hormone therapy failure in human prostate cancer: analysis by complementary … · 2000-08-03 ·...

Hormone Therapy Failure inHuman Prostate Cancer:Analysis by ComplementaryDNA and Tissue Microarrays

Lukas Bubendorf, Meelis Kolmer,Juha Kononen, Pasi Koivisto, SpyroMousses, Yidong Chen, EijaMahlamaki, Peter Schraml, HolgerMoch, Niels Willi, Abdel G.Elkahloun, Thomas G. Pretlow,Thomas C. Gasser, Michael J.Mihatsch, Guido Sauter, Olli-P.Kallioniemi

Background: The molecular mecha-nisms underlying the progression ofprostate cancer during hormonaltherapy have remained poorly under-stood. In this study, we developed anew strategy for the identification ofdifferentially expressed genes in hor-mone-refractory human prostate can-cer by use of a combination of comple-mentary DNA (cDNA) and tissuemicroarray technologies.Methods:Dif-ferences in gene expression betweenhormone-refractory CWR22R prostatecancer xenografts (human prostatecancer transplanted into nude mice)and a xenograft of the parental, hor-mone-sensitive CWR22 strain wereanalyzed by use of cDNA microarraytechnology. To validate the data fromcDNA microarrays on clinical prostatecancer specimens, a tissue microarrayof specimens from 26 prostates withbenign prostatic hyperplasia, 208 pri-mary prostate cancers, and 30 hor-mone-refractory local recurrences wasconstructed and used for immuno-histochemical detection of protein ex-pression. Results: Among 5184 genessurveyed with cDNA microarray tech-nology, expression of 37 (0.7%) wasincreased more than twofold in thehormone-refractory CWR22R xeno-grafts compared with the CWR22 xe-nograft; expression of 135 (2.6%) geneswas reduced by more than 50%. Thegenes encoding insulin-like growthfactor-binding protein 2 (IGFBP2)and 27-kd heat-shock protein (HSP27)were among the most consistently over-expressed genes in the CWR22R tu-mors. Immunohistochemical analysisof tissue microarrays demonstrated

high expression of IGFBP2 protein in100% of the hormone-refractory clini-cal tumors, in 36% of the primarytumors, and in 0% of the benign pros-tatic specimens (two-sidedP = .0001).Overexpression of HSP27 protein wasdemonstrated in 31% of the hormone-refractory tumors, in 5% of the pri-mary tumors, and in 0% of the benignprostatic specimens (two-sidedP =.0001). Conclusions: The combinationof cDNA and tissue microarray tech-nologies enables rapid identificationof genes associated with progressionof prostate cancer to the hormone-refractory state and may facilitateanalysis of the role of the encoded geneproducts in the pathogenesis of humanprostate cancer. [J Natl Cancer Inst1999;91:1758–64]

Despite the widespread use of prostate-specific antigen screening for early detec-tion, prostate cancer remains the secondleading cause of cancer-related deathamong men in western countries(1).Metastatic, hormone-refractory prostatecancer is the end-stage, lethal form of thedisease. Defining the molecular mecha-nisms underlying the transition of an an-drogen-responsive prostate cancer to ahormone-refractory prostate cancer repre-sents both an intriguing biologic questionand a critical clinical problem(2). It isimportant to better understand the bio-logic basis of prostate cancer progression,since no effective therapies exist for end-stage, hormone-refractory disease.

There are severalin vitro and in vivomodels for the study of hormone-refractory prostate cancer. For example,numerous hormone-independent strainsof the LNCaP human prostate cancer cell

Affiliations of authors:L. Bubendorf, M. Kolmer,J. Kononen, S. Mousses, Y. Chen, O.-P. Kallioni-emi, Cancer Genetics Branch, National Human Ge-nome Research Institute, Bethesda, MD; P.Koivisto, E. Mahlama¨ki, Laboratory of Cancer Ge-netics, Tampere University Hospital, Finland; P.Schraml, H. Moch, N. Willi, M. J. Mihatsch, G.Sauter (Institute of Pathology), T. C. Gasser (Uro-logic Clinics), University of Basel, Switzerland; A.G. Elkahloun, Research Genetics, Inc., Huntsville,AL; T. G. Pretlow, Institute of Pathology, CaseWestern Reserve University, Cleveland, OH.

Correspondence to:Olli-P. Kallioniemi, M.D.,Ph.D., National Institutes of Health, 49 Convent Dr.,MSC 4470, Rm. 4A24, Bethesda, MD 20892-4470(e-mail: [email protected]).

See“Notes” following “References.”

© Oxford University Press

1758 REPORTS Journal of the National Cancer Institute, Vol. 91, No. 20, October 20, 1999

line have been developed(3).Several hor-mone-refractory xenograft model systemsalso exist. Human xenografts are con-structed by the introduction of humanprostate tissue or cells into immunodefi-cient mice where they can be seriallytransplanted. For example, the CWR22xenograft tumor grows in nude mice andrecurs as hormone-refractory disease aftercastration of the mice(4). The availabilityof such model systems will become in-creasingly powerful, as high-throughputgenomic technologies, such as large-scaleparallel gene expression analysis withcomplementary DNA (cDNA) microar-rays or serial analysis of gene expression(5,6),become more widely available. Thequantity of information obtained fromthe analysis of the expression of thou-sands of genes at once creates uniqueopportunities for research but also posessubstantial challenges. For example,which of the hundreds of differentiallyexpressed genes identified in large-scalegene expression surveys are importantprimary events and which are down-stream or secondary changes? Further-more, are novel genes discovered fromexperimental model systems of cancerprogression also involved in the cancerprogression of human patients? By use oftraditional methods in molecular pathol-ogy, substantial work is required to ana-lyze the frequency of involvement or theclinical significance of just a single geneor protein. We recently developed a tissuemicroarray-based technology for high-throughput molecular analyses of humancancer(7). This tumor tissue microarray(“tissue chip”) technique is based on thearraying of cylindrical biopsy specimensfrom hundreds of different tumors into asingle paraffin block. Consecutive sec-tions of this tissue microarray block canthen be used for the analysis of multiplemolecular alterations at the DNA, RNA,and protein levels in hundreds of tumorsper experiment.

