The Rod cGMP-phosphodiesterase (cid:1) -Subunit Promoter Is a Specific Target for Sp4 and Is Not Activated by Other Sp Proteins or CRX*

The (cid:1) -subunit of cGMP-phosphodiesterase ( (cid:1) -PDE) is a key protein in phototransduction expressed exclu-sively in rod photoreceptors. It is necessary for visual function and for structural integrity of the retina. (cid:1) -PDE promoter deletions showed that the (cid:2) 45/ (cid:2) 23 region containing a consensus Crx-response element (CRE) was necessary for low level transcriptional activity. Overexpressed Crx modestly transactivated this promoter in 293 human embryonic kidney cells; how-ever, mutation of CRE had no significant effect on transcription either in transfected Y79 retinoblastoma cells or Xenopus embryonic heads. Thus, Crx is unlikely to be a critical (cid:1) -PDE transcriptional regulator in vivo . Inter-estingly, although the (cid:1) /GC element ( (cid:2) 59/ (cid:2) 49) binds multiple Sp transcription factors in vitro , only Sp4, but not Sp1 or Sp3, significantly enhanced (cid:1) -PDE promoter activity. Thus, the Sp4-mediated differential activation

The ␤-subunit of cGMP-phosphodiesterase (␤-PDE) is a key protein in phototransduction expressed exclusively in rod photoreceptors. It is necessary for visual function and for structural integrity of the retina. ␤-PDE promoter deletions showed that the ؊45/؊23 region containing a consensus Crx-response element (CRE) was necessary for low level transcriptional activity. Overexpressed Crx modestly transactivated this promoter in 293 human embryonic kidney cells; however, mutation of CRE had no significant effect on transcription either in transfected Y79 retinoblastoma cells or Xenopus embryonic heads. Thus, Crx is unlikely to be a critical ␤-PDE transcriptional regulator in vivo. Interestingly, although the ␤/GC element (؊59/؊49) binds multiple Sp transcription factors in vitro, only Sp4, but not Sp1 or Sp3, significantly enhanced ␤-PDE promoter activity. Thus, the Sp4-mediated differential activation of the ␤-PDE transcription defines the first specific Sp4 target gene reported to date and implies the importance of Sp4 for retinal function. Further extensive mutagenesis of the ␤-PDE upstream sequences showed no additional regulatory elements. Although this promoter lacks a canonical TATA box or Inr element, it has the (T/A)-rich ␤/TA sequence located within the ؊45/؊23 region. We found that it binds purified TBP and TFIIB in gel mobility shift assays with cooperative enhancement of binding affinity.
One of the key components of the phototransduction cascade that takes place in rod photoreceptors is the heterotetrameric (␣␤␥ 2 ) cGMP-phosphodiesterase (1). The gene encoding the ␤-subunit of the human enzyme (␤-PDE) 1 has been well characterized and consists of 22 exons encompassing ϳ43 kb of genomic DNA (2). Genetic defects in this gene have been linked to retinal degeneration in several animal species and human (3)(4)(5)(6)(7)(8)(9). There is increasing evidence that abnormalities in transcriptional regulatory components of different genes contribute significantly to or directly cause pathological phenotypes in the retina (10 -13). Therefore, further studies on the transcriptional regulation of rod-specific ␤-PDE gene will identify additional genes important for retinal function and structural integrity and will ultimately help to establish the molecular mechanisms crucial for retina-specific expression of this and perhaps some other genes.
We recently reported our initial results on the transcriptional control mechanisms that take place in the human ␤-PDE 5Ј-flanking region (14). Mutational analysis of the ␤-PDE promoter tested both in vitro and ex vivo, and confirmed by the generation of transgenic Xenopus expressing mutant ␤-PDE promoter/green fluorescent protein fusion constructs in vivo, revealed a minimal promoter region, from Ϫ93 to ϩ53, that supports high levels of rod-specific transcription (14). Two enhancer elements were localized within this minimal promoter, ␤Ap1/NRE and ␤/GC, that interact with nuclear factors and activate transcription from the ␤-PDE promoter.
