in each transfection was constant. Cells were harvested 48 h after transfection, and the luciferase activity of each sample was determined as described in the legend to  3. hMI-ER1␣ and ␤ repress activation of the minimal promoter by Sp1. HeLa cells were co-transfected with pGL3(Ϫ68) along with the indicated amount (g) of the plasmids listed below each histogram. In each case, the amount of pMyc was adjusted so that the total amount of DNA used in each transfection was constant. Cells were harvested 48 h after transfection, and the luciferase activity was determined as described in the legend to Fig. 1 An obvious question is whether hMI-ER1␣ and -␤ can affect transcriptional activation by Sp1. To investigate this, HeLa cells were co-transfected with pGL3(Ϫ68), pCR-Sp1, and increasing concentrations of pMyc, pMyc-␣, or pMyc-␤. Promoter activation by Sp1 was repressed in a dose-dependent manner by both hMI-ER1␣ and -␤ (Fig. 3), and repression resulted in a 50% reduction in luciferase activity at the highest hMI-ER1 levels. This repression was not due to a down-regulation of Sp1 protein, which remained constant at all hMI-ER1 levels (Fig. 3).
We have confirmed that this repression also occurs at the level of the endogenous hmi-er1 gene, using dox-inducible hMI-ER1 HeLa cell lines (Fig. 4A). We treated hMI-ER1␣-expressing (HT␣222), hMI-ER1␤-expressing (HT␤53), and control (HTC314) cell lines with 2 g/ml dox to induce hMI-ER1 expression and then extracted RNA for real time RT-PCR and for semiquantitative RT-PCR analysis. We compared changes in the steady state levels of endogenous hmi-er1 mRNA using primers in the noncoding region of hmi-er1 that are absent in the transfected construct. Fig. 4B shows that dox-induced expression of hMI-ER1 reduced steady state levels of endogenous hmi-er1 mRNA by 50% in HT␣222 cells and 42% in HT␤53 cells, relative to the expression level in HTC314 cells. This level of repression is consistent with the level of repression by hMI-ER1 observed in transient transfection approaches reported in Fig. 3A. A similar reduction in the expression level of endogenous hmi-er1 mRNA was observed using semiquantitative PCR (Fig. 4C). As expected, amplification using primers in the coding region revealed an increase in the total hmi-er1 mRNA expression level, due to dox induction of the transfected hmi-er1 cDNA, whereas no discernable effect of the steady state levels of endogenous ␤-actin was observed (Fig. 4C).

hMI-ER1 Is Associated with the Chromatin of Its Own Promoter and Interferes with Sp1
Binding-Although we have demonstrated that Sp1 can bind to recognition sequences in the hmi-er1 promoter, it is important to demonstrate that endogenous Sp1 can associate with the hmi-er1 promoter in vivo. Therefore, we performed ChIP assays using HeLa cells to investigate this possibility. Formaldehyde-cross-linked, sheared chromatin was isolated and subjected to immunoprecipitation using an anti-Sp1 antibody or nonimmune serum. Following purification and reversal of the cross-links, the DNA was amplified using primer sets flanking the minimal hmi-er1 promoter. Fig. 5A shows that the hmi-er1 promoter could be am-

FIG. 5. hMI-ER1 is associated with the chromatin of its own promoter and interferes with Sp1 binding.