In this study, we combined the cDNAand tissue microarray technologies toidentify molecular alterations associatedwith the progression of human prostatecancer. First, the CWR22/CWR22R hu-man prostate cancer xenograft model(4)was used to screen for differential mes-senger RNA (mRNA) expression of morethan 5000 genes between hormone-refractory and hormone-responsive pros-tate cancers. Two consistently overex-pressed genes, insulin-like growth factor-binding protein 2 (IGFBP2) and the 27-kd

heat-shock protein (HSP27), were thenvalidated to be involved in clinical pros-tate cancer progression on the basis of im-munohistochemical analysis of the en-coded proteins in a prostate cancer tissuemicroarray containing 264 clinical speci-mens from various stages of tumor pro-gression.

MATERIALS AND METHODS

Xenograft tumors. CWR22 is a serially trans-plantable, human prostate cancer that was derivedfrom a Gleason score 9 primary prostate cancer withosseous metastasis(8). CWR22 is highly responsiveto androgen deprivation, with marked tumor regres-sion after castration(4). About half of the treatedanimals develop recurrent tumors (CWR22R) over atime from a few weeks to several months. CWR22Ris not dependent on androgen and is able to grow incastrated animals(4). Nude mice were housed andcared for as described earlier(8,9).Their care was inaccord with institutional guidelines. Fresh-frozenhuman prostate xenograft tissues (one sample fromCWR22 and four independent hormone-refractoryCWR22R strains) were obtained.

Comparative genomic hybridization.Compara-tive genomic hybridization was used to characterizethe tumor progression in this model system and wascarried out essentially as described previously(10),with some modifications. In brief, tumor (test) andnormal male (reference) DNAs were labeled by nicktranslation incorporating either SpectrumGreen orSpectrumRed deoxyuridine diphosphates (VysisInc., Downers Grove, IL). Labeled DNAs were hy-bridized to denatured normal peripheral blood met-aphase slides. After acquisition of digital images onwavelengths matching the 48 ,6-diamidino-phenylindole, SpectrumGreen and SpectrumRedemissions, green-to-red-ratio profiles were quanti-tated with Quips XL program (Vysis Inc.). Greenand red intensities were normalized so that the av-erage green-to-red ratio in each metaphase was set to1.0. Chromosomal regions where ratios exceeded1.2 were considered as gained, and those regionswhere the ratio was less than 0.8 was considered aslost.

cDNA microarrays. RNA was prepared fromCWR22/CWR22R xenografts as described by Chir-gwin et al. (11), with minor modifications. mRNAwas purified with the use of oligo(dT) selection withDynaBeads (Dynamic Analysis Inc., Huntsville,AL) according to the manufacturer’s instructions.Two different cDNA microarray formats were used(Clontech Laboratories, Inc. [Palo Alto, CA], andResearch Genetics, Inc. [Huntsville, AL]). The Atlashuman cDNA expression array from Clontech Labo-ratories, Inc. contains 588 duplicate spots on a singlemembrane, each representing 8–10 ng of cDNA ofknown and sequence-verified genes. These arrayswere hybridized with [32P]deoxycytidine triphos-phate (dCTP)-labeled cDNA probes prepared from 2mg of polyadenylic acid–RNA. In addition, we usedcDNA array filters from Research Genetics, Inc.(Prostate array, version I), with transcripts known tobe expressed in the prostate on the basis of ex-pressed sequence tag (EST) sequences found in nor-mal or malignant cDNA libraries. These filters con-tained 5184 spots (each with 5 ng of cDNA) of

known genes (n4 1960) or expressed sequence tags(ESTs; n4 3224), which were not sequence veri-fied. These arrays were hybridized with [33P]dCTP-labeled cDNAs derived from 50mg of total RNA.After overnight hybridization at 68 °C in Ex-pressHyb solution (Clontech Laboratories, Inc.), thefilters were washed and exposed to a high-resolutionscreen (Molecular Dynamics, Sunnyvale, CA) for 3days and scanned on a Storm PhosphorImager®(Molecular Dynamics). The spot intensities reflect-ing gene expression levels on the Atlas humancDNA array filter were quantified with Image-Quant® software (Molecular Dynamics), and thoseon the Research Genetics prostate-specific filterwere quantified with a custom software (Dearraysoftware: Y. Chen). The normalization of the spotintensities within an experiment (CWR22R versusCWR22) was done on the basis of the average of theintensities of all spots. The gene expression profilesof the CWR22Rs were compared with the gene ex-pression profile of CWR22. To define genes/ESTsas underexpressed or overexpressed, an at least two-fold expression difference was required. In addition,visual confirmation of all differentially expressedspots on filters was performed. The gray-scale im-ages were pseudocolored (red for hormone refrac-tory and green for hormone responsive) and overlaidfor better visualization of the relative expression in-tensities with Adobe Photoshop software (AdobeSystems Inc., San Jose, CA).

