Differential Gene Expression in Human Lung Adenocarcinomas and Squamous Cell Carcinomas¹ Free

Despite significant progress toward the elucidation of human lung cancer tumorigenesis in the past 2 decades, it still remains the leading cause of cancer-related deaths in the United States, accounting for 25% (women) and 31% (men) of all cancer deaths annually (1). It is estimated that lung and bronchus cancer accounts for only 12–14% of new cancer cases per year in the United States, but its early metastatic spread results in a 5-year-patient-survival of only 10–15% (1, 2). Both genetic and environmental factors contribute to the development of lung cancer. Eighty percent of lung cancers develop in current or former smokers with perhaps an additional 5% arising from passive tobacco smoke exposure (3). Additional risk factors include asbestos, radon, and radiation exposure (4, 5). Susceptibility to lung cancer may be determined in part by the capacity to activate and detoxify inhaled procarcinogens from the environment (6, 7).

There are four major histological types of lung cancer: SCLC,³ adenocarcinoma, squamous cell carcinoma, and large cell carcinoma. Clinically and therapeutically, lung cancer is categorized as SCLC and NSCLC. SCLC comprises ∼20% of all lung cancers, is neuroendocrine in origin, and, although highly aggressive, can often be treated with combination chemotherapy. NSCLC consists of the three remaining subtypes and is predominantly treated with surgery. Moreover, chemotherapy or radiation therapy are often used in the treatment of late-stage NSCLC (3, 5).

The molecular pathogenesis of lung cancer routinely involves the activation of oncogenes (K-ras) and the inactivation of tumor suppressor genes (p53, p16^INK4a, and Rb). Defects in DNA repair pathways and cell cycle checkpoint pathways allow tumor cells to accumulate mutations that are advantageous to growth, invasion, and metastasis. Accumulation of these genetic alterations results in the phenotypic progression from normal to neoplastic lung tissue (8, 9).

Many of the genetic abnormalities that have been detected in human lung cancer occur in both SCLC and NSCLC. However, some alterations are more prevalent in specific histological subtypes. For example, the transmembrane receptor tyrosine kinase ERBB2 is expressed at high levels in ∼30% of NSCLCs, particularly in adenocarcinomas, and has been found to contribute to the tumorigenicity and metastatic potential of cells in vitro (2, 6). ERBB2 and its ligand neuregulin initiate a signal transduction cascade involving the mitogen-activated protein kinase pathway and may constitute a growth stimulatory loop in lung cancer (10). The mitogenic and motogenic hepatocyte growth factor is predominantly expressed in NSCLCs and, along with its MET proto-oncogene receptor, may form a second autocrine growth stimulatory loop (11, 12). Coexpression of c-kit proto-oncogene and its ligand, stem cell factor, is common in SCLCs and may mediate chemoattraction or stimulate tumor cell growth (13, 14). Ras mutations are detected frequently in NSCLCs (15–20%), commonly in adenocarcinomas (20–30%), but in <1% SCLCs (6, 15). p53 loss of heterozygosity with inactivating mutations in the retained allele has been detected in ∼50% of NSCLCs but in 75–100% of SCLCs (2, 6). Another reported difference is the more frequent detection of high Bcl-2expression levels in SCLCs (75–95%) compared with NSCLCs (10–35%). Overexpression of Bcl-2 is more commonly detected in squamous cell carcinomas (25–35%) than adenocarcinomas (10%; Refs. 2, 6). Other differential alterations include Rb inactivation in SCLCs and p16^INK4a inactivation in NSCLCs (2, 6, 16). Frequent chromosomal losses in SCLCs include 1p, 3p, 4q, 5q, 13q, 17p, and10q, whereas chromosomal losses in NSCLCs are common on 3p, 6q, 8p, 9p, 9q, 13q, 17p, 18q, 19p, 21q, and 22q (17, 18, 19).

CCLS (20) was used to examine genes that were differentially expressed in adenocarcinomas and squamous cell carcinomas compared with corresponding normal lung tissues of the same patient. Normalized comparisons of gene expression between adenocarcinomas and squamous cell carcinomas were also evaluated. Using this technique, we confirmed underexpression of 68 genes/sequences and overexpression of 4 genes/sequences in both NSCLC subtypes compared with normal lung tissues. We also detected 10 genes/sequences that were preferentially over- or underexpressed in only one NSCLC subtype. Finally, 5 genes and 1 putative gene locus were overexpressed in 1 NSCLC subtype while underexpressed in the other subtype. These differentially expressed genes may play pivotal roles in lung tumorigenesis and may potentially serve as biomarkers in both diagnosis and prognosis of human lung cancer.

