Regulation of Lung Cancer Metastasis by Klf4-Numb–like Signaling

The ability of tumor cells to colonize distant organs from a primary site, namely metastasis (1), heralds an incurable and almost invariably fatal disease course. It is now clear that this process involves broad remodeling of transcriptional programs, loss of tumor suppressor mechanisms, interplay with stromal components, and changes in cellular properties that control cell migration and invasion (2).

One expanding gene regulatory network implicated in tumor cell motility includes miRNAs (miR; ref. 3). These molecules can function as oncogenes or tumor suppressors (4), but changes in miR expression in disparate tumor types, with concomitant downstream modulation of gene transcription, have been often linked to the acquisition of metastatic traits (5). In some malignancies, changes in gene expression controlled by oncogenic miRs may also influence the pool of cancer-initiating, progenitor-like cells (6, 7). Although still phenotypically ill defined (8), these are considered rare cells with stem-like properties that may participate in the progression of at least certain tumors (9), for their resistance to drug-induced therapy (10), and propensity to metastatic dissemination (11).

In this context, decreased expression or complete loss of miR-296-5p (miR-296) has been observed in several types of human cancers (12–14) and often correlated with disease progression. Although elevated levels of miR-296 have been shown in tumor-associated angiogenic endothelial cells, and contributed to growth factor signaling in these settings (15), other data have proposed a pivotal role of miR-296 in tumor suppression, mechanistically linked to modulation of HMGA1 oncogene expression (16) and deregulated function of the Scribble (17) cell polarity module (18).

In this study, we sought to further map the miR-296 tumor suppressor network for potential regulation of novel metastatic traits, specifically in lung cancer.

Human lung carcinoma A549, H23, H460, H1299, H1437, and H1792 or MDA-MB-231 breast cancer cells were purchased from the American Type Culture Collection. Human embryonic kidney HEK293 cells were available in our laboratories. All cell lines were maintained in a 5% humidified incubator at 37°C and kept in culture as recommended by the supplier. Immortalized human bronchial epithelial cells (HBEC3) were a generous gift from Dr. Marcelo Kazanietz (University of Pennsylvania, Perelman School of Medicine, Philadelphia, PA). HBEC3 cells were cultured in keratinocyte-serum-free medium (SFM) containing 50 μg/mL bovine pituitary extract and 5 ng/mL EGF media until passage 7. All cell culture reagents were from Gibco-Invitrogen (Life Technologies).

Aliquots of lung, breast cancer, or HBEC3 cells were harvested 48 or 72 hours after transfection and solubilized in 150 μL radioimmunoprecipitation assay (RIPA) buffer supplemented with 1× complete protease and phosphatase inhibitors cocktails (Roche). Cell lysates (50 μg) were separated by electrophoresis on 10% to 12% SDS-PAGE, transferred to polyvinylidene difluoride (PVDF) membranes (Millipore) and probed with 1 μg/μL of antibodies against Numbl (Proteintech Group Inc.), Scrib (Santa Cruz Biotechnology), Numb (Proteintech), β-catenin (Thermoscientific), c-Src (Santa Cruz Biotechnology), Tyr416-phosphorylated Src (p-Src; Biosource International), fibronectin (H-300; Santa Cruz Biotechnology), p21^WAF1/Cip1 (Calbiochem, EMD Millipore Corporation), hemagglutinin (HA; Sigma-Adrich), laminin A (Santa Cruz Biotechnology), β-tubulin, or β-actin (all from Sigma-Aldrich). Antibodies to focal adhesion kinase (FAK), Tyr397-phosphorylated FAK (p-FAK), vimentin, Nanog, or Klf4 were from Cell Signaling. Reactive bands were visualized with ECL Plus reagents (GE Health Care).