Reverse transcription–polymerase chain reac-tion (RT–PCR). cDNA was prepared by reversetranscriptase reaction by use of oligo(dT) primer(Research Genetics, Inc.). PCR was carried out withspecific primers for the IGFBP2 (Gene Bank#M35410) and HSP27 gene (Gene Bank #M54079)at an annealing temperature of 55 °C for 27 cyclesgenerating 391-base-pair (bp) and 260-bp products,respectively. Aliquots of the reaction products weresubjected to electrophoresis on a 2% agarose gel andvisualized by staining with ethidium bromide. Am-plification of the human asparagine synthetase geneby use of specific primers was used as a control.

Prostate tissue microarray.Formalin-fixed andparaffin-embedded tumor and benign control speci-mens were obtained from the archives of the Insti-tutes for Pathology, University of Basel (Switzer-land) and the Tampere University Hospital(Finland). All sections of tumors and controls werereviewed by one pathologist (L. Bubendorf). Tumorgrading was performed according to the method ofGleason(12). The specimens included 208 primaryprostate cancers, 30 transurethral resection speci-mens from locally recurrent hormone-refractorycancers operated on from 1976 through 1997, and 26transurethral resections for benign prostatic hyper-plasia as benign controls. The group of primary(non-hormone-refractory) prostate cancers consistedof 56 incidentally detected tumors in transurethralresections for presumed benign prostatic hyperplasia(stage T1a or b), 137 radical prostatectomy speci-mens from patients with clinically localized disease(stage T2), and specimens from 15 patients withlocally extensive disease (stage T3 or T4)(13).Morethan one sample per tumor specimen was arrayed in34 of the 238 patients. In these cases, the samplewith the strongest immunohistochemical stainingwas chosen for the immunohistochemical classifica-tion. The array also included 114 autopsy specimensfrom hormone-refractory metastatic prostate can-

Journal of the National Cancer Institute, Vol. 91, No. 20, October 20, 1999 REPORTS 1759

cers. These were excluded from this analysis, sinceimmunohistochemistry is often unreliable in tissuesfrom routine autopsies because of protein degrada-tion. The prostate tissue microarray was constructedas previously described(7). In brief, core tissue bi-opsy specimens (diameter, 0.6 mm) were taken fromthe least differentiated regions of individual paraf-fin-embedded prostate tumors (donor blocks) andprecisely arrayed into a new recipient paraffin block(35 × 20 mm) with a custom-built precision instru-ment (Beecher Instruments, Silver Spring, MD). Af-ter the block construction was completed, 5-mm sec-tions were cut with a microtome by use of anadhesive-coated tape sectioning system (Instrumed-ics, Hackensack, NJ) to support the adhesion of thearray elements. The presence of tumor tissue on thearrayed samples was verified on an hematoxylin–eosin-stained section.

Immunohistochemistry. Antigen retrieval wasperformed by treatment in a pressure cooker for 5minutes. Standard indirect immunoperoxidase pro-cedures were used for immunohistochemistry(ABC-Elite; Vector Laboratories, Inc., Burlingame,CA). A goat polyclonal antibody, C-18 (1 : 1000;Santa Cruz Biotechnology, Inc., Santa Cruz, CA)was used for detection of IGFBP2. HSP27 proteinwas detected by use of a monoclonal mouse anti-body HSP27 (1 : 100; BioGenex Laboratories, SanRamon, CA). The reactions were visualized by di-aminobenzidine as a chromogen. The primary anti-bodies were omitted for negative staining controls.The intensity of the cytoplasmic IGFBP2 andHSP27 staining was classified into four groups(negative, weak, intermediate, and strong staining).The number of tumors that could be analyzed forIGFBP2 and HSP27 expression differed slightlyfrom each other because of loss of representativeprostate cancer tissue on consecutive sections ofsome punch samples.

Statistical analysis. Contingency table analysiswas used to analyze the relationship between immu-nohistochemical staining, grade, and stage (total chi-squared test). AllP values were two-sided.

RESULTS

Analysis of Chromosomal Alterationsby Comparative GenomicHybridization

The hormone-sensitive CWR22 xeno-graft contained five chromosomal aberra-tions, including gain of 1q, gain of wholechromosomes 7, 8, and 12, and loss of 2q.The same five aberrations were also pre-sent in the hormone-refractory CWR22Rxenograft, indicating that the recurrent tu-mor was a clonal derivative of the pri-mary CWR22. In addition, the CWR22Rshowed a gain of chromosome 14q, whichwas not present in the primary CWR22(data not shown).

cDNA Microarray Analysis of GeneExpression Changes

cDNA microarray experiments werefirst performed with a nylon filter-based588 clone array (Clontech Laboratories,

Inc.). This analysis revealed 10 overex-pressed and 14 underexpressed genes in atleast two or more of the four hormone-refractory CWR22R xenografts as com-pared with the hormone-responsiveCWR22 xenograft (Table 1). Amongthese, HSP27 was substantially overex-pressed in three of the four CWR22Rstrains (median ratio, 2.6) and IGFBP2in all four CWR22Rs (median ratio,2.6). Two other members of the insulin-like growth factor (IGF) pathway—insulin-receptor and IGF-II—were alsomarkedly overexpressed in two of thefour CWR22R xenografts. RT–PCRanalysis confirmed the finding that the ex-pression of IGFBP2 and HSP27 was in-creased in hormone-refractory CWR22Rstrains as compared with hormone-sensitive CWR22 strains (Fig. 1).

In addition to these consistently differ-entially regulated genes in two or morexenograft specimens, 47 genes were over-expressed and 89 genes were underex-pressed in only one of the four hormone-refractory CWR22R xenografts.

