Transcriptional Repression Mediated by the PR Domain Zinc Finger Gene RIZ *

The RIZ (G3B orMTB-Zf) zinc finger gene is structurally related to the myeloid leukemia gene, MDS1-EVI1, and the transcription repressor/differentiation factor, PRDI-BF1/BLIMP1, through a conserved amino-terminal motif, the PR domain. Similar toMDS1-EVI1, RIZ gene normally produces two protein products that differ by the PR domain. The smaller protein RIZ2 lacks the PR domain of RIZ1 but is otherwise identical to RIZ1. Here we show that RIZ proteins bind to GC-rich or Sp-1-binding elements and repress transcription. Both RIZ1 and RIZ2 repressed the herpes simplex virus thymidine kinase (HSV-TK) promoter, one of the best characterized eukaryotic promoters. Recombinant RIZ1 proteins were able to bind to HSV-TK promoter. This binding was mediated by the GC-rich Sp-1 elements of the promoter and the first three zinc finger motifs of RIZ1. RIZ also encodes a repressor domain that was mapped to the central region of the protein. Fusion of this region to the GAL4 DNA-binding domain generated GAL4 site-dependent transcriptional repressors. We also show that RIZ1 protein can efficiently repress the simian virus 40 (SV40) early promoter, which primarily consists of Sp-1 sites; RIZ2, however, only weakly repressed this promoter, suggesting a role for PR in modulating RIZ protein function. The data have implications for a role of RIZ proteins in the regulation of cellular gene promoters, many of which are characterized by GC-rich elements.

The RIZ zinc finger gene was isolated in a functional screening for proteins that can interact with the retinoblastoma tumor suppressor protein (1,2). The biological role of this interaction remains unclear because of the limited understanding of RIZ gene function. The predicted rat and human RIZ proteins are of 1706 and 1719 amino acids respectively, and are highly homologous (84% amino acid identity). RIZ gene normally produces two different products, RIZ1 and RIZ2, that are widely expressed (3). An internal promoter generates RIZ2, which is identical to RIZ1 except that it lacks the RIZ1 PR domain that defines a subclass of Krü ppel-like family of zinc finger genes. This RIZ gene structure is remarkably similar to a related PR domain gene MDS1-EVI1 involved in human and murine leukemia (4). An internal promoter within MDS1-EVI1 gene generates the EVI1 myeloid transforming gene product that lacks PR but is otherwise identical to MDS1-EVI1 (5). The PR domain of MDS1-EVI1 is a common target of viral insertions and chromosomal translocations in leukemogenesis (6 -8), suggesting it might play an important biological function. EVI1 has been shown to bind to DNA specifically and can function either as a repressor or a weak activator of transcription (9,10). The function of MDS1-EVI1 remains uncharacterized. It is of considerable interest to determine whether EVI1 and MDS1-EVI1, or RIZ1 and RIZ2, may function differently because of the PR domain.
Another PR domain gene is the PRDI-BF1/BLIMP1 transcription repressor/differentiation factor (11). It can repress the ␤-interferon gene promoter (12) and can induce B-lymphocyte maturation (13). BLIMP1 maps to human chromosome band 6q21-q22.1, a region often deleted in B cell non-Hodgkin lymphoma, suggesting that it might serve as a candidate tumor suppressor for this B cell tumor (14). In support of this notion, BLIMP1 has recently been shown to be a physiological transcriptional repressor of the c-myc oncogene (15). Like its related PR domain genes, RIZ also display properties of transcription factor with a potential role in cell growth and tumorigenesis. RIZ gene products encode DNA-binding as well as transcription factor-binding activities as evidenced by the independent isolation of RIZ as an retinoblastoma-binding protein (RIZ), a DNA-binding protein (MTB-Zf), or as a GATA3 transcription factor binding protein (G3B) (1,16,17). MTB-Zf is essentially identical to RIZ2 (3), and binds to the MTE DNA element GTCATATGAC of human heme-oxygenase-1 gene and can weakly activate transcription (16). RIZ gene maps to human chromosomal band 1p36, a region thought to harbor one or more tumor suppressor genes for a variety of human cancers including those of neurocrest, colon, liver, and breast tissues (16,18). Abnormal RIZ gene expression has been observed in human brain tumors (16). RIZ1 mRNA was commonly found absent or at reduced levels in tumor cell lines and tissues while RIZ2 was always found expressed, 1 consistent with RIZ1 as a potential tumor suppressor.
