(CANCER RESEARCH 51. 1418-1425. March 1. 1991]

Novel Gene Sequences Expressed by Human Melanoma Cells Identified byMolecular Subtraction1

Jeff T. 11uichins. Robert J. Deans, Malcolm S. Mitchell, Christopher Uchiyama, and June Kan-Mitchell2

Departments of Microbiology /J. T. //., R. J. D.. M. S. M.. C. U.J, Medicine fM. S. M.], and Pathology fJ. K-M./. University of Southern California School of Medicine,I.os Angeles, California 90033

ABSTRACT

Despite the variety of approaches used, only a limited number oftumor-associated antigens have been described for each histológica!typeof tumor. In this report, we present a new strategy involving molecularsubtraction to identify novel melanoma-associated gene sequences. Toward this end, 156 complementary DNA clones were isolated with asubtracted melanoma complementary DNA probe (melanoma minus lungcarcinoma) after screening 2 x HI1independent recombinants of a mela

noma expression library by in situ plaque hybridi/ation. These cloneswere then polymerase chain reaction amplified, screened for duplication,and categorized into 53 discrete genes. By applying poissonian distribution to the numbers of duplicate isolates, we found most of the genes tobe rare messages, present at <l copy/200 molecules of mRNA in atypical somatic cell. Messages specific for a type of tissue are usuallyexpressed in this range. The expression of the 53 genes was furtherstudied in human tumor cell lines and normal tissues. Partial sequencedata obtained for 20 complementary DNA clones revealed 8 novel humangenes. The mRNA transcripts for 5 of the novel genes were identified byNorthern blot analysis. Thus, molecular subtraction appears to be applicable for the identification of novel tumor-associated sequences. Some ofthe potential advantages and limitations of this technology are discussed,including its application to the molecular characterization of immunogenicmelanoma-associated antigens.

INTRODUCTION

Historically, the search for "tumor markers" was instigated

by their potential usefulness for diagnosis of cancer in man. In1962. the Bence-Jones proteins were identified as markers ofmultiple myeloma (1). The discovery of «-fetoprotein (2) andcarcinoembryonic antigen (3) soon followed, as did a host ofother tumor markers, including a variety of enzymes, hormones,serum proteins, and TAAs.' More recently, hybridoma tech

nology provided novel precise probes for identification as wellas molecular and functional analyses of many TAAs (4, 5). Ingeneral, tumor-reactive MAbs have been generated by immunizing rodents with human cells or tissues. However, it isbecoming clear that mouse antibodies generated by the typicalimmunization and screening procedures have readily identifiedonly a limited number of TAAs associated with each histológica! type of tumor (6). Recent gene-cloning and DNA-sequencingdata suggest that the mouse MAbs generated independently indifferent laboratories frequently recognize identical immuno-

Received 9/19/90: accepted 12/18/90.The costs of publication of this article were defrayed in part by the payment

of page charges. This article must therefore be hereby marked adrertiscmem inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.1This investigation was supported in part by USPHS Cirants C'A 43220. C'A

36233. and CAI 40146 awarded In the National Cancer Institute. NationalInstitute of Cieneral Medical Sciences. National Institutes of Health. Departmentof Health and Human Services; by a grant from the Concern Foundation: by giftsfrom Alan Glcitsman and the C'andle Foundation: and by a Research Fellowshipfor J. T. H. fron: the National Cancer C'enter. New York. NY.

2To whom requests for reprints should be addressed, at Department ofPathology, Norris 710. University of Southern California. 2025 /.onal Avenue.Los Angeles. CA 90033.

1The abbreviations used are: TAA. tumor-associated antigen: M Ab. monoclonal antibody; cDNA. complementary DNA; poly(A)* RNA. polyadenylated RNA;SDS. sodium dodecyl sulfate: SSC. standard saline citrate: PC'R. polymerase

chain reaction.

dominant epitopes, such as the repetitive sequences of thepolymorphic epithelial mucin found in carcinomas (7).

To detect other TAAs and to study the repertoire of thehumoral antitumor immune response in humans, we (8) andothers (9-11) have generated human MAbs from lymphocytes

that were spontaneously sensitized in vivo in cancer patients.Human MAbs of the IgM class were found to be predominantlyreactive against glycolipids. particularly gangliosides (12). Incontrast, our human IgG MAbs derived from regional lymphnode lymphocytes were reactive to internal epitopes apparentlyexpressed with neoplastic progression of the melanocyte (8) orthe colonie epithelial cell (13, 14). While the distribution ofthese antigens suggests biological as well as immunologicalimportance, the current technology for the production of humanantibodies falls short ofthat for mouse MAbs (15). Regardless,the number and type of antigens elucidated by a hybridomaapproach will always be limited to TAAs that elicit an antibodyresponse in the immunized host.

Molecular subtraction has been used to successfully isolategene sequences that are expressed in one tissue but not another.Usually, cDNAs derived from the cell type of interest wereextensively absorbed with mRNA sequences from a closelymatched cell type. In the mammalian system, the T-cell receptorwas thus identified by subtracting sequences between T- and B-lymphocytes (16). Similarly, a myoblast-specific sequence induced by 5-azacytidine that converts mouse embryonic fibro-blasts to myoblasts (17) and scrapie-modulated mRNAs were

described (18). Because molecular subtraction is independentof the immunogenicity of gene products, we propose that itmay complement the hybridoma approach for the detection ofTAAs. Indeed, it is potentially the most sensitive technologyavailable to date.

In this paper, we describe a new strategy involving molecularsubtraction to identify novel melanoma-associated gene se

quences. Here, subtraction was performed between 2 highlydisparate cell lines. Specifically. cDNA sequences from themelanoma cell line (MSM M-l) were exhaustively adsorbed

with mRNAs from a squamous cell carcinoma cell line of thelung (Lu-1). Since both are immortal human tumor cell lines,

housekeeping mRNAs and others associated with in vitro culture should be eliminated. On the other hand, subtraction withthe embryologically unrelated epithelial carcinoma cell lineshould not remove sequences characteristic of the mesenchymalmelanoma cells. Eventually, we hope that this technology maybe applied to the molecular characterization of immunologicallyimportant TAAs. In particular, we hope to identify TAAsrecognized by melanoma-reactive human IgG MAbs (19) orthose recognized by cytotoxic T-l\ mphocytes induced by active

specific immunotherapy in patients (20). This is an additionalrationale to use Lu-1 cells for adsorption, since both humanMAbs and cytolytic T-cells were not reactive to this cell line.

