Activation of Transcription of the Melanoma Inducing X mrk Oncogene by a GC Box Element*

Melanoma formation in Xiphophorus is caused by overexpression of the X mrk gene. The promoter region of the X mrk oncogene differs strikingly from the corre- sponding proto-oncogenic sequences and was acquired in the course of a nonhomologous recombination with another gene locus, D . In order to identify regulatory elements leading to the strong transcriptional activa- tion of X mrk in melanoma tissue and to contribute to an understanding of the role the regulatory locus R might play in suppressing the tumor phenotype in wild-type Xiphophorus , we performed functional analysis of the X mrk oncogene promoter. Transient transfections in melanoma and nonmelanoma cells revealed the exist- ence of a potent positive regulatory element positioned close to the transcriptional start site. Contained within this promoter segment is a GC-rich sequence identical to the binding site described for human Sp1. In vitro binding studies and biochemical characterizations demonstrated the existence of GC-binding proteins in fish that share immunological properties with members of the human Sp family of transcription factors and appear to be involved in the high transcriptional activation of the X mrk oncogene. Since the identified cis element is functional in both melanoma and nonmelanoma cells, additional silencer elements suppressing X mrk expres- sion in nonpigment cells must exist, thereby suggesting a negative regulatory function for the genetically de- fined R locus. Melanoma the teleost for genetic basis of tumor and the role of receptor kinases neoplastic

Melanoma formation in Xiphophorus is caused by overexpression of the Xmrk gene. The promoter region of the Xmrk oncogene differs strikingly from the corresponding proto-oncogenic sequences and was acquired in the course of a nonhomologous recombination with another gene locus, D. In order to identify regulatory elements leading to the strong transcriptional activation of Xmrk in melanoma tissue and to contribute to an understanding of the role the regulatory locus R might play in suppressing the tumor phenotype in wild-type Xiphophorus, we performed functional analysis of the Xmrk oncogene promoter. Transient transfections in melanoma and nonmelanoma cells revealed the existence of a potent positive regulatory element positioned close to the transcriptional start site. Contained within this promoter segment is a GC-rich sequence identical to the binding site described for human Sp1. In vitro binding studies and biochemical characterizations demonstrated the existence of GC-binding proteins in fish that share immunological properties with members of the human Sp family of transcription factors and appear to be involved in the high transcriptional activation of the Xmrk oncogene. Since the identified cis element is functional in both melanoma and nonmelanoma cells, additional silencer elements suppressing Xmrk expression in nonpigment cells must exist, thereby suggesting a negative regulatory function for the genetically defined R locus.
Melanoma formation in the teleost fish Xiphophorus provides an in vivo model for studies on the genetic basis of tumor induction and the role of receptor tyrosine kinases in neoplastic transformation (1,2). In these animals tumors are caused by a single, dominantly acting gene locus, Tu. According to the model developed to explain melanoma formation in Xiphophorus (3), the tumor-inducing potential of the Tu locus is suppressed in wild-type fish by an unlinked locus, R, which is proposed to function as a tumor suppressor and is progressively eliminated upon crossing with a parental fish containing nei-ther of the two loci. The stepwise displacement of R genes in Xiphophorus hybrids is thought to allow expression of the Tu phenotype, thus leading to benign or malignant melanoma depending on the copy number of R still present in the hybrid genome. The crossing data could, however, be similarly interpreted by assuming that instead of elimination of suppressor genes the introduction of intensifier genes leads to Tu-dependent melanoma formation. So far no genetic evidence could be obtained to decide between both possibilities.
The Xmrk 1 oncogene, which is the critical constituent of the Tu locus, encodes a receptor tyrosine kinase with high similarity to the human epidermal growth factor receptor and is responsible for the hereditary malignant melanoma in Xiphophorus (4). Besides the oncogenic version of Xmrk, a second copy exists in the Xiphophorus genome. This proto-oncogenic version of Xmrk is present in all individuals of the genus examined so far (5) and is expressed at low levels in a variety of tissues. 2 Expression of the Xmrk proto-oncogene is not associated with the tumor phenotype; however, the oncogene transcript is highly overexpressed in melanoma, whereas no expression is detectable in normal tissues in Northern blots (4,7), and only very low level expression is seen in reverse transcriptase-PCRs (8). The amount of oncogene transcript found in the tumors is positively correlated with their malignancy (4,7). It thus appears that it is the overexpression of the Xmrk oncogene which is primarily responsible for melanoma formation in Xiphophorus hybrids (9,10).
