MOLECULAR CLONING AND EXPRESSION OF A cDNA ENCODING A PROTEIN DETECTED BY THE K1 ANTIBODY FROM AN OVARIAN CARCINOMA (OVCAR-3) CELL LINE

MAb KI recognizes a cell-surface glycoprotein (MW -40 kDa) present in ovarian carcinomas, malignant mesotheliomas, squamous-cell carcinomas and normal mesothelial cells. In this study, expression screening was used to isolate cDNA clones encoding an antigen recognized by MAb KI from a cDNA library made from a human ovarian carcinoma cell line (OVCAR-3). Subsequently, other clones were isolated by DNA hybridization using a cDNA probe derived from one of the initial clones. The sequence of all the clones was similar. The longest cDNA contains 2,444 base pairs, and encodes a polypeptide of 263 amino acids with a calculated molecular weight of 30,51 I daltons. The nucleotide sequence and deduced amino-acid sequence of the protein show no homology to other sequences in current data bases. In vitro translation of RNA transcripts from the cDNA inserts yielded polypeptides of 29 and 30 kDa. Similar-sized proteins were obtained upon expression of the cDNA in Escherichio coli, and these proteins were reactive with MAb KI. The protein@) expressed in E. coli were purified and used to make rabbit or mouse antisera. These antisera reacted strongly with a soluble cytosolic protein in OVCAR-3 cells, but not with the membrane-bound antigen. Soluble cytosolic proteins of a similar size, recognized with MAb K I, were found in OVCAR-3 and N87 (gastric cancer) cells but not in 10 other cancer cell lines. These data indicate that the cloned cDNA encodes a cytosolic protein that reacted with MAb KI. This soluble protein is expressed only in cells containing the CAKl surface glycoprotein, suggesting that the 2 proteins could be structurally related.

MAb K I recognizes a cell-surface glycoprotein (MW -40 kDa) present in ovarian carcinomas, malignant mesotheliomas, squamous-cell carcinomas and normal mesothelial cells. In this study, expression screening was used to isolate cDNA clones encoding an antigen recognized by MAb K I from a cDNA library made from a human ovarian carcinoma cell line . Subsequently, other clones were isolated by DNA hybridization using a cDNA probe derived from one of the initial clones. The sequence of all the clones was similar. The longest cDNA contains 2,444 base pairs, and encodes a polypeptide of 263 amino acids with a calculated molecular weight of 30,51 I daltons. The nucleotide sequence and deduced amino-acid sequence of the protein show no homology to other sequences in current data bases. In vitro translation of RNA transcripts from the cDNA inserts yielded polypeptides of 29 and 30 kDa. Similar-sized proteins were obtained upon expression of the cDNA in Escherichio coli, and these proteins were reactive with MAb KI. The protein@) expressed in E. coli were purified and used to make rabbit or mouse antisera. These antisera reacted strongly with a soluble cytosolic protein in OVCAR-3 cells, but not with the membrane-bound antigen. Soluble cytosolic proteins of a similar size, recognized with MAb K I, were found in OVCAR-3 and N87 (gastric cancer) cells but not in 10 other cancer cell lines. These data indicate that the cloned cDNA encodes a cytosolic protein that reacted with MAb KI. This soluble protein is expressed only in cells containing the CAKl surface glycoprotein, suggesting that the 2 proteins could be structurally related. Q I994 Wile-y-Liss, Inc. * MAb K1 is a murine IgG1, monoclonal antibody (MAb) produced by a hybridoma obtained from mice immunized with the ovarian carcinoma cell line, OVCAR-3 (Chang et al., 1992d). This antibody recognizes a differentiation antigen present on normal mesothelium, but it also reacts strongly and homogeneously with most human epithelial ovarian cancers (Chang et al., 1992~1, d ) , many squamous-cell carcinomas of various origins (Chang et al., 1992c), and all tested malignant mesotheliomas of epithelial or mixed type (Chang et al., 1993). Further characterization of MAb K1 has demonstrated that MAb K1 reacts with an antigen different from those recognized by other MAbs that react with ovarian cancers. These include OC125 (Bast et al., 1981), B72.3 (Colcher et al., 198l), Movl8/Movl9 (Miotti et al., 1987)/MW207 (Mattes et al., 1987), OVB-1 (Kurrasch et al., 1989) and OVB-3 (Willingham et al., 1987).
The antigen recognized by MAb K1, CAK1, is a cell-surface glycoprotein with a molecular weight of -40 kDa . CAKl is associated with cell membrane via a glycosyl-phosphatidylinositol (G-PI) tail. Both PI-specific phospholipase C and n-butanol remove CAKl from the surface of OVCAR-3, HeLa and H-meso cells, as well as from human squamous-cell-carcinoma tissue samples (Chang et al., 1992a, b, c). Unlike many PI-anchored cell-surface proteins and tumor-associated antigens, CAKl is not shed into the serum of cancer patients or the medium from K1-positive cultured cells (Chang et al., 1992a). The epitope is protease-sensitive, and exoglycosidase-insensitive (Chang et al., 1992a), suggesting that MAb K1 reacts with the polypeptide backbone of the antigen. Furthermore, MAb K1 does not react with a panel of 34 neoglycoproteins whose carbohydrate residues are commonly present on tumor-associated antigens (data not shown).
The CAKl antigen is present in relatively low amounts on the membrane of cultured tumor cells; we estimate that cultured cell lines contain between 15,000 and 50,000 molecules/cell (data not shown). However, expression of CAKl is increased when tumor cells are grown in animals (Chang et al.,199%). In addition, the presence of the CAKl antigen was detected in the cytoplasm of normal mesothelia (data not shown) as well as in some human squamous-cell carcinomas (Gown, A., personal communication) by sensitive immunohistochemical methods, suggesting that other forms of CAKl may exist in the cytosol. The present study has 2 principal aims. One is to characterize the cell-surface-bound CAKl glycoprotein. The second is to investigate the relationship of the cell-surface CAKl with the cytoplasmic K1-reactive protein.

