Smoking is the most important known risk factor for the development of lung cancer. Tobacco exposure results in chronic inflammation, tissue injury, and repair. A recent hypothesis argues for a stem/progenitor cell involved in airway epithelial repair that may be a tumor-initiating cell in lung cancer and which may be associated with recurrence and metastasis. We used immunostaining, quantitative real-time PCR, Western blots, and lung cancer tissue microarrays to identify subpopulations of airway epithelial stem/progenitor cells under steady-state conditions, normal repair, aberrant repair with premalignant lesions and lung cancer, and their correlation with injury and prognosis. We identified a population of keratin 14 (K14)–expressing progenitor epithelial cells that was involved in repair after injury. Dysregulated repair resulted in the persistence of K14+ cells in the airway epithelium in potentially premalignant lesions. The presence of K14+ progenitor airway epithelial cells in NSCLC predicted a poor prognosis, and this predictive value was strongest in smokers, in which it also correlated with metastasis. This suggests that reparative K14+ progenitor cells may be tumor-initiating cells in this subgroup of smokers with NSCLC. Cancer Res; 70(16); 6639–48. ©2010 AACR.
Lung cancer has the highest mortality rate of all cancers, and patients with the diagnosis of metastatic non–small cell lung cancer (NSCLC) have a median survival of just 4 to 5 months (1). Lung cancer is a heterogeneous disease, and the natural history is still not well understood. The classic exponential growth model of tumor metastasis may not be relevant in some tumors, in which the biology of disease may affect prognosis more than the time and size of growth of the tumor (2). For example, as many as 40% of patients with completely resected stage I NSCLC will experience a recurrence of their disease, which suggests that a subpopulation of cells in these tumors is more prone to micrometastatic behavior (2).
The cancer stem cell (CSC) model of tumor development and progression refers to the presence of a population of rare cells in a tumor that have stem cell properties, namely they are capable of self-renewal and differentiation to their progeny. In this model, the self-renewal capacity of the CSCs are responsible for maintaining tumor growth indefinitely, and the other cells that make up most of the tumor are actively proliferating and differentiating, and therefore susceptible to current conventional cancer therapies (3–10). Consistent with this model, CSCs would be considered to be tumor-initiating cells (3–10). Recently, it has been found that CSCs may not necessarily represent rare cells in a tumor; that the tumor-initiating cell in a cancer reflects a cell with the property of indefinite self-renewal; and that this could be a rare stem cell, a progenitor cell, or a differentiated cell that has developed the ability to self-renew (11). These tumor-initiating cells are thought to arise from cells that have dysregulated repair resulting in indefinite self-renewal and are associated with relapse and recurrence of cancers, and a poor prognosis presumably due to resistance to chemotherapy and radiotherapy (3, 5–10). This model of CSCs leading to tumor resistance fits well with the natural history of lung cancer with its high incidence of recurrence and metastasis. In lung cancer, no population of dysregulated self-renewing cells has previously been found that correlates with poor prognosis.
Our understanding of stem and progenitor cells in the proximal airway epithelium is limited, but some populations have been identified with self-renewing and differentiation properties (12–15). Keratin 5 (K5)–expressing basal cells are considered to be progenitor cells in the adult large airways at steady state and during airway epithelial repair (12–15). In humans, unlike mice, K5-expressing basal cells have been found throughout the tracheobronchial tree (12). It was previously thought that keratin 14 (K14) is the obligate intermediate filament-binding partner of K5 in the basal cells of the airway epithelium (12, 16). However, although K14+ progenitor epithelial cells in the airway are important for repair, K14+ cells are rarely found in the airway epithelium under homeostatic conditions, whereas K5+ cells are relatively abundant (12, 16). We show here for the first time that K14-expressing cells are a unique subpopulation of airway epithelial cells that are mostly present in the submucosal gland/ducts in the steady state (17). Although K14+ progenitor epithelial cells in the airway are important for normal repair (12, 16), persistence of K14 expression is found in aberrant repair with premalignant lesions and in a subset of NSCLCs associated with injury from smoking. The primary objective of this study was to determine whether K14-expressing reparative progenitor airway epithelial cells within primary NSCLCs correlated with smoking injury, poor prognosis, and metastasis.
