Sp Family of Transcription Factors Is Involved in Valproic Acid-induced Expression of Gαi2 *

Valproic acid-induced gene expression has been attributed to the DNA-binding activity of the transcription factor activator protein 1 (AP-1). Using K562 cells, we have studied valproic acid-induced transcription from the human Gαi2 gene promoter, which lacks AP-1-binding motifs. We find that valproic acid-induced expression of Gαi2 is inhibited by mithramycin A, a compound that interferes with Sp1 binding to GC boxes in DNA. Three Sp1-binding sequences, located at +68/+75, −50/−36, and −92/−85 in the promoter, accounted for about 60% of this transcriptional effect, as judged by transient transfection assays. Electrophoretic mobility shift assays indicated that these sites bind members of the Sp family of transcription factors. Binding to DNA was inhibited by mithramycin A and was greater in nuclear extracts from cells treated with valproic acid than in control cells. Okadaic acid, calyculin A, and fostriecin, which are potent inhibitors of protein phosphatase, suppressed the transcriptional response to valproic acid. This inhibitory effect was not observed when promoter constructs containing mutations in the referenced Sp1-binding sites were used for transfections. In nuclear extracts from cells cultured in the presence of these inhibitors, the binding of Sp1/Sp3 to DNA probes was much less than in control cells. Alkaline phosphatase treatment of nuclear extracts resulted in enhanced binding of Sp proteins to the DNA probes. These results are consistent with the idea that dephosphorylating conditions enhanced Sp binding to the DNA probes as well as Sp-mediated transcription induced by valproic acid. This study demonstrates that the gene expression-inducing effect of valproic acid occurs, in part, through the Sp family of transcription factors.

Valproic acid (2-propylpentanoic acid), a branched shortchain fatty acid, is an anticonvulsant agent that is also clinically effective as a mood stabilizer in the treatment of manic depression (bipolar affective disorder) (1)(2)(3)(4)(5). The biochemical basis for the neurotrophic effects of valproic acid and other mood stabilizers has intrigued investigators for a long time, but at the present time the modes of action of these drugs are not clearly understood (4,6,7). Among the effects of valproic acid is an increase in gene expression, an effect that has been demonstrated for the genes for Bcl-2 and growth cone-associated p43 in SH-SY5Y cells (8), Tcf/Lef-dependent transcription in 293T cells (9), and the tyrosine hydroxylase gene (10,11).
Reports that valproic acid increases the expression of genes regulated by the transcription factor AP-1 1 (12)(13)(14) have led to the conclusion that the molecular mechanism of valproic acidinduced gene expression is via DNA binding activity of this transcription factor (8,14). However, whether AP-1 is the only transcription factor that can mediate valproic acid-induced gene expression is by no means resolved. A gene promoter that lacks AP-1 motifs, such as the G␣ i2 gene promoter (15)(16)(17), provides an excellent paradigm for testing the role of other transcription factors. Using K562 cells, we show, in this study, that valproic acid induces expression of G␣ i2 and that it activates transcription from the G␣ i2 gene promoter, in part, through the Sp family of transcription factors.

EXPERIMENTAL PROCEDURES
Chemicals and Reagents-Sodium salts of caproic and valproic acids were purchased from Sigma. Mithramycin A was purchased from Biomol Research Laboratories, Inc. (Plymouth Meeting, PA). Okadaic acid, calyculin A, and fostriecin were purchased from Alexis Biochemicals (San Diego, CA). Immobilon-P polyvinylidene difluoride transfer membranes for proteins were products of Millipore Corp. (Bedford, MA). Oligonucleotides used for mutagenesis experiments were purchased from Integrated DNA Technologies, Inc. (Coralville, IA). Oligonucleotides used for gel shift assays and antibodies to ␤-tubulin, G␣ i2 , Sp1, Sp2, Sp3, C/EBP␤, and c-Fos were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Anti-Sp1 antibody was also obtained from Upstate Biotechnology, Inc. (Lake Placid, NY). Dual luciferase assay systems and calf intestinal alkaline phosphatase were purchased from Promega (Madison, WI). Expression plasmid harboring the gene for dominant negative ERK2 was obtained from Dr. Melanie Cobb (University of Texas Southwestern Medical School, Dallas, TX). The sources of all other chemicals and reagents and the construction of the plasmid containing full-length human G␣ i2 gene promoter (pG␣ i2 (Ϫ1214/ϩ115)-luc) and mutants thereof have been previously described (18).