A, Sp1 is associated with the chromatin of the hmi-er1 promoter. Chromatin from HeLa cells was cross-linked, sheared, and subjected to imuunoprecipitation with anti-Sp1 or nonimmune antibodies (Non), as described under "Experimental Procedures." DNA from the immunoprecipitates (lanes 1 and 2) or from the chromatin input (lane 4) was amplified using primers that flank the hmi-er1 minimal promoter. Negative controls (lanes 3 and 5) consisted of samples that did not contain chromatin. B, hMI-ER1 is associated with the chromatin of the hmi-er1 promoter. Chromatin from hMI-ER1␤-expressing Tet-On cells was cross-linked with formaldehyde alone (ϪDTBP) or with DTBP followed by formaldehyde (ϩDTBP), as described under "Experimental Procedures." Cross-linked chromatin was immunoprecipitated with anti-pan-hMI-ER1 (lanes 3 and 7) or preimmune (lanes 2 and 6) antiserum, and DNA from the immunoprecipitates or from input chromatin (lanes 1 and 5) was amplified using primers flanking the minimal promoter. The negative control samples (lanes 4 and 8) were processed in the absence of added chromatin. C, overexpression of hMI-ER1 prevents Sp1 association with the hmi-er1 promoter. Chromatin from dox-induced control (lanes 1-4) or hMI-ER1␤-expressing Tet-On cells (lanes 5-8) was prepared, immunoprecipitated, and analyzed by PCR as described in the legend to Fig. 5A. hMI-ER1 SANT Domain Regulates Sp1 Activity plified from chromatin immunoprecipitated with anti-Sp1 (lane 1) but not with nonimmune serum (lane 2). Next ChIP assays were performed to determine whether hMI-ER1 was associated with the chromatin of the endogenous hmi-er1 promoter region. HT␤53 cells were induced with 2 g/ml dox and then fixed with either formaldehyde alone or with an initial protein-protein cross-linking step using DTBP, followed by formaldehyde. Sheared chromatin was isolated and subjected to ChIP using a pan-hMI-ER1 antibody (4). As can be seen in Fig. 5B, DNA sequence encompassing the hmi-er1 promoter co-immunoprecipitated with the hMI-ER1 protein, but only in the presence of DTBP (compare lanes 3 and 7). The specificity of this interaction was confirmed using preimmune serum (lanes 2 and 6). The failure to cross-link hMI-ER1 protein to its promoter with formaldehyde alone is not a technical difficulty, since Sp1 was always associated with the promoter under similar conditions (see Fig. 5A, lane 1). Finally, we utilized ChIP analysis to determine whether there was an interaction between hMI-ER1 and Sp1 proteins at the level of the promoter chromatin. We isolated formaldehyde cross-linked chromatin from dox-induced HT␤53 cells and HTC314 control cells and subjected it to immunoprecipitation with Sp1 or nonimmune antibodies, followed by amplification of the DNA from the hmi-er1 promoter region. The results in Fig. 5C demonstrate that Sp1 protein was associated with the hmi-er1 promoter in control cells (lane 1) but not in cells expressing elevated levels of hMI-ER1 protein (lane 5). Taken together, these results demonstrate that both hMI-ER1 and Sp1 are associated with the chromatin in the hmi-er1 promoter and that hMI-ER1 interferes with Sp1 binding to the endogenous hmi-er1 promoter.
hMI-ER1␣ and -␤ Do Not Bind to DNA but Physically Interact with Sp1 in Vitro and in Vivo-There are several possible mechanisms by which hMI-ER1 could interfere with Sp1 binding to the promoter. One possibility is that hMI-ER1 binds to specific DNA sites in the hmi-er1 promoter. The SANT domain present in hMI-ER1 is related to the DNA binding domain of the c-myb proto-oncogene (8), and therefore we investigated the possibility that hMI-ER1 binds to DNA sequences in its own promoter using EMSAs. EMSAs were performed using a probe containing the minimal promoter sequence (Ϫ68/ϩ144) and purified GST-Sp1, GST-hMI-ER1␣ (GST-␣), or ␤ (GST-␤), in the presence or absence of excess unlabeled probe. As shown in Fig. 6A, GST-Sp1 binds specifically to the minimal promoter, whereas GST-␣ and GST-␤ do not. These results are in agreement with our previous results, which could not identify a consensus DNA binding site for hMI-ER1 using CASTing techniques (2), and argue against hMI-ER1 binding to cis-acting elements as a possible mechanism for repression of its own promoter. These data in combination with the results of our ChIP assays suggest that hMI-ER1 is not bound to DNA but to protein component(s) of the chromatin containing the endogenous hmi-er1 promoter.
Next we examined whether this was a direct interaction between hMI-ER1 and Sp1 proteins by investigating the ability of hMI-ER1␣ and -␤ isoforms to physically associate with Sp1, using co-immunoprecipitation assays. 35 S-labeled Sp1 was synthesized in vitro, mixed with GST-␣ or GST-␤, and then subjected to pull-down assays (Fig. 6B). Sp1 protein was detected in pull-downs with both hMI-ER1␣ and -␤ (Fig. 6B, lanes 3 and  4), but not with GST alone (Fig. 6B, lane 2).
In vivo interaction between hMI-ER1 and Sp1 was examined by transiently expressing Myc, Myc-␣, or Myc-␤ in HeLa cells. Cell extracts were subjected to immunoprecipitation with anti-Sp1, followed by Western blotting with anti-Myc. As shown in Fig. 6C, both hMI-ER1␣ and -␤ co-immunoprecipitated with endogenous Sp1. To verify that endogenous, native complexes containing hMI-ER1 and Sp1 exist in the cell, co-immunoprecipitation analysis of extracts from nontransfected HeLa cells was performed, using preimmune serum or a pan-hMI-ER1 antibody. Sp1 protein was detected in the hMI-ER1 immunoprecipitate (Fig. 6D, lane 2), but not in the control (Fig. 6D, lane  1), demonstrating that endogenous hMI-ER1 and Sp1 proteins can physically associate in vivo.
