Cancer stem cells in lung cancer: Evidence and controversies

Abstract

The cancer stem cell (CSC) model is based on a myriad of experimental and clinical observations suggesting that the malignant phenotype is sustained by a subset of cells characterized by the capacity for self-renewal, differentiation and innate resistance to chemotherapy and radiation. CSC may be responsible for disease recurrence after definitive therapy and may therefore be functionally synonymous with minimal residual disease. Similar to other solid tumours, several putative surface markers for lung CSC have been identified, including CD133 and CD44. In addition, expression and/or activity of the cytoplasmic enzyme aldehyde dehydrogenase ALDH and capacity of cells to exclude membrane permeable dyes (known as the ‘side population’) correlate with stem-like function in vitro and in vivo. Embryonic stem cell pathways such as Hedgehog, Notch and WNT may also be active in lung cancers stem cells and therefore may be therapeutically targetable for maintenance therapy in patients achieving a complete response to surgery, radiotherapy or chemo-therapy. This paper will review the evidence regarding the existence and function of lung CSC in the context of the experimental and clinical evidence and discuss some ongoing controversies regarding this model.

INTRODUCTION

Lung cancer is the most common cause of cancer-related deaths worldwide.¹ It is a heterogeneous disease with two distinct pathological classes: non-small cell lung cancer (NSCLC, which makes up 80% of cases) and small cell lung cancer (SCLC, making up 20% of all lung cancers). The most common forms of NSCLC are adenocarcinoma (30–50% of NSCLC) and squamous cell carcinoma (30% of NSCLC). Future treatment is geared towards adapting management specific to the tumour gene profile. However, recent next generation sequencing studies show a marked degree of genomic heterogeneity within each histo-logical subtype, thus presenting ever greater challenges for personalized medicine based on targeted therapies.²

Although approximately 20% of NSCLC are operable at presentation, recurrence rates remain high at 30–50%.³ Locally advanced NSCLC can be treated with radical chemo-radiotherapy with intent to cure; however, overall 5-year survival rates remain low at around 7–20%.⁴ For most cases of advanced NSCLC, systemic chemotherapy response rates are below 20–50%, and the 5-year mortality is in the order of 95%.⁵ In a subset of patients with advanced adenocarcinoma, dramatic responses to tyrosine kinase inhibitors can be seen in tumours with activating mutations in epithelial growth factor receptor or analplastic lymphoma kinase. However, almost all patients succumb to drug-resistant recurrences within 2–3 years.^6,7 Although similar targeted agents have not been identified for use in SCLC, most cases are highly responsive to platinum-based chemotherapy, with or without radiotherapy in disease limited to the chest.⁸ Once again, despite complete responses in some, drug-resistant relapses result in a 5-year survival of only 5%.⁹

The remarkable ability of lung cancer to recur despite definitive local and/or systemic therapy suggests that minimal residual disease contains a population with enormous capacity for self-renewal and regeneration, a biological function generally limited to normal somatic stem cells. Termed ‘cancer stem cells’ (CSC), these cells are believed to be a phenotypically distinct population that possesses tumourigenic potential. Initially reported in haematological malignancies, and subsequently in several solid tumours, CSC represent potential therapeutic targets, and several strategies are already in clinical development. If the CSC hypothesis is correct, then their therapeutic targeting has the potential to delay or prevent disease recurrence. For the purposes of this review, we will not extend the definition of CSC to the question of cancer cell-of-origin, but rather focus on the question of stem-like cells in established lung cancers.

CANCER STEM CELLS

All complex multicellular organisms develop and regenerate from a somatic stem cell population that gives rise to a hierarchy of committed progenitors, leading ultimately to terminally differentiated cells in mature organs. The CSC hypothesis is based on the simple concept that cancers, like all other organs, contain a similar hierarchy with respect to self-renewal, differentiation and innate drug resistance.¹⁰ Thus, CSC give rise to highly proliferative progenitor cells and differentiated cells comprising the bulk of tumours and ultimately define the histological type of the cancer. Importantly, this hypothesis also predicts that, although these proliferating cells are responsive to treatment, their partially differentiated state prevents them from renewing the entire tumour from a minimal residual population. Numerous reports in haematological malignancies,¹¹ breast cancer,¹² brain tumours¹³ and colorectal cancer¹⁴ support this notion, and show that in each of these models, only a small, phenotypically distinct subpopulation of cells could recapitulate the tumour phenotype in vitro.