Molecular genetic analyses revealed that both versions of Xmrk, proto-oncogene and oncogene, are highly identical in their coding region but differ significantly in their promoter regions (11). This situation is explained by a nonhomologous recombination event between the Xmrk proto-oncogene and another gene locus (designated D) (12), giving rise to the oncogene as an additional copy of Xmrk with altered 5Ј sequences. This upstream region contains TATA-and CAAT-like sequences and could thus represent a "non-housekeeping gene" promoter (13) in contrast to the GC-rich sequences driving transcription of closely related receptor tyrosine kinases like the human epidermal growth factor receptor (14) or the rat HER2/neu gene (15). It is suggestive that these newly acquired upstream sequences account for the observed Xmrk overexpression in Xiphophorus hybrids and that the R locus might be involved in their transcriptional regulation. Analysis of the Xmrk oncogene promoter and its transcriptional control elements is therefore important to obtain insight into the mech-anisms leading to the tissue-specific overexpression of the Xmrk oncogene. These analyses might further help to decide whether a silencing or an enhancing mode of action has to be proposed for the regulatory gene(s) encoded by the R locus.
Here we report on the functional characterization of Xmrk promoter sequences leading to the identification of a potent cis element and its corresponding transacting factors enhancing transcription of the Xmrk oncogene. A GC box positioned close to the transcriptional start site was found to play a major role for Xmrk promoter activity in both melanoma and nonmelanoma cells. Analyses of the corresponding transcription factors point toward the involvement of fish homologues of human Sp proteins in the transcriptional control of Xmrk. Since transcriptional activation is not restricted to melanoma cells, the existence of additional potent silencer elements down-regulating Xmrk expression in nonpigment cells has to be postulated, thus supporting the possibility that the R locus might exert a negative regulatory function on the Xmrk oncogene.

Enzymatic DNA Amplification of the Xmrk Oncogene Sd Allele
Promoter--To obtain sequences of the Xmrk promoter upstream of the previously isolated region (11), a PCR was performed on genomic DNA from Xiphophorus maculatus of the genotype SdDr/SrAr using a downstream primer from the Xmrk breakpoint region (DA 11) and an upstream primer (JA 8) derived from the D locus (12). PCR amplification was performed in a total volume of 50 l with 200 ng of genomic DNA as template and 2.5 units of Taq polymerase. The buffer conditions were 100 mM Tris-Cl, pH 9.0, 50 mM KCl, 1.5 mM MgCl 2 , 0.1% gelatin, 1% Triton X-100. After initial denaturation for 4 min at 92°C, amplification was performed for 35 cycles in a two-step PCR with 70°C as annealing/extension temperature. The primers used had the following sequences. DA 11, 5Ј-CCTTTCTGTCCGGGTCTGTGCTGCAGCAG-3Ј; JA 8, 5Ј-CTCGGATCCCTCAAGGCAGACTGG-3Ј.
The resulting 0.8-kilobase amplification product was specific for the Xmrk oncogene Sd allele. 3 To minimize the risk of cloning DNA stretches containing mutated nucleotides as a result of polymerase errors only a BamHI/EcoRI fragment upstream of the previously identified promoter region was inserted into Bluescript II KS(ϩ) to create pBSXmrkϪ675/Ϫ272, and four independent plasmid clones were sequenced.
Insertion of a BamHI/HindIII Xmrk promoter fragment containing residues Ϫ277 to ϩ34 into pBLCAT5/BamHI/HindIII resulted in XmrktkCATϪ277/ϩ34 rev. All other XmrktkCAT fusions were constructed by inserting various Xmrk promoter fragments into the blunted SalI site of pBLCAT5.
The plasmid ptkTATA CATII was constructed by replacing the BamHI/BglII fragment from pBLCAT5 (16) containing the tk promoter region from Ϫ105 to ϩ51 with the corresponding fragment from a linker scanning mutant ending at Ϫ32 (17).
DNA Sequencing and Sequence Analysis-Double-stranded DNA sequencing was performed using the Sequenase-kit (USB) and ␣-35 S-dATP according to the manufacturer's recommendations. Sequence assembly and comparison was done using the UWGCG program package. 4 Cell Lines and Culture-The embryonic epithelial cell line A2 (18) and the melanoma cell line PSM (19), both from Xiphophorus, were cultured under the conditions described (20).
Transfections-Cells were collected and resuspended to a density of 1 ϫ 10 7 cells/ml (PSM) or 2 ϫ 10 7 cells/ml (A2) in F-12 containing 5% fetal calf serum and placed on ice. 350 l of this cell suspension were mixed with 50 l of DNA (0.6 mg/ml in TE) directly in a 4-mm electroporation cuvette and subjected to electroporation at 250 V/1200 F using an Easyject Plus electroporator (EUROGENTEC, Belgium) in single pulse mode. After the pulse, the cells were placed on ice for 5-15 min before they were seeded onto 6-cm Primaria culture dishes (Falcon) in 5 ml of regular growth medium and cultured for 2 days before harvesting.