To continue the systematic analysis of the ␤-PDE promoter structure, we have now carried out extensive mutagenesis of the proximal promoter and the 5Ј-untranslated region. The presence of a consensus CRE sequence in the minimal rodspecific ␤-PDE promoter prompted us to test whether Crx (cone, rod homeobox), a member of the Otx family of homeodomain-containing transcription factors, is involved in transcriptional regulation of the ␤-PDE gene. Previously, Crx had been shown to be important for the transcriptional control of several retina-specific genes, including rhodopsin (15,16). We report here that although Crx is capable of modest transactivation of the ␤-PDE promoter when overexpressed in 293 embryonic kidney cells, transfections in Y79 retinoblastoma cells and Xenopus embryonic heads showed that it is unlikely to be a major player in transcriptional regulation of the ␤-PDE gene. We also show that both purified TBP and TFIIB were able to bind to the ␤-PDE proximal promoter (Ϫ45/Ϫ16) with cooperative enhancement of binding. The interactions between the ␤-PDE promoter and the basal transcription factors were not completely disrupted by limited nucleotide substitutions in this region, which may be related to the complex, low affinity, basal transcription factor-promoter interactions over extended core promoter sequences described on other promoters (17,18).
The functionally important ␤/GC element is homologous to the consensus GC box that binds members of the Sp family of transcription factors including Sp1, Sp3, and Sp4. These nuclear factors share similar structural features and have highly conserved DNA binding domains that allow them to bind with identical affinity to the consensus GC box (19). We have previously shown that Sp1 and Sp4 can interact with the ␤-PDE promoter (14). Our intriguing finding that the predominantly central nervous system-expressed Sp4 is also abundantly present in the adult retina prompted us to further test its activation properties on the rod-specific ␤-PDE promoter under defined conditions in direct comparison to Sp1 and Sp3. We report here that only Sp4, but not Sp1 or Sp3, is a strong activator of transcription from the ␤-PDE promoter. Thus, the rod-specific ␤-PDE gene is the first specific gene target for the Sp4 transcription factor described to date.

EXPERIMENTAL PROCEDURES
Cell Cultures and Transient Transfections-Y79 human retinoblastoma cells and 293 human embryonic kidney cells were obtained from the American Type Culture Collection (Manassas, VA). Y79 retinoblastoma cells were cultured in RPMI 1640 (Invitrogen) supplemented with 15% (v/v) fetal calf serum (Invitrogen) as described previously (20). 293 kidney cells were propagated in Dulbecco's modified Eagle's medium/ Ham's F-12 (Invitrogen) supplemented with 10% fetal calf serum. These cells have been used previously for testing multiple retina-specific promoters and transcription factors including Crx (14,15).
Calcium phosphate-mediated transient transfections as well as luciferase and ␤-galactosidase assays were performed as described previously (20), except for the addition of a glycerol shock that improved the transfection efficiency of 293 kidney cells. For normalization of transfection efficiency, all transfection reactions included 5 g of the pSV-␤galactosidase expression plasmid as an internal control. In cotransfection experiments, the ratios of the reporter vector to expression plasmid were determined empirically. For each experiment, the total amount of transfected DNA per plate was kept constant (15 g per 60-mm plate) by addition of carrier plasmid DNA. Triplicate plates were used for all transfections, and experiments were repeated several times.
Plasmids-5Ј-End deletions in human ␤-PDE promoter were generated by PCR using sequence-specific primers as described previously (20). Deletion constructs were designated based on the inserted ␤-PDE sequences relative to the major transcription start site of the gene (e.g. Ϫ93 to ϩ53). A luciferase reporter construct containing the Ϫ132 to ϩ138 sequence of the human ␣Ј-PDE was also engineered in the pGL2-Basic vector.
