Establishment and Expression Profiling of New Lung Cancer Cell Lines from Chinese Smokers and Lifetime Never-Smokers

ORIGINAL ARTICLE

David C. L. Lam, MBBS,*† Luc Girard, PhD,‡ Wai-Sing Suen, MBBS,§ Lap-ping Chung, DPhil,† Vicky P. C. Tin, BSc,† Wah-kit Lam, MD,* John D. Minna, MD,‡ and Maria P. Wong, MD†

regulation of cell cycle and proliferation (CDKN1A and CDKN2A), Background: Bronchogenic adenocarcinoma is the predominant a putative oncogene (EEF1A2), and a gene related to metastasis histologic subtype of lung cancer, which ranks top in the cancer (MTSS1). mortality in both men and women. Female nonsmokers and adeno- Discussion: Four new lung adenocarcinoma cell lines were estab- carcinoma have emerged as a distinct combination in patients with lished from patients with different clinical characteristics. These lung cancer in recent decades. Lung cancer cell lines established characterized cell lines and their gene expression profiles will from patients with known clinical characteristics such as gender and provide resources for studies of lung cancer biology and in vitro smoking habit would be useful for future research on lung cancer. chemotherapeutic drug study. Methods: Four new lung adenocarcinoma cell lines (HKULC 1–4) were established from Chinese patients with primary lung adeno- Key Words: Lung adenocarcinoma, Cell line, Non-smokers. carcinomas and with different clinical characteristics with respect to (J Thorac Oncol. 2006;1: 932–942) age, gender, smoking habits, tumor staging, and previous therapy. They were characterized by immunohistochemical and growth kinetic studies, tests for tumorigenicity in nude mice, epidermal espite continuous efforts and clinical trials to develop growth factor receptor (EGFR) gene mutation analysis, and in situ Dnew diagnostic, prognostic, and therapeutic methods, hybridization, and gene expression profiling with Affymetrix Gene- lung cancer remains the leading cause of cancer morbidity Chip HG-U133A. and mortality in Hong Kong1 and worldwide.2 In Hong Kong, Results: The newly established HKULC lung adenocarcinoma cell lung cancer accounts for 31% of cancer deaths in men and lines were maintained for over 70 passages and demonstrated 25% in women, with a crude incidence rate of 56.8/100,000 morphologic and immunohistochemical features and growth kinetics population and a crude mortality rate of 48.6/100,000 popu- of tumor cell lines. One of the four HKULC cell lines, HKULC 3 lation.1 In recent decades, the combination of female, non- (derived from a female nonsmoking patient with lung adenocarci- smoker, and adenocarcinoma has emerged as prominent char- noma), was found to have a deletion at exon 19 of the EGFR gene. acteristics in a significant subgroup of lung cancer patients EGFR in situ hybridization showed no EGFR gene amplification in compared to male, smokers, and squamous cell carcinoma in these cell lines. HKULC 1 and 4 formed tumor xenografts after the past.3,4 A better understanding of the pathogenesis and inoculation in nude mice. A list of 71 genes that were differentially delineation of the major molecular pathways involved in lung expressed or showing class predictive significance was identified. cancer is essential for the identification of new molecular These genes included putative tumor suppressor genes (DKK3, targets for developing new diagnostic and targeted therapy of SERPINF1, CDH11, DSC3, and KLF6), genes involved in or related lung cancer. to the EGFR pathways (ERBB3, MUC1, VAV1), genes involved in Despite technical difficulties, lung cancer cell lines have been established,5–7 including sublines with resistance *University Department of Medicine and †Department of Pathology, Queen or sensitivity to defined chemotherapeutic agents.8–10 In Mary Hospital, University of Hong Kong, Hong Kong SAR, China; ‡Hamon Center for Therapeutic Oncology Research, University of Texas Hong Kong, we are encountering an increasing proportion of 1 Southwestern Medical Center at Dallas, Dallas, TX; §Cardiothoracic lung cancer patients who are females and nonsmokers. The Surgical Unit, Grantham Hospital, Hong Kong SAR, China. aim of this study was to report the establishment and char- The work described in this article was partially supported by grants from the acterization of four new lung adenocarcinoma cell lines Research Grants Council of the Hong Kong Special Administrative (named HKULC 1–4) from Chinese patients with known Region, China (Project No. HKU7468/04M) and the Lung Cancer SPORE Grant P50CA70907, United States. clinical characteristics of age, gender, smoking habits, and Address for correspondence: Maria P. Wong, MD, Department of Pathology, previous therapy. The morphologic, growth kinetics proper- Queen Mary Hospital, University of Hong Kong, Hong Kong SAR, ties, and epidermal growth factor receptor (EGFR) mutation China. E-mail: [email protected] status of the cell lines were described. Expression microarray Copyright © 2006 by the International Association for the Study of Lung Cancer studies were performed to document the gene expression ISSN: 1556-0864/06/0109-0932 profiles of these newly established HKULC cell lines. These