For immunofluorescence experiments, lung cancer or HBEC3 cells were grown on cover-glasses, fixed in 4% paraformaldehyde for 15 minutes, permeabilized in ice-cold methanol, and incubated with an antibody to Numbl or Numb (both 10 μg/μL; Proteintech) for 16 hours at 4°C, followed by a fluorescein isothiocyanate (FITC)-conjugated anti-rabbit secondary antibody (1:100; ThermoScientific) with or without an antibody to HA-tagged Klf4 (1:100; Sigma-Aldrich).

Slides were scored by light or fluorescent microscopy and photographed images were arranged with Adobe Photoshop CS5 for Windows. When confocal or 2-photons microscopy analyses were conducted, samples were imaged using a Leica TCS SP2 confocal or a Prairie Instruments Ultima 2 Photon microscopes, respectively.

Transfected A549 cells were labeled with Hoechst 33342 (Cell Signaling Technology, Inc.), as described previously (19, 20). Briefly, cells were suspended at 1 × 10⁶/mL in prewarmed Dulbecco's Modified Eagle's Medium (DMEM)-2% fetal calf serum (FCS) and 10 mmol/L HEPES buffer. Hoechst 33342 was added at a final concentration of 5 μg/mL in the presence or absence of reserpine (50 μmol/L; Sigma-Aldrich). Cells were incubated for 2 hours at 37°C with intermittent shaking, washed by centrifugation at 4°C with ice-cold Hank's Balanced Salt Solution (HBSS)-2% FCS and 10 mmol/L HEPES (HBSS+), and suspended in ice-cold HBSS+ at a final concentration of 2 × 10⁷/mL. Propidium iodide (PI; BD Biosciences) was added at a final concentration of 2 μg/mL to exclude dead cells. Before sorting, cells were filtered through a 40-μm cell strainer to obtain single cell suspension. All media reagents were from Gibco-Invitrogen (Life Technologies). Cell sorting and side population analyses were conducted on a FACSAria using the FACSDiva (version 6.1.2; BD Biosciences) or FlowJo software (version 7.6.5; Tree Star Inc.). The Hoechst 33342 dye was excited at 357 nm and its fluorescence was dual-wavelength analyzed (blue, 402–446 nm; red, 650–670 nm).

Subconfluent (70%) cells were transfected with miR-296 mimics, targeted siRNAs or controls, and cultured for 48 hours. Wounds in the cell monolayers were created using a P200 micropipette tip, and the migration distance (units) was determined as reduction in the wound's gap and quantified using NIH Image-J software, as described (17). For analysis of directional cell migration, monolayers of A549 cells were harvested 60 hours after transfection, scratched and examined for Golgi orientation relative to the nucleus and the migration front after additional 6 hours by immunofluorescence (17, 21). Golgi staining was conducted using Alexa Fluor555–conjugated anti-GM130 antibody (BD Biosciences). For analysis of stress fiber formation, F-actin was visualized by phalloidin–TRITC staining (Sigma-Aldrich), and samples mounted in 4′,6-diamidino-2-phenylindole (DAPI) I (Abbott) were imaged using an AxioImager Z1 microscope (Carl Zeiss).

A549 cells transfected with miR-296 mimics were seeded (5 × 10⁴ cells) after 48 hours in serum-free medium in Matrigel-coated chambers (8.0-μm pores; BD Biosciences). When siRNA or gene overexpression was conducted, HBEC3 or the indicated lung cancer cell type were seeded in Matrigel-coated inserts after 24 hours. FBS-containing medium was used as chemoattractant in the bottom chamber, and cells were allowed to migrate for 24 hours, as described previously (17, 22). Cells that had invaded the lower surface of the membrane were fixed with methanol, stained with DAPI, and quantified by fluorescence microscopy (Nikon E600, Nikon Instruments Inc.).

