Growth/Cell Cycle Regulation of Sp1 Phosphorylation*

Sp1 sites can mediate growth/cell cycle induction of dihydrofolate reductase in late G1 (Jensen, D. E., Black, A. R. Swick, A. G., and Azizkhan, J. C. (1997) J. Cell. Biochem. 67, 24–31). To investigate mechanisms underlying this induction, effects of serum stimulation on regulation of Sp1 were examined. In Balb/c 3T3 cells, serum stimulation did not affect Sp1 synthesis or the relative binding of Sp1 family members to DNA; however, it did result in a rapid, ∼2-fold increase in Sp1 levels and an ∼3-fold increase in specific Sp1 phosphorylation in mid-G1. In normal human diploid fibroblasts, serum stimulation also increased Sp1 phosphorylation in mid-G1 but did not affect Sp1 levels. Therefore, Sp1 phosphorylation is regulated in a growth/cell cycle-dependent manner which correlates temporally with induction of dihydrofolate reductase transcription. Further studies revealed a kinase activity specifically associated with Sp1 in a growth-regulated manner. This activity is distinct from purified kinases previously shown to phosphorylate Sp1 in vitro and phosphorylates Sp1 between amino acids 612 and 678 in its C terminus, a region also phosphorylated in mid-G1 in vivo. Therefore, this study indicates that phosphorylation of the C terminus of Sp1 may play a role in the cell cycle regulation of its transcriptional activity.

Expression of a large number of genes associated with DNA synthesis, such as dihydrofolate reductase (DHFR), 1 is tightly regulated with cell growth and the cell cycle. Many of these genes have promoters which lack a TATAA element but contain binding sites for the transcription factors Sp1 and E2F (1). Although the role of E2F sites in growth/cell cycle regulation of transcription and the regulation of E2F by retinoblastoma protein (pRB) and related pocket proteins have been extensively characterized (see Refs. 2-4, for review), a role for Sp1 sites in growth/cell cycle regulation of transcription is only beginning to emerge (e.g. Refs. 5 and 6). We have determined that Sp1 and E2F sites have distinct roles in the growth/cell cycle regulation of the hamster DHFR promoter (6). Although complete repression of DHFR transcription in G 0 and early G 1 requires E2F sites, its induction in late G 1 is mediated by Sp1 sites. A direct role of Sp1-dependent transcription in growth regulation of transcription is supported by targeting of Sp1 by viral oncoproteins (e.g. Refs. [7][8][9], down-regulation of Sp1 expression, and/or DNA binding activity upon differentiation in some systems (10), and increased Sp1 expression during events associated with transformation (11).
Sp1 is a ubiquitous, 778-amino acid transcription factor that recognizes GC-rich sequences present in many promoters (see Refs. 1 and 2, for review). Although Sp1 has been viewed as a constitutive transcriptional activator which acts as a basal factor for TATAA-less promoters, an increasing number of studies indicate that Sp1-dependent transcription is regulated in response to a variety of signals. For example, in addition to their role in growth/cell cycle regulation of transcription, Sp1 sites are involved in induction of DHFR transcription in response to methotrexate (12), induction of CYP11A transcription in response to cyclic AMP (13), and transforming growth factor ␤ induction of p15 (14).
Regulation of Sp1-dependent transcription could be effected by changes in Sp1 abundance, DNA binding activity, and/or transactivation activity. Sp1 is O-glycosylated and phosphorylated, and both of these modifications are likely to be important in its regulation. Phosphorylation has been implicated in changes in Sp1 binding and transcriptional activation (15)(16)(17), and changes in O-glycosylation alter the stability of Sp1 in vivo (18) and its interaction with other factors (19). An additional mechanism that could underlie the regulation of gene expression through Sp1 sites has come to light with the finding that Sp1 is a member of a multigene family. Of the four known members of the family, Sp2 does not recognize the same sequence as Sp1 and Sp4 expression is restricted to the brain. Sp3, on the other hand, is ubiquitously expressed and recognizes the same sequences as Sp1 but has a more complex transcriptional activity. Whereas Sp1 appears to be almost exclusively an activating transcription factor, Sp3 contains a transcriptionally repressive domain and can act as a transcriptional activator or repressor, dependent on the promoter and cell type (20 -25). A role for Sp3 in regulation of Sp1 site-dependent transcription is seen in its mediation of p21 induction during keratinocyte differentiation (26).
To determine the mechanism(s) underlying growth/cell cycleregulated induction of Sp1 site-dependent transcription, we characterized changes in Sp1 following serum stimulation of quiescent cells. Serum stimulation leads to an increase in Sp1 levels and phosphorylation, with the increased phosphorylation correlating with induction of Sp1-dependent activation of the DHFR promoter in mid-late G 1 . We also determined that this change in phosphorylation is accompanied by changes in the association of Sp1 with a novel Sp1 kinase activity which phosphorylates the C terminus of Sp1. MATERIALS  diploid fibroblasts (NHDF) from Clonetics were maintained in Dulbecco's modified Eagle's medium, 10% fetal bovine serum (Life Technologies, Inc.), penicillin/streptomycin at 37°C in a 10% CO 2 atmosphere (27) and used at less than 30 doublings. Cells were synchronized by serum starvation in medium containing 0.5% serum for 24 h (Balb/c 3T3) or 0.2% serum for 60 -72 h (NHDF) and induced to re-enter the cell cycle by addition of serum to 10 -20% (6,27).
