Study objective: We investigated the gene expression profiles of malignant pleural mesothelioma (MPM) specimens to identify novel genes that are potentially involved in the oncogenic transformation of human pleural cells.

Design: Complementary DNA (cDNA) microarray transcriptional profiling studies of 10 MPM cell lines and 4 MPM primary tumor specimens were performed using hierarchic clustering. To confirm microarray data, we used real-time polymerase chain reaction and immunoblotting.

Results: Cluster analysis differentiated among epithelial (E), sarcomatoid, and biphasic MPM variants. Expression profiling identified common overexpressed or underexpressed genes in MPM. Notably, matriptase messenger RNA was found to he overexpressed by 826-fold in E MPM, with protein expression subsequently confirmed by immunoblot analysis. This recently characterized trypsin-like serine protease has been implicated in tumor invasion and metastasis of E-derived cancers, but has not been described until now in MPM. We also identified other novel genes, such as insulin-like growth factor binding protein 5 and a cDNA clone similar to proteolipid MAL2.

Conclusions: Thus, further large-seale profiling of MPM may elucidate previously unrecognized molecular mechanisms by identifying novel genes that are involved in malignant transformation. Our study has now found matriptase to be one of these mesothelioma-associated genes, with potential pathogenic and therapeutic significance.

Key words: expression profiling; matriptase; mesothelioma; microarray; real-time polymerase chain reaction

Abbreviations: B = biphasic; cDNA = complementary, DNA; Ct = threshold cycle; E = epithelial; GUSB = [beta]-glueuronidase; IGF = insulin like growth factor; IGFBP = insulin like growth factor-binding protein; MPM = malignant pleural mesothelioma mRNA = messenger RNA; QRT-PCR = quantitative real time polymerase chain reaction; S = sarcomatoid

**********

Malignant pleural mesothelioma (MPM) is an aggressive neoplasm of the serosal lining of the pleural cavity arising from mesothelial cells (ie, from undifferentiated cells representing the adult remnants of the surface coelomic mesoderm). Mesothelial cells are biphasic (B) and may give rise both to the lining keratin positive epithelial (E) cells and to the paucicellular submesothelial layer. The biphasic nature of mesothelial cells results in the following three major forms of MPM: E; sarcomatoid (S), or fibrous; and B. MPM currently accounts for about 2,500 to 3,000 deaths per year in the United States. (1) Up to 80% of eases of MPM occur in patients 10 to 20 years after exposure to asbestos. (2) Due to this latency period between asbestos exposure and tumor development, the associated mortality rate in men, but not in women, continues to rise in industrialized countries at the rate of 5 to 10% per year, with a median survival time between 4 and 18 months. (3,4) This rising mortality rate is occurring despite the implementation of legislation limiting asbestos use and exposure in most industrialized countries.

Aside from asbestos exposure, other factors such as ionizing radiation or tumor DNA virus sinvian virus-40 may act synergistically in MPM pathogenesis. (5,6) Also, several well-defined acquired genetic targets have been identified in MPM, including the 9p21 locus (p161NK4a, p14ARF) and the 22q11-q13.1 locus (NF2). (2) However, the molecular mechanisms controlling the transformation of mesothelial cells remain poorly defined. This is underscored by the observation that these well-characterized etiologies incompletely account for the known incidences of MPM. About 10 to 20% of MPM occurrences have been documented in patients without previous exposure to asbestos, (2) and only 60% of MPM tumors are known to contain SV-40 viral DNA. (6) Accordingly, multiple active pathways are thought to be possible.

Only two studies (7,8) to date have used the microarray technique to search for additional genes and pathways that are potentially involved in MPM biology. These investigations were limited by the extremely small number of MPM cell lines examined and by the lack of additional analysis of primary tumor tissues. To identify novel candidate genetic targets, we used hierarchic cluster analysis in our current study to compare the gene profiles of a larger sampling to MPM cell lines and primary tumors vs nonmalignant mesothelium on complementary DNA (cDNA) microarrays.