All experiments involving animals were approved by an Institutional Animal Care and Use Committee. Female severe combined immunodeficient (SCID)/beige mice (6–8 weeks of age) were anesthetized with ketamine hydrochloride, the abdominal cavity was exposed by laparotomy, and animals were injected in the spleen with 4 × 10⁶ H460 or A549 cells previously transfected with control nontargeting or Numbl-directed siRNA. To avoid potential confounding effects due to variable growth of a primary tumor, the spleen was removed 24 hours after injection of the tumor cells, as previously described (23). On day 11, all mice in the 2 groups were sacrificed. Liver samples were formalin-fixed, paraffin-embedded, and analyzed histologically by hematoxylin and eosin staining.

A series of 209 consecutive patients surgically treated for non–small cell lung carcinoma (NSCLC) at Fondazione IRCCS Ca' Granda Hospital (Milan, Italy) between 2000 and 2004 was available for this study. This patient series included 149 cases of adenocarcinoma and 60 cases of squamous cell carcinoma (SCC) of the lung. Clinical outcome data were available for 172 patients (82%). NSCLC cases were staged according to the current tumor–node–metastasis (TNM) classification of malignant tumors (International Union Against Cancer, UICC, 7th edition, 2009). The clinical characteristics of the patient series analyzed in this study are summarized in Supplementary Table S1. An informed consent was obtained from all patients enrolled, and the study was approved by an Institutional Review Board of the Fondazione IRCCS Ca' Granda. The follow-up period ranged from 0 to 132 months (average 55.2 months). At the last follow-up (January 2011), 100 patients were deceased for progression of NSCLC, whereas 72 patients were alive. For 4 patients, matched distant metastatic lesions were available as well. Frozen tissues were available from an independent series of patients with NSCLC as previously described (24).

Differences among sample groups were analyzed using 1-sided Student t test, or Wilcoxon signed-rank test. For survival analysis, the Kaplan–Meier method was used. To examine a potential association of Numbl staining and overall survival, patients with NSCLC were assigned to 2 groups according to target expression. Cases whose immunoreactivity was scored below Numbl median value of immunoreactivity fell in the Numbl^low group, whereas cases whose immunoreactivity levels were above the median value fell in the Numbl^high group. The 2-sided log-rank test was used to compare survival curves. Statistical analyses were conducted using GraphPad Prism version 4 for Windows or Ministat 2.1 software. A P < 0.05 was considered as statistically significant.

We began this study by looking at additional targets of miR-296 potentially implicated in tumor suppression (17). In addition to Scribble (17), transfection of model A549 NSCLC cells with miR-296 inhibited the mRNA (Fig. 1A) and protein (Fig. 1B) expression of Numb-like (Numbl), a polarity protein originally described in neuronal progenitors (25, 26). In contrast, transfection of tumor cells with anti-miR-296 had no effect on Scribble or Numbl levels, compared with control cultures (Fig. 1A and B). Two putative miR-296-responsive sites were predicted in the Numbl 3′-untranslated region (UTR; Fig. 1C), and their combined mutagenesis reversed miR-296 repression of Numbl regulatory sequences (Fig. 1D), indicating that Numbl is a direct gene target of miR-296. Functionally, expression of miR-296 inhibited lung cancer cell migration in a wound closure assay (Fig 1E), blocked tumor cell invasion across Matrigel-coated inserts (Fig. 1F), and suppressed colony formation in soft agar (Fig. 1G). In reciprocal experiments, transfection of anti-miR-296 increased tumor cell migration, invasion, and colony formation (Fig. 1E–G). Consistent with a tumor-suppressive function, miR-296 levels were downregulated in lung cancer cells, compared with normal human bronchial epithelial HBEC3 cells (Fig. 1H), and were progressively lost during stepwise tumorigenesis in transgenic mice that express the mutant k-Ras-G12D oncogene in the lung (ref. 27; Fig. 1I). In addition, miR-296 expression was significantly reduced in patients with NSCLC, compared with normal lung (Fig. 1J).

