Identification of Differentially Expressed Genes in Human Prostate … · Identification of...

[CANCER RESEARCH 60, 1677–1682, March 15, 2000]

Identification of Differentially Expressed Genes in Human Prostate Cancer UsingSubtraction and Microarray 1

Jiangchun Xu,2 John A. Stolk, Xinqun Zhang, Sandra J. Silva, Raymond L. Houghton, Masazumi Matsumura,Thomas S. Vedvick, Kevin B. Leslie, Roberto Badaro, and Steven G. ReedCorixa Corporation, Seattle, Washington 98104 [J. X., J. A. S., X. Z., S. J. S., R. L. H., M. M., T. S. V., S. G. R.]; ImmGenics Pharmaceuticals Inc., Vancouver, British Columbia,V67 123 Canada [K. B. L.]; Department of Infectious Disease, Federal University of Bahia, Salvador, Bahia, Brazil [R. B.]; and Department of Pathobiology, University ofWashington, Seattle, Washington 98195 [S. G. R.]

ABSTRACT

We have identified human prostate cancer- and tissue-specific genesusing cDNA library subtraction in conjunction with high throughputmicroarray screening. Subtracted cDNA libraries of prostate tumorsand normal prostate tissue were generated. Characterization of sub-tracted libraries showed enrichment of both cancer- and tissue-specificgenes. Highly redundant clones were eliminated by colony hybridiza-tion. The remaining clones were selected for microarray to determinegene expression levels in a variety of tumor and normal tissues. Clonesshowing overexpression in prostate tumors and/or normal prostatetissues were selected and sequenced. Here we report the identificationof two genes,P503Sand P504S, from subtracted libraries and a thirdgene,P510S, by subtraction followed by microarray screening. Theirexpression profiles were further confirmed by Northern blot, real-timePCR (TaqMan), and immunohistochemistry to be overexpressed inprostate tissues and/or prostate tumors. Full-length cDNA sequenceswere cloned, and their subcellular locations were predicted by a bioin-formatic algorithm, PSORT, to be plasma membrane proteins. Thegenes identified through these approaches are potential candidates forcancer diagnosis and therapy.

INTRODUCTION

Prostate cancer is the second leading cause of death among men inthe United States. The American Cancer Society estimated that therewould be approximately 179,000 new cases diagnosed with prostatecancer and 37,000 deaths in 1999 (1). Little is known about thegenetic events in malignant transformation of prostatic cells. This isdue in part to the cellular heterogeneity of the prostate tissues and alack of genetic information from systematic analysis. Tissue- and/orcancer-specific genes can be used as markers for screening, diagnosis,prognosis, therapeutic monitoring, and/or follow-up for early indica-tion of relapse. These genes or gene products can also be targeted byvarious agents,e.g.,antibodies or drugs, for cancer therapy. Further-more, therapeutic vaccines may be developed from these. The mostcommonly used marker for diagnosis and prognosis of prostate canceris serum PSA.3 This has now replaced the previously used but lessspecific prostate acid phosphatase test (2). However, PSA levels arefrequently elevated in conditions such as BPH and prostatitis (3);therefore, PSA is not cancer specific. Other prostate markers reportedinclude prostate acid phosphatase (4), prostate-specific membraneantigen (5, 6), prostate inhibin peptide (7), PCA-1 (8), PR92 (9),prostate-associated glycoprotein complex (10), prostate-mucin anti-gen (11), 12-lipoxygenase (12), and p53 (13). The above-mentioned

markers are not entirely tissue and/or cancer specific; hence, improvedmarkers are needed for the diagnosis, prognosis, and therapy ofprostate cancer.

Numerous approaches have been used to dissect the genetic com-position of prostate and prostate cancer. Methods used to identifyoverexpressed genes include EST sequencing (14, 15), serial analysisof gene expression (16, 17), and differential display PCR (18). How-ever, none of these techniques provides a complete, systematic, andreliable comparison of gene expression between tissue types. Expres-sion cloning using cancer patient serum (19) or T-cell lines or clones(20) from patients has identified a panel of genes that may be immu-nologically relevant, although none have yet been validated, and theprocess is not very efficient. We have analyzed the genetic composi-tion of prostate and prostate cancer by combining a cDNA librarysubtraction method (21) and a microarray high throughput screeningprocedure (22). Here we report the discovery of three prostate tissue-and/or cancer-specific genes using these approaches.