CAT Assays-Transfected cells were washed once with PBS, harvested in 1 ml of TEN (40 mM Tris-Cl, pH 7.5, 1 mM EDTA, 150 mM NaCl) using a rubber policeman, transferred to 1.5-ml Eppendorf tubes, and spun down for 5 min at 5000 rpm. Cells were resuspended in 100 -150 l of 0.25 M Tris-Cl, pH 7.8, and lysed by three cycles of freezing/thawing. Cellular debris was removed by centrifugation for 15 min at 14,000 rpm, and the supernatants were used for all following assays. Protein concentrations of the lysates were determined according to Bradford (21), and equal amounts of protein were used to measure CAT enzyme activity.
CAT assays were performed according to Seed and Sheen (22) with modifications described by Crabb and Dixon (23). Lysates were complemented up to 88 l with 0.25 M Tris-Cl, pH 7.8, and mixed with 10 l of butyryl-CoA (5 mg/ml), 1 l 0.5 M EDTA, and 1 l of 2 mM [ring-3,5-3 H]chloramphenicol (0.25 Ci/l in EtOH). The reaction was allowed to proceed for 2-16 h at 37°C and stopped by addition of 200 l of mixed xylenes. After vortexing and centrifugation, the organic phase was reextracted twice with 100 l of TE. Conversion rates were calculated after counting the butyrylated chloramphenicol in the organic phase and the nonreacted chloramphenicol in the aqueous phase.
Microinjection into Early Fish Embryos-Medaka embryos were injected cytoplasmically with approximately 500 pl (equivalent to 25 pg) of supercoiled plasmid DNA into one cell of a two-cell stage embryo and assayed as described previously (24).
Nuclear Extracts-Nuclear extracts of PSM cells were prepared as described by Dignam et al. (25) with minor modifications. Cells were grown to near confluency, washed, and collected in PBS and pelleted by centrifugation for 5 min at 1000 ϫ g. All the following steps were performed at 4°C. The cells were resuspended in 5 packed cell volumes of 10 mM HEPES-KOH, pH 7.8, 0.1 mM EDTA, 0.1 mM EGTA, 0.75 mM spermidine, 0.15 mM spermine, 1 mM dithiothreitol, 1 mM benzamidine, and 0.5 mM PMSF. After 10 min incubation on ice, the cells were lysed by homogenization in a Dounce homogenizer with 12 strokes using a "B" pestle. After addition of 1 ⁄10 volume of 10 ϫ PBS to stabilize the nuclei, the lysate was centrifuged in a SS-34 rotor for 8 min at 4000 rpm. The nuclear pellet was resuspended in 3 packed cell volumes 50 mM Tris-Cl, pH 7.5, 10% sucrose, 420 mM KCl, 5 mM MgCl 2 , 0.1 mM EDTA, 20% glycerol, 1 mM dithiothreitol, 1 mM benzamidine, and 0.5 mM PMSF. The lysate was gently stirred for 1 h at 4°C and then centrifuged for 1 h at 40,000 rpm. The supernatant was passed over a PD-10 desalting column (Pharmacia Biotech Inc.), equilibrated with 20 mM HEPES-KOH, pH 7.8, 12.5 mM MgCl 2 , 1 mM EDTA, 100 mM KCl, 0.1% Nonidet P-40, 20% glycerol, 1 mM dithiothreitol, 1 mM benzamidine, and 0.5 mM PMSF, and eluted in the same buffer. The extract was aliquoted, frozen in liquid nitrogen, and stored at Ϫ80°C.
Electrophoretic Mobility Shift Assay (EMSA)-The oligonucleotides used as probes were annealed and labeled at the 5Ј-ends using T4 polynucleotide kinase and [␥-32 P]ATP (Ͼ6000 Ci/mmol). The doublestranded probes were purified over a 15% polyacrylamide gel, eluted, precipitated with ethanol, and resuspended in TE buffer. Binding reactions were performed in a final volume of 20 l containing PSM nuclear extract, 2 g of poly(dI⅐dC)⅐poly(dI⅐dC), 2 g/l bovine serum albumin, 90 mM KCl, and specific competitor DNA as indicated. The binding buffer consisted of 10 mM HEPES-KOH, pH 7.8, 2.5 mM EDTA, 5 mM spermidine, 2% Ficoll 400, 6% glycerol, 1 mM dithiothreitol, and 0.5 mM PMSF. After a 15-min preincubation step at 4°C, 20,000 cpm (1-2 fmol) of labeled oligonucleotide was added, and incubation was continued for 15 min. For supershift assays 1 l of either preimmune, anti-Sp1, or anti-Sp3 antiserum (26) was added to the binding reaction 10 min prior to loading of the gel. DNA-protein complexes were resolved on 4 -6% polyacrylamide gels containing 5% glycerol in 0.25 ϫ Tris borate buffer (22.5 mM Tris base, 22.5 mM boric acid, 0.5 mM EDTA). The gels were run at 4°C at a constant current of 12 mA with buffer recirculation, then dried, and autoradiographed.