932 Journal of Thoracic Oncology • Volume 1, Number 9, 2006 Journal of Thoracic Oncology • Volume 1, Number 9, 2006 Characterization and Profiling of Lung Cancer Cell Lines cancer cell lines (HKULC) could be useful tools for future ease was achieved for 12 months, but subsequently the tumor lung cancer research and investigation of molecular and thera- progressed with development of malignant pleural effusion. peutic targets for diagnostic and therapeutic intervention. Patient 4 (HKULC 4): Male Ex-Smoker, Surgically PATIENTS AND METHODS Resected Tumor, Stage IIIA HKULC 4 was cultured from a 52-year-old male ex- Clinical Characteristics smoker who presented with synchronous tumors of the left Patient 1 (HKULC 1): Female Nonsmoker, Stage lower lobe (adenocarcinoma, stage T2N2M0 with subcarinal IIIB, Pericardial Effusion lymph node metastasis) and the right upper lobe (squamous HKULC1 was derived from the pericardial fluid of a cell carcinoma diagnosed by percutaneous transthoracic nee- 63-year-old nonsmoking woman with primary adenocarci- dle biopsy) at the time of resection. This cell line was noma of the left lower lobe that invaded the pericardial cavity cultured from the resected left lower lobe adenocarcinoma. (T4N2M0). She presented with pleuropericardial effusion and a poor performance status. A computed tomography (CT) Methods of Collection scan of the thorax and abdomen showed no other tumor site. All the specimens were collected after informed written Pericardiocentesis was performed to relieve cardiac tampon- consents were given by the respective patients. Institutional ade, and no palliative chemoradiation was given (Table 1). review board approval had been obtained from the University of Hong Kong/Hospital Authority Hong Kong West Cluster Patient 2 (HKULC 2): Male Smoker, Stage IV, Research Ethics Committee for this study. For patients 1 Pleural Effusion through 3, pleural and pericardial effusions were obtained HKULC 2 was derived from the pleural fluid of a with aseptic techniques and centrifuged to collect the malig- 65-year-old male chronic smoker who presented with right- nant cells. For patient 4, fresh tissue from the resected tumor side pleural effusion and ascites. Right-side pleurocentesis was cut with fine scissors for cell culture and transplantation for pleural fluid examination confirmed metastatic adenocar- into nude mice. Fresh tissue adjacent to the area used for cinoma. CT scan showed a contrast-enhancing tumor in the culture was snap frozen and kept at Ϫ70°C until use. Col- right lower lobe and metastatic nodules in the right middle and left lower lobes of his lung and in the peritoneal cavity lected specimens were then seeded into two different types of (T2N3M1). There was no other mass lesion in the liver, culture medium for incubation, ACL4 or RPMI, supple- spleen, and kidneys. The patient refused palliative chemo- mented with 10% fetal bovine serum to maximize the chance therapy, and therapeutic pleurocentesis was performed for of establishment. ACL4 medium was prepared according to 6 symptomatic relief. the protocol published by Oie et al. Penicillin and strepto- mycin (105 U) were added as antibiotic prophylaxis. The cells Patient 3 (HKULC 3): Young Female Nonsmoker, were then incubated in 37°C and 5% carbon dioxide. At- Stage IIIB, Pleural Effusion, Previous Therapy tached cells reaching confluent density were subcultured and With Paclitaxel/Carboplatin growth maintained in either ACL4 or RPMI, depending on HKULC 3 came from the pleural fluid of a 38-year-old which medium the cells adapted to in vitro. nonsmoking woman who initially presented with chest pain. CT scan showed a 3-cm peripheral tumor in the right lower Growth Kinetics Study lobe abutting the pleura with satellite nodules within the same To determine the growth rate, 105 cells of each cell line lobe (T4N2M0). CT-guided aspiration of the main tumor were seeded into 25-cm2 culture flasks with 5 ml of growth showed adenocarcinoma. She received six cycles of chemo- medium. Cell counts were made on days 1, 3, 6, 9, and 12 in therapy with paclitaxel and carboplatin. Initially, stable dis- triplicate flasks.

TABLE 1. Summary of the Characteristics of the Four HKULC Cell Lines Cell Lines HKULC 1 HKULC 2 HKULC 3 HKULC 4

Sex/age Female/63 Male/65 Female/38 Male/52 Smoking habits Lifetime nonsmoker Chronic smoker Lifetime nonsmoker Ex-smoker (had stopped smoking for Ͼ1 yr at time of diagnosis Tissue origin Pericardial fluid Pleural fluid Pleural fluid Resected tumor Tumor cell type Adenocarcinoma Adenocarcinoma Adenocarcinoma Adenocarcinoma Staging IIIB IV IIIB IIIA Previous treatment None None 6 cycles of chemotherapy None Tumorigenicity in nude mice Yes No No Yes EGFR mutation in exons No No Deletion at exon 19 No 18–21 in respective cell line