MATERIALS AND METHODS

Tumor Samples and RNA Preparation. All normal and tumor tissuesamples obtained from various clinical sources were accompanied by clinicalinformation and pathological reports and were histologically confirmed bypathologists. The tissues were frozen in liquid nitrogen and homogenized witha polytron (Kinematica). Total RNA was prepared using Trizol reagent (LifeTechnologies, Inc.). Poly(A)1 RNA was then purified using a Qiagen oligotexspin column mRNA purification kit.

cDNA Library Subtraction. cDNA library subtraction was performedusing the protocol described by Haraet al. (21), with modifications. Briefly,cDNA libraries were first constructed with the Superscript Plasmid System forcDNA Synthesis and Plasmid Cloning Kit (Life Technologies, Inc.) usingpoly(A)1 RNA from prostate tumors, normal prostate tissue, and normalpancreatic tissue. The control cDNA library (70–150mg; driver) was digestedwith EcoRI, NotI, andSfuI, followed by filling in with the DNA polymeraseKlenow fragment. Driver DNA was labeled with Photoprobe biotin (VectorLaboratories) and dissolved in 23ml of H2O. To prepare prostate (tracer) DNA,cDNA libraries of normal or tumor prostate were digested withBamHI andXhoI, phenol chloroform-extracted, passed through Chroma spin-400 columns(Clontech), ethanol-precipitated, and dissolved in 5ml of H2O. Tracer DNAwas mixed with 15ml of driver DNA and 20ml of 23 hybridization buffer [1.5M NaCl, 10 mM EDTA, 50 mM HEPES (pH 7.5), and 0.2% SDS], overlaid withmineral oil, heat-denatured, and incubated at 68°C for 20 h. The reactionmixture was then incubated with streptavidin and extracted with phenol/chloroform four times. This hybridization process was repeated with an addi-tional 8ml of driver DNA at 68°C for 2 h. Subtracted cDNA was ligated intothe chloramphenicol-resistant pBC SK1 (Stratagene) and transformed intoElectroMax Escherichia coliDH10B cells by electroporation to generate atracer-specific subtracted cDNA library.

Library PCR. Library DNA (50 ng) was used as a template for PCRamplification of humanb-actin at 18, 23, 28, and 33 cycles. PCR productswere run on a 1% agarose gel and stained with ethidium bromide. Primers usedfor b-actin were as follows: (a) 59 primer, 59-ACCCCGTGCTGCTGACC; and(b) 39 primer, 59-AGGAAGGAAGGCTGGAAGAGT.

Colony Hybridization and Northern Blot Analysis. For colony hybrid-ization, lifts were prepared with randomly picked colonies from subtractedlibraries using 132 mm Hybond-N filters (Amersham Pharmacia Biotech). For

Received 9/1/99; accepted 1/19/00.The costs of publication of this article were defrayed in part by the payment of page

charges. This article must therefore be hereby markedadvertisementin accordance with18 U.S.C. Section 1734 solely to indicate this fact.

1 Supported in part by National Cancer Institute Grant CA80518.2 To whom requests for reprints should be addressed, at Corixa Corporation, 1124

Columbia Street, Suite 200, Seattle, WA 98104. Phone: (206) 754-5798; Fax: (206)754-5917.

3 The abbreviations used are: PSA, prostate-specific antigen; BPH, benign prostatehyperplasia; EST, expressed sequence tag; HGK-1, human glandular kallikrein 1; mAb,monoclonal antibody; poly(A)1 RNA, polyadenylated RNA; TCR, T-cell receptor;CCOII, cytochromec oxidase subunit II.

1677

on July 28, 2021. © 2000 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

http://cancerres.aacrjournals.org/

Northern blot analysis, 10mg of total RNA were run out on formaldehydedenaturing gel, transferred to Hybond-N membrane, cross-linked, stained withmethylene blue, and photographed. Radioactive32P-labeled cDNA probeswere prepared using a Ready-To-Go DNA labeling kit (Amersham). Filterswere prehybridized in hybridization solution [1% BSA, 1 mM EDTA, 0.5 M

NaHPO4 (pH 7.2), and 7% SDS] for 30 min at 65°C and replaced with freshhybridization solution containing radioactive cDNA probe and hybridizedovernight at 65°C. Washes were also carried out at 65°C with wash solution 13SCP (0.1M NaCl, 30 mM Na2HPO4, 1 mM EDTA, 1% N-lauroylsarcosine).

