A Novel Molecular Pathway for Snail-Dependent, SPARC-Mediated Invasion in Non–Small Cell Lung Cancer Pathogenesis

Lung cancer is the leading cause of cancer-related deaths worldwide, with a median survival of 8 months after diagnosis and only 16% of patients surviving more than 5 years (1). Understanding early lung cancer pathogenesis will facilitate targeted approaches for chemoprevention. Inflammatory mediators, such as TGF-β, eicosanoids, and interleukin-1β are overexpressed in the pulmonary microenvironment of smokers and patients with emphysema or pulmonary fibrosis; these patients have a heightened risk of developing lung cancer (2, 3). Recently described as one of the hallmarks of cancer, chronic inflammation is now considered a risk factor for the development of lung cancer (4). Although an active area of investigation, the molecular mechanisms underlying the association between inflammation and lung cancer initiation and progression remain largely undefined.

The zinc-finger transcription factor Snail, encoded by the SNAI1 gene, has been shown to be upregulated following exposure to inflammatory mediators such as prostaglandin E2 (5) and TGF-β (6). Snail exerts global effects on epithelial cell gene expression profiles, including regulation of EMT (7, 8). Snail plays a pivotal role in inducing EMT in several malignancies (9–11) and is expressed in a stem cell–like subpopulation within immortalized human mammary epithelial cells that are capable of transformation (12). Recent studies suggest that Snail may play a broader role in carcinogenesis (13, 14). We have shown that Snail is upregulated in NSCLC tissues, is associated with poor prognosis, and promotes NSCLC tumor progression in vivo (15). Furthermore, Snail overexpression in NSCLC is associated with differential gene expression related to diverse aspects of lung cancer progression, including angiogenesis (15). Identification of the mechanisms by which the inflammation-induced transcriptional repressor Snail contributes to lung cancer pathogenesis, specifically to the invasive phenotype, would be a step forward in understanding the contribution of inflammation to lung cancer development and progression.

SPARC, also known as osteonectin, is an extracellular matrix glycoprotein first identified as a major noncollagenous component of bovine bone (16). Its expression modulates reversible interactions between cells and the extracellular matrix (17). Upregulation of SPARC is associated with metastatic potential of melanomas and gliomas, as well as an invasive phenotype in breast, prostate, and colorectal carcinomas (18). Expression of SPARC in the NSCLC stroma is associated with poor prognosis (19), although its role in lung tumor progression, especially in relation to epithelial cell Snail expression, has not been evaluated.

To understand the molecular changes that occur during NSCLC initiation and development, we overexpressed Snail in both immortalized HBECs and NSCLC cells; the HBECs were previously established as a robust model of the pulmonary airway epithelium and its associated malignant transformation (20–24). We demonstrate that Snail upregulates SPARC and drives SPARC-dependent invasion in both models of premalignancy and established NSCLC. We also examined the mechanism for Snail-induced SPARC expression and identified key signaling pathways critical to this relationship.