Material
MAb K1 was isolated as previously described (Chang et al., 1992a, d). MOPC21 (mouse myeloma IgGl, protein) was used as a negative control antibody for MAb K1 (Sigma, St. Louis, MO). Peroxidase-conjugated goat anti-mouse IgG (H+L) was used in primary screening and Western blots (Jackson Immu-noResearch Laboratories, West Grove, PA). E. coli strain XL1-blue, poly d(T) cellulose columns, in virro transcription kits with T3 and T7 RNA polymerases, and helper phage were purchased from Stratagene (La Jolla, CA). E. coli strain D H h , agarose, protein standards, DNA markers, RNA ladders, isopropyl-P-thiogalactoside and the 5'-RACE kit were obtained from GIBCO (Gaithersburg, MD). LB broth, superbroth, SOC and NZCYM media were supplied by Digene (Silver Spring, MD), and the plasmid DNA purification kit by Qiagen (Chatsworth, CA).

Solubilization of the CAKl antigen from postnuclear siipematarlt or membrane preparation of OVCALR-3 cells
Approximately 1 X lo8 OVCAR-3 cells grown to 90% confluence were rinsed 3 times with cold PBS-and scraped into 50-ml conical tubes. The cells were then pelleted and washed once in PBS without Cat'-and Mg++ (PBS-) and once in hypotonic buffer (10 mM Tris-HC1, pH 7.4,2 mM MgClz and 1 mM EDTA). After resuspension in 5 ml hypotonic buffer for 10 min at 4"C, the cells were homogenized by 20 to 40 strokes of a tight Dounce homogenizer. The post-nuclear supernatant (PNS) was isolated by centrifugation at 200g at 4°C for 10 min. The membrane-rich preparation was made from the PNS by ultracentrifugation at 100,000 g. The soluble and cytosolic fraction was used directly for Western blotting and the PNS or the membrane-enriched preparation was further used for solubilization experiments.
To solubilize the CAKl protein, 10% octylglucoside in 0.25 M sodium phosphate, pH 7.5, 0.02% sodium azide, 0.05 M EDTA, 0.1% SDS, 1% p-mercaptoethanol, and 1 mM PMSF were addcd to the PNS or membrane preparation at a final detergent concentration of 1 to 1.5% and incubated with gentle rotation for 60 to 120 miri at 4°C. The octylglucosidesolubilized proteins were collected after ultracentrifugation at 100,OOOg for 30 min at 4°C. Both pellet and supernatant were adjusted to equal volumes before applying to SDS-PAGE.
Western blotting was performed as described (Chang et al., 1992a ) with minor modifications. After SDS-PAGE and transfer, nitrocellulose papers were first soaked in 3% milk blotto (3% Carnation non-fat milk powder, 2% glycine, 1 mM PMSF in PBS) at room temperature for 30 to 60 min, followed by incubation with 5 kg/rnI MAb K1 or MOPC-21 at 41°C for 12 to 18 hr. The blots were washed in PBS containing 0.05% Tween 20 (PBS/T) 5 times for 60 min, then incubated with 10 kg/ml of peroxidase-conjugated goat anti-mouse IgG (H+L) in blotto at 4°C for 8 to 16 hr, or at 23°C for 1 to 2 hr. After 5 washes in PBS/T for 60 min, the nitrocellulose blots were developed using 0.4 mg/ml of diaminobenzidine (Sigma) and 0.01% hydrogen peroxide in PBS for 10 min. The reaction was terminated by rinsing the filters in distilled water. Western analysis for expression screening of the Uni-:ZAP-XR OVCAR-3 cDNA library was performed in a similar manner.

