Physical and Functional Interactions between USF and Sp1 Proteins Regulate Human Deoxycytidine Kinase Promoter Activity*

Deoxycytidine kinase (EC 2.7.1.74, dCK) is central to drug activity of anticancer and antiviral agents such as cytosine arabinoside (araC) and gemcitabine. HepG2 hepatocellular carcinoma cells were used to study the transcriptional regulation of dCK. 5′-Deletion and site-directed mutagenesis of the dCK upstream region (positions –464 to –27) confirmed the importance of two GC-boxes (positions –317 to –309 and –213 to –206) and two E-boxes (positions –302 to –297 and –278 to –273). In vitro electromobility shift assays with HepG2 nuclear extracts and in vivo chromatin immunoprecipitation assays with HepG2 chromatin extracts confirmed the presence of bound Sp1/Sp3 and USF1/2. Co-transfections in HepG2 cells showed that USF1 and USF2a stimulated and Sp1 repressed promoter activity from a dCK-luciferase reporter gene construct. In Sp- and USF-null Drosophila Mel-2 cells, both Sp1 and USF1 stimulated dCK promoter activity in a dose-dependent manner, however, both Sp3 and USF2a were effectively inert. Combined Sp1 and USF1 showed additive transactivation at lower concentrations of Sp1. Sp1 was inhibitory at higher levels. Stimulation by combined USF1/USF2a with Sp1 was similar to that for USF1 alone with Sp1, whereas transactivation by Sp1 and USF2a without USF1 was synergistic. Physical interactions between USF and Sp proteins were confirmed by immunoprecipitations with Sp- and USF-specific antibodies. Domain mapping of USF1 and USF2a localized the functional interactions between USF and Sp proteins to the DNA binding domain of USF. Identifying the physical and functional interactions between Sp and USF proteins may lead to a better understanding of the basis for differential expression of the dCK gene in tumor cells and may foster strategies for up-regulating dCK gene expression and improving chemotherapy with araC and gemcitabine.

The product of dCK gene is a 30.5-kDa polypeptide that is present at low levels in most tissues (7)(8)(9)(10). The human dCK gene spans over 34 kb on chromosome 4 (4q13.3-q21.1) and includes a coding region consisting of seven exons, ranging in size from 90 to 1544 bp (11). Promoter activity was localized to a 697-bp upstream fragment, including 386 bp of 5Ј upstream region, 250 bp of exon 1, and 61 bp of intronic sequence (11). The dCK promoter is highly GC-rich and lacks a TATA-box but contains a transcription initiator region located adjacent to the major transcription start site at position Ϫ146 relative to the start of translation. The initiation region also contains an imperfect E2F binding site (12). In vitro DNase I footprint and electromobility shift assays demonstrated binding of Sp1 to two GC-boxes and upstream stimulatory factor (USF) to a critical E-box. Although site-directed mutagenesis of these cis-elements implied their transcriptional importance, the detailed mechanisms by which these binding associations regulate dCK gene expression have not been established.
The Sp and USF proteins are ubiquitously expressed, yet both are implicated in the regulation of genes characterized by tissue-specific or developmental patterns of expression (13,14). Sp1 and the related Sp3 proteins can exert activating or inhibiting effects on transcription, depending on the cell or promoter context, through binding to GC-or GT-box elements (13). The USF proteins belong to the class of b-HLH-ZIP (basic helixloop-helix leucine zipper) transcription factors, including the nuclear proteins Myc, Max, Mad, Mxi1, AP4, TFEB, TFE3, MiTF, and ADD1 (15,27). USF was first identified by its capacity to stimulate transcription from the adenovirus late promoter and was purified from HeLa cells as two polypeptides, designated USF1 (43 kDa) and USF2a (44 kDa) (16,17). Subsequently, a 38-kDa form, designated USF2b, was described (15). USF1 and USF2 proteins are encoded by two separate genes, and USF2a and 2b represent alternate splice forms of USF2 (15). USF1 and USF2 proteins have similar DNA binding domains (67% homology) but differ in their N-terminal transactivation domains, yet all forms bind to the E-box motif (CACGTG) as homo-and heterodimers (18,19).
