Are we Missing the Target? – Cancer Stem Cells and Drug Resistance in Non-small Cell Lung Cancer

Evidence for Stem Cells in the Normal Lung Epithelium

The lung epithelium consists of several different cell types whose composition varies with the anatomical region (Figure 3): The epithelium of the trachea and the proximal airways consists of ciliated cells and Clara-like cells producing secretoglobins, and moreover, a small number of neuroendocrine cells and mucus-producing submucosal glands containing ciliated and basal cells (7). In the more distal airways, Clara cells predominate over ciliated cells and there are more neuroendocrine cells, but no basal cells. The alveolar epithelium consists of pneumocytes type I and pneumocytes type II, producing surfactant factor.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

The lung epithelium. A: Tracheobronchial epithelium. The columnar tracheobronchial epithelium contains ciliated cells, goblet cells, basal cells and submucosal glands. Clara cells and neuroendocrine cells are not shown. Basal cells and a variant of Clara cells are considered stem/progenitor cells B: Bronchioalveolar duct junction (BADJ). As shown in rodent models, the junction between the cuboidal terminal bronchiole epithelium and the alveolar epithelium contains bronchioalveolar stem cells (BASC). C: Alveolar epithelium. The alveolar epithelium consists of pneumocytes type I and II and alveolar macrophages. Pneumocytes type II are able to regenerate and differentiate into pneumocytes type I. Microphotographs: Stain: hematoxylin-eosin, magnification: ×200, scale bar: 50 μm.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Driver mutations in lung adenocarcinoma. Mutation analysis of 422 patients with lung adenocarcinoma revealed driver mutations in 60% of the cases. K-ras (Kirsten rat sarcoma) and EGFR (epidermal growth factor receptor-1) mutations represent the most prominent alterations. Other mutations, ALK (anaplastic lymphoma kinase) rearrangements and MET (Met proto-oncogene) amplifications are present in <10% of the tumours. BRAF: B-Raf proto-oncogene serine/threonine-protein kinase, PIK3CA: phosphatidylinositol-3-kinase, catalytic, alpha polypeptide, HER2: epidermal growth factor receptor-2, MEK1: dual specificity mitogen-activated protein kinase kinase-1, NRAS: neuroblastoma RAS viral oncogene homolog. Adapted from M.G. Kris et al., J Clin Oncol 2011 (May 20 Supplement).

The first evidence for the presence of regenerating cells in the lung epithelium came from rodent injury models in the 1970s (7). Selective damage of specific cell types by inhaled oxidants such as NO₂ or ozone or intravenous administration of bleomycin revealed a regenerative potential of Clara cells and pneumocytes type II. In the nineties and 2000s, intraperitoneal injection of naphthalene showed that the regenerative potential of Clara cells is restricted to a specific subpopulation, so called variant Clara cells, which are lacking expression of cytochrome P450F2. Additional genetic labeling studies showed that cytokeratin 14-positive basal cells also possess regenerative potential in the proximal airways and give rise to basal cells, Clara cells and ciliated cells (10).

The distal lung epithelium appears to be regenerated by variant Clara cells and calcitonin gene-related peptide (CGRP)-expressing pulmonary neuroendocrine cells, both located in the neuroendocrine bodies (11). Moreover, in 2005 Bender Kim et al. showed that the junction between the cuboidal epithelium of the terminal bronchiole and the alveolar epithelium contains a population of stem cell antigen-1 (Sca-1)⁺, CD31^–, CD34⁺, CD-45^–, Clara cell antigen (CCA)⁺ cells which maintain the terminal bronchiolar epithelium (12). These so-called bronchioalveolar stem cells (BASC) gave rise to Clara-like, pneumocyte type II-like and peumocyte type I-like cells in vitro and showed expansion and differentiation towards the alveolar lineage upon induction of a Kirsten rat sarcoma (k-ras) mutation in vivo.

