Regulation of the Glial-specific JC Virus Early Promoter by the Transcription Factor Spl*

The regulation of glial-specific JC virus early gene expression was addressed by functional dissection of a previously uncharacterized form of the JC virus promoter (MH- 1). The MH- 1 promoter directed 31-fold higher reporter gene expression in U87MG glioma cells than in HeLa cells in a transient transfection assay. Transfection of promoter constructs containing proximal or proximal plus upstream regions revealed that reporter gene expression was activated by both proximal and tandem repeat regions in glioma cells. The proximal region contains a guanine-rich sequence, the GA box, which was found to regulate the promoter, and was recognized specifically by the transcription factor Spl. The GA box is also present in the promoters of three glial-specific cellular genes. Together with paired AP-1 and NF-1 sites in the tandem repeats, the GA box is part of a motif that is conserved between several glial-specific promoters, and is thus a potential determinant of glial-specific gene expression. These results delineate the promoter regions required for activation of the MH-1 JC virus promoter, suggest a new determinant of glial specificity, and establish a model for the investigation of combinatorial activation of a glial-specific viral promoter.

antigen transgene causes specific abnormalities of brain myelination in mice (Feigenbaum et al., 1992). Attempts to identify the molecular determinants of this specificity have resulted in the identification of several transcription factors that activate the JC virus early promoter, but their relevance * This work was supported in part by National Institute of Neurological Disorders and Stroke Grant NS 01650. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
to promoter specificity has either not been examined or is not clear (Amemiya et al., 1992;Ahmed et al., 1990;Tada and Khalili, 1992;Tamura et al., 1988;Wegner et al., 1993;Kumar et al., 1993). Furthermore, it is not known whether a single promoter region dominantly activates transcription in glial cells or whether multiple regions interact in a combinatorial fashion. Such a functional dissection has been difficult with the original, Mad-1 isolate, because the entire promoter region is duplicated in tandem.
Sequence analysis of numerous JC virus strains has shown that the Mad-1 type promoter is not commonly isolated from the brains of patients with PML (Henson et al., 1992;Martin et al., 1985). Unlike the Mad-1 strain, the promoter commonly found in PML brains contains a single TATA box in association with a guanine-rich, Spl-like site, and two upstream tandem repeats. Promoters vary in the exact junction between the two repeats, while the junction between the repeats and the proximal promoter region is perfectly maintained (Henson et al., 1992). Fig. lA shows that the structure of the MH-1 promoter, an example of the commonly isolated form, is very similar to that of the papovavirus SV40. Both MH-1 and SV40 promoters contain guanine-rich sequences (stippled) immediately upstream of a single TATA box (black), with the tandem repeats (open boxes) lying further upstream. Therefore, the structure of the MH-1 promoter lends itself to functional dissection. It can be divided between the proximal region containing the TATA box and guanine-rich region ("GA box"), and the upstream, tandemly repeated sequences.
Examination of the MH-1 JC virus promoter in this way suggests several new hypotheses. The GA box has sequence homology to the binding sites for the transcription factor Spl in the SV40 promoter. More interestingly, the promoters of two oligodendrocyte-specific cellular genes, myelin basic protein and proteolipid protein, contain GA boxes in their proximal regions, as shown in Fig. 1B (Gencic and Hudson, 1990;Devine-Beach et al., 1990). These homologies suggest that the GA box plays a role in the strength and specificity of the MH-1 promoter. The tandem repeats in the JC virus promoter, like those of the SV40 promoter, are also likely to have a major role in promoter strength, and possibly specificity. Here, the GA box is shown to regulate the JC virus promoter through interactions with the transcription factor Spl. Furthermore, the tandem repeats are shown to cooperate with the GA box to activate gene expression in glioma cells. These observations suggest a potential, combinatorial determinant for the activation of glial-specific gene expression.