Cell Cultures and Transfection Studies-K562 cells, obtained from the American Type Culture Collection (Manassas, VA), were maintained in culture as previously described (19). The cells (1 ϫ 10 5 cells/ well in 1 ml of medium) were seeded in 24-well plates; wells on the periphery of the plates were not used. After 24 h, the cells were transfected with 0.5 g of plasmid DNA (containing the G␣ i2 gene promoter or mutant) and 1.5 l of FuGENE TM 6 transfection reagent (Roche Molecular Biochemicals) 1 h before the addition of valproic or caproic acid. In co-transfection experiments described in the legend to Fig. 3, only 0.25 g of the reporter plasmid was used. When chemical inhibitors such as mithramycin A and okadaic acid were used, they were added 1 h prior to the addition of valproic acid. The cells were harvested 24 h later, by centrifugation at 12,000 ϫ g (45 s) in 1.5-ml microcentrifuge tubes and washed once with 1 ml of 1ϫ phosphate-buffered saline, pH 7.4. The cell pellets were then lysed with 150 l of 1ϫ cell lysis reagent (Promega, Madison, WI) and kept at room temperature for 10 min. The lysed material was then centrifuged for 2 min at 12,000 ϫ g, and the luciferase activity and protein content of each lysate were measured as previously described (18). In carrying out the dual luciferase assay protocol, we noted that valproic acid robustly induced the luciferase gene in conventional luc-reporter constructs (e.g. Promega's pRL-TK-luc and pRL-CMV-luc reporter vectors) usually used for normalization in transfection experiments, making it impractical to use such plasmids for this purpose. This problem was not encountered with promoterless constructs (e.g. phRG-B Renilla luciferase reporter vector from Promega). With such constructs, the dual luciferase assay resulted in similar -fold activation of transcription as when the luciferase activities were normalized to the protein content of the samples. Therefore, luciferase activity was routinely normalized to the protein content of each sample, after correcting for basal activity of GL3-basic and is expressed as -fold effect over cells that were not treated with valproic acid.
Western Immunoblotting-Immunoblotting of G␣ i2 was carried out with previously described protocols (18,19). The samples used for these assays were the same cell lysates prepared for the luciferase assay. The same blots were reprobed for ␤-tubulin, as loading control, to verify efficiency of transfer of protein to the polyvinylidene difluoride membrane. Immunoblotting of Sp1, Sp2, and Sp3 in nuclear extracts was carried out with 10 g of nuclear extract protein; the blots were reprobed for C/EBP␤ and c-Fos, as loading controls.
Northern Blot Analysis-Total RNA was isolated with TRI RE-AGENT TM (Sigma) according to the manufacturer. Northern blot analysis was performed as described previously, using 32 P-labeled G␣ i2 -specific oligonucleotide probe (19). To monitor differences in loading and transfer among samples, blots were stripped and rehybridized with an end-labeled oligonucleotide complementary to the human 28 S rRNA.
Preparation of Nuclear Extracts and Electrophoretic Mobility Shift Assays (EMSAs)-Nuclear extracts were prepared as described previously (18). For EMSA, annealed 5Ј-overhang oligonucleotide sequence, containing Sp1-binding motif, was labeled with [␣-32 P]dCTP, using the Klenow fill-in reaction, and purified (18). The labeled probe was used for the EMSA reaction as described in the legends to Figs. 5, 7, and 9. After electrophoresis, the gel was dried and then exposed to Eastman Kodak Co. XAR5 film at Ϫ80°C. The radiolabeled bands were detected by autoradiography.
Site-directed Mutagenesis-Specific nucleotides in the G␣ i2 gene promoter were mutated or deleted by using QuikChange TM mutagenesis kit from Stratagene, Inc. (La Jolla, CA). Mutations were confirmed by DNA sequencing.