Frozen lung cancer specimens and matching normal tissues were obtained from Cooperative Human Tissue Network through The Ohio State University Department of Pathology. Fourteen pairs of clinical samples (7 squamous cell carcinomas with their normal controls and 7 adenocarcinomas with their normal controls) were used in this study (Table 1). Frozen tumor tissues were microdissected to determine the tumor versus normal cell ratio for each specimen. Tissues were embedded in Tissue Tek OCT compound (VWR Scientific Products, West Chester, PA), cryostat sectioned, and stained with H&E for microscopy. Tumor tissue sections corresponding to the microscopic sections containing ≥70% tumor cells were isolated and stored at −80°C for subsequent RNA isolation. Matching normal tissues were also microdissected to ensure that specimens consisted of purely normal lung tissue. Fig. 1 shows the typical morphology of normal alveoli and an adenocarcinoma, as well as a normal bronchiole and a squamous cell carcinoma used in this study.

The Uni-ZAP XR vector human lung cDNA library that was used was derived from the normal lung bronchoalveolar cells of a 72-year-old male smoker (one-quarter packs per day; Stratagene, La Jolla, CA).

Extraction of total RNA from frozen tumor and corresponding normal tissues was carried out by ultracentrifugation over a cesium chloride cushion. Tissues were pulverized in liquid nitrogen and cells were lysed in guanidinium thiocyanate homogenization buffer (4 m guanidinium thiocyanate, 25 mm trisodium citrate, 2% sodium N-lauroylsarcosine, and 1% 2-mercaptoethanol) before separation. Samples were spun at 39 K and 20°C for 18 h. RNA pellets were then resuspended in guanidinium thiocyanate homogenization buffer and ethanol precipitated at −20°C overnight.

A portion of the cDNA library was plated onto 200 LB-agarose plates and transferred to nitrocellulose filters. Duplicate filters were made for each plate. DNA was then denatured in 1.5 m NaCl/0.5 n NaOH, neutralized in 2.5 m NaCl/1 m Tris (pH 7.5), and washed in 3× SSC.

Total RNA (2 μg) and 2.5 μg oligodeoxythymidylic acid primer was denatured at 65°C for 10 min and placed on ice. To this mixture the following components were added: 1 × reverse transcription buffer [50 mm Tris-HCl (pH 8.3), 75 mm KCl, and 3 mm MgCl₂]; 10 mm DTT; 1 mm of each dATP, dTTP, and dGTP; 100 μCi [α-³²P]dCTP or 1 mm dCTP; 60 units RNasin (Promega Corporation, Madison, WI); and 800 units Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc., Carlsbad, CA). Reaction mixtures were incubated at 37°C for 1 h.

Tumor and corresponding normal RNAs were reverse transcribed into [α-³²P]dCTP-labeled cDNA. Another reverse transcription of tumor RNA was performed and served as the cold competitor for both labeled tumor and labeled normal cDNAs. Probes were then purified by centrifugation through G50 Sephadex columns. Hybridization solution was composed of 5 × saline-sodium phosphate-EDTA, 1 × Denhardt’s solution, 0.1% SDS, 50% formamide, and 100 μg Herring sperm DNA. Filters were hybridized overnight at 42°C. After hybridization, filters were washed in 2× SSC/0.1% SDS twice for 10 min at room temperature and 0.1× SSC/0.1% SDS twice for 25 min at 55°C. Signal intensity differences between normal and tumor-hybridized filters suggested differential expression.

Additional confirmation of differentially expressed clones selected from the primary library screening was performed using dot blot analysis. pBluescript phagemid DNA was isolated using the mini prep method described by Sambrook et al. (21). DNA (∼2 μg) was denatured with 0.1 volume 2 n NaOH/2 mm EDTA and neutralized with 0.1 volume 3 m sodium acetate (pH 5.25). Denatured DNA was then spotted onto nylon membranes using the HYBRI.DOT Manifold Apparatus (Life Technologies, Inc.). A duplicate membrane was made for each set of clones screened. Normal and tumor sample cDNA was then hybridized to the fixed clone DNA using the same conditions described previously. GAPDH cDNA was spotted onto each set of nylon membranes to serve as a hybridization control. Signal intensities were compared by densitometry using ImageQuant software.

