Sp3 encodes multiple proteins that differ in their capacity to stimulate or repress transcription
Sp3 encodes multiple proteins that differ in their capacity to stimulate or repress transcriptionSarah B. Kennett, Ava J. Udvadia and Jonathan M. Horowitz*
Departments of Molecular Cancer Biology and Microbiology, Box 3686, Duke University Medical Center, Durham, NC 27710, USA
Received March 19, 1997;Revised and Accepted June 6, 1997
ABSTRACT
The product of the retinoblastoma (Rb) susceptibility gene (RB-1) regulates expression of a variety of growth control genes via discrete promoter elements termed retinoblastoma control elements (RCEs). We have previously shown that RCEs are bound and regulated by a common set of ubiquitously expressed nuclear proteins of 115, 95 and 80 kDa, termed retinoblastoma control proteins (RCPs). We have also previously determined that Sp3 and Sp1, two members of the Sp family of transcription factors, encode the 115 and 95 kDa RCPs respectively and that Rb stimulates Sp1/Sp3-mediated transcription in vivo. In this report we have extended these results by determining that the 80 kDa RCP arises from Sp3 mRNA via translational initiation at two internal sites located within the Sp3 trans-activation domain. Internally initiated Sp3 proteins readily bind to Sp1 binding sites in vitro yet have little or no capacity to stimulate transcription of Sp-regulated genes in vivo. Instead, these Sp3-derived proteins function as potent inhibitors of Sp1/Sp3- mediated transcription. Since cell cycle- or signal- induced expression of a variety of genes, including p21waf1/cip1, p15INK4B, CYP11A, mdr1 and acetyl-CoA carboxylase, have been mapped to GC-rich promoter elements that bind Sp family members, we speculate that alterations of the protein and/or DNA binding activities of internally initiated Sp3 isoforms may account in part for the regulation of such differentially expressed genes.
INTRODUCTION
Cell cycle progression is mediated in part by the carefully orchestrated transcription of growth control genes. One function of the retinoblastoma (Rb) protein, the product of a tumor suppressor gene (RB-1), is to couple transcription to progression of the cell cycle (1 ). Deletion or functional inactivation of RB-1 is associated with outgrowth of a variety of human cancers, including retinoblastoma, osteosarcoma and breast, bladder and small cell lung carcinomas (for reviews see 2 -5 ). Rb physically and/or functionally interacts with a number of transcription factors, including E2F, Elf-1, PU.1, ATF-2, Sp1/3 and myoD, resulting in repression or stimulation of trans-activation depending on the transcription factor and cell type examined (for reviews see 6 ,7 ).
Rb governs the synthesis of a subset of growth control genes, including c-fos, c-myc and TGF-[beta]1, via discrete portions of their promoters, termed retinoblastoma control elements (RCEs; 8 -10 ). Using RCEs derived from each of these Rb-regulated genes as probes, protein-DNA binding assays have identified a common set of three protein-DNA complexes (denoted 1A, 1B and 2) in mammalian cell nuclear extracts (11 ,12 ). The proteins (retinoblastoma control proteins; RCPs) bound to RCEs in vitro were shown to be ubiquitously expressed and to have apparent molecular weights of 115, 95 and 80 kDa. The physiological relevance of one or more of these RCE binding proteins was underscored by the observation that point mutations that abolish binding of RCPs in vitro also abrogate RCE-mediated transcription in vivo. Mutational analyses of RCEs and methylation-interference (`footprinting') assays resulted in the derivation of a 6 nt consensus sequence, 5'-GCCACC-3', that is essential for binding of RCPs in vitro and RCE function in vivo. This consensus RCP binding site was noted to be reminiscent of GC-rich sequences (GC boxes) bound by transcription factor Sp1, and by a variety of immunochemical and biochemical criteria the 95 kDa RCP was subsequently shown to be encoded by Sp1 (12 ,13 ).