Materials and Methods
Human and mouse tissue
Sections were obtained from uninjured C57Bl/6 mouse tracheas as well as from C57Bl/6 mouse syngeneic tracheal transplants. We used a well-established, reproducible model of tracheal epithelial regeneration using syngeneic subcutaneous (s.c.) tracheal transplants from wild-type C57Bl/6 mice into wild-type C57Bl/6 mice (The Jackson Laboratory; refs. 20, 21). For this model, donor wild-type C57Bl/6 mice were euthanized and the tracheas were dissected out, removing the blood supply to the tracheas and causing a hypoxic-ischemic injury. Recipient wild-type C57Bl/6 mice were sedated with ketamine, and an incision was made in the skin of the back of the mice. The donor tracheas were placed heterotopically under the skin of the recipient mice. Mice were euthanized at 7, 14, and 21 days after transplantation, and the tracheal transplants were harvested for fixation in formalin and then paraffin embedding. Animal use for these studies was approved by the Department of Laboratory Animal Medicine, David Geffen School of Medicine at the University of California at Los Angeles (UCLA). Tissue sections were obtained from human lung cancer specimens archived in the UCLA Lung Cancer Specialized Programs of Research Excellence tissue bank (IRB#02-07-011). The research protocol was approved by the UCLA Institutional Review Board (IRB), and all human participants gave written informed consent.
Dual immunofluorescence and immunohistochemistry
Dual immunofluorescence was performed as described (18). Briefly, tracheal tissue was fixed in 4% paraformaldehyde for 18 to 24 hours, was embedded in paraffin, and was sectioned. Sections (4 μm) were deparaffinized in xylenes, rehydrated in graded ethanols, and boiled in 10 mmol/L sodium citrate buffer for 10 minutes. Blocking was performed with serum-free protein block (Dakocytomation). The primary antibodies used were rabbit anti-mouse K5 (dilution, 1:500; Abcam), mouse anti-K14 (dilution, 1:20; Abcam), and rabbit polyclonal anti–proliferating cell nuclear antigen (PCNA; dilution, 1:50; Abcam).
For calculating the proportion of proliferating (PCNA+), K5+, and K14+ cells in the epithelia, premalignant lesions, or tumors, tissue immunofluorescence images were obtained using a Zeiss AxioImager microscope (Carl Zeiss). For mouse samples, cross-sections through the same level of each trachea were selected for measurement. Cells were manually counted at ×20 magnification. Total K14+K5+ cells and K14-K5+ cells in an epithelium or lesion were counted to determine the percentage of K14+ cells within all the K5-expressing cells. K14+PCNA+ cells and K14+PCNA− cells in a premalignant lesion or tumor lesion were counted to determine the percentage of K14-expressing cells that were proliferating.
Immunohistochemical analysis of human lung tissue was performed as described (18) with the K5 and K14 antibodies described above. The lung TMAs were sectioned just before use, and serial sections were stained for K14 or K5 using a two-step immunohistochemical protocol.
Histologic definitions: Reserve cell hyperplasia was defined as a continuous and double layer of basal cells. Squamous metaplasia requires the development of horizontally oriented squamous cells with intercellular bridges. Dysplasia was diagnosed in the setting of epithelial thickening with nuclear pleomorphism and partial loss of normal maturation from the basal to luminal surface. Carcinoma in situ has marked nuclear pleomorphism and coarse chromatin with no maturation from basal to luminal surface and the absence of frank invasion.