Test for Tumorigenicity in Nude Mice normalized by an in-house Visual Basic software MATRIX Tumor cells were harvested and counted with supravital (MicroArray Transformation in Excel) Version 1.30 written staining. An inoculum of at least2x105 cells was suspended by Luc Girard at the University of Texas Southwestern in 0.5 ml of the corresponding growth medium. The cell Medical Center at Dallas. The MATRIX program allowed suspension was then injected subcutaneously beneath the input of multiple CHP files (saved in text file format) from dorsal skin of a nude mouse, after which the mouse was MicroArray Suite 5.0 into an Excel spreadsheet where nor- maintained and observed for the next 12 months for the malization, comparison of arrays using log ratios and t tests, development of xenograft tumor. Where a palpable mass color display, and hierarchical clustering were performed. developed, the mass was harvested at a maximum of 1 cm in Captured gene expression signals were normalized to median external diameter by humanely killing the respective mouse. expression level across all chips in this experiment. Genes The harvested mass was fixed in 4% formalin, processed into with signals designated as absent in more than 50% of the paraffin sections, examined to confirm the presence of tumor samples were filtered out. The normalized gene expression xenograft, and analyzed for its morphologic characteristics. If signals were then log-transformed to the base 2, and log ratios there was no obvious tumor development at the end of 12 comparing different sample groups could be performed ac- ϭ months, the mouse was killed and dissected to look for tumor cording to the following formula: log ratio log2 (mean involvement in internal organs by gross and microscopic HKULC/mean NHBEC), where mean HKULC was the examination. If there was no tumor development with inoc- mean of signals of all four (HKULC 1–4) cell lines, and ulation content of 2 ϫ 105 tumor cells, a gradation of inocula mean NHBEC was the mean of signals of four normal containing 2 ϫ 106,2ϫ 107, and 2 ϫ 108 tumor cells were human bronchial epithelial cell lines (including NHBE attempted on three separate mice for each inoculum dose. [Clonetics], SAEC [Clonetics], BEAS-2B cell lines [ATCC CRL-9609], and HBEC1-KT12). A log ratio of Ն2.00 indicated increased expression of at least fourfold of mean Morphologic Characteristics and ՅϪ Immunohistochemical (IHC) Studies HKULC compared to mean NHBEC. A log ratio of 2.00 indicated reduced expression of at least fourfold in mean Harvested tumor cells of each cell line were centri- HKULC compared to mean NHBEC. fuged, fixed, and paraffin-embedded into a cell block accord- To compare the gene expression levels between ing to routine cytology procedures. A piece of tumor (DL31) HKULC 4 and DL31, log ratios of gene expression signals of adjacent to that used for cell culture establishment was also HKULC 4 compared to those of DL31 were obtained. The sampled from patient 4 for comparison with an HKULC 4 Pearson correlation coefficient of the gene expression levels xenograft. The cultured cells, xenografts, and the resected of HKULC and DL31 was also calculated. Unsupervised tumor were characterized for their morphologic characteris- hierarchical clustering was performed on the four HKULC tics by routine hematoxylin and eosin stain, mucin content, cell lines, DL31, and the four NHBEC lines to look for and their phenotypic patterns by IHC using a panel of similarity of gene expression profiles of the samples. monoclonal antibodies against cytokeratins (CK7 [DAKO, Class prediction with support vector machine with M7018], CK20 [DAKO, M7019]), the thyroid transcription leave-one-out analysis was performed using the BRB Array- factor TTF-1 (DAKO, M3575), and the proliferation marker Tools version 3.4.0 developed by Richard Simon and Amy MIB-1 (DAKO, M7240). Peng Lam from the National Cancer Institute (NCI),13 to identify a gene list that distinguishes cancer from normal Gene Expression Profiling Analysis with bronchial epithelial cells. To validate the prediction accuracy Affymetrix Genechip U133A of the identified gene list, Affymetrix GeneChip U133A data For the cultured cancer cells, at least 2 ϫ 107 cells were of a panel of 13 NCI adenocarcinoma cell lines (NCI H1437, harvested and total RNA was extracted with the RNeasy H1648, H1650, H1819, H1975, H1993, H2009, H2087, Miniprep (Qiagen) protocol. For the tumor block of DL31, H2126, H2347, HCC461, HCC515, HCC1833; data available normal tissue was trimmed away, resulting in Ͼ90% tumor from J.D.M.) were used as a separate test set in the analysis. by area. Fifty frozen sections (10 ␮m/section) were used for We first assigned available GeneChip data into two sets with total RNA extraction using the RNeasy Miniprep (Qiagen) set A including GeneChip data from the four HKULC cell protocol. The quality of the total RNA was checked with lines and the four normal human bronchial epithelial cell lines denaturing formamide gel electrophoresis, which showed two (eight samples), and set B with NCI adenocarcinoma cell sharp and distinct bands of 18S and 28S. Quality check was lines and the same four normal human bronchial epithelial also done by the Agilent Bioanalyzer with graphic analysis cell lines (17 samples). We first used set A as a training set to showing two distinct peaks of 18S and 28S without additional identify a group of classifier genes that distinguished between peaks of degradation. The total RNA was then hybridized HKULC cell lines and normal bronchial epithelial cell lines. onto Affymetrix GeneChip HG-U133A according to standard This group of genes was then used for validation with set B protocols.11 as the test set. The whole process was repeated, reversing set Absolute signals from individual chips were captured B as the initial training set and set A as the test set, until a list and processed with the standard MicroArray Suite 5.0 soft- of classifier genes that yielded 100% sensitivity and specific- ware (Affymetrix). Captured signals were further scaled and ity in either comparison were obtained.

FIGURE 1. Photomicrographs of the four newly established HKULC lung adenocarcinoma cell lines at the 70th passage. a: HKULC 1 showing piling up of tumor cells growing in the center. b: HKULC 2 showing monolayers of tumor cell growth. c: HKULC3 showing tumor cells clustering together without forming a confluent layer of tumor cells. d: HKULC 4 showing a confluent layer of tumor cell growth with some layering of cells at the center.

EGFR Mutation at Exons 18–21 Total genomic DNA was extracted from the respective cell lines according to standard protocol. Exons 18–21 of the EGFR gene were analyzed for mutations according to our previous protocol.14 In Situ Hybridization (ISH) for Detection of EGFR Overexpression EGFR gene copy number was analyzed by chromo- genic ISH, with digoxigenin-labeled EGFR and chromosome 7 DNA probes (SPoT-Light probes, Zymed) targeting the whole EGFR gene and chromosome 7 centromere, respec- FIGURE 2. Growth curves of the four newly established tively, on tumor cell blocks. At least 100 nonoverlapping HKULC lung adenocarcinoma cell lines at their 70th passage. tumor cells were scored. The number of signals for chromosome 7 control and EGFR probes, respectively, were scored for each counted tumor cell. formation of cell clusters in these cell lines before spreading to form confluent tumor cell layers. HKULC 1, 2, and 3 RESULTS showed similar doubling times, at the linear growth phase, of Cell Culture Characteristics about 25 hours, whereas HKULC 4 showed a doubling time of 33.5 hours. All cell lines were maintained in continuous culture for at least 70 passages. The four cell lines had different morphologic features. HKULC 1 formed tightly packed attached Test for Tumorigenicity in Nude Mice cells of medium size that tended to pile up before reaching HKULC 1 and 4 formed tumor xenografts of 1 cm in confluency (Fig. 1a), HKULC 2 (Fig. 1b) and HKULC 4 (Fig. about 6 to 8 weeks (Table 1). The parental tissue for HKULC 1d) formed confluent and mainly monolayer sheets of at- 4 was also transplantable and developed into xenograft. The tached cells of medium size. HKULC 3 formed colonies and xenograft obtained from these cell inoculations and the clusters of attached cells, with marked variation of cell size HKULC 4 parental tissue could further develop into tumors and frequent multiple or multilobated nuclei (Fig. 1c). after retransplantation into separate nude mice. The time for xenograft establishment was 3 to 4 weeks after inoculation Growth Kinetic Study (when a subcutaneous bulge was visible) and by 6 to 8 weeks HKULC 2 and 4 showed a short lag phase of 1 to 2 after inoculation, the xenograft would attain the external days. HKULC 2 showed the fastest growth rate, whereas dimension of 1 cm for xenograft harvesting. HKULC 4 grew more steadily (Fig. 2). Both HKULC 1 and HKULC 2 and 3 failed to produce xenograft 12 months 3 showed an initial lag phase of 4 to 5 days followed by an after tumor cell inoculation with at least three separate at- accelerated phase (Fig. 2). This feature might be related to the tempts for each dose of tumor cell inoculation.