An attractive biological feature of USF and Sp proteins is their abilities to mediate a wide range of transcriptional activities via protein-protein interactions with other families of transcription factors. Moreover, evidence for cooperative interactions between USF and Sp proteins has been described, via binding to juxtaposed E-box and GC-elements (20,21). Given the proximity of the essential GC-and E-box elements in the dCK promoter, co-operative interactions between USF and Sp proteins can, likewise, be envisaged to regulate expression of this gene. In this report, we significantly extend earlier reports of dCK promoter structure and function in lymphoid cells (12). We characterize the major cis-elements and transcription factors that regulate dCK in HepG2 human hepatoma cells and document the existence of physical interactions between the USF and Sp proteins that can potentially regulate dCK promoter activity over a wide range.
Cell Culture-The human HepG2 hepatocellular carcinoma cell line was obtained from the American Type Culture Collection (Rockville, MD) and was cultured as previously reported (21). Drosophila Mel-2 (D. Mel-2) cells were purchased from Invitrogen (Carlsbad, CA) and maintained in Schneider's insect medium supplemented with 10% fetal bovine serum and 2 mM L-glutamine plus 100 units/ml penicillin and 100 g/ml streptomycin at 28°C.
Promoter constructs harboring nucleotide substitutions in putative transcription factor binding elements were prepared using an overlap extension PCR protocol (22). Separate amplifications were performed for sense and antisense mutagenesis primers (Table I) with dCK/R2 or dCK/F2, respectively, using pdCK-464/ϩ75 as template. The two prod-ucts were mixed, and a second PCR was performed using the dCK/F2 and dCK/R2. The mutant amplicons were blunt-ended with T4 DNA polymerase and subcloned into the SmaI site of pGL3-Basic. All the deletion and site-directed mutagenesis constructs were sequenced to confirm the intended deletions or mutations.
Transient Transfections and Luciferase Assay-Transient transfections of dCK-luciferase reporter gene constructs (in pGL3-Basic) into HepG2 cells and reporter gene assays were performed as previously reported (21). Firefly luciferase activities were normalized to Renilla luciferase activities. D. Mel-2 cells were co-transfected with 1 g of the

Preparation of Nuclear Extracts and Electromobility Shift Assays-Nuclear extracts from HepG2 cells or D.
Mel-2 transfectants were prepared by standard methods (24). Electromobility shift assays with HepG2 and D. Mel-2 nuclear extracts were performed as previously described (21). For supershift experiments, reaction mixtures were pretreated for 30 min with 2 g of rabbit polyclonal antibodies (Sp1 (Geneka Biotechnology, Inc.) and USF-1 and USF2 (Santa Cruz Biotechnology)).
Immunoprecipitation and Western Blot Analysis-Nuclear extracts (300 g) from D. Mel-2 cells, co-transfected with pPacSp1 and pPacUSF1, or pPacSp1 and pPacUSF2a, were diluted to 1 g/l with immunoprecipitation buffer (10 mM Tris/HCl, pH 7.5, 150 mM NaCl, and 1% Nonidet P-40, containing protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1 mM EDTA, 1 g/ml leupeptin, 1 g/ml pepstain A, 1 g/ml aprotinin) and phosphatase inhibitors (1 mM sodium orthovanadate, 50 mM sodium fluoride)). Samples were pre-cleared with protein A-agarose beads for 1 h at 4°C, then treated overnight at 4°C with 10 g of rabbit USF1 or USF2 antibody (Santa Cruz Biotechnology), or normal rabbit IgG, followed by incubation with 100 l of protein Aagarose beads for 90 min at 4°C. The beads were collected by centrifugation (8000 rpm) and washed (four times) with immunoprecipitation buffer, and the immunoprecipitates were eluted by boiling for 5 min in 30 l of 2ϫ Laemmli sample buffer. The eluates were electrophoresed through 12% SDS-polyacrylamide gels, and transferred onto a polyvinylidene difluoride membrane. The blots were developed with Sp1 (Geneka Biotechnology, Inc.), USF1 or USF2 antibodies (in TTBS containing 0.5% fat-free dried milk powder) and Lumi-Light Western blotting Substrate (Roche Applied Science), and exposed to x-ray film.