Electrophoretic Mobility Shift Assay of Sp1 Family Members-Balb/c 3T3 cells were harvested at the indicated times following serum stimulation and nuclear extracts were prepared as described (28). Electrophoretic mobility shift assay of binding to a double-stranded 32 P-labeled probe corresponding to the first Sp1 site in the hamster DHFR promoter was performed as described (28,29). Assays were conducted with equal amounts of nuclear extract protein from serum-starved and serumstimulated extracts. Specificity of binding was confirmed by competition with a 50-fold excess of unlabeled probe.
Analysis of ATP Pools-Balb/c 3T3 cells were labeled with 32 PO 4 , washed with phosphate-buffered saline, and harvested by scraping in 1 M formic acid. Non-acid soluble material was removed by centrifugation and free 32 PO 4 was removed and samples were neutralized as described (30). Nucleotides were separated on a 3-m Supelcosil LC-18 column using a 15-min linear gradient of 0.1 M potassium dihydrogen phosphate, pH 6.0, 8 mM tetrabutylammonium, 0.1 M potassium dihydrogen phosphate, pH 6.0, 8 mM tetrabutylammonium, 30% methanol at a flow rate of 1.5 ml/min (31). Eluted nucleotides were detected by their absorbance at 254 nm using a Perkin-Elmer UV95 uv/visible in-line detector and radioactivity was detected by counting Cerenkov radiation with an INSUS Systems Inc. ␤-RAM in-line radiation detector. Levels of ATP in samples were estimated by comparison of peak height with that obtained with unlabeled standard ATP. Relative specific activity of ATP was determined by dividing radioactive peak height by absorbance peak height.
Phosphoamino Acid Analysis-Sp1 was immunoprecipitated from serum-stimulated 32 PO 4 -labeled cells, separated by 8% SDS-PAGE, and transferred to polyvinylidene difluoride membrane. The region of the membrane containing the 32 P-labeled Sp1 (as determined by autoradiography) was excised and phosphoamino acid analysis was performed by two-dimensional thin layer electrophoresis in the presence of unlabeled phosphoserine, phosphothreonine, and phosphotyrosine standards as described (32).
Assay of Promoter Activity-Balb/c 3T3 cells were transfected with 20 g of DNA (5 g of reporter constructs and 10 g of Flag-Sp1 constructs) by the calcium phosphate co-precipitation method and chloramphenicol acetyltransferase activity was assayed as described (6). Luciferase activity was assayed 24 h after serum stimulation using the Promega Luciferase Assay System. Correction for differences in transfection efficiency between precipitates used the activity of a co-transfected Rous sarcoma virus long-term repeat-chloramphenicol acetyl-transferase construct which is unaffected by Sp1 overexpression. DHFR⌬E2F and 5XSp1⌬53MLP constructs and adenovirus major late promoter constructs were as described (6,27) except where the chloramphenicol acetyltransferase cDNA was replaced with that of firefly luciferase.
GST-Sp1 Constructs and Expression of GST Proteins-GST-Sp1 and GST-Sp1(1-612), in which Sp1 cDNA was fused to glutathione Stransferase cDNA in pGEX1, were kind gifts from Dr. J. Horowitz and are described elswhere (27). Other GST-Sp1 constructs were generated using standard cloning techniques by deletion between the 5Ј and 3Ј end of the cDNA and the internal XmnI and Eco47III sites of Sp1. Bacteria, freshly transformed with GST-Sp1 expression vector, were grown to mid-log and protein was induced for 4 h with 1 mM isopropyl ␤-Dthiogalactopyranoside. Cells were harvested by centrifugation and extracted with 6 M urea. Following removal of particulate material by centrifugation, proteins were renatured by removing urea through dialysis against 10 mM Tris-HCl, pH 7.5, 0.2 M NaCl, 50 M ZnCl 2 , 5% (w/v) glycerol, 1 mM 2-mercaptoethanol, and stored at Ϫ90°C.
In Vitro Phosphorylation of GST Fusion Proteins-GST fusion proteins were bound to glutathione-Sepharose for 30 min at 4°C and unbound proteins were removed by washing the beads with RIPA (4 ϫ 1 ml) and kinase buffer (3 ϫ 0.5 ml). Glutathione-Sepharose bound GST-fusion proteins were incubated in 50 l of kinase buffer containing 5 l of 1 M NaCl extract of Sp1 immunoprecipitates and 10 Ci of [␥-32 P]ATP or [␥-32 P]GTP (6000 Ci/mmol) for 30 min at 37°C. Where indicated, inhibitors were added to the GST fusion proteins prior to addition of 1 M NaCl extracts and ATP. Phosphoproteins were analyzed by SDS-PAGE and autoradiography.
Analysis of Phosphorylation of Flag-Sp1 Fusion Proteins-Flag-Sp1 fusion constructs were generated by subcloning Sp1 cDNA into pFlag-CMV2 (Kodak). pFlag-Sp1(1-612) and pFlag-Sp1(612-778) were produced likewise except that 5Ј and 3Ј deletions of the Sp1 cDNA were generated using the internal BamHI site in Sp1 cDNA. Balb/c 3T3 cells were transfected with these plasmid constructs (20 g), serum-starved and -stimulated as described (6). Cells were harvested 9 h after serum stimulation and Flag fusion proteins were immunoprecipitated from the pre-cleared extracts with a mouse monoclonal anti-Flag antibody (M2; Kodak) and protein G-Sepharose as described above. Proteins were analyzed by SDS-PAGE and autoradiography. Expression of Flag fusion proteins was confirmed by Western blot analysis with a rabbit anti-Flag antibody (Santa Cruz).

