Differentially expressed genes in nonsmall cell lung cancer: expression profiling of cancer-related genes in squamous cell lung cancer

Cancer Genetics and Cytogenetics 149 (2004) 98–106

Lead article

Differentially expressed genes in nonsmall cell lung cancer: expression profiling of cancer-related genes in squamous cell lung cancer Eeva Kettunena, Sisko Anttilaa,b, Jouni K. Seppa¨nenc, Antti Karjalainenb, Henrik Edgrena, Irmeli Lindstro¨md, Reijo Salovaaraa, Anna-Maria Nisse´na, Jarmo Salod, Karin Mattsond, Jaakko Hollme´nc, Sakari Knuutilaa,*, Harriet Wikmana,b a

Departments of Pathology and Medical Genetics, Haartman Institute, University of Helsinki, and Laboratory of Cytomolecular Genetics, Helsinki University Central Hospital, Haartmanink 3, 4th floor, PO Box 400, Helsinki FIN-00029 HUS, Finland b Departments of Occupational Medicine and Epidemiology, Finnish Institute of Occupational Health, Helsinki, Finland c Laboratory of Computer and Information Science, Helsinki University of Technology, Espoo, Finland d Departments of Internal Medicine, Division of Pulmonary Diseases and Cardiothoracic Surgery, Section of General Thoracic and Esophageal Surgery, Helsinki University Central Hospital, Helsinki, Finland Received 9 May 2003; received in revised form 8 July 2003; accepted 15 July 2003

Abstract

The expression patterns of cancer-related genes in 13 cases of squamous cell lung cancer (SCC) were characterized and compared with those in normal lung tissue and 13 adenocarcinomas (AC), the other major type of nonsmall cell lung cancer (NSCLC). cDNA array was used to screen the gene expression levels and the array results were verified using a real-time reverse-transcriptase– polymerase chain reaction (RT-PCR). Thirty-nine percent of the 25 most upregulated and the 25 most downregulated genes were common to SCC and AC. Of these genes, DSP, HMGA1 (alias HMGIY), TIMP1, MIF, CCNB1, TN, MMP11, and MMP12 were upregulated and COPEB (alias CPBP), TYROBP, BENE, BMPR2, SOCS3, TIMP3, CAV1, and CAV2 were downregulated. The expression levels of several genes from distinct protein families (cytokeratins and hemidesmosomal proteins) were markedly increased in SCC compared with AC and normal lung. In addition, several genes, overexpressed in SCC, such as HMGA1, CDK4, IGFBP3, MMP9, MMP11, MMP12, and MMP14, fell into distinct chromosomal loci, which we have detected as gained regions on the basis of comparative genomic hybridization data. Our study revealed new candidate genes involved in NSCLC. 쑖 2004 Elsevier Inc. All rights reserved.

1. Introduction Lung cancer is the leading cause of cancer deaths worldwide, with more than one million deaths each year [1]. Squamous cell carcinoma (SCC), adenocarcinoma (AC), and large cell carcinoma (LCC) are the main types of non–small cell lung cancer (NSCLC), small cell lung cancer (SCLC) being the other main category of lung cancer. SCC is the predominant type of lung cancer among caucasian men, whereas AC is the most common type among women, nonsmokers and among most oriental populations. Cytogenetic and molecular genetic aberrations in NSCLC are diverse and complex [2–4]; however, different NSCLC types share many cytogenetic changes and also some characteristic molecular genetic changes. For instance, deregulation * Corresponding author. Tel.: ⫹358-9-19126527; fax: ⫹358-9-19126788. E-mail address: [email protected] (S. Knuutila). 0165-4608/04/$ – see front matter 쑖 2004 Elsevier Inc. All rights reserved. doi:10.1016/S0165-4608(03)00300-5

of MYC [5], activation of the MET/HGF pathway [6], and inactivation of the RB1/TP16/CCND1 (RB/p16/cyclin D1) pathway [7,8] suggest some common features in the pathogenesis of NSCLC. Nevertheless, different types of lung cancer also have recurrent characteristic genetic abnormalities. In SCC, the implicated molecular genetic changes include activation of the epidermal growth factor receptor gene EGFR, which is thought to predispose to the development of SCC if coexpressed with TP53 [9]. A higher frequency of TP53 mutations has been found in SCC than in AC [10]. Changes overrepresented in AC include KRAS mutations, in 20–30% of cases [11,12], and ERBB2 (alias HER2/neu) overexpression (30–35%) [13,14]. Because most of the above-mentioned genes are deregulated in only a limited number of lung cancer cases, more broadly useful marker genes are needed. To reveal new candidate genes for diagnostic, prognostic, and therapeutic purposes in lung cancer, we conducted a study using cDNA

E. Kettunen et al. / Cancer Genetics and Cytogenetics 149 (2004) 98–106

array technology to detect specific expression patterns in pulmonary adenocarcinomas [15]. Our objective was to characterize the expression patterns of cancer-associated genes in SCC compared with both normal lung tissue and AC of the lung. The differences in the expression patterns were evaluated using three different statistical methods to obtain a broad view of deregulated genes in SCC, and both commonly and specifically expressed genes were detected. We also considered and here discuss conceivable associations of gene over- and underexpression with the DNA copy number changes detected previously using comparative genomic hybridization (CGH).