Valproic Acid Induces Transcription from the Human G␣ i2
Gene Promoter-We measured the transcriptional effect of valproic acid by reporter gene assay, using the full-length human G␣ i2 gene promoter linked to a luciferase reporter gene, pG␣ i2 (Ϫ1214/ϩ115)-luc, transfected into K562 cells. For comparison, this parameter was also measured in cells treated with another short-chain fatty acid (i.e. caproic acid (hexanoic acid)). Valproic acid caused a dose-dependent increase in transcription from the G␣ i2 gene promoter, whereas caproic acid had very little effect compared with valproic acid (Fig. 1A). We assessed expression of the endogenous G␣ i2 gene by using Western blotting assays to measure G␣ i2 protein levels and Northern blotting assays to measure G␣ i2 mRNA levels. As shown in Fig. 1B, increasing concentrations of valproic acid resulted in corresponding increases in G␣ i2 levels that were about 4-, 6-, and 10-fold at 1, 2, and 5 mM valproate, respectively. Valproic acid also induced a severalfold increase in G␣ i2 mRNA levels (Fig. 1C, left panel), and a dose-dependent increase was evident (Fig. 1C, right panel). Taken together, the data in Fig. 1 demonstrate that valproic acid induces expression of G␣ i2 not only from the transfected promoter but also from the endogenous promoter.
Mithramycin A Inhibits Valproic Acid-induced G␣ i2 Gene Expression-Because we have previously shown that trichostatin A, a histone deacetylase inhibitor, induces transcription from the G␣ i2 gene promoter (18), we were intrigued by the report by Phiel et al. (9) that valproic acid mimics trichostatin A in its ability to inhibit histone deacetylase. The transcriptional action of trichostatin A is known to involve Sp1 (20 -22). Therefore, we tested whether mithramycin A, a potent inhibitor of Sp1 binding to GC boxes in DNA, which is capable of interfering with Sp1-mediated gene transcription (23)(24)(25)(26)(27), would alter valproic acid-induced G␣ i2 gene expression. First, we tested whether mithramycin A would influence expression of the endogenous gene by assessing G␣ i2 levels via immunoblotting assays. Second, we tested whether mithramycin A FIG. 1. Effects of valproic and caproic acids on the expression of G␣ i2 . K562 cells (1 ϫ 10 5 cells/well in 1 ml of medium) were grown in 24-well plates for 24 h and then incubated with varying concentrations of valproic or caproic acid for another 24 h, as described under "Experimental Procedures." A, transcription from the G␣ i2 gene promoter: dose-dependent effects of valproic and caproic acids. K562 cells were transfected with pG␣ i2 (Ϫ1214/ϩ115)-luc, treated with valproic or caproic acid, and harvested 24 h later. Luciferase activities were normalized to protein content of each sample and are expressed as -fold stimulation, relative to cells that were not treated with either valproic or caproic acid. All promoter activities were corrected for basal activity of GL3-basic. Values shown are means Ϯ S.E. for triplicate assays from four different cell cultures. B, dose-dependent effects of valproic and caproic acids on G␣ i2 levels. G␣ i2 levels were determined by Western blotting, using 10 g of protein for each sample. The same blot was reprobed for ␤-tubulin as a loading control. The data are representative of two similar Western blots. C, G␣ i2 mRNA levels (measured by Northern blots) in control cells and cells treated with valproic acid. G␣ i2 mRNA levels were measured at 15, 20, and 24 h after treatment of the cells with 5 mM valproic acid (left panel). The right panel shows dosedependent effects of valproic acid; for this analysis, the cells were harvested for RNA preparation 24 h after the addition of valproic acid to the cultures. For both panels, the blots were stripped and rehybridized with an end-labeled oligonucleotide complementary to the human 28 S rRNA, in order to monitor differences in loading and transfer among samples. The data are representative of two experiments. VPA, valproic acid; CPA, caproic acid. would influence promoter activity of transfected G␣ i2 promoter. For comparison, butyrate was included as a positive control in these experiments, because we previously showed that butyrate increased G␣ i2 expression in K562 cells via Sp1-binding to DNA (18). Fig. 2A shows a 6.4-fold increase in the endogenous expression of G␣ i2 when 2 mM valproic acid was used; this -fold increase is similar to that shown in Fig. 1B at this concentration of valproate. As judged by the decreased intensity of G␣ i2 signal on Western blots ( Fig. 2A), treatment with mithramycin A not only decreased valproate-induced expression of the endogenous G␣ i2 gene (compare lane 3 with lane 2) but also the butyrate-induced effect (compare lane 5 with lane 4). By itself, mithramycin A had no effect on basal G␣ i2 expression. In transfection experiments with pG␣ i2 (Ϫ1214/ϩ115)-luc, mithramycin A drastically inhibited (87%) valproate-induced as well as butyrate-induced G␣ i2 gene promoter activity (Fig. 2B). The results of these experiments with mithramycin A suggest the involvement of Sp1-mediated gene transcription in the action of valproic acid in the expression of G␣ i2 .