Localization of the Major cis-Elements in the dCK Promoter by 5Ј-Deletion and Mutation
Analysis-A previous report demonstrated that two GC-boxes (positions Ϫ317 to Ϫ309 and Ϫ213 to Ϫ206; designated GCa and GCb, respectively) and one E-box (positions Ϫ302 to Ϫ297; designated E-box (a)) ( Fig. 1) are important for dCK promoter activity in human lymphoblast cell lines (12). To further investigate the interrelationships between these elements and their bound transcription factors, a 539-bp upstream fragment of the dCK gene was amplified from HL-60 genomic DNA (including 464 bp of 5Ј-flanking sequence and 75 bp of the first coding exon) and subcloned into pGL3-basic to generate pdCK-464/ϩ75 (Fig. 1). Alignment of our dCK promoter sequence to the previously published sequence (12) revealed variations at three positions (not shown). However, no difference was observed between our dCK promoter sequence and the human genome sequence (GenBank TM accession AC093851). In addition to GCa and GCb-boxes and the E-box (a) elements, data base analysis revealed a previously unrecognized E-box (positions Ϫ278 to Ϫ273, designated E-box (b)) ( Fig. 1).
Similar to results with lymphoid cells, site-directed mutagenesis of either the GCa and E-box (a) elements in pdCK-464/ϩ75 was notably inhibitory in reporter gene assays (decreases of 33 and 45%, respectively, relative to the wild type construct). However, the effect of mutating both the GCa and E-box (a) elements was essentially the same as those for the individual mutants. Mutation of E-box (b) resulted in ϳ20% decrease of dCK promoter activity. Mutation of GCb resulted in increased luciferase activity (ϳ17%) over the wild type pdCK-464/ϩ75 construct, albeit less than reported for lymphoblast cells (12). The double GCa/GCb and E-box (a)/GCb mutants were also slightly activating. Interestingly, double mutation of the two E-boxes resulted in additive loss of dCK promoter activity (to 35% of the wild type). The double GCa/E-box (b) and E-box (b)/GCb mutants were slightly repressive (30 and 15% decrease, respectively).
Collectively, these results implicate a 61-bp stretch beginning 298 bp upstream of the translation start and including the GCa and E-box (a) elements as important to dCK promoter transactivation. Furthermore, they imply that the downstream GCb-box is repressive and that mutating this element can override the inhibitory effects resulting from mutation of either the GCa or the E-box elements. The failure of the double GCa/E-box (a) mutant to repress activity to an extent similar to the pdCK-298/ϩ75 deletion construct may reflect (i) the presence of E-box (b) between the GCb and E-box (a) and/or (ii) the repressive effects of GCb on pdCK-298/ϩ75 activity.

In Vitro and in Vivo Binding of Transcription Factors Bound to the GC-box and E-box Elements in the dCK Promoter-Gel
shifts were used to identify the major transcription factors that bound to the dCK promoter. When a labeled oligonucleotide (dCK-328/-289, positions: Ϫ328 to Ϫ289) containing both the GCa and the E-box (a) elements was incubated with nuclear extracts from HepG2 cells, five distinct DNA-protein complexes (labeled 1-5) were detected (Fig. 3A, lane 2), which were completely abolished by treatment with excess unlabeled dCK-328/-289 probe (Fig. 3A, lane 3). Competition and supershift results suggested a complex interplay between USF binding to E-box (a), and Sp proteins binding to the GCa element. Thus, unlabeled Sp1 consensus oligonucleotide completely abolished complexes 1, 2, 3, and 5 (lane 4), whereas a USF1 competitive  6 and 7), whereas USF1 or USF2 antibodies supershifted complexes 1, 2, and 4 and significantly decreased complexes 3 and 5 (lanes 8 and 9). When a labeled oligonucleotide spanning the E-box (b) (dCK-291/-260, positions: Ϫ291 to Ϫ260) was incubated with HepG2 nuclear extracts, one major DNA-protein complex was detected, which was completely abolished by excess unlabeled dCK-291/-260 or a USF1 consensus oligonucleotide (Fig. 3B,  lanes 2-4). However, competition was ineffective with the dCK-291/-260 oligonucleotide, including a mutated E-box (b) (E-box (b) mt, lane 7). Addition of USF1 or USF2 antibodies supershifted the DNA-protein complex (lanes 5 and 6), confirming that this contains both USF1 and USF2 proteins.