1418

on April 23, 2020. © 1991 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

MELANOMA mRNAs DETECTED BY MOLECULAR SUBTRACTION

MATERIALS AND METHODS

Human Tumor Cell Lines. The MSM M-l (abbreviated as M-l inthis paper), M-2. M-3. and M-4 melanoma cell lines, the Lu-1 cell line,and the skin fibroblast culture were established from biopsied tissues.Their cellular origin and tumorigenicity of the tumor cell lines wereconfirmed by standard histopathological evaluations of s.c. xenograftsestablished in nude mice. All cell lines were maintained in RPMImedium (GIBCO, Grand Island, NY) supplemented with 10% heat-inactivated fetal calf serum (HyClone, Logan, UT). The M21 melanomacell line was a generous gift from Dr. Donald L. Morton (University ofCalifornia at Los Angeles, Los Angeles, CA). All other cell lines(H138MG. Daudi. K562, 734B, HT-29, Hutu 80) were obtained fromthe American Tissue Culture Collection. The cell lines were tested every2 months and found to be free of Mycoplasma.

RNA Extraction and First Strand cDNA Synthesis. Cells were grownto confluence and detached with phosphate-buffered saline containing0.02% EDTA. Cytoplasmic RNA was isolated as described (21).Poly(/\r RNA was made by selecting over oligo(dT) cellulose (Collaborative Research, Inc., Bedford, MA).

Ten to 20 ^g of mRNA was dissolved in water and heated to 65°C

for 5 min and cooled on ice. cDNA was synthesized with the avianmyeloblastosis virus reverse transcriptase (Seikagaku America, Inc., St.Petersburg. FL), with oligo-dT at O.I mg/ml as primer (22). Integrityof first and second strand cDNA was tested by Southern hybridizationusing mouse actin and thymidine kinase probes.

Subtracted cDNA Probe Preparation. Subtractive hybridization wasperformed as described previously (23). Briefly, cDNA from the M-Icell line and a 10- to 60-fold excess mRNA from Lu-1 were coprecipi-tated in ethanol. Hybridization was performed in 16 n\ of 0.0375 Mpiperazine-/V,/V'-bis(2-ethane-sulfonicacid) (pH 6.9)-0.2% SDS-1.125M NaCl-0.19 M EDTA at 68°Cunder paraffin oil. Reactions achieved

C0t values of 1800 or greater. To remove single-stranded RNAs thatare in great excess, the reaction mixtures were digested with 10 ¿ig/mlribonuclease A and T, (Boehringer Mannheim, Indianapolis, IN) in0.24 Msodium phosphate buffer (pH 7.0). Selection for single-strandedmaterial was performed over hydroxyapatite equilibrated in 0.12 MNaH:PO4/Na2HPO4-0.1% SDS at 68°Cin a water-jacketed column.

Double-stranded material was eluted with 0.5 M NaH2PO4/Na2HPO4-0.1% SDS at 68°C.Authenticity of single-stranded nucleic acid was

confirmed by sensitivity to SI nuclease (23).For in situ screening, subtracted cDNA probes were radioactively

labeled by random primer synthesis (24).Library Screening. The M2I Agtll expression library was prepared

as previously described (25). The library was assumed to be representative, since 4% of the recombinants were shown to be reactive with ahuman 7-actin cDNA probe (data not shown; 26). Initial in situ plaquehybridizations, of unamplified library, were performed with duplicatenitrocellulose filters from 100-mm plates containing 1 x IO4 clones.DNA was denatured with 0.2 N NaOH-0.15 M NaCI for 2 min. Afterneutralizing with 0.4 M Tris (pH 7.5), the filters were rinsed with 2xSSC (Ix SSC is ISO mM NaCl-15 m\i sodium citrate) and baked at80°Cfor 2 h under vacuum. Hybridizations were carried out for 24 hat 42°Cin 0.25 M NaH2PO4/Na2HPO4 (pH 7.0)-0.25 M NaCl-50%

formamide-10% polyethylene glycol (M, 8,000-10,000; Fischer Scientific, Pittsburgh. PA) with 5% dextran sulfate and 7% SDS. Themembranes were washed briefly in 2x SSC and then incubated for 15min in 2x SSC-0.1% SDS at room temperature twice. Next, 2 highstringency 15-min washings were performed with 0.2x SSC-0.1% SDSat 42 and 60°C.Autoradiography was performed with an intensifyingscreen at -70°Cfor 24-72 h.

Selection of cDNA Clones. The cDNA recombinants selected fromthe first replicate plates of 2 x IO4were replated at single plaque density(100/100-mm plate) and reactive clones were reselected with the subtracted cDNA probe to ensure clonality and specificity of the in situcolony hybridization.

Purification of the Xgtll Phage DNA and Its Amplification. Xgtllphages were precipitated with 0.8 M NaCl-6.5% 0.25 M NaH2PO4/Na2HPO4 (pH 7.0)-0.25 M NaCl-50% formamide-10% polyethyleneglycol from the liquid lysates of infected Escherichia coli cells, strain

Y1090, grown on 100-mm Petri dishes (25). Phage DNA was isolatedby phenol:chloroform extraction followed by ethanol precipitation.

The cDNA inserts of phage DNA were amplified by the PCR withcommercial primers flanking the £coRI-cloning site (New EnglandBiolabs, Beverly, MA). The reaction mixtures consisted of 200 ^M eachof the 4 deoxynucleotide triphosphates, 1 n\\ primer, and 2.5 units ofthe Taq polymerase (Perkin-Elmer Cetus, Norwalk, CT) in 50 mMKC1-10 mM Tris-HCl (pH 8.3)-15 mM MgCl2-0.01% gelatin. Amplification was attained with 40 cycles in the thermocycler; each cycleconsisted of 30 s at 92°C,30 s at 56°C,and 3 min at 72°C.Under these

conditions, we were able to amplify inserts as large as 2 kilobases fromcontrol cDNA clones (data not shown) and all cDNA clones surveyedby PCR yielded inserts. The average fragment size was between 700and 900 base pairs in length.