2. Materials and methods 2.1. Lung specimens Thirteen new, freshly snap-frozen lung tissue specimens of SCC from Finnish caucasian patients were analyzed using cDNA array, in addition to 13 previously described adenocarcinoma samples (see [15] and Table 1). One of the original fourteen AC samples was excluded due to its unreliable array result (high background). To confirm the cDNA array result, four additional SCC samples were analyzed using reverse transcriptase polymerase chain reaction (RT-PCR) (Table 1). To exclude the effect of individual variation in gene expression levels, four different normal lung tissue specimens were used as reference controls (Table 1) [15]. All of them were histologically verified as nontumorous tissues from patients with tuberculoma, intrabronchial granuloma or lung cancer. All the samples were examined and classified for histologic type and grade according to WHO standards (1999) by the same pathologist (S.A.). The previous CGH data from five SCC cases are shown in Table 2 [16]. The Ethical Review Board of the Department of Thoracic and Cardiovascular Surgery of the Helsinki University Central

99

Hospital approved the study protocol and the present study was approved by the Ethics Committee for Research in Occupational Health and Safety. 2.2. RNA extractions and cDNA array hybridizations RNAs were extracted according to standard methods described in Wikman et al. [15]. Tumor content of the samples was confirmed from adjacent sections, when available (not in cases 133, 139, 164, and 264). Only samples with more than 50% tumor cells were chosen. The good quality and integrity of RNA was confirmed using ethidium bromide– agarose gel electrophoresis and spectrophotometry. Atlas Human Cancer Gene Filter 1.2 (Clontech, Palo Alto, CA) including 1176 tumor-relevant genes was used for the cDNA array experiments (a gene list is available at http://atlasinfo.clontech.com/atlasinfo). Total RNA (3.5 µg) was reverse-transcribed into cDNA with labeled [α33P]-dATP (Amersham Pharmacia Biotech, Buckinghamshire, UK) using the Clontech cDNA array labeling kit. Purification of the probe, hybridization (68⬚C, overnight), and washings were done according to the manufacturer’s instructions. No filter was used more than three times. The scanned raw images of the arrays (Bio-Imaging Analyzer model BAS2500; Fuji, Nakanuma, Japan) were analyzed and the expression levels determined using the AtlasImage 2.0 software (Clontech). 2.3. Statistical analyses To identify genes that were differentially expressed in SCC compared with normal lung, the raw expression data obtained with AtlasImage 2.0 were analyzed with two complementary statistical techniques, principal component analysis (PCA) and permutation test. These methods have been described in detail elsewhere [15]. Our objective was to find

Table 1 Main characteristics of the study population and non–small cell lung cancer tumor specimens (squamous cell carcinoma and adenocarcinoma) Characteristic

Reference controls

SCCa

SCCb

AC

Sample size (no.) Female/Male Age, mean (years) Age, range (years) Histologic grade I Histologic grade II Histologic grade III Clinical stage I Clinical stage II Clinical stage III Never smoked (no.) Ex-smoker (no.) Current smoker (no.) Pack years, mean ⫾ SD Smoking years, mean ⫾ SD

4 1/3 56.8 48–70 — — — — — — 1 0 3 47.6 ⫾ 2.9 44.7 ⫾ 7.1

13 2/11 60.8 46–77 1 5 7 6 5 2 0 5 8 39.6 ⫾ 20.8 39.7 ⫾ 12.1

4 2/2 68.0 56–74 1c 1c 0c 3 0 1 0c 0c 2c 43.5 ⫾ 10.6c 61.5 ⫾ 14.8c

13 1/12 59.3 35–69 8 2 3 0 4 9 1 0 12 46.0 ⫾ 15.7 39.4 ⫾ 12.1

a b c

Used in cDNA arrays and RT-PCR analyses. Used in RT-PCR analyses only. Histologic grade and smoking data available only for two of the four cases.

100

E. Kettunen et al. / Cancer Genetics and Cytogenetics 149 (2004) 98–106

Table 2 CGH data for five SCC tumors analyzed with the cDNA array Case no.a

Losses (⭐0.85)

Gains (⭓1.17)

55 (3)

—

139 (6)

11p15, 17p

164 (13)

4

185 (15)

10p [COPEB], 17p, 19p

189 (16)

11p15, 17p

2p10~p16, 3q21~q24, 3q25~qter, 4q10~q23, 5p10~p15.2, 5p15.3, 6q24~qter, 7p [IGFBP3], 8q, 9p13~qter, 11q10~q14, 12p10~p12, 12p13 [MRP-1, GAPDH], 12q13~q15 [CDK4, IGA7B], 14q10~q12 [MMP14], 16p, 18 [DSG2, DSC3], 20 [MMP9], 22q [MMP11, MIF] 1q24~q42 [IRF6], 3q13.3~q25, 3q26.1~q26.3, 3q27~qter, 4p15.3~q23, 8q10~q23, 13q21~q22, 13q22~q31, 13q32~qter [COL4A2], 15q10~q22 3q, 5p, 7p10~p21 [IGFBP3], 8q, 15q21~qter, 22q [MMP11, MIF] Xq10~q21, Xq22~qter, 2p13~p22, 2q14.1~q31, 3q10~q13.3, 3q13.3~qter, 4p14~p15.3, 5p, 7p21~qter [IGFBP3], 9p22~pter, 14q13~q31 Xq22~qter, 1q [IRF6], 2p13~pter [RRM2], 3q10~q24, 3q25~qter, 4q26~q31.2, 5p, 6p10~p12, 6p12~p21.1 [HMGAI], 6p21.2~p23, 7p10~p21 [IGFBP3], 8q, 11q10~q14, 11q14~qter [MMP12], 12p [NT-3, MRP-1, GAPDH], 12q10~q23 [CDK4, CK2E, IGA7B], 15q22~qter

Data from Bjo¨rkqvist et al., 1998 [16]. The up- and downregulated genes located at the distinct regions of gains and losses are given in brackets. High-level amplifications (⭓1.5) are highlighted in boldface type. a The case number in parenthesis refers to the sample number in the source report [16].