Morphologic and Differentiation lines showed 30% to 50% proliferating cells as indicated by Characteristics of Tumor Cells and Xenografts the MIB1 proliferation marker (Fig. 3f). HKULC 1 showed fairly uniform population of medi- Mutation Status and In Situ Hybridization um-sized cells containing single nuclei. Its xenograft showed Analysis (ISH) of EGFR high-grade tumor cells arranged in solid clusters with cribri- HKULC 3 showed deletion of amino acids 756–760 of form formations (Fig. 3a). Intracytoplasmic mucin globules exon 19 (Table 1). No mutation was found in HKULC 1, 2, were present in occasional cells. HKULC 4 showed mild and 4 at exons 18–21. ISH showed that HKULC 1 and 4 were variation in cell size with occasional multinucleate forms. near disomy, whereas HKULC 2 and 3 were near tetrasomy The xenograft consisted of high-grade tumor cells arranged in for chromosome 7 (Fig. 4; Table 2). A slight increase in compact solid clusters with scattered areas showing acinar EGFR to chromosome 7 copy number resulting in a mean differentiation (Fig. 3b). Compared to the xenograft, the ratio of 1.11–1.31 was observed in the cell lines (Table 2). resected parent tumor showed more frequent acinar differentiation, but solid areas displaying large clusters of tumor cells Expression Profiling with Affymetrix Genechip were also present (Fig. 3c). Mild inflammatory cell infiltra- When compared to normal bronchial epithelial cells, tion was present, most prominent around areas of necrosis. genes that showed either (1) a higher or lower expression by No intracytoplasmic mucin formation was observed in the fourfold (log ratio Ն2.00 or ՅϪ2.00) and statistically sig- parent or transplanted tumor. HKULC 2 consisted of pleo- nificant difference of mean expression level across all four morphic tumor cells with eccentric nuclei and no mucin HKULC cell lines at p Յ 0.05 by t tests or (2) class-predictive formation. The cancer cells of HKULC 3 showed marked significance (adenocarcinoma cell lines, either the HKULC variation in cell size with frequent large bizarre multinucleate group or the NCI adenocarcinoma group versus normal bron- forms and occasional fine granules of intracytoplasmic mu- chial epithelial cell lines) by support vector machine with cin. For tumor cell markers, all the cell lines and xenografts leave-one-out analysis, were identified (detailed in supple- were CK7 positive (Fig. 3d) and CK20 negative. HKULC 2, mentary document sheet 1 available online). By criterion 1, 3, and 4 showed TTF1 expression in at least 10% of tumor there were 84 genes that showed fourfold altered expression cells (Fig. 3e), whereas HKULC 1 was negative. All the cell levels (supplementary document sheet 2). Among them, 61

FIGURE 3. Photomicrographs of xenografts and cell blocks derived from the four HKULC cell lines. a: HKULC 1 xenograft (hematoxylin and eosin staining) showing solid tumor clusters and gland forms. b: HKULC 4 xenograft (hematoxylin and eosin staining) showing solid tumor clusters separated by narrow connective tissue septa. c: Parent tissue of HKULC 4 (hematoxylin and eosin staining) showing solid tumor clusters and high tumor-to-stroma ratio. d: HKULC 3 cell block, positive immunohistochemical staining for CK7. e: HKULC 4 xenograft positive TTF1 staining in about 50% of tumor cells. f: HKULC 2 cell block, positive MIB1 staining in Ͼ50% of tumor cells.

FIGURE 4. In situ hybridization for evaluation of chromosome 7 and endothelial growth factor receptor (EGFR) gene copy number in HKULC 2 cell line. a: Chromosome 7 probe showing three to four signal copies per cell. b: EGFR probe showing four to five signal copies per cell.

TABLE 2. The Average Copy Number of Endothelial Growth Factor Receptor (EGFR) Gene Compared to Chromosome 7 for the Four HKULC Cell Lines chr 7 Signal/Cell EGFR Signal/Cell EGFR Signal/Cell

HKUL C1 1.95 2.55 1.31 FIGURE 5. Unsupervised hierarchical clustering of the HKUL C2 3.93 4.63 1.18 HKULC cell lines, DL31, and the normal bronchial human HKUL C3 4.31 4.79 1.11 epithelial cell lines. Note that HKULC 4 and DL31 cluster to- HKUL C4 2.19 2.73 1.25 gether. NHBE (Clonetics), normal human bronchial epithelial cells; SAEC (Clonetics), smallairway epithelial cells; BEAS-2B (ATCC CRL-9609), a SV40 T antigen-immortalized bronchial epithelial cell line; HBEC 1-KT, a Cdk4/hTERT-immortalized bronchial epithelial cell line.12 demonstrated statistical significance of p Յ 0.05 by t tests (1610 genes showed statistical significance at this level by t tests). By criterion 2, 17 classifier genes were found that were When the gene expression data of HKULC4 and its able to predict the class phenotype of both HKULC cell lines parental tissue (DL31) were compared, the Pearson coefficient and NCI adenocarcinoma cell lines, respectively, from nor- of correlation of the gene expression data from the two samples mal bronchial epithelial cell lines with 100% sensitivity and was r ϭ 0.92. Unsupervised hierarchical clustering analysis of specificity. Seven genes listed by criteria 1 and 2 overlapped the four cancer cell lines, four normal bronchial epithelial cells, with each other and a final list of 71 genes (Table 3) that fit and DL31 showed that HKULC 4 and DL31 clustered together, either criteria 1 or 2 or both was obtained. supporting a high correlation of their expression profiles (Fig. 5).

