Structure and Promoter Analysis of the Gene Encoding the Human Melanoma-inhibiting Protein MIA*

We have recently described the isolation of a novel protein, MIA, which is secreted from malignant melanoma cells and elicits growth inhibition on melanoma cells in vitro (Blesch, A., Bosserhoff, A. K., Apfel, R., Behl, C., Hessdo¨rfer, B., Schmitt, A., Jachimczak, P., Lotts- peich, F., Schlingensiepen, H., Buettner, R., and Bogdahn, U. (1994) Cancer Res. 54, 5695–5701). Here, we re- port the structure of the human MIA gene locus, describe its expression pattern in melanocytic tumors in vivo , and provide an initial characterization of the MIA promoter. The MIA gene is encoded by four exons, and the mRNA initiation site was identified 70 base pairs upstream from the translation start codon. MIA mRNA expression in vivo correlated with progressive malig-nancy of melanocytic lesions and was inducible in other cells by phorbol esters. To investigate mechanisms mediating this melanoma-associated expression pattern, we analyzed the promoter activity of the 1.3-kilobase genomic sequences located 5 (cid:42) -upstream of the MIA gene. The MIA promoter conferred high levels of gene activa- tion specifically in human and murine melanoma cells, and its activity was further enhanced by treatment with phorbol esters. Site-directed mutation of an NF-kB site within the same mu- tation of the MIA -NF-kB site that was generated for the gel shift competition experiment was also introduced into the CAT3 plasmid containing the full 1361-bp MIA promoter using a site-directed mutagenesis kit (Clontech) Nickoloff, 1992).

Growth and expansion of tumor cells including malignant melanomas are modulated by a complex network of growth factors, which regulate proliferation and cell-matrix interaction through a variety of different signal transduction pathways. Therefore, the net growth or regression of melanomas in vivo reflects integration of many different stimulatory or inhibitory factors produced both by the tumors cells and their environment. Well studied examples of growth regulatory proteins in melanoma cells include members of the tumor growth factor-␤ and platelet-derived growth factor families, transferrin, basic fibroblast growth factor, epidermal growth factor, and tumor growth factor-␣ (Herlyn and Malkowicz, 1991;Halaban et al., 1991;Rodeck et al., 1991;Shih and Herlyn, 1993).
We have recently isolated and cloned a novel protein that is secreted by malignant melanoma cell lines and exerts autologous growth inhibitory effects on melanoma cells in vitro (Bogdahn et al., 1989;Apfel et al., 1993;Blesch et al., 1994). Due to the growth inhibitory effect, which allowed purification by means of a bioassay, this protein was designated MIA (melanoma inhibitory activity). Isolation of fully encoding human and murine MIA cDNA clones revealed that MIA is translated as a 131-amino acid precursor protein and secreted into the tissue culture supernatant of melanoma cells after cleavage of a 24-amino acid signal peptide. MIA appears to constitute a unique protein since no significant sequence homology to any other known protein was detected. Initial characterization of MIA expression by Northern blot analyses indicated that MIA is expressed in all melanoma cell lines that we tested and infrequently in glioma cell lines but not in fibroblast or epithelial cell lines. This interesting melanoma-associated expression pattern prompted us to examine in more detail skin biopsies along with benign and malignant melanocytic tumors in vivo for expression of MIA mRNA. We further aimed to isolate the entire genomic locus of the human MIA gene, to determine its exon-intron organization, and to provide an initial characterization of the melanoma cell typespecific function of the MIA gene promoter.

EXPERIMENTAL PROCEDURES
Isolation of the Human MIA Gene-A commercially available human placental genomic library in the phage FixII (Stratagene) was screened using the fully encoding human MIA cDNA (Blesch et al., 1994) as a probe. Two positive clones were obtained from 6 ϫ 10 6 phages, one of which was used for restriction analysis of the MIA gene locus (phage UB1 shown in Fig. 1). Two adjacent XbaI fragments (UB1-1 and UB1-3) hybridizing with the MIA cDNA probe were subcloned into the plasmid pBluescript.