true up- and downregulated genes; thus, an arbitrarily chosen cutoff level of 25 genes was used. PCA has been used for analysis of array data [17–19]. PCA is a linear signal decomposition technique, which finds a set of orthogonal basis vectors in the data space so that the variance of the data is maximal along these directions; the first direction or first principal component (PC) explains most of the variance, the second PC explains most of the remaining variance, and so on [20]. Using the basis vectors, original gene expression profiles can be transformed to a lowerdimensional representation. In particular, we refer to the projection of a gene expression profile on the first principal component as the PCA score. Genes whose expression level differs consistently among the tumor samples compared with the mean of reference controls have a high absolute PCA score. The sign of the PCA score corresponds to lower or higher expression level in tumor samples compared with the reference controls. To quantify how different two groups of measurements are, we used a second technique, which we refer to here as g-score. Briefly, the score assesses the probability of each measurement having arisen from a normal distribution estimated from the opposite set, validated using a permutation test. As a measure of the diagnostic accuracy of each gene, we used a statistic based on the relative operating characteristic

(ROC) curve. We pooled the measurements corresponding to AC patients and reference controls as one group and used the SCC patients as another group. For each individual gene, we consider the diagnostic test of using an expression threshold: subjects whose expression is lower than the threshold are diagnosed as belonging to one of the two groups, and vice versa. Different thresholds yield different trade-offs between correct and incorrect classifications, and plotting the fraction of correctly classified SCC patients against the fraction of incorrectly classified nonpatients yields the ROC curve. We use the area under the ROC curve, called the AUROC statistic. In an ideal case, some threshold allows 100% accuracy. The ROC curve will then have a point corresponding to no false positives and all true positives, so the area will be 1; in the case of absolutely no diagnostic value, the curve will be a straight line and the area will be 0.5. Intuitively, the area corresponds to the probability of passing the following test [21]: given a randomly chosen SCC patient and a randomly chosen nonpatient, does the expression level of the given gene allow classifying the two patients correctly? Empirical P-values corresponding to AUROC values were obtained using a permutation test. We have earlier used a similar method in a leukemia study [19]. 2.4. Real-time semiquantitative RT-PCR Semiquantitative RT-PCR was used to verify the array data of four genes (CCNB1, CAV1, COPEB [alias CPBP], and DSC3) in 17 SCCs, 5 ACs, and 4 controls. AMV-RT enzyme was used to convert 800 ng of template total RNA (for case 464, 100 ng of mRNA was used) into cDNA according to the manufacturer’s instructions (first strand cDNA synthesis kit; Roche Diagnostics, Indianapolis, IN). The amplifications were performed using a LightCycler (Roche Diagnostics, Mannheim, Germany) with the Roche LightCycler Fast-start DNA Master SYBR green kit. The PCR conditions for CCNB1, CAV1, and COPEB were as described by Wikman et al. [15]. For DSC3, the total volume of 10 µL PCR reaction included the master-mix, 0.5 mmol/L of each primer (forward: 5′-AGT GGG GTC AAA GAT CAA CG-3′, reverse: 5′-CTT TGT CTA TTG CCA GGA CTG T-3′; TIB MolBiol, Berlin, Germany), and 2.5 mmol/L MgCl2. The following cycling conditions were used: initial denaturation at 95⬚C for 7 minutes, followed by 45 cycles with denaturation at 95⬚C for 0 seconds, annealing at 64⬚C for 9 seconds, and elongation at 72⬚C for 9 seconds, with a ramping rate of 20⬚C/s. To verify the amplification specificity, melting curve analysis was performed. The estimations of the concentrations were acquired with the second derivative maximum method provided by the LightCycler software. Relative standard curves were obtained from serial dilutions of one of the samples showing the highest intensity values in the cDNA array [22]. To normalize the PCR results, the relative concentration values of the housekeeping gene encoding phospholipase 2A (PL2A) were used.

E. Kettunen et al. / Cancer Genetics and Cytogenetics 149 (2004) 98–106

3. Results The expression levels of 1,176 cancer-related genes were compared between 13 SCC and four different normal lung tissue samples, as well as between SCC and non-SCC (including normal lung and 13 AC) for which we had previously performed similar gene expression profiling [15]. Three different statistical methods were used to reveal the differentially expressed genes. All score values are available on request. 3.1. Gene expression patterns in SCC With a cutoff level of 25 genes (i.e., the 25 most upregulated and the 25 most downregulated), PCA and g-score revealed a total of 79 aberrantly expressed genes, as shown in Fig. 1. Fifteen of the 25 most upregulated genes in SCC compared with normal lung tissue were detected with both PCA and g-score. Altogether, 36 genes were upregulated in SCC compared with normal lung tissue according to PCA, g-score, or both. Several genes encoding cell-to-cell adhesion, growth factor, and cell-cycle regulator-associated proteins were among the upregulated genes. As expected, several genes encoding epithelial cell-specific markers, such as keratins 10, 14, 19, and 2E, were upregulated. Seven out of the 25 most downregulated genes in SCC compared with normal lung tissue were found with both statistical methods; altogether, with either or both methods 43 genes were found to be downregulated. In addition to the previously reported lung cancer–related gene CAV1 [23], the genes COPEB, BENE, BMPR2, PRL-1, FCRN, and TYROBP were found to be similarly downregulated with both analyses in SCC. As could be expected, combined gene expression data from separate comparisons of SCC or AC [15] versus normal lung reference controls showed similar gene expression profiles of SCC and AC. Of the 25 most upregulated and the 25 most downregulated genes calculated applying either PCA or g-score, 43 (39%) out of the 111 genes appearing on either list were common to SCC and AC, using both PCA and g-score (data not shown). Among the commonly upregulated genes in NSCLC were HMGA1 (alias HMGIY), SFN (alias 14-33 sigma), TN, DSP, CCNB1, PLK1, MIF, TIMP1, MMP12, and MMP11; the downregulated genes included AKAP12 (alias gravin), BMPR2, COPEB, SOCS3, BENE, TIMP3, CAV1, CAV2, and TYROBP. 3.2. Marker genes for SCC To specify positive SCC marker genes, we used the AUROC statistic to compare gene expression levels in SCC with those in non-SCC (i.e., AC and normal lung). The 13 most upregulated (AUROC value ⭓0.8 and P ⬍ 0.05) genes in SCC versus non-SCC are listed in Table 3. They included two sets of genes located at close proximity to each other. Besides cytokeratins 14 and 10 expressed in normal epidermis [24], 17q21~q22 harbors also ITGB4 and NGFR, all upregulated in SCC. The other group of genes, located at