To further explore the differential geneexpression patterns in hormone-refractoryprostate cancer, we analyzed the same tu-mors with a much larger cDNA microar-ray (5184 spots, Research Genetics, Inc.)containing a comprehensive collection ofgenes and ESTs found to be expressed in

Table 1.Most consistently overexpressed and underexpressed genes in the complementaryDNA microarray experiments and the ratios of gene expression in hormone-refractory human prostate

cancer xenografts (CWR22Ra–d) compared with gene expression in a xenograft of the hormone-sensitivestrain CWR22

Gene nameChromosomal

location

Ratios

CWR22Ra

CWR22Rb

CWR22Rc

CWR22Rd Median

OverexpressedIGFBP2 2q33–q34 2.7 2.4 2.6 5.3 2.6Heat-shock 27-kd protein 7q 2.6 2.7 1.5 4.8 2.6Insulin receptor 19p13.3–p13.2 1.8 1.5 2.9 5.3 2.4Transcription factor LCR-F1 7q32 3.2 3.1 0.8 0.9 2.0BSP-1 4q28 2.8 2.9 1.1 1.1 2.0P14-cyclin dependent kinase

inhibitor9p21 2.4 2.7 1.2 1.3 1.9

Insulin-like growth factor-II 11p15.5 1.1 0.7 2.7 2.1 1.6Homeobox protein HOX-A4 7p15–p14 2.0 2.1 1.2 1.1 1.6Tumor suppressor protein DCC 18q21.1 2.2 2.6 0.7 0.7 1.5ETS variant gene 3 1q21–q23 2.1 2.1 0.9 0.9 1.5

UnderexpressedOncostatin M 22q12.1–q12.2 0.2 0.3 1.1 0.5 0.4Integrin alpha 2B 17q21.32 0.5 0.3 1.4 0.2 0.4T-lymphocyte-secreted

protein I-30917 0.3 0.6 0.7 0.5 0.5

CD40 ligand Xq26 0.4 0.6 0.8 0.3 0.5Acyl-COA-binding protein 2q12–q21 0.4 1.2 0.5 0.4 0.5Interleukin 9 receptor Xq28 or Yq12 0.4 0.4 0.8 0.6 0.5E-selectin 1q22–q25 0.8 0.6 0.3 0.4 0.5Fms-like tyrosine kinase 4 5q34–q35 0.3 0.7 0.5 0.4 0.6Interleukin 2 receptor alpha chain 10p15–p14 0.4 0.5 1.1 1.3 0.8Hepatoma-derived growth factor Xq25 0.4 0.5 1.0 1.8 0.8Interleukin 7 receptor alpha chain 5p13 0.4 0.5 1.0 1.9 0.8Cyclin H 5q13.3–q14 1.0 1.7 0.5 0.3 0.8SHB adaptor protein 9p12–p11 1.4 1.7 0.5 0.5 1.0Clusterin 8p21–p12 0.4 0.3 2.0 2.0 1.2

*IGFBP2 4 insulin-like growth factor-binding protein 2; LCR-F14 locus control region F1; BSP-14transforming growth factor-b signaling protein-1; DCC4 deleted in colorectal carcinoma; ETS4 E-twenty-six specific; SHB4 src homology B.

Fig. 1. Reverse transcription–polymerase chain re-action analysis of insulin-like growth factor-bindingprotein 2 (IGFBP2), 27-kd heat-shock protein(HSP27), and asparagine synthetase (internal con-trol) expression in one hormone-responsive(CWR22) and in four hormone-refractory prostatecancer xenografts (CWR22Ra–d).


cDNA libraries from normal or malignantprostate. Altogether, 172 overexpressedor underexpressed genes or ESTs (ap-proximately 3%) in at least three of thefour hormone-refractory derivatives werediscovered as compared with the un-treated, hormone-sensitive human pros-tate cancer xenograft. Thirty-seven tran-scripts (0.7%) were substantially (ratio>2) elevated and 135 (2.6%) were under-expressed (ratio <0.5) in the CWR22R xe-nografts. A pseudocolored overlay of oneCWR22/CWR22R comparison and thecorresponding ratio distribution areshown in Fig. 2.

Histology and Immunohistochemistry

To evaluate whether the gene expres-sion changes seen in the hormone-refractory CWR22R tumors reflectedmolecular changes involved in tumor pro-gression in patients with prostate cancer,we created a tissue microarray to analyzethe expression of two overexpressedgenes, IGFBP2 and HSP27, at the proteinlevel in 238 different human prostate can-cers and in 26 benign prostate tissues. Thetotal number of evaluable specimenson the tissue microarray was 264 for the

IGFBP2 and 258 for the HSP27 immuno-staining.

In these arrayed clinical specimens, astrong association was seen between in-creased IGFBP2 and HSP27 protein ex-pression and the progression of prostatecancer to hormone-refractory disease(Fig. 3). A strong cytoplasmic IGFBP2staining was present in all of the 30 lo-cally recurrent, hormone-refractory pros-tate cancers, in 74 (36%) of the 208primary tumors, and in none of the 26benign prostate specimens (Fig. 3;P 4.0001, two-sided). HSP27 was stronglyexpressed in nine (31%) of 29 recur-rent tumors, in 11 (5%) of 204 primarytumors, but never in the secretory pros-tate epithelial cells of 25 benign pros-tatic hyperplasia specimens (Fig. 3;P 4.0001, two-sided). There was no statisti-cally significant association between IG-FBP2 or HSP27 expression and tumorgrade or T stage in the primary tumors(data not shown). A subgroup of 36patients had received primary neoadju-vant endocrine therapy before radicalprostatectomy, but their IGFBP2 andHSP27 expression data were similar tothose of the untreated patients (data notshown).