MAPKs Are Not Involved in Valproate-induced Transcription-Because butyrate-induced, but not trichostatin A-induced, transcription from the G␣ i2 gene promoter was previously shown to involve the MEK-ERK signaling pathway (18), we tested whether this pathway might be involved in the valproic acid effect by carrying out co-transfection experiments with an expression plasmid harboring the gene for dominant negative ERK2. Transfection with this expression plasmid inhibited butyrate-induced transcription but had no effect on valproate-induced transcription, indicating lack of involvement of the MEK-ERK signaling pathway in the valproic acid effect (Fig. 3A). We also found that, unlike butyrate (18), valproic acid-induced transcription was not affected by U0126, a selective inhibitor of the MEK-ERK (Fig. 3B). Furthermore, selective inhibitors of p38 and c-Jun N-terminal kinase MAPKs (i.e. SB 203580 and PD 169316) also had no effect on valproateinduced transcription (Fig. 3B). These data indicate that the transcriptional effect of valproate is not influenced by any of the standard MAPK modules and therefore may be accounted for, most likely, by its inhibition of histone deacetylase.
Sp1-binding Sites Are Involved in Valproate-induced Transcription from the G␣ i2 Gene Promoter-Inhibition of histone deacetylase, a transcriptional repressor, can release an inhibitory constraint on Sp1 (22), thereby allowing this transcription FIG. 2. Mithramycin A inhibits valproate-and butyrate-induced G␣ i2 gene expression. K562 cells (1 ϫ 10 5 cells/well in 1 ml of medium), grown in 24-well plates for 24 h, were transfected with pG␣ i2 (Ϫ1214/ϩ115)-luc, as described under "Experimental Procedures." Mithramycin A (100 nM) was added to the cells 1 h after transfection with the reporter plasmid, followed by the addition of sodium valproate (2 mM) or sodium butyrate (2.5 mM) 1 h later. The cells were harvested 24 h after the addition of valproate or butyrate and processed for luciferase assay and Western blotting, as described under "Experimental Procedures." A, mithramycin A inhibits butyrate-and valproate-induced G␣ i2 levels. G␣ i2 levels were determined by Western blotting, using 10 g of protein for each sample. The same blot was reprobed for ␤-tubulin as a loading control. The plot is a densitometric quantification of three Western immunoblots similar to that shown in the inset. B, mithramycin A inhibits butyrate-and valproate-induced transcriptional activity. Luciferase activities were normalized to protein content of each sample and are expressed as -fold stimulation, relative to cells that were not treated with valproic or butyric acid. All promoter activities were corrected for basal activity of GL3-basic. Values shown are means Ϯ S.E. for triplicate assays from three different experiments. Con, control; B, sodium butyrate; Mit, mithramycin A; V, valproic acid. factor to influence transcription. To further test the idea that Sp1 is involved in valproic acid-induced transcription from the G␣ i2 gene promoter, truncation mutants of the G␣ i2 gene promoter (18) were used in transient transfection assays. We found that valproic acid-induced promoter activity was unaffected when deletions occurred in the Ϫ1214/Ϫ184 region (data not shown), a region that contains four of the seven putative Sp1-binding sites in the human G␣ i2 gene promoter (15)(16)(17). However, when the truncation was extended to Ϫ79 bp, involving only one (Ϫ92/Ϫ85) of the remaining three Sp1-binding sites, we observed a 30 -35% depression of transcription. Therefore, several promoter constructs that were mutated at these sites were used in further transfection experiments, as indicated in Fig. 4. In this figure, the letters M and D are used to designate point mutations (M) and sequence deletions (D), respectively. Single point mutations at sites 1 (ϩ68/ϩ75), 2 (Ϫ50/Ϫ36), and 3 (Ϫ92/Ϫ85) inhibited transcription by 34, 21, and 36%, respectively, whereas deletion of Sp1 sequences at sites 2 and 3 (mutants pD2 and pD3 in Fig. 4) inhibited transcription by 42 and 39%, respectively. Double sequence and point mutations at sites 2 and 3 (mutants pD2,3 and pM2,3 in Fig. 4) resulted in 59 and 46% inhibition of transcription, respectively. When all three sites were mutated (i.e. triple mutant, pM1,2,3), transcription was inhibited by 55-60% (Fig. 4), indicating that all three putative Sp1 sites (Ϫ92/Ϫ85, Ϫ50/Ϫ36, and ϩ68/ϩ75) contribute to the transcriptional response to valproic acid.