Lung cancer cell lines A549, H292, H358, H441, and H1437 were obtained from American Type Culture Collection. All HBEC lines were provided by Drs. J.D. Minna and J.W. Shay (University of Texas, Southwestern Medical Center, Dallas, TX). The cells were immortalized in the absence of viral oncoproteins via ectopic expression of human telomerase reverse transcriptase (hTERT) and cyclin-dependent kinase 4 under control of puromycin and geneticin, respectively (22). Four parental cell lines derived from four patients, HBEC2, HBEC3, HBEC4, and HBEC7, were used. A mutated HBEC3 cell line designated H3mutP53/KRAS (or H3mut) was also used. This cell line was derived by stably transfecting HBEC3 with an short hairpin RNA (shRNA) construct targeting the tumor suppressor P53 and a construct overexpressing the oncogenic protein KRAS with an activating mutation under control of zeocin and blasticidin, respectively. All cell lines were routinely tested for the presence of Mycoplasma using the MycoAlert Mycoplasma Detection Kit (Lonza). Cell lines were authenticated in the University of California at Los Angeles (UCLA) Genotyping and Sequencing Core utilizing Promega's DNA IQ System and Powerplex 1.2 System according to the manufacturer's instructions. All cells were used within 10 passages of genotyping. Lung cancer cell lines were grown in RPMI-1640 (Mediatech Inc.) supplemented with 10% FBS (Gemini Biological Products), 1% penicillin/streptomycin (Life Technologies), and 2 mmol/L glutamine (Life Technologies). HBEC lines were grown in keratinocyte serum-free media (Life Technologies) supplemented with 30 μg/mL bovine pituitary extract and 0.2 ng/mL recombinant EGF 1–53 (Life Technologies). Treatments were carried out in 6-well plates at a density of 1.25 × 10⁵ cells per well in serum-free medium, unless stated otherwise. The MEK/ERK inhibitor U0126 (Cell Signaling Technology) was prepared in sterile dimethyl sulfoxide at a concentration of 10 mg/mL, and cells were treated at a final concentration of 15 μg/mL for 24 hours. Recombinant human TGF-β1 (PeproTech) was prepared in sterile 0.1% bovine serum albumin (BSA) at a concentration of 5 ng/μL, and cells were treated at a final concentration of 5 ng/mL for 24 hours.

Cells were stably transduced as follows: wild-type Snail cDNA pcDNA3 (a gift from Dr. E. Fearon, University of Michigan, Ann Arbor, MI) was excised from the plasmid with HindIII and EcoRV and subcloned into the retroviral vector pLHCX (Clontech) that includes a drug resistance (hygromycin B) marker. All constructs were verified by restriction endonuclease digestion. For virus production, 70% confluent 293T cells were cotransfected with pLHCX–Snail or pLHCX (vector alone). The HBEC and NSCLC cells were then transduced with high titer supernatants producing either Snail or pLHCX virus. Following transduction, the cells were selected with hygromycin B (Life Technologies).

SPARC shRNA plasmids on the pLKO.1-Puro vector backbone and relevant controls were obtained from Sigma-Aldrich. Using the viral transduction method described above, NSCLC cell lines were stably transduced with one of five shRNA sequences: −shSPARC1, −shSPARC2, −shSPARC3, a nonsilencing sequence (−shNS), or the pLKO.1 vector backbone sequence (−shV). The cell lines were selected with optimized concentrations of puromycin (EMD Chemicals). For the HBEC lines, as the overexpression of hTERT was under puromycin selection, the puromycin selection marker in the shSPARC vectors was replaced with the inosine-5′-monophosphate dehydrogenase (IMPDH) gene, encoding resistance to mycophenolic acid. After viral transduction, the cells were selected with optimized concentrations of mycophenolic acid (Alfa Aesar).

Cells were plated in 6-well plates at 1.25 × 10⁵ cells per well and cultured for 24 hours in growth media. For transient overexpression of miR-29b, complete media were replaced before transfection with the miRVana miR-29b mimic (Life Technologies) or miRVana miRNA mimic negative control #1 (Life Technologies). For transient inhibition of TGF-β1, complete media were replaced with serum-free media overnight before transfection with target siRNA (Integrated DNA Technologies) or Silencer Negative Control #1 siRNA (Life Technologies). Transfections were carried out using the Lipofectamine RNAiMAX Transfection Reagent (Life Technologies) in serum-free media for 4 hours before replacement with fresh serum-free media and an additional 20-hour incubation.