To determine whether Numbl functioned in the miR-296 tumor suppressor network, we next silenced its expression by siRNA. Transfection of lung cancer cells with Numbl-directed siRNA suppressed Numbl mRNA (Fig. 2A) and protein (Fig. 2B) levels, compared with nontargeting siRNA. Numbl knockdown reproduced the effect of miR-296 expression and suppressed the migration (Fig. 2C) and invasion (Fig. 2D) of lung cancer cells (Supplementary Fig. S1A), compared with control transfectants. Similarly, siRNA knockdown of Numbl in breast adenocarcinoma MDA-MB231 cells (Supplementary Fig. S1B) inhibited tumor cell migration (Fig. 2E) and invasion (Supplementary Fig. S1C). In contrast, silencing of Numbl in prostate adenocarcinoma PC3 cells had no effect (Supplementary Fig. S1D), suggesting a tumor-specific response. Consistent with a potential loss of polarity function, silencing of Numbl in lung cancer cells resulted in extensive cytoskeletal changes, with reduced content of oriented actin fibers, compared with control transfectants (Fig. 2F). In addition, loss of Numbl impaired directional cell migration, characterized by defective Golgi orientation relative to nuclei and the migration front of tumor cells (Fig. 2G).

The biochemical requirements of Numbl regulation of tumor cell motility were next investigated. First, Numbl silencing in lung cancer cell types was associated with decreased expression of fibronectin, β-catenin and vimentin (Fig. 3A), thus differently from conventional hallmarks of epithelial–mesenchymal transition (10). Conversely, loss of Numbl induced dephosphorylation, that is, inactivation of FAK on Tyr397 (Fig. 3B) and of Src on Tyr416 (Fig. 3C), 2 pivotal cell motility effectors.

Next, we asked whether Numbl-directed random cell motility influenced the metastatic propensity of tumor cells, and we first looked at potential changes in anoikis. This is a form of cell death caused by detachment of epithelial cells from the extracellular matrix and considered a key determinant of metastasis (28). Numbl knockdown in lung cancer cells maintained in suspension was associated with nearly completely loss of cell viability, whereas adherent cells were not significantly affected by PKH26 vital staining and fluorescence microscopy (Fig. 3D). Consistent with these data, siRNA silencing of Numbl in attached lung cancer cell types had negligible effects on cell viability, by flow cytometry quantification of hypodiploid, that is, sub-G₁ DNA content (Fig. 3E). In contrast, knockdown of Numbl under conditions of ultra-low cell attachment resulted in increased cell death, compared with control transfectants (Fig. 3E). These cells exhibited increased reactivity for Annexin V, consistent with apoptosis, by multiparametric flow cytometry (Fig. 3F). To test whether this pathway was important, in vivo, we next injected lung cancer cells transfected with control or Numbl-directed siRNA in the spleen of immunocompromised mice and looked at their ability to form liver metastasis within an 11-day interval. In these experiments, control transfectants formed large metastatic foci in the liver of all reconstituted animals (Fig. 3G). In contrast, siRNA silencing of Numbl nearly completely abolished the ability of lung cancer cells to form liver metastasis in this model (Fig. 3G).