Substitution mutations by nucleotide transversions were generated in the context of the Ϫ72 to ϩ53 ␤-PDE sequence in the pGL2-Basic vector as described previously (20). The construct designation refers to the positions of the mutated nucleotides in the ␤-PDE promoter relative to the transcription start site. For example, the Ϫ51/Ϫ49m construct contains mutations in nucleotides Ϫ51 through Ϫ49. All inserted sequences were fully sequenced in both directions to confirm identity and desired alterations.
The pcDNA-Crx expression plasmid containing a full-length bovine Crx cDNA and the pcDNA3.1 vector (Invitrogen) was obtained from Dr. Donald Zack (15). The pMT-Nrl expression plasmid encoding the fulllength human Nrl, pMT-DD10 encoding the truncated Nrl mutant that has its N-terminal acidic domain deleted, and the empty pMT3 vector were kindly supplied by Dr. Anand Swaroop (21). Dr. Guntram Suske kindly provided us with pRC/CMV-Sp1, pRC/CMV-Sp3, and pRC/CMV-Sp4 encoding the full-length Sp1, Sp3, and Sp4, respectively (22). In cotransfection experiments, empty expression vectors were used to keep the amount of DNA constant for each transfected plate.
Gel Mobility Shift Assays-Gel mobility shift assays (GMSA) were performed essentially as described elsewhere (18). Briefly, 5% polyacrylamide gel (1.7% cross-linking) contained 0.5ϫ TBE, 5% glycerol, 2 mM MgCl 2 , and 1 mM dithiothreitol. 10-l binding reactions included 50 ng of poly(dG-dC), 0.1 ng of the probe (0.5-1 ϫ 10 5 cpm), 20 -25 nM TBP, and/or 20 -25 nM TFIIB. The reactions were incubated at 30°C for 30 min and promptly loaded onto the gel. Electrophoresis was carried out for 20 -30 min at 400 V due to the rapid dissociation rate of TBP-DNA complexes. The temperature of the electrophoresis running buffer (0.5ϫ TBE, 2 mM MgCl 2 , prechilled and submerged in ice) in both compartments of the gel apparatus (Mini-Protein II tank, Bio-Rad) and the temperature of the gel plates were measured at the beginning and the end of each experiment and was maintained at less than 21°C with continuous cooling. Gel images were captured and quantified using a PhosphorImager (Molecular Dynamics).
Purified TBP and TFIIB were a kind gift of Dr. Branden S. Wolner. The ␤-PDE oligonucleotide probe was prepared by annealing singlestranded oligonucleotides radiolabeled at the 5Ј-end. The labeled probe was purified with MicroSpin G-25 columns (Amersham Biosciences). The adenovirus major late (AdML) promoter probe has been described previously (18).
Preparation of Xenopus Embryos and Transient Transfections ex Vivo-Preparation of Xenopus embryo heads and the transient transfection procedure using DOTAP (Roche Molecular Biochemicals) were described previously (23). Briefly, embryos at stages 24 -28 were selected and prepared by severing embryos at the ear placode. Transfection of each of the tested DNA constructs was carried out in groups of 10 -12 dissected embryo heads. Three to six independent groups of dissected embryo heads were used for testing each construct. After 20 h of ex vivo organ culturing, the DNA-containing medium was replaced by a fresh culture medium. Incubation was continued until 78 h postfertilization when the embryo heads were harvested and homogenized. Protein extracts from the embryonic heads were prepared, and duplicate aliquots were used to measure luciferase activity. Comparison of sample means was performed in a large number of experimental trials as described previously (23). A construct containing the Ϫ508 to ϩ41 region of the Xenopus opsin promoter was used as positive control (24) and the empty pGL2-Basic vector (Promega) as negative control. In addition, to control for specificity of retinal cell type expression, all ␤-PDE promoter constructs were transfected in dissected Xenopus embryo trunks that contain many cell types (14,23) and showed no activity (data not shown).