MATERIALS AND METHODS
Plasmid Construction-Inserts described below were ligated into the HindIIIISmaI site of pA,PLUC, upstream of the firefly luciferase gene (gift from Aviva Symes, Molecular Neurobiology Laboratory, Massachusetts General Hospital; Maxwell et al. (1989)). pAaPLUC contained a triplet of SV40 polyadenylation signals immediately upstream of the cloning sites, such that spuriously initiated transcripts would not be translated, resulting in low background. Plasmids MH-1LUC (called plasmid C in Fig. 2) and Mad-1LUC were constructed by ligating a HindIIIIMscI restriction fragment of MH-1 or Mad-1 JC virus into pA,PLUC. The MH-1 JC virus promoter fragment used to construct MH-1LUC (plasmid C) was derived from plasmid p2 (Henson et al., 1992); the Mad-1 JC virus promoter fragment used to construct Mad-1LUC was derived from plasmid pl. The remaining inserts (A, B, and D) were produced using plasmid p2 as a template for polymerase chain reaction (PCR) amplification. Plasmid A contained the TATA box and the single GA box of the wild-type promoter, and was constructed using a pair of primers (JC7 and JC21, see Table I) that amplified this segment of the wild-type promoter. Plasmid B was produced using JC7 and a 60-base-long upstream primer (JC27) that contained four tandem GA boxes (underlined in Table I), the 3' end of which was complimentary to the template. PCR generated a product that contained all four GA boxes as well as a noncomplimentary SmaI restriction site. Each GA box was 10-11 bases apart, as measured from the central adenine, thereby separating each by one integral turn of the double helix, and placing the four binding sites on the same face of the helix. Plasmid D was the wild-type JC virus promoter except that 4 point mutations have been introduced into the GA box (Ho et al., 1989). In brief, two primer pairs (JC7 and JC19m; JC6 and JC2Om) were used to generate PCR products that overlapped across the GA box. Two 30-base-long primers overlying the GA box region each contained four mismatched bases. Both PCR products were purified in low melt agarose gels, and were added to a third PCR reaction containing primers JC6 and JC7. In this manner, using high template concentrations and low numbers of PCR cycles to avoid polymerase errors, full-length JC virus promoter was synthesized containing the four point mutations. Plasmid SV/JC-LUC was prepared by amplification of the tandem repeats of SV40, blunt-ending the PCR product with T4 DNA polymerase (New England BioLabs) and ligating the product into the S m d site upstream of the GA box in plasmid A. Plasmids were purified using column chromatography (Quiagen). All constructs were verified by restriction analysis and by dideoxy DNA sequencing (U. S. Biochemical Corp.). Plasmids were analyzed on agarose gels to assure purity and supercoiling. Transfections-HeLa or U87MG glioma cells were plated at approximately 75% confluence in 100-mm plastic dishes. 12 h later the medium was changed and incubated for 1 h. 15 pg of each plasmid DNA were mixed with 3 pg of SVCAT DNA per dish, and transfected according to the calcium phosphate method (Kingston, 1991). 12 h later 15% glycerol solution was added for 30 s, the cells were washed with serum-free medium and transferred to serum-containing medium. Forty-eight hours later the cells were washed with ice-cold PBS and harvested by scraping in 300 pl of luciferase lysis buffer which is 1% Triton X-100 and 1 mM dithiothreitol in 1 X luciferase buffer (4 mM EGTA, 25 mM glycylglycine, 15 mM MgSO,) at 4 "c. Lysates were spun to pellet cell debris. 100 pl of lysate were added to 370 pl of complete buffer solution (300 pl of 1 X luciferase buffer, 62 pl of 0.1 M potassium phosphate, pH 7.8, 4 p1 of 0.1 M dithiothreitol, and 4 pl of 0.2 M ATP). To determine luciferase activity, samples were placed in a luminometer (LKB), and 200 pl of luciferin substrate (200 p M luciferin, Sigma, in 1 X luciferase buffer) were added. The luminometer reports arbitrary light units (LU), which is a measure of the area under the curve of emitted light. The luciferase assay is sensitive over a 100,000-fold range (from 60 to 1,000,000 LU). pA3PLUC (promoterless) plasmid was transfected as a measure of background activity. Expression is linear over the 48-h incubation period (data not shown). Within each assay, LU values were normalized using CAT activity. CAT activity was measured in a standard two-phase partition assay, using butyryl-CoA (Pharmacia LKB Biotechnology Inc.) and ['4C]chloramphenicol (DuPont) as substrate (Kingston, 1991). CAT activities of the samples were at least 5-fold above background and were within the linear range of the assay. Each experiment was performed in duplicate at least three times.