Gel shifts were also performed using HepG2 nuclear extracts and a labeled oligonucleotide spanning the GCb box (dCK-223/-181, positions: Ϫ223 to Ϫ181). Two DNA-protein complexes (complexes 1 and 2) were readily detected (Fig. 3C, lane 2) and were effectively competed by either excess unlabeled dCK-223/-181 or a Sp1 consensus oligonucleotide (lanes 3 and 4). However, competition was ineffective if a dCK-223/-181 oligonucleotide was used in which the GCb element was mutated (GCb mt) (lane 7). Addition of Sp1 or Sp3 antibody supershifted both complexes ( lanes 5 and 6), thus confirming the competition results and establishing that both Sp1 and Sp3 were bound to the GCb element.
In vivo binding of the transcription factors to the dCK promoter in HepG2 chromatin extracts was confirmed by the ChIP assays (Fig. 3D). These results demonstrate binding of the Sp and USF families of proteins to the critical GC and E-box elements in the dCK promoter.
Co-transfections of Sp1 and USF1/USF2a in HepG2 Cells-To determine the functional significance of Sp1/3 and USF1/2 binding to the dCK promoter, transient co-transfec-tions were performed in HepG2 cells using the pdCK-464/ϩ75 reporter construct with Sp1, USF1, or USF2a expression constructs. Sp1 alone effected a 36% decrease of dCK promoter activity (Fig. 4). Overexpression of USF1 or USF2a, in the absence of Sp1, resulted in a potent stimulation of luciferase activity (12-and 9-fold, respectively). When USF1 or USF2a was co-expressed with Sp1, the maximal transactivation response compared with USF1 or USF2a, alone, was decreased by ϳ25%.
These results document an apparent repressor role for Sp1 binding, consistent with the findings of our mutagenesis experiments with the GCb element (Fig. 2). Moreover, they support a functional interaction between Sp and USF proteins, consistent with the results of our gel shift experiments.
Functional Interactions of Sp1/Sp3 and USF1/USF2a in D. Mel-2 Cells-To further explore the functional relationships between Sp and USF proteins in regulating the dCK promoter, additional experiments were performed in Drosophila Mel-2 cells, which provide a null background for these transcription factors (23,25,26). The pdCK-464/ϩ75 reporter gene construct was co-transfected with expression constructs for Sp1 (pPacSp1) or Sp3 (pPacSp3 and pPacUSp3), and USF1 (pPacUSF1) or USF2a (pPacUSF2a). Luciferase activities were compared with those obtained with the empty pPacO expression vector.
The potent transactivation by Sp1 in D. Mel-2 cells showed a strong dose dependence from 10 to 50 ng. At 50 ng of Sp1, a net 105-fold stimulation of promoter activity was observed over empty pPacO vector (Fig. 5A). Likewise, USF1 alone (100 -500 ng) showed a dose-dependent transactivation with a maximum stimulation of ϳ50-fold. Transactivation by combinations of Sp1 and USF1 was completely additive at levels of Sp1 to 25 ng, however, at 50 ng of Sp1, promoter activity progressively declined with increasing USF1 (Fig. 5A). This effect became even more pronounced at higher levels of Sp1 (200 -500 ng), because under these conditions inhibition of promoter activity occurred at all doses of USF1 (not shown). In contrast to USF1, USF2a up to 500 ng was incapable of transactivating the pdCK-464/ ϩ75 reporter gene construct (Fig. 5B). However, when Sp1 was co-transfected with USF2a, a strong synergistic transactiva-

FIG. 2. Functional analysis of the cis-regulatory elements in HepG2 cells.