Sequencing of clone DNA confirmed by dot blot analysis was initially performed using Sequenase version 2.0 DNA sequencing kit purchased from Amersham Pharmacia Biotech (Piscataway, NJ). Sequencing primers used were sequences corresponding to the cloning vector arms of the pBluescript SK(+/−) phagemid: SK primer: 5′ CGCTCTAGAACTAGTGGATC 3′ and KS primer: 5′ TCGAGGTCGACGGTATC 5′. DNA sequences were entered into GenBank Blast search (22) for comparison with known gene sequences. Subsequent automated sequencing was performed on the remaining clone DNA as follows. Single-stranded DNA (100 ng) and 3.2 pmol SK primer was added to 4 μl of Ready Reaction Mix (Applied Biosystems, Foster City, CA) and 2 μl 5 × sequencing buffer (400 mM Tris HCl, 10 mM MgCl₂, pH 9.0) (Applied Biosystems) to a total volume of 10 μl and subjected to 25 cycles of PCR at 96°C for 10 s, 50°C for 5 s, and 60°C for 4 min. Extension products were then purified by ethanol precipitation and dissolved in 4 μl of loading buffer (2:1 ratio of formamide to dextran-EDTA loading dye). Clone DNA was electrophoresed and analyzed on the ABI PRISM 377 DNA sequencer.

Primers for each differentially expressed gene were designed and ordered from Life Technologies, Inc. Each primer was diluted to 1 optical density/100 μl and forward primers were then end-labeled with [γ-³²P] dATP using T4 polynucleotide kinase (New England Biolabs, Beverly, MA). Normal and tumor tissue total RNA (4 μg) was reverse transcribed to cDNA as described previously. PCR reactions using 1 μl cDNA, 1 × reaction buffer, 2 mm MgCl₂, 0.2 mm deoxynucleotide triphosphates, 0.75 units of Taq DNA Polymerase (Promega), labeled forward primers, and unlabeled reverse primers were subjected to 18–23 cycles of amplification at 95°C for 1 min, 55°C for 1 min, and 72°C for 1 min/cycle. GAPDH was used as an internal control in each PCR amplification. PCR products were then electrophoresed on 8% polyacrylamide gels with urea and exposed to Molecular Dynamics (Sunnyvale, CA) PhosphorImager screens for analysis. Densitometry was performed by ImageQuant software.

A normal human lung cDNA library from Stratagene was plated out on 200 LB-agarose plates containing approximately 500–700 plaques/plate and covering an estimated 100,000–140,000 clones expressed in the library. Fig. 2 shows schematically the CCLS procedure. Briefly, cDNA clones were transferred in duplicate to nitrocellulose membranes, and each membrane was hybridized with reverse transcribed-³²P-labeled mRNA from either normal or tumor tissue obtained from the same patient. Unlabeled reverse transcripts of tumor mRNA served as the “cold” competitor and ensured that labeled signals were not saturated during hybridization. Tissue availability constraints required the use of tumor “cold” competitors and resulted in the detection of primarily under-expressed clones in the cDNA library. Both NSCLC subtypes and their matching normal tissues were hybridized to 100 filter pairs each. As shown in Fig. 3, each filter contained ∼128 spots representing individual clone DNAs that were detected in the tumor or normal tissues. Approximately 22 clones/plate appeared to be differentially expressed between tumor and matching normal tissues (Fig. 3).

A total of 2790 putative differentially expressed clones were selected from the 200 LB-agarose plates for additional confirmation by dot blot analysis (Fig. 4). Of the 2758 clones from which DNA was successfully isolated, 1163 were confirmed as putative positive clones (42%). Clone DNA was then sequenced and entered into GenBank Blast search (22) for comparison with known gene sequences. The frequency of detection of differentially expressed genes was as follows: surfactant proteins, 34%; immunoglobulins, 20%; mitochondrial genes, 13%; other less frequently found known genes, 22%; unknown sequences that did not show significant homology to genes in the database, 11%; and empty vector sequences, <1%. Primers were designed for those sequences with the highest homology to known genes or DNA sequences in GenBank. Final confirmation of differential expression was determined by RT-PCR analysis (Fig. 5). Of the 121 genes/sequences subjected to RT-PCR, 113 were differentially expressed in at least 1 or more NSCLC tissue samples. However, differential expression in ≥50% of the adenocarcinomas and squamous cell carcinomas was used as the cutoff limit for additional analysis. As shown in Table 2, 92 genes were analyzed for differential expression in all 7 adenocarcinomas and 7 squamous cell carcinomas compared with their corresponding normal lung tissues. Genes/sequences that are located in chromosomal regions that are frequently lost in NSCLCs are listed in Table 3. Additional verification of differential gene expression in either squamous cell carcinomas or adenocarcinomas by Northern blot analysis could not be performed because of the limited amount of microdissected lung tissues available for the present study.