Comparison of Sp1 from G 0 and G 2 Balb/c 3T3 Cells-To
elucidate mechanisms underlying the growth/cell cycle regulation of Sp1-dependent transcription, changes in various aspects of Sp1 regulation were examined in serum-starved Balb/c 3T3 cells and in these cells following 8 h of serum stimulation (i.e. cells in G 0 and late G 1 , respectively (6)). Levels of expression, rate of synthesis, and phosphorylation of Sp1 were examined by labeling equal numbers of cells with [ 35 S]methionine or 32 PO 4, followed by immunoprecipitation of Sp1 and SDS-PAGE/autoradiography and Western blotting. Serum stimula-tion resulted in an ϳ2-fold (1.9 Ϯ 0.4, n ϭ 8) increase in Sp1 levels ( Fig. 1A), whereas Sp1 synthesis remained essentially unchanged (G 1 /G 0 ϭ 0.8 Ϯ 0.1, n ϭ 4; Fig. 1B). Incorporation of 32 P into Sp1 increased ϳ6-fold (6.1 Ϯ 1.5, n ϭ 4) with serum stimulation (Fig. 1C). Since the increase in phosphorylation was greater than the increase in the level of protein, the increase in 32 P incorporation not only reflects the increased amounts of Sp1 but also an increase (3.1 Ϯ 0.4-fold, n ϭ 4) in the specific activity of Sp1 molecules.
Growth factor stimulation of quiescent Swiss 3T3 cells leads to a rapid increase in phosphate uptake, ATP turnover, and ATP pool size (33,34). To determine if the changes in incorporation of 32 P into Sp1 were due to differences in the specific activity of ATP pools, cells were labeled as above and ATP specific activities were measured (31). In Balb/c 3T3 cells, serum stimulation led to a 1.8-fold increase in cellular ATP pools, but only a slight (1.3-fold) increase in the specific activity of ATP (Table I). Since the change in specific activity of ATP is significantly lower than the increase in specific activity of Sp1, there must be an increase in Sp1 phosphorylation during the G 0 to S phase transition.
To determine if the ratio of Sp1 family members changed following serum stimulation, electrophoretic mobility shift assays of nuclear extracts were performed. These assays detected no change or a slight decrease in Sp1 family binding activity following serum stimulation (Fig. 1D). This apparent inconsistency with the Western blot data (Fig. 1A) arises from different methods of normalization. Western blots were normalized to cell number and, therefore, reflect a change in the level of Sp1 per cell; however, electrophoretic mobility shift assays were normalized to the amount of protein in the extracts, which increases on a per cell basis following serum stimulation. Western blot analysis of nuclear extracts indicated that, when normalized to total nuclear protein, the relative levels of Sp1 in G 1 extracts were also the same or slightly less than those in G 0 extracts (data not shown). Thus, although the level of Sp1 per cell increases following serum stimulation, its specific binding activity remains unaltered. These electrophoretic mobility shift assay results also indicate that the relative levels of Sp1 and Sp3 do not change significantly following serum stimulation (Fig. 1D, arrowhead and arrows). Therefore, of the parameters tested, only changes in Sp1 levels and/or phosphorylation could account for the induction of Sp1-dependent transcription seen upon serum stimulation.
Time Course for Serum Induction of Sp1 Levels and Phosphorylation-Since serum induction of Sp1-dependent transcription from the DHFR promoter occurred in late G 1 (6), Sp1 levels and phosphorylation were assessed in Balb/c 3T3 cells at various times following serum stimulation. As seen in Fig. 2A (lower panel), induction of Sp1 levels was apparent by 2 h and was essentially maximal by 4 -6 h following serum stimulation. In keeping with the increased levels of Sp1, 32 P incorporation into Sp1 also increase by 2 h; however, a clear difference between the timing of the increase in Sp1 levels and its increased phosphorylation became apparent when the changes in Sp1 levels were taken into account (Fig. 2B). Following serumstimulation, specific Sp1 phosphorylation changed little for the first 4 h, but increased to maximal levels by 6 -8 h. Since S-phase occurs 10 -12 h following serum stimulation of these cells, Sp1 phosphorylation is induced in mid-G 1 and, therefore, occurs concomitant with or slightly precedes the induction of Sp1-dependent DHFR transcription (6).
The delayed induction of DHFR following serum stimulation  would indicate that it is relatively insensitive to Sp1 levels. This is supported by experiments in which overexpression of Sp1 had little (1.34 Ϯ 0.44-fold, n ϭ 4) effect on DHFR⌬E2Fluciferase activity in serum-stimulated cells, whereas it led to a 2.89 Ϯ 0.69-fold induction of transcription from a promoter containing 5 Sp1 sites upstream of a TATAA box (5XSp1⌬53MLP). The higher sensitivity of the latter promoter to Sp1 levels is consistent with its more rapid induction following serum stimulation of Balb/c 3T3 cells (6).