Microarray. mRNA expression of cDNA clones from subtracted librarieswas determined using a high throughput microarray approach (22). Coloniesthat were negative from prescreening were randomly picked from subtractedlibraries and PCR-amplified for 30 cycles with vector-specific primers accord-ing to a protocol suggested by Incyte. PCR products were then arrayed ontoglass slides using Incyte patented chemistry. The arrayed cDNA clones werehybridized with a 1:1 mixture of Cy3- or Cy5-labeled first-strand cDNAsgenerated from poly(A)1 RNA from various tissues, including both normaland tumor tissues, using the protocol provided by Incyte. The fluorescenceintensity was scanned, and data were analyzed using Incyte GEMTOOLSsoftware. Results were also analyzed by normalizing fluorescence intensitiesbetween experiments using a subset of cDNA clones. This enabled us tocompare data between different experiments.

Quantitative Real-Time PCR (TaqMan). Total RNA was treated withDNase I (Ambion) in the presence of RNasin (Promega) to remove DNAcontamination before cDNA synthesis. cDNA was synthesized with oligode-oxythymidylic acid primer (Boehringer Mannheim) and Superscript II reversetranscriptase (Life Technologies, Inc.). Real-time PCR (TaqMan) analysis wasperformed on a Perkin-Elmer/Applied Biosystems 7700 Prism. Matching prim-ers and fluorescence probes (see below) were designed for each of the genes(P503S, P504S, andP510S) according to the Primer Express program providedby Perkin-Elmer/Applied Biosystems. Primer and probe concentrations wereoptimized with a pool of cDNAs from prostate tumors. ForP503SandP504S,both forward and reverse primers were 900 nM. For P510S, both forward andreverse primers were 300 nM. In all cases, the final probe concentration was160 nM. The PCR reaction was performed in 25ml with dATP, dCTP, anddGTP at 0.2 mM and dUTP at 0.4 mM; 0.625 unit of Amplitaq Gold; 0.25 unitof Amperase uracil-N-glycosylase (Perkin-Elmer/Applied Biosystems); 5 mM

MgCl2; trace amounts of glycerol, gelatin, and Tween 20 (Sigma); and 2ml ofcDNA template.b-Actin primers and probes were obtained from Perkin-Elmer/Applied Biosystems.

The following primers and probes were used: (a)P503S, TGCCCTCGT-GACGTTCTTCT (forward primer), TCTTTCTTGATGGCAGGCACTAC(reverse primer), and CACCACAATGGCTGAGCACTTCCTGA (probe); (b)P504S, AAATGGTTATCATTAGGGCTTTTGA (forward primer), TTCC-TTTTTCACTAGAACCCATTCA (reverse primer), and TATCAAGC-AAACTGGAAGGCAGAATAACTACCATAATT (probe); and (c) P510S,TTGAACAGCTACTACGGTCAATGTATT (forward primer), GCAG-AGAGCAACCGATGTTTT (reverse primer), and TTGAGTGAAGCCTTA-AAAAGCACACACCACA (probe).

To quantitate the amount of specific mRNA in the samples, a standard curvewas generated for each run using the plasmid containing the gene of interest(dilutions ranging from 20 to 23 106 copies). In addition, a standard curve wasgenerated forb-actin ranging from 200 fg to 2 ng. This enabled standardizationof the initial RNA content of a tissue relative to the amount ofb-actin.

Bioinformatic Analysis. Protein localization was predicted by the PSORTalgorithm using the amino acid sequences ofP503S,P504S, andP510S.

Protein Expression, mAb Generation, and ImmunohistochemistryStaining. TruncatedP503S(amino acid 113–241) was cloned into pET32b(Novagen) and expressed inE. coli as thioredoxin fusion proteins with ahistidine tag.P503Swas purified by nickel chromatography, digested withthrombin, and further purified by reverse-phase chromatography. Full-lengthP504Swas cloned into pTrcHisC (Invitrogen) and expressed inE. coli with ahistidine tag. The protein was purified by nickel chromatography, followed byion exchange chromatography. Rabbit mAbs were generated from rabbitsimmunized withP503SandP504Sprotein by ImmGenics using the previouslypublished protocol (23). Immunohistochemistry was performed on formalin-fixed, paraffin-embedded tissues by QualTek Molecular Laboratories usingrabbit mAbs raised againstP503SandP504Sprotein.