Primer Extension Analysis and Rapid Amplification of 5Ј-cDNA Ends-Polymerase Chain Reaction-A primer extension experiment was performed following standard procedures (Sambrook et al., 1989) using a phospholabeled oligonucleotide (MIA extension, 5Ј-CAAGGGGGT-GCTGGGTCTCCAATTT-3Ј) complementary to nucleotides Ϫ26 to Ϫ51 of the MIA cDNA and 10 g of total RNA isolated from the melanoma cell line Mel Im (Jacob et al., 1995). RACE-PCR 1 was performed using the AmpliFinder RACE kit (Clontech) precisely as described previously (Bauer et al., 1994). Briefly, an antisense primer (5Ј-CAGCCATGGAGATAGGGT-3Ј) matching residues ϩ57 to ϩ75 of the MIA cDNA was used for reverse transcription of 2 g of poly(A) ϩ -selected RNA isolated from Mel Im melanoma cells. After hydrolysis of the template mRNA, an anchor primer was ligated to the 3Ј-end of the cDNA, and then a PCR reaction was performed using the anchor primer and the same oligonucleotide that was used for the primer extension analysis as a nested MIA-specific primer.
Reverse Transcriptase-mediated Polymerase Chain Reaction (RT-* This work was supported by grants from the Deutsche Forschungsgemeinschaft (to R. H., and R. B.) and from the Deutsche Krebshilfe and Boehringer-Mannheim (to U. B., and R. B.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EMBL Data Bank with accession number(s) X84707.
The genomic region 5Ј adjacent to the MIA gene from residue Ϫ1361 to Ϫ1 was amplified by PCR, inserted into the promoterless chloramphenicol acetyltransferase plasmid pBLCAT3 (Luckow and Schü tz, 1987), resequenced, and then a series of 5Ј-deleted constructs was generated using nested deletion (Henikoff, 1984). 2 ϫ 10 5 cells were seeded into 90-mm dishes and transiently transfected with 5 g of plasmid DNA using DOTAP transfection reagent (Boehringer) the following day. To normalize transfection efficiency, 2.5 g of an LTR-lacZ plasmid was cotransfected and an enzyme-linked immunosorbent assay (Boehringer) was used to quantify CAT activities.
Gel Mobility Shift Assays and Site-directed Mutagenesis-Two complementary synthetic oligonucleotides spanning the NF-kB site at position Ϫ207 to Ϫ198 in the MIA promoter (5Ј-ACCTTATCTGG-GAATTTCCTTGGGCCTTAC-3Ј) were hybridized and phospholabeled. Nuclear extracts were prepared from B16 melanoma cells, and gel shifts were performed as described previously (Buettner et al., 1993). Competition experiments were performed using the same MIA-NF-kB binding site and further a 25-fold excess of an unrelated oligonucleotide ("170oligo" matching residues Ϫ195 to Ϫ160 of the MIA promoter), a mutated binding site (5Ј-ACCTTATCTGAAGCTTTCCTTGGGCCTTAC-3Ј), or a perfect bona fide NF-kB site from the HIV-1 LTR (5Ј-ATCAGG-GACTTTCCGCTGGGACTTTCCG-3) (Wu et al., 1988). The same mutation of the MIA-NF-kB site that was generated for the gel shift competition experiment was also introduced into the CAT3 plasmid containing the full 1361-bp MIA promoter using a site-directed mutagenesis kit (Clontech) (Deng and Nickoloff, 1992).