To determine if growth regulation of Sp1 phosphorylation is unique to Balb/c 3T3 cells, a similar experiment to that in Fig.  2A was performed using NHDF which enter S-phase 10 -14 h after serum stimulation (27). These cells were chosen because (a) they represent primary human cells, and (b) Sp1 levels do not change significantly during the G 0 to S phase transition in these cells (Ref. 27 and Fig. 3A). Serum stimulation of NHDF led to an increase in Sp1 phosphorylation (2.4 Ϯ 0.4-fold, n ϭ 4) while neither Sp1 levels (G 1 /G 0 levels ϭ 1.2 Ϯ 0.1, n ϭ 7) nor the specific activity of ATP pools was affected (Fig. 3A, Table I). Thus, induction of Sp1 phosphorylation in response to growth stimulation is not cell type-specific and can be seen in the absence of significant changes in Sp1 protein levels. Induction of Sp1 phosphorylation was also delayed following serum stimulation of NHDF, with increased phosphorylation first seen 6 -8 h following addition of serum.
The vast majority of Sp1 phosphorylation is on serine in cycling HeLa cells (35). To determine if this was the case following serum stimulation of serum-starved cells, phosphoamino acid analysis was performed on 32 P-labeled Sp1 from late G 1 cells. Since only phosphoserine was detected (Fig. 3B), the increased phosphorylation of Sp1 seen in mid-to late G 1 must be due to serine phosphorylation.
Identification of a Growth-regulated Sp1-associated Kinase Activity-It has been proposed that association of kinases with their substrate proteins can be an important factor for their specificity and activity (36); therefore, the possibility that increased Sp1 phosphorylation could be due to its association with a cellular kinase was investigated. Sp1 was immunoprecipitated from extracts of G 0 or G 1 cells and the immunoprecipitates were subjected to an in vitro kinase reaction. As seen in Fig. 4A, kinase activity immunoprecipitated with Sp1. Furthermore, the kinase activity associated with Sp1 was 2-3-fold higher in G 1 cells than in G 0 cells. Similar results were obtained when Sp1 was immunoprecipitated from whole cell extracts ( Fig. 4A) or nuclear extracts 2 and were seen with extracts from NHDF (Fig. 4) and Balb/c 3T3 cells. 3 In these reactions, although a phosphorylated band corresponding in size to Sp1 was sometimes seen (Fig. 4A, arrow), this was not the major phosphoprotein detected. The low level of Sp1 phosphorylation is not unexpected since (a) antibody, rather than 2 A. R. Black and J. C. Azizkhan, unpublished data. 3 S. Pajovic and J. C. Azizkhan, unpublished data.

FIG. 2. Analysis of Sp1 levels and phosphorylation in Balb/c 3T3 cells.
A, time course for serum induction of Sp1 levels and phosphorylation in Balb/c 3T3 cells. Equal numbers of Balb/c 3T3 cells were serum-starved for 24 h prior to re-addition of serum to 10%. Cells were labeled with 32 PO 4 for 2 h prior to harvest at the indicated times (in hours) after addition of serum (0 indicates no re-addition of serum). Cells were extracted, Sp1 was immunoprecipitated, and subjected to 8% SDS-PAGE and transferred to nitrocellulose. Incorporation of 32 P into Sp1 was detected by autoradiography (upper panel) and Sp1 levels were determined by Western blot analysis (lower panel). Arrowheads indicate the migration of molecular weight standards (116 and 83 kDa). B, increase in specific phosphorylation of Sp1. Levels of Sp1 protein and phosphorylation were quantified by densitometric analysis of autoradiograms and Western blot data as above. Specific phosphorylation was calculated by dividing the densitometric values for the band corresponding to Sp1 in each lane of the autoradiograms by those from the Western blots and are plotted relative to specific phosphorylation seen in serum-starved cells. Data are the mean (Ϯ S.E.) of values from two independent experiments.

FIG. 3. Analysis of Sp1 levels and phosphorylation in NHDF.
A, time course for serum induction of Sp1 phosphorylation in NHDF. Equal numbers of normal human diploid fibroblasts were serumstarved for 72 h prior to re-addition of serum to 10%. Cells were labeled with 32 PO 4 and Sp1 levels and incorporation of 32 P into Sp1 was determined as described in the legend to Fig. 2. Arrowheads indicate the migration of molecular mass standards (116 and 83 kDa). B, phosphoamino acid analysis of late G 1 phosphorylated Sp1. 32 P-Labeled Sp1 from normal human diploid fibroblasts which had been serum stimulated for 14 h was immunoprecipitated, subjected to 8% SDS-PAGE, transferred to polyvinylidene difluoride, and subjected to acid hydrolysis as described under "Materials and Methods." Material released from the membrane was analyzed by two-dimensional thin layer electrophoresis and autoradiography. Circled areas indicate the migration of unlabeled phosphoamino acid standards (detected by ninhydrin staining) which were co-electrophoresed with the samples and of free 32 PO 4 released during hydrolysis. Sp1, represents the major protein component of the immunoprecipitates (c.f. Fig. 1A), and (b) Sp1 that is associated with the kinase activity is likely to be already phosphorylated and would not, therefore, represent a substrate in these reactions. The identity of other bands phosphorylated in these extracts is unknown, but they are likely due to phosphorylation of proteins in the antibody preparations used for immunoprecipitation and possibly Sp1-associated proteins. Nevertheless, the data clearly demonstrate that a kinase activity does co-immunoprecipitate with Sp1 and that this activity is growth-regulated.