A series of 5Ј-deletion and site-directed mutational constructs were prepared, as described under "Materials and Methods," and transiently transfected into HepG2 cells. Numbering is based on the dCK promoter sequence in Fig. 1. Luciferase activities for the deletion and mutation constructs were expressed relative to the activity for the full-length pdCK-464/ϩ75. The data are presented as the means Ϯ S.E. for three experiments.  Table I. NS indicates nonspecific complex. Free indicates free probe. D, in vivo binding of Sp1 and USF proteins to the dCK promoter was confirmed by ChIP assays, as described under "Materials and Methods." tion of promoter activity was observed at all levels of Sp1 and USF2a (Fig. 5B).
Additional experiments were performed with combined USF1 and USF2a, thus permitting comparisons between transactivating potentials of the USF1 and USF2a homodimers, and USF1/USF2a heterodimers. In this series of experiments, combined USF1/USF2a (250 ng each), without Sp1, was essentially equivalent to 500 ng of USF1. Together with Sp1, combined USF1/USF2a was ϳ25% less transactivating than homodimeric USF2a (500 ng) and approximately equivalent to the effects of homodimeric USF1 (500 ng) (data not shown).
Although Sp3 binding to the GCa and GCb elements was detected on gel shifts with HepG2 nuclear extracts (Fig. 3, A and C), neither the long (USp3) nor the short forms of Sp3 (25 ng) transactivated dCK promoter activity when transfected individually into D. Mel-2 cells. Whereas a small (1.2-fold) stimulation of promoter activity was observed when Sp3 or USp3 was combined with USF1, there was no effect of combined Sp3/USp3 with USF2a on promoter activity (data not shown).
To verify the binding function of these ectopically expressed transcription factors in these experiments, nuclear extracts were prepared from D. Mel-2 cells transfected with USF and Sp1, and used for gel shifts with the labeled dCK-328/-289 probe. Predictably, no specific DNA-protein complexes were detected with extracts from mock transfected cells (Fig. 6, lane  2). For the dCK/USF1/Sp1 transfectants, specific protein binding was detected (lanes 3, 5, 7, 9, and 12). The identities of the bound proteins were confirmed as Sp1, USF1, and USF2a by competitions with Sp1 and USF consensus oligonucleotides (lanes 4, 6, 8, 10, 11, 13, and 14). Notably, in co-transfections with Sp1 and USF proteins, competition and supershift assays for USF1 and USF2a binding not only perturbed the complexes identified as USF1/USF2a, but also decreased the binding of Sp1 (lanes 11 and 14).
These results directly support the notion that functional antagonistic or synergistic relationships exist between Sp and USF proteins with the dCK promoter (Fig. 5) and that these effects are reflected in physical interactions between these proteins detected on gel shifts. The net transcriptional response observed appears to be specific to Sp1, because Sp3 was functionally inert, and thus varies in both magnitude and direction with the different Sp proteins and with the heteroversus homodimeric composition of the USF complex.
Co-immunoprecipitation Assays of Sp1 and USF Proteins-To provide evidence for the existence of direct physical associations between Sp1 and USF1 and/or USF2a, co-immunoprecipitations were performed from nuclear extracts prepared from D. Mel-2 cells co-transfected with expression constructs for these transcription factors. As shown in Fig. 7 (lower  panels for both A and B), USF1 and USF2a were immunoprecipitated with antibodies to these proteins. Moreover, both USF1 (panel A, lower) and USF2 (panel B, lower) antibodies also immunoprecipitated Sp1, demonstrating that Sp1 physically interacts with the USF proteins.
Identification of the USF1 and USF2a Domains Responsible for E-box-dependent dCK Transactivation or Repression with Sp1-Our functional studies with HepG2 and D. Mel-2 cells suggested a potential for wide ranging regulation of the dCK promoter by the Sp and USF families of proteins. For instance, whereas USF1 and Sp1 were both potently transactivating, USF2a and Sp3 were effectively inert. For USF1, promoter activity was repressed at elevated levels of Sp1, and Sp1 combined with USF2a was highly synergistic. Direct binding associations between Sp1 and USF1/USF2a proteins were strongly implied by gel shifts and were confirmed by co-immunoprecipitations with USF-specific antibodies.