A list of 202 genes differing by fourfold or more in either were compared to the four normal human bronchial epithelial direction was found (supplementary document sheet 3). More cell lines and 13 NCI adenocarcinoma cell lines to look for than 50% of these genes were adhesion molecules, immuno- any candidates of differentially expressed genes. globulin or immunoglobulin receptor genes, cell surface binding Among the 71 genes listed (Table 3), some were proteins, chemokine receptor genes, collagen, and major histo- putative tumor suppressor genes showing four- to eightfold compatibility molecules. lower expression levels, namely, DKK3 (log ratio of individual HKULC cell lines: Ϫ2.43 to Ϫ4.25), SERPINF1 (log ratio: Ϫ3.78), CDH11 (log ratio: Ϫ3.76), DSC3 (log ratio: DISCUSSION Ϫ2.73 to Ϫ3.33), and Kruppel-like factor 6 (KLF6) (log The four lung adenocarcinoma cell lines (HKULC 1–4) ratio: Ϫ2.28 to Ϫ2.75), in HKULC cell lines compared to were all newly established from patients with primary lung normal human bronchial epithelial cell lines. DKK3 has been adenocarcinomas. They displayed a variable degree of nu- reported to demonstrate tumor-suppressing activities in a clear pleomorphism, with HKULC 3 showing admixtures of variety of tumors including tumors of the lung, kidney, large tumor cells with multinucleation (Fig. 1c). HKULC 2 urinary bladder, pancreas, and prostate and in different tumor and 4 had a shorter initial lag phase and formed confluent cell lines.15 DKK3 has been found to be down-regulated in tumor cell layers, whereas HKULC 1 and 3 displayed tumor lung cancer cell lines by promoter hypermethylation, and cell aggregation (Fig. 1a and c). The tendency for tumor cells treatment of these cell lines with 5-aza-cytidine restored its to form clusters was apparently not associated with transplant- tumor-suppressing ability.16 SERPINF1, also known as pig- ability, as only HKULC 1 but not HKULC 3 formed tumor ment epithelium-derived factor, was also thought to possess xenografts in nude mice (Table 1). Although only four cancer tumor-suppressing properties via its antiangiogenic effects. cell lines were available for analysis, comparison of expression HeLa cells bearing the wild-type allele show tumor-suppress- profiles between cell lines with differences in transplantabil- ing abilities, whereas HeLa cells bearing the mutant allele ity and tendency of aggregation did not yield genes that demonstrate loss of growth inhibition.17 Transfection of showed significant differential expression (data not shown). SERPINF1 into colonic adenocarcinoma cell lines restored EGFR mutations have been found to be more common tumor-suppressing properties with reduced tumor growth and in lung adenocarcinomas in nonsmoking women, especially tumor microvessel densities18 in the pulmonary metastatic well-differentiated tumors with bronchioloalveolar differen- tumor model. The expression of SERPINF1 at mRNA and tiation. Among the four cancer cell lines, EGFR exon 19 protein levels had also been found to be reduced in lung tumors, deletion was found only in HKULC 3 (Table 1), which was and this reduction in the level of expression was correlated derived from a female nonsmoker whose tumor cells ex- 19 pressed the TTF1 pneumocyte differentiation marker. The with tumor microvessel density and poor survival. other tumor cells, including HKULC 1 from a female non- Two members of the cadherin superfamily were found smoker but which did not show TTF1 expression, and in this list of 71 genes. CDH11 has been proposed as a tumor 20 HKULC 2 and 4, which were from current or previous suppressor gene in retinoblastoma. Its expression in pulmo- smokers, did not show EGFR mutations. ISH showed near nary metastasis model of osteosarcoma cell lines restored 21 tetrasomy for chromosome 7 in HKULC 2 and 3 and near tumor-suppressing function. Desmocollin (DSC3) is also a disomy for HKULC 1 and 4. Amplification of EGFR was not member of the cadherin molecules. DSC3 had been reported observed, with all cell lines showing not more than 1.31 copies to show loss of expression in breast tumor cells via promoter of EGFR to every chromosome 7 control probe (Table 2). hypermethylation.22 KLF6 had been shown to exert its tumor- Unsupervised hierarchical clustering of the HKULC suppressing activities via inhibition of c-Jun proto-oncogene– and normal bronchial epithelial cell lines showed that the mediated cellular proliferation.23 The expression of KLF6 had expression profile of HKULC 3 clusters separately from been found to be reduced in prostate24 and colorectal cancer the other cell lines. Several factors could have accounted for cells.25 All these genes, namely, DKK3, SERPINF1, CDH11, the separation of HKULC 3 from the other cell lines, such as DSC3, and KLF6, are potential tumor suppressor genes that EGFR exon 19 deletion, previous chemotherapy, and the were found to be down-regulated in the newly established presence of numerous large, pleomorphic, and multinucleated HKULC cell lines 1–4 compared to normal human bronchial tumor cells. To evaluate the effects of EGFR mutation on the epithelial cells. The new cancer cell lines could act as suitable clustering pattern, the microarray data set included EGFR handles for further studying the role of these putative tumor mutant-bearing cell lines, such as H1650 and H1975, for suppressor genes in lung cancer. hierarchical clustering analysis. The results showed that Some of the genes could be related to elements in the HKULC 3 and these EGFR mutant cell lines did not cluster EGFR signaling pathway, including ERBB3, MUC1, and VAV1. together, indicating that EGFR mutation alone was not the The EGFR gene and chemotherapeutic agents targeting ele- major factor determining the clustering pattern (result not ments in its downstream signaling pathway are under intensive shown). investigations as possible therapeutic interventions for manage- In this study, the gene expression profiles were document of lung cancer.26 The EGFR gene family includes EGFR mented as part of the characterization of these cell lines so (ERBB1 or HER1), HER2 (EGFR2 or ERBB2/NEU), HER3 that the information would be available as reference for future (EGFR3 or ERBB3), and HER4 (EGFR4 or ERBB4).27 Each of research studies using these cell lines. As a preliminary them displays distinct characteristics with ERBB2 showing in- analysis, the expression profiles of these HKULC cell lines tense kinase activity, but ERBB3 showing no kinase activity at