Binding of Sp Family of Transcription Factors to DNA Probes-Electrophoretic mobility shift assays were performed with three different labeled double-stranded DNA probes containing corresponding nucleotide sequences at locations 1 (ϩ68/ ϩ75), 2 (Ϫ50/Ϫ36), and 3 (Ϫ92/Ϫ85) in the G␣ i2 gene promoter. For each DNA probe, the results showed three binding complexes that are typical of nuclear extract binding to Sp1/Sp3binding element (28); therefore, only one such EMSA is shown (Fig. 5). A consensus Sp1 oligonucleotide completely abolished the binding of nuclear proteins to the three DNA probes (Fig. 5,  lane 3). Corresponding unlabeled Sp1 oligonucleotides also abrogated binding to these probes (lane 5), whereas a mutated Sp1 oligonucleotide (lane 4) and an unrelated (NF-B) oligonucleotide (lane 6) had no effect. As expected, mithramycin A interfered with this binding (lane 15).
To determine the protein composition of these complexes, antibodies to members of the Sp family of transcription factors were added to the DNA-binding assay. For all three DNA probes, supershifted protein-DNA complexes were observed with antibodies to Sp1, Sp2, and Sp3, not only in nuclear extracts from control cells (Fig. 5, lanes 7-10) but also in nuclear extracts from cells treated with valproic acid (lanes  12-14). These results confirm that these three sites (Ϫ92/Ϫ85, Ϫ50/Ϫ36, and ϩ68/ϩ75) in the human G␣ i2 gene promoter bind to members of the Sp family of transcription factors. Although Sp2 is not known to recognize the same sequence (GC-box) as Sp1 or Sp3 (27)(28)(29), it has been reported by others to bind to DNA probes designed to detect Sp1/Sp3-binding sites (30). This binding is probably attributable to the flanking GT-containing sequences in the probes used. Taken together with the data in Fig. 4, we conclude that these Sp1/Sp3-binding sites are relevant to the valproic acid-induced transcription from the G␣ i2 gene promoter. Given that mutations in these critical Sp1/Sp3-binding sites did not completely suppress valproic acid-induced transcription (Fig. 4), the possibility that other transcription factor(s) may also be involved in the valproic acid effect cannot be ruled out.
Dephosphorylating Conditions Enhance Sp1/Sp3 Binding to DNA as Well as Valproic Acid-induced Transcription from the G␣ i2 Gene Promoter-In Fig. 5, it is clear that the intensity of the protein-DNA signal was greater in cells treated with valproic acid than in control cells (lane 11 versus lane 2; also compare lanes 12 -14 with lanes 7-10). This might suggest either an increased affinity of nuclear proteins for the labeled DNA or an increased nuclear content of these transcription factors in the valproate-treated cells. We did not detect any change in the nuclear content of Sp1, Sp2, or Sp3, as measured by Western blotting of nuclear extract samples (Fig. 6). Because transcriptional action of Sp1 can be influenced by its phosphorylation state (31)(32)(33), we reasoned that the increased signal intensity probably resulted from post-translational modification of the transcription factor(s).