Better characterization of RIZ gene function is clearly needed to elucidate its potential role in transcriptional regulation and in cell growth control. Here, we have found that RIZ proteins function as transcription repressors and can bind to GC-rich Sp-1 elements. Using the herpes simplex virus thymidine kinase (HSV-TK) 2 promoter as a model system (19), we show that RIZ proteins repress this promoter and can bind to it in vitro. Repression requires DNA-binding domain as well as a repressor domain that is transferable to a heterologous DNAbinding domain. We also show that RIZ proteins repress the simian virus 40 (SV40) promoter, which is primarily governed by GC-rich Sp-1 elements (19). Given that many cellular gene promoters, especially those of growth regulated genes, are characterized by GC-rich elements (20), these results have implications for a potential role of RIZ in the transcriptional regulation of a broad spectrum of cellular genes.

MATERIALS AND METHODS
Cell Culture and Transfections-NIH 3T3 cells were grown in Dulbecco's modified Eagle's medium plus 10% calf serum. For transfections of these cells, a calcium phosphate precipitation procedure was used (21). One day before transfection, the cells were split to a density of 2 ϫ 10 5 cells/60-mm dish. The cells were given fresh media 0.5 h prior to transfection. The standard co-transfection included 1.5 g of reporter gene, 8 g of expression construct, and 1 g of a ␤-galactosidase expression construct driven by the CMV promoter (pCMV␤) (22). The DNA precipitate was left on the cells for 16 -20 h, and then the medium was changed. Cells were processed 24 h after withdrawal of DNA for immunoblot and chloramphenicol acetyltransferase (CAT) or luciferase analysis. Equivalent amounts of cell extracts, made by freeze-thaw lysis, were incubated with acetyl-CoA and [ 14 C]chloramphenicol for 1-2 h, and percent conversion of the chloramphenicol to the acetylated form was measured by thin layer chromatography followed by quantitation on a phosphoimager radioanalytic imaging system scanner. Luciferase activity in cell extracts was determined by measurement of chemiluminescence using D-luciferin-potassium salt as substrate (Analytical Luminescence Laboratory) and the MicroLumat LB96P microplate luminometer system (EG&G Berthold). The cell extracts were also assayed for ␤-galactosidase activity using standard procedures (23). All CAT or luciferase values were normalized based on the respective ␤-galactosidase activities. The data represent the average of at least three independent experiments. Multiple independent transfections did not show any correlation between RIZ protein expression and ␤-galactosidase activities, showing that RIZ did not affect CMV promoter, which drives ␤-galactosidase expression.
Immunoblot and Immunoprecipitation Analysis-GAL4-(1-147) antiserum (Santa Cruz Biotechnologies, Inc.) was used for immunoblot analysis to verify the expression of GAL4-RIZ fusion proteins. Monoclonal antibodies 2D7 and P4E1 were used for immunoblot analysis to verify expression of RIZ proteins (1,3). Nuclear extracts were prepared from transfected cells by rapidly suspending cells in lysis buffer for 3 min (0.5% Nonidet P-40, 0.15 M NaCl, 10 mM Tris-HCl, pH 7.9, and 1 mM EDTA). Nuclei were then pelleted at 2000 ϫ g for 5 min and washed once with lysis buffer without Nonidet P-40. Nuclear proteins were then extracted with high salt buffer (10 mM Hepes, pH 7.9, 0.42 M NaCl, 0.1 mM EDTA, 1.5 mM MgCl 2 , 1.0 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 1 mM leupeptin, and 1 mM aprotinin) by shaking at 4°C for 30 min. The extracts were then cleared by centrifugation at 10,000 ϫ g for 10 min at 4°C, which were then analyzed by SDS-gel electrophoresis. Immunoblot was performed on Immobilon P filters (Millipore) using RIZ or GAL4 antibodies and alkaline phosphataseconjugated goat anti-mouse or anti-rabbit IgG. Immunoprecipitation of nuclear extracts was performed using 2D7 or P4E1 antibody and protein A-Sepharose as described (3).
