INTRODUCTION

90K is a secreted glycoprotein of 90 kDa, originally identified in the supernatant of human breast cancer cells (1, 2). Purification and cloning revealed its identity to the binding protein of human Mac-2 (3), a soluble and membrane-associated lectin thought to play a role in cell adhesion and immune response, and high similarity to a mouse cyclophilin C-associated protein (4, 5). Besides a scavenger receptor cysteine-rich domain (6), no similarity was observed to other proteins.

90K was detected in serum of healthy persons and, at elevated levels, in serum of patients suffering from various types of cancer (1, 10), AIDS (7-9), or autoimmune diseases.¹ In many mammary carcinomas an inverse correlation was found between expression of 90K and mRNA levels of the HER2 receptor tyrosine kinase, a marker for poor prognosis for patients with breast cancer (2).

Activation of the immune system was suggested by increased natural killer and lymphokine-activated killer cell activity of peripheral blood lymphocytes after treatment with 90K (2). Natural killer cells destroy cells lacking major histocompatibility complex class I molecules, a common consequence of virus infection or malignant transformation (11). Furthermore, an increase of 90K expression in mouse mammary carcinoma and glioblastoma cells significantly reduced their tumorigenicity in nude mice (12). The tumor suppressive activity of 90K was local as well as systemic, since a 90K expressing tumor could inhibit the growth of a distally implanted tumor. Perhaps as part of the tumor suppressive mechanism, 90K expression led to strong induction of intercellular adhesion molecule-1 and vascular adhesion molecule-1 in the tumor endothelium (12). Based on these observations, we hypothesized that 90K plays an important role in the body defense against cancer, viral infections, and possibly other pathogens.

How expression of 90K is regulated is only poorly understood. Previous studies reported up-regulation of its secretion in cancer patients by exogenous IFN² (13) and in human tumor cell cultures by IFN and IFN (13, 14). In mouse macrophages the 90K mRNA was shown to be induced by IFN, tumor necrosis factor , and adherence (5).

In order to define the molecular basis of transcriptional regulation of 90K, we isolated in this study the mouse 90K gene and analyzed its structure. The promoter region was characterized, and elements mediating induced transcriptional induction after treatment with IFN or poly(I·C), a polynucleotide mimicking the double-stranded RNA of viruses, were identified.

MATERIALS AND METHODS

An oligo(dT)-primed cDNA library of 3T3 L1 adipocytes in UniZap (Stratagene, prepared by Dr. I. Sures) was screened using a full-length human 90K cDNA probe, random-primed labeled with [-³²P]dATP (Amersham Int., UK) (15). Hybridization was at relaxed stringency in 30% formamide, 5 × SSC at 42 °C, washing at 42 °C in 0.2 × SSC, 0.1% SDS. A full-length cDNA clone of mouse 90K was obtained and checked by sequencing using the dideoxy chain termination method (16), comparing it with the published sequence (5). This cDNA was random-primed labeled with [-³²P]dATP and applied as a probe screening a genomic mouse library of 129 SVJ mice in  FixII (Stratagene) at stringent hybridization conditions (50% formamide, 5 × SSC). Positive clones were isolated and further checked by polymerase chain reaction (PCR; Ref. 21), amplifying short sequences from 5-noncoding, coding, and 3-noncoding regions of the mouse 90K cDNA to confirm isolation of a complete genomic clone. Restriction mapping and Southern blot analysis were carried out according to standard procedures (15). Phage DNA was digested with EcoRI and NotI. A 0.8-kilobase pair (kb) and a 12-kb NotI-EcoRI fragment were subcloned into pBluescript II KS (+) (Stratagene). A 4.6-kb EcoRI-EcoRI fragment was introduced into pLXSN (17). All exons, the exon-intron borders, and the 5-flanking region were sequenced using the dideoxy chain termination method. Intron sizes were confirmed by PCR, using oligonucleotide primer pairs flanking the introns.

Mouse chromosomes were prepared according to the published procedure (18). Chromosome slides were made by conventional method (hypotonic treatment, fixation, and air dry). The DNA probe, the 12-kb EcoRI-NotI fragment of the mouse 90K genomic clone, was biotinylated using the BRL BioNick labeling kit (19). The procedure for FISH detection was performed according to Heng et al. (19) and Heng and Tsui (19). FISH signals and DAPI banding pattern were recorded separately by taking photographs, and the assignment of the FISH mapping data with chromosomal bands was achieved by superimposing FISH signals with DAPI-banded chromosomes.