So far, it is unclear whether these stem and progenitor cells are organized in a hierarchical structure, as known from the hematopoietic system, or represent different types of region-specific stem cells. An analysis of Qualter et al. in epithelial cell adhesion molecule (EpCAM)⁺, Sca-1⁺, CD24^dim, CD31^–, CD45^–, CD49⁺, CD104⁺ murine lung epithelial cells provided first evidence for a stem cell hierarchy in the lung (13). These cells were able to self-renew in cytokine-supplemented, stroma-free cultures and generated colonies of airway, alveolar and mixed lung epithelial lineages upon co-culture with pulmonary mesenchymal cells.

Although the use of different stem and progenitor cell markers in the various studies hamper an uniform view so far, there is strong evidence that the lung contains various stem and progenitor cells that regenerate specific regions of the pulmonary epithelium and might be organized in a hierarchical structure (14). The characteristics of these putative stem and progenitor cells are summarized in Table I.

Evidence for Stem Cells in NSCLC

First evidence for the presence of stem cells in NSCLC came from a mouse strain with inducible k-ras G12D mutation developed by Bender Kim et al. (12). In 2005 this group demonstrated an increased expansion and reduced differentiation of BASC upon activation of k-ras resulting in an atypical adenomatous hyperplasia. Combination of k-ras activation and naphthalene treatment led to an increase in tumour number and areas, strongly indicating that BASC might represent the progenitor cells of pulmonary adenocarcinoma. Subsequent analyses of Xu et al. demonstrated that not only BASC, but also pneumocytes type II and Clara cells in the terminal bronchioles proliferate in response to k-ras activation. In this study, only pneumocytes type II progressed to adenocarcinoma (15).

As for human lung cancer cell lines and primary cells derived from lung cancer samples, Ho et al. demonstrated the presence of a Hoechst dye 33342 extruding subpopulation, the so-called side population (SP), which accounted for 0.03%-6.1% of the tumour cells (16). In contrast to non-SP cells, this small subpopulation was able to regenerate SP and non-SP cells and showed an increased tumorigenicity in mice. In line with their resistance to multiple cytostatic drugs and enhanced proliferation capacity, these cells showed an elevated expression of various ATP-binding cassette transporters and human telomerase reverse transcriptase (hTERT).

Eramo et al. used the common stem cell antigen CD133 to prospectively isolate human lung cancer cells with stem cell properties. These cells were able to form tumour spheres, showed co-expression of the ATP-binding cassette transporter breast cancer resistance protein-1 (BCRP1) and the ESC transcription factors octamer binding protein (Oct) 3/4 and Nanog. Moreover, these cells were highly tumorigenic in mice, able to regenerate the original tumour histology and resistant to various cytostatic drugs (17). Chen et al. reported very similar results for CD133-positive NSCLC cells and moreover showed that these cells lost their stem cell characteristics when the expression of Oct4 was knocked down by siRNA (18).

Levina et al. who used a reverse approach to isolate lung cancer stem cells also reported corresponding findings (19). Exposure of the NSCLC cell line H460 to cytostatic drugs enriched a Hoechst dye 33342-extruding SP. The antigen profile and functional properties of this SP was basically consistent with the stem cell population described by Eramo et al. and Chen et al. In addition to previous analyses, Levina et al. demonstrated that these cells had an activated β-Catenin pathway known to be involved in stem cell self-renewal, and moreover, that tumours originating from these cells show a high expression of angiogenic cytokines and growth factors.

Bertolini et al. confirmed an enrichment of CSC after cytostatic treatment in vivo (20). When mice carrying lung cancer xenografts were treated with cisplatin, the tumours showed an enrichment of epithelial cells expressing CD133 and BCRP1. These cells shared several functional and molecular properties with the CSC described in previous studies including an extraordinary high tumorigenicity in mice and an increased expression of the transcription factors Oct3/4 and Nanog, adhesion and motility genes and various ATP-binding cassette transporters.

As for human pulmonary CSC, CD133 and BCRP1 appear to be the most appropriate markers so far. However, a recent study of Curtis et al. demonstrated that the phenotype of lung cancer propagating cells varies with the type of genetic alteration present in the tumour (21). Since recent analyses in lung adenocarcinoma revealed the presence of driver mutations in up to 87% of all cases, several phenotypes of pulmonary CSC might exist. This suggests that rather functional properties than surface antigens represent appropriate targets for specific therapies against CSC (22, 23).