Cells were washed with PBS and whole-cell lysates were collected over ice using lysis buffer prepared according to standard methods (25). Protein concentrations were measured with a bicinchoninic acid (BCA) protein assay reagent (Pierce). Proteins were resolved by 10% SDS–PAGE and analyzed by Western blot analysis using polyvinylidene difluoride membranes (Millipore) according to standard methods. Membranes were blocked with 5% nonfat dry milk or 5% BSA in TBS plus 0.05% Tween 20. The membranes were probed overnight at 4°C with anti-Snail, anti-SPARC, anti-TGFBR2, anti-ERK1/2, and anti-phospho-ERK1/2 (all from Cell Signaling Technology). Secondary antibodies, goat anti-mouse (Bio-Rad) and goat anti-rabbit (Santa Cruz Biotechnology), were incubated at room temperature for 60 minutes. Membranes were developed using Supersignal Chemiluminescence System (Pierce) or Western Lightning Plus-ECL (PerkinElmer) and exposed to X-ray film (Life Sciences Products, Inc.). Equal loading of samples was confirmed by probing the membranes with α-tubulin (Cell Signaling Technology). Antibody catalog numbers and blotting conditions are detailed in Supplementary Table S1. Each experiment was carried out at least three times, and one representative experiment or image is shown.

Secreted TGF-β1 levels were quantified using the eBioscience Human/Mouse TGF-β1 ELISA Ready-SET-Go! Kit (Cat #88-8350). Cells were plated in 6-well plates at a density of 1.25 × 10⁵ cells per well; triplicates of each condition were plated. Once adherent, cells were washed and media were replaced with 1 mL serum-free media. Media supernatants and cell lysates were collected after 24 hours. Supernatant TGF-β levels were evaluated following the manufacturer's instructions and normalized to cell lysate protein concentrations. Each experiment was carried out at least three times, and one representative experiment or image is shown.

Cells were serum-starved and plated at a density of 2 × 10⁴ (cancer cells) or 1 × 10⁴ (HBEC) cells per well in a Corning HTS Transwell-96 Permeable Support Plate (Sigma-Aldrich); six replicates of each condition were plated. Before plating, the transwells were coated with type I Rat Tail Collagen (BD Biosciences) to create a “membrane” for the cells to degrade and invade. Cells were allowed to invade for 48 hours into a lower chamber containing media with 20% FBS (cancer cells) or 2% FBS (HBEC) as a chemoattractant. The upper chamber was aspirated and washed with PBS to remove noninvasive cells. The lower chamber was washed with PBS, and invasive cells were released from the underside of the top chamber with Cell Dissociation Solution (Trevigen). Calcein AM (Life Technologies) was used to stain viable cells in the lower chamber only, and fluorescence was quantified and compared with measurements from day 0 control plates that contained the total number of cells plated for each cell line to generate “percentage of input invasion” values. Each experiment was carried out at least three times, and one representative experiment or image is shown.

RNA was isolated using the miRNeasy Mini Kit (Qiagen), and cDNA was prepared using the High Capacity RNA-to-cDNA Kit (Life Technologies) according to the manufacturers' protocols. Transcript levels of miR-29b and SPARC were measured by quantitative real-time reverse transcription PCR (qRT-PCR) using the TaqMan Probe-based Gene Expression system (Life Technologies) in a MyiQ Cycler (Bio-Rad) following the manufacturers' protocols. Amplification was carried out for 40 cycles of 15 seconds at 95°C and 60 seconds at 60°C. All samples were run in triplicate, and their relative expression levels were determined by normalizing the expression of each target to RNU6b (miR-29b) or GUSB (SPARC). These levels were then compared with the normalized expression levels in a reference sample using the |$2^{-\Delta \Delta C_t}$| method to calculate fold change (FC) values (26). Each experiment was carried out at least three times, and one representative experiment or image is shown.

Serial sections obtained from human NSCLC clinical specimens archived in the UCLA (Los Angeles, CA) Lung Cancer Specialized Programs of Research Excellence tissue bank were collected from patients under informed consent in accordance with the Declaration of Helsinki (Institutional Review Board #10-001096). Antigen retrieval was accomplished with sodium citrate 10 mmol/L (pH 6.0). Sections were blocked with 10% normal goat serum and then probed with an antibody against Snail (Cat #ab85931; Abcam) or SPARC (Cat #AON-5031; Haematologic Technologies) using a working dilution of 1:500 for tissue staining. Primary antibodies were incubated overnight at 4°C. After incubation with secondary antibody (Vector Laboratories), staining was developed using the DAB Substrate Kit for Peroxidase (Vector Laboratories). Snail and SPARC expression was evaluated by a pathologist (M.C. Fishbein) specializing in cardiopulmonary disease. Evaluation of the clinical specimens was based on epithelial cell staining intensity and correlation of staining between serial sections; stromal and immune cell staining patterns were also evaluated but were of secondary importance to this study. Photomicrographs were obtained using an Olympus BX50 microscope with Plan Apochromat objective lenses. An Olympus DP11 camera and Olympus Camedia software were used to produce the images.