all.27 Studies have shown that the female gender, nonsmoker to show synergistic action with EGFR and demonstrated status, Asian ethnic origin, and carrying activating EGFR mu- proto-oncogenic properties by promoting pancreatic cancer tation are predictors of good clinical response to EGFR tyrosine cell proliferation.42 kinase inhibitors.28,29 Like EGFR mutation, ERBB2 mutations ERBB3, MUC1, and VAV1 were the discriminatory are found to be associated with Asian origin, female gender, and genes common to all four HKULC cell lines. Thus, apart nonsmoker status,30 and increased copy number of HER2 has from mutation of EGFR, these could interact with and assist been reported to be associated with response to EGFR tyrosine in the activation of the EGFR pathway. kinase inhibitor.31 ERBB2 overexpression has been found to From the list of genes in Table 3, other genes that could increase the sensitivity of lung cancer cells to gefitinib via represent potential targets for manipulation in these HKULC heterodimerization to form the ERBB2/ERBB3 complex.32 Rel- cell lines include genes that are involved in control of cell atively fewer studies focus on ERBB3 itself in lung cancer. One cycle and proliferation, including CDKN1A (log ratio: Ϫ2.28 recent study reports on ERBB3 overexpression and association to Ϫ3.94, indicating four- to eightfold lower levels of expres- with response to gefitinib in lung cancer patients.33 ERBB3 could sion in HKULC cell lines compared to normal human bron- thus be one potential target for therapeutic intervention in lung chial epithelial cell lines) and CDKN2A (log ratio: Ϫ3.56, cancer. Among the ERBB gene family, ERBB3 showed higher i.e., more than eightfold lower levels of expression in level of expression of at least threefold (log ratio of individual HKULC cell lines compared to mean NHBEC). CDKN1A has HKULC cell line: 1.84–3.36) in all the four HKULC cell lines been found to interact with MUC1.43 Suppressed expression when compared to normal human bronchial epithelial cell lines of CDKN2A by promoter hypermethylation has been well (Table 3). The HKULC cell lines could be suitable materials for reported in lung cancer.44,45 The low expression of CDKN1A in vitro drug sensitivity tests, for example, in testing the effects and CDKN2A may reflect their roles in the control of cell of anti-HER3 antibody on the growth of lung cancer cells. cycles and proliferation in these HKULC cell lines. Other Expression levels of ERBB1, ERBB2, and ERBB4 in all the four targets with potential roles in lung cancer development requiring HKULC cell lines were not more than twofold altered (higher or further investigation include EEF1A2 and MTSS1. EEF1A2 is a lower) compared to the NHBEC cell lines, except that HKULC putative oncogene in ovarian cancer46 and is also highly ex- 3 showed about a threefold higher level of ERBB2 expression pressed in all the HKULC cell lines. The expression of the (log ratio of HKULC 3: 1.27, HKULC 1: Ϫ0.41, HKULC 2: MTSS1 gene has been reported to inhibit metastasis in bladder 0.57, HKULC 4: 0.71) than NHBEC cell lines. HKULC 3 was cancer47 and MTSS1 gene showed low expression in the also the cell line bearing EGFR gene deletion at exon 19 (Table HKULC cell lines. 1). The observation that the HKULC 3 cell line showed high One of the theoretical concerns regarding the use of expression levels of both ERBB2 and ERBB3 could imply, as cultured cells as research material is that they could acquire suggested in a previous study of breast cancer, that ERBB2/ genetic changes during in vitro propagation so that they no ERBB3 could form oncogenic functional heterodimers to pro- longer represent the original tissues from which they were mote tumor proliferation.34 derived. In this study, we have demonstrated a close histo- MUC1 overexpression has also been reported to be logic resemblance of HKULC 4 xenograft tumor and its associated with ERBB1, ERBB2, ERBB3, and ERBB4 expres- parent tumor tissue (Fig. 3). We have also shown a high sion.35 It has been shown that MUC1 interacts with various correlation coefficient (0.92) between the expression profile ERBB proteins to potentiate signal transduction along the of HKULC 4 and its parent tumor DL31. Most of the genes EGFR pathway by activation of mitogenic mitogen-activated that showed at least fourfold difference were immunoglobulin protein kinases.35 MUC1 is also highly expressed in the four genes, adhesion molecules, or genes found in the cellular HKULC cell lines (log ratio of HKULC versus mean NH- matrix. The differential expression of genes between a cell BEC: 3.54–5.36, i.e., eight- to 30-fold overexpression). line and its parent tissue could reflect the repertoire of genes Overexpression of MUC1 has been reported in various tumor involved in cell-matrix interaction in the in vivo environment. and cancer cell lines including various subtypes of lung The 92% correlation of gene expression levels between the adenocarcinomas36 and breast and colorectal cancer, and its HKULC 4 cell line and its parent tissue and their clustering expression has been correlated with malignant transformation together indicates a close resemblance in the biological ac- and tumor invasiveness.37,38 MUC1 encodes for a tumor- tivities engaged in the maintenance of in vitro propagation associated antigen that is currently under evaluation as a and validates the use of cancer cell cultures as surrogate target for tumor vaccine development in non-small cell lung models of cancer growth. Similar observations have been cancer.39,40 The four HKULC cell lines with MUC1 overex- made in other cancer types. When breast cancer cell lines pression could be useful for in vitro testing of MUC1 as a were compared to their parental tissues, a satisfactory repre- potential therapeutic target. sentation of the parental tumor characteristics in terms of The VAV1 gene was found to be highly expressed in the gene mutations and expression was reported.48 HKULC 3 and 4 cell lines. VAV1 is a proto-oncogene mainly In summary, we have established four new primary expressed in hemopoietic cells and the VAV1 protein belongs lung adenocarcinoma cell lines from Chinese patients, includ- to a family of VAV proteins that could bind to various cell ing female nonsmokers, with different clinical characteristics, surface receptors including EGFR, resulting in EGFR activa- culture, and expression profiles. These cell lines could serve tion and promotion of cellular growth and proliferation as useful tools for further functional studies and drug sensi- through downstream signaling.41 VAV1 has also been found tivity testing in translational research in lung cancer.