Specifically, we detected underexpression of 68 genes and overexpression of 4 genes in both adenocarcinomas and squamous cell carcinomas compared with normal tissues. We also found several genes that were differentially expressed in one NSCLC subtype but rarely in the other subtype. Eight of these genes were underexpressed (TS, locus Xq21.1, Tβ4, RPS3, UBC, mybpc3, NADH dehydrogenase III, and Wnt-13), whereas only two genes were overexpressed (hepatocyte growth factor A inhibitor and UBC9) compared with normal lung tissues. Another subset of genes were generally underexpressed in one NSCLC subtype but overexpressed in the other subtype (casein kinase II, locus 8p21.3, Ly-GDI, CFI, RPL18, and EF1G) compared with normal controls. All of these gene transcript levels except casein kinase II were lower in squamous cell carcinomas, whereas they were higher in adenocarcinomas with respect to normal tissues.

Changes in gene expression affect the biological functions of all living cells. Cell growth, differentiation, development, apoptotic cell death, and cellular senescence are some of the normal cell processes regulated by changes in gene expression patterns. Dysregulated gene expression often results in pathological conditions such as cancer (23, 24). To examine the gene expression changes that occur in adenocarcinomas and squamous cell carcinomas of the human lung, we have used the CCLS method. Approximately 100,000 clones were screened from a human lung cDNA library derived from the normal bronchoalveolar cells of a 72-year-old male smoker. Of the ∼2,800 putative positive clones selected from the preliminary screening, ∼1,200 were confirmed by dot blot analysis and sequenced. Approximately 120 known genes and DNA sequences were then subjected to RT-PCR analysis for final confirmation.

There are many advantages in the use of CCLS to determine gene expression changes in cancer. For example, expression differences for both known and novel genes can be detected. Because the cDNA library is not normalized to ensure approximately equal representation of poly(A)⁺ RNA sequences, detection frequencies of differentially expressed genes can be determined, indicating the relative frequency of mRNA expression in the normal lung tissue. Additionally, in-depth sampling of gene expression changes for >100,000 clones is possible. Although CCLS is laborious and time-consuming, screening data are extensive and allow for the additional characterization and functional analysis of unknown genes as well as examination of the potential roles of known genes in lung tumorigenesis.

The use of CCLS methodology to screen a normal human lung cDNA library yields a number of implications. First, if a gene that is found in tumor tissue is not expressed at all in the normal lung, it would not be detected in this study. Secondly, genes that are expressed at very low levels in normal lung were probably missed using this method. Thirdly, because the cDNA library is not normalized, clones associated with significantly overexpressed genes may be repeatedly selected. Thus, there were hundreds of independent clones identified that proved by sequencing to be either surfactants or Clara cell gene −CC10. Lastly, because of the use of a cold tumor competitor, most differentially expressed genes were underexpressed in NSCLCs compared with normal lung tissues. Therefore, this methodology focused on the identification of tumor suppressor genes, differentiation-specific genes, and genes involved in escape from immune surveillance and cytoskeletal restraints that were expressed at high enough levels for initial screening detection. Moreover, activated oncogenes were poorly represented in the normal lung library and were, therefore, rarely detected in this study.

A total of 76 genes/sequences were predominantly underexpressed, and 6 genes were predominantly overexpressed in squamous cell carcinomas, adenocarcinomas, or both. Ten genes/sequences were preferentially over- or underexpressed in only 1 NSCLC subtype. Five genes and 1 putative gene locus exhibited opposite expression patterns in squamous cell carcinomas and adenocarcinomas compared with normal lung tissues. Therefore, although most gene expression differences were similar for both tumor types, there were some distinct differences that suggested that the NSCLC subtypes might follow subtly different pathways to tumorigenesis.