RESULTS
The Ϫ45 to Ϫ23 ␤-PDE Region Contains Regulatory Sequences-Significant reduction of transcriptional activity occurs with the 5Ј-end deletion of ␤-PDE promoter sequences from Ϫ72 to Ϫ45 (14). However, the Ϫ45 to ϩ53 construct showed residual promoter activity about 8 -10-fold that of the promoter-less control (Fig. 1A), which suggests the presence of additional regulatory sequences. Thus, as an initial step toward identifying the potential control elements located in this region, further deletions of the proximal promoter were performed. The Ϫ45 to ϩ53, Ϫ23 to ϩ53, and ϩ4 to ϩ53 ␤-PDE promoter/luciferase fusion constructs were transiently transfected first in cultured Y79 human retinoblastoma cells and then ex vivo in dissected Xenopus embryo heads. These human retina-derived cell culture and amphibian in situ transfection systems have been employed previously for studying the regulation of photoreceptor-specific gene expression, including that of ␤-PDE (14). Luciferase activities were measured and normalized to the ␤-galactosidase activities obtained with a control plasmid in Y79 cells or expressed per embryo and averaged statistically as described previously for Xenopus transfections (23). Further reduction in promoter activity was observed in both transfection systems when the Ϫ45 to Ϫ23 region was deleted (Fig. 1). The activity level of the Ϫ23 to ϩ53 promoter was not significantly different from that observed with the promoter-less vector when tested in Y79 cells or Xenopus embryos. The ϩ4 to ϩ53 promoter construct carrying further 5Ј-end deletion past the major transcription start site (25) showed that luciferase activity remained low. High evolutionary conservation of the Ϫ45 to Ϫ23 region (25) that composes the consensus CRE motif (Ϫ41/Ϫ36) and the T/A-rich ␤/TA sequence located at a consensus position for the TATA box is evident between mouse and human also suggesting its functional importance.
Overexpressed Crx Enhances ␤-PDE Transcription and Has Additive Effect with Nrl-The CRE motif ((C/T)TAATC) interacts with Crx and plays an important regulatory role in the transcription of rhodopsin and several other retina-specific genes (15,26). Recently, mutations in Crx have been linked to various forms of human retinal degeneration (10,11). To investigate whether Crx is able to directly transactivate the ␤-PDE promoter, we transiently overexpressed Crx in 293 human embryonic kidney cells in cotransfections with different deletion mutants of the ␤-PDE promoter. These cells do not endogenously produce rod-specific phosphodiesterases includ-ing ␤-PDE and have been used previously for transient transfections to study transcriptional regulation of the ␤-PDE gene (14,15,20,27). Fig. 1B shows that although the uninduced activity of the Ϫ45 to ϩ53 promoter was significantly lower than that of the Ϫ72 to ϩ53 construct, overexpressed Crx caused similar (ϳ4-fold) transactivation of both promoters. In contrast, the Ϫ23 to ϩ53 construct produced luciferase activity comparable with that of the promoter-less control and failed to show any transactivation potential when coexpressed with Crx.
For comparison, we also tested the promoter region of the cone photoreceptor-specific ␣Ј-PDE gene that contains two sequences homologous to consensus CRE, one at position Ϫ95/ Ϫ89 (TTAATCC) and the other at Ϫ118/Ϫ112 (GATTTAG). Cotransfections of the Ϫ132 to ϩ138 ␣Ј-PDE/luciferase reporter construct with the Crx expression plasmid resulted in ϳ8-fold increase in promoter activity compared with the uninduced promoter (Fig. 1B).