Three sequences of the 5-flanking region of the mouse 90K gene, overlapping the putative transcription start region, were amplified by PCR and subcloned into pBluescript II KS (+) (RMD2: 66  126, 192 base pairs (bp); RMD5: 66  91, 157 bp; RMD7: 112  91, 203 bp). To facilitate subcloning, the PCR primers contained 5-overhangs, creating a HindIII site at the upstream, and a XbaI site at the downstream end of the amplified fragment. Plasmid DNA was prepared using Qiagen midi-prep columns (Diagen, Germany), linearized by HindIII, and purified by Qiaex (Diagen, Germany).

For preparation of RNA probes, 1 µl of linearized template (approximately 200 ng) was mixed with 1 µl of 10 × transcription buffer (Boehringer Mannheim, Germany), 1 µl of 40 units/µl RNasin (Boehringer Mannheim, Germany), 1 µl of 125 µM UTP (Pharmacia, Sweden), 1 µl of 5 mM (GTP + ATP + CTP) (Pharmacia, Sweden), 4 µl of [-³²P]UTP (3000 Ci/mmol; Amersham, UK) and 1 µl of 20 units/µl T7 RNA polymerase (Boehringer Mannheim, Germany) and incubated for 1 h at 37 °C. Then, 100 µl of H₂O and 1 µl of 10 mg/ml Escherichia coli tRNA (Boehringer Mannheim, Germany) were added. After extraction with 1 volume of Chloropane (CHCl₃/phenol, 1:1, saturated with 10 mM sodium acetate, pH 6.0, 0.1 M NaCl, 1 mM EDTA), nucleic acids were precipitated by adding 0.7 volumes of 5 M ammonium acetate and 2.5 volumes of ethanol. Samples were incubated for 30 s in liquid nitrogen and centrifuged for 10 min at 10,000 × g. The pellet was washed with 70% ethanol, resuspended in 5 µl of formamide dye, denatured for 2 min at 95 °C, and applied on a sequencing gel (5% polyacrylamide, 8 M urea). After running, the RNA probe was located by autoradiography, excised, and extracted from the smashed gel piece by shaking for 2 h in 200 µl of elution buffer (0.5 M ammonium acetate, 0.1 M EDTA, 0.1% SDS, 50 µg/ml E. coli tRNA). Gel pieces were removed by spinning through a mini glasswool column, and the RNA was precipitated as described above.

200,000 cpm of radiolabeled RNA probe were mixed with 30 µg of total RNA or 0.5-2 µg of poly(A)⁺ RNA and dried completely in a vacuum centrifuge. The pellet was resuspended in 10 µl of hybridization buffer (80% formamide, 40 mM PIPES, pH 6.7, 0.4 M NaCl, 1 mM EDTA), heated for 5 min at 90 °C, and incubated overnight at 50 °C. Then 100 µl of RNase digestion buffer (10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 200 mM NaCl, 100 mM LiCl, 40 µg/ml RNase A, 2 µg/ml RNase T₁ (both enzymes from Boehringer Mannheim, Germany)) were added. After 60 min incubation at 37 °C 2 µl of 10% SDS and 3 µl of 10 mg/ml proteinase K (Boehringer Mannheim, Germany) were added for another 15 min incubation at 37 °C. The digest was transferred to a fresh tube, extracted with chloropane, and precipitated as described above. The pellet was resuspended in 5 µl of formamide loading buffer and analyzed on a denaturing sequencing gel. The length of the protected fragments was estimated by comparison to DNA fragments of a sequencing reaction run in parallel. Size estimation of several undigested RNA probes of known length by this method was accurate (data not shown).

A 4.6-kb EcoRI-EcoRI fragment containing exon 1 and exon 2 was subcloned into pLXSN. Deletion mutants were generated from this clone by PCR (21). To facilitate subcloning of PCR fragments, upstream primers contained a 5-overhang creating a HindIII site and downstream primers a 5-overhang creating a XbaI site. The PCR fragments were cloned into pCAT-Basic (Promega) digested with HindIII and XbaI. The sequence of the amplified fragments was checked by DNA sequencing (16). To generate MD3B, a 0.62-kb HindIII-EcoRI fragment of the mouse 90K gene, directly upstream of the 4.6-kb EcoRI-EcoRI piece, was introduced into a pBluescript construct containing already 1.7 kb of the upstream side of the 4.6-kb fragment, beginning at the EcoRI site. The fused 2.3-kb fragment was subcloned into pCAT-Basic. Sequencing of a PCR fragment derived from DNA of the genomic phage, which overlapped the fusion EcoRI site, proved the validity of the construct. Constructs with point mutations were generated by PCR using mismatch primers.