Pulmonary surfactant mRNA transcripts were the most frequently detected differentially expressed genes in this study. Pulmonary surfactant is a phospholipid-protein complex that functions in normal respiration to lower the air-liquid interface surface tension in lung alveoli (25). SPA, in particular, stabilizes surfactant, controls recycling of surfactant by type II cells, and boosts antimicrobial host defenses. It also regulates the transcription of surfactant proteins, surfactant receptor, and c-jun through a receptor-mediated pathway. It is considered to be a marker of epithelial differentiation in the lung (26). NSCLCs are derived from respiratory epithelium and, therefore, occasionally express SPA (27). However, adenocarcinoma of the lung is the predominant subtype of NSCLC that expresses SPA and SPB, and expression is only detected in approximately 30–50% of cases (28, 29, 30). SPA and SPB were detected in normal lung tissues but not in lung carcinoma cell lines or primary NSCLCs by mRNA differential display and magnet-assisted subtraction techniques, respectively (31, 32). This suggested either dedifferentiation of the tumor cells or tumor origination from a cell type that did not express surfactant. In contrast, Pilling et al. (33) detected SPA, SPB, and SPC mRNA expression in all of the B6C3F1 non-neoplastic lung tissues and mouse lung tumors examined. SPA mRNA levels were also increased in urethane-induced A/J, BALB/c, and (A/J/C3H/He)F1 mouse lung tumors compared with normal controls (34).

Several genes involved in cell adhesion and cytoskeletal integrity were found to be decreased in expression in NSCLCs. Cellular fibronectin is a high molecular weight extracellular matrix glycoprotein that can affect cell adhesion, morphology, surface architecture, migration, and differentiation in numerous cell types (35, 36). It is absent or greatly reduced in many transformed cells resulting in loss of contact inhibition and potential migration (37, 38). Occludin is an integral membrane protein that functions as a component of tight junctions. Tight junctions seal adjacent cells together and allow for membrane domain differentiation (39). Receptor protein tyrosine phosphatase-μ is abundantly expressed in the lung (40) and associates with the cadherin-catenin complex involved in homophilic cell adhesion at adherens junctions. Protein phosphatase-μ specifically promotes this cell adhesion through dephosphorylation of the cadherin-catenin complex (41). Ezrin (cytovillin) is a cytoplasmic peripheral membrane protein that is highly expressed in lung, intestine, and kidney, and is present in microvilli and other cell surface structures (42). It is capable of binding to actin and links the actin cytoskeleton to the plasma membrane (43). Ezrin family members are involved in cell adhesion, cell migration, and maintenance of surface architecture (44, 45).

Another differentially expressed group of genes function in the immune or inflammatory response, including CFI, β-2 microglobulin, MHC I and II, serglycin, and MCP-3. Serglycin is a widely distributed proteoglycan and ligand for the transmembrane glycoprotein CD44 (46). It acts as a packaging scaffold and chaperone for extracellularly secreted granzymes that possess apoptotic activity (47). MCP-3 is a chemokine that recruits macrophages, leukocytes, and eosinophils during inflammation, cancer invasion, and allergen challenge (48, 49). Gene transfer of MCP-3 into P815 mastocytoma cells resulted in reduced tumorigenicity and increased leukocyte infiltration through the activation of type I T-cell-dependent immunity (50).

Several MHC class I and class II genes were differentially expressed in both adenocarcinomas and squamous cell carcinomas, including HLA-DRβ1, HLA-DRα, HLA-DPα1, HLA-E, and β-2 microglobulin (HLA class I light chain). These genes code for antigen-presenting proteins that present processed cellular or foreign antigens to T cells, potentially evoking an immune response (51). Reduced or absent surface expression of HLA class I molecules, in particular, has been observed in colorectal, melanoma, and Burkitt lymphoma cell lines (52, 53, 54).

Several enzymes of the electron transport chain in mitochondria were underexpressed in these NSCLC subtypes compared with normal lung tissues. Most squamous cell carcinomas examined in this study exhibited decreased expression of these genes. However, differential expression in adenocarcinomas was detected in <50% of tumors. Previous studies have reported elevated activities of cytochrome c oxidase in bladder cancer and malignant breast tumors, and a significantly higher expression level of ND5 in a highly metastatic large cell lymphoma cell line (55, 56, 57). However, Wong et al. (58) reported that down-regulation of ND3contributed to the resistance of Al0A cells to reactive oxygen species-mediated doxorubicin-induced apoptosis. Finally, the flat, revertant cell line Rl expressed high transcript levels for cytochrome b, cytochrome c oxidase subunit II, and NADH dehydrogenases 1 and 4 compared with a human H-ras oncogene-transformed NIH3T3 cell line, reflecting decreased enzyme expression in the transformed morphology (59).

TGF-β2 is secreted by several tumor types, including breast carcinomas, melanomas, and glioblastomas (60, 61, 62). Because TGF-β receptors are often nonfunctional on tumor cells, the growth arrest or apoptotic effects elicited by TGF-β often affect normal host cells, such as T cells, and result in immunosuppression in the host (60, 63). Although differential expression of TGF-β itself was not detected in these normal and tumor tissues, numerous TGF-β activators and effectors were identified.