RESULTS

cDNA Library Subtraction Is an Effective Way to EnrichTissue- and/or Cancer-specific Genes.We generated subtractedcDNA libraries to enrich prostate tissue- and/or cancer-specific genes(Table 1). Subtraction one was done by subtracting prostate tumor(Gleason score5 7; 25 ng/ml PSA) with normal pancreas. Subtractiontwo was done by subtracting prostate tumor with normal prostate.Subtraction three was done by subtracting normal prostate with nor-mal pancreas. Subtractions four and five were designed to enrich lessabundant genes by spiking several abundant sequences identified fromprevious subtractions (e.g., PSA) into the driver. Subtraction four wasalso performed with a pool of three prostate tumors (tumor 1, Gleasonscore5 7, 100 ng/ml PSA; tumor 2, Gleason score5 8, 35 ng/mlPSA; tumor 3, Gleason score5 7, 59 ng/ml PSA) that differed fromthose used for previous subtractions. Normal pancreas was used inseveral subtractions as the driver control because both prostate andpancreas are secretory organs with glandular structures. Use of pan-creas as a driver control should be effective in eliminating genesshared by both tissues.

Fig. 1A shows subtraction one, where several discrete bands are

Fig. 1. cDNA library subtraction.A, subtraction one. The subtracted cDNA library wasdigested withBamHI andXhoI and run on a 1% agarose gel (Lane 2).Lane 1, 1-kb DNAladder.B, subtraction five. The subtracted cDNA library was digested withBamHI andXhoI and run on a 1% agarose gel (Lane 4).Lane 1is a 1-kb DNA ladder;Lanes 2and3 are prostate tumor and normal pancreas cDNA libraries, respectively.

Fig. 2. Subtraction efficiency showed by depletion of humanb-actin in subtractedlibraries. DNAs from nonsubtracted and subtracted cDNA libraries were PCR-amplifiedfor 18, 23, 28, and 33 cycles with humanb-actin-specific primers. PCR products were runout on a 1% agarose gel and stained with ethidium bromide.S-1, subtraction one;S-2,subtraction two;S-3, subtraction three;S-4, subtraction four;S-5, subtraction five.

Table 1 Subtraction library summary

Subtractionlibrary Tracer Driver Spikinga

No. of clones onmicroarray

Subtraction 1 PTb Normal pancreas None 92Subtraction 2 PT NP None 122Subtraction 3 NP Normal pancreas None 120Subtraction 4 PT pool Normal pancreas Yes 114Subtraction 5 PT Normal pancreas Yes 78a Spiking was done with PSA, HGK-1, and CCOII.b PT, prostate tumor; NP, normal prostate.

1678

PROSTATE CANCER GENE IDENTIFICATION



seen. Sequence analysis of this library showed abundant clones to bePSA, HGK-1, CCOII, and TCR germ-lineg chain that comigrate withthe enriched bands. Less abundant tissue- and/or cancer-specific geneswere enriched in the subtraction five library (Fig. 1B), where the threemost abundant species were depleted by spiking them into the driver.Several discrete bands are seen in the subtraction five library (Fig. 1B,Lane 4). Abundant clones that comigrate with the enriched bands aretumor expression enhanced gene (TEEG), autonomously replicatingsequence (ARS), a novel gene described in another study,4 and the twogenesP503SandP504S.Lanes 2and3 in Fig. 1Bare nonsubtractedprostate tumor and normal pancreas library in which a smear wasseen.

The efficiency of subtraction was also demonstrated by depletion ofthe abundant housekeeping geneb-actin (Fig. 2). In nonsubtractedlibraries,b-actin-specific PCR product is visible by eighteenth cycleof amplification and becomes saturated at 28–33 cycles. Subtractedlibraries require a higher number of amplification cycles forb-actin tobe detected, indicating thatb-actin is preferentially depleted in sub-tracted cDNA libraries. Depletion ofb-actin is more significant in thesubtraction one, two, and three libraries, suggesting that these threesubtractions are more complete.

Microarray Screening of cDNA Clones from Subtracted Li-braries for Assessment of Tissue and Cancer Specificity.Sub-tracted cDNA libraries were prescreened by colony hybridization withcDNAs (PSA, HGK-1, CCOII, ARS, TEEG, aldehyde dehydrogenase

6, TCR germ-lineg chain, and sequence 3 from patent number5565323) to eliminate redundant cDNA clones. A total of 500 pre-screen negative clones, includingP503Sand P504S, were selected,and their mRNA expression levels were determined by microarrayanalysis. Clones showing overexpression in normal and/or tumorprostate tissues were sequenced to determine their identity.

Fig. 3 shows mRNA expression levels determined by microarray

4 J. Xu, J. A. Stolk, M. Kalos, E. Zasloff, A. Zhang, R. L. Houghton, and S. G. Reed.Identification and characterization of prostein, a novel prostate-specific antigen, manu-script in preparation.