RESULTS
Structure of the Human MIA Gene-To isolate the genomic MIA locus, a FixII human placental DNA library was screened using the entire human cDNA (Blesch et al., 1994) as a probe. Two positive recombinant phages were identified, one of which was referred to as UB-1 and subjected to restriction and Southern blot analyses. The insert of phage UB-1 contained five different XbaI fragments spanning approximately 15 kb of genomic DNA. Two of these five XbaI fragments (UB1-1, 2.2-kb in size and UB1-3, 1.4 kb in size) hybridized to the MIA cDNA probe and were subcloned into the plasmid pBluescript and then fully sequenced on both strands. Comparison of the genomic and cDNA sequences revealed that the MIA gene consists of four small exons interrupted by three introns ( Fig. 1) with consensus splice sequences at the intron-exon junctions. Further analysis by a PCR reaction confirmed that the two XbaI fragments were located adjacently in the genomic DNA. Thus, the entire locus encoding the MIA protein is encompassed within a small region of approximately 2 kb.
Determination of the Transcription Start Site-Three different methods were employed to identify the transcription initiation site. First, a series of MIA cDNA clones obtained from a HTZ-19 melanoma cell library was sequenced, and the one clone extending 5Ј the most started at position Ϫ70 with respect to the ATG protein start codon (data not shown). Further, a primer extension assay was performed on RNA isolated from Mel Im melanoma cells using an oligonucleotide complementary to MIA cDNA upstream of the protein start site. Analysis of the extension products on a sequencing gel ( Fig. 2A) revealed a specific band of a size between 52 and 54 bp, suggesting that the 5Ј terminus of the MIA mRNA is at residue Ϫ70. These results were then confirmed by RACE-PCR (Frohman et al., 1989), resulting in a specific product of approximately 70 bp (Fig. 2, B and C). The amplification product was subcloned into the plasmid pCRScript, sequenced, and thereby shown to start at nucleic acid residue Ϫ70. The location of the primers and the expected sizes of the reaction products are graphically summarized in Fig. 2C. In conclusion, all three results are consistent with an mRNA initiation site at nucleic acid residue Ϫ70 upstream from the translation start codon.
Expression Pattern of MIA in Melanoma Cell Lines, Skin, and Melanocytic Tumors-We have recently described the molecular cloning of the MIA cDNA and have shown by Northern blot analyses that MIA mRNA is widely expressed in malignant melanoma cell lines in vitro, including HTZ-19 and Mel Im cells (Blesch et al., 1994). Therefore, we have used RNA isolated from these melanoma cell lines to establish conditions for RT-PCR to determine the MIA expression pattern in a series of cell lines and in benign and malignant melanocytic lesions in vivo.
As shown in Fig. 3A, specific MIA RT-PCR products were readily amplified from both melanoma cell lines but not from two different benign melanocyte cultures. To control the specificity of these reactions, we subcloned the PCR products into a plasmid vector and confirmed by sequencing that they represented MIA cDNA fragments (data not shown). In parallel ␤-actin mRNA was amplified to verify equal amounts and integrity of different RNA preparations.
We then used the same PCR conditions to perform PCR reactions on RNA samples from a series of different cell lines. As summarized in Table I and in good agreement with previously performed Northern blot analyses (Blesch et al., 1994), we detected high levels of MIA mRNA expression in every melanoma cell line we tested, including B16 murine melanoma cells. In contrast, we did not detect any significant expression in other skin-derived cells including normal fibroblasts and HaCaT keratinocytes (Fig. 3B). Also, other epithelial cell lines such as COS, HeLa, and HepG2 cells, DU 145 (human prostate cancer) and J82 cells (human bladder cancer), or PA-1 teratocarcinoma cells did not express MIA. Interestingly, significant MIA expression was induced by treatment with phorbol esters in skin fibroblasts, HaCaT, COS, and HeLa cells.