EGR-1 is a DNA-binding transcription factor that functions in cell differentiation, development, proliferation, cell cycle regulation, and apoptosis in a cell type-specific manner. It induces fibronectinand TGF-β1 transcription, which results in increased expression of plasminogen activator inhibitor 1 (64, 65). The secretion of fibronectin and plasminogen activator inhibitor 1 enhances cell attachment and suppresses transformation (65). Decreased or absent EGR-1 expression was detected in 72% of 101 human NSCLC cases compared with normal bronchial tissues (66). Primary human breast carcinomas and breast carcinoma cell lines showed a similar lack of EGR-1 expression (67).

CTGF is a secreted, cysteine-rich protein that is induced by TGF-β in connective tissue cells. It is a potent angiogenic factor that promotes adhesion, proliferation, and migration of vascular endothelial cells in vitro (68). Overexpression of CTGF has been reported in melanomas, chondrosarcomas, and pancreatic cancer (69, 70, 71). However, Igarashi et al. (72) reported an inverse correlation between CTGFexpression levels and the malignant transformation of fibroblast and endothelial cell-derived tumors.

Another set of genes that were differentially expressed in these NSCLCs helps to control cell migration and/or metastasis. Platelet endothelial cell adhesion molecule 1 is a member of the immunoglobulin superfamily. It is normally present on the surface of circulating platelets, neutrophils, monocytes, and endothelial cells, and functions in angiogenesis, integrin activation, and transendothelial migration (73). It is highly expressed at endothelial intercellular junctions and functions as a vascular barrier (74). It can also associate with several intracellular protein tyrosine phosphatase signaling molecules such as SHP1, a protein that negatively regulates Kit receptor tyrosine kinase signaling in a cell-type specific manner in vivo (75, 76).

A2M is a protease inhibitor that binds and inhibits proteases involved in coagulation, inflammation, and fibrinolysis, such as trypsin, thrombin, and collagenase (77, 78). Endopeptidases that are secreted by tumor cells and digest basement membrane contribute to tumor invasion and metastasis (79). A2M also plays a role in growth regulation by binding to cytokines and growth factors such as TNF-α, interleukin 6, platelet-derived growth factor, TGF-β1, TGF-β2, and nerve growth factor (77). Although we found decreased levels of A2M in both squamous cell carcinomas and adenocarcinomas, Marchandise et al. (80) reported increased levels in lung cancer compared with normal lung tissue. Metastatic colon carcinoma cells also produced and secreted large amounts of A2M, which seemed to correlate with their tumorigenicity (81).

Several chromosomal locations harboring differentially expressed DNA sequences were detected using this method. Many of these sites cover regions that have been reported previously to exhibit frequent chromosomal losses in NSCLC: 6q12, 6q24, 8p21.3-p22, 9q34, 21q22.3-qter, 22q12-qter, and chromosome 17 (17, 19). These regions are thought to harbor putative tumor suppressor genes.

We performed preliminary studies on 121 known genes and DNA sequences to determine their differential expression in human lung adenocarcinomas and squamous cell carcinomas using a CCLS technique. Although most expression differences were detected in both NSCLC subtypes, several changes were more frequently detected in one subtype compared with the other subtype. Additionally, some expression differences were opposing changes in squamous cell carcinomas and adenocarcinomas compared with normal tissue controls. These discrepancies suggest subtle differences in their progression to the neoplastic state. Future in-depth studies of these expression alterations need to be pursued to determine their functional significance in lung tumorigenesis. Several of these changes may have arisen as by-stander effects to truly critical genetic alterations and may, therefore, serve as potential biomarkers for early cancer detection and intervention.

Our preliminary results provide many possible avenues to explore in the discovery and understanding of the complex pathological changes that lead to lung tumor progression. It is interesting to note that most of the genes and DNA sequences detected using this method were underexpressed in tumor tissues compared with normal lung tissues and are putative tumor suppressor genes. Although several chromosomal loci are still thought to harbor unknown tumor suppressor genes involved in lung cancer, a more limited number of genes such as p53, Rb, p16^INK4a, and fragile histidine triad have been positively identified as contributors to the tumorigenic process. Additional investigation of the genes and DNA sequences detected in this study may elucidate additionally important genetic alterations and may supply novel gene therapy targets for cancer chemoprevention and treatment.