Plasmid Constructions-pBLCAT2 contains the herpes simplex virus thymidine kinase promoter sequence Ϫ105 to ϩ51 linked to the CAT gene (24). pGL3-promoter (Promega) contains SV40 early promoter linked to luciferase gene. pGL3-control (Promega) contains SV40 early promoter plus enhancer linked to luciferase gene. RIZ1 and RIZ2 protein expression plasmids p3RIZr and p3RIZrKK were previously described (3). For carboxyl-terminal-half deletion mutant p3RIZrB (deleting amino acid 901-1706), p3RIZr plasmid was linearized by BstEII digestion. The linearized plasmid was treated with Klenow and self-ligated. The elimination of BstEII site generated a frameshift and a stop codon at the destroyed BstEII site. For mutant p3RIZrA (deleting amino acids 659 -1706), p3RIZr plasmid was linearized by AflII followed by Klenow treatment and self-ligation generating a frameshift and a stop codon. For deletion of zinc finger 1-3 region in p3RIZrB (deleting amino acids 216 -575), the fragment from StuI to Spe1 was deleted from p3RIZrB to generate p3RIZrB⌬Zf1-3.
Expression and Purification of Recombinant Protein-Plasmids bearing log-phase Escherichia coli XL1-blue cells were induced with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside (isopropyl-␤-D-thiogalactopyranoside) for 3 h at 30°C. Protein extract was made and recombinant GST fusion protein was purified as described in (23).
DNA Immunoprecipitation Assay-This assay was performed essentially as described (27). The 100-base pair TK promoter probe, as used in EMSA, and protein A-Sepharose-bound RIZ (P4E1) immunoprecipitation products were incubated for 1 h at 4°C in 200 l of binding buffer as described for EMSA. The reaction was washed three times with binding buffer, processed as described (27), and analyzed by electrophoresis in a 5% polyacrylamide TBE gel or 1% agarose TBE gel, followed by autoradiography.

RESULTS
During the course of experiments examining the transcriptional regulatory function of RIZ proteins, we found that a commonly used reporter plasmid (pBLCAT2) was consistently repressed by RIZ gene products in transient transfections in 3T3 cells (Fig. 1). Both RIZ1 and RIZ2 proteins can similarly repress the promoter in a dose-dependent manner. The plasmids (p3RIZr and p3RIZrKK) used for transient expression of RIZ1 and RIZ2 proteins have been previously described; the amount of RIZ1 protein produced from plasmid p3RIZr is similar to that of RIZ2 from plasmid p3RIZrKK (3). The construct pBLCAT2 contains the Ϫ105 to ϩ51 sequences of the HSV-TK S-transferase; EMSA, electrophoretic mobility shift assay; TBE, Tris-borate-EDTA. promoter (24). Inspection of the sequence revealed no MTE element (GTCATATGAC), the known binding site for RIZ2/ MTB-Zf. The data suggest that RIZ can repress transcription independent of binding to the MTE DNA element.
RIZ proteins contain eight zinc finger motifs of the C2-H2 and C2-HC classes that are well known DNA-binding motifs; one such motif can be sufficient for sequence-specific DNA binding (28). It is likely that the MTE site might represent only one of the DNA-binding sites to which RIZ could bind. We thus examined whether RIZ proteins could directly bind to HSV-TK promoter, which might, at least in part, be responsible for the repression in Fig. 1. For these experiments, several GST fusion proteins containing various RIZ zinc finger motifs were expressed in E. coli and purified as shown in Fig. 2B. GST fusion protein of zf7-8 migrated slower than predicted in SDS gel; this is likely a result of anomalous migration that is also displayed by full-length RIZ proteins (1).