The PCR products were deproteinized with phenol, precipitated withethanol, and «dissolvedfor digestion with the restriction endonucleasefcoRI. After electrophoresis through 1% low melting agarose gels, thecDNA inserts were identified and used without further purification forboth randomly primed radiolabeling and subcloning into M13.

Subcloning of cDNA Inserts into M13 for DNA Sequence Analysis.Amplified cDNA inserts were subcloned into the M13 vector for DNAsequence analysis (27). Briefly, the PCR insert was ligated into EcoR\-digested M14 mpl8 phage DNA with T7 DNA ligase according tomanufacturer's specifications (New England Biolabs). The DNA was

then transfected into competent E. coli, strain Xl-1. A clone containinga cDNA insert of the appropriate size (determined from the double-stranded replicative forni) was isolated for sequencing. Single-strandedtemplate DNA was prepared from phage particles in liquid lysates.

Automatic DNA sequencing was performed by the dideoxy-sequenc-ing method (model 370A sequencer. Applied Biosystems, Foster City,CA). Sequence analyses were performed with the GCG Sequence Analysis Package (University of Wisconsin, Madison, WI) on a VAX8250minicomputer (Digital Equipment Corp.).

Northern Blot Hybridization Analysis. Poly(A)+ mRNA from cultured

M21 cells were analyzed in 1.2% agarose denaturing gel prior toNorthern transfer. Briefly, RNA was dissolved in a loading bufferconsisting of 10% glycerol-0.2 mM EDTA-50% formamide-2.2 Mform-aldehyde-0.1% xylene cyanol. The gel was electrophoresed in 2.2 Mformaldehyde containing 20 mM boric acid-1 mM EDTA (pH 8.3).When electrophoresis was complete (6 h at 70 V or overnight at 20 V),the gel was soaked in running double-distilled H2O for 15 min. RNAwas then transferred to a nylon membrane (Amersham, ArlingtonHeights, IL) with lOx SSC transfer buffer for 18 h. Transferred RNAswere cross-linked to the filter by UV light (Stratagene, La Jolla, CA)for hybridization.

RESULTS

Hydroxyapatite Chromatograph)1 and Generation of Sub

tracted cDNAs. First strand cDNA was synthesized frompoly(A)+ mRNA derived from the M-l melanoma cell line. In

general, the yield was 3 pg from 10 //g mRNA. Because verysmall double-stranded molecules (<200 base pairs) did not bindwell to hydroxyapatite and eluted with single-stranded materialsat low ionic strength, they were removed from the startingcDNA preparation prior to subtraction. This was achieved bypassing the M-l cDNA.-mRNA hybrid molecules generated by

first strand synthesis over the hydroxyapatite column, recovering only the double-stranded material bound to the matrix. TheDNA molecules were then freed from their mRNA templatesby alkaline hydrolysis.

To prepare the subtracted probes that are enriched for melanoma-associated sequences, the first strand M-l cDNA wassubjected to 2 cycles of subtraction: first with a 10-fold excessand then a 60-fold excess by weight of poly(A)+ Lu-1 mRNA.Table 1 summarizes the recovery of subtracted M-l cDNAs.The ratio of single- to double-stranded material separated by

1419

on April 23, 2020. © 1991 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

MELANOMA mRNAs DETECTED BY MOLECULAR SUBTRACTION

Table I Recorery of first slrund cD\.4 after each cycle suhtraclire hybridizationRatio*

Suhtraclive M-l cDNA Lu-l mRNA (single-to Yield of cDNAh\ In nii/.u uin (jjg) (Mg)° douhle-siranded) (fig)'

First cycleSecond cycle

101.7

100(10)100(60)

0.22

1.7(17)0.5.1 (31)

°Numbers in parentheses, fold in excess.* Ratio of single- rersus double-stranded materials separated by hydroxyapatite

alter the reaction achieved a C()t (concentration in mol/liter x time in s) value of1800.

' Numbers in parentheses, '<

the hydroxyapatite column increased from 0.2 after the firstcycle to 2 after the second, indicating progressive enrichmentof M-l sequences with each reaction. Approximately 0.53 ^g(5%) was recovered from 10 ^g of M-l cDNA. Assuming totalrecovery, the resulting cDNA probe was enriched 20-fold forM-l sequences.

Fig. 1 is an autoradiogram of a 1% alkaline agarose gel ofthe M-l cDNAs at each step. The average size of the cDNAprobes was unaffected by alkaline hydrolysis (Fig. 1, lanes Aand lì),ranging from 300 to 4000 nucleotides in length. However, there was progressive reduction in size after each cycle ofsubtraction, possibly due to nonspecific strand breakage of thecDNA molecules induced by the elevated temperature in thehybridization reaction. After the first cycle, the cDNA molecules were about 2000 residues long (Fig. 1, lanes C and D). Incontrast. M-l cDNAs after the second subtraction (Fig. 1, laneE) were 300-500 bases long, as determined by electrophoresison a 3.5e; polyacrylamide gel (data not shown). This cDNA

material was radiolabeled by random primer synthesis and usedto screen the Xgtl l M21 melanoma cDNA library.

Selection of cDNA Clones. One hundred and fifty-six cDNAclones were selected from screening 2x10'' independent recom

binants of the M21 library with the subtracted cDNA probe.To identify clones that may be homologous for the samemRNAs. the cDNA insert of each clone was used in turn toprobe the other 155 clones by in situ hybridization. This was

ABODE

7.23-4.02-

2.41-

1.14-

0.52-0.22 •

Fig. 1. ¡MitesA-E. cDNA generated from the mRNA of M-l melanoma cellswas si/e fractionated on a l'i agarose gel under alkaline conditions. The sizemarkers are expressed in kilobases. Lane A. Ml cDNA/mRNA hybrid moleculesafter first strand cDNA synthesis and isolation on hydroxyapatite; this wasfollowed In alkali treatment to remove the mRNA strand in lane B. Lanes ( andI), single-stranded and double-stranded cON'As. respectively, after hybridi/ationwith l.u-l mRNA and separation on hydroxxapatite. Lain' /:'. M-l single-stranded

cPNA after a second round of hybridi/alion with I.u-l and fractionation usinghydroxyapatite. Si/e fraclionation of the M-l cDNA material on lane ¿"onan

acrvlamide gel documented a si/e range of 200-400 base pairs (data noi slum n).