101

12q11~q13 and upregulated in SCC, includes a basic cytokeratin CK2E, IGA7B, RARG (alias RAR-g1), and COL2A1. MIF was found to be an SCC marker in our analysis, even though it was also upregulated in AC compared with the normal lung reference controls. Thus, this gene can indicate either type of cancer, but the AUROC statistic reveals it to have diagnostic value specifically in differentiating SCC patients from AC patients and controls. 3.3. Array results compared with CGH results When cDNA array results for five SCC cases were compared with the CGH data obtained from them, we found that some of the up- or downregulated genes were located at distinct gained or lost areas, which suggests their potential involvement in amplification or loss of DNA and consequent over- or underexpression. Table 2 lists the genes located in these areas. 3.4. RT-PCR analysis The cDNA array results for four genes (CCNB1, CAV1, COPEB, and DSC3) were confirmed with real-time RT-PCR in 17 SCC and 4 normal lung tissue specimens; for DSC3, the PCR reactions also included 5 AC cases. The cDNA from normal lung tissue reference controls were used as separate samples as well as pooled cDNA. Fig. 2 shows the PCR results. All SCC results were confirmed, and PCR verified the clear difference in DSC3 expression between SCC and AC. For SCC, the number of significant cases (intensity difference ⬎ 5,000 and normalized ratio ⬎ 2 in arrays; normalized ratio ⬎ 5 in RT-PCR) were as follows, for cDNA arrays and RT-PCR, respectively: 6 of 13 cases and 13 of 17 cases for CCNB1, 11 of 13 and 11 of 17 for CAV1, and 10 of 13 and 12 of 17 for COPEB. For DSC3, the significant cases were more frequent in SCC: 8 of 13 cases and 16 of 16 cases in SCC (in array and RT-PCR, respectively), compared with 1 of 5 and 0 of 5 in AC. Thus, the overexpression of DSC3 exclusively in SCC was verified with PCR.

4. Discussion We characterized the expression profiles of known cancerrelated genes in 13 SCC specimens by using cDNA array. The expression levels were compared between SCC and four normal lung tissue samples, and also with the gene expression levels of 13 previously characterized AC samples. For detection of general gene expression profiles we combined two statistical methods, PCA and g-score, because this approach has been shown to give a more extensive view of the deregulated genes in cancer than either of the methods alone [15]. In addition, with the help of the AUROC statistic we could identify a set of genes that can separate SCC from non-SCC with 80% probability (and P ⬍ 0.05). The identification performance of AUROC has

102

E. Kettunen et al. / Cancer Genetics and Cytogenetics 149 (2004) 98–106

E. Kettunen et al. / Cancer Genetics and Cytogenetics 149 (2004) 98–106 Table 3 The most upregulated genes in SCC as calculated using the AUROC statistic AUROC valuea

GenBank accession

Protein name (gene symbol)

Chromosomal localization

0.95

M99061

12q11~q13

0.94

X83929

0.91

J00124

0.90

M19156

0.85

X74295

0.85

M24857; M38258

0.84 0.83

X53587; X52186 M25639

0.82

M65062

0.81 0.81

AF028593 M14764

0.81

X53586; X59512 X16468

Type II cytoskeletal 2 epidermal keratin (CK2E) Desmocollin 3A/3B precursor (DSC3) Type I cytoskeletal 14 keratin (CK14) Type I cytoskeletal 10 keratin (CK10) Integrin alpha 7B precursor (IGA7B) Retinoic acid receptor gamma 1 (RARG; alias RAR-g1) Integrin beta 4 (ITGB4); CD104 antigen Macrophage migration inhibitory factor (MIF) IGF-binding protein 5 (IGFBP5; alias IBP5) Jagged 1 (JAG1; alias HJ1) Low-affinity nerve growth factor receptor (NGFR) Integrin alpha 6 precursor (ITGA6; alias VLA6) Procollagen 2 alpha 1 subunit precursor (COL2A1)

0.80

a

18q12.1

103

identify new candidate genes deregulated in SCC and thereby to discover useful tumor markers. To avoid the interference of individual variation in gene expression, our reference controls consisted of normal lung tissue specimens from four individuals who had similar age, gender, and exposure distribution as the patients. Although the reference individuals were not disease-free, the tissue samples were carefully chosen to represent normal lung.