Cell Transfection, CAT Enzyme-linked Immunosorbent Assay, and -Galactosidase Assay

NIH 3T3 cells (ATCC CRL 1658) and CRL 6468 (ATCC CRL 6468) were maintained in Dulbecco's modified Eagle's medium high glucose (Life Technologies, Inc., UK) supplemented with 10% fetal calf serum (Sigma), 1 mM sodium pyruvate, 100 units/ml penicillin, and 0.1 mg/ml streptomycin at 37 °C, 5% CO₂. All plasmids used for transfection were purified by CsCl gradient (21). Subconfluent cultures (approximately 80% confluent) were transfected using the BBS method (15), applying 3.5-5 µg of 90K CAT construct and 0.8 µg of pCMV (Clontech) per 35-mm well. After 72 h cells were washed once with PBS (137 mM NaCl, 2.7 mM KCl, 1.5 mM KH₂PO₄, 8.1 mM Na₂HPO₄, pH 7.4) and lysed with 100 µl of CAT lysis buffer supplied with the CAT enzyme-linked immunosorbent assay kit (Boehringer Mannheim, Germany). 10-50 µl of the lysate were tested for CAT protein in the CAT enzyme-linked immunosorbent assay according to the instructions of the manufacturer. -Galactosidase activity was assessed following the method of by Nolan et al. (22), adjusted to microtiter plates. Briefly, 10 µl of lysate were mixed in a microtiter plate with 190 µl of PM-2 buffer (20 mM NaH₂PO₄, 80 mM Na₂HPO₄, 0.1 mM MnCl₂, 2 mM MgSO₄, 40 mM -mercaptoethanol, pH 7.3). Then 50 µl of 4 mg/ml o-nitrophenyl--D-galactopyranoside (Sigma) in PM-2 were added to each well. After incubation for 10-30 min at room temperature, optical density was measured at 420 nm. The CAT lysis buffer did not influence the -galactosidase activity as tested in independent experiments previously. The relative promoter activity was calculated by dividing the amount of CAT protein by the -galactosidase activity.

For the stimulation experiments, cells of two 35-mm culture dishes were trypsinized after transfection, combined, and reseeded into two 15-mm culture dishes. After overnight incubation one of the dishes was treated with 375 units/ml mu IFN (Sigma) or poly(I·C), whereas the other dish was used as a control. For treatment with poly(I·C), cells were washed two times with PBS+ (PBS supplemented with 0.9 mM CaCl₂ and 0.5 mM MgCl₂) and incubated for 1 h with serum-free Dulbecco's modified Eagle's medium containing 100 µg/ml DEAE-dextran (Pharmacia, Sweden) and 10 µg/ml poly(I·C) (Sigma). Afterwards, cells were washed again two times with PBS+ and incubated further for 72 h with normal growth medium. Control cells were treated with growth medium without IFN or with serum-free Dulbecco's modified Eagle's medium without DEAE-dextran and poly(I·C). DEAE-dextran alone did not induce 90K message (data not shown). Cells were lysed and assayed for CAT protein as described above. Stimulation was calculated by dividing amount of CAT enzyme of treated cells by that of untreated cells.

10-cm culture dishes with confluent NIH 3T3 were stimulated as described above. After indicated times cells were transferred on ice, washed once with PBS, and lysed by adding 1 ml of GTC solution (4 M guanidine thiocyanate, 25 mM sodium citrate, 0.5% sarcosyl, 100 mM -mercaptoethanol, pH 7.0). The lysate was refluxed through a 27-gauge needle to shear genomic DNA. Then, 0.1 volumes of 2 M sodium acetate, pH 4.0, 1 volume of phenol equilibrated with H₂O, and 0.2 volumes of CHCl₃ were added sequentially, mixing after each addition. After 15 min incubation on ice, the samples were centrifuged for 10 min at 10,000 × g. The aqueous phase was transferred to a new tube. RNA was precipitated by addition of an equal volume of isopropyl alcohol, incubation for 30 s in liquid nitrogen, and centrifugation for 10 min at 10,000 × g. The RNA pellet was resuspended in 300 µl of GTC solution, and RNA was precipitated again with 1 volume of isopropyl alcohol. After washing with 70% ethanol, RNA was taken up in 50 µl of H₂O. Northern blotting was carried out following standard procedures (15). As a probe, the complete mouse 90K cDNA was used, random primed labeled with [-³²P]dATP. To control the amount of RNA applied, the blot was hybridized with a TaqI-StyI fragment of the mouse glyceraldehyde-3-phosphate dehydrogenase cDNA.