MATERIALS AND METHODS

Cell Lines and Tumor Tissues

A total of 11 cell lines and 4 primary tumor specimens were studied (Table 1). The CRL-2081 MPM cell line, the SV40 virus-transformed mesothelial cell line CRL-9444, and the breast cancer cell line MCF-7 (a positive control (9) for the matriptase messenger RNA [mRNA] and protein expression studies) were used (American Tilde Culture Collection; Manassas, VA). An additional nine MPM cell lines developed and previously characterized by Harvey Pass were obtained from the National Cancer Institute. (10) The mesothelial cell line was grown and propagated in M199 medium (BioSource International; Camarillo. CA), and was supplemented with 10% fetal calf serum according to instructions. All remaining cell lines were maintained in Roswell Park Memorial Institute-1640 mechum (Invitrogen Life Technologies; Carlsbad, CA) supplemented with 10% fetal calf serum and antibiotics. Separate paraffin cell blocks were prepared corresponding to each cell line, and our surgical pathologist confirmed the correct subtype classification.

All available primary tumor specimens (E = 3; B = 1) were obtained from the Tissue Procurement Facility at the University of Minnesota Cancer Center, in accordance with the policies of our Institutional Review Board. These surgical specimens were colleted flesh, were immediately snap-frozen in liquid nitrogen. and were stored at -80[degrees]C until use. All tissue specimens were verified by histopathologic studies and immunohistochemical studies (cytokeratin 5/6, 7. and 20: calretinin: E-cadherin; Ber-EP4; CD15; carcinoembryonic antigen; TTF-1; and B72.3) as containing relatively pure tumor.

RNA Isolation

Cells growing asynchronously were lysed in reagent (Trizol; Invitrogen Life Technologies) when they reached about 80% confluence and then were processed as described previously, (11) with modifications. Before RNA precipitation by isopropyl alcohol, the RNA aqueous phase was mixed with an equal volume of 70% ethanol, and total RNA was subsequently extracted on a silica gel-based membrane spin column (Qiagen; Valencia, CA) per the manufacturer's instructions. On-column deoxyribonuclease digestion was performed with all samples to eliminate potential genomic DNA. RNA yield and purity, were determined by, spectrophotometry. Integrit was verified on 1.5% agarose-formaldehyde gels stained with ethidium bromide. Tissue specimens were processed in a similar fashion, each starting from 100 tug frozen tumor.

Microarrays

Microarray experiments were performed (MieroMax Human cDNA Microarray System II TSA; Perkin Elmer LifeSciences; Boston, MA) according to the manufacturer's instructions. Briefly, 1 [micro]g total RNA of each MPM sample was reverse transcribed and simultaneously labeled with fluorescein-deoxyuridinetriphosphate to produce target cDNA. CRL-9444 target cDNA was labeled with biotin-deoxyuridinetriphosphate and served as the reference "normal" sample in each microarray experiment, as validated by others. (7) Target cDNA was hybridized on 4.800 gene MicroMax microarrays. The washing and detection steps were based on a sequential fluorescence detection process with horseradish peroxidase-conjugated antibodies and tyramide-linked cyanine-3 (Cy3) or cyanine-5 (Cy5) dyes. (12,13) Microarray slides were scanned at 532 nm (Cy3, MPM samples) and 635 nm (Cy5, reference) [ScanArray Express confocal laser scanner; Perkin Elmer. Some microarray experiments were selected for duplicate processing to assure reproducibility

Microarray Analysis

Each gene was represented on a microarray in duplicate, so that a corresponding raw expression ratio was defined as the mean signal intensity of the Cy3/Cy5 replicates. Expression ratios then were filtered per microarray, according to the following algorithm, (14) with modifications. Briefly, the 50th percentile of all measured expression ratios was used as a positive control for each gene. Every gene measurement was divided by this synthetic positive control, assuming that it was at least 0.01. The bottom 10th percentile of expression ratios was used as a test for correct background subtraction. This was never less than the negative of the synthetic positive control. Across all 14 MPM specimens, the threshold values used to define significant relative expression changes were set at 3.0 for overexpression and 0.30 for underexpression.