Poly (A)+ RNA isolation, and cDNA Iibraly construction and screening
Total cellular RNA from OVCAR-3 cells was extracted as described (Sambrook et al., 1989~). Poly (A)+ RNA was isolated with oligo(dT)-cellulose affinity chromatography according to the Stratagene protocol. The cDNA synthesis and library constructions were performed by Stratagene. IPoly (A) + RNA ( 5 kg) were used as templates for synthesis of the cDNA using the ZAP cDNA@ synthesis kit. The cDNAs containing 3' XhoI and 5' EcoRI cohesive termini were size-fractionated, ligated into phosphorylated Lambda ZAP I1 arms, and packaged with Gigapack I1 packaging extracts, and 3.5 X lo6 primary phages were obtained.
Both 2.5 x lo6 plaques of an unamplified and 1 x lob pfu of an amplified library were screened at approximately 50,000 pfu/l50-mm plate by the method of Young and Davis (1983) using protein-A-purified MAb K1 (5 yg/ml) and peroxidaseconjugated goat anti-mouse IgG (H+L) (10 k*g/ml)I. Positive plaques were isolated and the phages were purified to homogeneity by at least 3 rounds of screening. Iii vivo excision of the positive phage clones with R408 helper phage was carried out as described (Arcot and Deininger, 1992), and the circularized phagemid DNAs were extracted using the Qiageri plasmid DNA isolation kit and protocol. Restriction mapping using XhoI, EcoRI, SalI, BamHI and NcoI revealed thal. 4 clones were identical. The fifth clone contained an internal EcoRI site and reacted weakly with the control MOPC-21 antibody. One clone, g17(3), which was chosen for DNA sequence analysis, contained a cDNA insert of 2,100 bp with a 668-bp open reading frame sequence at its 5' terminus. A cDNA probe (specific activity = 8.5 x lo5 cpm/pl) spanning the 668-bp ORF sequence was made by random priming and the same OVCAR-3 library was re-screened to isolate a full-length cDNA clone using the filter hybridization method (Benton and Davis, 1977) with minor modifications.

Sequencing analysis, restriction mapping, and Northern and Southern analyses
Using T3, T7 and 12 17-bp synthetic primers, the entire cDNAs were sequenced according to the method of Sanger (1977). Restriction sites were determined with several restriction endonucleases (XhoI, SalI, NcoI and EcoRI) and the nucleotide sequence and the deduced amino-acid sequence were analyzed for homology to sequences in GenBank/EMBL.
To determine the size of the mRNA, total RNA (20 kg) and/or poly(A)+ mRNA (2 kg) were electrophoreseld on a 1% agarose gel containing MOPS buffer (0.04 M 3-morpholinopropanesulfonic acid, 0.01 M sodium acetate, 0.01 bl EDTA) and 16.6% formaldehyde. All gels were stained with ethidium bromide to assess the integrity of the RNA and the quantity of each loading. The RNA was then transferred to nitr80cellulose paper in lox SSC for 18 to 36 hr and cross-linked u:sing a?UV cross-linker (Stratalinker 1800, Stratagene). The blots were pre-hybridized, then hybridized with the appropriate cDNA probe. Northern blots were also probed with radiolabeled human p-actin cDNA as an internal control. In Southern-blot analyses, human placental genomic DNA was dige:sted with EcoRI, HindIII, BamHI, Pstl and BglII, electrophoresed on a 0.7% agarose gel and transferred to nitrocellulose paper (Human Geno-Blot, Stratagene). After prehybridization with salmon sperm DNA, the blots were hybridized with the probe as described above.