To identify the USF domains responsible for dCK transactivation and the cooperative or antagonistic functional interactions between Sp1 and USF proteins, a series of USF deletion clones were prepared (8 for USF1; 11 for USF2) and analyzed by transient co-transfections of D. Mel-2 cells with pdCK-464/ ϩ75, with or without Sp1. Both USF1 and USF2a bind to DNA as homo-and heterodimers, involving the leucine zipper and helix-loop-helix regions (18,19). Thus, except for the ⌬223-242 mutant of USF2a, all of our USF mutant constructs were designed to contain the basic region and the helix-loop-helix domains, two cysteines (229 and 248 for USF1; 265 and 284 for USF2a) required for oligomerization, and the C-terminal leucine zipper.
Among the transfections with USF1 constructs, levels of USF1 and Sp1 proteins were essentially equal on Western blots (Figs. 8, B and C). Deletion of the first 15 amino acids from the N terminus of USF1 (⌬1-15) significantly increased USF1 transactivation, with and without Sp1, suggesting a repressor function for this region (Fig. 8A). For both conditions, promoter activity was unchanged accompanying loss of an additional 24 amino acids (⌬1-39). Further deletions to position 80 (⌬1-80) resulted in a ϳ60% decrease in the extent of USF1 transactivation (but only in the absence of Sp1). Loss of 20 more amino acids (⌬1-100) was accompanied by a nearly total loss of response to USF1, alone, and completely abolished the additive effect of USF1 with Sp1. Thus, it appears that amino acids 80 -100 are involved in a functional interaction between USF1 and Sp1. The USF1 ⌬1-130, ⌬1-156, ⌬1-175, and ⌬1-196 mutants (including the basic, helix-loop-helix, and leucine zipper domains) were transcriptionally inert. Moreover, they all exerted dominant negative effects on Sp1 transactivation, suggesting that the interaction between Sp1 and USF1 also involves the DNA binding domains.
For the transactivations with assorted USF2a deletion constructs, USF2a and Sp1 levels on Western blots were constant (Fig. 9, B and C). With the exception of ⌬76 -143 mutant (USF2b), none of the USF2 constructs by themselves significantly transactivated the dCK promoter (Fig. 9A). The effects of full-length USF2a with Sp1 on dCK promoter activity were again synergistic (Fig. 9A, right panel), and this was only slightly affected by deletion of exons 1-4 (⌬1-143). However, with further deletions (⌬1-193 and ⌬1-223), activity somewhat decreased and plateaued at ϳ70% of the level for full-length USF2a. FIG. 4. Functional interactions between Sp1 and USF1/USF2a in HepG2 cells. The full-length dCK promoter construct pdCK-464/ ϩ75 (1 g) was co-transfected into HepG2 cells along with 10 ng of pRLSV40 and expression constructs for Sp1 (100 ng), USF1 or USF2a (500 ng), or Sp1 (100 ng) with USF1 or USF2a (500 ng). Luciferase activity is expressed relative to a mock transfection with the empty vector. For all transfections, constant plasmid was maintained at 600 ng with empty pcDNA3 plasmid. Data are presented as the means Ϯ S.E. from three independent experiments.
Interestingly, internal deletion of exon 4 (⌬76 -143; USF2b) not only completely eliminated the synergistic activation with Sp1 but also decreased promoter activity below that with Sp1, alone (although higher than for empty pPacO and the other USF2 deletion mutants). Internal deletion of exon 5 (⌬143-193) resulted in 30% decreased activity compared with the full-  5 g), pPacUSF1 or pPacUSF2a (30 g each), or co-transfected with both pPacSp1 and pPacUSF1 or pPacUSF2a. Nuclear extracts were prepared and incubated with 32 P-labeled dCK-328/-289 probe in the absence or presence of Sp1 or USF1 consensus oligonucleotides or commercial antibodies for Sp1, USF1, or USF2. Specific DNA-protein complexes are noted. NS indicates nonspecific complex. Free indicates free probe. that DNA binding of USF2a is essential for its functional interaction with Sp1.