17q12~q21 17q21~q22

4.1. Genes upregulated specifically in SCC

12q13.13

Several genes encoding epithelial cell-specific markers were found upregulated only in SCC, including cytokeratins 14, 10, and 2E and epithelial cell adhesion (hemidesmosomal) proteins ITGB4, IGA7B, and ITGA6. The hemidesmosome is a specialized cell–extracellular matrix (ECM) contact that mediates the attachment of stratified and complex epithelium to the ECM, binding externally to laminins and collagens and internally to the keratin cytoskeleton (CK14, for instance) [34,35]. Hemidesmosomes include the heterodimeric α6β4 integrin (ITGA6 and ITGB4). In agreement with a study of NSCLC showing α6 and β4 integrin proteins more strongly expressed in SCC than in AC [36], we found the same pattern at the mRNA level; another study, however, reported no differences between SCC and AC concerning these proteins [37]. Note that CK14 was upregulated only in our SCC samples. As intracellular bridges are characteristic of SCC but not of AC, one could expect to find proteins involved in these bridges, such as desmosomal proteins, to be upregulated in SCC [38]. In agreement with this, we found desmocollin 3 (DSC3) upregulated in SCC exclusively, as verified in RTPCR. DSC3 has been shown to be expressed in the epithelium of trachea [39], but so far it has not been reported in NSCLC.

12q13

17q11~qter 22q11.2 2q33~q36 20p12 17q21~q22 2q31.3 12q13.11~q13.2

All values significant at P ⬍ 0.05.

been suggested to be particularly good to characterize diagnostic accuracy [21]. In general, SCC had a gene expression profile similar to that of AC [15], with HMGA1, COPEB, AKAP12, and SOCS3 as novel genes involved in NSCLC. Nevertheless, several genes belonging to distinct protein families (cytokeratins and hemidesmosomal proteins) had a different expression pattern in SCC than in AC and the controls. Several gene expression profiling studies of the major lung cancer types have been published [25–33]; however, it is difficult to compare these studies, which have been conducted using different methods (dissimilar array types; cDNA subtraction; serial analysis of gene expression, SAGE), different types of specimens (primary tumor tissue versus tumor cell lines), and different analytical and statistical methods. Furthermore, the aim has often been to classify lung cancer types into distinct subclusters. Our goal was to

4.2. Deregulated genes common to SCC and AC The gene expression patterns of SCC and AC were closely similar. Abnormal expression (mRNA, protein, or both) of CCNB1, PLK1, MIF, MMP11, TIMP1, TIMP3, DP, TN, and caveolins (CAV) have all previously been associated with NSCLC, whereas HMGA1, COPEB, AKAP12, BENE, and SOCS3 are novel findings (discussed in more detail in [15]). The expression differences of CCNB1, CAV1, and COPEB between SCC and normal lung were verified using RT-PCR. CCNB1 was upregulated and CAV1 and COPEB were downregulated in SCC.

䉳 Fig. 1. Seventy-nine genes were aberrantly expressed in SCC, compared with normal lung tissue; these were detected using PCA, g-score, or both methods. The study samples are listed at the top. Gray squares represent cDNA signals with missing values. The 25 most upregulated (26 for the g-score, because two genes had the same value) and the 25 most downregulated genes are listed as follows. Panels A and F: genes significantly deregulated according to both methods. Panels B and E: genes significantly deregulated according to PCA only. Panels C and D: genes significantly deregulated according to gscore only. The normalized expression difference was achieved by subtracting from each cDNA signal first the background value, then the adjusted intensity of four controls (mean), and finally the mean intensity of the array.

104

E. Kettunen et al. / Cancer Genetics and Cytogenetics 149 (2004) 98–106

Fig. 2. Gene expression results obtained with RT-PCR and cDNA array of COPEB and CAV1 (downregulated) and CCNB1 and DSC3 (upregulated) in 17 SCC cases. For DSC3, five AC cases are included. The gene expression levels were normalized to a housekeeping gene (PL2A) and then divided by the mean of gene intensity value of four controls, to calculate the ratio. In two AC cases (47, 280), the DSC3 RT-PCR results were below the detection threshold. In four SCC cases (137, 294, 1255, 464), cDNA array results were not available; in SCC case 464, the RT-PCR result for DSC3 was not available.

4.3. Several deregulated genes in SCC locate at distinct areas of DNA gains detected with CGH A common mechanism for up- or downregulation of genes in cancer is their amplification or deletion and gains or losses of whole chromosomal regions. Some distinct areas of chromosomal changes in SCC of Finnish caucasian patients have previously been revealed with CGH analysis [16]. Five of these cases are included in the present study (Table 2). Examples of upregulated genes mapped to regions of gains are HMGA1 (6p21), IGFBP3 (7p13~p12), CDK4 (12q14), MMP12 (11q22), MMP14 (14q11~q12), MMP9 (20q11.2~q13.1), and MMP11 (22q11.2). Chromosomal rearrangements of HMGA1, with subsequent activation of the gene, have been found in a variety of benign mesenchymal tumors, such as pulmonary chondroid hamartomas [40,41]. HMGA1 is worthy of note also in solid tumors of epithelial origin, in that elevated protein levels have been found in at least colorectal and pancreatic carcinomas, as well as in prostate cancer [42,43]. The expression pattern of HMGA2 (alias HMGIC) but not HMGA1 in NSCLC has been studied [44]. CDK4 amplification, often coamplified with MDM2, has been reported in several cancer types [45–47], but not yet in NSCLC. Nevertheless, studies using immunohistochemistry have demonstrated high expression of CDK4 in NSCLC tumors and association with poor survival [48,49]. In four out of five samples, upregulation of IGFBP3 coincided with an amplification of 7p10~p21. An extensive microarray study showed that IGFBP3 was upregulated at both mRNA and protein levels in lung adenocarcinomas [50]; however, promoter methylation status of IGFBP3 has been shown to correlate with poor prognosis in stage I