RESULTS

Screening of a mouse genomic DNA library identified six different recombinant clones. After initial characterization by restriction mapping and PCR amplification of short sequences from the 5-noncoding, coding, and 3-noncoding region of the mouse 90K cDNA, a clone was chosen containing all sequences represented in the cDNA clone (2) and more than 2 kb of 5-flanking sequence. This clone was analyzed further by restriction mapping, Southern blotting, and PCR (Fig. 1). Following digestion with EcoRI and NotI, three fragments of 0.8, 4.6, and 12 kb were subcloned, and the complete sequences represented in the 90K mRNA and 5-flanking sequence were characterized.

Using FISH mapping, the chromosomal location of the mouse 90K gene was determined. As a probe, a 12-kb fragment of the 90K gene was used. Under the conditions used, the hybridization efficiency was 60% for the probe (among 100 checked mitotic figures, 59 of them showed signals on one pair of the chromosomes). Upon identification of the chromosomes by DAPI banding and evaluation of 10 preparations, the 90K gene was located on chromosome 11, region E (Fig. 2). This corresponds well to the location of human 90K to 17q25 (23).

The murine 90K gene was determined to contain 6 exons separated by 5 introns, defining a gene of about 8.8 kb in size (Fig. 1). Exon sizes ranged from 74 bp to 1.4 kb and intron sizes from 0.13 to 2.4 kb (Table I). All exon-intron borders were consistent with the reported consensus sequence for 5-splice donor and 3-splice acceptor junctions in mammalian genes (24, 25). While exon 1 contains noncoding sequences, exon 2 is composed of a noncoding stretch of nucleotides, the initiation codon, and most of the leader peptide coding region. The only identified subdomain of 90K, the scavenger receptor cysteine-rich repeat, is encoded by exons 3 and 4, which are separated by a small intron of 0.13 kb.

Table I. Exon/intron organization of the murine 90K gene The exon sequence is capitalized, and the intron sequence is in lowercase letters. Exon Size cDNA position 5-splice donor Intron Size 3-splice acceptor bp kb 1 163 1 -163 CTCCCAG gtgagtctccac 1 2.2 tttctgctgcag GGTTGGG 2 74 164 -237 ACTCAAG gtgagtcctgcc 2 0.8 ctctcccgacag GTACAGA 3 192 238 -429 GGGCCAG gtgtggctgagg 3 0.13 ttgtgcccacag GAAAGGG 4 132 430 -561 TCCAACG gtgagctgccct 4 1.3 tgtccatcccag ATACCAC 5 252 562 -813 TCCTCAG gtaagccaaacc 5 2.4 tccctccctcag GTACTTT 6 1373+ 814 -2185

Transcription start sites were assessed using RNase protection assays. Three different probes overlapping the putative start region (RMD2, RMD5, and RMD7; Fig. 3A) were applied on poly(A)⁺ RNA prepared from two unstimulated mouse mammary tumor cell lines (ATCC CRL 6374 and ATCC CRL 6468). In addition, total RNA derived from unstimulated NIH 3T3 cells, showing a very low level of 90K mRNA, and from NIH 3T3 treated for 24 h with IFN or poly(I·C), as described under "Materials and Methods," was used. Both agents induced high levels of 90K message in NIH 3T3 cells (cf. Fig. 6). RMD2 resulted in protected fragments of 126 nucleotides (nt; major signal), 93, 95, and 129 nt (Fig. 3B, left panel: lane 1 (CRL 6468), lane 2 (CRL 6374); right panel: lane 10 (poly(I·C), lane 11 (IFN)). RMD5 and RMD7 showed an identical pattern of protected fragments of 91 (major signal), 58, 60, and 94 nt (Fig. 3B, left panel: lanes 4 and 7 (CRL 6468), lanes 5 and 8 (CRL 6374), of RMD5 and RMD 7, respectively; right panel: lanes 4 and 7 (poly(I·C), lanes 5 and 8 (IFN), of RMD7 and RMD 5, respectively). Since the probes RMD5 and RMD7 share the same 5-end, which is hybridizing to the mRNA of 90 kDa, this identity was expected (Fig. 3A). The 5-end of RMD2 is 35 nt further downstream. On that account, the protected fragments detected with RMD2 should be 35 nt longer than the fragments detected with RMD5 or RMD7, which exactly agrees with the observed size differences. Induction of 90K mRNA in NIH 3T3 correlated with increased amounts of protected fragments, demonstrating the specificity of the signals (Fig. 3B, right panel: compare lanes 6, 9, and 12 (untreated) to lanes 4, 7, and 10 (poly(I·C) and lanes 5, 8, and 11 (IFN)). tRNA, used as negative control, did not yield any protected fragments (Fig. 3B, left panel: lanes 3, 6, and 9).