Mobility Shift and Western Analysis of U87MG and HeLa Nuclear Extracts-Nuclear extracts were prepared as described (Dignam et al., 1983). Protein concentrations were determined by the Bradford assay (Bio-Rad). Mobility shift assays were performed using doublestranded oligonucleotides with sequences as shown in Table I total volume containing 25 mM Tris, pH 8.0,6.25 mM M&12,0.5 mM EDTA, 50 mM KC1,0.5 mM dithiothreitol, 10% glycerol, 2% polyvinyl alcohol, 1 pg of poly(dI.dC), and 50,000 cpm of labeled doublestranded oligonucleotide (approximately 1 pmol). 1 pl of nuclear extract (3 pg of U87MG protein and 1 pg of HeLa protein) was added, and the mixture was incubated for 20 min at 4 "C. Reactions were terminated by adding gel-loading buffer (0.4% Nonidet P-40, 2% glycerol, 0.05% TBE (Trisboric acid/EDTA), and 0.04% bromphenol blue) and products were separated on a 4% polyacrylamide gel (30:l acry1amide:bis ratio, 0.5 X TBE). Gels were dried and exposed to X-OMAT film (Kodak). For the supershift studies, 1 pg of an affinitypurified rabbit polyclonal antibody directed against Spl (PEP-2 X Transcruz, Santa CNZ Biotechnology) was incubated with nuclear extract under the above conditions for 10 min, labeled doublestranded oligonucleotides were added and the mixture was incubated for an additional 10 min, terminated, and separated on a gel. For Western analysis, 60 pg of nuclear extract were boiled in 2% SDS loading buffer and separated by 8% SDS-polyacrylamide gel electrophoresis. Following transfer of proteins onto nitrocellulose (Schleicher & Schuell), the membrane was allowed to dry for 30 min and was blocked in blotto. The filter was incubated with affinitypurified rabbit polyclonal antibodies against Spl (PEP2, Santa Cruz) at a concentration of 10 pg/ml. Secondary antibody (Vector) was detected with ECL (Amersham Corp.) on autoradiography film (Kodak).

MH-1 JC Virus Early Promoter
Directs Glial-specific Expression-Previous experiments demonstrating the glial specificity of the JC virus promoter in transient transfections assays have employed the Mad-1 type promoter. To test whether the MH-1 JC virus promoter can direct glial-specific gene expression, the promoter was ligated into a reporter plasmid upstream of the coding region of the firefly luciferase gene and transfected into U87MG glioma cells and HeLa cells. The U87MG glioma cell line has been used in similar previous experiments with the Mad-1 promoter, and the HeLa cell line was chosen as a non-glial cell control (Tada and Khalili, 1992). Table I1 shows that the MH-1 promoter stimulated 31-fold higher levels of luciferase expression in U87MG glioma cells than in HeLa cells. The Mad-1 promoter was also tested in this assay and it stimulated 40-fold more luciferase expression in U87MG glioma cells than in HeLa cells. These results demonstrate that the MH-1 promoter can direct glial-specific gene expression in a transient transfection assay.

GA Box Activates Transcription from the MH-1 Promoter-
The GA box has sequence and positional homology to the Spl binding sites in the SV40 promoter, and the latter are known to strongly stimulate the SV40 promoter (Gidoni et al., 1985). To determine whether the GA box activates transcription in glial cells, four promoter constructs were prepared and ana-

MH-1 and Mad-1 JC virus promoters stimulate higher levels of reporter gene expression in U87MG glioma cells than in HeLa cells
The derivation of this result from the raw data is shown. LU (see "Materials and Methods") represent the average of duplicate transfections, N is thc normalization factor based on the CAT values.  (Fig. 2). Plasmid A contained JC virus early promoter sequence proximal to the tandem repeats. This proximal promoter region gave 5-fold activation of luciferase expression over a promoterless plasmid in U87MG cells (data not shown). The GA box was multimerized four times in tandem (plasmid B) to increase the effect of transcription factors binding to this promoter element (Gidoni et al., 1985). The spacing between the central adenine of each box is 10 or 11 bases. Therefore each GA box is separated by an integral turn of the double helix (see primer JC27, Table I), placing the four elements on the same face of the helix (Su et al., 1991). Plasmid B produced 5-fold more luciferase expression in U87MG cells than did plasmid A (Fig.   2).