In vitro transcription and translation
To study the polypeptide encoded by the cDNA clones, the purified g5 and g26 pBluescript plasmid DNAs were linearized by digestion with XhoI or EcoRI. The in vitro synthetic and capped RNAs were transcribed with T3 or T7 R N A polymerase in the presence of m7G(5')ppp(5')G analog of GTP to obtain either sense or anti-sense R N A transcripts following the method supplied by the manufacturers (Stratagene and Promega). The yield of the synthetic RNAs ranged from 40 to 50 pg per reaction and the integrity and size of the transcripts were determined by gel electrophoresis on a 1% agarose gel and ethidium bromide staining. Synthetic RNAs (2-4 pgiml) were translated in vitro in a SO-pI reaction solution containing 10 pI (0.8 mCi/ml) of L-[35S]-methionine, 10 pl of cocktail, 4 pl of 1M potassium acetate. 1 pI of 32.5 mM magnesium acetate, 1 p1 RNasin and 20 p1 of reticulocyte lysate in the presence and absence of 0.5 pl of dog pancreatic microsomal membrane following the protocol provided by the manufacturer (New England Nuclear and Promega). Translation products were resolved on a 12.5% SDS-PAGE reducing gel, the proteins were fixed and the unincorporated radioactivity was removed by soaking the gel 3 times in 200 ml of fixation buffer (40% methanol and 10% acetic acid in deionized water) for 30 min. The gels were then soaked in 200 ml of Amersham's Amplify enhancement solution for 30 min. After drying, the translated products were visualized by autoradiography.

Expression of the cloned cDNA in E. coli
Oligonucleotide primers (1. S'ACGTTGCAACGTCATATG GGACCAC TTCACAAAAGC3' and 2. S'ACGTTGCAAC-GTGAATTCTTAGGTCAGCTT CAAGCC3') were synthesized by automated phosphoramidite chemistry on a model 380A DNA synthesizer (Applied Biosystems). NdeI and EcoRI restriction sites were introduced 5' to the first initiation codon and 3' to the termination codon of the open reading frame (ORF) of g26 (2,444 bp) using a recombinant PCR method.
PCR was performed for 35 cycles (denaturation at 94°C for 40 sec, annealing at 55°C for 40 sec and elongation at 72°C for 60 sec), the final elongation step being extended by 3 min, using a DNA Thermal Cycler (Perkin Elmer Cetus, Nonvalk, CT). The reaction products were filtered through a Centricon 100 and extracted with phenol/chloroform. The amplified DNA fragment was analyzed in a 1% agarose gel stained with ethidium bromide. pVEX11 plasmid DNA and the PCR-engineered ORF of g26 DNA were digested with NdeI and EcoRI. This was followed by 2 units of calf intestine alkaline phosphatase (CIAP) treatment at 37°C for 30 min (pVEX11 only) and phenol/chloroform extraction. The DNA fragments were then isolated and purified to homogeneity using LMP agarose (SeaPlaque-GTG from FMC, Rockland, ME) as described (Sambrook et al., 1989a). The NdeI/EcoRI containing O R F of g26 (referred to as g26NE) fragment was directionally cloned to the pVEX1l vector with T4 ligase in an insert:vector molar ratio of 3:l using a rapid ligation method (Sambrook et a/., 19896). The ligated product (pAPKI) was transformed into competent DH5a E. coli cells and the insert-containing clones were isolated and verified by restriction mapping of DNA minipreps. The cDNA was then used to transform BL21 (XDE3) E. coli in an effort to express the cloned gene following the procedure described (Studier and Moffatt, 1986).
A single colony of the pAPK1, or pVEXl1-transformed BL21 cells was inoculated into a 5-ml LB culture containing 100 pg/ml of ampicillin, and grown at 37°C until it reached an ODm0 of 0.8 to 1.0; this was followed by addition of 1 mM IPTG and continuous growth at 37°C for 1 to 2 hr. An aliquot of each culture was dissolved in SDS-PAGE sample buffer and the rest of the culture was pelleted. The bacterial pellet was resuspended in 250 pl of TE buffer, frozen and thawed 3 times on dry ice and sonicated for 30 sec ~2 at SO w. The lysate was microcentrifuged and the pellet was suspended in 500 p1 1 x sample buffer, while the supernatant was mixed with 250 pl of 2~ SDS-PAGE sample buffer. The whole cell extract, and insoluble and soluble fractions were analyzed by 10% SDS-PAGE and Western blotting as described above.