Induction of 90K mRNA in NIH 3T3 cells by IFN and poly(I·C).

All three probes indicated one major site of transcription initiation (designated +1) and three minor ones 3 bp upstream and 31 and 33 bp downstream of the major start site (Fig. 4). These start sites were identical in unstimulated mammary tumor cells and in NIH 3T3 cells stimulated with IFN or poly(I·C), suggesting the same promoter was responsible for transcription of the 90K gene in induced and noninduced cell lines (Fig. 3B, left panel: compare lanes 1-3 (CRL 6374) to lanes 4, 5, 7, 8, 10, and 11 (NIH 3T3)). This notion was confirmed by long exposures of the RNase protection gels, which revealed a protected fragment corresponding to the major transcription start site even with RNA of untreated NIH 3T3 cells.

Structure of the 5-flanking region of the murine 90K gene.

The single major transcription start site suggested the presence of an initiator element guiding the transcription machinery to that nucleotide (26). However, alignment of the DNA sequence surrounding the major transcription start site of the mouse 90K gene with known initiator sequences did not reveal strong similarities. Only a certain degree of similarity was detected to the downstream part of the porphobilinogen deaminase (PBGD) initiator element (Fig. 3C; Ref. 26). The sequence "TCCTGG" is found in the human PBGD initiator element and, 6 bp downstream of the major start site, in the 90K gene. Mutations in this sequence prevented accurate transcription initiation of the human PBGD gene in vitro. The upstream part of the human PBGD initiator element, however, is not found in the 90K gene. Most importantly, the "C" at position 1 of the PBGD initiator, shown to be essential for initiator function, was replaced by an "A" in the 90K gene.

Sequence Analysis of 5-Flanking Region

2.3 kb of 5-flanking DNA of the mouse 90K gene were sequenced and screened for potential transcription factor binding sites by comparison to the data base "TFSITES" using the GCG program "FASTA" (Fig. 4). Several TATA box-like sequences were found, all of them, however, in reverse orientation and located far upstream of the major transcription start site and therefore unlikely to be functional. It was shown that the sequence TACAAA can behave like a low affinity TATA box (28). This sequence motif is present exactly at the major transcription initiation site. However, since mRNA synthesis is usually initiated 25-30 bp downstream of the TATA box, this TACAAA sequence is very likely not responsible for the major transcription start. Still, it might result in mRNA synthesis beginning at the two minor start sites downstream (+32, +34).

The 90K gene, therefore, seems to belong to the class of TATA-less promoters. The majority of these promoters characterized to date display a very high GC content, possess multiple SP1 binding sites, and have a CpG island covering the transcription start (29). None of these features was found in the mouse 90K gene (GC content and CpG islands were checked using the GCG program "BASE PAIRPLOT").

Sequence analysis identified potential binding sites for NF-I (2103; Ref. 30), AP-1 (1512, 815; Ref. 31), AP2 (491; Ref 32), and progesterone receptor (1455; Ref. 33). There is no CAAT box close to the transcription start. At position 28 a sequence was detected shown to bind the transcription factors IRF-1 and IRF-2 in vitro (IRF-E; Ref. 34). IRF-1 is an activating transcription factor induced by IFN as well as by IFN and IFN. The transcription factor IRF-2 is mostly inhibiting but can be activating as well, as shown recently. Overlapping with this motif is a sequence "TTTCTGAAA" (33  25), similar to the GAS consensus element "TTNCNNNAA" (35). GAS elements mediate transcriptional activation by IFN. These sites are interesting, since 90 kDa is known to be up-regulated by IFN and IFN (5, 13, 14). Also tumor necrosis factor  was shown to induce 90K gene expression (5). However, no NFB site was found in the 2.3-kb 5-flanking region.