Lung cancer tissue microarray
The tissue microarrays (TMA) were constructed under appropriate IRB and Health Insurance Portability and Accountability Act (HIPAA) regulations using formalin-fixed, paraffin-embedded archival lung samples from the UCLA Department of Pathology and Laboratory Medicine and the lung cancer Specialized Program of Research Excellence tissue bank at The University of Texas M.D. Anderson Cancer Center. The characteristics of these TMAs have been previously described in detail (19). The TMA was scored in a semiquantitative fashion by a pathologist (MA) and spot checked by a second pathologist (VM), both of whom were blinded to clinical and outcomes information. K5 and K14 cytoplasmic staining was quantified based on the intensity and frequency of cell staining, similar to previously described methods (19). A total of 399 patients from the UCLA TMA and 505 patients from the M.D. Anderson TMA were used in these studies.
Analyses were performed using the open source R software (http://www.R-project.org) including survival, Design, and Hmisc packages. Pooling criteria were similar to those previously described (19). K5 and K14 expression differences among various subgroups were determined using the Wilcoxon signed-rank test or Kruskal-Wallis rank sum test. For dichotomized (positive versus negative staining for K5 and K14) expression, the Fisher exact test was used for analysis with categorical variables such as stage, grade, smoking history, and presence of metastasis. Survival curves were calculated using the Kaplan-Meier method, and comparisons were made using the log-rank test. The Cox proportional hazards model (univariate and multivariate) was used to determine the significance of various factors related to survival. LogRank and Fisher exact P values were two sided, and a P < 0.05 was considered significant.
Methods used for the experiments found in the Supplementary Data are in the Supplementary Data Methods section.
Identification of K14+K5+, K14-K5+ cell populations in the steady-state airway epithelium and submucosal glands
Dual immunofluorescent staining of the steady-state proximal airway epithelium, submucosal glands, and submucosal gland ducts showed the presence of K14+K5+ cells throughout the submucosal gland myoepithelial cells and submucosal gland ducts. Dual K5 and K14 expression was found in only 10.7% ± 3.4% of basal cells of the mouse pseudostratified columnar epithelium (Supplementary Table S3; Fig. 1A) and 1.3% ± 0.8% of basal cells in the human pseudostratified columnar epithelium (Supplementary Table S3; Fig. 1B). K5+K14− cells comprised the remainder of the basal cells of the pseudostratified columnar epithelium.
K14+K5+ cells are reparative in the context of airway epithelial injury
We further examined the relative abundance and location of K14+K5+ cells in a model of airway epithelial injury. To do this, we performed heterotopic, syngeneic tracheal transplants in mice and examined the repairing tracheal airways for K14 and K5 expression after hypoxic ischemic injury (20, 21). We found K14+K5+ cells in the submucosal glands, submucosal gland ducts, as well as in cells on the basement membrane repairing the surface airway epithelium. These K14+K5+ cells persisted in the airway epithelium during all stages of repair and represented 85.6% ± 5.3% of all cells of the mouse repairing surface airway epithelium (Supplementary Table S1; Fig. 2Ai and ii). In the repaired pseudostratified columnar epithelium, only K5 expression was present in the basal cells (Fig. 2A, iii and iv).
K14+K5+ cells populate preneoplastic and neoplastic lesions
We further explored the expression of K14+K5+ cells in human disease representing chronic injury and repair after smoking. For this, we performed dual immunofluorescent staining of airway tissue from patients with chronic obstructive pulmonary disease. As in the mouse airway injury model, we observed a persistence of K14+K5+ cells in repairing areas of reserve cell hyperplasia (Fig. 2B). We also observed a predominance of K14+K5+ cells in potentially preneoplastic lesions represented by squamous metaplasia, dysplasia, and carcinoma in situ (Fig. 2B and C). We found K14+K5+ cells in all premalignant lesions from all patients examined to date, and in the premalignant lesions, K14+K5+ cells represented 75.3% ± 3.4% of cells in the lesions (Supplementary Table S1; Fig. 2B and C).