Copper-Phenanthroline Footprinting-According to Papavassiliou (27) an upscaled EMSA reaction was performed, using a DNA fragment as probe labeled exclusively at one of its 5Ј-ends. Following electro-phoresis the gel was immersed in 200 ml of 10 mM Tris-Cl, pH 8.0. After addition of 20 ml 2 mM 1,10-phenanthroline monohydrate, 0.45 mM CuSO 4 the chemical nuclease reaction was initiated by adding 20 ml of 58 mM 3-mercaptopropionic acid. The reaction was quenched after 7 min at 4°C by adding 20 ml of 28 mM 2,9-dimethyl-1,10-phenanthroline monohydrate. After 2 min incubation the gel was rinsed four times in distilled water and subsequently electroblotted on DE-81 membrane for 5 h at 500 mA in 0.5 ϫ TBE. The membrane was autoradiographed overnight, and bands corresponding to free and bound probe were cut out. The membranes were washed twice with 100 l of LS wash buffer (50 mM Tris-Cl, pH 8.0, 150 mM NaCl, 10 mM EDTA, pH 8.0) and then eluted twice with 100 l of HS wash buffer (50 mM Tris-Cl, pH 8.0, 1 M NaCl, 10 mM EDTA, pH 8.0) for recovery of DNA. Following sequential extractions with phenol/chloroform (1:1 (v/v)) and chloroform the DNA was precipitated with ethanol. After resuspension in 5 l of loading buffer (50% formamide, 0.05% bromphenol blue, 0.05% xylene cyanol), equal amounts (as determined by Cherenkov counting) were loaded onto a 12% 1 ϫ TBE, 50% urea polyacrylamide gel together with a G ϩ A sequencing ladder, which was prepared according to Maxam and Gilbert (28). The gel was run at 60 watts for 3-4 h, dried on Whatman 3MM paper, and autoradiographed.
Western Analysis-For Western analysis equal amounts of nuclear proteins were boiled for 5 min in Laemmli buffer and separated on SDS-polyacrylamide gel electrophoresis (10%). Using a transfer buffer containing 4 mM NaH 2 PO 4 and 57 mM Na 2 HPO 4 , the proteins were electroblotted to nitrocellulose membranes (Schleicher & Schü ll, Dassel, Germany) by standard procedures. Filters were blocked for 1 h at room temperature with NETG (50 mM Tris-Cl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 0.05% Triton X-100, 0.25% gelatin). For protein detection anti-Sp1 and anti-Sp3 (26) as well as anti-Sp1(PEP2) (Santa Cruz Biotechnology, Inc.) antisera and as a control preimmune serum were used as primary antibodies. After incubation with primary antibody overnight at 4°C and washing three times for 20 min at room temperature, the filters were incubated with horseradish peroxidase-coupled anti-mouse or anti-rabbit secondary antibodies for 1 h at room temperature and washed again as above. Nonradioactive detection (enhanced chemiluminescence, Amersham Corp.) was performed according to the supplier's recommendations. To reprobe Western blots with different antibodies, the membrane was incubated in STRIP buffer (62.5 mM Tris-Cl, pH 6.7, 2% SDS, and 100 mM 2-mercaptoethanol) for 1 h at 50°C. To remove any remaining SDS, filters were washed three times in PBS and then blocked again for 1 h with NETG.

Identification of Regulatory Elements within the Xmrk Promoter Proximal
Region-Sequences further upstream of the previously isolated promoter region of the Xmrk oncogene (11) were obtained by PCR using primers deduced from the D locus (12) and the Xmrk coding region. This segment containing 675 bp upstream of the transcriptional start site was inserted into a CAT reporter gene vector to yield XmrkCAT Ϫ675/ϩ34. Transient transfection of this promoter construct into two different Xiphophorus cell lines of melanoma (PSM) and nonmelanoma (A2) origin revealed similar high activity in both cell types (Fig.  1). The seemingly different transcriptional activity of Xmrk Ϫ675/ϩ34 observed upon comparison with a HSV tk promoter construct was only due to a differential transcriptional potential of the viral promoter in the two cell lines, which was demonstrated using a construct containing only the tk-derived TATA box as a reference. Thus, the existence of one or several strong, positive regulatory element(s) within the Xmrk Ϫ675/ ϩ34 region has to be proposed, driving transcription of the Xmrk oncogene in a nontissue-specific manner.
Gene transfer of the Xmrk Ϫ675/ϩ34 CAT fusion into early embryos of medaka fish (Oryzias latipes), a genus closely related to Xiphophorus, revealed that the investigated promoter fragment was not only active in tissue culture cells but was also functional in vivo in a whole animal system (Fig. 2). CAT activity observed in embryos 3 days after injection with Xmrk Ϫ675/ϩ34 CAT was on average about 50% of CAT expression driven by one of the strongest enhancer-promoter combinations in this in vivo assay, CMV-tk, indicating the high transcriptional activity of the Xmrk regulatory sequences. A promoter-less CAT construct sharing the plasmid backbone with Xmrk Ϫ675/ϩ34 CAT led to expression at background level comparable with that of noninjected embryos.