We next examined the DNA-binding activities of these GST fusion proteins using HSV-TK promoter probe. As shown in Fig. 2C, dose-dependent binding was observed for GST-RIZ-Zf1-3 protein containing the amino-terminal three zinc finger motifs. No binding activity was observed for zinc finger 4 -6 or 7-8 GST fusion proteins. The results show that the aminoterminal three zinc finger motifs of RIZ can directly bind to HSV-TK promoter. To examine whether full-length RIZ1 protein could also bind to HSV-TK, we performed DNA immunoprecipitation assay. Nuclear extracts prepared from p3RIZr-or pcDNA3-transfected 3T3 cells were first immunoprecipitated with P4E1 monoclonal antibody or M73 anti-E1A antibody serving as negative control. The protein A-Sepharose bound products were then incubated with HSV-TK DNA probe. DNAbinding was specifically observed for RIZ1 but not for vectortransfected cells (Fig. 2D), suggesting that full-length RIZ1 can bind to HSV-TK. Whether this is a direct interaction requires future experiments using purified full-length RIZ proteins.
The HSV-TK promoter is one of the best characterized eukaryotic promoters. Two types of DNA elements, Sp-1 or GCrich (GGGCGG), and CTF (CCAAT) sites, have been well established to be essential for the expression of HSV-TK (19) (Fig.  3A). Inspection of the sequence also revealed an E-box like element CAGATG that partly resembles the core sequence of MTE element GTCATATGAC. We next asked whether the binding as shown in Fig. 2 could be mediated by one or more of these elements. Double-stranded oligonucleotides representing each site were used to compete for binding to HSV-TK promoter probe in EMSA assays (Fig. 3B). A consensus Sp-1 site oligonucleotide inhibited binding in a dose-dependent manner. A mutant oligonucleotide with a "GG" to "TT" substitution in the Sp-1 binding site weakly inhibited binding only at high dose. Both the CTF and the E-box oligonucleotides derived from HSV-TK did not significantly inhibit binding. The results show that GC-box Sp-1 site mediates binding of HSV-TK promoter to RIZ. Binding of RIZ protein to Sp-1 site was also more directly demonstrated by EMSA assays using Sp-1 oligonucleotide as probe (Fig. 3C).
A typical transcriptional repressor contains both a DNAbinding domain and a repressor domain. Results described above mapped a GC-rich DNA-binding domain to zinc finger 1-3 of RIZ. To determine whether RIZ1 protein also encodes a repressor domain, we made several deletion mutants of RIZ1. Mutant p3RIZrB (amino acid 1-901) expressing the aminoterminal half of RIZ1 protein (amino acid 1-901) was found to be sufficient for repression; mutant p3RIZrA (amino acid 1-659) showed 2-4-fold less repression activity than p3RIZrB (Fig. 4A). The result suggested that amino acid 659 -901 might encode transcriptional repressor activity. To show that repression by p3RIZrB also requires DNA-binding, we constructed mutant p3RIZrB⌬Zf1-3 that deleted zinc finger 1-3 region (amino acid 216 -575) from p3RIZrB. This deletion completely abolished repression activity (Fig. 4A). Similar levels of expression of these mutant proteins were confirmed by immunoblot analysis (Fig. 4B). The DNA binding activities of these mutant proteins were also examined by DNA immunoprecipitation assay using antibody P4E1, which can immunoprecipitate all these mutant proteins. As expected, mutant p3RIZrA and p3RIZrB retained DNA binding but p3RIZrB⌬Zf1-3 showed no activity (Fig. 4C). The results suggest that repression by RIZ1 requires an intact DNA-binding domain as well as a repressor domain.

FIG. 2. Binding of RIZ proteins to HSV-TK promoter.