Table 2 Selection ofcl).\A clones with the M-l subtracted cl).\A probe

Number of clones screened 20.000 ( 100%)Clones selected 156(0.78%)Genes identified 53Genes identified" 1 (24). 1 (15). I (II), 2 (7), 3 (6). I (5). 4 (4).2(3). 10(2). 28(1)

" Number of genes with this number of duplicate isolates. Numbers in paren

theses, numbers of duplicate isolates.

achieved by radiolabeling. using random primer synthesis, purified inserts from phage DNA amplified by the PCR (data notshown). In this manner. 53 discrete sequence populations withno apparent homology to one another were isolated. Table 2summarizes the steps involved and the number of duplicateisolates for each gene sequence.

The frequency of duplication for these genes ranged from 1to 24 (Table 2). Because 1-24 isolates/20.000 are consideredrare events, they follow poissonian statistics (28). Assumingthat the melanoma cell, as do other somatic cells, expresses IO5

molecules of mRNAs, the number of mRNA molecules per cellcould be estimated for each gene (29; Table 3). The calculatedranges of the number of each type of mRNA per cell arepresented in Table 3. These values are within the 99'V confi

dence limit. Strikingly, 55% (29 of the 53) of the genes hadonly 1 isolate, equivalent to 1-37 mRNA copies/melanomacell.

Expression of Sequence Families in Human Cells and Tissues.Of the IO5 mRNA molecules in a typical somatic cell, one-half

or more are of the rare class. Not all of these, however, aretissue-specific sequences. For example, c-myc and hypoxan-thine-guanine phosphoribosyl transferase are rare messages, at3-4 copies/IO5 cDNA clones. To better understand what typesof melanoma-associated genes could be identified with thismolecular approach, the cell- and tissue-specific expression ofthese 53 cDNA clones was examined in 19 different types ofhuman cells and tissues. In these experiments, the cDNA cloneswere grid plated, and in situ colony hybridization was performedwith radiolabeled cDNA probes derived from the human cellsand tissues. First strand cDNAs were synthesized from 2 ^g ofthe poly(A)* mRNAs with oligo-dT as primer (30). The average

size of these cDNA probes was 1 kilobase.All 53 genes were screened with cDNAs derived from the

melanoma M-l and M21 as well as the squamous cell carcinoma Lu-l cell lines. In addition, cultured human cells ofdifferent origins were studied, including 4 recently establishedmelanoma cell lines derived from métastases(M-2, M-3, M-4),a finite culture of human skin fibroblast (F4), a glioblastomacell line (U l38MG), 3 carcinoma cell lines (breast, 734B; colon,HT-29; and Hutu 80), a myeloid leukemia (K562) and a Burk-itt's lymphoma (Daudi). To evaluate the expression of these

genes in normal cells. cDNAs were prepared from different

Table 3 Estimated number of niR\A molecules tor the Kcne sequences identifiedin a typical melanoma cell

No. ofduplicateisolates1234567II1524Range

for20.000 cDNAclones0.01-7.40.10-9.30.3-11.00.7-12.61.1-14.21.5-15.72.0-17.14.3-22.86.9-28.213.3-39.8Kstimalcdcopies/cell"0.03-370.5-461.7-553.4-635.5-717.6-7810-8616-10031-13466-199

' The melanoma cell was assumed to contain 10' mRNAs.

1420

on April 23, 2020. © 1991 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

MELANOMA mRNAs DETECTED BY MOLECULAR SUBTRACTION

types of normal human tissues. Specifically, we used placenta(as a source of fetal tissues), brain (neuroectodermal derivation),spleen (a lymphatic organ), and liver (foregut and reticuloen-dothelial derivation). Furthermore, lymphoblasts generated byin vitro activation with concanavalin A from peripheral bloodmononuclear cells were included as an example of highly pro-

liferative but nonmalignant cells.The ;'/; situ plaque hybridization was performed with the

phage CND in excess to any complementary sequences thatmay be present in the cDNA probes from the cell lines andtissues. Therefore, this design directly compared the expressionof the 53 genes in each cell type. Differential expression of thecloned sequences among the cell types was approximated fromthese data. This procedure proved to be logistically more practical than Northern blot analyses, providing some basis toidentify potentially interesting cDNA clones.

All 53 genes were expressed by the M-l and the M21 melanoma cell lines. In contrast, only 23 cDNA clones were notreactive with the cDNA probe derived from the Lu-1 cell line(Table 4). The remaining 30 genes were positive (Table 5),although the intensities tended to be much lower than the othercell types. These most likely represent transcriptional geneproducts expressed by both M-l and Lu-1 but are especiallyabundant in the former. In addition, most of the 53 genes werefound in several types of human cells or tissues. However, as agroup, the genes that were absent from the Lu-1 cell line (Table4) were more restricted in their distribution than those thatwere present (Table 5) (x2 = 32.3, with 1 d.f.; for P = 0.005, x2

is 7.88).Partial Sequence Analyses of Selected cDNA Clones. To better

understand the type of mRNAs that were identified by thesubtracted M-l cDNA probe, partial DNA sequencing wasperformed on 20 cDNA clones. This information most directlyidentifies a gene sequence to be novel (not yet described), known(complete identity to a described gene), or related by partialhomology to an existing gene or gene family. In general, 300-450 bases were sequenced for each of the selected cDN A clones,which was sufficient for determining homology to known genes.To accomplish this, gene sequence information from each ofthe cDNA clones in Tables 6 and 7 was compared to Genbankand EMBL databases with the GCG Sequence AnalysisPackage.

Table 6 summarizes the data concerning 12 clones that werepresent in our panel of melanoma cell lines but were absent inthe Lu-1 cell line. We selected cDNA clones that were represented with frequencies between 1 and 11 duplications. Six ofthese were novel, representing genes expressed by melanomasthat have not been previously described. Six cDNA clonescorresponded with known proteins (cyclophilin, a poly(A)-bind-

ing protein, and 4 mitochondrial proteins). Their importancein melanoma will be discussed below. For comparison. 8 genesexpressed in both melanomas and the Lu-1 cell line were alsosequenced (Table 7). Six had 1-2 duplicate isolates. Six of the8 were previously described genes, encoding for proteins suchas tropomyosin and fibronectin. Two of these, clones 102 and127, appeared to be novel sequences.