Predictive and Prognostic Significance of CSC in Lung Cancer

In recent years, several immunohistochemical and a few microarray studies analyzed the prognostic and predictive significance of CSC antigens and gene signatures in NSCLC.

As for the most common human pulmonary CSC antigens CD133 and BCRP1, expression frequencies between 15%-have been 63% and 15%-51%, respectively, were reported (24-29). While our group and Salnikov et al. found no prognostic significance of CD133 and BCRP1 expression in 133 stage I/II and 88 stage I-III patients who underwent complete pulmonary resection, two Asian studies in 145 stage I NSCLCs and 177 stage I AC patients found an up to 9.3-fold increased risk of postoperative disease recurrence in CD133/BCRP1 and CD133-positive cases, respectively (25, 28-30).

In patients with locally-advanced or metastatic disease who received platinum-based therapy studies of Yoh et al. and Ota et al. demonstrated a predictive and prognostic significance of BCRP1 (26, 27): while Yoh et al. reported a higher response rate (44% vs. 24%, p=0.08) of patients with BCRP1 negative tumours and a 2.8-fold (p=0.001) and 1.8-fold (p=0.08) reduced risk of relapse and disease-related mortality, respectively, Ota et al. found no association between BCRP1 expression and response to therapy or progression-free survival (PFS), but an association with a significantly reduced median survival time (7.8 vs. 14.9 months, p=0.03) in BCRP1-positive cases.

Also in neoadjuvant treatment, a recent study in 50 patients who received preoperative chemoradiotherapy, demonstrated that the 5-year overall survival (OS) was worse in patients with CD133 or ALDH1 expression (45% vs. 90%, p=0.042) (31).

Authors who used other stem cell markers such as the basal cell antigen cytokeratin 14 or the ESC transcription factor Oct4, found a 1.6-fold increased risk of disease-related mortality in two large datasets of 399 (p=0.003) and 505 stage I-IV patients (p=0.004) and a 2.2-fold increased risk of disease-related mortality in a collective of 112 operated stage IB-IIIA AC patients (p=0.015), respectively (32, 33).

Onaitis et al. applied a 10-gene progenitor cell signature derived from a transgenic mouse model to a dataset of resected human AC patients and found a 3.5-fold increased risk of disease-related mortality (p=0.0004) when this gene signature was expressed (34).

All in all, these studies provide strong evidence that the presence of tumour cells expressing antigens or gene signatures associated with stem cells might play a prognostic and predictive role in NSCLC.

Significance of Drug Resistance in Current Lung Cancer Therapy

In recent years, increasing efforts were undertaken to identify therapy responders in advance. The presence of specific targets and the absence of resistance mechanisms represent the two pillars of tailored therapy.

Overexpression of key enzymes of the folate synthesis, particularly thymidylate synthase (TS), and components of the mitotic spindle apparatus such as tubulin beta III (TUBB3) show a high level of evidence for an association with response to folate antagonist such as pemetrexed and microtubule-inhibiting taxanes, respectively (35-37).

Particularly targeted therapies with tyrosine kinase inhibitors (TKI) such as erlotinib, gefitinib and crizotinib rendered impressive results when response predictors were considered. While conventional chemotherapy shows response rates of maximal 30% in the first-line treatment and about 10% in subsequent therapy lines, treatment with the epidermal growth factor receptor (EGFR)-TKI erlotinib or gefitinib or the echinoderm microtubule-associated protein-like 4-anaplastic lymphoma kinase (EML4-ALK)-TKI crizotinib in patients with activating EGFR mutations or EML4-ALK rearrangements shows response rates of ≥70% and 57%, respectively (38, 39).

Also DNA repair proteins such as excision repair cross-complementation group 1 (ERCC1), breast cancer 1, early onset (BRCA1) and ribonucleotid reductase 1 (RRM1) gained growing significance for tailoring cisplatin or gemcitabine-containing therapy regimens. Operated patients who received cisplatin-based, adjuvant treatment showed a significantly longer survival, when their tumours were ERCC1-negative. Also in advanced disease, lower ERCC1 mRNA levels were highly predictive for longer survival after cisplatin-based treatment (40-42). In metastatic disease, low BRCA1 mRNA levels predicted a higher response rate, longer PFS and median OS towards platinum-containing second-line therapies (43). As for gemcitabine, studies of Bepler et al. demonstrated that patients with low RRM1 mRNA levels experienced better response and longer PFS and OS to this antimetabolite (44, 45). However, none of these candidate biomarkers have been established to date since large, prospective trials are just getting under way.