Fig. 4. Northern blot analysis of prostate gene expression in prostate and other tissues.Tenmg of total RNA were run on a formaldehyde denaturing gel, blotted, and probed withrandom-primed cDNA probes ofP503S,P504S, andP510S.Lanes 1–4, prostate tumors;Lanes 5and6, normal prostates;Lanes 7and8, BPH; Lane 9, normal colon;Lane 10,normal kidney;Lane 11, normal liver;Lane 12, normal lung;Lane 13, normal pancreas;Lane 14, normal skeletal muscle;Lane 15, normal brain (P510S,Lane 15, normal skeletalmuscle); Lane 16, normal stomach;Lane 17, normal testis;Lane 18, normal smallintestine;Lane 19, normal bone marrow. Loading control is the RNA gel stained withmethylene blue showing 18S and 28S bands.

Fig. 3. Prostate gene expression determined by microarray. cDNA clones were PCR-amplified, arrayed onto glass slides, and probed with a 1:1 mixture of Cy3-labeled probe 1 andCy5-labeled probe 2. Prostate tumors used had Gleason scores of 3–8 with a PSA value of 1.7–35 ng/ml. Fluorescence scans represented in pseudocolor correspond to hybridizationintensities. Fluorescence intensity values were normalized using a subset of cDNA clones. Background fluorescence value is 2156 238, which is defined by the fluorescence valuesof empty spots (spots with no DNA target on glass slides,n 5 8) for all probe pairs (n5 22).

1679




for P503S,P504S, and another gene,P510S, which was identifiedthrough microarray screening.P503S is overexpressed in prostatetumors, BPH, and normal prostate. It is also expressed in normal colonand is slightly elevated in normal kidney and bladder. Other normaltissues tested had low or undetectable levels ofP503Sexpression.P504Sis overexpressed in about 30% of prostate tumors and is low toundetectable in normal tissues tested, including normal prostate, sug-gesting thatP504Sis a prostate cancer-specific gene.P510Sis pri-marily overexpressed in a subset of prostate tumors and BPH, whereasthe expression levels in nonprostate normal tissues are low to unde-tectable. Although microarray is a rapid and effective way to screenhundreds to thousands of clones in parallel, expression values may becompressed due to the limitation of detection sensitivity for raremessage and the saturation of signal for highly abundant genes.Therefore, other independent methods are necessary to give a moreaccurate assessment of mRNA expression levels.

Tissue and/or Cancer Specificity Was Determined by NorthernBlot Analysis. Consistent with the microarray analysis,P503Swasexpressed in prostate tumors, normal prostate tissues, and normalcolon and to a lesser extent in normal kidney and lung (Fig. 4). It wasundetectable in other normal tissues tested including liver, pancreas,skeletal muscle, brain, stomach, testis, small intestine, and bonemarrow.P504Swas overexpressed in 50% (two of four) of prostatetumors and was low or undetectable in other tissues tested, includingnormal prostate.P510Swas expressed in 50% of normal and tumorprostate tissues but was absent from other tissues tested. Therefore,Northern blot analysis confirmed the expression patterns observedwith the microarray approach, supporting the use of microarray as avalid high throughput screening approach.

Gene Expression Measured by Quantitative PCR (TaqMan)Analysis. An additional more sensitive and quantitative approach,real-time PCR (TaqMan), was used to assess the overexpression ofprostate candidates (Fig. 5). Using real-time PCR (TaqMan),P503Sishighly expressed in prostate tumors and normal prostate, as well as ina subset of other tumors (breast tumors, ovarian tumors, and colontumors but not lung tumors). It was also expressed by normal kidney

and stomach but detected at a very low level in other normal tissues.P504S is overexpressed in about 60% (four of seven) of prostatetumors. With the exception of normal liver that shows a low level ofmessage, other normal tissues including normal prostate show littleexpression ofP504S. P510Swas overexpressed in three of eightprostate tumors tested. It was expressed at lower levels in two otherprostate tumors and two of seven normal prostate tissues. Othernormal tissues did not have detectableP510Smessage. Thus, quan-

Fig. 5. Real-time PCR (TaqMan) analysis ofP503S,P504S, andP510S. PCR was performed on a Perkin-Elmer/Applied Biosystems 7700 Prism instrument using cDNAreverse-transcribed from RNA of various tissues.b-Actin was used as an internal reference control. Tumors have Gleason scores of 5–8 and PSA levels of 8.1–66 ng/ml. Results arerepresentative of three experiments for each gene.

Fig. 6. Immunohistochemistry staining forP503SandP504S. Prostate tumor (A) andBPH (B) were stained with rabbit mAb againstP503S. Prostate tumor (C) and BPH (D)were stained with rabbit mAb againstP504S. A total of five prostate carcinomas, grade3–5, and one BPH have been examined with these antibodies, and results are similar onall prostate carcinomas.