Promoter Activity of 5-Flanking Region of Mouse 90K Gene

To examine the promoter activity of the 5-flanking region of the mouse 90K gene, a series of CAT expression plasmids were constructed containing various deletions of the 5-flanking region. Using PCR, fragments were generated and cloned into the plasmid pCAT-Basic immediately upstream of the CAT encoding sequence. Promoter activity was assessed measuring the amount of CAT enzyme in NIH 3T3 cells transfected with these constructs. To compensate for different transfection efficiencies, CAT values were normalized with -galactosidase activity, as described above.

The 2.3-kb piece of 5-flanking region demonstrated strong promoter activity (MD3B; Fig. 5), being significantly above the negative control of pCAT-Basic. This activity was comparable with that of the SV40 early promoter. Interestingly, however, 90K gene expression is expressed in NIH 3T3 cells only at a very low level. This apparent discrepancy might suggest the existence of additional negative elements in the 90K gene or of other regulatory mechanisms controlling 90K mRNA levels.

Deletion analysis of 5-flanking region of murine 90K gene by transient transfection.

5-Deletion up to 1235 (MD8) led to a small increase in promoter activity, whereas further deletion to 133 (MD21) had no significant effect on promoter strength. While deletion to 67 (MD2) increased activity about 2-fold, indicating negatively acting elements in the region 132  67, removal of additional 51 bp led to complete loss of promoter activity (MD22). Since this mutant still contains the sequence around the major transcription start site, this region apparently does not possess any promoter activity. This is in contrast to several described initiator sequences, which by themselves can initiate low level transcription (26).

3-Deletions from +126 to +92 (MD5) and to 15 (MD1N), including the transcriptional start site, decreased promoter activity, pointing to positive elements in these parts of the 90K gene. The reduced activity of MD1N compared with MD2 could also be due to the loss of the endogenous transcriptional start site. Further deletion up to 98 (MD12) completely abolished promoter activity, confirming the nonfunctionality of the upstream TATA box-like sequences.

Together the 5- and 3-deletions allowed to identify a short region from 66 to 16 to be essential for promoter activity. Testing this sequence by itself, a significant promoter activity was measured (MD1N). This minimal promoter, therefore, seems to be essential as well as sufficient for transcriptional activity. Remarkably, this sequence does not comprise the major transcription start site.

Previous works reported induction of 90K mRNA in various cell lines by IFN and IFN (5, 13, 14). Poly(I·C), a substance mimicking the double-stranded RNA of viruses, was shown to elicit gene induction similar to that observed after infection with Newcastle disease virus, probably via a common pathway triggered by double-stranded RNA (36). Since we speculated that 90K might play a role in host defense against viruses, we tested IFN and poly(I·C) for their effect on 90K mRNA in NIH 3T3 cells. As shown in Northern blotting (Fig. 6), both substances could induce 90K message levels. The kinetics of the induction was similar. After 8 h an increase was clearly detectable, becoming rather pronounced the following 16 h, especially with poly(I·C).

Stimulation of 5-Flanking Region Promoter Activity

To test whether the isolated 5-flanking region of the mouse 90K gene can mediate an increased promoter activity after stimulation with IFN and poly(I·C), NIH 3T3 cells were transfected with various 90K gene CAT fusion constructs. Transfected cells were split, and one-half was treated with the stimulating agents, while the other was used as a control. The extent of stimulation was determined by dividing the amount of CAT protein of treated cells by that of untreated.

IFN increased the promoter activity of the 2.3-kb 5-flanking region of the 90K gene 3-5-fold, in different experiments (Fig. 7). 5-Deletions up to 67 (MD2) and 3-deletion to 15 (MD1N) did not significantly change this induction. Poly(I·C) enhanced promoter activity of MD3B 6-10-fold, in different experiments (Fig. 7). 5-Deletion to 1235 decreased inducibility by half, indicating a poly(I·C) responsive element in the region 2183  1235. Further 5-deletions up to 66 and 3-deletions to 16 had no influence on the inducibility. These results show that the minimal promoter is able to mediate enhanced transcription after stimulation with poly(I·C) or IFN.