DISCUSSION

The transition from a hormone-sensitive human prostate cancer to a hor-mone-refractory recurrent strain in theCWR22 xenograft model system re-sembles the clinical progression of humanprostate cancer(4). As shown in thisstudy by comparative genomic hybridiza-tion, there was a close clonal genetic re-lationship between the primary and recur-rent xenograft tumors. Furthermore, manyof the alterations seen by comparative ge-nomic hybridization in this model system,such as gains of chromosome 7 and 8, aresimilar to those commonly found in clini-cal specimens from patients with prostatecancer. The cDNA microarray technologyallows rapid, large-scale screening of ex-pression of hundreds or thousands ofgenes in a single experiment(5). Here, upto 170 genes (3.3%) were identified to bedifferentially expressed between the pri-mary and recurrent (hormone-sensitiveand hormone-refractory) xenograft tu-mors. This high number of differentiallyexpressed genes illustrates the complexmolecular basis of prostate cancer pro-gression. The regrowth of the hormone-refractory tumor during androgen depri-vation therapy may necessitate a complexreprogramming of multiple key regula-tory mechanisms involving cell growth,apoptosis (i.e., programmed cell death),and other signaling pathways. It will beimportant to identify the molecularmechanisms that contribute to the devel-opment of recurrent tumors and to exam-ine if some of the signaling pathways in-volved would provide starting points forthe development of novel diagnostic ortherapeutic approaches for patients withadvanced, hormone-refractory prostatecancers.

The translation of gene-expressionfindings from model systems to humanpatients with cancer presents several chal-lenges. First, although this xenograftmodel system displayed phenotypic prop-erties resembling human prostate cancerprogression, it remains important to vali-date whether the same alterations of geneexpression and the same signaling path-ways contribute to the disease progressionin human cancer patients. Second, to uti-lize the cDNA microarray data for the de-velopment of improved diagnostic ortherapeutic approaches, it remains criti-cally important not only to screen for ex-pression of many different genes but alsoto screen many different tumor tissuesand establish an accurate frequency of in-

Fig. 2. Hybridization of the pros-tate complementary DNA microar-ray containing 5184 genes (Re-search Genetics, Inc.). A colorimage overlay of the CWR22 hy-bridization (green) and CWR22Rrecurrent xenograft (red) is shown.Spots with morered color repre-sent transcripts overexpressed inthe hormone-refractory tumor incomparison to the primary tumor,yellow spots indicate genes thatwere equally abundant, andgreenspots indicate underexpressedgenes in CWR22R. Genes thatwere not expressed in either of thetwo tissues appear in theblackbackground color. Inset (histo-gram) shows a normal frequencydistribution of the log10 intensityra t ios for CWR22R versusCWR22 for all of the 5184 spotson the microarray. The ratios aredisplayed on thex-axis, and therelative frequency of genes withthe given ratios is indicated on they-axis. Ratios of genes that have atwofold or higher expression in therecurrent than in the primary xeno-graft tumor are shown asred barsand those with a 50% or more re-duction asgreen bars.


volvement of these genes in differentstages of the prostate cancer progression.A substantial amount of work is requiredto fully explore the role of just a singlegene in cancer. Before performing full-length cDNA cloning, functional analy-ses, and other tedious experiments, onewould have to prioritize the long list ofpotential target genes that always emerges

from cDNA microarray experiments andto perform large-scale studies of clinicalspecimens. In this study, we first took ad-vantage of the fact that the pattern of geneexpression in the recurrent xenograft tu-mors was different from one animal toanother. Therefore, we decided to firstconcentrate on those genes that were dif-ferentially expressed in two or more re-

current xenograft tissues. One would ex-pect that such genes are more likely to beassociated with hormone therapy failure,whereas genes that are only overex-pressed in one case may be importantonly for that particular tumor. The deci-sive step was the evaluation of gene ex-pression patterns in clinical specimens byuse of our newly developed tissue micro-array technology.

Evaluation of all candidate genesemerging from the present cDNA micro-array experiments in a large series of un-cultured clinical tumors would take yearsif traditional methods were used. Further-more, after a few hundred genes had beenanalyzed, one would run out of the avail-able tumor tissues. Tissue microarraytechnology substantially facilitates thetranslation of basic research findings toclinical applications(7) and makes it pos-sible to performin situ analysis of hun-dreds of tumors either at the DNA, theRNA, or the protein level. This study wasdone with immunocytochemical tech-niques, but expression analyses of newlyidentified genes could also be analyzed bymRNA in situ hybridization when anti-bodies are not available. Such a strategyallows one to quickly validate and furtherexplore in a large number of clinicalspecimens thein vivo significance of can-didate genes discovered with the cDNAmicroarrays. Only minute amounts of tis-sues are required to make the tissue mi-croarray blocks, causing minimal damageto the original tumor blocks. Since onecan generate multiple replicate tissue mi-croarray blocks, each of which can be sec-tioned 200–300 times, one could easilygenerate thousands of tissue microarraysections from the same set of clinical tu-mor material. Each section can be utilizedfor the analysis of a different molecularmarker.

The small size of the samples makestissue microarrays a powerful screeningtool. However, the small tissue samplesmay not always be representative of thewhole tumor and, therefore, the preva-lence of a molecular alteration in a tissuemicroarray analysis may be underesti-mated. However, sampling bias may notbe a serious concern if the tumor areas arecarefully selected for punching. In ourprevious studies(7,14), we found a highconcordance between gene-amplificationfrequencies on tissue microarrays whencompared with the data from the litera-ture. The representativeness of tissue mi-croarray data could be improved by in-