To elucidate the mechanism(s) of Numbl regulation of tumor cell motility and invasion, we next used an inducible luciferase reporter array of 45 transcription factors (Fig. 4A). siRNA silencing of Numbl under these conditions resulted in increased activity of 7 transcription factors (Fig 4B) implicated in pluripotency and differentiation (Klf4, VDR, or c-Myc; ref. 29), or stress-sensing (Nrf1-2, AP-1, ATF6, or YY1) responses (30). In particular, derepression of Kruppel-like factor-4 (Klf4)-dependent transcription was the most significant change observed in response to Numbl knockdown (Fig. 4B). To validate these changes, we next looked at potential modulation of Klf4 target gene expression in response to Numbl targeting. In these experiments, transfection of lung cancer cells with Numbl-directed siRNA, but not control nontargeting siRNA, increased the expression of p21^Cip1/WAF1 mRNA (Fig. 4C) and protein (Fig. 4D) and downregulated Sparc levels (ref. 31; Fig. 4C), 2 known Klf4 target genes. Overall Klf4 protein levels did not change and a control siRNA had no effect (Fig. 4C and D). To further test the specificity of Numbl regulation of Klf4-dependent transcription, we next looked at the levels of several other “stemness” molecules under conditions of Numbl-Klf4 targeting. In these experiments, siRNA silencing of Klf4, Numbl, or the potentially related molecule, Numb (see later), did not significantly affect the mRNA levels of receptor–ligand pairs in the Notch (Notch-Jagged 1) or Sonic hedgehog (SHH-Gli1) pathway (Fig. 4E). Overall levels of c-Myc were also not significantly different between the variously silenced cells and control cultures (Fig. 4E). We next studied a potential reciprocal relationship between Numbl and Klf4 in normal and lung cancer cells. Endogenous Numbl was present more abundantly in nuclei of tumor cells, with a smaller pool of the molecule expressed in their cytosol (Fig. 4F). In contrast, endogenous Klf4 was almost undetectable in lung cancer cells, and became expressed solely in nuclei, but not cytosol, upon plasmid transfection (Fig. 4F). In gene silencing studies, knockdown of Numbl or Klf4 did not affect the expression or subcellular localization of the other molecule (Fig. 4G). In contrast, expression of Klf4 in normal or tumor cell types was associated with Numbl colocalization in nuclei of transfected cells, by confocal microscopy (Fig. 4H).

Numbl has been proposed as a protein related to Numb, a regulator of cell fate specification and Notch1 signaling (32), and this potential relationship in lung cancer was next investigated. siRNA silencing of Numbl did not affect the expression of Numb in lung cancer cells (Fig. 5A). At variance with the results obtained with Numbl, silencing of Numb mRNA (Fig. 5B) or protein expression (Fig. 5C) increased lung cancer cell invasion across Matrigel-coated inserts (Fig. 5D). The response to anoikis was also different between the 2 molecules, as siRNA knockdown of Numb did not affect the viability of lung cancer cells under attachment conditions or forced to remain in suspension, by DNA content analysis and flow cytometry (Fig. 5E). With respect to potential modulation of Klf4 target genes, knockdown of Numb did not affect p21^Cip1/WAF1 levels, whereas Sparc expression was actually increased (Fig. 5F).

Klf4 orchestrates a pivotal “stemness” transcriptional program (33), which may participate in context-specific tumor suppression (31), including in lung cancer (34). Consistent with a potential tumor suppressive function, Klf4 was nearly undetectable in lung cancer cells, compared with normal HBEC3 (Fig. 6A), and its reintroduction in tumor cells by plasmid transfection (Supplementary Fig. S2A) was sufficient to inhibit cell invasion across Matrigel inserts (Fig. 6B). siRNA silencing of Klf4 (Supplementary Fig. S2B and S2C) had no effect in tumor cells (Fig. 6B). In contrast, Klf4 knockdown enhanced the invasive potential of normal HBEC3 cells (Supplementary Fig. S2C and Fig. 6C). In parallel experiments, transfection of Numbl (Supplementary Fig. S2A and S2C) had no effect on tumor cell motility (Fig. 6B) and modestly increased HBEC3 invasion (Fig. 6C).