Since the Ϫ45/Ϫ23 region was found to be important for the ␤-PDE promoter transactivation by overexpressed Crx, we tested whether the consensus CRE located within this region (Ϫ41/Ϫ36) was responsible for this transactivation. We also tested whether Crx-mediated transactivation of the ␤-PDE promoter had functional synergy with Nrl that had been previously shown to bind and transactivate this promoter (14). The activity of the wild-type Ϫ72 to ϩ53 ␤-PDE promoter was compared with the CRE-mutant construct Ϫ41/Ϫ38m. Approximately 9 -10-fold increase in luciferase activity by coexpressed Crx and Nrl was observed compared with the ϳ4-fold increase caused by Crx alone and a 3-fold increase produced by Nrl alone (Fig. 1C). These results are consistent with an additive or a modest synergistic effect, which differs from the rhodopsin promoter that shows significant synergistic transactivation by the combination of Crx and Nrl (15). The Ϫ41/Ϫ38 mutation disrupting the consensus CRE caused substantial reduction in overexpressed Crx-or Crx/Nrl-mediated transactivation of the ␤-PDE promoter (Fig. 1C). In contrast, the Nrl-mediated transactivation of CRE-mutant promoter remained unaffected by this mutation providing a positive control for sequence specificity of the Crx-CRE-mediated transactivation.
To determine whether the transactivation of the ␤-PDE promoter/luciferase constructs by Crx and Nrl was specific for the ␤-PDE promoter, an SV40 promoter/luciferase construct (pGL2-Promoter, Promega) was tested as control and showed no significant induction of luciferase activity (1.3-fold increase) by the mixture of Nrl and Crx (data not shown). This is consistent with the previous report of the inability of Crx to transactivate a non-retinal collagenase promoter/luciferase construct (15) and suggests that there is promoter selectivity in the Nrl/Crx transactivation mechanism.
Functional Analysis of the Consensus CRE, the Proximal Promoter, and the 5Ј-Untranslated Region in Retina-related Transfection Systems-Based on the results described above, we further investigated whether CRE and its flanking sequences were functionally relevant to the transcriptional regulation of the ␤-PDE promoter in vivo in the context of a retina-relevant environment. A series of ␤-PDE promoter mutants carrying substitutions in the Ϫ45 to Ϫ23 region was transfected in Y79 retinoblastoma cells and then in Xenopus embryos maintained ex vivo (summarized in Fig. 2, top). Contrary to our initial expectations, the Ϫ41/Ϫ38m mutation that completely disrupted the consensus CRE had little effect on the ␤-PDE promoter activity in both transfection systems ( Fig. 2A). In addition, no significant changes were seen with mutations in Ϫ37/Ϫ36, Ϫ35/Ϫ34, and Ϫ33/32. The Ϫ30/Ϫ27m mutant con- Although neither the mutations in consensus CRE or ␤/TA affected promoter activity, a cooperative interaction of transcription factors at both sites located in close proximity of each other could not be ruled out. Therefore, a double mutant was constructed that contained both Ϫ30/Ϫ27m and Ϫ41/Ϫ38m. Transient transfections of Y79 cells using the double mutant showed no significant alterations in promoter activity compared with the wild-type ␤-PDE promoter (Fig. 2B). To search for other regulatory sequences in this TATA-and Inr-less gene, additional ␤-PDE promoter mutants (n ϭ 14) containing nucle-otide substitutions spanning the proximal 5Ј-flanking and the 5Ј-untranslated regions (Ϫ23 to ϩ53; Fig. 2, top) were tested in transient transfections of Y79 retinoblastoma cells. Promoter activity determined in these mutants ranged between ϳ0.5and 1.5-fold that of the wild-type control (Fig. 2B). A 3Ј-end deletion mutant (Ϫ72 to ϩ4) lacking most of the 5Ј-UTR showed ϳ3-fold reduction of promoter activity.
Taken together, these results suggest that the ␤-PDE promoter does not have well defined core elements responsible for basal transcription in Y-79 cells or Xenopus embryo heads.
Rather, it appears that the transcription factors responsible for maintaining low level expression from this promoter do not require a rigid sequence for interactions, but can accommodate a range of nucleotides.