Fig. 3. A) Hematoxylin–eosin and immunohistochemical staining of insulin-like growth factor-bindingprotein 2 (IGFBP2) and 27-kd heat-shock protein (HSP27) on the prostate cancer tissue microarray (originalmagnification ×200). Benign prostate glands show no immunoreactivity. Primary untreated prostate cancer(PRCA) demonstrates weak immunostaining of IGFBP2 but no immunoreactivity of HSP27. In contrast,hormone-refractory prostate cancer with local recurrence (Hr PRCA) shows strong expression of bothIGFBP2 and HSP27.B) Frequency distribution of expression of IGFBP2 and HSP27 during progression tohormone-refractory prostate cancer as measured by immunohistochemistry on a prostate cancer tissue mi-croarray.


cluding several samples from differentsites of a tumor on each array. Further-more, comparisons of the involvement ofone gene against another on the same ar-ray or comparisons of one molecular al-teration between two different stages oftumor progression will generate relativefrequency estimates that are not biased bythe sampling method. Nevertheless, tissuemicroarray technology should be re-garded as a rapid, high-throughput tool tosurvey many different genes and markersto identify those that are most promisingfor clinical applications. These wouldthen have to be tested on conventionaltissue specimens before clinical applica-tion.

The tissue microarray results validatedthat overexpression of IGFBP2 may be animportant event in hormone-refractoryprostate cancer, not only in the CWR22xenograft model system but also in pa-tients who had developed a recurrent tu-mor during androgen deprivation therapy.This finding is in agreement with recentexperimental and clinical studies(15–19)indicating that the IGF system may be akey growth regulatory pathway in pros-tate cancer. IGFBP2 is a member of theIGF growth factor system, which involvestwo growth factors (IGF-I and IGF-II),two IGF receptors (type I and II), sevenIGF-binding proteins (IGFBP1–7), aswell as IGFBP proteinases(16,20).IGF-Istimulates growth and inhibits apoptosisin normal and transformed epithelial cells(21–24). High plasma levels of IGF-Iwere recently shown to be associated withincreased risk of getting prostate cancer(17). Moreover, IGF-I has been shown toenhance androgen receptor-mediatedgene transcription in the prostate cancercell lines DU 145 (after cotransfectionwith an androgen-inducible reporter geneand an androgen receptor expression vec-tor) and LNCaP in the absence of andro-gen, suggesting that IGF-I may drive theandrogen-signaling pathway in hormone-refractory prostate cancer(25). IGFBPscan enhance or inhibit the bioactivity ofIGFs (IGF-I and IGF-II) by modulatingthe availability of free IGFs for their re-ceptors (26,27). IGFBP2 has also beensuggested to be an enhancer of IGF-Ifunction (22). It can be speculated thatoverexpression of IGFBP2 promotes sur-vival and androgen-independent growthof prostate cancer by increasing the bio-availability of IGFs. Members of the samepathway (IGF-II and insulin receptor)were also overexpressed in some of the

hormone-refractory xenograft tissues.However, IGFBP2 was systematicallyand most highly overexpressed, suggest-ing that it may perhaps have a central rolein modulating the IGF signaling in hor-mone-refractory prostate cancer. Alter-ations of IGFBP2 may also play a role inthe development and progression of othertumor types, such as breast, colorectal,and ovarian cancers(28–30).Overexpres-sion of IGFBP2 has also been observed incell lines established from several solidtumors(31,32).

The overexpression of HSP27 in aboutone third of hormone-refractory prostatecancers but in only 5% of primary tumorsis intriguing in light of the fact thatHSP27 has been shown to increase resis-tance to apoptosis induced by severaldrugs such as doxorubicin(33–36).Blockage of apoptosis may be an impor-tant feature of hormone-refractory pros-tate cancer and has been associated withthe differential expression of the Bcl-2gene family(37–39).It was recently sug-gested that HSP27 and Bcl-2 act at differ-ent levels to prevent apoptosis in immor-talized embryo fibroblasts, depending onthe type of apoptotic stimulus(40). Therole of HSP27 as a predictor of patientoutcome or response to therapy has re-ceived attention in breast cancer(41–44),but it has not been extensively studied inprostate cancer. In one study(45), vari-able HSP27 immunostaining was found in13 prostate tumors derived from transure-thral resection specimens, but no informa-tion about the hormonal treatment statuswas provided. Another study(46) did notfind HSP27 immunoreactivity in radicalprostatectomy specimens from patientswith clinically localized disease. On thebasis of this study, HSP27 expression isunlikely to play a major role in primaryprostate cancer but may be important inhormone therapy failure.

In summary, we describe a new strat-egy based on the combination of cDNAand tissue microarray technologies to ex-plore the molecular basis of human pros-tate cancer progression. Our results indi-cate that multiple gene expressionchanges may contribute to prostate cancerprogression and hormonal therapy failureand that at least some of the mechanismsinvolved in the CWR22 xenograft modelsystem may be similar to those contribut-ing to therapy failure and hormone-refractory prostate cancer growth in pa-tients. We detected an associationbetween increased expression of IGFBP2

and HSP27 and the hormone therapy fail-ure in both the xenograft model systemand in patients’ specimens. Further stud-ies are needed to evaluate these moleculesas well as dozens of other differentiallyexpressed genes as diagnostic or thera-peutic targets for hormone-refractoryprostate cancer.

REFERENCES

(1) Wingo PA, Ries LA, Giovino GA, Miller DS,Rosenberg HM, Shopland DR, et al. Annualreport to the nation on the status of cancer,1973–1996, with a special section on lung can-cer and tobacco smoking. J Natl Cancer Inst1999;91:675–90.

(2) Konety BR, Getzenberg RH. Novel therapiesfor advanced prostate cancer. Semin Urol On-col 1997;15:33–42.

(3) Wu HC, Hsieh JT, Gleave ME, Brown NM,Pathak S, Chung LW. Derivation of androgen-independent human LNCaP prostatic cancercell sublines: role of bone stromal cells. Int JCancer 1994;57:406–12.