These results prompted us to study the expression pattern of MIA mRNA in normal skin and skin-derived melanocytic tu- mors. As shown in Table II and as examples in Fig. 4A, we did not detect MIA mRNA in normal skin except for two cases, in which minute mRNA levels were amplified when 32 rather than 25 PCR cycles were performed. Low or moderate MIA mRNA levels were detected in 8 of 15 benign melanocytic nevi, and high levels were found in one case. In all specimens taken from primary malignant melanomas (7 cases) and from lymph node metastasis of malignant melanomas (3 cases), abundant MIA transcripts were amplified. From all of these specimens, ␤-actin cDNA was coamplified to control for equivalence and integrity of RNA preparations. In summary, we detected high levels of MIA mRNA in all malignant melanoma biopsies and cell lines, low or moderate MIA mRNA levels in most benign melanocytic nevi, and very low or no MIA mRNA in nonneoplastic skin biopsies, melanocytes, fibroblasts, and keratinocytes. In addition, we did not detect any MIA mRNA in a panel of normal mouse tissues including skin, spleen, brain, thymus, kidney, intestine, lung, and skeletal muscle (Fig. 4B).
MIA Promoter Analysis-A computer analysis of the genomic sequence located 5Ј adjacent to the MIA mRNA start revealed very few canonical sequences as putative binding sites for transcription factors. As annotated in Fig. 5, consensus SP-1 (Jones and Tjian, 1985), NF-kB (Lenardo and Baltimore, 1989), and CTF/NF-1 (Jones et al., 1989) binding sites are located 35, 130, and 555 bases upstream from the mRNA start site, respectively. No TATA box or any other known cis-regulatory motif was detected in the sequence.
To determine whether the MIA promoter is activated specifically in melanoma cells, we cloned the 1386-bp fragment shown in Fig. 5 in front of a promoterless CAT plasmid and tested its activity in several human melanoma cell lines in comparison to non-melanocytic cancer cells. We found that the MIA promoter confers high levels of gene expression specifically in human or murine melanoma cell lines but not in HeLa, HepG2, PA-1, and COS cells (Fig. 6, A and B). To map in more detail cis-regulatory elements mediating MIA mRNA expression in melanoma cells, we transfected a series of 5Ј-deleted CAT reporters both into B16 and COS cells. B16 cells were chosen for this experiment because they were transfected much more efficiently, and therefore small changes in promoter activity could be monitored reliably. Fig. 6B gives a summary of the promoter constructs and CAT activities obtained from transiently transfected B16 and COS cell cultures. Maximal CAT activity was observed when a promoter fragment ranging from Ϫ493 to Ϫ1 with respect to the ATG protein start codon was used. This promoter fragment conferred 14-fold activation to FIG. 2. Determination of the MIA mRNA initiation site. A, primer extension assay. A 25-mer synthetic oligonucleotide was phospholabeled and used to extend the N-terminal end of the MIA mRNA. Extended products were size-fractionated on a 5% urea/polyacrylamide gel next to a sequencing reaction as a size marker. The largest extended product is 54 bp in length matching nucleic acid residue Ϫ70 relative to the ATG protein start codon. B, direct cloning of the N-terminal MIA cDNA end by RACE-PCR. First strand cDNA was synthesized from Mel Im poly(A) ϩ RNA using an antisense primer in the first exon. An anchor primer was ligated to the 3Ј-cDNA end, and the resulting template was amplified by PCR using the anchor primer and a nested MIA primer. Shown is an ethidium-stained agarose gel of the PCR product next to a molecular size standard. C, graphic summary indicating relative location of RACE-PCR primers and the size of the expected PCR product.  Significant changes in CAT activities were observed when a series of fragments extending further 5Ј-upstream was analyzed, indicating that silencer and enhancer elements are located between residues Ϫ1200 and Ϫ761 and Ϫ761 and Ϫ493. Further deletion of the promoter to Ϫ212 decreased significantly CAT activities, and a promoter fragment starting at Ϫ170 was entirely inactive as compared with the promoterless pBLCAT3 plasmid. In summary, these CAT assays led to the conclusion that residues located between Ϫ493 and Ϫ1 in the MIA promoter are necessary and sufficient to mediate high levels of cell typespecific gene expression. Activation of the MIA Promoter Involves Binding of NF-kB-Since a consensus NF-kB binding site at residue Ϫ207 was detected within the most active promoter fragment, we explored further the functional role of NF-kB in regulating MIA expression. Gel shift analyses were performed to investigate whether NF-kB binds to this site in melanoma cells. When we used a synthetic 30-mer binding site spanning residues Ϫ216 to Ϫ187 including the NF-kB site, one specific bandshift was observed (Fig. 7A). Binding was not competed when a similar binding site mutated at four critical residues in the NF-kB core sequence or an unrelated sequence of the MIA promoter (Ϫ170 oligonucleotide) were used in contrast to a synthetic binding site matching the NF-kB element in the HIV-1 LTR, which competed specifically the MIA-NF-kB bandshift activity.