To determine if the Sp1-associated kinase activity could phosphorylate Sp1, it was eluted from Sp1 immunoprecipitates and used in an in vitro kinase reaction containing bacterially expressed glutathione S-transferase Sp1 fusion protein (GST-Sp1). The 1 M NaCl extraction eluted the majority of the kinase activity from the immunoprecipitates (Fig. 4A) and the eluted kinase was able to phosphorylate GST-Sp1 in a growth-regulated manner (Fig. 4B). The specificity of the association of this Sp1 kinase activity with Sp1 was confirmed, since very little phosphorylation of GST-Sp1 could be detected when a blocking peptide (recognized by the anti-Sp1 antibody) was added to the cell extracts prior to immunoprecipitation of Sp1, and no activity was detected when anti-Sp1 antibody was omitted from the immunoprecipitation (Fig. 4C). Phosphorylation did not occur in the GST portion of the fusion protein since the activity was unable to phosphorylate GST (data not shown) or some of the GST-Sp1 truncation mutants tested below (c.f. Fig. 6). Phosphorylation of Sp1 by a contaminating bacterial kinase can also be excluded since kinase reactions performed in the absence of eluate from Sp1 immunoprecipitations revealed no phosphorylation of GST-Sp1 (see Fig. 6A, lane 1). In these reactions, a band which migrated below GST-Sp1 at ϳ80 kDa could be seen; this band was seen to varying extents in different preparations of GST proteins and is presumably due to a bacterial kinase.
The Sp1-associated Kinase Activity Represents a Novel Sp1 Kinase-Sp1 can be phosphorylated in vitro by protein kinase CK2 (15,37), DNA-dependent protein kinase (35,38), and protein kinase A (PKA) (17). Therefore, various substrates and inhibitors were used to determine if the Sp1-associated kinase activity corresponded to any of these enzymes. CK2 can utilize GTP as a phosphate donor essentially as efficiently as ATP (39). However, GTP was a very poor phosphate donor for the Sp1-associated kinase (Fig. 5, lanes 1 and 2); therefore it is unlikely to represent CK2. DNA-dependent protein kinase requires Sp1 and the kinase to be bound to DNA (38). Since disruption of protein-DNA interaction with 50 g/ml ethidium bromide (40) did not affect Sp1 phosphorylation (Fig. 5, lane 3), the Sp1-associated kinase activity cannot be due to DNA-dependent protein kinase. Finally, the ability of staurosporine to inhibit the Sp1-associated kinase activity was tested. Staurosporine is a competitive inhibitor of ATP binding to a number of kinases and has a K i for PKA of 5 nM (41). Although stauro- FIG. 4. Association of Sp1 with a kinase activity. A, association of Sp1 with a growth-regulated kinase activity. Equivalent numbers of NHDF were serum-starved for 74 h (G 0 ) or serum-starved for 60 h prior to re-addition of serum to 10% for 14 h (G 1 ). Cells were then extracted and Sp1 was immunoprecipitated as described under "Materials and Methods." Immunoprecipitates were extensively washed and kinase activity was assayed by incubation in the presence of [␥-32 P]ATP followed by 8% SDS-PAGE/autoradiography (Unextracted). Alternatively, kinase reactions were performed with the immunoprecipitates which were extensively washed and extracted with 1 M NaCl (Extracted). Reactions were analyzed by 8% SDS-PAGE and autoradiography. The arrow indicates the migration of Sp1 and the numbers to the right indicate the migration of molecular weight standards. B, phosphorylation of GST-Sp1 by the Sp1-associated kinase activity. The presence of kinase activity capable of phosphorylating Sp1 in a 1 M NaCl eluate of Sp1 immunoprecipitates from G 0 or G 1 cells was assessed in in vitro kinase reactions containing glutathione-Sepharose-purified, bacterially expressed GST-Sp1 with 5 l of eluate and 10 Ci of [␥-32 P]ATP. Reactions were analyzed by 8% SDS-PAGE and autoradiography. The left panel shows autoradiographic detection of 32 P-labeled protein and the right panel shows the Coomassie Blue-stained gel. C, as in B, except that immunoprecipitation reactions from G 1 cells were carried out in the presence of anti-Sp1 antibody (␣Sp1), anti-Sp1 antibody and a blocking peptide recognized by the antibody (␣Sp1 ϩ peptide) or in the presence of no antibody (No antibody).

FIG. 5. Characterization of the Sp1-associated kinase activity.
In vitro kinase assays of the Sp1-associated kinase activity were carried out as described in the legend to sporine was able to inhibit the Sp1-associated kinase activity (Fig. 5, lanes 4 and 5), the IC 50 of this inhibition was 40 -50 nM under conditions where the IC 50 for PKA would be Ͻ10 nM; therefore, the Sp1-associated kinase activity also differs from PKA. Inhibition of the activity by 50 nM staurosporine further argues against the kinase being CK2 which has a K i of 6 M for staurosporine (41). Thus, the growth-regulated Sp1-associated kinase activity differs from kinases known to phosphorylate Sp1 and presumably represents a novel Sp1 kinase.