Sp1 is a ubiquitously expressed nuclear protein that was initially identified as a protein that binds and stimulates transcription of the SV40 early promoter (14 ,15 ). Since this initial observation, Sp1 has been shown to play a central role in trans-activation of a multitude of cellular promoters, including the aforementioned RCE-containing promoters, dihydrofolate reductase (DHFR), thymidine kinase (TK), p21waf1/cip1, p15INK4B, GM-CSF, EGFR, IL2-R[alpha], IGF-II, SIS/PDGF-B and hIR (13 ,16 -25 ). Sp1 physically interacts with other sequence-specific DNA binding proteins; interactions with transcription factor PU.1 appear to impart cell specificity to Sp1-mediated transcription, whereas interactions with proteins such as E2F-1, BPV E2 protein or p53 result in synergistic trans-activation (22 ,26 -30 ). In turn, physical interactions between Sp1 and components of the basal transcription machinery, such as the TATA box binding protein (TBP) and TBP-associated factors (TAFs) 110 and 55, catalyze initiation of transcription by RNA polymerase II (31 -33 ). We and others have shown that the functional consequence of co-expression of Rb and Sp1 is stimulation (`superactivation') of Sp1-mediated transcription (12 ,13 ). Moreover, this functional interaction is likely to be physiologically significant, as Rb amino acids required for tumor suppression are also required for `superactivation' of Sp1-mediated transcription (34 ). The precise mechanism(s) whereby Rb stimulates Sp1 trans-activation remains to be resolved (6 ).
Sp1 is a member of a burgeoning family of related transcription factors (35 -37 ). Sp2, Sp3 and Sp4 share extensive structural and sequence homology with Sp1 and each of these Sp family members binds specifically to DNA. Sp3, like Sp1, is a ubiquitously expressed protein, whereas the distribution of Sp2 and Sp4 appears to be more limited (35 ,36 ). Given that Sp family members share similar cognate DNA binding sites, we hypothesized that Sp3, a 110 kDa protein, was likely to encode the 115 kDa RCP. This hypothesis was subsequently confirmed and we and others have shown that Sp3 stimulates transcription of RCE- and GC box-containing promoters in vivo (21 ,34 ). Moreover, as previously noted for Sp1, Rb co-expression leads to `superactivation' of Sp3-mediated transcription (34 ). To further our understanding of Rb-mediated transcription, we have pursued identification and characterization of the remaining (80 kDa) RCE binding protein identified in nuclear extracts. Here we report that this 80 kDa RCP is encoded by Sp3 and arises via internal translational initiation within Sp3 mRNA. We also demonstrate that internally initiated Sp3-derived proteins have little or no inherent trans-activation activity. Instead, these novel Sp3-derived proteins function as potent inhibitors of Sp3- and Sp1-mediated trans-activation.
MATERIALS AND METHODS
Cell culture
Human cervical carcinoma line C-33A and Drosophila SL2 cells were cultured as previously described (11 ,12 ).
Plasmid constructions
pCR-M1/flu and pCR-M2/flu were prepared from plasmid pSP72-Sp3 (36 ; provided by Astar Winoto, University of California-Berkeley, Berkeley, CA) using PCR and one of two 5' oligonucleotides (M1, 5'-GGGGGATCCATGACTGCAGGCATTAATG-3'; M2, 5'-GGGGGATCCATGGATAGTTCAGACAATTCA-3') and a 3' oligonucleotide (5'-GGGGGATCCCTAGCT- AGCGTAATCTGGAACATCGTATGGGTACTCCATTGTCTCATTTCCA-3'). Resulting M1 and M2 Sp3-derived cDNAs carried N-terminal truncations of 217 and 234 amino acids respectively and an influenza hemagglutinin (HA) epitope tag at their C-termini. PCR-amplified, epitope-tagged M1 and M2 cDNAs were cloned into pCR-ScriptCamSK(+) (Stratagene Inc., La Jolla, CA) for in vitro transcription/translation experiments. pPacM1 and pPacM2 were constructed by subcloning M1 and M2 cDNAs from pCR-M1/flu and pCR-M2/flu into pPac, an insect cell expression vector driven by the Drosophilaactin5C promoter (38 ). pBSK-Sp3/flu was prepared by subcloning an Sp3 cDNA carrying a C-terminal HA epitope tag from CMV-Sp3/flu (34 ) into pBSK(+) (Stratagene Inc.). pPacSp1 and pPacSp3 expression vectors have previously been described (12 ,34 ). A GAL4-VP16 expression vector and GAL4-CAT and DHFR-CAT reporter plasmids have been previously described (16 ,39 ). A glutathione S-transferase-M1 fusion protein, GST-M1/flu, was created by subcloning the M1 cDNA from pCR-M1/flu into the BamHI site of pGEX-2TK (Pharmacia Inc., San Fransisco, CA).