Thus, as with USF1, for USF2a, the DNA binding domains are sufficient to functionally interact with Sp1. However, in the case of USF2a, the effects of the DNA binding domains involve a synergistic transactivation rather the repression. The Nterminal exon 5 is clearly an activator, whereas exon 4 functions as a repressor in the presence of exon 5. Both exon 4 and the basic region are critical for the synergistic response with Sp1.

DISCUSSION
Phosphorylation of a variety of nucleoside analogs (e.g. araC and gemcitabine) that are important to cancer chemotherapy by dCK is a rate-limiting step in drug activation (1,2). Moreover, dCK levels in leukemia cells correlate with clinical responses to araC (28). Interestingly, our previous study of determinants of araC response in DS myeloblasts compared with non-DS myeloblasts, described a significant median 2.6-fold increased dCK expression in DS myeloblasts accompanying dramatically enhanced araC sensitivities (29). dCK promoter activity was previously localized to a 697-bp upstream fragment, including 386 bp of 5Ј sequence flanking 250 bp of exon 1, and 61 bp of intronic sequence (11,12). The dCK promoter is highly GC-rich and lacks a TATA-box. Binding of Sp1 to two GC-boxes and USF to a single E-box appeared to be critical for promoter activity in lymphoblast cultures (12). Based in part on the detection of high levels of dCK transcripts in HepG2 cells by RT-PCR (not shown), we used this cell model to further explore the transcriptional regulation of the dCK gene. Deletion and mutation analysis confirmed that the two GC-boxes (designated GCa and GCb) and the E-box element (E-box (a)) previously reported (12) comprised important regulatory regions and that GCa and E-box (a) were activators and that GCb was repressive. An additional functional E-box (E-box (b)) was also identified, which appeared to play an important transactivating role.
An attractive biological feature of USF and Sp proteins involves their abilities to mediate a wide range of transcriptional activities via protein-protein interactions with other families of transcription factors including Ets-1 (30,31), MTF-1 (32), hepatic nuclear factor-4 (33), and the basal transcription factors TFIID (34,35) and TFII-I (36,37). In the dCK promoter, the GC-box elements and the E-box elements are juxtaposed. Thus, co-operative functional interactions between Sp1/Sp3 proteins bound to the GC-boxes, and USF1/USF2 bound to the E-box elements can easily be envisaged to regulate gene expression over a wide range.
To test this hypothesis, we performed gel shifts with the dCK Ϫ328/-289 oligonucleotide, including both the GCa and E-box (a) elements and nuclear extracts from HepG2 cells. In addition to the three major complexes previously reported and identified as containing Sp and USF proteins (12), two slower migrating DNA-protein complexes (1 and 2) were identified that by competition and supershift assays were found to contain both the Sp (Sp1 and Sp3) and USF (USF1 and USF2a) families of proteins, suggesting physical interactions between these two families of proteins. USF1/USF2 binding to the E-box (b) and Sp1/Sp3 binding to the GCb box were demonstrated on gel shifts with the dCK-291/-260 and dCK-223/-181 probes, respectively. In vivo binding of Sp1, USF1, and USF2 to the dCK promoter between positions Ϫ464 and ϩ75 was confirmed by ChIP assays. Our transient co-transfections of HepG2 cells with a dCK promoter reporter gene construct and Sp1, USF1, and USF2a expression vectors suggested that functional interactions occurred between Sp1 and USF1/USF2.
To further investigate the physical and functional interactions between Sp proteins and USF1/USF2a, experiments were extended to D. Mel-2 cells. Both Sp1 and USF1 could each transactivate the dCK promoter, however, Sp3 and USF2a were largely inert. Combinations of Sp1 and USF1 were additive at low levels of Sp1, but promoter activity decreased at higher Sp1 levels. In contrast, the effects of Sp1 and USF2a on dCK promoter activity were highly synergistic. The addition of USF1 (thus forming USF1/USF2a heterodimers) decreased the synergism between Sp1 and USF2a to a level essentially indistinguishable from that for USF1 and Sp1. Thus, USF and Sp proteins could potentially regulate dCK promoter activity over a wide range, reflecting relative levels and tissue distributions of Sp1 and Sp3, as well as of USF1 and USF2a.