Rabbit and mouse sera against recombinant APKl protein
Cytosolic recombinant antigenic protein (APKl) encoded by pAPKl was precipitated with 20% ammonium sulphate and resolved in a 10% SDS-PAGE preparative gel and stained with Coomassie brilliant blue. The APKl bands were electroeluted. APKl (0.3 mg) in 0.5 ml was mixed with 0.5 ml complete Freund's adjuvant (CFA) and injected into rabbits S.C. at multiple sites on day 1, then a second boost of antigen in incomplete Freund's Adjuvant (IFA) was given on day 21 and a third boost of 0.3 mg/0.5 ml APKl on day 42. The sera were collected on day 49.
Renatured APKl (20 pg in CFA) was injected into BALB/c mice both i.p. and S.C. Three weeks later each mouse was boosted with 10 pg of APKl in IFA, then 5 boosts of 10 pg/mouse of APKl in PBS were given once a week. Blood was collected within 3 days after the final boost, and the serum was saved.

Characterization of CAKl from OVCAR-3 cells
We have previously used intact OVCAR-3 cells to show that CAKl is a glycoprotein probably anchored to the cell membrane by phosphotidylinositol (Chang et aL, 1992a, b, c). To further characterize the protein, we needed to release it in a soluble form. We have previously tried several detergents without success (Chang et al., 19924~). In this study, we evaluated octylglucoside. Figure 1 shows that over 80% of CAKl was released into the supernatant after treatment with 1% or 1.5% octylglucoside (Fig. 1, lanes 1 and 3 respectively), while less than 20% of CAKl remained associated with membranes ( Fig. 1, lanes 2 and 4). The CAKl band was often a doublet. However, treatment with endoglycosidase F generates a single, smaller band (lanes 6 and 9) indicating that the doublet may be due to different glycosylated forms. In addition, an abundant membrane protein with an MW -39 kDa is evident upon Coomassie blue staining which may contribute to the generation of the 2 bands. We then subjected the octylglucoside solubilized CAKl to a series of exo-and endoglycosidases. Extensive digestion with neuraminidase or neuraminidase and P-galactosidase or 0-glycanase had no effect on the size of CAKl, whereas diges,tion with endoglycosidase F (Fig. 1, lanes 6-8) or peptide-N-glycosidase F (Fig. 1, lanes  9-1 1) decreased the apparent molecular weight of the glycoprotein. This indicates that CAKl is an N-linked glycoprotein. The apparent MW of CAKl decreased from 40 kDa to 30 kDa after deglycosylation with either of the endoglycosidases (Fig.  1, lane 7 or 8 VS. 6, and lane 10 or 11 vs. 9, respectively). Peptide-N-glycosidase F deglycixylated CAKl more completely than endoglycosidase F (Fig. 1, lane 9 vs 6). After removal of carbohydrates, MAb K1 reacted with the deglycosylated form of CAKl with the same intensity as with glycosylated CAKl (Fig. l, lanes 7, 8, 10, and 11). This result confirms our previous data that MAb K1 recognized the core peptide rather than the carbohydrate moiety present on the protein (Chang et al., 1992a). Based on the observation that the antibody reacted with a denatured form of CAKl on an immunoblot, we screened a h cDNA expression library for a cDNA encoding the 30-kDa protein reactive with MAb K1.

Isolation arid characterization of the pAPK1 cDNA clories
The human ovarian carcinoma cell line, OVCAR-3, was chosen as a source of poly (A)+ mRNA for construction of a lambda ZAP cDNA library because OVCAR-3 cells were the original immunogens for generation of MAb K1 and they contain CAKl on their surface as demonstrated by immunofluorescence (Chang et al., 1992a, d)i. We screened about 2 x lo6 recombinant phages with MAb K1 and obtained 5 positive clones from an amplified library. These K1-positive clones were purified and screened simultaneously with MAb K1 and a control antibody, MOPC-21, of the same isotype, to verify their specificity. Four of the 5 clones were specifically reactive with MAb K1. One clone showed weak reactivity with the control MOPC-21 antibody and was not further investigated. All 4 K1-spccific cDNA clones had inserts about 2,100 bp long (Fig.  2) and had similar restriction patterns when analyzed by the restriction enzymes XhoI, EcoRI, SalI, BglI, BamHI and NcoI, indicating they may have come from a single primary phage clone.
The complete nucleotide sequence was determined on both DNA strands. The g17(3) clone is 2,100 bp long, and contains a 668-bp open reading frame followed by a 3' non-coding segment of 1,432 bp. Neither a 5' non-coding sequence nor a consensus initiation sequence (Kozak, 1987) could be found in the cDNA fragment. Therefore, we assumed that this clone is missing sequences at the 5' end of the cDNA. To determine the size of the messenger RNA, the RNAs extracted from tumor-cell lines and normal tissues were resolved by electrophoresis and a Northern-blot was performed. This experiment showed that the RNA was about 3,800 bp in size (Fig. 3). To obtain the sequence of the 5' end, RNA from OVCAR-3 cells was rctrotranscribed into cDNA, which was then extended in the 5' direction using the 5' RAlCE method (Frohman et al., 1988). The result showed that a segment of about 1,000 to 1,400 bp was missing at the 5' end of the g17(3) insert (data not shown). To obtain a full-length cDNA, the OVCAR-3 cDNA library was rescreened with an EcoRIiBamHI cDNA fragment as a probe [g17(3)EB]. This probe came from the 5' end of the g17(3) cDNA insert. Using a random-primed cDNA probe, we obtained 30 candidate clones from the primary screening: 11  1141  1201  1261  1321  1381  1441  1501  1561  1621  1681  1741  1801  1861  1921  1981  2041  2101  2161  2221  2281  2341  2401 TGGATAGAGCGAGGAGAGGTCAACCGTCGTAGCGCCAATAACTTCTACTCCj\TGATCCAG