Protein phosphatase had previously been shown to dephosphorylate Sp1 and thereby increase the binding of this transcription factor to GC boxes on DNA (31, 32, 34). Also, alkaline phosphatase treatment of nuclear extracts has been shown to dephosphorylate Sp1 in vitro (35). As shown in electrophoretic mobility shift assays (Fig. 7), alkaline phosphatase treatment of nuclear extracts increased binding of Sp proteins to the same DNA probes used in Fig. 5. Nonspecific bands were not affected by this treatment. If this increased binding is relevant to the transcriptional activity of valproate, then a decrease in valproate-induced transcription should be expected when transfection experiments are carried out under conditions that would engender dephosphorylation in situ. Therefore, we used calyculin A and okadaic acid, two potent inhibitors of protein phosphatase 1 (PP1) and 2A (PP2A) (36,37) to inhibit protein phosphatase and then measured the effect of this inhibition on valproate-induced promoter activity. The results show that these compounds potently inhibited valproic acid-induced transcription in a concentration-dependent manner; furthermore, fostriecin, another protein phosphatase inhibitor that exhibits much greater inhibitory potency against PP2A than PP1 (38), also inhibited valproic-acid induced transcription (Fig. 8A). The protein phosphatase inhibitors were effective only against the wild-type promoter but not against promoter constructs in which Sp1-binding sites had been mutated (Fig. 8B). These results are consistent with the idea that modulation of the phosphorylation status of Sp family of transcription factors is a mechanism by which valproic acid induces G␣ i2 gene expression. This idea was further supported by electrophoretic mobility shift and supershift assays using nuclear extracts from cells treated with valproic acid and okadaic acid (Fig. 9). In these assays, the binding of nuclear extract proteins from cells treated with valproic acid and okadaic acid (Fig. 9, lanes 11-13) was significantly attenuated compared with nuclear extracts from cells that were treated with valproic acid alone ( lanes  8 -10). At the concentration used, okadaic acid, by itself, had no effect on nuclear extract binding to the probes used (Fig. 9,  compare lanes 5-7 with lanes 2-4). Taken together with the FIG. 5. EMSA. K562 cells (2 ϫ 10 6 cells) were grown for 24 h in 20 ml of medium, using 100-mm Petri dishes. Valproic acid (2 mM) was then added to some dishes, and the cells were harvested after 24 h. Nuclear extracts were prepared as previously described (18). Annealed oligonucleotides, used as probes, were labeled with [␣-32 P]dCTP by using the Klenow fill-in reaction. Using nuclear extracts from control and valproic acid-treated cells, EMSA was performed as described in detail elsewhere (18). The reactions were carried out with 2 g of nuclear extract protein for each lane. Competition experiments were carried out with a 50-fold excess (or more) of Sp1 consensus oligonucleotide (lane 3), mutated unlabeled oligonucleotide (lane 4), unlabeled oligonucleotide identical (in sequence) to the labeled oligonucleotide (lane 5), and an unrelated (NF-B) oligonucleotide (lane 6). Mithramycin A (final concentration ϭ 5 ϫ 10 Ϫ4 M) was incubated with the labeled probe for 30 min at 25°C before adding nuclear extract (lane 15). The anti-Sp1 antibodies used in lanes 7 and 8 were from Santa Cruz Biotechnology and Upstate Biotechnology, respectively. All other antibodies were from Santa Cruz Biotechnology. The EMSA shown was performed with the probe 5Ј-GTGGGTCGGGCGGGGCCGAGCCG-3Ј, which contains the putative Sp1-binding site (underlined) designated as site 1 throughout. EMSAs were also performed with 5Ј-ACCCCCGGCCCGCCCCGC-CGTCG-3Ј and 5Ј-GCCTGCAAGCCCGCCCCGGCCCAGTCACA-3Ј, which contain the putative Sp1-binding sites (underlined) designated as sites 2 and 3, respectively, throughout. With all three probes, the EMSA patterns were similar. NE, nuclear extract; Ab, antibody; oligo, oligonucleotide.