Single samples of RNA were collected from Snail-overexpressing and vector control H292, H358, H441, and H1437 cell lines with the miRNeasy Mini Kit (Qiagen). One microgram of total RNA was labeled using the miRCURY LNA microRNA Array Power Labeling Kit by the UCLA Clinical Microarray Core. The labeled microRNAs (miRNA) were hybridized to Exiqon miRCURY LNA microRNA Array-6th Generation according to the manufacturer's instructions. This array includes 927/648/351 human/mouse/rat miRNAs, as well as 438 miRPlus miRNAs. The miRNA array slides were scanned using Axon GenePix 4000B scanner (Axon Instruments) and processed by using the GenePix Pro 6.0 software (Axon Instruments). The raw data were normalized by using a combination of housekeeping miRNA, spike-in miRNA, and invariant endogenous miRNAs. Harvard dChip software (27) was used for data analysis. Microarray data discussed in this publication are deposited in National Center for Biotechnology Information's Gene Expression Omnibus (accession #GSE48922; ref. 28).

Samples were plated and run in triplicate, unless otherwise indicated, and all experiments were carried at least three times. Results from one representative experiment or image are shown. Probability values were calculated using a two-tailed nonpaired Student t test for most count-based data, and ANOVA models were used where pairwise comparisons were made. Results were considered significant as follows: *, P ≤ 0.05; **, P ≤ 0.001; and ***, P ≤ 0.0001.

Recently, we found that forced overexpression of Snail in NSCLC cell lines leads to global expression changes, including increased expression of genes implicated in angiogenesis (15). Gene expression profiling and bioinformatic analysis indicated that a panel of Snail-overexpressing NSCLC cell lines has significantly increased expression of SPARC (15). FC values of more than 10 were observed in three of four cell lines tested (H358 FC = 32, H441 FC = 193, and H1437 FC = 1276). qRT-PCR and immunoblot analyses of NSCLC cell lines with and without Snail overexpression confirmed the relationship between Snail and SPARC in established cancer cell lines (Fig. 1A and B, respectively). To model early pathogenesis of lung cancer, the same Snail overexpression plasmid was introduced into HBECs. qRT-PCR and immunoblot analyses of these cell lines show that Snail overexpression leads to enhanced SPARC expression in this model as well (Fig. 1C and D). Taken together, these data suggest that SPARC expression is positively associated with Snail in models of both premalignancy and established lung cancer.

To confirm the relationship between Snail and SPARC in human NSCLC neoplasms in situ, we stained serial sections of paraffin-embedded human lung adenocarcinoma (ADC; n = 10) and squamous cell carcinoma (SCC; n = 9) for Snail and SPARC protein expression. Immunolocalization revealed a pattern of uniform cytoplasmic SPARC staining of the epithelial component of the neoplasm in association with intense nuclear Snail staining of the epithelia (Fig. 1E and Table 1). The immunostained sections were reviewed by a pathologist (M.C. Fishbein) and scored on the basis of Snail or SPARC staining intensity (low, medium, or high) and the Snail:SPARC correlation (yes or no; ADC = 9/10 yes, SCC = 7/9 yes). In regions of more poorly differentiated and infiltrative tumor, there tended to be increased SPARC expression, whereas necrotic neoplasm and benign airways were not positive for either protein. The stromal and immune cell components of the tumor microenvironment were characterized by positive cytoplasmic SPARC staining of stromal spindle cells (likely fibroblasts and myofibroblasts) and alveolar macrophages and absence of staining of the infiltrating lymphocytes. In contrast, large numbers of infiltrating lymphocytes were positive for nuclear Snail staining, as were alveolar macrophages and stromal spindle cells. No significant difference was noted when comparing ADC and SCC staining of the three compartments (epithelial, stromal, and immune). These results suggest a relationship between epithelial cell Snail and SPARC in human NSCLCs.