NSCLC [51], and the intriguing role of IGFBP3 in NSCLC seems to be proapoptotic [52]. Interestingly, in a human prostate adenocarcinoma cell line, MMP9, a product of one of the four overexpressed genes encoding matrix metalloproteinases located in the area of DNA gains, was indicated as a regulator of IGFBP3 [53]. We know of no reports of amplifications of the other overexpressed genes located at regions of DNA gain (Table 2) in human lung carcinomas. Compared with profiling based on the level of gene expression, however, the resolution achieved with the CGH technique is lower, and discrepancies probably arise from sensitivity differences between these two methods. Further studies are needed, perhaps using the new CGH array technique [54], to assess the possible association between DNA copy number changes and differences in gene expression levels. 4.4. Conclusions In summary, by using the cDNA array technique we were able to reveal marked gene expression differences between SCC and normal lung tissue. Many of the differentially expressed genes were shared with AC. Nevertheless, several genes involved specifically in SCC were revealed, for instance encoding members of distinct protein families, such as cytokeratins and hemidesmosomal proteins. This study showed that the cDNA array technique is a suitable method for screening new candidate genes for further studies to assess their potential value as tumor type-specific markers and indicators of biologic clues to pathogenic pathways in cancer. Acknowledgments We are grateful to Pa¨ivi Tuominen for excellent technical assistance. J.K.S. and J.H. wish to thank Prof. Heikki

E. Kettunen et al. / Cancer Genetics and Cytogenetics 149 (2004) 98–106

Mannila for many helpful discussions. This work was supported by grants from the K. Albin Johanssons Stiftelse, Sigrid Juse´lius Foundation, Ida Montin Foundation, Finnish Cancer Foundation, and the Academy of Finland.

[16]

[17]

References [1] Ferlay J, Bray F, Pisani P, Parkin D. GLOBOCAN 2000: cancer incidence, mortality and prevalence worldwide. Lyon: IARCPress; 2001. [2] Mitelman F, Johansson B, Mertens F, editors. Mitelman database of chromosome aberrations in cancer (2003). Available at: http://cgap. nci.nih.gov/Chromosomes/Mitelman. [3] Knuutila S, Bjo¨rkqvist AM, Autio K, Tarkkanen M, Wolf M, Monni O, Szymanska J, Larramendy ML, Tapper J, Pere H, ElRifai W, Hemmer S, Wasenius VM, Vidgren V, Zhu Y. DNA copy number amplifications in human neoplasms - review of comparative genomic hybridization studies. Am J Pathol 1998;152:1107–23. [4] Knuutila S, Aalto Y, Autio K, Bjo¨rkqvist AM, El-Rifai W, Hemmer S, Huhta T, Kettunen E, Kiuru-Kuhlefelt S, Larramendy ML, Lushnikova T, Monni O, Pere H, Tapper J, Tarkkanen M, Varis A, Wasenius VM, Wolf M, Zhu Y. DNA copy number losses in human neoplasms. Am J Pathol 1999;155:683–94. [5] Broers JL, Viallet J, Jensen SM, Pass H, Travis WD, Minna JD, Linnoila RL. Expression of c-myc in progenitor cells of the bronchopulmonary epithelium and in a large number of non–small cell lung cancers. Am J Respir Cell Mol Biol 1993;9:33–43. [6] Olivero M, Rizzo M, Madeddu R, Casadio C, Pennacchietti S, Nicotra MR, Prat M, Maggi G, Arena N, Natali PG, Comoglio PM, Di Renzo MF. Overexpression and activation of hepatocyte growth factor/ scatter factor in human non-small-cell lung carcinomas. Br J Cancer 1996;74:1862–8. [7] Jin M, Inoue S, Umemura T, Moriya J, Arakawa M, Nagashima K, Kato H. Cyclin D1, p16 and retinoblastoma gene product expression as a predictor for prognosis in non–small cell lung cancer at stages I and II. Lung Cancer 2001;34:207–18. [8] Otterson GA, Kratzke RA, Coxon A, Kim YW, Kaye FJ. Absence of p16INK4 protein is restricted to the subset of lung cancer lines that retains wildtype RB. Oncogene 1994;9:3375–8. [9] Rusch V, Klimstra D, Linkov I, Dmitrovsky E. Aberrant expression of p53 or the epidermal growth factor receptor is frequent in early bronchial neoplasia and coexpression precedes squamous cell carcinoma development. Cancer Res 1995;55:1365–72. [10] Husgafvel-Pursiainen K, Ridanpa¨a¨ M, Anttila S, Vainio H. p53 and ras gene mutations in lung cancer: implications for smoking and occupational exposures. J Occup Environ Med 1995;37:69–76. [11] Husgafvel-Pursiainen K, Hackman P, Ridanpa¨a¨ M, Anttila S, Karjalainen A, Partanen T, Taikina-Aho O, Heikkila¨ L, Vainio H. K-ras mutations in human adenocarcinoma of the lung: association with smoking and occupational exposure to asbestos. Int J Cancer 1993;53: 250–256. [12] Mao L, Hruban RH, Boyle JO, Tockman M, Sidransky D. Detection of oncogene mutations in sputum precedes diagnosis of lung cancer. Cancer Res 1994;54:1634–7. [13] Harpole DJ Jr, Herndon J 2nd, Wolfe WG, Iglehart JD, Marks JR. A prognostic model of recurrence and death in stage I non–small cell lung cancer utilizing presentation, histopathology, and oncoprotein expression. Cancer Res 1995;55:51–6. [14] Hirsch FR, Varella-Garcia M, Franklin WA, Veve R, Chen L, Helfrich B, Zeng C, Baron A, Bunn PA Jr. Evaluation of HER-2/neu gene amplification and protein expression in non–small cell lung carcinomas. Br J Cancer 2002;86:1449–56. [15] Wikman H, Kettunen E, Seppa¨nen J, Karjalainen A, Hollme´n J, Anttila S, Knuutila S. Identification of differentially expressed genes