Stimulation of promoter activity by treatment with IFN or poly(I·C).

The minimal promoter contains at its downstream end an IRF-E, which could mediate stimulation by IFN or poly(I·C), e.g. via the anti-oncogenic transcription factor IRF-1. To assess the function of this IRF-E several mutants of the minimal promoter have been prepared (Fig. 8). MD1 contains two point mutations not interfering with the IRF-E consensus, and MD24 four point mutations destroying the IRF-E motif. The IRF-E is deleted partially in MD23 and completely in MD25 and MD26.

Comparison with the wild type minimal promoter (MD1N) revealed that destruction of the IRF-E significantly reduced basal promoter activity (Fig. 9A, MD23-MD26). Still, even the biggest truncation of the minimal promoter tested (MD25) retained transcriptional activity at least 3-fold above background (Basic). Integrity of the IRF-E consensus sequence was also essential for the inducibility by IFN or poly(I·C), since disruption of the IRF-E motif leads to complete loss of induction (Fig. 9B, MD23-MD26). Stimulation values below 1, as observed for MD23-MD26, might be due to a general inhibitory effect of the treatments on protein synthesis. Modification of the IRF-E sequence without destroying the consensus (MD1) did not influence stimulation by poly(I·C), while a slightly reduced activation was observed with IFN. In conclusion, these data indicate that the IRF-E within the minimal promoter contributes to the basal activity and is mediating transcriptional activation by IFN or poly(I·C).

We found recently that overexpression of 90K protein in mouse mammary cell lines significantly reduced their ability to form tumors in nude mice (12). In order to see whether the mouse 90K promoter elements characterized in NIH 3T3 are similarly functional also in other cell lines, we transfected four promoter constructs (MD3B, MD10, MD1N, and MD23) into the easily transfectable mouse mammary carcinoma cell line CRL 6468. This cell line has a basal amount of 90K mRNA higher than NIH 3T3, which is about 5-fold increased by treatment with poly(I·C) or IFN (data not shown;).³

As in NIH 3T3, the minimal promoter (MD1N) was able to mediate induced transcription after stimulation with poly(I·C) (Fig. 10C) or IFN (Fig. 10B). This induction was dependent on the IRF-E within the minimal promoter since partial deletion of this sequence completely abolished stimulation (Fig. 10, B and C, MD23). A poly(I·C) responsive element not responsive to IFN was detected far upstream of the minimal promoter (Fig. 10, B and C; compare MD3B with MD10). The relatively low degree of induction might be explained partly by the fact that for technical reasons (reseeding, growing to complete confluency) the stimulation was started only 36 h after transfection. The relative promoter activity of MD3B and MD10 compared with MD1N was higher than in NIH 3T3, which could be related to the higher basal level of 90K mRNA observed in CRL 6468. As in NIH 3T3, partial deletion of the IRF-E significantly reduced promoter activity (Fig. 10 A, compare MD1N with MD23). Taken together, these data indicate that 90K promoter elements found in NIH 3T3 are similarly functional in the mammary carcinoma cell line CRL 6468. 

DISCUSSION

The promoter of the mouse 90K gene is a TATA-less promoter of unusual structure. Promoters without a TATA box have previously been found in constitutively expressed "housekeeping" genes, which in addition are GC-rich, contain several SP1 sites, and a CpG-rich island, and initiate transcription at multiple sites (29). A second group of TATA-less genes is not GC-rich and possesses only one or few tightly clustered transcription start sites. These discrete transcription initiation points are encompassed by an initiator element, containing all information necessary for determining specific initiation of transcription, both in vivo and in vitro (26). The 90K promoter, in contrast, is not GC-rich, initiates transcription mainly at one single site, and is lacking so far characterized initiator elements. The partial similarity to the PBGD initiator (27) does not include the upstream part of the motif, shown to be essential for initiation for that gene. Furthermore, deletion of the region around the transcription start site did not abolish promoter activity in the mouse 90K gene (Fig. 5, MD1N), in contrast to the PBGD gene, where the initiator is essential for strong promoter activity (27).