Sp and USF expression constructs were co-transfected into D. Mel-2 cells, and nuclear extracts were prepared for gel shift and co-immunoprecipitation studies. Physical interactions between Sp1 and USF1 or USF2a were suggested by the decreased levels of Sp1 binding to GCa on gel shifts in the presence of competitive oligonucleotides and by supershifts with USF1 and USF2a antibodies. Direct evidence for an association between Sp1 and USF1/USF2a was provided by co-immunoprecipitations of Sp1 with USF1 and USF2a. Taken together, these results strongly suggest that both Sp and USF proteins are essential for high level transactivation of the dCK promoter and that direct physical interactions between these proteins contribute to this response.
Cooperative interactions between Sp1/Sp3 and USF proteins were previously suggested to be required for the full activation of human transcobalamin II promoter (20). Likewise, our previous studies of the human cystathionine-␤-synthase-1b promoter suggested that cooperative interactions between USF1 binding to the E-box and Sp1/Sp3 binding to the GC-c box element were involved in the transcriptional regulation of this important gene (21). Our recent studies on the human reduced folate carrier B promoter have also implicated Sp1-USF interactions as critical to high level transactivation. 2 Thus, cooperative interactions between Sp1/Sp3 and USF proteins can easily be envisaged to be a global mechanism for regulating tissuespecific gene expression, presumably by protein-protein interactions. Notably, our finding that Sp1 and USF proteins can be co-immunoprecipitated from D. Mel-2 co-transfectants provides the first direct evidence that a physical interaction between these proteins may contribute to a functional response.
The domains of USF1 and USF2a responsible for dCK transactivation and functional interactions with Sp1 were mapped by deletional mutagenesis and transient co-transfections with pdCK-464/ϩ75 in Sp1-and USF-null D. Mel-2 cells. Our results with USF1 identified both repressor (amino acids 1-15) and activator (residues 39 -100) domains and suggested that amino acids 80 -100 also were critical to functional interactions between USF1 and Sp1. Interestingly, the USF1 DNA binding domains (including the basic region, helix-loop-helix, and leucine zipper domains), alone, significantly repressed the activating effects of Sp1 on the dCK promoter, strongly implying a functional interaction between this region and Sp1.
For USF2a, two repressor (exons 1, 2, and 4) and two activator (exons 3 and 5) domains were identified, yet the DNA binding domains, by themselves, were sufficient to mediate a synergistic transactivation response in combination with Sp1. This difference from USF1 was surprising in light of the close homology between the USF2a and USF1 DNA binding domains (67%), yet further documents significant functional differences between these closely related transcription factors. Interestingly, USF2b, a naturally occurring USF2 splice form, resulting from internal deletion of exon 4 (⌬76 -143), not only completely eliminated the synergistic activation with Sp1 but also decreased promoter activity below that with Sp1, alone. This suggests another regulatory component that may come into play for tissues in which USF2b is present at high levels, involving a significant repressive effect by USF2b on the extent of dCK transactivation by Sp1, and USF1 and USF2a. Indeed, USF2b has been suggested to modulate E-box-mediated transcriptional activity of the major histocompatibility complex class I gene in this fashion (38).
In summary, our results significantly extend earlier studies of transcriptional regulation of dCK in lymphoid cells. They document transcriptionally important roles for Sp and USF1/ USF2a proteins, via binding to essential GC-and E-box elements, and provide the first direct evidence for a physical association between these proteins that results in promoter transactivation/repression. Co-operative interactions between Sp and USF proteins, including Sp1, Sp3, USF1, USF2a, and USF2b, can be envisaged to effectively regulate dCK expression over a wide range and contribute to tissue-specific patterns of expression of this critical gene. Better understanding of the major determinants of dCK gene expression may lead to opportunities for therapeutic interventions by modulating dCK activity at the transcriptional level in combination with cytotoxic antiviral and antitumor nucleosides. Given our interest in the basis for the disparate therapeutic responses of Down syndrome acute myeloid leukemia patients to chemotherapy, including araC (29), it seems reasonable that alterations at the level of dCK transcription may, in part, provide a molecular explanation for this unique clinical finding (29), as well.