FIG. 6. hMI-ER1␣ and -␤ do not bind to DNA but physically associate with Sp1 protein. A, hMI-ER1 does not bind to the minimal promoter in vitro. EMSAs were performed as described under "Experimental Procedures," using 32 P(Ϫ68) and the indicated GST fusion protein, in the presence (lanes 1, 3, 5, and 7) or absence (lanes 2, 4, 6, and 8) of excess unlabeled probe. The position of the free probe is indicated. B, hMI-ER1 interacts with Sp1 in vitro. 35 S-labeled TNTs programmed with cDNA encoding Sp1 were incubated with GST, GST-␣, or GST-␤ and subjected to pull-downs with glutathione-Sepharose beads. One-twentieth volume of the TNT used for pull-downs was loaded into lane 1. Proteins were visualized by SDS-PAGE and autoradiography. C, hMI-ER1 interacts with Sp1 in vivo. Cell lysates from HeLa cells transiently transfected with pMyc, pMyc-␣, or pMyc-␤ were prepared, and equivalent amounts of protein from each sample were subjected to IP with anti-Sp1 (lanes 5, 7, and 9) or with nonimmune serum (Non) (lanes 4, 6, and 8). Input lanes (lanes 1-3)  The SANT Domain of hMI-ER1 Is Required for Interaction with Sp1-To determine which region of the hMI-ER1 protein is required for interaction with Sp1, a series of GST-␣ and -␤ deletions was constructed (Fig. 7A), and pull-down assays were performed using 35 S-labeled Sp1. Our deletion analysis revealed that the sequence required for interaction with Sp1 maps to aa 287-357 (Fig. 7B), a region that contains the SANT domain (aa 288 -332). Further deletion of this region at the C-terminal end to remove Lys 331 -Lys 332 (KK), the last 2 aa of the predicted SANT domain (8), abolished interaction with Sp1 (Fig. 7B, lane 8), demonstrating the importance of an intact SANT domain for interaction with Sp1.
hMI-ER1 ␤ Interferes with GC Box Recognition by Sp1-The functional consequence of hMI-ER1 interaction with Sp1 was investigated by examining the effect of hMI-ER1 on Sp1 binding to the hmi-er1 promoter. EMSAs were performed using GST-Sp1 and the minimal promoter sequence (Ϫ68 to ϩ144), in the presence or absence of various GST-hMI-ER1 fusion proteins. As shown in Fig. 7C, the addition of GST-␣ or GST-␤ to the EMSA reaction mixtures resulted in loss of Sp1-DNA complexes (lanes 3 and 4), whereas the addition of GST alone had no effect (lane 2). The addition of GST-(287-357), containing an intact SANT domain, also resulted in loss of Sp1-DNA complexes (lane 5), whereas the addition of GST-(287-330), containing a deletion of the last 2 aa of the SANT domain, did not (lane 6). These results demonstrate that interaction of hMI-ER1 with Sp1 intereferes with the latter's ability to bind DNA.
Identical results were obtained when a consensus Sp1 binding site oligonucleotide was used (Fig. 7D), demonstrating that hMI-ER1 interference with Sp1 binding to DNA is not specific to the hmi-er1 promoter and suggests that other Sp1-regulated genes may be repressed by a similar mechanism.

DISCUSSION
The transcription of a eukaryotic gene is regulated by the combined action of multiple sequence-specific transcription factors, general transcription factors, histone modifiers, cofactors, and mediators that regulate transcription factor activity and chromatin structure. Our previous studies revealed the hMI-ER1 is a potent transcriptional regulator that can repress transcription from a heterologous promoter and that this activity is dependent upon recruitment of HDAC1 activity by the conserved ELM2 domain (2). In contrast, we show here that the activity of the hmi-er1 promoter was repressed by hMI-ER1 in an HDAC-independent manner and involved interference with Sp1 binding.