L E K R W K F C G F E G L K L T '
CCTAACAACTTGGGACTCCTGAAGATAAATATGTGTTGGACAAGCATAGAAAGTGATTTA TATTTTTAATGGTTTTCAAGTGGAAGTTCCTTTGAATTTGTCAGTTCATTC1:TGGAAAAT CTTTTGAGTTAAAATAAGGATCCTAGGACAGCACCTCGAACTACAGGCCCT,4AAGAGAAA TTGCCTCAAACCACAAGTGCTGTAACTTCCTCCCCTTTCTGTCAATTGGTTI~TCTTTAAA TATTGCAAAAGTCCTGATGCTAAACAGTATTTGGAGTGTTTTCAGTGTCTG'rACTACTGT TGTAGACCTTGGTATTTTTTTAAACACTGTTAACTGAAATGTTTTGATGAT'rTGTATGTG ATTTGTGTTTCTAAACTTCTCTTTACATTAATGTTGTTACTGGTGAAAGGC,~TGAGAGCA GCACTAAGTCCTCTGTGTAACTGCCATTGTCTTTCCAATCCCCAGTAGACC,4GTAAATAA clones were subjected to a second round of subcloning and all had inserts that ranged from 400 bp to 2,500 bp in size. Two of the longest clones (g5 and g26) were sequenced as shown in

Northernand Southern-blot analysis
As shown in Figure 3a, a single 3,800 mRNA species from cells of OVCAR-3. A431 and N87 (lanes 1 , 2 and 3) was found to hybridize with the cDNA probe, but it was not detectable in FEMX melanoma cells or W138 human fibroblast cells (lanes 4 and 5). Based on ethidium bromide staining of the gel and on similar hybridization with the control p-actin probe (lanes 1-5, bottom), the R N A from each of these sources appeared to be intact and equally loaded and transferred. Using the same probe, we also tested poly(A)+ mRNA from various normal human tissue samples (Fig. 3b). mRNA of the same size was present in heart, placenta and skeletal muscle (lanes 1,2 and 5 , respectively). It was also present, but in lower amounts, in lung, liver, kidney and pancreas (lanes 3, 4, 6 and 7, respectively). Small amounts of mRNA, about 5.5 kb in size, hybridized with the cDNA probe in all the normal tissues tested. The origin of this signal is not known. A Southern blot of the digested genomic DNA from human placental tissue is shown in Figure 4. The probe hybridizes strongly with a prominent band of about 16 kb in genomic DNA cleaved with HindIII (lane 2). Other less prominent bands are present in other digestions. The simple pattern of the Southern blot suggests that the CAKl antigen is not a member of a large gene family .

Amino-acid sequence analysis
The deduced amino-acid sequence of the cloned cDNA is shown in Figure 2. The protein is 263 amino acids in length with a calculated MW of 30,511 kDa. It is an acidic protein  with a PI = 4.9 and contains 3 potential N-linked glycosylation sites, Asn-X-Ser or Asn-X-Thr as underlined in Figure 2 and multiple potential phosphorylation sites (Ser, Thr, and Tyr). An amino-acid hydropathy plot of the deduced amino acid identifies one very hydrophobic region from amino-acid residues 185 to 200 (Fig. 5). Homology analysis was performed for both nucleotide and amino-acid sequences using FASTA and TFASTA of the GCG program against EMBLiGenebank. No significant homology was found with any known sequence at either the nucleotide or amino-acid level. At the carboxyl terminus of proteins that are attached to the plasma membrane by PI, there are usually non-polar amino acids forming a hydrophobic region (Low and Saltiel, 1988). However, we could not identify such a region in the deduced peptide sequence of the cDNA we had isolated. Furthermore, a typical signal peptide sequence could not be recognized at the N-terminus of the coding sequence, even though CAKl is known to be a cell-surface-associated glycoprotein (Chang et aL, 19927, b, c, d).