To verify that these cDNA sequences represent functionallyactive genes, the mRNA transcripts of 11 clones were identifiedby Northern blotting. The molecular weights of the transcriptsare summarized in Tables 6 and 7. Fig. 2 shows the autoradiogram with the transcripts for 5 of the novel cDNA clones(clones 50, 58, 127, 151, and 159). The DNA sequence information concerning these clones is consistent with their beingfunctionally active genes. Long open reading frames were identified in all of these clones.

I ranscriptional Analysis of Clone 50. Verification of the authenticity of the subtractive hybridization probe generation wasaccomplished by Northern blot analysis of selected cDNAclones. An example of this, for cDNA clone 50. is illustratedby Northern blot analysis of several human tissues (Fig. 3).

Consistent with the initial expression survey data representedin Table 6, clone 50 hybridizes to a transcript of 6 kilobases(arrowhead) found in the melanoma cell lines tested and absentfrom the squamous cell lung carcinoma. Although equalamounts of poly(A)-selected RNA were hybridized in this experiment, overexposure of the blot to normalize relative actinintensities still revealed no hybridization (data not shown).

In addition, it can be seen that clone 50 is expressed inprimary fibroblasts from one of the melanoma patients (Fig. 3;M-4 compared to F4 fibroblasts), as well as in two tumor cell

lines, U138MG (glioblastoma) and B734 (breast carcinoma).Experiments are in progress to characterize the product of thisgene and its complete gene sequence.

Table 4 Expression of the 23 gene sequences absent in Lit-l in human tumor cell lines and normal tissttea

cDNA clones

10 14 16 25 27 37 38 50 55 58 61 77 84 96 106 113 124 125 139 142 151 154 159

Cell linesMelanomaFibroblast,

F4Colon

carcinomaHT-29Myeloid

leukemia.K562Normal

tissuesPlacentaSpleeno

o •o••

•••o•O • •00

• • ••o

o •ooooooooooooo00o00oooC'O

(ooo

to

t}

Oo»

0»

ooooo00oo0oo0oo•0o•C0•fjCC0C)0ooo0o•

•.bindingofradiolabe

1421

; O. apparent absence of binding.

on April 23, 2020. © 1991 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

MELANOMA mRNAs DETECTED BY MOLECULAR SUBTRACTION

1"5ÈSi1~~~113tu5£5ì5•E="»••ft^-.511"C-C^||V)Uzi(«OJoZOow,r

1!f*ì2r-ri£i/-,C30(N=£ZINOv00Xp=s00

sO2$r*ìSMN0

nX«N-T••••••O«•ooo»ooo•••••oo«••••••o«*

*•*••ooo»ooo•••••••••••••00«••••••00•••••ooo••••••oo*

• •»****•*•••••••••ooo»oo«*

* ••*•

••••ooo«o»»••oo«»»«••oo«o««••*•••••••*••*••••••ooo««««••••••••••o««««3OoaZ

2,"ttTf

'oîa.!*jjliiü22¿üí"§'Sí=

7 il3á•

oo«o«oo•

•o«oo«oo«oo•

oo«•

000•

•oooooo•

ooo•

ooo•

o«*•

•0«•

•••••o*o«oo•

•oo•

•oo0»00•

•••o»»

o•

•0*•

••o•••••

••••

•••0««

0rl

sV,"i

ilU0.'S

Ä ^>.3

?,1 73.•'-.o*~^ ~'5E

2?&Of.se00000•

0«00•

•••o•

o«oo•

oooo•

•••0•

oooo•

••00•

••00•

o«oo•

••••••••••o«««o«ooo•

••••o«o««•

oo•••

o ••••

o••••

•*•

•0•••

o**«•

•o««•

««««0«0««•

O A ftA\^J WW'A

rioÄ*W W vJW!í!

«5

S.i« O

*- ÎÕ ^-Cä

a.s||s=J25 "E >..=i a.x y:—-JzobC•5C"Hs1—=

ur:c.AdC~2uo^oÕor

tissuilo—¡l\•=r;00Cc"^•3

DISCUSSION

The concept of what constitutes "tumor antigens" has evolved

over the years, depending in part upon the technologies available for their detection at the time. The first tumor markerswere produced and secreted by tumor cells in sera at levelsdetectable by biochemical procedures. Subsequently, motivatedby their potential application for therapy and diagnosis, surfaceTAAs were specifically sought by serological techniques including the use of xenoantisera rendered specific by absorption andthe generation of mouse and human MAbs. Studies with humanMAbs revealed that internal rather than surface antigens arethe most immunogenic in human (13-15, 19). Indeed, themajority of IgG MAbs derived from spontaneously sensitizedlymphocytes of patients with malignant melanoma (8) or coloncarcinoma (14) appeared to define a different class of TAAs.These internal antigens are present at levels lower than otherspreviously described with other technologies (13). Interestingly,some are apparently expressed with neoplastic transformation(8, 14).

Recently, recombinant DNA technology has been applied tothe characterization of genes differentially expressed by onetissue but not another. In general, an expression library waseither differentially screened or screened with a subtractedprobe. Because of the unprecedented level of sensitivity of amolecular approach, most investigators had restricted theiranalysis to comparing closely related cell types, to limit thenumbers of potentially interesting cDNA clones that need tobe characterized. In fact, despite its proven applicability, thelaboriousness of the screening procedure has cast serious doubtsin many as to the general usefulness of a molecular approach.

In this paper, we report our experience using molecularsubtraction for the isolation of melanoma-associated gene sequences. In particular, we have taken advantage of severaltechnical advances to screen and subsequently characterize arelatively large number of clones. We describe here a prototypicscreening strategy that is rapid and practical.

Because the cell types we have selected are so disparate intheir derivation, a large number of cDNA clones (0.76%) wereidentified by the subtracted probe. The first level of screeningwas to identify redundant clones encoding the same sequences.To accomplish this, the cDNA insert of each clone was amplified by the PCR reaction as described in "Materials and Methods." Each cDNA insert was then used in turn to probe the

other 156 clones in situ hybridization. In this manner, 53discrete gene sequences were identified.