[18]

[19]

[20] [21] [22]

[23]

[24] [25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]

105

in pulmonary adenocarcinoma by using cDNA array. Oncogene 2002; 21:5804–13. Bjo¨rkqvist AM, Husgafvel-Pursiainen K, Anttila S, Karjalainen A, Tammilehto L, Mattson K, Vainio H, Knuutila S. DNA gains in 3q occur frequently in squamous cell carcinoma of the lung, but not in adenocarcinoma. Genes Chromosomes Cancer 1998;22:79–82. Armstrong SA, Staunton JE, Silverman LB, Pieters R, den Boer ML, Minden MD, Sallan SE, Lander ES, Golub TR, Korsmeyer SJ. MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia. Nat Genet 2002;30:41–7. Hilsenbeck SG, Friedrichs WE, Schiff R, O’Connell P, Hansen RK, Osborne CK, Fuqua SAW. Statistical analysis of array expression data as applied to the problem of tamoxifen resistance. J Natl Cancer Inst 1999;91:453–9. Niini T, Vettenranta K, Hollme´n J, Larramendy M, Aalto Y, Wikman H, Nagy B, Seppa¨nen J, Ferrer Salvador A, Mannila H, Saarinen-Pihkala U, Knuutila S. Expression of myeloid-specific genes in childhood acute lymphoblastic leukemia - a cDNA array study. Leukemia 2002;16:2213–21. Hand DJ, Mannila H, Smyth P. Principles of data mining. Cambridge, MA: MIT Press; 2001. Swets JA. Measuring the accuracy of diagnostic systems. Science 1988;240:1285–93. Rajeevan MS, Vernon SD, Taysavang N, Unger ER. Validation of array-based gene expression profiles by real-time (kinetic) RT-PCR. J Mol Diagn 2001;3:26–31. Racine C, Belanger M, Hirabayashi H, Boucher M, Chakir J, Couet J. Reduction of caveolin 1 gene expression in lung carcinoma cell lines. Biochem Biophys Res Commun 1999;255:580–6. Milisavljevic V, Freedberg IM, Blumenberg M. Close linkage of the two keratin gene clusters in the human genome. Genomics 1996;34:134–8. Anbazhagan R, Tihan T, Bornman DM, Johnston JC, Saltz JH, Weigering A, Piantadosi S, Gabrielson E. Classification of small cell lung cancer and pulmonary carcinoid by gene expression profiles. Cancer Res 1999;59:5119–22. Bhattacharjee A, Richards WG, Staunton J, Li C, Monti S, Vasa P, Ladd C, Beheshti J, Bueno R, Gillette M, Loda M, Weber G, Mark EJ, Lander ES, Wong W, Johnson BE, Golub TR, Sugarbaker DJ, Meyerson M. Classification of human lung carcinomas by mRNA expression profiling reveals distinct adenocarcinoma subclasses. Proc Natl Acad Sci U S A 2001;98:13790–5. Garber ME, Troyanskaya OG, Schluens K, Petersen S, Thaesler Z, Pacyna-Gengelbach M, van de Rijn M, Rosen GD, Perou CM, Whyte RI, Altman RB, Brown PO, Botstein D, Petersen I. Diversity of gene expression in adenocarcinoma of the lung. Proc Natl Acad Sci U S A 2001;98:13784–9. Hellmann GM, Fields WR, Doolittle DJ. Gene expression profiling of cultured human bronchial epithelial and lung carcinoma cells. Toxicol Sci 2001;61:154–63. McDoniels-Silvers AL, Stoner GD, Lubet RA, You M. Differential expression of critical cellular genes in human lung adenocarcinomas and squamous cell carcinomas in comparison to normal lung tissues. Neoplasia 2002;4:141–50. Nacht M, Dracheva T, Gao Y, Fujii T, Chen Y, Player A, Akmaev V, Cook B, Dufault M, Zhang M, Zhang W, Guo M, Curran J, Han S, Sidransky D, Buetow K, Madden SL, Jen J. Molecular characteristics of non–small cell lung cancer. Proc Natl Acad Sci U S A 2001;98: 15203–8. Petersen S, Heckert C, Rudolf J, Schluns K, Tchernitsa OI, Scha¨fer R, Dietel M, Petersen I. Gene expression profiling of advanced lung cancer. Int J Cancer 2000;86:512–7. Wang T, Hopkins D, Schmidt C, Silva S, Houghton R, Takita H, Repasky E, Reed SG. Identification of genes differentially over-expressed in lung squamous cell carcinoma using combination of cDNA subtraction and microarray analysis. Oncogene 2000;19:1519–28. Virtanen C, Ishikawa Y, Honjoh D, Kimura M, Shimane M, Miyoshi T, Nomura H, Jones MH. Integrated classification of lung tumors and

106

[34]

[35] [36]

[37]

[38]

[39]

[40]

[41]

[42]

[43]