Antibodies
Anti-Sp3 trans-activation domain antiserum was prepared as described (34 ) using the N-terminal 300 amino acids of Sp3 as immunogen. Anti-Sp3 DNA binding domain antiserum (a kind gift of Astar Winoto) was prepared against a GST fusion protein that carries the C-terminal 415 amino acids of Sp3 (34 ). A polyclonal anti-Sp1 antiserum prepared against the Sp1 trans-activation domain has previously been described (12 ). Epitope-tagged proteins were detected with the anti-HA monoclonal antibody 12CA5 (12 ) employed as hybridoma supernatants.
Protein-DNA binding assays
Nuclear extracts from C-33A and SL2 cells were prepared as previously described (11 ,12 ). Synthetic oligonucleotides employed in protein-DNA binding assays were synthesized and purified as described (11 ). A c-fos RCE-derived oligonucleotide carrying a high affinity binding site for RCPs (5'Fos-4, 5'-CCCTTGCGCCACCCCTCT-3') and a mutated derivative lacking this site (5'Fos-2, 5'-CCCGCGCTTCACCCCTCT-3') have previously been described (11 ). Oligonucleotides were radiolabeled with [[alpha]-32P]GTP and [[alpha]-32P]CTP (ICN Inc., Costa Mesa, CA) and Klenow polymerase (New England Biolabs, Beverly, MA) and employed in electrophoretic mobility shift assays as described (11 ).
Western blotting and immunoprecipitation
Western blots using anti-Sp1 antiserum were performed as described (40 ). Western blots with anti-Sp3 antiserum were performed as for anti-Sp1, using a 1:2000 dilution of primary antibody and a 1:40 000 dilution of secondary antibody. Antibody-antigen complexes were detected using an enhanced chemiluminescent system (ECL; Amersham Corp., Arlington Heights, IL). Immunoprecipitations were performed as previously described (40 ).
In vitro transcription/translation
In vitro transcribed/translated proteins were produced using T3 or T7 RNA polymerases (New England Biolabs), pBSK-Sp3/flu, pCR-M1/flu or pCR-M2/flu as template and a coupled reticulocyte lysate system (TNT; Promega Inc., Madison, WI). In vitro translated proteins were radiolabeled with a proprietary cocktail of radiolabeled amino acids (Express; New England Nuclear Inc., Boston, MA).
Transfections and CAT assays
Transient transfection of SL2 cells was performed and nuclear extracts were prepared as previously described (11 ,12 ). Chloramphenicol acetyltransferase (CAT) assays were performed using a liquid scintillation assay as previously described (41 ).
RESULTS
Retinoblastoma control elements (RCEs) are bound by Sp3 and Sp3-related proteins in vitro
We have previously shown, by photoaffinity labeling techniques, that RCEs derived from the c-fos, c-myc and TGF[beta]-1 promoters are bound by three proteins in vitro, termed retinoblastoma control proteins (RCPs), of 115, 95 and 80 kDa (11 ). Using immunochemical and molecular genetic techniques we have also previously determined that two members of the Sp family of transcription factors, Sp3 and Sp1, encode the 115 and 95 kDa RCPs respectively (12 ,34 ). These results relied in part on an anti-Sp3 rabbit antiserum ([alpha]Sp3 TAD) we had prepared against a GST fusion protein carrying the Sp3 trans-activation domain. Surprisingly, subsequent immunizations of this rabbit with the Sp3 trans-activation domain resulted in production of antiserum that also reacts with the 80 kDa protein that forms RCE-RCP complex 2 (Fig. 1 A). Whereas immune serum collected following initial immunizations (Fig. 1 A, lane I-1) greatly diminished the abundance of an Sp3-DNA complex (RCP 1B), aliquots of rabbit antiserum collected subsequently (Fig. 1 A, lane I-5) were noted to abrogate both RCP complexes 1B and 2, leading to a novel `supershifted' complex of decreased mobility (Fig. 1 A, asterisk). These data suggested that the Sp3 trans-activation domain shares one or more antigenic determinants with the 80 kDa RCP. To determine whether the DNA binding domain of Sp3 is also antigenically related to that of the 80 kDa RCP, we employed a polyclonal rabbit antiserum ([alpha]Sp3 DBD), prepared against the C-terminal 415 amino acids of Sp3, in protein-DNA binding assays (34 ). As for later aliquots of [alpha]Sp3 TAD antiserum, this distinct anti-Sp3 antiserum also abrogates the detection of RCP complexes 1B and 2 (Fig. 1 A). In accord with previous results, [alpha]Sp3 TAD preimmune serum does not appreciably alter the abundance of RCP-RCE complexes and anti-Sp3 immune sera do not cross-react with Sp1-DNA complexes (34 ; Fig. 1 A, RCP 1A). Taken together with our previous observation (11 ) that the nucleotide binding specificities of each RCP are identical, results using [alpha]Sp3 TAD and [alpha]Sp3 DBD antisera support the contention that the 80 kDa RCP is antigenically as well as functionally related to Sp3.