1680




titative PCR confirmed the results obtained by microarray and North-ern blot.

Protein Expression Determined by Immunohistochemistry.mRNA expression profiles ofP503Sand P504Sin prostate tumorsand normal prostate tissues were confirmed by immunohistochemistryusing rabbit mAbs raised againstP503Sand P504Sproteins. mAbagainstP503Sreacted with both prostate tumors and BPH (Fig. 6),whereasP504SmAb stained prostate tumors but not BPH. Immuno-histochemistry also indicated thatP503Sstaining was cell surface andcytoplasmic, whereasP504S showed strong granular cytoplasmicstaining.

Full-length cDNAs and Prediction of Protein Structures. Usingvarious approaches, including bioinformatics and colony hybridiza-tion, we cloned full-length sequences for all three genes. Localizationof protein in different subcellular compartments was predicted byPSORT. The full-length cDNA ofP503S, with 1228 bp and an openreading frame of 241 amino acids, was identified as human tetraspanNET-1 (GenBank accession number 3152700), a member of thetetraspanin/TM4SF family shown to be involved in cancer metastasis(24). Consistent with what has been shown for other family members,P503Sis predicted to be a type IIIa plasma membrane protein withtwo transmembrane spans and a cleavable signal sequence.P504Sisa 1621-bp cDNA with an open reading frame of 382 amino acids. Ithas been identified as humana-methylcacyl-CoA racemase (Gen-Bank accession number 4204097). AlthoughP504Sis predicted to bea type Ib plasma membrane protein with one transmembrane span andno signal sequence, the rat homologue has been shown to be expressedin the cytosol and mitochondria (25).P510Shas been identified ashuman ABC transporter MOAT-B (GenBank accession number3335173). It is predicted to be a plasma membrane protein with ninepotential transmembrane spans with no NH2-terminal signal sequence.

DISCUSSION

Coupling subtraction and microarray is a rapid and efficient way toidentify differentially expressed genes. cDNA library subtraction is apowerful approach to enrich tissue- or cancer-specific genes. Theefficiency of subtractions was shown by depletion of the highly andubiquitously expressed geneb-actin. Furthermore, the presence ofother known prostate-specific genes (e.g.,PSA and HGK-1) at highfrequency in subtracted libraries also suggests that the subtractionprocess worked effectively. Enriched tissue- and/or cancer-specificlibraries can serve as ideal target sources for microarray screening.Target selection is especially important for microarray. It can be veryinefficient to use nonnormalized or nonsubtracted libraries. On theother hand, arrays with a limited number of preselected genes, whichonly constitute a small percentage of the human genome, will beneither comprehensive nor systematic. Other problems associatedwith microarray include the use of cell line RNAs as probes that maynot reflect natural biological processes, and the use of a limitednumber of probes that may not cover enough cell types and tissuesfrom different organs in body. In our study, we used primary tissues,both normal and tumor. We also used a broad spectrum of normaltissues to determine the tissue distribution of genes. We are currentlyperforming studies with a larger number of tumor samples withwell-documented histology information to further correlate gene ex-pression and tumor histology.

Vasmatziset al. (15) have reported the discovery of three genesspecifically expressed in human prostate by using an electronic sub-traction approach with EST databases. Three of the five most abun-dant genes discovered by electronic subtraction, PSA, HGK-1, and theTCR g chain C region gene, are also abundantly present in oursubtracted libraries. This indicates that the two approaches could

generate similar but not identical results. It is postulated that thereasons for not identifying prostate-secreted seminal plasma proteinand prostatic acid phosphatase gene in our experiment are as follows:(a) tumor samples used in our studies did not contain these two genesat high levels; and (b) normal pancreas may also express these genesor their homologues at some level; therefore, they were eliminated bysubtraction process. Electronic subtraction is a rapid and powerfulapproach to identify potential differentially expressed genes; however,as the authors pointed out, it has some disadvantages. The biggestlimitation is that it depends on the availability of EST sequences. Theincompleteness of the EST database and the random nature of ESTclones may cause the specificity determined by electronic subtractionto be false positive or false negative. Information regarding specificityis especially unreliable when gene expression is low. Furthermore, thealgorithm the authors used can generate false EST clusters becauseone cluster can contain several different genes, and one gene can haveseveral clusters. This may also cause specificity information to beinaccurate. Therefore, information obtained by electronic subtractionneeds to be confirmed by experimental approaches.