E. Kettunen et al. / Cancer Genetics and Cytogenetics 149 (2004) 98–106 cell lines by expression profiling. Proc Natl Acad Sci U S A 2002;99: 12357–62. Dowling J, Yu QC, Fuchs E. Beta4 integrin is required for hemidesmosome formation, cell adhesion and cell survival. J Cell Biol 1996;134: 559–572. Nievers MG, Schaapveld RQ, Sonnenberg A. Biology and function of hemidesmosomes. Matrix Biol 1999;18:5–17. Patriarca C, Alfano RM, Sonnenberg A, Graziani D, Cassani B, de Melker A, Colombo P, Languino LR, Fornaro M, Warren WH, Coggi G, Gould VE. Integrin laminin receptor profile of pulmonary squamous cell and adenocarcinomas. Hum Pathol 1998;29:1208–15. Smythe WR, LeBel E, Bavaria JE, Kaiser RL, Albelda SM. Integrin expression in non–small cell carcinoma of the lung. Cancer Metastasis Rev 1995;14:229–39. Young GD, Winokur TS, Cerfolio RJ, Van Tine BA, Chow LT, Okoh V, Garver RI Jr. Differential expression and biodistribution of cytokeratin 18 and desmoplakins in non–small cell lung carcinoma subtypes. Lung Cancer 2002;36:133–41. Nuber UA, Scha¨fer S, Stehr S, Rackwitz HR, Franke WW. Patterns of desmocollin synthesis in human epithelia: immunolocalization of desmocollins 1 and 3 in special epithelia and in cultured cells. Eur J Cell Biol 1996;71:1–13. Kazmierczak B, Meyer-Bolte K, Tran KH, Wockel W, Breightman I, Rosigkeit J, Bartnitzke S, Bullerdiek J. A high frequency of tumors with rearrangements of genes of the HMGI(Y) family in a series of 191 pulmonary chondroid hamartomas. Genes Chromosomes Cancer 1999;26:125–33. Tallini G, Vanni R, Manfioletti G, Kazmierczak B, Faa G, Pauwels P, Bullerdiek J, Giancotti V, Van den Berghe H, Dal Cin P. HMGIC and HMGI(Y) immunoreactivity correlates with cytogenetic abnormalities in lipomas, pulmonary chondroid hamartomas, endometrial polyps, and uterine leiomyomas and is compatible with rearrangement of the HMGI-C and HMGI(Y) genes. Lab Invest 2000;80:359–69. Abe N, Watanabe T, Sugiyama M, Uchimura H, Chiappetta G, Fusco A, Atomi Y. Determination of high mobility group I(Y) expression level in colorectal neoplasias: a potential diagnostic marker. Cancer Res 1999;59:1169–74. Chiappetta G, Manfioletti G, Pentimalli F, Abe N, Di Bonito M, Vento MT, Giuliano A, Fedele M, Viglietto G, Santoro M, Watanabe T, Giancotti V, Fusco A. High mobility group HMGI(Y) protein expression in human colorectal hyperplastic and neoplastic diseases. Int J Cancer 2001;91:147–51.

[44] Rogalla P, Drechsler K, Schroder-Babo W, Eberhardt K, Bullerdiek J. HMGIC expression patterns in non–small lung cancer and surrounding tissue. Anticancer Res 1998;18:3327–30. [45] Cheung TH, Yu MM, Lo KW, Yim SF, Chung TK, Wong YF. Alteration of cyclin D1 and CDK4 gene in carcinoma of uterine cervix. Cancer Lett 2001;166:199–206. [46] Simon R, Struckmann K, Schraml P, Wagner U, Forster T, Moch H, Fijan A, Bruderer J, Wilber K, Mihatsch MJ, Gasser T, Sauter G. Amplification pattern of 12q13-q15 genes (MDM2, CDK4, GLI) in urinary bladder cancer. Oncogene 2002;21:2476–83. [47] Zhao J, Roth J, Bode-Lesniewska B, Pfaltz M, Heitz P, Komminoth P. Combined comparative genomic hybridization and genomic microarray for detection of gene amplifications in pulmonary artery intimal sarcomas and adrenocortical tumors. Genes Chromosomes Cancer 2002;34:48–57. [48] Lingfei K, Pingzhang Y, Zhengguo L, Jianhua G, Yaowu Z. A study on p16, pRb, cdk4 and cyclinD1 expression in non–small cell lung cancers. Cancer Lett 1998;130:93–101. [49] Volm M, Kooma¨gi R. Relevance of proliferative and pro-apoptotic factors in non-small-cell lung cancer for patient survival. Br J Cancer 2000;82:1747–54. [50] Beer DG, Kardia SL, Huang CC, Giordano TJ, Levin AM, Misek DE, Lin L, Chen G, Gharib TG, Thomas DG, Lizyness ML, Kuick R, Hayasaka S, Taylor JM, Iannettoni MD, Orringer MB, Hanash S. Gene-expression profiles predict survival of patients with lung adenocarcinoma. Nat Med 2002;8:816–24. [51] Chang YS, Wang L, Liu D, Mao L, Hong WK, Khuri FR, Lee HY. Correlation between insulin-like growth factor-binding protein-3 promoter methylation and prognosis of patients with stage I non– small cell lung cancer. Clin Cancer Res 2002;8:3669–75. [52] Lee H, Chun K, Liu B, Wiehle S, Cristiano R, Hong W, Cohen P, Kurie J. Insulin-like growth factor binding protein-3 inhibits the growth of non–small cell lung cancer. Cancer Res 2002;62:3530–7. [53] Man˜es S, Llorente M, Lacalle RA, Gomez-Mouton C, Kremer L, Mira E, Martı´nez AC. The matrix metalloproteinase-9 regulates the insulin-like growth factor-triggered autocrine response in DU-145 carcinoma cells. J Biol Chem 1999;274:6935–45. [54] Monni O, Hyman E, Mousses S, Barlund M, Kallioniemi A, Kallioniemi OP. From chromosomal alterations to target genes for therapy: integrating cytogenetic and functional genomic views of the breast cancer genome. Semin Cancer Biol 2001;11:395–401.