The 78-80 kDa Sp3-related proteins are isoforms of Sp3 that arise via internal translational initiation
Since Northern analyses indicate that Sp3 encodes a single ubiquitously expressed mRNA of 4.2 kb, we reasoned that the 78-80 kDa Sp3-related proteins were unlikely to result from alternative splicing of Sp3 message or to be encoded by closely related genes (35 ,36 ). Additionally, since we have previously shown that in vivo expression of an epitope-tagged Sp3 cDNA leads to synthesis of a single protein of ~110 kDa (34 ), we concluded that the 78-80 kDa Sp3-related proteins are unlikely to arise via proteolytic degradation of Sp3. Consequently, we focused our attention on the possibility that the 78-80 kDa Sp3-related proteins might arise via internal translational initiation within Sp3 mRNA.
Scanning the Sp3 cDNA sequence, we noted two closely juxtaposed methionine residues (Fig. 2 A, M1 and M2) embedded within nucleotide sequences compatible with efficient translational initiation and synthesis of proteins of ~80 kDa (43 ). To assess whether translation can occur from these internal methionine residues, we prepared Sp3 cRNA and translated this message in rabbit reticulocyte extracts. As shown in Figure 2 B, in vitro translation of cRNA derived from an epitope-tagged (influenza hemagglutinin, HA) Sp3 cDNA gives rise to three prominent proteins of 115, 80 and 78 kDa. Each of these Sp3-derived proteins are precipitated by antisera prepared against the Sp3 trans-activation domain ([alpha]Sp3 TAD) and the C-terminal epitope tag (Fig. 2 B, [alpha]HA). To confirm that the 80 and 78 kDa Sp3-derived proteins initiate from the predicted internal methionine residues, we utilized Sp3 oligonucleotides and PCR to synthesize epitope-tagged, partial Sp3 cDNAs whose translation initiates at M1 and M2. As expected, translation of cRNA produced from the M1 Sp3 construct leads to synthesis of the 80 and 78 kDa Sp3-derived proteins, whereas M2-derived cRNA directs synthesis of only the 78 kDa isoform of Sp3 in vitro (Fig. 2 B). Thus, internal translational initiation can occur at M1 and M2 in vitro producing proteins that are similar in size to the 78-80 kDa Sp3-related proteins synthesized in vivo.