The presence of K14+ cells in NSCLC tumor samples confers a worse prognosis
Based on the overrepresentation of K5+K14+ cells in preneoplastic lesions, we further assessed whether the presence of K14 in primary NSCLC tumors was associated with lung cancer development and/or progression. To do this, we examined protein expression on a population basis using high-density lung TMAs. We first examined 399 patients from the UCLA TMA (237 adenocarcinoma, 19 adenosquamous, 100 squamous cell carcinoma, 7 neuroendocrine, 32 large cell carcinoma, and 4 other). Levels of K14 and K5 were found to be similar in all NSCLC with the notable exception of tumors with squamous differentiation. In squamous cell carcinoma, 90% of cells were K5 positive and 60% were K14 positive compared with 57% and 18% of cells in adenocarcinomas, respectively (positivity defined by 5% cut point; Supplementary Fig. S1). These results were verified by quantitative real-time PCR (Supplementary Fig. S2A), review of publicly available lung cancer microarray expression data sets (Supplementary Table S2; Supplementary Fig. S2B; refs. 22, 23), and Western blot analysis on frozen adenocarcinoma and squamous lung cancer samples (Supplementary Fig. S2C). The percentage of K14+ or K5+ cells in NSCLC did not correlate with stage, and although lower grade tumors tended to have somewhat higher percentages of K5+ and/or K14+ cells, a significant association was only seen for K14 in squamous carcinomas (data not shown). Tumor samples from male subjects had slightly higher percentages of both K5+ and K14+ cells than did samples from female subjects (Supplementary Table S3).
We further examined whether tumors expressing K14 represented a more aggressive substratum of tumors. Consistent with this, patients with NSCLC that expressed K14 were found to have a significantly worse prognosis than patients with NSCLC in which K14 was below the level of detection (P = 0.004, hazard ratio = 1.58; Fig. 3A). We also validated this TMA data with an independent TMA obtained from the M.D. Anderson Cancer Center. We found identical results to those found on the UCLA TMA: patients with K14-expressing tumors had a worse prognosis (P = 0.003, hazard ratio = 1.60; Fig. 3B).
The presence of K14+ cells in NSCLC tumor samples confers a worse prognosis in smokers and is associated with metastasis
It is generally accepted that cigarette smoking has a causal relationship with lung cancer. Smoking results in chronic airway epithelial injury, and dysfunctional repair is commonly seen (24). We hypothesized that the presence of K14+ reparative cells that predict poor prognosis might be associated with chronic smoking injury. Consistent with this hypothesis, we found a striking increase in the predictive value of K14-expressing NSCLC tumors in individuals who were current or former smokers (P = 0.001, hazard ratio = 1.77; Fig. 4A). In all smokers, K14 positivity (>5%) was an independent predictor of poor prognosis (P = 0.027) in a multivariate Cox proportional hazards model, which also included stage, grade, and age (Supplementary Table S4). When separating current from former smokers, the predictive value of K14 positivity was more pronounced in current smokers (P = 0.01, hazard ratio = 2.11; Fig. 4B). Current smokers were defined as those patients who were currently smoking or who had quit within 1 year of when their tissue sample was collected. However, K14 positivity was still predictive of poor prognosis in former smokers as well (P = 0.04; hazard ratio = 1.68; Fig. 4C). Former smokers were defined as those patients who had quit more than a year before their tissue sample was collected. This was true for individuals with either squamous cell carcinoma or adenocarcinomas. In never smokers, the presence of K14+ cells had no predictive value for outcome (Fig. 4D). Never smokers were defined as having smoked <100 cigarettes over their lifetime.
Validation of these results was performed on an independent TMA from the M.D. Anderson Cancer Center and smoking was again associated with poor prognosis (P = 0.004, hazard ratio = 1.59; Supplementary Fig. S4A), but again, there was no association between K14 expression and prognosis in nonsmokers (P = 0.356, hazard ratio = 2.51; Supplementary Fig. S4B).
We found that the presence of K14+ cells in the primary tumors of current smokers was associated with metastatic disease (P = 0.02; Table 1A). We further found that nonadenocarcinoma primary NSCLCs from smokers with metastases had a higher percentage of K14+ cells. (P = 0.004; Table 1B). Examination of K14 expression in distant metastatic sites revealed a significant increase in the number of K14+ cells in metastases compared with the primary sites in squamous lung cancer (P < 0.001), but not in other histologic subtypes (Supplementary Fig. S5).