REFERENCES 26. Isobe T, Herbst RS, Onn A. Current management of advanced non-small 1. Hospital Authority (Hong Kong China). Hong Kong Cancer Stat 2003. cell lung cancer: targeted therapy. Semin Oncol 2005;32:315–328. Hong Kong: Hong Kong Cancer Registry Hospital Authority, 2003:v. 27. Shigematsu H, Gazdar AF. Somatic mutations of epidermal growth 2. Parkin DM. International Agency for Research on C, International factor receptor signaling pathway in lung cancers. Int J Cancer 2006; Association of Cancer R. Cancer Incidence in Five Continents, Volume 118:257–262. VIII. Lyon, France: International Agency for Research on Cancer, 2002. 28. Herbst RS, Kies MS. Gefitinib: current and future status in cancer 3. Brownson RC, Chang JC, Davis JR. Gender and histologic type variations therapy. Clin Adv Hematol Oncol 2003;1:466–472. in smoking-related risk of lung cancer. Epidemiology 1992;3:61–64. 29. Tsao MS, Sakurada A, Cutz JC, et al. Erlotinib in lung cancer— 4. Zang EA, Wynder EL. Differences in lung cancer risk between men and molecular and clinical predictors of outcome. N Engl J Med 2005;353: women: examination of the evidence. J Natl Cancer Inst 1996;88:183–192. 133–144. 5. Sugaya M, Takenoyama M, Osaki T, et al. Establishment of 15 cancer 30. Shigematsu H, Takahashi T, Nomura M, et al. Somatic mutations of the cell lines from patients with lung cancer and the potential tools for HER2 kinase domain in lung adenocarcinomas. Cancer Res 2005;65: 1642–1646. immunotherapy. Chest 2002;122:282–288. 31. Cappuzzo F, Varella-Garcia M, Shigematsu H, et al. Increased HER2 6. Oie HK, Russell EK, Carney DN, et al. Cell culture methods for the gene copy number is associated with response to gefitinib therapy in establishment of the NCI series of lung cancer cell lines. J Cell Biochem epidermal growth factor receptor-positive non-small-cell lung cancer Suppl 1996;24:24–31. patients. J Clin Oncol 2005;23:5007–5018. 7. Tai AL, Fang Y, Sham JS, et al. Establishment and characterization of 32. Hirata A, Hosoi F, Miyagawa M, et al. HER2 overexpression increases a human non-small cell lung cancer cell line. Oncol Rep 2005;13:1029– sensitivity to gefitinib, an epidermal growth factor receptor tyrosine 1032. kinase inhibitor, through inhibition of HER2/HER3 heterodimer forma- 8. Chikamori M, Takigawa N, Kiura K, et al. Establishment of a 7-ethyl- tion in lung cancer cells. Cancer Res 2005;65:4253–4260. 10-hydroxy-camptothecin-resistant small cell lung cancer cell line. An- 33. Cappuzzo F, Toschi L, Domenichini I, et al. HER3 genomic gain and ticancer Res 2004;24:3911–3916. sensitivity to gefitinib in advanced non-small-cell lung cancer patients. 9. Koizumi F, Shimoyama T, Taguchi F, et al. Establishment of a human Br J Cancer 2005;93:1334–1340. non-small cell lung cancer cell line resistant to gefitinib. Int J Cancer 34. Holbro T, Beerli RR, Maurer F, et al. The ErbB2/ErbB3 heterodimer 2005;116:36–44. functions as an oncogenic unit: ErbB2 requires ErbB3 to drive breast 10. Sharifzadeh S, Owji SM, Pezeshki AM, et al. Establishment and char- tumor cell proliferation. Proc Natl Acad SciUSA2003;100:8933–8938. acterization of a human large cell lung cancer cell line with neuroendo- 35. Schroeder JA, Thompson MC, Gardner MM, et al. Transgenic MUC1 crine differentiation. Pathol Oncol Res 2004;10:225–230. interacts with epidermal growth factor receptor and correlates with 11. Affymetrix. Microarray Suite User Guide 5.0. mitogen-activated protein kinase activation in the mouse mammary 12. Ramirez RD, Sheridan S, Girard L, et al. Immortalization of human gland. J Biol Chem 2001;276:13057–13064. bronchial epithelial cells in the absence of viral oncoproteins. Cancer 36. Awaya H, Takeshima Y, Yamasaki M, et al. Expression of MUC1, Res 2004;64:9027–9034. MUC2, MUC5AC, and MUC6 in atypical adenomatous hyperplasia, 13. Simon R, Lam AP. BRB ArrayTools version 3.3.0, 2005. bronchioloalveolar carcinoma, adenocarcinoma with mixed subtypes, 14. Tam IY, Chung LP, Suen WS, et al. Distinct epidermal growth factor and mucinous bronchioloalveolar carcinoma of the lung. Am J Clin receptor and KRAS mutation patterns in non-small cell lung cancer Pathol 2004;121:644–653. patients with different tobacco exposure and clinicopathologic features. 37. Baldus SE, Monig SP, Huxel S, et al. MUC1 and nuclear beta-catenin Clin Cancer Res 2006;12:1647–1653. are coexpressed at the invasion front of colorectal carcinomas and are 15. Hsieh SY, Hsieh PS, Chiu CT, et al. Dickkopf-3/REIC functions as a both correlated with tumor prognosis. Clin Cancer Res 2004;10:2790– suppressor gene of tumor growth. Oncogene 2004;23:9183–9189. 2796. 