A, schematic representation of RIZ zinc finger GST fusion proteins; B, expression of GST fusion proteins of RIZ zinc finger motifs. Purified GST fusion proteins containing various zinc finger motifs as indicated were analyzed by SDS-gel electrophoresis followed by Coomassie Blue staining. C, RIZ zinc finger 1-3 region binds to the HSV-TK promoter. GST fusion proteins (either 50 ng or 250 ng) were used for EMSA assays. The probe was a 32 P-labeled 100-base pair HSV-TK promoter fragment derived from BamHI and MluI digestion of pBLCAT2. D, DNA-binding by full-length RIZ1 protein. Nuclear extracts of 3T3 cells transfected with p3RIZr or pcDNA3 vector were immunoprecipitated with P4E1 antibody or M73 E1A antibody. Proteins bound to protein A-Sepharose were incubated with HSV-TK probe. Lane 1 contains 5% of the total input counts used for each of the other samples. Bound DNA was analyzed on 1% agarose TBE gel. We next asked whether the repressor domain can also function when linked to a heterologous DNA-binding domain. The GAL4 system for mapping transcriptional regulatory motifs was used (26). Several GAL4-(1-147)-DNA-binding domain fusion proteins of RIZ1 were constructed. These fusion protein constructs were assayed by using pGAL4-TK-CAT construct as reporter, which contains five copies of GAL4 binding sites linked to the HSV-TK promoter. Similar to results of Fig. 4, pGRIZ1-(18 -901) fusion protein strongly repressed transcription but pGRIZ1-(18 -659) did so only weakly (Fig. 5A). Amino acids 738 -948 were found to be sufficient to confer repression function to GAL4-(1-147) protein, demonstrating that this region encodes repressor activity. This region has been noted to contain putative GTPase and SH3 motifs (1). To examine whether some of these conserved sequence motifs might play a role in repression function, two point mutations were generated. The point mutation K755N altered a conserved residue in the G1 motif of the putative GTPase domain and did not affect repression. The point mutation L745P that alters the SH3 motif completely disrupted repression activity. Immunoblot analysis of transfected cells showed that these different GAL4 fusion constructs expressed similar levels of proteins (Fig. 5B). The results suggest that sequences encoding the putative SH3 motif but not the GTPase motif might be involved in repressor function.
The results described above suggest that RIZ proteins may function to repress promoters containing GC-rich Sp-1 sites. To test a distinct promoter regulated by Sp-1, we analyzed the effect of RIZ on the SV40 early promoter, which is primarily controlled by Sp-1 sites (19,29). Two reporter constructs were tested: pGL3-promoter, which contains the SV40 early promoter linked to the luciferase gene, and pGL3-control, which also contains the SV40 enhancer in addition to the promoter. As shown in Fig. 6A, RIZ1 efficiently repressed both reporters (3-4-fold), whereas RIZ2 repressed weakly (ϳ1.5-fold). Immunoblot analysis showed that similar levels of RIZ1 and RIZ2 proteins were produced by transient transfection (Fig. 6B). The results suggest a role for PR in regulating the transcriptional repressor function of RIZ proteins.

DISCUSSION
The predicted protein structure and the ways through which the RIZ/G3B/MTB-Zf gene was isolated suggest that this gene might function as a DNA-binding transcription factor. MTB-Zf or RIZ2 has previously been shown to bind to the MTE element GTCATATGAC and to function as a weak transcription activator (16). Here, we provide evidence that RIZ gene

FIG. 3. RIZ binding to HSV-TK promoter is mediated by Sp-1.
A, DNA sequence of HSV-TK promoter. The sequence of HSV-TK promoter (Ϫ105 to ϩ51) in pBLCAT2 plasmid is shown here flanked by BamHI and XhoI sites. Sp-1, CTF, E-box, and TATA box sites are underlined. MluI enzyme site is also underlined. Transcription start site is marked ϩ1. B, Sp-1 consensus oligonucleotide competes for RIZ binding to HSV-TK promoter. EMSA assays using pGST-RIZ1-Zf1-3 protein (50 ng) and HSV-TK promoter probe were performed in the absence (lane 1) or presence of various oligonucleotides representing Sp-1, mutant Sp-1, CTF, and E-box elements as indicated. The amounts of each competitor used was in 25-, 100-, and 400-fold excess over the probe. C, RIZ binding to Sp-1 consensus oligonucleotide. Sp-1 oligonucleotide probe was incubated with GST protein (lane 1) or GST-RIZ1-Zf1-3 protein (lane 2) and analyzed by EMSA assays.