Consistent with the physiological relevance of the Sp1-associated kinase activity, treatment of serum-stimulated cells with 100 nM staurosporine for 6 h led to a 58% (Ϯ5%, n ϭ 5) reduction in activity of DHFR⌬E2F-luciferase. However, it is not possible to attribute this effect to direct inhibition of the Sp1-associated kinase since this level of staurosporine inhibits a broad range of kinases (41), including cyclin-dependent kinases which are involved in cell-cycle progression. G 1 Phosphorylation of the C Terminus of Sp1-Since Sp1 is multiply phosphorylated (35), the region of Sp1 preferentially phosphorylated in G 1 was investigated by comparing the region phosphorylated by the Sp1-associated kinase with phosphorylation of Sp1 mutants in vivo. In addition to phosphorylating full-length GST-Sp1 (GST-Sp1(1-778)), the Sp1-associated kinase activity could phosphorylate C-terminal truncations of Sp1 which contained amino acids 1-658 or 1-687 (Fig. 6A). However, no phosphorylation of the fusion protein containing amino acids 1-612 of Sp1 was detected, indicating that the phosphorylation is in the C-terminal region of the protein. The band seen in the 1-612 lane migrates below GST-Sp1(1-612) and corresponds to the band produced by bacterial kinase seen in lane 1 (Fig. 6A, arrowhead). When N-terminal truncation mutants were examined, the kinase was able to phosphorylate GST-Sp1(612-778), consistent with the phosphorylation occurring within this C-terminal region. Since no phosphorylation of GST-Sp1(687-778) was detected, the sites phosphorylated by the Sp1-associated kinase activity appear to lie between amino acids 612 and 687 of Sp1. Interestingly, the activity could phosphorylate both GST-Sp1(658 -778) and GST-Sp1(1-658) indicating that at least two sites are phosphorylated by this kinase activity in these in vitro assays.
Studies of Sp1 phosphorylation in vivo used Sp1 truncation mutants containing amino acids 1-612 or 612-778 since they have been shown to correctly localize to the nucleus (35). Nterminal Flag-tagged Sp1 (Flag-Sp1), Flag-Sp1(1-612), or Flag-Sp1(612-778) were transfected into Balb/c 3T3 cells and their phosphorylation was determined using an anti-Flag antibody to immunoprecipitate the proteins from 32 PO 4 -labeled late G 1 cells. SDS-PAGE/autoradiographic analysis clearly detected phosphorylation of not only full-length Flag-Sp1 and Flag-Sp1(1-612) but also of Flag-Sp1(612-778) (Fig. 6B). Although phosphorylation of Flag-Sp1(612-778) in late G 1 was less than that of full-length Flag-Sp1 protein, it was not possible to determine how the phosphorylation of these transfected proteins varies with serum stimulation because Western blot analysis has revealed that their expression (which is driven by the cytomegalovirus promoter) is very low in serum-starved cells (data not shown). However, the data clearly demonstrate that the C-terminal portion of Sp1 is phosphorylated in late G 1 cells in vivo and support the physiological relevance of the growth-regulated phosphorylation of this region seen in vitro.

DISCUSSION
To elucidate possible mechanisms underlying the serum induction of Sp1 site-dependent transcription, we have examined the levels, synthesis, phosphorylation, and DNA binding of Sp1. Serum stimulation of Balb/c 3T3 cells leads to an increase in Sp1 levels and phosphorylation, while neither Sp1 synthesis nor its specific DNA binding activity change. The increase in Sp1 phosphorylation was greater than that of protein, indicating that it results, at least in part, from an increase in the level of phosphorylation of individual Sp1 molecules. This was confirmed in NHDF where increases in Sp1 phosphorylation were seen with no significant increase in protein levels. The ratios of the different Sp1 site binding complexes were not affected by serum stimulation, indicating that the induction of transcription is not simply due to changes in the relative binding of stimulatory (Sp1) and inhibitory (Sp3) factors to the Sp1 sites. Therefore, of the parameters tested, only changes in Sp1 levels and phosphorylation could account for the growth-related changes in Sp1-dependent transcription.
Analysis of the kinetics of these events revealed that induction of Sp1 protein levels occurs rapidly following serum stimulation of Balb/c 3T3 cells ( Fig. 2A), whereas the increase in phosphorylation is delayed (Figs. 2B and 3). Previously, we have reported that the kinetics of serum induction of Sp1-dependent transcription is promoter dependent in Balb/c 3T3 cells (6), occurring either with delayed kinetics (the DHFR promoter) or rapidly (5XSp1⌬53MLP) following serum stimulation. In vivo footprinting data demonstrating that the DHFR promoter is constitutively occupied during the G 0 -S phase transition (42) suggest that Sp1 is not limiting at this promoter. The finding that the 5XSp1⌬53MLP promoter shows greater sensitivity to Sp1 overexpression than the DHFR promoter (see "Results") supports this notion and indicates that Sp1 levels are more limiting at the more rapidly induced 5XSp1⌬53MLP promoter. This leads to a model for differential regulation of promoter activity by Sp1: at promoters where Sp1 levels are limiting, the rapid induction of Sp1 levels by serum would lead to rapid induction of transcription, whereas with promoters which are regulated by Sp1 phosphorylation and where Sp1 levels are not limiting, induction would be delayed.