So far, the association between these putative biomarkers and CSC is widely unknown. The findings of Ota et al. who demonstrated a correlation between protein expression of ERCC1 and the stem cell antigen BCRP1 (p=0.012) and the results of Salnikov et al. who demonstrated a correlation between protein expression of TS and the stem cell antigen CD133 (p<0.01), indicate that tumour cells with predominant expression of these enzymes might bear stem cell characteristics (27, 29).

Mechanisms of Drug Resistance in CSC

Hitherto, little heed has been taken of the specific drug resistance mechanisms of CSC in antineoplastic treatment. The ability of asymmetric cell division, the slow division kinetics or even quiescence and the expression of several drug efflux pumps and DNA repair proteins represent the major escape mechanisms of stem cells (46-48).

In 2010 Pine et al. demonstrated for the first time that a small subpopulation of lung cancer cells which were present in commercially available cell lines and primary tumour cell cultures was able to undergo asymmetric cell division: While the differentiation markers prosurfactant protein C and pan-cytokeratins were passed to one daughter cell, the stem cell marker CD133 was passed to the other and exclusively characterized cells with tumour repopulating capability (49). A subsequent study of Dey-Guha et al. in breast cancer cell lines showed that asymmetric division generated slowly cycling G₀-like cells which were enriched in tumour tissue after neoadjuvant therapy (50). The presence of such a quiescent and drug-resistant subpopulation has also been shown in lung cancer (51). Several pathways and molecules such as v-akt murine thymoma viral oncogene homolog 1 (AKT1), insulin-like growth factor receptor-1 (IGF-1R), protein kinase C (PKC), aurora kinase A (AURAKA), Notch, phosphatidylinositol 3-kinase (PI3K), p53 or phosphatase and tensin homolog deleted on chromosome 10 (PTEN) were found to be involved in the regulation of division symmetry and quiescence of stem cells (50-53). As most classical cytostatics require cell division for being effective, quiescence or slow division kinetics protects tumour cells against lethal DNA damage. Moreover, novel molecular drugs such as AKT-TKI might rather support than suppress CSC as they favor asymmetric cell division and the generation of slowly-cycling, drug-resistant tumour cells (50, 54).

Besides division symmetry and slow division kinetics, CSC show an increased expression of DNA repair proteins and drug efflux transporters. Salnikov et al. and Ota et al. demonstrated that overexpression of TS and ERCC1 which might predict resistance against pemetrexed and cisplatin, respectively, correlated with the expression of the stem cell antigens CD133 and BCRP1 (27, 29). Moreover, several other proteins involved in detoxification and drug resistance such as glutathione-S-transferase, the drug transporter ABCC2 (multidrug resistance 1 [MDR1], p-glycoprotein [p-gp]), metallothionein, catalase or O⁶-methylguanine-DNA methyltransferase showed an overexpression in CD133-positive cells. Ho et al. reported an increased expression of the drug transporters ABCG2 (BCRP1), ABCA2, MDR1 and ABCC1 (multidrug resistance protein-1 [MRP1]) in the SP of various NSCLC cell lines (16). In other tumour entities, glutathione-S-transferase and MRP1 were shown to confer resistance to gemcitabine, while MDR1 was associated with resistance to paclitaxel and BCRP1 with resistance to the EGFR-TKI gefitinib (55-57).

All in all, CSC possess several mechanisms to overcome irreversible damage by cytostatic drugs. These mechanisms have neither been studied sufficiently nor have they been considered in cancer therapy so far.

Potential Targets and Approaches for CSC-directed Therapy

The ability of asymmetric cell division and the generation of slowly cycling or quiescent progenies represent important escape mechanim of stem cells. Thus, pivotal molecules and mechanisms involved in the regulation of division symmetry might provide one of the Achilles’ heels to target these cells in NSCLC.