FIG. 4. Mutational analysis of RIZ transcriptional repression function.
A, analysis of transcriptional activity of RIZ mutant proteins. 3T3 cells were transfected with the reporter plasmid pBLCAT2, internal control plasmid pCMV-␤, and RIZ mutant plasmids p3RIZrB, p3RIZrA, or p3RIZrB⌬Zf1-3, whose structures are schematically shown. CAT activity obtained by co-transfection of pBLCAT2 and pcDNA3 was set at 100%. B, immunoblot analysis of RIZ mutant proteins. 3T3 cells were transfected with RIZ mutant protein expression constructs. Nuclear extracts prepared from transfected cells were resolved on SDS gel followed by immunoblot analysis using RIZ antibody P4E1. Lanes 1-4 represent extracts from mock, p3RIZrB⌬Zf1-3, p3RIZrB, and p3RIZrA transfected cells, respectively. Stars mark the positions of expressed proteins. C, DNA immunoprecipitation assay. Nuclear extracts of 3T3 cells transfected with p3RIZrA (lane 2), p3RIZrB (lane 3), p3RIZrB⌬Zf1-3 (lane 4), and pcDNA3 vector (lane 5) were immunoprecipitated with P4E1 antibody. Proteins bound to protein A-Sepharose were incubated with HSV-TK probe. Bound DNA was analyzed on 0.5% polyacrylamide-TBE gel. Lane 1 contains 5% of the total input counts used for each of the other samples. products can bind to a different type of DNA element and function to repress transcription.
We show that RIZ gene products can efficiently repress the transcriptional function of HSV-TK promoter, one of the best characterized eukaryotic promoters. The Sp-1 and CTF sites are well established essential elements for basal expression of this promoter (19). Disruptions of either element impaired promoter function (30). The mechanisms of repression of HSV-TK by RIZ probably involves direct binding of RIZ proteins to the Sp-1 sites of the promoter. We show that RIZ proteins could directly interact with this promoter which can be blocked by consensus Sp-1 oligonucleotide. We also show that the DNA-binding motif of RIZ is required for repression of HSV-TK. Definitive proof for a role of binding to GC-rich elements requires demonstration of loss of RIZ-mediated repression as a result of Sp-1 site mutation. Here, the requirement of Sp-1 sites for basal expression as demonstrated previously (30) could make such an experiment uninformative. An impaired basal activity caused by Sp-1-site mutation may obscure RIZ repression effects. Nevertheless, demonstrating site dependence represents an important future study. It should be noted that several factors other than Sp-1 can also bind to GC-rich motifs such as E2F1 (31) and others (20).
If DNA binding is directly involved in RIZ repression of HSV-TK, a simple mechanism for repression might be through competition with Sp-1 proteins for binding sites. While our results do not rule out this passive mechanism, two lines of evidence suggest that RIZ exerts repression through an active mechanism. First, truncation mutant p3RIZrA showed significantly impaired repression relative to truncation mutant p3RIZrB. While they differ by the central region of RIZ which encodes repressor activity, both mutant proteins contain zinc finger motifs 1-3, which were sufficient for binding to HSV-TK. The observation suggests that DNA-binding alone cannot confer maximal repression.