Few promoters of mammalian genes have been analyzed that like the 90K promoter do not fit into the groups described above, being TATA-less, not GC-rich, and without known initiator sequences: the promoters of cytosolic phospholipase A2 (37), which is important for prostaglandin and leukotriene synthesis of c-mos (38), a germ cell restricted proto-oncogene of C4 (39), a complement component, and of mouse DNA methyltransferase (Ref. 40). Whereas the cPLA2 promoter has a long run of CA repeats, the methyltransferase promoter contains a high amount of GT dinucleotides. Both the CA repeats and the GT-rich region could be deleted without affecting the promoter activity. In all of these genes a small region around the transcriptional start site was essential for promoter activity. No further similarities could be identified. The 5-flanking region of the mouse 90K gene did not contain a long run of CA repeats or a region rich in GT dinucleotides. There is a stretch of 88 bp (923  836), enriched in G and T (28.5 and 62.5%, respectively). Deletion of this region had no influence on transcriptional activity. In contrast to the other promoters, the region around the transcription start of 90 kDa was dispensable for promoter activity.

The lack of features in the 90K promoter common to other genes suggests a rather distinct way of linking the 90K promoter to the basal transcription machinery. Conceivably, this could be a feature defining specific and characteristic regulation of 90K expression.

A small sequence of 51 bp (66  16) upstream of the major start site was found to be essential and sufficient for transcriptional activity. Sequences upstream (2183  1235, 132  113, 112  67) decreased, and sequences downstream (17  +91, +92  + 126) increased promoter activity. To this minimal promoter proteins should bind that directly or indirectly interact with RNA polymerase II, similar to the TATA box binding protein in TATA box containing promoters (26). TFII-I, a 120-kDa protein specifically binding to the terminal deoxynucleotidyltransferase initiator, might play this role in certain TATA-less promoters (41).

The minimal promoter also mediated transcriptional induction by IFN or poly(I·C). Point mutations and deletions within the minimal promoter showed the integrity of the IRF-E at 28  16 to be essential for the stimulation. This activation may involve binding of IRF-1 to the IRF-E. Similar to 90K protein the expression of this transcription factor is increased by treatment of cells with IFNs, poly(I·C), and by viral infection (36, 42, 43) and also by other stimuli such as tumor necrosis factor , interleukin-1, and interleukin-6 (44-46). Experiments with IRF-1 knockout mice, however, suggested that at least poly(I·C) can induce the binding of activating transcription factors besides IRF-1 to the IRF-E (47). Moreover, IRF-2, normally an inhibiting transcription factor binding to the IRF-E, could be involved in stimulation as reported lately (48). It was recently shown that specific activation of the IFN gene requires a defined set of regulatory elements, including two IRF-1 binding sites, which are precisely arranged on the DNA surface (49). It will be interesting to test if similar cooperative effects take place in the minimal promoter of 90 kDa.

The isolated 2.3-kb 5-flanking region of the 90K gene possessed strong promoter activity and mediated inducibility by IFN and poly(I·C). Still, additional regulatory regions upstream or downstream, or different regulatory mechanisms, like chromatin structure or mRNA stability, for example, might modulate the amount of 90K mRNA. The high promoter activity in low expressing NIH 3T3 cells suggests a negatively acting mechanism in these cells independent from the 2.3-kb 5-flanking region.

Induced transcription by poly(I·C), a double-stranded RNA mimicking viral infections, indicated that virus infection may increase 90K levels not only via induction of IFNs but also possibly directly involving IRF-1. Another pathway seemed to be IRF-1-independent, since the region 2183  1235 increased poly(I·C) induction, while not containing an IRF-E. Moreover, this element did not respond to IFN.

We recently reported the inhibitory effect of 90K protein on tumor growth of mammary cell lines in nude mice (12). Based on this result it would be conceivable that certain oncogenes block, and certain anti-oncogenes stimulate, 90K expression. Interestingly, IRF-1 has been described as a tumor suppressor gene (50-52). Embryonic fibroblasts from mice with an inactivated IRF-1 gene can be transformed by expression of c-Ha-ras oncogene. Expression of IRF-1 reverted this phenotype. Furthermore, overexpression of IRF-2, an antagonist of IRF-1, in NIH 3T3 cells enhanced their tumorigenicity in nude mice. Concomitant overexpression of IRF-1 prevented this effect. IRF-1 is very likely activating a set of genes whose products are required for the negative regulation of cell growth. It is tempting to speculate that 90K belongs to this group of genes.

In summary, the functional analysis of the 90K promoter appears to be in line with our hypothesis of 90K as a part of the body defense against viral infections and cancer. Characterization of proteins interacting with the 90K promoter and comparison of promoter activity in normal and malignant cells will be carried out to further our understanding on how 90K expression is regulated.