To determine whether Numbl-Klf4 signaling affects a tumor-initiating progenitor phenotype, we next examined the side population of lung cancer types, which is enriched in stem-like cells (20). Transfection of Klf4 significantly reduced the side population of A549 cells (Fig. 6D). In contrast, transfection of these cells with a Numbl cDNA had the opposite effect and expanded the side population compartment of A549 cells by approximately 2-fold (Fig. 6D). Consistent with the results above, siRNA silencing of Klf4 had no effect on the side population of A549 cells as compared with control transfectants (Fig. 6D). Next, we looked at the individual cell populations isolated by fluorescence sorting. Numbl was preferentially expressed in the side population of A549 cells by Western blotting (Fig. 6E) and quantitative real-time PCR (qRT-PCR; Fig. 6F). Mirroring these results, its upstream repressor, miR-296, was nearly exclusively present in the nonside population fraction of A549 cells (Fig. 6F). With respect to other developmentally regulated “stemness” markers, only c-Myc was preferentially enriched in the side population of A549 cells (Fig. 6G). In contrast, expression of Notch, SHH, or Klf4 was not statistically different in side population or nonside population fractions of A549 cells (Fig. 6G). Functionally, siRNA knockdown of Numbl depleted the side population of A549 cells by up to 87%, compared with control transfectants (Fig. 6H). This was associated with nearly complete inhibition of tumor cell invasion across Matrigel inserts (Fig. 6I). Consistent with the data reported earlier, silencing of Numb did not modulate the side population of A549 cells (Supplementary Fig. S3).

On the basis of the data above, we next looked at the potential impact of a Numbl-Klf4 signaling axis in human tumors. Consistent with its oncogenic properties, in vitro, the expression of Numbl was significantly increased in many human tumors by immunohistochemistry (Fig. 7A and B and Supplementary Fig. S4), and analysis of public databases (Supplementary Fig. S5). When compared with other “stemness” markers, Nanog expression was also found broadly upregulated in various patient-derived human tumors (Supplementary Fig. S6A), whereas the levels of Oct4 were more restricted to testicular and breast cancer (Supplementary Fig. S6B).

With respect to NSCLC, the tumor cell population stained intensely positive for Numbl, compared with the adjacent non-neoplastic parenchyma (Supplementary Fig. S7A), with higher levels in SCC than adenocarcinoma (Fig. 7C). In both human samples (Supplementary Fig. S7B) and mouse k-Ras–mutant lung tumors (ref. 27; Fig. 7D), Numbl levels increased with disease progression from localized lesions to metastatic foci. In terms of disease outcome, Numbl was associated with shorter overall survival in patients with NSCLC (n = 209; HR, 1.7, 95% confidence interval (CI), 1.037–2.439; P = 0.03; Fig. 7E), and its expression in these cases correlated with the presence of other stem cell markers, including Nanog and CD44 (ref. 35; Fig. 7F). In addition to deregulated expression, Numbl was subcellularly mislocalized in human tumors from a membranous staining in the normal colonic mucosa (Fig. 7G) to a diffuse cytoplasmic and perinuclear distribution in colorectal adenocarcinoma (Fig. 7H) and matched liver metastasis (Supplementary Fig. S7C).

In this study, we have shown that a developmentally regulated polarity protein originally described in neuronal progenitors, Numbl (25, 26), becomes subcellularly deregulated and overexpressed in various human cancers. This results from loss of a tumor-suppressive miR, miR-296 (17), which actively represses Numbl expression. In turn, aberrantly increased levels of Numbl support multiple metastatic traits in lung cancer, including enhanced cell invasion, resistance to anoikis, maintenance of cancer-initiating, progenitor-like cells, and metastatic competency, in vivo. Mechanistically, this pathway involves Numbl inhibition of the Klf4 “stemness” transcriptional program (33, 36), resulting in shortened overall survival in patients with lung cancer.

Polarity proteins are known for maintaining the integrity of epithelia and directional cell migration (18). Perturbation of their function in cancer, especially with respect to gap junction assembly, cell orientation, and cell invasion has also been postulated (37), suggesting that these molecules may participate in an evolutionary conserved pathway of tumor suppression (38). However, recent data point to a more complex scenario, as polarity proteins become broadly overexpressed and subcellularly mislocalized in various malignancies (39) and exploited for random, as opposed to directional cell migration, tumor cell invasion, and disease maintenance (17). The results presented here further support the model of a general exploitation of polarity proteins for tumor progression and identify Numbl as a novel effector of aberrant lung cancer cell motility, invasion, and metastasis, in vivo. Although Numb (32) and Numbl have been proposed as related gene products with potentially overlapping functions in cell fate specifications (25), our results show that these molecules have instead entirely distinct functions in cancer, with no role for Numb in tumor cell motility.