The cDNA library subtraction approach described here systemati-cally compares cDNAs from two different tissue types and preferen-tially depletes sequences common in both tissue types, thereby en-riching sequences specific for one tissue type. This procedureeliminates the problems associated with electronic subtraction such asincompleteness of the EST database and false clustering. We havealso performed additional subtractions that incorporated the followingmodifications to the procedure: (a) pooling and using more tumorsamples to identify cancer-associated genes present only in a smallsubpopulation of cancer patients; (b) adjusting the driver:tracer ratioto identify less abundant genes and reduce the redundancy of sub-tracted libraries; and (c) using a pool of normal tissues as drivers,especially the ones that share similar cellular components with pros-tate tumor to eliminate nonspecific clones more efficiently; subtrac-tions performed in this manner generated results similar to thosereported herein. Alternative subtraction approaches, such as PCR-based subtraction (Clontech), have also been used to identify lessabundant genes.

Tissue and cancer specificity is one of the key requirements for amarker to be used for diagnosis and therapy. We have shown thatgenes identified by subtraction and microarray approaches can primehuman immunological responses, including both humoral and cell-mediated responses.5 Therefore, these genes can be used as potentialvaccine candidates for cancer immunotherapy. Our approaches pro-vide alternative methods for cancer antigen identification. This isextremely important for cancers other than melanoma, where cancerimmunogenicity is weak, and it is difficult to use immunologicalapproaches for cancer antigen identification.

ACKNOWLEDGMENTS

We gratefully acknowledge Incyte for help with microarray technology. Wethank Dr. Michael Kalos for valuable suggestions regarding the manuscript.Some tissue samples were obtained from the Cooperative Human TissueNetwork, which is funded by the National Cancer Institute, and from theNational Disease Research Interchange.

REFERENCES

1. Bostwick, D. G., MacLennan, G. T., and Larson, T. R. Prostate Cancer: What EveryMan—and His Family—Need to Know, revised edition. New York: Villard Books,1999.

5 E. Zasloff, R. Friedman, A. G. Spies, K. Grabstein, and M. Kalos. Generation ofprostein-specific CD8 and CD4 cell responses byin vitro priming with dendritic cellsinfected with viruses that express prostein, manuscript in preparation.

1681




2. Lowe, F. C., and Trauzzi, S. J. Prostatic acid phosphatase in 1993. Its limited clinicalutility. Urol. Clin. N. Am., 20: 589–595, 1993.

3. Guinan, P., Bhatti, R., and Ray, P. An evaluation of prostate specific antigen inprostatic cancer. J. Urol.,137: 686–689, 1987.

4. Oesterling, J. E., Brendler, C. B., Epstein, J. I., Kimball, A. W. J., and Walsh, P. C.Correlation of clinical stage, serum prostatic acid phosphatase and preoperativeGleason grade with final pathological stage in 275 patients with clinically localizedadenocarcinoma of the prostate. J. Urol.,138: 92–98, 1987.

5. Horoszewicz, J. S., Kawinski, E., and Murphy, G. P. Monoclonal antibodies to a newantigenic marker in epithelial prostatic cells and serum of prostatic cancer patients.Anticancer Res.,7: 927–935, 1987.

6. Israeli, R. S., Powell, C. T., Fair, W. R., and Heston, W. D. Molecular cloning of acomplementary DNA encoding a prostate-specific membrane antigen. Cancer Res.,53: 227–230, 1993.

7. Teni, T. R., Sheth, A. R., Kamath, M. R., and Sheth, N. A. Serum and urinaryprostatic inhibin-like peptide in benign prostatic hyperplasia and carcinoma of pros-tate. Cancer Lett.,43: 9–14, 1988.

8. Edwards, J. J., Anderson, N. G., Tollaksen, S. L., von Eschenbach, A. C., andGuevara, J. J. Proteins of human urine. II. Identification by two-dimensional electro-phoresis of a new candidate marker for prostatic cancer. Clin. Chem.,28: 160–163,1982.

9. Kim, Y. D., Robinson, D. Y., Manderino, G. L., Tribby, I. I., and Tomita, J. T.Molecular characterization of the epitope in prostate and breast tumor-associatedPR92 antigen. Cancer Res.,49: 2379–2382, 1989.

10. Wright, G. L. J., Beckett, M. L., Lipford, G. B., Haley, C. L., and Schellhammer, P. F.A novel prostate carcinoma-associated glycoprotein complex (PAC) recognized bymonoclonal antibody TURP-27. Int. J. Cancer,47: 717–725, 1991.

11. Beckett, M. L., Lipford, G. B., Haley, C. L., Schellhammer, P. F., and Wright, G. L. J.Monoclonal antibody PD41 recognizes an antigen restricted to prostate adenocarci-nomas. Cancer Res.,51: 1326–1333, 1991.