As SPARC is known to be associated with invasive cancers and is correlated with poor patient prognosis in NSCLC (18, 19), we evaluated both the HBEC and NSCLC cell lines utilizing in vitro invasion assays. In both the HBEC and the NSCLC cell lines, Snail overexpression led to increased invasion (Fig. 2A and B). Our previous studies demonstrated that Snail overexpression confers a proliferative disadvantage to NSCLC cells in vitro (15), suggesting that the increased invasion is independent of proliferation. Having established that Snail overexpression leads to both elevated SPARC expression and invasion, we hypothesized that the increased invasion is mediated by SPARC. The HBEC line HBEC3mutP53/KRAS and cancer cell line H1437 with stable Snail overexpression (−S) or vector control (−V) were stably transfected with a plasmid containing an shRNA sequence specific to the 3′untranslated region (UTR) of SPARC or nonsilencing controls (−NS). Coordinate protein level expression of Snail and SPARC was confirmed in all cell lines by Western blot analysis (Fig. 2C and D). Utilizing the modified Boyden chambers described in Materials and Methods, we plated the cells and allowed them to invade for 48 hours. SPARC knockdown by shRNA in both HBEC and NSCLC cells reversed Snail-mediated invasion (Fig. 2E and F), indicating that SPARC is at least partially responsible for increased invasion downstream of Snail. Similar results were obtained for HBEC3mutP53/KRAS and HBEC3 (data not shown), suggesting that the SPARC effect in HBECs in not mutation-dependent for these mutations. The observation that Snail continues to drive low-level invasion of HBEC cells in the absence of SPARC, whereas SPARC knockdown totally abrogates Snail-driven invasion of NSCLC cells, suggests a differential reliance on Snail and SPARC for the invasion program at different time points during lung carcinogenesis.

The promoter region of SPARC does not contain a binding site for Snail, indicating that Snail must upregulate SPARC by an indirect mechanism. Total RNA was isolated from a panel of Snail-overexpressing NSCLC cell lines and vector controls and subjected to miRNA array. Array results were combined with mRNA array results previously published (15) and analyzed by Ingenuity Pathway Analysis Software. Combinatorial analysis revealed multiple potential intermediary molecules and signaling pathways that could be responsible for Snail-mediated upregulation of SPARC; most prominently that Snail may interact with SPARC via TGF-β, ERK1/2, and miR-29b. We evaluated the secreted protein levels of one potential candidate, the cytokine TGF-β1, utilizing ELISA. TGF-β1 secretion was significantly increased (HBEC2 FC = 3.1, HBEC3 FC = 2.4, HBEC7 FC = 3.6, and H3mut FC = 5.0; P ≤ 0.0001) in Snail-overexpressing HBEC lines compared with vector controls (Fig. 3A), indicating that TGF-β is upregulated by Snail and may be upstream of SPARC. Treatment of the parental HBEC lines with recombinant TGF-β1 resulted in increased expression of Snail, SPARC, and phosphorylation of ERK1/2 (Fig. 3B), suggesting an autocrine signaling mechanism for Snail and TGF-β1 expression. Inhibition of TGF-β1 expression by siRNA abrogated Snail-mediated ERK1/2 phosphorylation and SPARC expression (Fig. 3C), further suggesting that TGF-β1 is necessary for Snail-mediated SPARC expression. The knockdown efficiency of the TGF-β siRNAs was measured by ELISA following inhibition (Fig. 3D). Immunoblot analysis of a panel of HBEC lines indicated that TGF-βR2 is also elevated in Snail-overexpressing lines (Fig. 1D), suggesting that both the ligand and the receptor are upregulated by Snail and are intermediates in the Snail-to-SPARC pathway.