To determine the position of the regulatory elements responsible for this high level activation, the reporter gene activity of a set of 5Ј deletions spanning the Ϫ675/ϩ34 segment of the Xmrk upstream region was quantitated in the melanoma cell line PSM. Whereas deletion to Ϫ194 led to a reduction in reporter activity to about 60% of Ϫ675/ϩ34 (Fig. 3A), an even larger decrease in promoter activity to approximately 20% of Ϫ675/ϩ34 was observed in a construct lacking sequences upstream of Ϫ49. These results indicate the presence of probably two positive cis regulatory elements positioned in the promoter region downstream of Ϫ277 and Ϫ125, respectively. For a more precise localization of the regulatory elements various Xmrk promoter fragments were then fused to the tk promoter and Reporter constructs containing the CAT gene without promoter (pBLCAT6, no promoter) and under the control of Xmrk Ϫ675/ϩ34 (Xmrk) or a combination of the human cytomegalovirus enhancer with the HSV tk promoter (CMV-tk), respectively, were microinjected cytoplasmically into one cell of two-cell stage medaka embryos. Three days after injection extracts from individual embryos were prepared and assayed for CAT enzymatic activity. Each bar represents CAT expression in one single embryo, which was normalized to background expression observed in noninjected control embryos (not inj.). The inset is a blow-up to demonstrate expression obtained with the promoterless reporter gene plasmid. tested for their transcriptional activation potential (Fig. 3B). Since fragments Ϫ277/ϩ34 rev and Ϫ194/Ϫ20 did not differ significantly regarding their CAT expression, obviously no enhancer element was detectable in this assay within the region between Ϫ277 and Ϫ194. However, a fragment containing the region between Ϫ67 and Ϫ20 proved to be sufficient to activate a heterologous promoter to a similar level as all longer Xmrk fragments demonstrating the presence of functional activating elements within these 48 bp.
DNA-Protein Interactions with Xmrk Ϫ67/Ϫ20 -To test whether the in vivo observed positive regulatory activity within Ϫ67 and Ϫ20 was also reflected by in vitro DNA-protein interactions, EMSA was performed using an 88-bp HindIII/BamHI fragment of XmrktkCAT Ϫ67/Ϫ20 as probe and nuclear extract of PSM cells (Fig. 4A). One major complex was formed with this fragment. To precisely localize the position of the complex we performed copper-phenanthroline footprint analysis applying the same conditions as in the previous EMSA. The complex produces one clear footprint positioned between Ϫ50 and Ϫ41 with a hypersensitive site at position Ϫ46 (Fig. 4B). Interestingly, the protected region coincides with a GC-rich region within positions Ϫ48 and Ϫ41 of the Xmrk promoter that has the sequence 5Ј-CCCGCCCC-3Ј, thus resembling the previously determined Sp1 consensus binding motif (29). These data suggest that the complex is formed over the Sp1 consensus within Ϫ67/Ϫ20, implying the possible involvement of DNAbinding proteins related to Sp1 in complex formation.
To test further whether Sp1-like proteins were involved in complex formation, EMSA was performed using an oligonucleotide as probe, which was derived from the Xmrk sequence between Ϫ67 and Ϫ32 (Xmrk Ϫ67/Ϫ32) (Fig. 5A), in order to avoid undesired side effects from the endogenous Xmrk TATA box positioned within residues Ϫ34 to Ϫ27. High resolution gel analysis of the DNA-protein complex revealed two bands of slightly different mobility (C, CЈ) (Fig. 5B). Both complexes proved to be specific since unlabeled Xmrk Ϫ67/Ϫ32 inhibited complex formation, whereas an oligonucleotide mutated in two positions within the GC box (Xmrk Ϫ67/Ϫ32 mut) did not, suggesting the involvement of Sp1-related transcription factors in the complex. This observation was confirmed by using an Sp1 consensus oligonucleotide as competitor (30) sharing only the Sp1 core sequence with Xmrk Ϫ67/Ϫ32, which also inhib- FIG. 4. Binding of PSM nuclear components to Xmrk ؊67/؊20. A, EMSA with an 88-bp HindIII/BamHI fragment of XmrktkCAT Ϫ67/ Ϫ20 containing the Xmrk Ϫ67/Ϫ20 promoter region and PSM nuclear extracts. B, copper-phenanthroline footprint with Xmrk Ϫ67/Ϫ20. The EMSA reaction depicted in A was scaled up; the EMSA gel was subjected to the chemical nuclease reaction and the products were separated on a 12% sequencing gel. Lanes containing DNA from free probe (Ϫ) and complex (ϩ) as well as the products of a purine-specific cleavage reaction (G ϩ A) are indicated. Bp positions relative to the transcriptional start site are shown on the right, and the black bar indicates the position of the GC box. The hypersensitive site observed within the protected region is marked with an arrow.
ited formation of complexes C and CЈ. The unrelated binding consensus of transcription factor AP1 (31) as competitor showed no effect. Accordingly, no protein binding was detectable in an EMSA using the Xmrk Ϫ67/Ϫ32 mutant as a probe (Fig. 5C).