While the current study was ongoing, it was reported that serum stimulation of CHO K1 cells leads to a delayed induction of Sp1 levels and that this change correlates with the increase in Sp1-dependent transcription of DHFR (43). The discrepancy between this study and our findings presumably resides in the different cell types and/or promoters used (the studies in Chinese hamster ovary cells used a truncated DHFR promoter containing only one Sp1 site which has shown very low activity in other studies (28)). Although the possible contribution of phosphorylation to the induction seen in Chinese hamster ovary cells was not addressed, it appears that Sp1-dependent transcription may be regulated at multiple levels dependent on the promoter and cell type studied. This is supported by the finding that Sp1 sites play a greater role in growth regulation of the rep3 promoter than the DHFR promoter in NIH 3T3 cells (5), and that Sp1 levels and/or binding activity do not change following serum stimulation of NIH 3T3 cells or human fibroblasts (8,27,44,45).
Although not examined here, another potential mechanism for regulation of Sp1 is through O-glycosylation which has been shown to play a role in the association of Sp1 with other factors (19) and in regulation of its degradation (18). Our findings are suggestive of a possible role for O-glycosylation since Sp1 levels change in Balb/c 3T3 cells without concomitant changes in Sp1 synthesis, indicating that its levels may be regulated by changes in degradation.
In keeping with the serum stimulation of the phosphorylation of Sp1, we have observed an Sp1-associated kinase activity, whose interaction with Sp1 is growth-regulated. Although rigorous proof that the Sp1-associated kinase activity represents a physiologically relevant Sp1 kinase will require identification of the kinase(s) involved and determination of the effects of its inhibition in vivo (46) (and is, therefore, beyond the scope of the present study), several factors argue in favor of its physiological relevance. The level of the activity is growthregulated, indicating that either the activity and/or association with Sp1 of the relevant kinase(s) is likewise growth-regulated. Furthermore, at least the majority of the activity is specifically associated with Sp1 (Fig. 4B). The specificity of the interaction is reinforced by the finding that this activity binds strongly to FIG. 6. Region of Sp1 phosphorylated by the Sp1-associated kinase activity. A, in vitro labeling of GST-Sp1 truncation mutants. Autoradiogram and Coomassie-stained 10% SDS-PAGE gel of in vitro kinase assays of the Sp1-associated kinase activity carried out as in the legend to Fig. 4B. In lane 1 (1-778, no eluate), kinase reactions were carried out with GST-Sp1 but without addition of eluate from Sp1 immunoprecipitations. The substrates in the other kinase reactions (which contained eluate from Sp1 immunoprecipitations) were full-length GST-Sp1 (1-778) or various truncation mutants of Sp1 fused to GST. The numbers above the lanes indicate the amino acids of Sp1 present in these fusion proteins. The arrowhead indicates the migration of a phosphoprotein due to nonspecific phosphorylation. B, in vivo labeling of Sp1 truncation mutants. Flag-tagged, full-length Sp1 (Flag-Sp1), and Flag-tagged truncation mutants containing the N-terminal 612 (Flag-Sp1(1-612)) or C-terminal 167 (Flag-Sp1(612-778)) amino acids of Sp1 were transfected into Balb/c 3T3 cells prior to serum starvation for 24 h. Cells were then serum stimulated for 9 h with 32 P labeling for 2 h prior to harvest. Flag-tagged proteins were immunoprecipitated from cell extracts with anti-Flag antibody and 32 P incorporation into the proteins was determined by 8% (Flag-Sp1 and Flag-Sp1(1-612)) or 15% (Flag-Sp1(612-778)) SDS-PAGE and autoradiography. For clarity, panels show only the region of the autoradiogram where the fusion proteins migrate. With the exception of material trapped at the top of the 15% gel, no other bands were apparent on the autoradiograms. C, amino acid sequence of the C-terminal region of Sp1. Underlined regions delineate the zinc fingers of Sp1, while the bold letters show the region phosphorylated by the Sp1-associated kinase in the in vitro assays. The 3 serines in the labeled region are shown in bold italics.
GST-Sp1 (it remains bound following extensive washing of glutathione-Sepharose-bound GST-Sp1 which had been preincubated with eluate from Sp1 immunoprecipitates 2 ). Finally, the Sp1-associated kinase activity phosphorylates Sp1 in its C-terminal domain, a region phosphorylated in late G 1

cells in vivo.
Although these studies have identified an Sp1-associated kinase that is likely to be involved in regulation of its transcription, the identity of the kinase is unknown at present. Sp1 has been shown to be phosphorylated in vitro by CK2, PKA, and DNA-dependent protein kinase; however, the Sp1-associated kinase differs from these kinases in its ability to utilize GTP and its sensitivity to staurosporine and ethidium bromide, respectively. Although it is possible that the Sp1-associated kinase activity represents more than one kinase (it phosphorylates at least two sites on Sp1 in vitro), the low level of its GTP usage, minimal inhibition by low concentration of staurosporine, and insensitivity to ethidium bromide indicate that, if it contains any of the above kinases, they represent only a small proportion of the total activity. During the preparation of this manuscript, it was reported that DNA-dependent protein kinase leads to an increase in Sp1 activity through phosphorylation of serine 131 in the N terminus of Sp1 (47). Although this phosphorylation could be enhanced by HIV-TAT, it remained dependent on the presence of DNA. Since the Sp1associated kinase did not phosphorylate the N terminus of Sp1 and was insensitive to disruption of protein-DNA complexes by ethidium bromide, these findings further emphasize that the kinase identified here is distinct from DNA-dependent protein kinase. Therefore, the Sp1-associated kinase activity identified here contains a novel growth-regulated Sp1 kinase(s).