(4) Nagabhushan M, Miller CM, Pretlow TP, Gia-conia JM, Edgehouse NL, Schwartz S, et al.CWR22: the first human prostate cancer xeno-graft with strongly androgen-dependent and re-lapsed strains bothin vivo and in soft agar.Cancer Res 1996;56:3042–6.

(5) DeRisi J, Penland L, Brown PO, Bittner ML,Meltzer PS, Ray M, et al. Use of a cDNA mi-croarray to analyse gene expression patterns inhuman cancer. Nat Genet 1996;14:457–60.

(6) Velculescu VE, Zhang L, Vogelstein B, Kinz-ler KW. Serial analysis of gene expression.Science 1995;270:484–7.

(7) Kononen J, Bubendorf L, Kallioniemi A, Bar-lund M, Schraml P, Leighton S, et al. Tissuemicroarrays for high-throughput molecularprofiling of tumor specimens. Nat Med 1998;4:844–7.

(8) Wainstein MA, He F, Robinson D, Kung HJ,Schwartz S, Giaconia JM, et al. CWR22: an-drogen-dependent xenograft model derivedfrom a primary human prostatic carcinoma.Cancer Res 1994;54:6049–52.

(9) Pretlow TG, Wolman SR, Micale MA, PelleyRJ, Kursh ED, Resnick MI, et al. Xenografts ofprimary human prostatic carcinoma. J NatlCancer Inst 1993;85:394–8.

(10) Tirkkonen M, Tanner M, Karhu R, KallioniemiA, Isola J, Kallioniemi OP. Molecular cytoge-netics of primary breast cancer by CGH. GenesChromosomes Cancer 1998;21:177–84.

(11) Chirgwin JM, Przybyla AE, MacDonald RJ,Rutter WJ. Isolation of biologically active ri-bonucleic acid from sources enriched in ribo-nuclease. Biochemistry 1979;18:5294–9.

(12) Gleason DF. Histologic grading and clinicalstaging of prostatic carcinoma. In: Tannen-baum M, editor. Urologic pathology: the pros-tate. Philadelphia (PA): Lea & Febiger; 1977.p. 171–8.

(13) International Union Against Cancer. Classifi-cation of malignant tumours. 5th ed. New York(NY): Wiley-Liss; 1997.

(14) Schraml P, Kononen J, Bubendorf L, Moch H,


Bissig H, Nocito A, et al. Tissue microarraysfor gene amplification surveys in many differ-ent tumor types. Clin Cancer Res 1999;5:1966–75.

(15) Lee AV, Hilsenbeck SG, Yee D. IGF systemcomponents as prognostic markers in breastcancer. Breast Cancer Res Treat 1998;47:295–302.

(16) Nunn SE, Gibson TB, Rajah R, Cohen P.Regulation of prostate cell growth by the insu-lin-like growth factor binding proteins andtheir proteases. Endocrine 1997;7:115–8.

(17) Chan JM, Stampfer MJ, Giovannucci E, GannPH, Ma J, Wilkinson P, et al. Plasma insulin-like growth factor-I and prostate cancer risk: aprospective study. Science 1998;279:563–6.

(18) Hankinson SE, Willett WC, Colditz GA,Hunter DJ, Michaud DS, Deroo B, et al. Cir-culating concentrations of insulin-like growthfactor-I and risk of breast cancer. Lancet 1998;351:1393–6.

(19) Burfeind P, Chernicky CL, Rininsland F, IlanJ, Ilan J. Antisense RNA to the type I insulin-like growth factor receptor suppresses tumorgrowth and prevents invasion by rat prostatecancer cellsin vivo. Proc Natl Acad Sci U S A1996;93:7263–8.

(20) Baxter RC, Binoux MA, Clemmons DR,Conover CA, Drop SL, Holly JM, et al. Rec-ommendations for nomenclature of the insulin-like growth factor binding protein superfamily.Endocrinology 1998;139:4036.

(21) Harrington EA, Bennett MR, Fanidi A, EvanGI. c-Myc-induced apoptosis in fibroblasts isinhibited by specific cytokines. EMBO J 1994;13:3286–95.

(22) Cohen P, Peehl DM, Lamson G, RosenfeldRG. Insulin-like growth factors (IGFs), IGF re-ceptors, and IGF-binding proteins in primarycultures of prostate epithelial cells. J Clin En-docrinol Metab 1991;73:401–7.

(23) Cohen P, Peehl DM, Baker B, Liu F, Hintz RL,Rosenfeld RG. Insulin-like growth factor axisabnormalities in prostatic stromal cells frompatients with benign prostatic hyperplasia. JClin Endocrinol Metab 1994;79:1410–5.

(24) Rajah R, Valentinis B, Cohen P. Insulin-like growth factor (IGF)-binding protein-3 in-duces apoptosis and mediates the effects oftransforming growth factor-beta1 on pro-grammed cell death through a p53- and IGF-independent mechanism. J Biol Chem 1997;272:12181–8.

(25) Culig Z, Hobisch A, Cronauer MV, RadmayrC, Trapman J, Hittmair A, et al. Androgen re-ceptor activation in prostatic tumor cell linesby insulin-like growth factor-I, keratinocytegrowth factor, and epidermal growth factor.Cancer Res 1994;54:5474–8.

(26) Jones JI, Clemmons DR. Insulin-like growthfactors and their binding proteins: biologicalactions. Endocr Rev 1995;16:3–34.

(27) Kelley KM, Oh Y, Gargosky SE, Gucev Z,Matsumoto T, Hwa V, et al. Insulin-likegrowth factor-binding proteins (IGFBPs) andtheir regulatory dynamics. Int J Biochem CellBiol 1996;28:619–37.