Recent studies indicate that the zinc finger protein Sp1 is one of several members of a differentially expressed gene family (32,33). In order to determine which of their corresponding fish homologues might bind to the Xmrk oncogene promoter, we performed supershift analyses with different antibodies raised against human Sp proteins. Whereas addition of a polyclonal anti-Sp1 antiserum as well as an antibody raised against residues 520 -538 of the human Sp1 protein (data not shown) left both complexes unaltered, addition of an anti-Sp3 antiserum led to loss of complex C and a strong reduction of CЈ, indicating the presence of a fish Sp3 homologue in the complexes (Fig. 6). Addition of an antiserum against human Sp4, a nuclear factor with a very limited expression pattern (32), as well as preimmune serum as control also showed no effect on complex formation (data not shown).
Sp-related Proteins in Xiphophorus-In various studies Sp1 has been reported to be ubiquitously expressed (29), and therefore its lack of detectability in the supershift assay was unexpected. The inability of both anti-Sp1 antisera to reduce or supershift complexes C and CЈ could either result from the absence of Sp1 in the nuclear extracts or a lack of crossreactivity of the antibodies toward fish Sp1 under EMSA conditions. To test this we conducted Western analyses of nuclear extracts prepared from PSM cells using different antibodies for detection (Fig. 7). An antibody raised against a peptide outside the zinc finger region of human Sp1 (␣Sp1(PEP2)), which identifies the human Sp1 95-and 106-kDa polypeptide species, recognized one major protein of about 80 kDa in melanoma cell nuclear extracts. A polyclonal serum against full-length Sp1 (␣Sp1) also detected a protein of approximately 80 kDa and in addition species of 85 and 75 kDa molecular mass. Immunodetection using anti-Sp3 antiserum also identified a prominent band of about 85 kDa. No such bands were observed using preimmune sera in control experiments. These results demonstrate the existence of fish homologues of human Sp1 and Sp3 in PSM cells. Hence, it seems likely that the inability of anti-Sp1 antibodies to supershift the complex formed over Xmrk Ϫ67/Ϫ32 was rather due to a lack of cross-reactivity under EMSA conditions than to the general absence of Sp1 in the fish cells.

FIG. 5. Detection of GC-binding activity in PSM nuclear extracts.
A, upper strand sequences of oligonucleotides used in EMSAs presented in B and C. Substituted bases in Xmrk Ϫ67/Ϫ32 mut are indicated by asterisks. The GC-rich sequence shared between Xmrk Ϫ67/Ϫ32 and Sp1 consensus oligonucleotide (Sp1) is boxed. B, specificity of protein binding to Xmrk Ϫ67/Ϫ32. PSM nuclear extracts were incubated with radiolabeled Xmrk Ϫ67/Ϫ32 in the absence (Ϫ) or presence of increasing amounts (20-and 200-fold molar excess) of unlabeled double-stranded competitors Xmrk Ϫ67/Ϫ32 (wt), Xmrk Ϫ67/Ϫ32 mut (mut), Sp1, and an unrelated oligonucleotide comprising an AP1 binding site (AP1), respectively. The two complexes forming over Xmrk Ϫ67/Ϫ32 (C, CЈ) are indicated by arrows; the control lane contains no protein in the EMSA reaction. C, protein binding to wild-type and mutated Xmrk GC sequences. Radiolabeled oligonucleotides containing the Xmrk Ϫ67/Ϫ32 region (wt) or the corresponding mutant (mut) altered within the GC box were used in an EMSA with PSM nuclear extracts.
FIG. 6. EMSA detection of Sp-related factors in PSM nuclear extracts. Nuclear extracts from PSM cells were incubated with radiolabeled Xmrk Ϫ67/Ϫ32. One microliter of serum against human Sp1 (␣Sp1), serum against Sp3 (␣Sp3), or a mixture of sera against Sp1 and Sp3 (␣Sp1 ϩ ␣Sp3) was included in the binding reaction as indicated. Specific complexes C and CЈ are indicated on the right, a nonspecific complex sometimes seen under certain experimental conditions is marked by an asterisk. The control lane contains no protein in the EMSA reaction.

FIG. 7. Western detection of SP proteins in PSM cells.