In vitro translation analysis
Even though the cDNA clone we had isolated did not have a typical signal sequence as expected for a membrane-associated protein, we wanted to determine if the sequence at the amino end might still function as a sign,al sequence and enable the protein to reach the cell surface. Therefore, we used a reticulocyte lysate system and pancreatic microsomes to determine if the protein encoded by clone g26 could be introduced into microsomes. First, mRNAs were prepared, using T3 or T7 RNA polymerase; both sense and anti-sense transcripts were detected as a single band (data not shown) with apparent lengths of 2,400 bp for the g26 transcripts. As shown in Figure  6, in vitro translation of the synthetic sense R N A of g26 yields a polypeptide with an apparent MW of approximately 29-30 kDa in the absence (lane 1) or presence (lane 2) of microsomal membranes, whereas in vitro translation products of the anti-sense RNA transcript (lane 3) or a negative control without addition of exogenous R N A (lane 4) did not produce detectable quantities of the 29-to 30-kDa radiolabeled polypeptide. There was no apparent change in MW of the in vitrotranslated g26 in the presence of microsomal membranes, indicating that neither translocation nor glycosylation of the cDNA encoded product had occurred (Fig. 6, lane 2). In addition, we performed protease protection and microsome sedimentation experiments (data not shown) which verified that the cDNA-encoded product was not translocated into microsomes.

Expression of recombinant APKl antigen in E. coli
To prepare a sufficient amount of APKl for further characterization, we expressed the antigen in E. coli using a T7-based expression system (Studier and Ivloffatt, 1986). The g26NE segment was ligated into a T7-based vector (pVEX11) as

PNS Membrane
CVrn<"l Cell line FIGURE 8 -Antisera raised from mouse immunized with recombinant APKl found in OVCAR-3 cells. Fifty micrograms of octylglucoside-solubilized membrane (lanes 1-3) and cytosolic preparations (lanes 4-5) of OVCAR-3 cells were electrophoresed and transferred to nitrocellulose, and a Western blot was performed as described in text. Lanes 1 and 4, incubated with MAb Kl; lanes 2 and 5 , incubated with the mouse antiserum to recombinant APK1; lanes 3 and 6, incubated with normal mouse serum. Large arrowhead: cytosolic protein reactive with both MAb K1 and mouse antiserum; small arrowhead: another cytosolic species reactive only with MAb K1. Right: kDa values.
frame. The BL21 T7 RNA polymerase may recognize this internal SD sequence. This was confirmed by immunoblotting which showed that both bands reacted with MAb K1 with a similar intensity (Fig. 7b, lanes 1 and 3, large arrowheads). The reactivity of MAb K1 with the 29-kDa and 30-kDa recombinant proteins was specific, since an isotype-matched control antibody MOPC-21 did not react with these bands (lane 5). Other minor bands (small arrowheads) of a higher MW also reacted with MAb K1, but these bands were also present in BL21 lysates transformed with either pVEX1l (Fig. 7a, lane 2) or TGFa-PE40 (data not shown), suggesting they may be products related to transformation of BL21 cells and unrelated to pAPK1-specific expression. Therefore, we conclude that MAb K1 reacts with a 29-kDa and 30-kDa protein encoded by g26NE.

Reactivity of rabbit and mouse antisera with tumor-cell lines as compured with MAb KI
To investigate the nature of the cloned antigen, both rabbit and mouse antisera against the recombinant antigen encoded by pAPKl were raised as described above. These antisera were then analyzed for their immunoreactivity with membrane and soluble proteins present in several different human tumor cell lines. As shown in Figure 8 and Table I, the mouse antiserum strongly reacted with a cytosolic protein(s) found in both OVCAR-3 (Fig. 8, lane 5, large arrowhead) and N87 cells ( Table I). The mouse antiserum was poorly reactive with the membrane-associated octylglucoside-solubilized 40-kDa CAKl (lane 2). As expected, MAb K1 reacted with the membrane (CAK1) antigen (lane 1). In addition, MAb K1 reacted with 2 bands of 34 kDa and 39 kDa, respectively, in the cytosolic fraction of OVCAR-3 cells (lane 4). One of these bands (MW -39 kDa) was in the same location as the band reacting with the mouse antisera. The 39-kDa protein that reacted with the mouse antiserum was not detected in any of the 10 other tumor-cell lines. These cells also failed to react with MAb K1 ( Table I). Neither the control MOPC 21 antibody (data not OVCAR-3 Immunoblotting analyses were performeJas described in Changeid. (1992d) and in the text (this report). + ++, Very strong reactivity; +, moderate reactivity and -, no reactivity. The values for reactivity with MAb K l (CAK1) are on the left and reactivity with mouse anti APKl are on the right in each column.
shown) nor the normal mouse serum reacted with membrane (lane 3) or cytosolic (lane 6) proteins from OVCAR-3 cells. It is evident that clone g26 encodes a soluble protein found in OVCAR-3 and N87 tumor cells but not in many other human cancer-cell lines including A431, KB3-1, HUT-102, Daudi, HTB103, HepG2, Huh7, MCF-7, MB-MDA231 and JMN (Table I). Both OVCAR-3 and N87 cell lines also contain substantial amounts of CAKl, while the other lines do not.