The Notch, Wnt and Hedgehog pathway represent key mechanisms of stem cell regulation (58). In 2008 Tsao et al. demonstrated that disruption of Notch signaling by inhibition of gamma-secretase cleavage of Notch receptors dramatically expanded the population of distal pulmonary progenitors and prevented formation of proximal structures in murine embryonic lungs (59). Two years later, Sullivan et al. showed that the proliferation and clonogenicity of self-renewing ALDH-positive CSC in primary human NSCLC cell lines was significantly decreased by treatment with either a γ-secretase inhibitor or stable expression of shRNA against Notch3 (60). A large study of Westhoff et al. in tumour tissue of more than 400 NSCLC patients demonstrated an activated Notch pathway in about one third of the cases and a significantly worse survival of these patients when concomitant p53 mutations were lacking (61). Altogether, these studies indicate not only a pivotal pathogenetic role, but also a prognostic relevance of Notch signaling in NSCLC.

Wnt signaling represents another pathway that is critically involved in the regulation of CSC (62). Levina et al. demonstrated that NSCLC cells surviving cytostatic treatment show stem cell properties and an activated Wnt pathway (18). Similar findings were reported by Teng et al. in cisplatin-selected A549 NSCLC cells: Activation of Wnt signaling led to a significantly enhanced proliferation, colony formation, migration and drug resistance (63). In a murine model, Zhang et al. demonstrated that Gata6-regulated Wnt signaling controls the balance between progenitor cell expansion and epithelial differentiation required for both, lung development and regeneration (64).

As for Hedgehog signaling, recent studies of Shi et al. and Tian et al. in a pulmonary LCC and an AC cell line showed that the proliferation of the SP was significantly impaired by inhibition of the Hedgehog pathway with Cyclopamine or Vismodegib (GDC-0449), respectively (65, 66).

Another molecule which has repeatedly been reported to be involved in the regulation of CSC in NSCLC is the tumour suppressor gene PTEN. In a murine model with constitutive loss of PTEN, epithelial cells residing in the bronchioalveolar duct junction underwent proliferation and formed uniform masses (67). Similar results were derived from a previous study analyzing a murine model with bronchioalveolar epithelium-specific null mutation of PTEN: The numbers of BASC, the putative progenitors of pulmonary adenocarcinoma, were likewise increased (68). In addition, PTEN is an inhibitor of the PI3K/AKT/mTOR (mammalian target of rapamycin) signal transduction axis and loss of function leads to an activation of these molecules which have all been shown to be involved in the maintenance of CSC (50, 52, 69).

Currently, several phase I and II studies testing inhibitors of the Wnt, Notch and Hedgehog pathway as well as the PI3K/AKT/mTOR signal transduction axis are in progress (Table II). However, it remains to be elucidated whether targeting these pathways and molecules will lead to an increased therapeutic efficacy or even have an opposite effect in lung cancer (54).

As for drug efflux transporters, pharmacodynamic studies with the MDR1 inhibitors CBT-1(R) and tariquidar showed an increased uptake of the radionuclid ^99mTc-sestamibi, a substrate of MDR1 (70-72). However, two phase III studies in stage IIIB/IV NSCLC combining tariquidar with first line single-agent vinorelbine or a carboplatin/paclitaxel doublet showed no advantage and had to be terminated prematurely (ClinicalTrials.gov ID NCT00042315, NCT00042302). Recent in vitro and in vivo studies analyzing the effect of TKI such as axitinib, gefitinib, vandetanib, pelitinib or crizotinib on the function of ABCG2 (BCRP1) and ABCB1 (MDR1) demonstrated an inhibition of these drug efflux transporters and an enhanced efficacy of cytostatic drugs in multidrug-resistant cell lines and murine xenograft models (51, 73, 74). These TKI might be worthwhile combination partners for specific cytostatics such as anthracyclines, Vinca alkaloids, taxanes or epipodophyllotoxins in subsequent clinical trials.