Nuclear extracts from PSM cells were fractionated on a 10% SDS-polyacrylamide gel and immobilized on nitrocellulose filter. The filters were incubated with antisera against a peptide containing residues 520 to 538 of human Sp1 (␣Sp1 PEP2), against full-length Sp1 (␣Sp1), or human Sp3 (␣Sp3).

Functional Analysis of the GC Box within Xmrk
Ϫ67/ Ϫ32-To determine whether the GC box constitutes an important functional cis element within the Xmrk promoter, the region between Ϫ67 and Ϫ32 was analyzed in more detail in a set of transient transfections in PSM cells (Fig. 8). Oligonucleotide Xmrk Ϫ67/Ϫ32 and its corresponding mutant carrying two nucleotide exchanges within the GC box (Xmrk Ϫ67/Ϫ32 mut) were inserted directly upstream of the tk TATA box and evaluated functionally in this heterologous promoter context. To avoid interference with the endogenous Sp1 site within the HSV tk promoter, the tk-derived TATA box was used. Xmrk Ϫ67/Ϫ20 and Xmrk Ϫ67/Ϫ32 in either orientation stimulated CAT activity to comparable levels, demonstrating that these sequences are sufficient to enhance reporter gene expression. Mutation of two base pairs within the GC box (Xmrk Ϫ67/Ϫ32 mut) resulted in a complete loss of activity. In accordance with these findings an isolated Sp1 site (30) sharing only an 8-bp core sequence with Xmrk Ϫ67/Ϫ32 (Fig. 5A) was able to drive CAT expression to a level which did not differ significantly from the Xmrk promoter fragment. These results clearly indicate that the GC box is the functional element within Xmrk Ϫ67/Ϫ32 and therefore critical for the activity of the Xmrk oncogene promoter.

DISCUSSION
Overproduction of the Xmrk-encoded receptor tyrosine kinase has been shown to be responsible for melanoma formation in Xiphophorus hybrids (7,11). Resulting from high steady state levels of the corresponding mRNA, this elevated expression is most likely due to a transcriptional deregulation of the Xmrk promoter in tumor cells. In order to contribute to an understanding of the mechanisms leading to tumor formation in Xiphophorus, it was important to identify regulatory elements that account for this high promoter activity. Preliminary experiments had revealed that a fragment comprising 277 bp upstream of the transcriptional start site of the Xmrk oncogene was able to drive the expression of a reporter gene (11). In this report we have isolated sequences further upstream of the originally described promoter and identified a potent cis element and its corresponding transacting factors enhancing tran-scription of the Xmrk oncogene. An Xmrk promoter fragment spanning the region Ϫ675/ϩ34 was shown to be highly active in two different Xiphophorus cell lines of melanoma and nonmelanoma origin. In addition, the analyzed upstream sequences proved not only to be functional in tissue culture cells but also in fish embryos after transient gene transfer. Comparison with CMVtkCAT demonstrated the high level activation potential of the Xmrk promoter in vivo in a developing embryo. The variability of CAT activity observed in individual embryos could be explained by the mosaic distribution of the injected plasmid (10) leading to CAT expression only in a subset of tissues exhibiting responsiveness to the Xmrk promoter.
Whereas deletion analysis suggested the presence of two distinct positive regulatory elements downstream of positions Ϫ277 and Ϫ125, respectively, subsequent evaluation of various promoter fragments in a heterologous promoter context revealed the existence of only one potent positive regulatory element. The failure to detect enhancer activity within the more distal region in this type of analysis could be due to the requirement for an authentic Xmrk promoter context of this putative regulatory element to function. Such specific enhancer-promoter interactions have been described for several other genes and may represent a mechanism ensuring transcriptional specificity (34,35). Within the proximal region, however, a 48-bp segment adjacent to the Xmrk TATA box proved to be able to confer strong transcriptional activity to a heterologous promoter.
By copper-orthophenanthroline footprinting analysis the DNA-protein complex formed over this region was shown to be positioned over a core sequence, 5Ј-CCCGCCCC-3Ј, that is identical to the binding motif described for the zinc finger protein Sp1 (29). Mobility shift assays substantiated the finding that GC-binding proteins interacting specifically with the Xmrk GC box are present in Xiphophorus nuclear extracts. The cross-competition observed between the Xmrk-derived binding site and the Sp1 consensus oligonucleotide, which are unrelated outside the core region, clearly demonstrates the crucial role of the GC box for the protein-DNA interaction. This is supported by the complete loss of protein binding upon mutation of two base pairs within the core consensus. In addition to the members of the Sp protein family (32,33), other GC box binding proteins have been identified. Whereas the BTE binding factors seem to recognize a variant sequence motif deviating from the Sp1 consensus site analyzed here (36 -38), the only other proteins described so far binding to this consensus are the yeast MIG1 repressor (39) and its fungal homologue CREA (40). MIG1 interacts specifically with the pentanucleotide motif 5Ј-GCGGG-3Ј; however, it requires flanking AT-rich sequences adopting a particular geometry for high affinity binding (41), making it unlikely that a factor of this nature binds to the Xmrk GC box.