NF-kB activity results from a gene family that is expressed in a large number of different cell types and tissues, and consequently the MIA-NF-kB site was also shifted when COS cell extracts were used (data not shown). To address whether binding of the NF-kB site is necessary for cell type-specific function of the MIA promoter, we introduced the same four bases that abolished binding of NF-kB in gel shift experiments by sitedirected mutagenesis in the CAT reporter plasmid under the control of the full 1361-bp MIA promoter. This promoter con-struct mutated at the NF-kB site was transiently transfected into B16 melanoma cells in parallel with the wild-type promoter construct. As shown in Fig. 7B, we observed approximately 2-fold decreased activity in comparison to the wild-type promoter, whereas the mutation did not affect significantly the stimulating effect of PMA on CAT expression. These results indicate that the NF-kB contributes to the MIA promoter activity in melanoma cells but is dispensable for stimulation in response to PMA.
To test whether the induction by PMA represents a primary response or a late event, we determined the time course of mRNA induction and promoter activation. RT-PCR analyses of HeLa and COS cells revealed that the mRNA was first detected 8 h after the onset of PMA treatment. These results were in good agreement with CAT activities obtained from HeLa and

RT-PCR study of MIA expression in biopsies from skin and pigmented skin tumors
RT-PCR results were semi-quantified from ethidium bromide-stained agarose gels in parallel to control PCR reactions of ␤-actin mRNA. Four groups were defined: negative (Ϫ), weakly positive after 32 PCR cycles (ϩ), weakly positive after 25 PCR cycles and strongly positive after 32 cycles (ϩϩ), and strongly positive after 25 cycles (ϩϩϩ). The diagnosis of melanoma was based on the assessment of two independent pathologists, and immunohistochemical stainings were positive in 100% of all cases for S-100 and at least focally positive for HMB-45. COS cells transfected transiently with the MIA promoter-CAT plasmid and treated with an inhibitor of RNA synthesis at various points after PMA induction. Stimulation of CAT activity was not observed when actinomycin D was added to the cell cultures earlier than 9 h after PMA, indicating that it does not represent a primary response (data not shown). DISCUSSION Here, we report the molecular cloning of the human genomic MIA locus, describe its organization, and provide an initial characterization of cis-regulatory elements within the 5Јgenomic region mediating high levels of gene expression in melanoma cells. The complete exon-intron structure was determined by sequencing two adjacent genomic XbaI fragments that cover four small exons interrupted by three intervening introns. The coding nucleic acid residues matched perfectly to the cDNA sequence obtained recently from a malignant melanoma cDNA library (EMBL Library accession no. X75450). The 5Ј-mRNA start was mapped by a primer extension experiment and was further cloned by RACE-PCR using poly(A) ϩ RNA from Mel Im melanoma cells. These experiments revealed that the mRNA is initiated 70 bases upstream from the protein coding region downstream of a pyrimidine-rich sequence motif followed by the nucleic acid residues AC. As frequently observed with TATAA-less genes, the initiator sequence is flanked 5Ј by a consensus SP-1 binding site. The polyadenylation signal AAATACAA is located 43 bases 3Ј downstream from the protein stop codon. Assuming a tail of approximately 200 -250 adenines, the sizes of the predicted transcript and the mRNA (approximately 750 bases) observed on Northern blots (Blesch et al., 1994) are in good agreement.