TBP and TFIIB Bind the ␤-PDE Promoter-Although the ␤/TA sequence located at Ϫ31/Ϫ25 of the ␤-PDE promoter is significantly different from the consensus TATA box, it has a high T/A content and is located in the proximity of the transcription start site with no other consensus core promoter elements present in this gene. Thus, we tested whether the ␤/TA sequence was able to bind purified TBP separately or in complex with TFIIB in GMSAs. As a control, we compared the binding of TBP, TFIIB, and the TFIIB-TBP combination to the AdML promoter. Shifted bands were observed with the addition of either TBP alone or TFIIB alone to the ␤/TA probe ( Sp4 Activates Rod-specific ␤-PDE Promoter 3A). Addition of the combination of TBP and TFIIB resulted in a slower migrating complex with about 3-fold increase in band intensity compared with TBP alone, producing a characteristic supershifted pattern described previously for the AdML promoter (18). These results suggest an enhanced cooperative binding by the TFIIB-TBP complex to the ␤-PDE promoter compared with TBP alone. In contrast, when comparable protein concentrations were used, the AdML promoter interacted with TBP and TFIIB-TBP, but did not form a stable TFIIB-DNA complex in GMSA as demonstrated previously (Ref. 18 and data not shown). Although the addition of a 200-fold molar excess of the wild-type Ϫ45/Ϫ16 competitor to the binding reaction prevented the shifted complex formation, the mutant Ϫ30/Ϫ27m and Ϫ35/Ϫ27m competitors also showed some competition with the wild-type sequence for TFIIB-TBP binding. These results further corroborate our functional transfection data that a well defined core promoter sequence could not be found in the ␤-PDE 5Ј-flanking region.
Sp4 but Not Sp1 or Sp3 Specifically Transactivates the ␤-PDE Promoter-Mutational analyses and in vitro protein-DNA binding studies have defined the ␤/GC element (Ϫ55/Ϫ46) as an important enhancer of the ␤-PDE promoter that binds transcription factors of the Sp family (14). We compared different members of the Sp protein family (Sp1, Sp3, and Sp4) that share similar DNA-binding characteristics (19) for their effects on transcription from the ␤-PDE promoter. Increasing amounts of expression plasmids (0.08, 0.4, and 2 g) each carrying a full-length cDNA for either Sp1, Sp3, or Sp4 were transiently coexpressed with the wild-type minimal rod-specific ␤-PDE promoter (Ϫ93 to ϩ53, 2 g). Compared with other members of the Sp family, Sp4 was the only transcription factor that showed significant dose-dependent effect on the ␤-PDE promoter (Fig. 4A). Promoter specificity of the Sp4-mediated transactivation was confirmed by comparing its effect on transcription from the ␤-PDE promoter (ϳ21-fold enhancement) to that on the SV40 promoter (no significant change) relative to the uninduced transcription, respectively, when tested in the 293 cell transfection system (Fig. 4B). The maximum activation was observed using 2 g of pRC/CMV-Sp4 and was comparable with that seen using 5 g of pRC/CMV-Sp4 indicating a saturation effect. In contrast, neither Sp1 nor Sp3 showed any significant effect on transcription from either ␤-PDE or SV40 promoter compared with the dramatic difference between Sp4mediated transactivation of the two promoters. DISCUSSION Sequence analysis of the 5Ј-flanking region of the ␤-PDE gene showed that it has several sites homologous to known response elements (14,28). These include an E box, an AP1/ NRE-like sequence (␤Ap1/NRE), a GC box-like site (␤/GC), and a sequence identical to the consensus CRE. We recently demonstrated that ␤Ap1/NRE and ␤/GC were cis-acting elements FIG. 4. Sp4, but not Sp1 or Sp3, selectively activates the ␤-PDE promoter. The fold induction of the ␤-PDE or SV40 reporter constructs was determined relative to the uninduced reporter activity. A, the Ϫ93 to ϩ53 minimal rod-specific ␤-PDE promoter (2 g) was cotransfected with increasing amounts of Sp1-, Sp3-, and Sp4-containing plasmids and compared with the uninduced promoter cotransfected with an empty plasmid. B, the SV40 promoter/luciferase vector (2 g, pGL2-Control, Promega) was cotransfected with 2 g of the plasmid containing either Sp1, Sp3, or Sp4 cDNA and compared with an empty plasmid. Luciferase activity was measured in cell lysates and normalized to the corresponding ␤-galactosidase activity for each sample. The results are expressed as the fold induction of the mean activity of the uninduced Ϫ93 to ϩ53 ␤-PDE reporter construct Ϯ S.D. Each transfection was done in triplicate and repeated at least twice.