Next, we utilized the in situ plaque hybridization assay toidentify cDNA clones that are "melanoma specific" from those

that are clearly expressed by both cell lines. Although this assaywould only provide a first approximation of the differentialexpression of these cDNA clones by the various cell types, itwas very easy to perform and allowed us to eliminate commonclones. In fact, more than one-half (30 of the 53) of the discretecDNA clones were found to be commonly expressed. Thus,serial exhaustive hybridization with excess mRNAs is by itselfnot sufficient to generate tissue-specific probes.

To develop additional insight into the clones selected, weperformed limited sequence analysis on both melanoma-restricted (Table 6) and common clones (Table 7). The knowngenes restricted to melanoma appeared as predicted by ourstatistical analysis to be rare proteins: mitochondria! proteins,a poly(A)-binding protein, and cyclophilin. a protein usuallyassociated with T-lymphocytes (31 ). Incidentally, Northern blot

1422

on April 23, 2020. © 1991 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

MELANOMA mRNAs DETECTED BY MOLECULAR SUBTRACTION

Table 6 Identification of selected gene sequences absent in the Lu-1 cell line

The RNA transcripts for the gene sequences selected were identified by Northern blot analysis. The cDNA inserts of purified phage DNAs were amplified by thePCR. After radiolabeling by random primer synthesis, they were used to probe 2 ng of M-l poly(A)1"mRNA resolved on a 1.2% agarose gel.

No. of cDNA mRNA transcript(s)cDNA duplicate insert in M-l cellsclone isolates (base pairs) Homologous gene(kilobases)50

113151

15425586127

125159139 (124 1500

Novel 6400 Human cvclophilin (31)" 1.0.2.1,3.1

450 Novel >10650 Novel ND*

400 Human mitochondria! protein (32) ND700 Novel 2.4

1000 Novel NDt 900 Human mitochondria! protein (32) 2.7\ 700 Human mitochondria! protein (32) 1.2.2.4,3.7i 800 Novel 0.8,1.9,3.55 500 Human polv( <\)-binding protein (33)09,25,41400

Human mitochondria! protein (32)ND"Ref. number.

h ND. not determined.

Table 7 Identification oj selected gene sequences present in the l.u-l cell line

The RNA transcripts for the gene sequences selected were identified by Northern blot analysis. The cDNA inserts of purified phage DNAs were amplified by thePCR. After radiolabeling by random primer synthesis, they were used to probe 2 ¿igof M-l poly(A)* mRNA resolved on a 1.2°cagarose gel.

cDNAclone94951021036310192127No.

ofduplicateisolates1]1122415cDNAinsert(basepairs)1500700400500400800800500HomologousgeneHuman

fibroblast tropomyosin(34)"Rat

clathrin(35)NovelHuman

fibronectin(36)Humanglyceraldehyde-3-phos-phate

dehydrogenase(37)Humanribosomal SI 1 protein(38)HumantissueplasminogenactivatorNovelmRNA

transcript(s)in M-l cells(kilobases)ND*NDNDND1.41.2ND2.8

" Ref. number.* ND. not determined.

analysis with the cyclophilin clone (113) reveal a unique melanoma-associated mRNA transcript4 (Table 6). In contrast, at

least 3 of the known genes common to both cell types werestructural proteins: tropomyosin, fibronectin, and ribosomalSII protein. Interestingly, glyceraldehyde-3-phosphate dehydrogenase, also identified, was previously noted to be differentially expressed by human adenocarcinomas, as compared totheir normal counterparts (39).

The amounts of cDNA that could be amplified by PCR fromthe cDNA clones were sufficient for subcloning and DNAsequence analysis. Sequencing was performed by the AdvancedBiosystems automated DNA sequenator, yielding 350-500 basepairs/run, which was well suited for homology searches andgene analysis. The combined PCR and automated DNA sequence analysis resulted in low levels of error. For example,<0.1% variation in sequence data was observed in the amplifiedmaterials from the clone encoding the human cyclophilin gene(31). The combined use of these technologies enabled us toidentify 8 of the 20 cDNA clones to be novel at relatively earlystages of clone characterization.

Our studies confirmed that molecular subtraction is an extremely powerful tool and demonstrate that it can be adaptedto identify a variety of new rare-class genes expressed by melanoma cells. A molecular approach offers several advantagesover those previously developed: (a) it does not require any apriori sequence information of the genes to be sought. Thus,molecular subtraction can be applied to identify TAAs in thesame manner as hybridoma technology: (b) molecular tech-

4J. T. Hutchins. R. J. Deans, M. S. Mitchell. C. Uchiyama. and J. Kan-

Mitchell, unpublished information.

niques are intrinsically the most sensitive to date. Indeed, themajority of cDNA clones that we detected were rare mRNAsequences (Tables 2 and 3); (c) partial sequence information ofthe cDNA clones was readily attainable, thus permitting directidentification of 8 novel melanoma-associated genes after

5kb-

2kta-

Fig. 2. Northern analysis of poly(A) selected RNA from M21 melanoma cellsusing hybridization probes derived from isolated melanoma cDNA clones. Priorto hybridization. 2 ¿¡gof poly(A)* mRNA from M21 cells was electrophoresed/lane on a 1.2cr agarose gel. ran under denaturing conditions in formaldehyde,

prior to nylon membrane transfer. Each lane was hybridized to separate hybridization probes representing known cDNA sequences (clone 101) and novel cDNAsequences (clones 50, 58. 127. 151. 159). Transcript sizes were estimated by theirrelative migration to 28S (5.1 kilobases) and 18S (1.9 kilobases) ribosomaÕRNAbands, kb. kilobase.

1423

on April 23, 2020. © 1991 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

MELANOMA mRNAs DETECTED BY MOLECULAR SUBTRACTION

- SÌ_a.91

I SS 2 *3 pr» m i"o •-h»(—o z> ODi

¡••{^••••••••••••lpActin

Fig. 3. Northern analysis of mRNA from various human cell lines and tissue.The ¿VoRlfragment of clone 50 and an actin cDNA hybridization probe (26)were used as hybridization probes. The cDNA clone 50 fragment identifies atranscript of approximately 6 kilobases in length (arrow).

screening only 2 x IO4 recombinants of the Xgtll expression

library (Tables 6 and 7); (d) the gene products can be analyzedon both the RNA transcript level and protein level. Oligonucle-otides specific for these genes can be synthesized to be used fortheir characterization. Additionally, fusion proteins can bereadily produced which are potentially useful in immunodi-agnostic surveys.