ERK1/2 was also indicated as a potential intermediate between Snail and SPARC (29–31). Immunoblot analysis of a panel of HBEC lines indicated that ERK phosphorylation was increased in Snail-overexpressing lines (Fig. 1D) and may be an intermediate in the Snail-to-SPARC pathway. Suppression of ERK activation by the MEK1/2 phosphorylation inhibitor U0126 led to decreased mRNA and protein expression of SPARC in Snail-overexpressing HBEC lines (Fig. 4A and B). In addition, U0126 exposure led to increased expression of the miRNA miR-29b (Fig. 4C).

Our miRNA array data analysis indicated that miR-29b levels were significantly reduced in H441 and H1437 Snail–overexpressing cell lines compared with vector controls (FC = 2.3 and 2.6, respectively). On the basis of these findings, we hypothesized that miR-29b could be an intermediate regulator in the Snail-to-SPARC pathway. Analysis of the 3′UTR of SPARC mRNA by TargetScan software revealed three putative binding sequences for miR-29b. We compared the levels of miR-29b in three NSCLC and four HBEC cell lines ectopically expressing Snail to levels in the corresponding vector control lines by qRT-PCR using a TaqMan miRNA assay. In both types of cell lines, miR-29b expression levels were lower in Snail-overexpressing lines compared with vector controls (Fig. 5A and B), indicating that miR-29b is downregulated by Snail and may be involved in SPARC regulation. Furthermore, transfection of the cells with a miR-29b precursor mimic led to downregulation of SPARC protein in a panel of Snail-overexpressing HBEC lines (Fig. 5C). In addition, miR-29b expression was significantly upregulated following MEK/ERK inhibition (Fig. 4C), indicating a direct link between Snail, TGF-β, MEK/ERK, and miR-29b upstream of SPARC. We propose a regulatory pathway wherein Snail upregulates TGF-β in an autocrine or paracrine fashion, leading to activation of the MEK/ERK pathway, downregulation of miR-29b, and finally upregulation of SPARC (Fig. 5D).

This is the first report demonstrating that Snail upregulates SPARC in models of both pulmonary premalignancy and established NSCLC. The necessity of SPARC for robust Snail-mediated invasion in both models suggests a role for both Snail and SPARC in the pathogenesis of NSCLC. Furthermore, we have identified a number of critical intermediates in this pathway. The deregulation of both the TGF-β and MEK/ERK pathways in malignancy is well known (32–34); however, this is the first description of their deregulation in the context of Snail-mediated lung carcinogenesis.

To progress from in situ to metastatic disease, tumors must acquire characteristics that allow them to degrade and invade their local basement membrane and then migrate before forming a micrometastatic focus. As part of this process, tumors undergo a series of events comprising the EMT program, wherein cells may transiently lose epithelial characteristics and become more mesenchymal in phenotype and molecular profile. The classical model of “linear” tumor progression proposes that a small population of cells within the invasive edge of an established tumor acquires characteristics necessary for EMT. However, recent findings suggest that micrometastases may occur very early in the course of the disease, supporting a recently proposed model of “parallel” tumor progression in which metastatic dissemination occurs throughout the course of primary tumor development (35–39). This phenomenon may be especially relevant in the clinical subset of lung cancers that apparently disseminate early, leading to the clinical problem of metastatic disease following surgery for early-stage disease. According to this model, metastases arise from a subpopulation of stem cell–like cells present at tumor initiation, which express proteins that induce EMT and confer putative stem cells with migratory and invasive capacity (36). The clinical course of patients presenting with metastatic disease following surgery is consistent with this model. Parallel progression seems to be a frequent and major clinical problem in the treatment of NSCLC; as many as 40% of patients will have recurrence of lung cancer at metastatic sites following lung cancer resection (1), which is thought to be due to micrometastatic disease that is below the level of detection by imaging studies at the time of surgery (35, 37, 40). Thus, our finding of the necessity of SPARC for Snail-mediated invasion in premalignant and established NSCLC models suggests that the Snail–SPARC axis may play a role in this process.