12. Tang, D. G., and Honn, K. V. 12-Lipoxygenase, 12(S)-HETE, and cancer metastasis.Ann. NY Acad. Sci.,744: 199–215, 1994.

13. Thomas, D. J., Robinson, M., King, P., Hasan, T., Charlton, R., Martin, J., Carr,T. W., and Neal, D. E. p53 expression and clinical outcome in prostate cancer. Br. J.Urol., 72: 778–781, 1993.

14. Pennisi, E. A catalog of cancer genes at the click of a mouse. Science (WashingtonDC), 276: 1023–1024, 1997.

15. Vasmatzis, G., Essand, M., Brinkmann, U., Lee, B., and Pastan, I. Discovery of threegenes specifically expressed in human prostate by expressed sequence tag databaseanalysis. Proc. Natl. Acad. Sci. USA,95: 300–304, 1998.

16. Velculescu, V. E., Zhang, L., Vogelstein, B., and Kinzler, K. W. Serial analysis ofgene expression. Science (Washington DC),270: 484–487, 1995.

17. Zhang, L., Zhou, W., Velculescu, V. E., Kern, S. E., Hruban, R. H., Hamilton, S. R.,Vogelstein, B., and Kinzler, K. W. Gene expression profiles in normal and cancercells. Science (Washington DC),276: 1268–1272, 1997.

18. Liang, P., and Pardee, A. B. Differential display of eukaryotic messenger RNA bymeans of the polymerase chain reaction. Science (Washington DC),257: 967–971,1992.

19. Tureci, O., Sahin, U., Zwick, C., Koslowski, M., Seitz, G., and Pfreundschuh, M.Identification of a meiosis-specific protein as a member of the class of cancer/testisantigens. Proc. Natl. Acad. Sci. USA,95: 5211–5216, 1998.

20. van der Bruggen, P., Traversari, C., Chomez, P., Lurquin, C., De Plaen, E., Van denEynde, B., Knuth, A., and Boon, T. A gene encoding an antigen recognized bycytolytic T lymphocytes on a human melanoma. Science (Washington DC),254:1643–1647, 1991.

21. Hara, T., Harada, N., Mitsui, H., Miura, T., Ishizaka, T., and Miyajima, A. Charac-terization of cell phenotype by a novel cDNA library subtraction system: expressionof CD8 a in a mast cell-derived interleukin-4-dependent cell line. Blood,84:189–199, 1994.

22. Schena, M., Shalon, D., Davis, R. W., and Brown, P. O. Quantitative monitoring ofgene expression patterns with a complementary DNA microarray. Science (Wash-ington DC),270: 467–470, 1995.

23. Babcook, J. S., Leslie, K. B., Olsen, O. A., Salmon, R. A., and Schrader, J. W. Anovel strategy for generating monoclonal antibodies from single, isolated lympho-cytes producing antibodies of defined specificities. Proc. Natl. Acad. Sci. USA,93:7843–7848, 1996.

24. Claas, C., Seiter, S., Claas, A., Savelyeva, L., Schwab, M., and Zoller, M. Associationbetween the rat homologue of CO-029, a metastasis-associated tetraspanin moleculeand consumption coagulopathy. J. Cell Biol.,141: 267–280, 1998.

25. Reichel, C., Brugger, R., Bang, H., Geisslinger, G., and Brune, K. Molecular cloningand expression of a 2-arylpropionyl-coenzyme A epimerase: a key enzyme in theinversion metabolism of ibuprofen. Mol. Pharmacol.,51: 576–582, 1997.

1682




2000;60:1677-1682. Cancer Res Jiangchun Xu, John A. Stolk, Xinqun Zhang, et al. Prostate Cancer Using Subtraction and MicroarrayIdentification of Differentially Expressed Genes in Human

Updated version

http://cancerres.aacrjournals.org/content/60/6/1677

Access the most recent version of this article at:

Cited articles

http://cancerres.aacrjournals.org/content/60/6/1677.full#ref-list-1

This article cites 24 articles, 16 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/60/6/1677.full#related-urls

This article has been cited by 33 HighWire-hosted articles. Access the articles at:

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

Subscriptions

Reprints and

[email protected] 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/60/6/1677To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/content/60/6/1677.full#ref-list-1

http://cancerres.aacrjournals.org/content/60/6/1677.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]



Identification of Differentially Expressed Genes in Human Prostate … · Identification of...

Documents

Transcript of Identification of Differentially Expressed Genes in Human Prostate … · Identification of...