Sp1 is a sequence-specific transcription factor that binds GC and GT boxes to activate a wide range of viral and cellular genes (reviewed in Ref. 20). Sp1 is important both in transcription initiation and activation, and it can be regulated by multiple mechanisms in a cell type-specific and promoter contextspecific manner (reviewed in Ref. 21). Sp1 has been linked to the maintenance of methylation-free CpG islands (22), and hypermethylation around Sp1 binding sites has been reported to reduce Sp1 binding, thereby decreasing transcription (23). There are several other mechanisms that serve to regulate transcription through Sp1. Transcription regulators, such as Sp3, Sp4, and BTEB3, compete with Sp1 for binding to core cis-elements and repress transcription (24,25). Sp1 activity can also be regulated through protein-protein interactions. Factors that interact with Sp1 include E2F1, GATA1, and YY1, all of which act synergistically with Sp1 on DNA to increase transcriptional activity (26 -29). Another set of Sp1-interacting transcription factors that impair Sp1-mediated transcriptional activity includes p107, PML, FBI-1, TAF-1, and MDM2 (30 -34). Furthermore, HDAC1-containing complexes can be recruited directly by Sp1 for transcriptional repression (35). Whereas Rb has not been shown to bind Sp1 directly, it can increase Sp1 activity by releasing Sp1 from MDM2-Sp1 complexes (34). Post-translational modifications of Sp1, such as phosphorylation, are also critical for regulating Sp1 activity (reviewed in Ref. 21). Thus, there are several different kinds of mechanisms for regulating Sp1 transcriptional activity.
The present study showed that hMI-ER1 could form complexes with Sp1 in vitro and in vivo and that this interaction interferes with Sp1 recognition and binding to GC boxes. Thus, hMI-ER1 represses transcription of its own promoter, most likely by interfering with the chromatin association and DNA binding activity of Sp1 in vivo. This transcriptional regulatory mechanism is distinct from that reported previously for the hMI-ER1, in which repression was dependent upon TSA-sensitive HDAC1 recruitment through the ELM2 domain (2). This indicates that hMI-ER1 can function as a transcription repressor through both HDAC-dependent and -independent mechanisms, utilizing distinct domains. Both HDAC-dependent and -independent transcriptional repression mechanisms have been reported for other transcription regulators, such as Rb (36), Stra13 (37), and LCoR (38).
The mechanism of hMI-ER1 interference with Sp1 required an intact SANT domain, and this was necessary and sufficient to prevent Sp1 binding to its cognate site. The SANT domain is present exclusively in nuclear proteins that have an important role in the regulation of transcription (8). This domain is related in primary and secondary structure to the DNA binding motif of the Myb oncoprotein and other Myb-like domains, including the homeodomains (8). A subset of SANT-containing proteins possess two or three related repeats of this domain, each of which serve distinct functions. For example, the nuclear hormone co-repressor, N-CoR, contains two SANT domains; its N-terminal SANT1 domain serves as a deacetylase activation domain that binds and activates HDAC3, whereas the C-terminal SANT2 domain functions as a histone interaction domain, preferentially targeting nonacetylated histones and inhibiting histone acetyltransferase activity (12). Thus, the two domains function synergistically to repress specific regions of chromatin by targeting HDACs and repressing associated histone acetyltransferase activity. The SANT domains of single SANT proteins, like hMI-ER1 and the transcription regulator ADA2, are predicted to be involved in protein-protein interactions. Interestingly, the single SANT domain of ADA2 has been shown to subserve two different functions in yeast. The N-terminal portion is critical for efficient acetylation of histone targets by GCN5 in the context of the SAGA transcription regulatory complex, whereas the C-terminal portion functions as an interaction domain for GCN5 or GCN5containing complexes (11). Our C-terminal deletion of the hMI-ER1 SANT domain in this study eliminates hMI-ER1 interaction with Sp1, suggesting that this is an important interaction domain for several transcription regulatory factors and complexes. The majority of single SANT domain proteins also possess an ELM2 domain that functions to recruit HDAC1 activity (2) in a similar manner to the SANT1 domain in N-CoR. One might have expected then that hMI-ER1 would displace Sp1 from its binding site by recruiting HDAC1 to modify the chromatin in and around the GC box in the promoter or by modifying Sp1 itself. However, our data show that hMI-ER1-mediated repression of Sp1 activity does not involve HDAC but direct binding to Sp1 itself. Together, these data reveal a novel mechanism for the negative regulation of Sp1 target promoters. The role of hMI-ER1 in regulating the expression of other cellular genes, particularly known Sp1 target genes, is currently being investigated.
Thus hMI-ER1, like other SANT-containing proteins including N-CoR, SMRT, and MTA1-3, is part of a multiprotein transcriptional regulatory complex with several critical activities from modulating DNA binding to altering chromatin structure (2). The function of these complexes and the molecules in them is context-dependent and is determined by the presence, stoichiometry, and modifications of these regulatory molecules in different cell types to precisely regulate cellular responses and behavior. Feedback regulation of hMIER1 levels provides a precise mechanism for regulating a critical component of the transcriptional machinery.