FIG. 6. Valproic acid has no effect on the expression of Sp family of proteins. Nuclear extracts (10 g of protein) from valproic acid-treated and control cells were analyzed by immunoblotting for Sp1 (A), Sp2 (B), and Sp3 (C) protein content. The nuclear extracts used for these Western blots were the same as those used for EMSAs in Fig. 5. The blots were reprobed for C/EBP␤ (D) and c-Fos (E) as loading controls. The blots shown are representative of results obtained with at least two separate preparations of nuclear extracts. VPA, valproic acid. FIG. 7. Alkaline phosphatase treatment of nuclear extracts increases binding to Sp1-binding DNA probes. Nuclear extracts (1 g of protein) from K562 cells were incubated in buffer with calf intestinal alkaline phosphatase for 20 min at 37°C. The amounts of enzyme used are indicated as units of enzyme activity. At the end of the 20 min, annealed oligonucleotide labeled with [␣-32 P]dCTP was added to the tubes, and the mixture was used for the gel mobility shift assay. The data shown are representative of three such experiments, using two different DNA probes identical to those used in Fig. 5. Alk. Phos., alkaline phosphatase; NE, nuclear extract. data in Figs. 7 and 8, these results support the idea that dephosphorylating conditions enhanced Sp1/Sp3 binding to DNA as well as transcriptional activity of these transcription factors.

DISCUSSION
Valproic acid is the active ingredient in divalproex sodium (depakote) that is used clinically in the treatment of epilepsy. In addition to its anticonvulsant activity, valproic acid also exhibits teratogenic effects in humans and animals (9, 39 -44). The mechanism by which valproic acid brings about these effects is not known. A recent study involving induction of differentiation in F9 embryocarcinoma cells has suggested that valproic acid induces the expression of peroxisome proliferating activating receptor ␦ and has associated this induction with the occurrence of valproic acid-induced teratogenicity (45). Valproic acid-induced gene expression (8 -11) has been attributed to the DNA-binding activity of the transcription factor AP-1.
However, it is unclear whether the gene expression-inducing effect of valproic acid can be attributed solely to its activation of AP-1 binding to DNA. In fact, Phiel et al. (9) had noted that valproic acid robustly activated transcription from Renilla luciferase reporter gene driven by the cytomegalovirus promoter (pRL-CMV), which they indicated does not contain AP-1 sites.
In the present study, we have used the human G␣ i2 gene promoter, which does not contain AP-1 sequence motifs (15)(16)(17), to illustrate involvement of the Sp family of transcription factors in the transcriptional activity of valproic acid. Our results show that in K562 cells, valproic acid-induced transcription from this promoter was inhibited by mithramycin A, a potent inhibitor of Sp1 binding to GC boxes in DNA (23)(24)(25)(26)(27). Valproic acid-induced transcription was also substantially depressed when three Sp1-binding sequences in the human G␣ i2 gene promoter were mutated. Electrophoretic mobility shift assays indicated that these sequences bind not only Sp1 but also Sp2 and Sp3, three closely related members of the Sp family of transcription factors. Together with the results of our reporter gene assays, these data support the conclusion that one or more members of the Sp family of transcription factors is involved in the transcriptional effect of valproic acid on the human G␣ i2 gene promoter. This study provides a substantially different perspective from reports (8,13) that defined the AP-1 response element as the modality by which valproic acid induces gene expression. The finding that mutations at these critical Sp1 sites from the G␣ i2 gene promoter did not com- FIG. 8. Effect of protein phosphatase inhibitors on valproateinduced promoter activities of wild-type and mutant G␣ i2 gene promoters. K562 cells (1 ϫ 10 5 cells/well in 1 ml of medium), grown in 24-well plates, were transfected with wild-type or mutant G␣ i2 gene promoters, as described under "Experimental Procedures." The mutant promoters contained mutations in Sp1-binding sites and were described in Fig. 4. The protein phosphatase inhibitors were added to the cultures 1 h before the addition of valproic acid (2 mM), and the cells were harvested for luciferase assay 24 h later. For each inhibitor concentration, promoter activity in the absence of valproic acid was equated to 1.0 and used as control to calculate -fold stimulation in the presence of valproic acid. Values shown are means Ϯ S.E. for triplicate assays for four different cell cultures in A and three in B. All values are corrected for basal activity of GL3-basic. A, dose-dependent effect of three protein phosphatase inhibitors (okadaic acid, calyculin A, and fostriecin) on the activity of the wild-type G␣ i2 gene promoter (pG␣ i2 (Ϫ1214/ϩ115)-luc). B, lack of effect of the protein phosphatase inhibitors on mutant G␣ i2 gene promoters. Con, control (no VPA added); VPA, valproic acid; WT, wild type; pD2,3, mutant promoter construct in which Sp1-binding motifs at sites 2 and 3 were deleted (see Fig. 4); pM1,2,3, mutant promoter construct containing nucleotide substitution mutations at Sp1-binding sites 1, 2, and 3, as shown in Fig. 4.   