Internally initiated isoforms of Sp3 function as potent inhibitors of Sp-mediated transcription in vivo
Since M1 and M2 lack substantial portions of the Sp3 trans- activation domain, we speculated that their potency as trans-activators was likely to be significantly compromised. Furthermore, given that M1/M2 appear to have the same DNA binding specificity as Sp1/Sp3, we hypothesized that co-expression of M1 or M2 with Sp1 or Sp3 might interfere with Sp-mediated transcription. To examine these possibilities we utilized a well-characterized promoter-CAT construct prepared from the hamster DHFR promoter as a reporter gene and Drosophila SL2 cells as recipients to monitor Sp-mediated trans-activation in vivo. We and others have previously shown that Sp1 is a potent trans-activator that induces dose-dependent transcription of DHFR and a variety of additional mammalian promoters in SL2 cells (12 ,16 ,38 ,44 ). We have also shown that Sp3-mediated DHFR transcription approximates that of Sp1 (34 ). Consistent with the supposition that truncation of the Sp3 trans-activation domain might negatively effect M1/M2-mediated transcription, transient expression of M1 or M2 in SL2 cells led to little or no DHFR-mediated transcription above that of the reporter gene alone (Table 1 and data not shown). As might be predicted given their common nucleotide binding specificities, co-expression of M2 with Sp1 or Sp3 resulted in a profound reduction in Sp-mediated DHFR transcription (Table 2 ). Similar results were noted in parallel co-transfection experiments with M1 (data not shown). Interestingly, although Sp1 and Sp3 are expressed at equivalent levels in vivo (Fig. 3 B), it is apparent that Sp1 and Sp3 differ in their relative sensitivity to transcriptional inhibition; Sp3-mediated transcription is at least 10-fold more sensitive to transcriptional inhibition by M1/M2 than is Sp1 (Table 2 ). For example, transfection of 50 or 100 ng M2 with 50 ng Sp3 was sufficient to reduce Sp3-mediated transcription by 74 and 91% respectively. In contrast, 1000 ng M2 were required to diminish Sp1-mediated transcription to equivalent levels. The negative effect of M1 and M2 co-expression on Sp-mediated transcription is not restricted to the DHFR promoter, as similar results were obtained using promoter-CAT constructs dependent on RCEs derived from the c-fos, c-myc and TGF-[beta]1 genes (Table 2 ; 11 ,12 ). To confirm that transcriptional inhibition by M1/M2 is specific and not due to generalized disruption of transcription in vivo, co-transfection experiments with a GAL4-CAT reporter construct and a GAL4-VP16 fusion protein, a potent trans-activator, were performed. Consistent with the notion that Sp3-mediated transcriptional inhibition is specific, co-expression of M1 or M2 with GAL4-VP16 did not have a dose-dependent effect on VP16- mediated transcription (Table 2 and data not shown). A slight increase in VP16-mediated transcription was noted at low input amounts of M1/M2 that was not apparent with increasing concentrations of DNA. We conclude from these results that internally initiated isoforms of Sp3 do not function as stimulatory transcription factors in SL2 cells. Instead, such isoforms can function as potent inhibitors of transcription when co-expressed with Sp1 or Sp3. Furthermore, we conclude that Sp3 is significantly more sensitive to transcriptional inhibition by internally initiated Sp3 isoforms than is Sp1.
. Trans-activation of hamster DHFR promoter by Sp-family membersa
Reporter construct
Additional DNA(s)
Relative CAT activity (+-SE)b
nc
DHFR
-
1.0
10
Sp1 50 ng
130 (31)
14
Sp3 50 ng
244 (98)
18
M1 10 ng
1.5 (0.8)
4
M1 50 ng
1.0 (0.1)
4
M1 100 ng
1.2 (0.2)
4
M1 500 ng
2.0 (1.7)
4
M2 10 ng
3.0 (0.7)
4
M2 20 ng
3.2 (1.0)
4
M2 50 ng
1.9 (0.5)
10
M2 100 ng
2.9 (0.9)
4
M2 200 ng
3.0 (0.9)
4
aDrosophila SL2 cells were transiently transfected with a DHFR-CAT reporter construct alone or in conjunction with the indicated amounts of expression constructs for Sp1 (pPacSp1), Sp3 (pPacSp3), M1 (pPacM1) or M2 (pPacM2). Resulting levels of CAT activity were quantified as previously described (12).bShown are mean values for percentage acetylation (+-SE) of chloramphenicol per A600 of 1 [mu]l total cell extract normalized to the amount of activity recorded for the DHFR-CAT reporter construct.cn is the number of independent plates of transfected cells analyzed.
DISCUSSION
Rb has previously been shown to govern transcription of a subset of growth control genes, such as c-fos, c-myc and TGF-[beta]1, via discrete promoter elements termed RCEs (6 ,7 ). To determine the mechanism(s) by which Rb regulates RCE function, we have sought to identify and characterize nuclear RCE binding proteins in vitro, determine if such proteins are required for RCE activity in vivo and assess the functional consequence of their interaction with Rb. We have previously determined that RCEs are bound in vitro by Sp1 and Sp3, two members of the Sp family of transcription factors, and that RCE-mediated transcription is stimulated by these factors in vivo (12 ,34 ). Moreover, co-expression of Rb and Sp1 or Sp3 leads to marked `superactivation' of Sp1/Sp3-mediated transcription (12 ,34 ). In this report we demonstrate that: (i) Sp3 encodes at least three distinct transcription factors; (ii) each of these Sp3-related proteins binds RCEs in vitro; (ii) two Sp3-related proteins arise via internal translational initiation within Sp3 mRNA; (iv) internally initiated Sp3 isoforms are potent inhibitors of Sp-mediated transcription. Taken together with recent evidence that Sp1 is an alternatively spliced gene (45 ), our results indicate that transcriptional control of RCEs and, by extension, other promoter elements regulated by Sp family members may be significantly more complex than previously suspected.