Consistent with the data obtained in vitro, it is also unlikely that CK2, PKA, or DNA-dependent protein kinase are directly responsible for the increased phosphorylation of Sp1 in G 1 in vivo. Phosphorylation of Sp1 by CK2 leads to a reduction in Sp1 DNA binding and occurs on threonine (15,37). Therefore, since G 1 phosphorylation of Sp1 is on serine and no differences in DNA binding were observed between G 0 and G 1 cells, CK2 is unlikely to be directly responsible for the growth/cell cycleregulated phosphorylation of Sp1. Although PKA activation leads to activation of Sp1 site-dependent transcription at a number of promoters, its precise role in this phenomenon is unclear. PKA phosphorylates Sp1 in vitro and this phosphorylation leads to increased Sp1 DNA binding (17). In contrast, activation of PKA in vivo leads to enhanced Sp1 site-dependent transcription of the sgk promoter without changes in Sp1 levels or DNA binding activity (48), and enhanced binding to Sp1 sites at the NF-L promoter following PKA activation is due to a factor other than Sp1 (49). The increased DNA binding induced by PKA and the presence of only threonine (Thr 366 ) in the PKA consensus site in Sp1 preclude it from being the kinase responsible for the G 1 phosphorylation seen here. DNAdependent protein kinase requires DNA ends for activity and has been implicated in responses to DNA damage (50); therefore, it is unlikely to be relevant to the G 1 phosphorylation of Sp1. The involvement of kinases other than DNA-dependent kinase in phosphorylation of Sp1 is supported by its phosphorylation in G 1 SCID mouse fibroblasts 2 which lack DNA-dependent kinase activity (51). Thus, in keeping with the presence of Sp1 sites in multiple promoters, Sp1 is phosphorylated at multiple sites by multiple kinases which are likely to be important in regulation of its activity in response to diverse signals. Ongoing studies are aimed at identifying the Sp1associated kinase(s) and determining the precise physiological roles of its phosphorylation of Sp1.
Sp1 is multiply phosphorylated and migrates as a doublet on SDS-PAGE due to different mobilities of the phosphorylated and unphosphorylated forms (35) (Fig. 1A). Interestingly, serum stimulation did not change the ratio of these forms in either Balb/c 3T3 cells or NHDF (Figs. 1-3 and Ref. 27). Therefore, the enhanced phosphorylation of Sp1 in mid-late G 1 presumably reflects increased phosphorylation of only one or a small subset of the Sp1 phosphorylation sites. In keeping with this, the Sp1-associated kinase phosphorylated a limited region of Sp1 within its C terminus (Fig. 6). Furthermore, in addition to phosphorylation of the N terminus of Sp1, our studies revealed phosphorylation in the C-terminal 167 amino acids of Sp1 in G 1 cells in vivo. Unfortunately, due to low expression of transfected proteins in serum-starved cells, it was not possible to determine how phosphorylation of this fragment is related to the up-regulation of Sp1 phosphorylation in G 1 . Interestingly, while studies in asynchronously growing HeLa cells readily detected phosphorylation of the N-terminal portion of Sp1, they failed to detect significant phosphorylation of the C-terminal domain (35). Although this discrepancy may reflect cell type differences, together with the phosphorylation of the C terminus of Sp1 by a growth-regulated, Sp1-associated kinase activity, it further argues that the C terminus of Sp1 is preferentially phosphorylated in G 1 cells.
The region of Sp1 phosphorylated in vitro lies in the Nterminal half of the DNA-binding domain of the protein. This region contains only 3 serines (Fig. 5C); however, due to the relaxed kinase substrate specificity observed in vitro, it is not possible to say which of these are physiological sites of phosphorylation. Nonetheless, these studies delineate the region of Sp1 which can be phosphorylated by the Sp1-associated kinase activity. The N-terminal zinc finger contributes weakly to the high affinity binding of Sp1 to DNA (52), which may account for the finding that enhanced Sp1 phosphorylation in G 1 does not correlate with significant changes in the specific DNA binding activity of Sp1.
Lack of correlation between DHFR transcription and increased levels or DNA binding of Sp1 indicate that the enhanced Sp1 activity is likely to involve alterations in its association with other proteins. With regard to enhanced phosphorylation of Sp1 in G 1 , it is noteworthy that the C terminus of Sp1 has been implicated in its interaction with other transcription factors. This region of Sp1 is involved in synergistic activation with sterol regulatory element-binding protein (53,54) and mediates its interaction with TAF II 55 (55) and with both YY1 and E2F1 (27,56), two factors which can mediate serum induction and cell cycle regulation of transcription (2,57). Indeed, the region of Sp1 that interacts with E2F1 maps precisely to the region phosphorylated by the Sp1-associated kinase activity. 4 Since interaction with YY1 or E2F1 can affect Sp1-dependent transcription in the absence of their direct binding to DNA (27,56), modulation of these interactions by phosphorylation of Sp1 in mid-late G 1 could mediate changes in Sp1 site-dependent transcription. Sp1 has also been shown to physically or functionally interact with other cell cycle regulatory proteins such as retinoblastoma protein and p107 (58 -61). Current efforts are directed at determining how mutation of this C-terminal region of Sp1 affects its interaction with other factors and the growth/cell cycle regulation of Sp1dependent transcription.