Second, conforming to the commonly observed modular nature of transcriptional regulatory domains, the repressor region of RIZ (amino acid 738 -948) could act in a heterologous context. Fusion of this region to the GAL4-DNA-binding domain generated a GAL4-binding site-dependent repressor. As noted previously (1), this region of RIZ bears sequence similarity to the consensus sequence of SH3 and GTPase motifs. However, functional significance of this similarity remains unclear. It appears that a GTPase function is unlikely involved in repressor function. The mutant protein p3RIZrB-(1-901), lacking part of the conserved GTPase structure (missing the G4 motif, TQPD), retained full repressor function. A point mutation (K755N), changing the conserved Lys residue in the G1 motif of GTPase domain, did not affect repressor function but is known to disrupt GTP binding of other GTPases (32). In contrast, another point mutation (L745P) outside the GTPase domain completely abolished repressor function. Although substituting to Pro may appear to represent a drastic change, this region is Pro-rich to begin with (Fig. 5C). The result seems to implicate a role for SH3 sequence because the mutated residue is conserved among different SH3 motifs. Evidence for a role of SH3 requires future demonstration of SH3 function for this region, such as binding to proline-rich peptides (33).
Consistent with binding to Sp-1 site, we show that RIZ1 can also efficiently repress the SV40 early promoter, another well characterized eukaryotic promoter whose function is primarily governed by Sp-1 sites (19,29). Using SV40 promoter as a model, we also show that RIZ1 is a stronger repressor than RIZ2. This demonstrates a functional difference between a PR-plus and a PR-minus "twin" protein and suggests a role for PR in modulating protein function. How PR domain might confer a stronger repression function on RIZ1 protein represents an important future investigation. The PR domain appeared not to have intrinsic repressor function: the deletion mutant p3RIZrA protein contains PR and DNA-binding domain and yet is a much weaker repressor than the mutant p3RIZrB, which contains repressor domain. Also, no repression activity was found for a GAL4-PR fusion protein (data not shown). 3 These observations suggest that PR domain might be indirectly involved in transcriptional repression.
It is not clear why the difference between RIZ1 and RIZ2 as seen for the SV40 early promoter was not observed for the HSV-TK promoter. This may indicate a dependence for RIZ protein function on promoter context. Indeed, some promoters were not affected by RIZ proteins, such as the CMV promoter in pCMV␤ reporter mentioned in Experimental Procedures. Certain promoter can even be activated by RIZ as in the case of human heme-oxygenase-1 gene promoter (16). We note that this promoter contains a single Sp-1 site, but this site appears dispensable for basal expression (34). These observations are consistent with the emerging theme that promoter function is determined by specific combination or interaction of multiple factors assembled on a specific promoter (35).
It is interesting to note that other PR domain genes also function in transcription repression. The PRDI-BF1/BLIMP1 gene has been shown to repress transcription through competition for DNA-binding sites with transcription activators (12). It remains to be examined whether it also contains a repressor domain. The EVI1 gene has been shown to repress transcription through competition with positive factors for DNA binding sites (36). It has also been shown recently to contain a repressor domain that is required for its transforming activity (10). It will be important to examine whether MDS1-EVI1 might function as a better or worse repressor relative to EVI1, which might provide a molecular rational for the specific disruption of MDS1-EVI1 PR in leukemogenesis. The repressor domain of EVI1 bears no similarity to that of RIZ and has been noted to be proline-rich (10), which is a common feature of some repressor domains (37). The repressor region of RIZ has multiple prolines but is clearly not the most proline-rich region of the protein. It is intriguing that the RIZ repressor region resembles SH3 domain, which is known to bind to proline-rich peptides. As transcriptional activation or repression domains are known to act through protein-protein contact, it is not inconceivable that a SH3-proline-type interaction could be involved in the mechanisms of transcriptional repressor domain action.
Because GC-rich elements are extremely common in cellular gene promoters, the results presented here suggest a role for RIZ proteins in the regulation of a potentially large number of cellular genes. In particular, many growth-regulated genes are characterized by GC-rich and TATA-less promoters including oncogenes, growth factors, and their receptors, transcription factors, and housekeeping enzymes (20). In many cases where it has been studied, the GC-rich element appears to be essential for either basal expression or for activation in response to growth regulatory signals. Future investigations will be needed to determine whether RIZ proteins might repress cellular growth-regulated genes and in turn exert control over cell growth.