Data summarized in Tables I and II  lignancy of pigmented skin tumors and is not expressed in normal tissues of adult mice. By means of RT-PCR results, we detected no or very little MIA mRNA in non-neoplastic skin biopsies, moderate levels in the majority of non-malignant melanocytic nevi, and very high levels in every biopsy from malignant melanomas or metastases from melanomas. Interestingly, the two skin biopsies that expressed very low levels of MIA mRNA were taken from sun-exposed facial skin, and therefore MIA expression might result from subtle activation of melanocytes not detectable on microscopic examination. In the small number of biopsies examined in this study, we were not able to correlate levels of MIA mRNA in benign melanocytic nevi with a certain histological type of nevi. Therefore, it will be necessary to explore in a larger study whether MIA expression provides a prognostic parameter to define nevi at risk for malignant progression. Analyses of other S100 positive tumors including astrocytomas, oligodendrogliomas, and glioblastomas indicate that MIA expression is highly associated with melanocytic tumors and can be detected only occasionally in other neuroectodermally derived tumors (Blesch et al., 1994). 2 The melanoma-associated expression pattern of MIA was further substantiated by RT-PCR amplifications and Northern blot analyses of cell cultures in vitro. Together with data published previously (Blesch et al., 1994), we have now tested 10 different malignant melanoma cell lines, every one of which expressed very high levels of MIA mRNA. In contrast, all cultures of non-neoplastic skin cells including fibroblasts, keratinocytes (HaCaT cells), or melanocytes did not express MIA mRNA. The close correlation between MIA expression and melanocytic tumors or tumor cell lines raises questions about the function of MIA in regulating growth and invasion of malignant melanomas in vivo. We have observed recently that treatment of malignant melanoma cells with purified MIA protein in vitro results in growth inhibition paralleled by a significant change in cell morphology (Apfel et al., 1993;Blesch et al., 1994). Melanoma cells round up within 2 h after the addition of MIA protein to the tissue culture supernatant. It is therefore possible that secretion of MIA in vivo leads to decreased adhesiveness of melanocytic cells and thereby promotes melanoma progression and invasion. Ambivalent functions in regulation of tumor cell growth, invasion, and metastasis have been described for a number of different signal molecules including tumor necrosis factor-␣ and tumor growth factor-␤ (Orosz et al., 1995;Rodeck, 1993). Clearly, more functional studies are needed to assess the effect of MIA protein on melanoma cell growth and progression in vivo and further to define its physiological role in non-neoplastic cells.
Although our molecular analyses of the MIA promoter are still preliminary, our experiments point to a region of less than 500 base pairs, which is sufficient to mediate high levels of gene expression in malignant melanoma cells and which is much less active in benign pigmented skin tumors or normal skin. It will be important to define sites of specific protein interaction within this promoter region to elicit transcriptional changes associated with progression from benign melanocytes to malignant melanoma cells. A number of genes specifically expressed in melanocytes or melanoma cells have been described recently including melanotransferrin (Duchange et al., 1992), tyrosin-ase (Ganss et al., 1994), and MART-1 (Kawakami et al., 1994), and a promoter fragment of the tissue plasminogen activator gene mediating expression in melanoma cells has been identified (Fujiwara et al., 1994). A careful comparison of the melanocyte-specific promoter regions in these genes did not reveal any obvious cis-regulatory element in common with the MIA promoter.
Another very interesting finding is the activity of an NF-kBdependent cis-regulatory element present in the MIA promoter. NF-kB is a key mediator of a broad spectrum of signal molecules involved in inflammatory processes, and therefore NF-kB-mediated activation of MIA gene expression is likely to occur in response to the host immune defense. It is well known that melanoma cells frequently elicit a strong inflammatory host response at their site of invasion (Rodeck, 1993;Böcker et al., 1988), and partial regression of melanomas in areas of inflammatory infiltration is a quite common finding. These data point to a molecular link between classical host immune mechanisms in tumor rejection and specific regulation of growth regulatory genes in melanoma cells.