functionally important for ␤-PDE transcriptional regulation and that the E box seemed to have no significant role in ␤-PDE transcription (14,28). To gain further insight into transcriptional regulation of this rod photoreceptor-specific gene, we performed a detailed mutagenic screen of the ␤-PDE promoter and the 5Ј-UTR. The functional relevance of the consensus CRE motif (Ϫ41/Ϫ36) was also investigated. In addition, we tested the effects of different members of the Sp family of transcription factors on activity of the ␤-PDE promoter that contains the functionally important ␤/GC element. Because the ␤-PDE gene does not have a TATA box, Inr sequence, or other known core promoter elements, we extended our studies to include the basal promoter region.
Crx has been shown to interact with the upstream region of the ␤-PDE gene in DNase I footprinting assays (15). We observed a modest increase in ␤-PDE promoter activity by cotransfected Crx that was significantly reduced after substituting CRE nucleotides Ϫ41/Ϫ38 and virtually eliminated with the Ϫ23 to ϩ53 construct that completely lacks CRE and the surrounding sequences. Although the luciferase gene has been reported to show promoter-independent enhancement of its expression under certain conditions, e.g. induced by IFN-␥ (29), this was not the case for Crx-and Crx/Nrl-activated ␤-PDE promoter/luciferase transcription in 293 kidney cells. The SV40 promoter/luciferase control was not affected by overexpressed Crx either alone or with cotransfected Nrl excluding the possibility of promoter-independent activation of luciferase or a general, nonspecific effect on transcription in 293 cells. Thus, our data show that Crx can interact and moderately transactivate the ␤-PDE promoter when overexpressed at high concentrations in 293 kidney cells and that the Ϫ41/Ϫ36 sequence functions as CRE.
However, in transfections of both Y79 retinoblastoma cells and Xenopus embryos, ␤-PDE mutants with disrupted CRE (Ϫ41/Ϫ38m and Ϫ37/Ϫ36m) showed promoter activity comparable with that of the wild-type promoter. The lack of effect of an in vivo concentration of Crx on the ␤-PDE transcription in a retinal system differs from the transcriptional activation of the ␤-PDE promoter by cotransfected Crx in 293 kidney cells. The latter could be caused by Crx overexpression and, possibly, by the context of a non-retinal cell line that may contain additional cofactors interacting with Crx. Therefore, although Crx is involved in the regulation of several other photoreceptorspecific genes, our results suggest that the CRE-like sequence located in the ␤-PDE proximal promoter is unlikely to be a functional element important for the transcriptional activation of this gene in retinal cells. Nevertheless, under certain conditions, Crx may be able to modulate the effect of other transcription factors.