Characteristics of the cloned cDNA arid its encoded protein
The cDNA clones encoding a protein recognized by MAb K1 were isolated from an expression library made from OVCAR-3 cells. These clones differ in size, but all encode an identical peptide of 263 amino-acid residues. A computerassisted search at both the nucleotide and amino-acid levels showed no homology of the cloned molecule with known sequences. The protein encoded by the cloned cDNA, now designated APK1, has the following characteristics: (1) the 2,444-bp cDNA contains a 789-bp open reading frame which encodes a polypeptide of 263 amino acids; (2) both iii vitro translation and in vivo E. coli expression of the K1-reactive clone g26 yield polypeptides of 29 and 30 kDa, which agrees with the calculated MW value of the peptide of ORF; (3) recombinant antigen(s) expressed in E. coli has an MW of 29 and 30 kDa and is specifically reactive with MAb K1; and (4) mouse antisera to the recombinant protein reacted strongly with the same cytosolic proteins in OVCAR-3 with which MAb K1 reacts, and reacted weakly with a 39-to 40-kDa membrane protein that migrates in the same location as the MAb-K1reactive CAKl band. These characteristics show that the isolated cDNA sequence encodes an antigenic protein reactive with MAb K1. However, CAKl is a membrane-bound protein whereas APKl is a soluble protein.

Relationship of the CAKl to APKl
The complexity of expression of oncoproteins, including tumor antigens and oncogene products, has been explored in many human malignancies. Tumor antigen can be an oncoprotein resulting from an over-expressed oncogene, as in the case of HER-2/neu in breast carcinomas (Slamon et a/., 1987;Chang et al., 1991). A normal functional gene can be mutated during carcinogenesis leading to "creation" of tumor-specific epitope(s) such as mutant p53 (Finlay et a/., 1989;Lane and Benchimol, 1990;Gannon et al., 1990;Bartek et al., 1990;Chang et a/., 1991). Tumor antigcn can share epitope(s) with a normal protein as was shown in CEA and its related antigens in colon cancer and normal colon (Kuroki et al., 1981;Henslee ef al., 1992). It is not surprising that 3 protein species were found to b e immunologically reactive with M A b K1, even though the complicated mechanism which lies behind them remains to be discovered.
We have found that the c D N A we isolated encodes a 39-t o 40-kDa cytosolic protein which is recognized by MAb K1. The protein detected by M A b K1 on immunoblots was only detectable in 2 of 12 tumor-cell lines examined (Table I). Moreover, both of these (OVCAF1-3 and N87) strongly reacted with M A b K1 when analyzed by immunofluorescence. In addition, the amount of APKl expression correlated with the level of CAKl in these tumor-cell lines. Another K1-reactive 34-kDa species (now designated as CAK2) is also present in the cytosol of both OVCAR-3 and N87 cells, but it does not react with either rabbit or mouse sera raised against recombinant A P K l . CAK2 a n d APKl are usually present together in cells that are strongly K1-positive, but absent in imost K1negative or weakly positive cells, indicating that they may b e coexpressed with the CAKl antigen in these tumor cells.
In summary, we have isolated a cDNA that encodes a 39-to 40-kDa protein (CAK1) found in OVCAR-3 and N87 tumorcell lines. APKl is a cytosolic protein whereas CAKl is bound to the plasma membrane. Nevertheless, APKl is immunologically related to the CAKl antigen. Now that a KL-reactive antigen has been cloned, epitope mapping can be carried out t o pin-point the K1-reactive peptide sequence and should b e very useful in cloning the other antigens such as C A K l or CAK2 that are recognized by M A b K1.