K14 expression is not a marker of proliferation
We next assessed whether K14 expression was prognostic only because it might be a surrogate marker for cell proliferation. Therefore, to determine whether the poor prognosis in K14+ tumors was related to increased proliferation in these tumors, we performed dual immunostaining for K14 and PCNA to assess the percentage of K14+ cells that are also proliferating in premalignant lesions and NSCLC. In premalignant lesions, we found that 57.8% ± 5.1% of K14+ cells also expressed PCNA (Fig. 5i and ii). In squamous lung cancer patient samples, we found that 67.3% ± 7.3% of K14+ cells also expressed PCNA (Fig. 5iii and iv). We also found many other cell populations, which were K14 negative that expressed PCNA. There was also clearly a subpopulation of K14+ cells that were not proliferating (Fig. 5). K14+ cells are therefore not a unique marker of proliferating cells, as many other cell populations are proliferating in lung cancer. This is consistent with the point that K14 is a marker of poor prognosis but may not functionally be important for proliferation.
In addition, we performed K14 knockdown studies in BEAS2B-immortalized normal human bronchial epithelial cells. We used small interfering RNA (siRNA) technology to reduce expression of K14 in BEAS2B cells by 90% at 5 days posttransfection compared with control siRNA–transfected cells. The MTS proliferation assay showed no effect on cell proliferation in the K14 siRNA–transfected cells compared with the control siRNA–transfected cells, and there was also no effect on cell morphology. Similarly, PCNA expression in transfected cells was found to be equivalent by Western blot analysis in K14 siRNA and control siRNA–transfected cells (Supplementary Fig. S6).
Cigarette smoking causes cycles of injury and repair of the airway, and is a known cause of lung cancer (25). We, and others, have shown that K14+ progenitor cells are a reparative cell population, and contribute to repair of the epithelium of the cartilaginous airways and in the more distant bronchioles after injury, such as hypoxic ischemic injury, naphthalene injection, and sulfur dioxide inhalation (12, 16). Here, we propose that in the context of injury, K5+K14+ cells originate from the submucosal gland/duct K5+K14+ cells and/or from the K5+K14− basal cells that then acquire K14 expression on the repairing surface airway epithelium. However, once normal repair is completed, K14 expression is no longer seen in the mature basal cells of the pseudostratified columnar epithelium. This implies that K14 expression is tightly regulated at steady state, and the persistence of K5+K14+ cells on the surface airway epithelium after injury represents self-renewing cells that do not differentiate to mature airway epithelial cell types and represent dysregulated repair. Our data are, therefore, consistent with the development of dysregulated repair after injury leading to a self-renewing K14+ progenitor cell population in premalignant lesions. These cells could therefore potentially survive long enough to accumulate the genetic mutations and epigenetic changes that are thought to be necessary to develop a tumor (3). We found that the presence of dysregulated K14+ progenitor cells in NSCLC after chronic smoking injury was associated with increased mortality from lung cancer. This implies that there could be a novel putative tumor-initiating cell population in a subset of smoking-related NSCLCs with a poor prognosis.
A self-renewing tumor-initiating cell population associated with poor prognosis in human NSCLC has not yet been described. Kim and colleagues (26) isolated a putative lung stem cell termed the bronchoalveolar stem cell or BASC, which expressed markers of both Clara cells (CCSP) and type II pneumocytes (SP-C), proliferated for repair, were seen in the earliest cancerous lesions, and increased as the tumors advanced. However, these studies were performed in mice, and it is not clear what the equivalent human cell surface markers are that would enable the purification and propagation of these cells in xenograft models to determine whether these are truly CSCs in lung cancer patients. In addition, the heterogeneity of lung cancers suggests that there are likely to be multiple tumor-initiating cell populations for different lung cancer histologic subtypes and locations. K14+ cells have been found for repair in the distal bronchioles (16), and we found K14 mRNA and protein expression in adenocarcinomas as well as squamous cell cancers. In addition, K14 expression correlated with poor prognosis in all NSCLC histologic subtypes, although it only correlated with metastases in nonadenocarcinoma histologies.