The full-length MH-1 promoter (plasmid C) directed 81fold stronger activity than plasmid A. Four point mutations in the GA box (plasmid D) reduced expression 3-4-fold compared to the full-length, wild-type promoter. These results demonstrate that the proximal promoter is regulated by the GA box. In addition, it is clear that the GA box acts in concert with elements in the tandem repeat region, suggesting a combinatorial mechanism of transcriptional activation of the JC virus early promoter.
Determinants of Glial Specificity-The proximal region of the proteolipid protein, and myelin basic protein and JC virus promoters contain GA boxes (Fig. lB), suggesting that the proximal promoter might confer specificity in addition to promoter strength. It also seems likely that the tandem repeats act as an enhancer and regulate promoter specificity. Having shown that both promoter regions cooperate to activate reporter gene expression in glial cells, the promoter constructs shown in Fig. 2 were transfected into HeLa cells to determine whether either region activated transcription in nonglial cells. Although the levels of expression were 30-fold lower than in U87MG glioma cells, both the proximal promoter and tandem repeats activated luciferase expression in HeLa cells (plasmids A = 1; B = 5.7; C = 42; D = 22). The cooperation between the proximal and upstream regions was 2-fold weaker in HeLa cells than in U87MG cells. Thus, while the quantity of activation was much higher in glial cells, both promoter regions activated expression in glial and nonglial cells.
These findings suggest that transcription factors common to glial and nonglial cells activate expression weakly but that in glial cells other factors act in a combinatorial fashion to

FIG. 3. A, mobility shift assay shows that similar proteins in HeLa
and U87MG nuclear extracts bind to the GA box. The labeled oligonucleotide binding site in lanes 1-6 contained the sequence of strongly activate expression. An alternative explanation could be a repressor acting at the basal promoter in nonglial cells, or an antiterminator acting in glial cells. In the former situation, transcription factors in both cell types could activate the promoter, but transcription would be greatly reduced in nonglial cells by the action of a repressor on the basal promoter. To test this possibility, a chimeric promoter was constructed which placed the two tandem repeats from the SV40 promoter upstream of the proximal JC virus promoter (pSV/ JC-LUC). If a repressor was operating at the basal promoter region, the SV/JC-LUC promoter should still be more active in U87MG cells than in HeLa cells. However, the chimeric promoter activated luciferase expression strongly in both cell types (data not shown). This result suggests that glial specificity is not due to the action of a repressor in nonglial cells.