FIG. 9. Treatment of K562 cells with okadaic acid decreases binding of Sp1/Sp3 to DNA. K562 cells (2 ϫ 10 6 cells) were grown for 24 h in 20 ml of medium, using 100-mm Petri dishes. In cells that were treated with valproic acid and okadaic acid, okadaic acid (20 nM) was added 1 h before the addition of valproic acid (2 mM). The cells were harvested 24 h later, and nuclear extracts were prepared. EMSA was performed as described in detail in the legend to Fig. 5, using DNA probes specific to binding sites designated 1 and 2 in the legends to Fig.  5 (see "Results"). The results with both probes were similar. The representative figure shown here was generated with the probe (5Ј-GT-GGGTCGGGCGGGGCCGAGCCG-3Ј), which contains the putative Sp1-binding site (underlined), designated as site 1. All reactions were carried out with 2 g of nuclear extract protein. Supershift assays were carried out with 2 g of each antibody used. Control, nuclear extracts from cells that were not treated with either valproic acid or okadaic acid; Ab, antibody; VPA, valproic acid; OA, okadaic acid; NE, nuclear extract. pletely suppress transcription in our cellular transfection system suggests that other transcription factor(s) besides Sp family members may also be involved in the demonstrated transcriptional effect of valproic acid.
Histone deacetylase (HDAC) inhibitors are increasingly gaining attention in cancer research (46 -50) because of their ability to inhibit cell proliferation. Valproic acid is a new addition to this list and was shown recently to inhibit histone deacetylase in Xenopus and human embryonic kidney (293T) cells (9). Göttlicher et al. (49) also demonstrated that valproic acid is a powerful inhibitor of HDAC, relieves HDAC-dependent transcriptional repression, and causes hyperacetylation of histones in vivo and in cultured cells. Taken together with the studies of Doetzlhofer et al. (22) that show that inhibition of HDAC by HDAC inhibitors releases an inhibitory constraint on Sp1, making it possible for this transcription factor to associate with other accessory proteins to effect transcription, our present study provides an important dimension to further understanding of mechanisms underlying valproic acid-induced gene transcription.
Valproic acid-induced transcription demonstrated in this study was sensitive to three powerful inhibitors of PP1 and PP2A (i.e. okadaic acid, calyculin A, and fostriecin), indicating that it requires the action of protein phosphatase. These chemical inhibitor studies, however, do not precisely pinpoint which protein phosphatase may be involved. That the transcriptional action of Sp1 can be influenced by its phosphorylation state has been amply demonstrated (31-33, 52, 53). In this context, it should be noted further that Lacroix et al. (34) recently demonstrated, in interleukin-2-treated T lymphoma cell line Kit225, that PP2A but not PP1 dephosphorylates Sp1, resulting in increased binding of Sp1 to DNA as well as up-regulation of its transcriptional activity with respect to the human immunodeficiency virus type 1 long terminal repeat promoter as well as a chloramphenicol acetyltransferase reporter gene under the control of six tandem Sp1-binding sites from SV40 early promoter (SV40-tkCAT). Also, Li et al. (54) recently showed in studies with mouse embryonal carcinoma P19 cells that retinoic acid-induced transcription from the mouse -opioid receptor gene promoter is mediated by increased binding of dephosphorylated Sp1 to GC box elements in that promoter. We can infer from their studies and from our protein phosphatase experiments (Figs. 7-9) that dephosphorylated Sp1/Sp3 mediated valproic acid-induced transcription from the G␣ i2 gene promoter observed in our present study. Besides Sp1, not much is known about post-translational modification(s) that might influence the transcriptional activity of Sp family of transcription factors. Our present work suggests that, like Sp1, phosphorylation state of Sp3 might influence its DNA binding/ transcriptional activity. Interestingly, two research groups (55,56) showed recently that another post-translational modification (i.e. sumoylation) influences whether Sp3 functions as a repressor or activator of transcription.