Classic validation of a CSC tumor–initiating cell population involves reconstituting the human tumor in an immunodeficient mouse, followed by the indefinite serial xenotransplantation of these CSCs. Eramo and colleagues (17) found CD133 expression in small cell and non–small cell lung tumors. High numbers of CD133+epCAm+ cells isolated from fresh lung tumor specimens were capable of generating tumor xenografts upon s.c. injection. However, the self-renewal capacity of CD133+ cells was not evident and CD133 expression was found not to be prognostic in NSCLC, although it did correlate with the expression of chemotherapy-resistant genes (27). To show the tumor-initiating potential of K14+ cells in NSCLC, by current definitions, the development and serial transplantation of NSCLC in immunodeficient mice is required (28). However, no surface markers have as yet been identified to allow the isolation of live K14+ cells from tumors. It is therefore not currently possible to evaluate the tumor-initiating ability of K14+ cells in NSCLC in a serial xenotransplantation model. We are therefore not functionally able to test the tumor-initiating potential of K14+ cells.
Precursor lesions of squamous lung cancer are known to have high levels of K14 expression, from basal/reserve cell hyperplasia to squamous metaplasia, dysplasia to carcinoma in situ, as well as invasive carcinoma itself (29). Our data suggest that K14-expressing cells in the airway epithelium in premalignant lesions may represent self-renewing, reparative progenitor cells that may have the potential to be tumor-initiating cells. We also believe that K14 expression alone is not sufficient to generate a malignancy, and that subsequent genetic and epigenetic changes are needed to develop NSCLC. This is illustrated by work from Dakir and colleagues (29) who used a mouse Clara cell–specific 10-kDa protein promoter (CC10) to constitutively express human K14 in bronchial epithelium. The CC10-hK14–expressing transgenic mouse developed a squamous differentiation program in the mouse lung but failed to promote squamous maturation with rare squamous metaplastic lesions and squamous carcinomas in old age mice. This supports the idea that K14 expression in airway epithelial cells is a marker of a self-renewing progenitor cell and is a putative tumor-initiating cell, which requires genetic and/or epigenetic changes to be sufficient for carcinogenesis. Although we found no difference in the proliferative capacity of K14+ cells compared with K14- cells in premalignant lesions and in NSCLC, it is possible that the K14+ cells are an important subset of tumor cells as the keratin 14 cytoskeletal protein may allow for changes in cell shape and motility with an increased potential for cell migration.
In summary, the presence of K14+ cells in NSCLC is a biomarker of tumors with a worse prognosis. This was especially predictive in smokers, and furthermore, these patients had an increased likelihood of metastases. K14 expression in NSCLC in smokers may therefore be useful as a biomarker of poor prognosis and of metastases in squamous lung cancer. Furthermore, identifying the genetic and epigenetic changes that occur in the K14+ cell population that lead to dysregulated repair may result in the discovery of novel biomarkers and therapeutic targets for chemoprevention in smokers (30).
Disclosure of Potential Conflicts of Interest
A. Spira is consultant to and has equity in Allegro Diagnostics, Inc. S.M. Dubinett serves on the scientific advisory board of Tragara Pharmaceuticals. The other authors declared no potential conflicts of interest.
We thank Dr. Talal Chatila for critical review of the manuscript.
Grant Support: CIRM RN2-00904-1, K08 HL074229, American Thoracic Society/COPD Foundation ATS-06-065, The Concern Foundation, The UCLA Jonsson Comprehensive Cancer Center Thoracic Oncology Program/Lung Cancer Specialized Programs of Research Excellence, the Gwynne Hazen Cherry Memorial Laboratories (B.N. Gomperts), Early Detection Research Network NCI CA86366 (L. Goodglick and D. Chia), and Department of Defense W81XWH-04-1-0142 (I.I. Wistuba).
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.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
L. Goodglick and B.N. Gomperts, cosenior authors.
- Received February 4, 2010.
- Revision received June 7, 2010.
- Accepted June 14, 2010.
- ©2010 American Association for Cancer Research.