. Inhibition of Sp-mediated, but not VP16-mediated transcription by an internally initiated isoform of Sp3a
Reporter construct
Trans-activator
Additional DNA
Relative CAT activity (+-SE)b
nc
DHFR
Sp1 50 ng
-
1.0 (0)
14
M2 10 ng
1.5 (0.3)
8
M2 50 ng
1.7 (0.4)
10
M2 100 ng
1.2 (0.2)
14
M2 250 ng
0.6 (0.1)
2
M2 500 ng
0.3 (0.1)
4
M2 1000 ng
0.1 (0)
2
Sp3 50 ng
-
1.0 (0)
18
M2 10 ng
0.8 (0.1)
12
M2 50 ng
0.3 (0.4)
14
M2 100 ng
0.1 (0)
18
M2 250 ng
0 (0)
2
M2 500 ng
0 (0)
4
FOS
Sp3 50 ng
-
1.0 (0)
3
M2 100 ng
0.2 (0)
3
MYC
Sp3 50 ng
-
1.0 (0)
3
M2 100 ng
0.2 (0)
3
TGF-[beta]1
Sp3 50 ng
-
1.0 (0)
3
M2 100 ng
0.2 (0)
3
GAL4
VP16 10 [mu]g
-
1 (0)
4
M2 50 ng
1.9 (0.1)
2
M2 100 ng
1.2 (0.1)
4
M2 500 ng
1.3 (0)
2
aDrosophila SL2 cells were transiently transfected with DHFR-CAT, FOS-CAT, MYC-CAT, TGF-[beta]1-CAT or GAL4-CAT reporter constructs in conjunction with the indicated amounts of expression constructs for Sp1 (pPacSp1), Sp3 (pPacSp3), M2 (pPacM2) or GAL4-VP16. Resulting levels of CAT activity were quantified as previously described (12).
bShown are mean values for percentage acetylation (+-SE) of chloramphenicol per A600 of 1 [mu]l total cell extract normalized to the amount of activity recorded for DHFR or GAL4 reporter constructs in the absence of M2.
cn is the number of independent plates of transfected cells analyzed.
We and others have previously reported that Sp3 stimulates transcription in vivo, however, this observation has not proven to be universal (21 ,34 ,42 ,46 -51 ). Using a cDNA isolated by Hagen et al. (35 ), a number of laboratories have reported that Sp3 has little or no capacity to function as a transcriptional trans-activator. Instead, based on a wide variety of transient co-transfection experiments, Sp3 has been reported to function as a repressor of Sp1-mediated transcription. Since it is now apparent that Sp3 encodes at least three distinct proteins in vivo and that internally initiated Sp3 isoforms function as transcriptional inhibitors, it is likely that these once disparate observations may be easily reconciled, i.e. experiments that employ Sp3 expression vectors that preferentially lead to biogenesis of full-length Sp3 protein will score Sp3 as a transcriptional trans-activator while expression vectors that largely yield internally initiated Sp3 isoforms will score Sp3 as a transcriptional repressor. This conclusion appears to be entirely consistent with the sizes of Sp3 proteins expressed following transient transfections with Sp3 cDNAs (34 ,50 ).