Spl Binds to the GA Box-The GA box has sequence homology to the consensus binding site for the zinc finger transcription factor, Spl. Recombinant HeLa Spl binds to the GA box (Henson et aL, 1992), but the identity of binding proteins in nuclear extracts has not been established. To characterize these proteins, nuclear extracts from U87MG and HeLa cells were prepared, incubated with [32P]-labeled oligonucleotides containing the GA box sequence, and analyzed by mobility shift assay. Lanes 1-6 of Fig. 3A employed a labeled oligonucleotide containing the JC virus GA box sequence (see Table I). Lanes 2 and 3 show that two distinct bands bind as a doublet to the GA box from both HeLa and U87MG nuclear extracts. The doublet was competed out by a 20-fold excess of cold S p l consensus oligonucleotide (lane 4 ) , but not by an oligonucleotide containing an AP-1 site (lane 5 ) nor by an oligonucleotide of irrelevant sequence (lane 6). Point mutations in the GA box abolished binding of the doublet (lanes 7 and 8). Fig. 3A suggested that Spl is one of the GA box binding proteins in the nuclear extracts. T o determine if S p l is in fact producing the retarded bands, antibodies to Spl were added to the mobility shift assay (Fig. 3B). The upper band of the doublet in HeLa (lane 4 ) and U87MG (lane 6 ) extracts was further retarded ("supershifted") by anti-Spl antibodies. These results show that Spl from U87MG and HeLa cell nuclei binds to the GA box in an identical fashion. The lower band of the doublet was not supershifted and probably represents binding of a recently identified S p l family member, Sp3 (Kingsley and Winoto, 1992). Western analysis of the U87MG and HeLa nuclear extracts showed two protein bands of expected M,, representing differentially phosphorylated species of S p l (data not shown; Jackson et al., 1990). Thus, postranslational modification of S p l is similar in the two cell lines. Since the GA box regulates the MH-1 promoter, it is a functional S p l binding site. analysis than are the promoters of most cellular genes. This study has utilized the previously uncharacterized, commonly occurring form of the JC virus promoter to examine which promoter regions are responsible for transcriptional activation, an analysis that was difficult to perform with the original, Mad-1 promoter. The results demonstrate that transcription is activated by cooperative interactions between two promoter regions. Spl is a newly described participant in JC virus promoter regulation, and by binding to a site (GA box) which is conserved in several glial-specific promoters, appears to play a role in promoter strength and specificity. Fig. 4 shows the locations of sites at which transcription factors are known to regulate or bind to the JC virus promoter.
Examination of the proximal and upstream promoter regions as diagrammed in Fig. 4 reveals an interesting homology with several brain-specific cellular gene promoters. The promoters of JC virus, myelin basic protein and proteolipid protein each contain a GA box in their proximal regions, and adjacent AP-1 and NF-1 sites in their upstream promoter regions (see above; Amemiya et al. (1992)). Paired AP-1 and NF-1 sites are present in the promoters of several other genes that are selectively expressed in the nervous system, including the human glial fibrillary acid protein promoter, the mouse neurofilament L promoter, and the human proenkephalin promoter (Amemiya et al., 1992). There is evidence that the NF-1 sites contribute to the glial specificity of the JC virus promoter (Kumar et al., 1993). These observations suggest the hypothesis that the combination of a GA box with paired AP-1 and NF-1 sites in a promoter may be an determinant of glial specific transcriptional activation. The evidence of cooperative interactions between the GA box and upstream regions presented in this paper is consistent this hypothesis. Because both proximal and upstream elements activate expression in nonglial cells, although at a much lower level than in glial cells, a repressor could be operating in the basal promoter region (Levine and Manley, 1989). However, a chimeric promoter containing the JC virus proximal promoter attached to the SV40 tandem repeats activates expression strongly in both U87MG and HeLa cells. Therefore it is unlikely that a repressor or transcriptional antitermination is playing a role in glial-specific regulation. The promoter constructs utilized in this report do not contain the first large T antigen binding site (LTa-I) site by virtue of the position of the HindIII site used in cloning, making it unlikely that a cellular factor in nonglial cells negatively regulates the promoter at the LTa-I site. An MH-1 JC virus promoter construct which contains the LTa-I site functioned identically to the MH-1 contruct reported here, except that it acquired the ability for repression by JC virus large T antigen.' The absence of basal transcriptional repressor is also consistent with the glial-specific motif proposed above.
The GA box appears to contribute to gene expression in the glial-specific JC virus promoter and is recognized by the transcription factor Spl. A role for Spl in specificity may appear inconsistent with the function of a widely expressed transcription factor. However, recent evidence has shown that Spl expression is regulated. Spl distribution varies during development and Spl expression is up-regulated during SV40 infection (Saffer et al., 1990(Saffer et al., , 1991. Immunohistochemical observations suggest that in developing mouse brain and adult human brain, Spl expression is restricted to oligodendrocytes and a limited number of other cell types (Saffer et al., 1991;Henson et al., 1992). Therefore, by acting in a combinatorial manner with other transcription factors in glial cells, Spl may contribute to promoter specificity.