As with all technologies, this approach may not be able todetect all TAAs. Genes whose expression is regulated quantitatively rather than qualitatively will be missed. Genes withminor alterations in sequences such as point mutations will alsonot be detected, although we are not aware of an alternativesystematic method for their detection either. Clones containingreiterated sequences, e.g., an Alu repeat in the 3'-untranslated

region, would also be eliminated on that basis. Finally, messagessmaller than 100 base pairs or those with a large amount ofsecondary structure would not be represented.

In summary, we have shown that novel tumor-associatedgene sequences, restrictedly expressed by a particular type ofneoplastic cell, can be identified by a molecular approach. Inour studies, eight new human gene sequences were identifiedand expressed in melanoma. We are currently characterizingthese novel cDNA clones to understand the nature of theirtissue-specific expression, regulation, and function and to determine whether any are immunogenic to humans.

ACKNOWLEDGMENTS

We would like to thank Dr. Linda Walling for her help in establishingthe conditions for molecular subtraction, without which the projectwould not have been initiated. We would also like to thank Dr. MimiC. Yu for her help with the statistical analysis. We would like toacknowledge the excellent technical help of JoAnne Goodnight for thelibrary construction and Steve Chamberlin for assistance in automatedDNA sequencing.

REFERENCES

1. Edelman. G. M.. and Galley. J. M. The nature of Bence-Jones proteins.Chemical similarities to polypeptidc chains of myeloma globulins and normal7-globulins. J. Exp. Med.. Ì16:207-227. 1962.

2. Abelev, G. I.. Perova. S. D.. Khramkova. N. !.. Postnikova. Z. A., and Irlin.I. S. Production of embryonal «-globulinby transplantable mouse hepatomas.Transplantation (Baltimore). /: 174-180. 1963.

3. Gold. P.. and Freedman. S. Ü.Demonstration of tumor-specific antigens inhuman colonie carcinoma by immunological tolerance and absorption techniques. J. Exp. Med.. 121: 439-462. 1965.

4. Metzgar. R. S.. and Mitchell. M. S. (eds). Human Tumor Antigens andSpecific Tumor Therapy. UCLA Symposia on Molecular and Cellular Biology. New Series. Vol. 99. New York: Alan R. Liss. Inc.. 1989.

5. Reisfeld. R. A., and Cheresh. D. A. Human Tumor Antigens. Adv. Immunol..40:323-377, 1987.

6. Kan-Mitchell. J. Human monoclonal antibodies. In: R. S. Metzgar and M.S. Mitchell (eds.). Human Tumor Antigens and Specific Tumor Therapy,UCLA Symposia on Molecular and Cellular Biology. New Series. Vol. 99,pp. I05-Ì14.New York: Alan R. Liss. Inc.. 1989.

7. Gendler. S.. Burchell, J. M.. Duhig. T.. Lamport. D.. White, R.. Parker. M.,and Taylor-Papadimitriou. J. Cloning of partial cDNA encoding differentiation and tumor-associated mucin glycoproleins expressed by human mammary epithelium. Proc. Nati. Acad. Sci. USA. 84: 6060-6064. 1988.

8. Kan-Mitchell. J.. White. W. L.. and Mitchell, M. S. Tumor-reactive humanIgG monoclonal antibody from a melanoma patient. Cancer Res.. 49: 4536-4541. 1989.

9. Irie. R. F.. Chandler. P. J., and Morton. D. L. Melanoma, gangliosides andhuman monoclonal antibodies. In: R. S. Metzgar and M. S. Mitchell (eds.).Human Tumor Antigens and Specific Tumor Therapy, UCLA Symposia onMolecular and Cellular Biology. New Series, Vol. 99. pp. 115-126. NewYork: Alan R. Liss, Inc.. 1989.

10. Livingston. P. O., Natali, E. J., Calves, M. J., Stocken. E.. Oettgen, H. F.,and Old. L. J. Vaccines containing purified GM2 ganglioside elicit GMzantibodies in melanoma patients. Proc. Nati. Acad. Sci. USA. 84: 2911-2915. 1987.

11. Haspel, M. V.. McCabe. R. P.. Pomato. N.. Hoover, H. C., and Hanna, M.G.. Jr. Corning full circle in the immunotherapy of colorectal cancer: vaccination with autologous tumor cells-lo human monoclonal antibodies-to development and application of a generic tumor vaccine. In: R. S. Metzgar andM. S. Mitchell (eds.). Human Tumor Antigens and Specific Tumor Therapy.UCLA Symposia on Molecular and Cellular Biology, New Series, Vol. 99.pp. 335-344. New York: Alan R. Liss.. 1989.

12. Lloyd. K. O.. and Old, L. J. Human monoclonal antibodies to glycolipidsand other carbohydrate antigens: dissection of the humoral immune responsein cancer patients. Cancer Res.. 49: 3445-3451. 1989.

13. Kan-Mitchell. J., Forment!. S. C., and Mitchell. M. S. Human monoclonalantibodies reactive with colon carcinoma: identification by a novel screeningprocedure. J. Clin. Lab. Anal., 3: 41-49. 1989.

14. Formenti, S. C., Mitchell. M. S., Taylor, C. R., Lipkin. M., Jernstrom, P.H.. and Kan-Mitchell. J. Reactivity of a human monoclonal antibody againstcarcinomas and other lesions of the colon. Cancer Immunol. Immunother.,28: 296-300. 1989.

15. Kan-Mitchell. J.. and Mitchell. M. S. Human monoclonal antibodies andtheir usefulness in diagnosis. In: H. Z. Kupchik (ed.).. In I'itro Diagnosis of

Human Tumors using Monoclonal Antibodies. Immunology Series, Vol. 39,pp. 289-304. New York: Marcel Dekker. Inc., 1988.

16. Hedrick, S. M., Cohen, D. !.. Nielsen, E. A., and Davis, M. M. Isolation ofcDNA clones encoding T cell-specific membrane-associated protein. Nature(Lond.). 308: 149-153, 1984.