Although our observation of the highly motile and invasive capacity of epithelial cells relative to tumor cells may seem counterintuitive, these findings are actually consistent with those in the recent literature that document the profoundly motile capacity of epithelial cells (38, 39, 41). This motile capacity is consistent with the movement characteristics of epithelial cells in embryonic development and wound closure. In teleologic terms, these profound capacities (movement and invasion) endowed by the genetic program EMT are inherent in embryonic development and wound closure which are requisite phenotypes whose evolutionary histories apparently far exceed that of malignancy.

The studies described here delineate a Snail-to-SPARC pathway involving a set of intermediates, several of which have been implicated in carcinogenesis. The molecules and pathways described here are likely not the only intermediaries in the Snail-to-SPARC pathway, and further studies will be required to identify additional pathways, as well as possible epigenetic changes involved. Although we propose a largely linear pathway for Snail-to-SPARC expression, it is likely that the interactions among all of the involved intermediates are more complex. For example, SPARC is known to increase expression of Snail and therefore repress E-cadherin in melanoma progression (42). In hepatic stellate cells, TGF-β and SPARC are known to cooperate in an autocrine feedback loop revealed by the finding that SPARC knockdown reduces TGF-β1 secretion, whereas TGF-β1 treatment increases SPARC gene expression (43). As we demonstrate here, TGF-β1 and Snail also cooperate in an autocrine or paracrine feedback loop in lung cancer. Furthermore, miR-29b likely has many targets in addition to SPARC, with as many as 7,000 predicted by computational algorithms. These signaling pathways are known to have numerous effects on processes independent of invasion and metastatic progression, including cell-cycle regulation, proliferation, and survival. Understanding the contribution of the Snail–SPARC axis to these phenotypes may yield a more thorough understanding of the pathogenesis of NSCLC.

Here, we have identified molecular pathways in which altered genetic expression or activation in an in vitro system lay the groundwork for additional investigation of the potentially important Snail–SPARC pathway in human lung cancer.

No potential conflicts of interest were disclosed.

Conception and design: J.L. Grant, K. Krysan, J.D. Minna, T.C. Walser, S.M. Dubinett

Development of methodology: J.L. Grant, M.C. Fishbein, L.-S. Hong, K. Krysan, J.D. Minna, T.C. Walser

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.L. Grant, M.C. Fishbein, J.D. Minna, J.W. Shay, T.C. Walser

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.L. Grant, M.C. Fishbein, J.W. Shay, T.C. Walser, S.M. Dubinett

Writing, review, and/or revision of the manuscript: J.L. Grant, M.C. Fishbein, K. Krysan, J.D. Minna, J.W. Shay, T.C. Walser, S.M. Dubinett

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.L. Grant, M.C. Fishbein, L.-S. Hong, T.C. Walser

Study supervision: T.C. Walser, S.M. Dubinett

The authors thank Ying Lin and Zhe Jing for their technical assistance, as well as the UCLA Genotyping and Sequencing Core, the Clinical Microarray Core, and UCLA Core Facility funding sources.

This study was financially supported in part by the National Cancer Institute (#T32-CA009120-36; to S.M. Dubinett and J.L. Grant; #U01CA152751; to S.M. Dubinett, M.C. Fishbein, and T.C. Walser), Department of Defense (#W81XWH-10-1-1006; to S.M. Dubinett, M.C. Fishbein, K. Krysan, J.D. Minna, and T.C. Walser), UCLA Clinical and Translational Science Institute (#UL1TR000124; to S.M. Dubinett), Merit Review Research Funds from the Department of Veteran Affairs (#5I01BX000359; to S.M. Dubinett), Lung Cancer SPORE (#P50CA70907; to J.D. Minna and J.W. Shay), NASA NSCOR (#NNX11AC54G; to J.D. Minna and J.W. Shay), and Tobacco-Related Disease Research Program (#18FT-0060; to T.C. Walser; #20KT-0055; to T.C. Walser; and #21DT-0013; to J.L. Grant).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.