Since all Sp3 cDNAs cloned to date are partial cDNAs lacking 5'-untranslated sequences and the extreme N-terminus of Sp3 protein, we can only speculate on the mechanism governing translational initiation within Sp3 mRNA. One possibility is that the translational start site at the 5'-end of the Sp3 message is partially obscured by RNA secondary structure or embedded within a sequence that is weak relative to internal sites of translational initiation. Presumably this arrangement ensures that appropriate levels of stimulatory and inhibitory Sp3 proteins are synthesized in vivo. That a single mRNA can encode multiple functionally distinct transcription factors may be a more common theme for transcriptional regulation that previously realized. Like Sp3, the C/EBP[alpha] and C/EBP[beta] CCAAT/enhancer binding proteins have previously been noted to encode full-length proteins that function as trans-activators as well as internally initiated isoforms that act as inhibitors of C/EBP function (52 -55 ). As for Sp3, internal translational initiation within C/EBP[alpha] and C/EBP[beta] mRNA results in elimination of a significant portion of their respective trans-activation domains. C/EBP[beta] encodes liver-enriched activating (LAP) and inhibiting (LIP) proteins and the LIP/LAP ratio, which has been shown to be developmentally and hormonally regulated, influences transcription of a variety of cellular genes. Although the precise mechanism by which LIP inhibits LAP function remains to be determined, it is likely that LIP dimerizes with LAP via their respective basic region/leucine zipper domains forming transcriptionally inactive complexes that compete with transcriptionally active LAP dimers for occupancy of DNA binding sites. It is worth emphasizing that our results provide little insight into the mechanism(s) by which internally initiated Sp3 isoforms abrogate Sp-mediated transcription. In protein-DNA binding assays we have previously shown that: (i) Sp1 and full-length and internally initiated Sp3 isoforms are equally abundant in extracts prepared from >20 mammalian cell types grown under a variety of growth conditions; (ii) Sp1 and each Sp3 isoform require identical RCE nucleotides for formation of protein-DNA complexes; (iii) on- and off-rate measurements indicate that each protein-RCE complex is equally stable in vitro (A.J.Udvadia and J.M.Horowitz, unpublished observations). These results would suggest that competition for binding site occupancy between internally initiated Sp3 isoforms and Sp3 or Sp1 could account for diminished Sp-mediated transcription. Yet, if binding site competition were the sole mechanism by which Sp-mediated transcription is inhibited, equivalent levels of inhibition of Sp1 and Sp3 would be expected. Instead, Sp3-mediated transcription is reproducibly at least 10-fold more sensitive to inhibition than is transcription directed by Sp1. Given this intriguing observation, we speculate that disparate degrees of protein-protein interactions may account for the differential sensitivities of Sp1 and Sp3 to inhibition by internally initiated isoforms of Sp3. As one possibility in this regard, we note that the D domain of Sp1 has previously been shown to play an important role in its multimerization and this region is strikingly dissimilar in Sp3. Thus, amino acid differences within the D domains of Sp1 and Sp3 may facilitate higher affinity interactions between full-length Sp3 and its internally initiated isoforms. Yet another possibility springs from the recent identification of a `repression domain' immediately upstream of the Sp3 DNA binding domain (56 ). It is conceivable that interactions between Sp3 molecules via this domain may be favored over those involving Sp1 and Sp3 proteins. Once complexed with Sp1 or Sp3, internally initiated Sp3 isoforms may diminish the capacity of multimeric complexes to bind DNA or prevent formation of transcriptionally active protein-DNA complexes. Consistent with the notion that multimerization of Sp proteins is required for transcriptional repression, Sp3-mediated repression has been reported to be dependent on the number of Sp3 binding sites within cellular promoters (51 ).
Although the mechanism by which internally initiated Sp3 isoforms inhibit Sp1/Sp3-mediated transcription remains obscure, we speculate that this mode of regulation likely plays an important role in cell cycle- and signal-induced transcription. Indeed, cell cycle- or signal-induced expression of a variety of genes, including p21waf1/cip1, p15INK4B, CYP11A, mdr1 and acetyl-CoA carboxylase, has been mapped to GC-rich promoter elements that bind Sp family members in vitro (18 ,19 ,23 ,57 -60 ). We predict that cell cycle- and signal-induced alterations of the protein and/or DNA binding activities of internally initiated Sp3 isoforms may account in part for regulation of such differentially expressed genes.
ACKNOWLEDGEMENTS
The authors are indebted to members of the Horowitz laboratory, Michael Datto and Xiao-Fan Wang for helpful discussions and to Astar Winoto for Sp3 antiserum and plasmid DNA used in this study. This work was supported in part by National Cancer Institute, National Institutes of Health grant CA53248 and Faculty Research Award A-73970 from the American Cancer Society to J.M.H. J.M.H. is a member of the Pew Scholars Program in the Biomedical Sciences.