17. Davis, R. L., Weintraub, H.. and Lassar, A. B. Expression of a singletransfected cDNA converts fibroblasts to myoblasts. Cell. SI: 987-1000,1987.

18. Duguid. J. R.. Rohwer, R. G., and Seed, B. Isolation of cDNAs of scrapie-modulated RNAs by subtractive hybridization of a cDNA library. Proc. Nail.Acad. Sci. USA. 85: 5738-5742, 1988.

19. Kan-Mitchell. J.. Imam. A., Kempf. R. A.. Taylor, C. R., and Mitchell, M.S. Human monoclonal antibodies directed against melanoma tumor-associated antigens. Cancer Res., 46: 2490-2510, 1986.

20. Mitchell. M. S.. Kan-Mitchell. J.. Kempf. R. A.. Harel. W.. Shau. H.. andLind. S. Active specific ¡mmunotherapy for melanoma: phase 1 trial ofallogene«:Usâtesand a novel adjuvant. Cancer Res.. 48: 5883-5893. 1988.

21. Maniatis. T.. Fritsch, E. F.. and Sambrook, J. Molecular Cloning: A Laboratory Manual. Cold Spring Harbor. NY: Cold Spring Harbor Laboratory.1982'

22. Krug. M. S., and Berger. S. L. First-strand cDNA synthesis primed witholigo(dT). Methods Enzymol., 152: 316-325. 1987.

23. Timberlake. W. E. Developmental gene regulation in Aspergillus nidulans.Dev. Biol., 78: 497-510. 1980.

24. Feinberg. A. P.. and Vogelstein. B. A Technique for radiolabeling DNArestriction endonuclease fragments to high specific activity. Anal. Biochem.,1)2: k-n. 1983.

25. Huynh. T. V., Young. R. A., and Davis. R. W. Constructing and screeningcDNA libraries in XgtlO and Xgtll. In: D. M. Glover (ed.), DNA Cloning,Vol. 1, pp. 4-78. Washington, DC: 1RL Press. Ltd., 1985.

26. Chou, C. C., Davis, R. C., Fuller. M. L., Slovin. J. P.. Wong, A.. Wright. J.,Kania, S.. Shaked. R., Gatti, R. A., and Salser, W. A. i'-Actin: unusualmRNA 3'-untranslated sequence conservation and amino acid substitutionsthat may be cancer related. Proc. Nati. Acad. Sci. USA. 84: 2575-2579.1987.

27. Messing. J. Newr M13 vectors for cloning. Methods Enzymol.. 101: 20-78,

1983.28. Diem. K.. and Lentner. C. Scientific Tables, p. 108. Ardsley. NY: Ciba-Geigy

Ltd.. 1971.29. Hoel. P. G.. S. C.. and Stone, C. J. Introduction to Probability Theory, p.

75. Boston: Houghton Mifflin Co.. 1971.30. Gubler. V.. and Hoffman. B. J. A simple and very effective method for

generating cDNA libraries. Gene. 25: 263-269. 1983.31. Haendler. B.. Hofer-VVarbinck. R.. and Hofer, E. Complementary DNA for

human T-cell cyclophilin. EMBO J.. 6: 947-950. 1987.32. Anderson. S.. Bankier. A. T.. Barrell. B. G.. deBruijn. M. H. L., Coulson. A.

1424

on April 23, 2020. © 1991 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

MELANOMA mRNAs DETECTED BY MOLECULAR SUBTRACTION

R.. Drouin. J.. Eperon. I. C Nicrlich. D. P.. Roe. B. A.. Sanger. F.. Schreier.P. II.. Smith. A. J. H.. Siaden. R.. and Yound. I. G. Sequence and organization of the human mitochondria! genome. Nature (Lond.). 290:457-465,

1981.33. Grange. T.. Martins de Sa. C'.. Odds. J.. and Pictet. R. Human mRNA

polyadenylate binding protein: evolutionary conservation of a nucleic acidbinding motif. Nucleic Acids Res.. IS: 4771-4787. 1987.

34. Macleod. A. R.. Talbot. K.. Smillie. L. B.. and Houlker. C. Characterizationof cDNA defining a gene family encoding TM30pl. a human fibroblasttropnmyosin. J. Mol. Biol.. 194: 1-10. 1987.

35. Kirchhausen. T.. Harrison. S. C.. Chow. E. P.. Mattaliano. R. J.. Ramachan-

dran. K. L.. Smart. J.. and Brosius. J. Clathrin heavy chain: molecular

cloning and complete primary structure. Proc. Nati. Acad. Sci. USA. 84:8805-8809. 1987.

36. Kornblihtt. A. R.. Umezawa, K.. Vibe-Pedersen. K., and Baralle. F. E.Primary structure of human fibronectin: differential splicing may generate atleast 10 polypeptides from a single gene. EMBO J.. 4: 1755-1759. 1985.

37. Accari. P., Martinelli. R.. and Salvatore. F. The complete sequence of a fulllength cDNA for human liver glyceraldehyde-3-phosphate dehydrogenase:evidence for multiple mRNA species. Nucleic Acids Res.. 12: 9179-9189,1984.

38. Loti. J. B.. and Mackie. G. A. Sequence of a cloned cDNA encoding humanribosomal protein SII. Nucleic Acids Res.. 16: 1205. 1988.

39. Schek. N., Hall. B. L.. and Finn. O. J. Increased glyceraldehyde-3-phosphatedehydrogenase gene expression in human pancreatic adenocarcinoma. CancerRes!. 48: 6354-6359. 1988.

1425

on April 23, 2020. © 1991 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

1991;51:1418-1425. Cancer Res   Jeff T. Hutchins, Robert J. Deans, Malcolm S. Mitchell, et al.   Identified by Molecular SubtractionNovel Gene Sequences Expressed by Human Melanoma Cells

  Updated version

  http://cancerres.aacrjournals.org/content/51/5/1418

Access the most recent version of this article at:

   

   

   

  E-mail alerts related to this article or journal.Sign up to receive free email-alerts

  Subscriptions

Reprints and

  .pubs@aacr.orgDepartment at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

  Permissions

  Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/51/5/1418To request permission to re-use all or part of this article, use this link

on April 23, 2020. © 1991 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from