16. Kobayashi K, Ouchida M, Tsuji T, et al. Reduced expression of the 38. Canbay E. Erb-B2 homodimerization inhibits MUC1 transcription in REIC/Dkk-3 gene by promoter-hypermethylation in human tumor cells. cultured human mammary epithelial cells. Cell Biol Int 2003;27:477– Gene 2002;282:151–158. 481. 17. Hosomichi J, Yasui N, Koide T, et al. Involvement of the collagen 39. Liu M, Acres B, Balloul JM, et al. Gene-based vaccines and immuno- I-binding motif in the anti-angiogenic activity of pigment epithelium- therapeutics. Proc Natl Acad SciUSA2004;101 Suppl 2:14567–14571. derived factor. Biochem Biophys Res Commun 2005;335:756–761. 40. O’Mahony D, Kummar S, Gutierrez ME. Non-small-cell lung cancer 18. Mahtabifard A, Merritt RE, Yamada RE, et al. In vivo gene transfer of vaccine therapy: a concise review. J Clin Oncol 2005;23:9022–9028. pigment epithelium-derived factor inhibits tumor growth in syngeneic 41. Moores SL, Selfors LM, Fredericks J, et al. Vav family proteins couple murine models of thoracic malignancies. J Thorac Cardiovasc Surg to diverse cell surface receptors. Mol Cell Biol 2000;20:6364–6373. 2003;126:28–38. 42. Fernandez-Zapico ME, Gonzalez-Paz NC, Weiss E, et al. Ectopic 19. Zhang L, Chen J, Ke Y, et al. Expression of pigment epithelial derived expression of VAV1 reveals an unexpected role in pancreatic cancer factor is reduced in non-small cell lung cancer and is linked to clinical tumorigenesis. Cancer Cell 2005;7:39–49. outcome. Int J Mol Med 2006;17:937–944. 43. Wei X, Xu H, Kufe D. Human MUC1 oncoprotein regulates p53- 20. Marchong MN, Chen D, Corson TW, et al. Minimal 16q genomic loss responsive gene transcription in the genotoxic stress response. Cancer implicates cadherin-11 in retinoblastoma. Mol Cancer Res 2004;2:495– Cell 2005;7:167–178. 503. 44. Jarmalaite S, Kannio A, Anttila S, et al. Aberrant p16 promoter meth- 21. Kashima T, Nakamura K, Kawaguchi J, et al. Overexpression of cad- ylation in smokers and former smokers with nonsmall cell lung cancer. herins suppresses pulmonary metastasis of osteosarcoma in vivo. Int J Int J Cancer 2003;106:913–918. Cancer 2003;104:147–154. 45. Shimamoto T, Ohyashiki JH, Hirano T, et al. Hypermethylation of 22. Oshiro MM, Kim CJ, Wozniak RJ, et al. Epigenetic silencing of DSC3 E-cadherin gene is frequent and independent of p16INK4A methylation is a common event in human breast cancer. Breast Cancer Res 2005;7: in non-small cell lung cancer: potential prognostic implication. Oncol R669–680. Rep 2004;12:389–395. 23. Slavin DA, Koritschoner NP, Prieto CC, et al. A new role for the 46. Anand N, Murthy S, Amann G, et al. Protein elongation factor Kruppel-like transcription factor KLF6 as an inhibitor of c-Jun proto- EEF1A2 is a putative oncogene in ovarian cancer. Nat Genet 2002; oncoprotein function. Oncogene 2004;23:8196–8205. 31:301–305. 24. Narla G, Heath KE, Reeves HL, et al. KLF6, a candidate tumor suppressor 47. Lee YG, Macoska JA, Korenchuk S, et al. MIM, a potential metastasis gene mutated in prostate cancer. Science 2001;294:2563–2566. suppressor gene in bladder cancer. Neoplasia 2002;4:291–294. 25. Reeves HL, Narla G, Ogunbiyi O, et al. Kruppel-like factor 6 (KLF6) is 48. Wistuba II, Behrens C, Milchgrub S, et al. Comparison of features of a tumor-suppressor gene frequently inactivated in colorectal cancer. human breast cancer cell lines and their corresponding tumors. Clin Gastroenterology 2004;126:1090–1103. Cancer Res 1998;4:2931–2938.

TABLE 3. List of 71 Genes Identified in Comparison between the HKULC Cell Lines and Normal Bronchial Epithelial Cell Lines

(Continued)

TABLE 3. (Continued)

HKULC1, HKULC2, HKULC3, HKULC4 vs Mean NHBEC signals: values represent the log ratios of gene expression signals from individual HKULC cell lines compared with mean NHBEC signals. Positive values (higher expression in HKULC cell lines compared with NHBEC cell lines) Ն2.00 are highlighted in red. Negative values (lower expression in HKULC cell lines compared with NHBEC cell lines) ՅϪ2.00 are highlighted in green. T-test Mean HKULC signals vs Mean NHBEC Signals: listing of the p value of student t tests of the respective genes when comparing all 4 HKULC cell lines with the 4 NHBEC cell lines, with two-tails and unequal variance assumed. Class Predictors: those genes found to be class predictors with respect to either HKULC cell lines against NHBEC cell lines or NCI adenocarcinoma cell lines against NHBEC cells are labeled with ‘P’. Color Display: the gene expression levels are represented with a white to blue scale so that the lighter colors represent low expression and deeper colors represent high expression. NHBEC: Normal Human Bronchial Epithelial cell line.