The possibility of an additional regulatory sequence(s) in the ␤-PDE basal promoter region or the 5Ј-UTR is suggested by the tight regulation of the transcriptional initiation site selection in this gene. There are only one major and one minor transcription start sites in both human and murine ␤-PDE genes (25). This indicates the assembly of the basal transcription machinery at a specific core promoter element rather than random binding to a variety of sequences. However, there are no consensus core promoter elements in the ␤-PDE gene. The T/A-rich ␤/TA sequence located 25 bp upstream from the major transcription start site of the ␤-PDE gene (Ϫ31/Ϫ25) seemed to be a likely site for interactions with basal transcription factors, although this sequence (TAAGAAA) is not predicted from crystallographic studies to form a stable TBP-DNA complex (30). However, our protein binding studies suggest that the ␤-PDE promoter forms stable interactions in vitro in GMSA with purified TBP as well as with the TFIIB-TBP complex, with a cooperative enhancement of binding. Interestingly, a stable ␤-PDE promoter-TFIIB complex was also observed in the absence of TBP with a relatively modest concentration of TFIIB. Because the ␤-PDE promoter lacks the consensus BRE, (G/ C)(G/C)(G/A)CGCC (30), proposed to be a binding site for TFIIB, it is likely that TFIIB interacts with an alternative DNA sequence in the ␤-PDE promoter. In addition, deletion of the Ϫ45/Ϫ23 region reduced the promoter activity virtually to the level of the promoter-less vector suggesting the presence of nucleotides critical for the ␤-PDE transcription. However, we observed no effect of a 4-nucleotide substitution in ␤/TA (Ϫ30/ Ϫ27m) on promoter activity in either Y79 retinoblastoma cells or Xenopus embryos. Both Ϫ30/Ϫ27m and Ϫ35/Ϫ26m showed significant competition with the wild-type sequence for TFIIB-TBP binding in GMSA. These results suggest the lack of a well defined core element in this promoter necessary for basal transcription machinery binding and low level basal transcription.
The most significant finding of the present investigation was the demonstration of the functional involvement of members of the Sp family in transcriptional regulation of the ␤-PDE promoter. Interestingly, Sp1, Sp3, and Sp4 transcription factors showed differential effects on the ␤-PDE promoter activity. Sp4-mediated transactivation was significantly higher than that produced by Sp1 or Sp3, which suggests the importance of this transcription factor for the ␤-PDE gene transcription. Whereas Sp4 is predominantly restricted to the central nervous system and retina in vivo, it is expressed in many cell lines (19) including the 293 kidney cells. 2 Sp1 and Sp3 are ubiquitous in mammalian cells. Therefore, all three proteins are expressed in 293 cells. Nevertheless, the difference between Sp4-mediated transactivation of the ␤-PDE and SV40 promoters was dramatic compared with Sp1 and Sp3.
Members of the Sp family bind GC-rich DNA sequences through three zinc finger motifs. The residues involved in the determination of the target site specificity and binding affinity are highly conserved between Sp1, Sp3, and Sp4. In fact, all of these transcription factors bind GC and GT boxes with equal affinity in vitro (19). Thus, a relative abundance of any one of the Sp proteins would lead to its increased competition with the others for the ␤-PDE promoter binding. However, DNA binding by single proteins may not be the key molecular basis to explain the Sp4, Sp1, and Sp3 functional differences in vivo. The ␤-PDE promoter contains other regulatory sequences. Thus, the differential Sp4-mediated stimulation of this gene transcription is likely to be dependent on its promoter context. Sp4 has been the least characterized member of the Sp family partly because of its restricted pattern of expression in vivo. Here we demonstrate the first natural target gene for Sp4 that also seems to lack transcriptional regulation by Sp1 and Sp3. The fact that Sp4 was the only transcription factor to transactivate significantly the rod-specific ␤-PDE promoter supports our previous finding of this highly restricted protein, compared with Sp1 or Sp3, being abundantly expressed in retina (14). The lack of other known Sp4 targets combined with our finding of its regulation of a very specific rod-restricted ␤-PDE gene implies that this transcription factor functions in a relatively narrow promoter-specific manner.
In addition, Sp4 could have a more universal role in cell type-specific expression of certain genes in rods and possibly other retinal cell populations by interacting with different arrays of transcription factors. We have shown previously that another nuclear factor, Nrl, regulates transcription from the ␤-PDE promoter (14). Considering the additional ␤-PDE tran-scriptional mechanism described in this study, we can suggest that a unique combination of molecular interactions may be required for rod-specific transcription from this TATA-and Inr-less promoter (Fig. 5). This is consistent with the combinatorial model of transcriptional regulation of cell-specific gene expression.