One of the emerging roles of polarity proteins is in the regulation of stem cell–related functions (40), including cancer stem cells (41). The role of these rare, potentially cancer-initiating cells in disease maintenance is far from elucidated (42), but there is evidence from clinical correlates (9) and genetic disease models in mice (43) that they may play a role in the pathogenesis and progression of lung cancer. Here, Numbl expression emerged as a novel requirement to maintain, and potentially expand the side population of lung cancer cell types, a subset enriched in the stem cell phenotype (20) that carries elevated metastatic potential (44). A key mechanistic requirement of this pathway was the ability of Numbl to antagonize Klf4-dependent transcription, a pivotal “stemness” program (29, 33), implicated in context-dependent tumor suppression (31), including in lung cancer (34).

Although the molecular requirements of how Numbl inhibits Klf4-dependent gene expression need to be further elucidated, the 2 molecules were shown to colocalize in nuclei of tumor cells, and expression of Klf4, alone, was sufficient to shut off tumor cell motility and eliminate the progenitor-like population of lung cancer cells. Altogether, these data suggest a dual role of deregulated Numbl in lung cancer, promoting aberrant cell migration and invasion when localized in the cytosol, and antagonizing Klf4-dependent suppression of the cancer stem cell phenotype in the nucleus. Consistent with this model, Klf4 was virtually undetectable in lung cancer cell types, whereas it potently antagonized migration of normal bronchial epithelial cells. The ability of Numbl to prevent anoikis (28) and maintain phosphorylated levels of known survival kinases, including FAK and Src, may also elevate the antiapoptotic threshold in the cancer-initiating stem-like compartment, thus further enhancing their metastatic propensity. Irrespectively, Numbl-directed tumorigenesis emerged here as a major determinant of unfavorable outcome in patients with lung cancer, correlating with expression of other stem cell markers, Nanog and CD44 (8), but not Oct4, and overall supporting an important pathogenetic role of the stem/progenitor cell compartment in lung cancer progression (9).

In sum, this study identified novel molecular determinants of lung cancer metastasis, uncovering a broad oncogenic role for the Numbl polarity protein in tumor cell invasion coupled to the maintenance of progenitor-like cancer-initiating cells (41). Despite gaps in our understanding of cancer stem cells (42) and their interplay with pathways of cell motility and drug resistance (10), targeting the Numbl pathway described here may open new therapeutic prospects to limit the metastatic potential of advanced lung cancer.

No potential conflicts of interest were disclosed.

Conception and design: V. Vaira, S. Bosari, D.C. Altieri

Development of methodology: V. Vaira, S. Ferrero

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.M. Martin, D.S. Garlick, M. Nosotti, J.L. Kissil

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V. Vaira, A. Faversani, D.S. Garlick, S. Ferrero, J.L. Kissil, S. Bosari, D.C. Altieri

Writing, review, and/or revision of the manuscript: V. Vaira, M. Nosotti, S. Bosari, D.C. Altieri

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Nosotti, D.C. Altieri

Study supervision: S. Bosari, D.C. Altieri

The authors thank Frederick Keeney (Wistar) for help with imaging studies and Jeffrey Faust (Wistar) for help with flow cytometry-sorting analyses.

This work was supported by NIH grants CA140043, HL54131, CA78810, and CA118005 (to D.C. Altieri) and by grants from Fondazione Cariplo (2010-0846), Fondazione Berlucchi and Ministero della Salute “5 × 1000” (S. Bosari). A. Faversani was supported by a fellowship of the Doctorate School of Molecular Medicine at Università degli Studi di Milano. Support for Core Facilities used in this study was provided by Cancer Center Support Grant (CCSG) CA010815 to The Wistar Institute.

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.