Supporting evidence for the identity of the GC box binding factors in PSM cells is provided by the Western blot experiments indicating the presence of proteins that share structural features with members of the mammalian Sp family of transcription factors even outside the highly conserved zinc finger region. Consistent with these findings are supershift assays in which addition of an anti-Sp3 antiserum led to reduced formation of the respective DNA-protein complexes, indicating the involvement of different forms of a fish Sp3 homologue in complexes C and CЈ, which is in accordance with observations made in mammalian cells (26). Anti-Sp1 antiserum was, in contrast, not able to exhibit an effect in this type of assay, although Sp1-related proteins are present in Xiphophorus. It is conceivable, however, that an antibody may only under certain conditions be able to cross-react with its corresponding polypeptide, in particular when crossing species border. It is likely that the lack of immunoreactivity of anti-Sp1 antiserum in the supershift experiments was rather due to its failure to cross-react with the corresponding fish protein in this specific type of experiment than to the general absence of Sp1 in PSM cell extracts.
In summary, it can be assumed that homologues of human Sp proteins, namely Sp1 and Sp3, are present in Xiphophorus and that at least one of them is able to bind to the GC box contained within the Xmrk promoter. To this date homologues of human Sp proteins have only been described for mouse and rat (37,42), and several distant relatives have been isolated from Drosophila (6,43,44). Whereas all known mammalian Sp1 proteins are nearly identical to human Sp1 in their complete primary sequence, the respective Drosophila proteins show fairly low overall similarity except for their zinc finger region. These proteins are functionally related to Sp1 (44) but may not necessarily represent the Drosophila Sp1 orthologue. To our knowledge, we report here for the first time that at least some members of the Sp family may be highly conserved in a nonmammalian vertebrate. Notably, this conservation seems to extend even outside the DNA-binding domain, as is demonstrated by immunoreactivity toward an antibody raised against a region outside the zinc finger domain of human Sp1. The size difference between fish and human Sp proteins observed hereby is not unusual considering the evolutionary distance between these two species.
It cannot be ruled out that there might even be a larger number of GC-binding proteins present in fish that were not detected under our experimental conditions. This speculation is supported by the inability of either antibody used to completely prevent complex formation in the supershift assays even at high concentrations.
Transient transfection analyses of wild-type and mutant oligonucleotides containing residues Ϫ67 to Ϫ32 of the Xmrk promoter demonstrated that the GC box is not only able to interact with nuclear proteins in vitro but is also functional in an in vivo system. Since mutation within the GC box completely abolished the transcriptional activation potential of Ϫ67/Ϫ32, it seems to be the only functionally relevant element within this region. Our data provide evidence that Sp-like transcription factors bind to this sequence within the Xmrk promoter in vitro and might thus play a role in the observed transcriptional activation. However, we cannot exclude the possibility that other so far uncharacterized GC box binding proteins are involved, for which a similar DNA target recognition would have to be postulated.
It is conceivable that other, possibly closely positioned, positive and negative elements went undetected by our set of deletions. Based upon the observation that, when compared with the tk promoter, Xmrk Ϫ675/ϩ34 activated CAT expression in PSM cells by about 16-fold, but fragment Ϫ67/Ϫ20 enhanced transcription of a heterologous promoter only 5-7fold, it can be assumed that the coordinate action of a greater number of regulatory elements outside the region investigated in detail is required for full promoter activity. It will be interesting in the future to determine the quantitative effect of the GC box on the overall promoter activity, e.g. by mutagenizing this sequence within the Xmrk Ϫ675/ϩ34 construct. This, however, will only be fully informative if other functional elements present upstream of the Xmrk gene have been characterized.
The notion that high transcriptional activity of the Xmrk promoter accounts for the overexpression of the Xmrk oncogene in melanoma tissue is strongly supported by the presence of a potent positive regulatory element identified within the Xmrk Ϫ675/ϩ34 promoter region. Since this element is not only functional in transformed but also nontransformed cells that are normally not susceptible to melanoma development, the existence of additional silencer elements positioned outside the investigated region leading to transcriptional inactivation of Xmrk in nonpigment cells has to be postulated. It is conceivable that the genetically defined R locus, suppressing the melanoma inducing potential of the Xmrk oncogene in wild-type Xiphophorus, might act through such regulatory elements. We suggest that subscribers photocopy these corrections and insert the photocopies at the appropriate places where the article to be corrected originally appeared. Authors are urged to introduce these corrections into any reprints they distribute. Secondary (abstract) services are urged to carry notice of these corrections as prominently as they carried the original abstracts.