A recent analysis of Bartucci et al. in CSC derived from primary NSCLC cultures demonstrated a strong activation of the DNA damage checkpoint protein kinase 1 (Chk1) upon exposure to cisplatin, gemcitabine or paclitaxel (75). In contrast, differentiated, chemosensitive NSCLC cells showed only a weak activation of Chk1. When the Chk1 inhibitor AZD7762 and chemotherapy were co-administered tumour growth in vivo was suppressed, whereas chemotherapy alone was scarcely effective. Thus, AZD7761 might represent an interesting agent for future clinical trials directing CSC in NSCLC.

Poly [ADP-ribose] polymerase 1 (PARP1), telomerase and mitogen-activated protein kinases (MAPK) respresent other promising targets for the eradication of CSC. In patients with squamous NSCLC, low expression of the DNA repair protein PARP1 showed a significant association with an increased chemotherapeutic efficacy (76). Although analyses on PARP1 expression in pulmonary CSC are so far lacking, recent data on the pivotal significance of this enzyme in the regulation of pluripotency and differentiation of murine ESC suggest a general role in stem cells (77). As for human telomerase, Ho et al. demonstrated an increased expression of hTERT in the SP of various NSCLC cell lines and primary NSCLC cells corresponding to the enhanced proliferation capacity of this subpopulation (19, 78, 79). Also MAPK is involved in the regulation of pulmonary stem cell proliferation (80). However, the efficacy of MAPK inhibitors might depend on the type of MAPK that is blocked: While inhibition of rat fibrosarcoma/mitogen-activated protein kinase kinase 1/2 (Raf/ERK1/2)-mediated downstream signaling might impair the self-renewal of NSCLC cells, blocking of the p38 MAPKα might have the opposite effect: In a murine model of Ventura et al., inhibition of p38 MAPKα generated an immature and hyperproliferative lung epithelium which was highly susceptible to k-ras-induced tumorigenesis (80-82).

Presently, several early phase studies analyzing the efficacy of MAPK inhibitors in lung cancer and various phase II and III trials with the PARP1 inihibitor Iniparib and agents against telomerase are in progress (Table II). High throughput screening methods such as those developed by Gupta et al. for breast cancer stem cells, might represent an effective approach for the identification of novel compounds against CSC (83).

All in all, consideration of CSC and their specific mechanisms of drug resistance might pave novel ways for more effective treatment and long-lasting therapeutic success in lung cancer.

Future Perspectives

Although analyses in lung cancer and various other tumour entities provide strong evidence for a pivotal role of CSC in drug resistance and relapse, several questions remain to be answered.

So far, various markers and methods have been used to identify and isolate cancer cells with self-renewing and multilineage differentiation capability. However, comparative functional and molecular analyses of these differentially isolated cells are so far lacking. It is unclear, whether they represent the same cell population, a hierarchy of stem and progenitor cells or different cell types which share some common functional and molecular properties. Moreover, the origin and development of these cells is not yet understood: Do CSC derive from a neoplastic stem cell or a neoplastic tumour cell which acquired stem cell charcteristics? Does their development require specific mutations that occur in an orchtestrated manner or does it represent a clonal selection of randomly mutated cells by microenvironmental cues? The identification of specific markers and gene signatures of various differentiation stages might represent a first step towards better understanding of pulmonary stem cells. However, established differentiation assays are the prerequisite for such analyses and are yet lacking.

Moreover, several fundamental results on lung stem cells derive from murine models. So far, it is uncertain, whether these findings can be transferred onto human conditions. The common murine stem cell marker Sca-1, for instance, has no human homologue. In addition, the life expectancy of mice and men is very different and might affect the proliferation capacities of murine and human stem cells. As for transplantation studies, the growth behaviour of human xenografts strongly varies under the murine background and thus, results from in vivo studies might in a large part reflect the experimental conditions and less the genuine behaviour of the human tumour cells.

The development of novel imaging techniques to identify and track stem cells in their natural niche might help to better understand their significance in drug resistance, tumour regeneration and metastasis (84-86). However, these methods are still in early stages of development.

Altogether, several in vitro and in vivo studies provide strong evidence for the presence of drug-resistant and highly reg-nerative tumour cells in NSCLC. Eradication of these cells might contribute to better response and prolonged survival in lung cancer.