(28) Helle SI, Jonat W, Giurescu M, Ekse D, HollyJM, Lonning PE. Influence of treatment withonapristone on the IGF-system in breast cancerpatients. J Steroid Biochem Mol Biol 1998;66:159–63.

(29) el Atiq F, Garrouste F, Remacle-Bonnet M,Sastre B, Pommier G. Alterations in serum lev-els of insulin-like growth factors and insulin-like growth-factor-binding proteins in patientswith colorectal cancer. Int J Cancer 1994;57:491–7.

(30) Kanety H, Kattan M, Goldberg I, Kopolovic J,Ravia J, Menczer J, et al. Increased insulin-likegrowth factor binding protein-2 (IGFBP-2)gene expression and protein production lead tohigh IGFBP-2 content in malignant ovariancyst fluid. Br J Cancer 1996;73:1069–73.

(31) Reeve JG, Morgan J, Schwander J, BleehenNM. Role for membrane and secreted insulin-like growth factor-binding protein-2 in theregulation of insulin-like growth factor actionin lung tumors. Cancer Res 1993;53:4680–5.

(32) Menouny M, Binoux M, Babajko S. IGFBP-2expression in a human cell line is associatedwith increased IGFBP-3 proteolysis, decreasedIGFBP-1 expression and increased tumorige-nicity. Int J Cancer 1998;77:874–9.

(33) Samali A, Cotter TG. Heat shock proteins in-crease resistance to apoptosis. Exp Cell Res1996;223:163–70.

(34) Mehlen P, Hickey E, Weber LA, Arrigo AP.Large unphosphorylated aggregates as the ac-tive form of hsp27 which controls intracellularreactive oxygen species and glutathione levelsand generates a protection against TNFalpha inNIH-3T3-ras cells. Biochem Biophys ResCommun 1997;241:187–92.

(35) Oesterreich S, Weng CN, Qiu M, HilsenbeckSG, Osborne CK, Fuqua SA. The small heatshock protein hsp27 is correlated with growthand drug resistance in human breast cancer celllines. Cancer Res 1993;53:4443–8.

(36) Huot J, Houle F, Spitz DR, Landry J. HSP27phosphorylation-mediated resistance againstactin fragmentation and cell death induced byoxidative stress. Cancer Res 1996;56:273–9.

(37) Koivisto P, Visakorpi T, Rantala I, Isola J. In-creased cell proliferation activity and de-creased cell death are associated with the emer-gence of hormone-refractory recurrent prostatecancer. J Pathol 1997;183:51–6.

(38) McDonnell TJ, Troncoso P, Brisbay SM,Logothetis C, Chung LW, Hsieh JT, et al. Ex-pression of the protooncogene bcl-2 in theprostate and its association with emergence ofandrogen-independent prostate cancer. CancerRes 1992;52:6940–4.

(39) Raffo AJ, Perlman H, Chen MW, Day ML,Streitman JS, Buttyan R. Overexpression ofbcl-2 protects prostate cancer cells from apop-tosisin vitro and confers resistance to androgendepletion in vivo. Cancer Res 1995;55:4438–45.

(40) Guenal I, Sidoti-de Fraisse C, Gaumer S, Mig-notte B. Bcl-2 and Hsp27 act at different levelsto suppress programmed cell death. Oncogene1997;15:347–60.

(41) Love S, King RJ. A 27 kDa heat shock proteinthat has anomalous prognostic powers in earlyand advanced breast cancer. Br J Cancer 1994;69:743–8.

(42) Conroy SE, Sasieni PD, Amin V, Wang DY,Smith P, Fentiman IS, et al. Antibodies to heat-shock protein 27 are associated with improvedsurvival in patients with breast cancer. Br JCancer 1998;77:1875–9.

(43) Vargas-Roig LM, Fanelli MA, Lopez LA,Gago FE, Tello O, Aznar JC, et al. Heat shockproteins and cell proliferation in human breastcancer biopsy samples. Cancer Detect Prev1997;21:441–51.

(44) Vargas-Roig LM, Gago FE, Tello O, Aznar JC,Ciocca DR. Heat shock protein expression anddrug resistance in breast cancer patients treatedwith induction chemotherapy. Int J Cancer1998;79:468–75.

(45) Thomas SA, Brown IL, Hollins GW, HockenA, Kirk D, King RJ, et al. Detection and dis-tribution of heat shock proteins 27 and 90 inhuman benign and malignant prostatic tissue.Br J Urol 1996;77:367–72.

(46) Storm FK, Mahvi DM, Gilchrist KW. Hsp-27has no diagnostic or prognostic significance inprostate or bladder cancers. Urology 1993;42:379–82.

NOTES

Present address:M. Kolmer, National PublicHealth Institute, Department of Human MolecularGenetics, Helsinki, Finland.

Supported by the Swiss National Science Foun-dation (81BS-052807) (to L. Bubendorf); by theAcademy of Finland and by the Tampere UniversityHospital Foundation (to P. Koivisto); and by PublicHealth Service grant CA57179 from the NationalCancer Institute, National Institutes of Health, De-partment of Health and Human Services (to T. G.Pretlow).

We thank Rita Epper, Martina Mirlacher, MartinaStorz, and Heidi Oggier, University of Basel (Swit-zerland), for their excellent technical support; DarylLeja, National Human Genome Research Institute(Bethesda, MD), for his illustration support; andSteve Leighton, Beecher Instruments (Silver Spring,MD), for his help in tissue microarray instrumenta-tion.

Manuscript received March 1, 1999; revised Au-gust 10, 1999; accepted August 18, 1999.


hormone therapy failure in human prostate cancer: analysis by complementary … · 2000-08-03 ·...

Documents