Isolation of CSC

The capacity of CSC to undergo self-renewal and differentiation has been demonstrated experimentally in various assays. Purified cancer cells isolated from freshly dissected tissues are able to grow as spheres in non-adherent cultures, form colonies more efficiently than their differentiated progeny and can be continually repassaged at clonal density.¹⁵ However, the current gold standard assay for CSC tumourigenic potential is their ability to grow in vivo as serially transplantable tumours in immunodeficient hosts. This strategy has successfully identified CSC in several tumour types, such as brain, breast, haemato-logical malignancies^12,13,16 as well as in lung cancer (Table 1).

Table 1.

Experimental studies performed to study putative cancer stem cell markers in lung cancer

Study	Sample	Marker to isolate stem cells	Main results
Ho et al.11	Cell lines	SP	SP cells showed high invasiveness and resistance to chemotherapy and high telomerase activity. In vivo high tumourigenicity compared with non-SP cells
Levina et al.18	Cell lines	SP	Enrichment of cells with SP and CD133 after chemotherapy treatment. Drug surviving cells were enriched in self-renewal, differentiation and tumourigenic capacity.
Chen et al.19	Cell lines/tumour specimens	CD133	CD133-positive cells formed spheres, expressed ABCG2 and Oct-4 CD133+ cells possessed self renewal capacity and were resistant to chemotherapy and radiotherapy
Eramo et al.20	Tumour specimens	CD133	CD133-positive cells showed sphere formation, differentiation and chemoresistance CD133-positive cells expressed Oct-4, NANOG, EPCAM and NCAM
Bertolini et al.21	Cell lines/tumour specimens	CD133	CD133-positive cells expressed Oct-4, NANOG, α-6 integrin and CXCR4 CD133-positive cells were resistance to cisplatin and in patient's samples predicted poor response to treatment.
Jiang et al.22	Lung cancer/cell lines	ALDH	ALDH-positive cells had increased growth and formed large colonies in culture. ALDH-positive cells showed differentiation and invasion potential. ALDH-positive patients had poor survival 60% patient samples were also positive for CD133
Shi et al.23	Cell lines	SP	SP cell formed spheres and expressed ABCG2 and Hedgehog pathway transcript (SMO). Cyclopamine inhibited SP cell proliferation

ABCG2, ATP-binding cassette G2; ALDH, aldehyde dehydrogenase; SMO, Smoothened; SP, side population.

However, this model has been criticized at both technical and theoretical levels. For example, the use of profoundly immunodeficient non-obese diabetic/severe combined immune deficiency/interleukin-2 knockout mice as a host for human melanoma xenografts shows that more than one in four cells can clonally generate tumours, regardless of surface marker expression.²⁴ However, when Matsui and colleagues repeated these experiments in other solid tumour models, including lung cancer, the distinct cellular phenotype of tumour-initiating cells was preserved.²⁵ These studies suggest that in some highly aggressive tumours, such as melanoma, the stem cell phenotype is a defining feature of the tumour, whereas more differentiated tumours maintain a stem cell hierarchy despite the level of immunodeficiency in xenograft assays.

Phenotypic characterization of CSC

Marker expression is a key element to categorize cancer cells into tumourigenic and nontumourigenic. Cells sorted for expression of CSC markers are passaged through various assays to determine tumourigenicity. One such method is the detection of side-population (SP) phenotypes. SP phenotyping is a functional assay based on the differential ability of the cancer cells to efflux Hoechst 33342 dye, as imparted by the ATP-binding cassette family of transporter proteins present on the cellular membrane.²⁶ The SP assay has the advantage of measuring a functional parameter of the cells. However, the process is difficult to perform on tumour cells from intact clinical samples. Cancer cells as well as many stromal cells may possess dye exclusion properties, which could make the interpretation of results difficult. Moreover, the Hoechst dye is toxic to the cells.²⁷

Aldehyde dehydrogenase (ALDH) activity is an important functional marker of normal and malignant stem/progenitor cells. Through oxidation of retinol to retinoic acid, ALDH is involved in early stem cell development.²⁸ ALDH activity forms the basis of a fluorescence-activated cell sorter-based assay. Initially used to sort haematopoietic stem cells,²⁹ a BIODIPY-aminoacetaldehyde substrate is oxidized intracellularly by ALDH, causing cells to become highly fluorescent. ALDH contributes to drug resistance through detoxification of many cytotoxic agents³⁰ and has been reported as a reliable CSC marker in several tumour types.^31,32

Another commonly used strategy is the isolation of CSC by flow cytometry according to the expression of certain surface markers. Commonly used markers include CD133 and CD44. CD133 (prominin-1 or AC133) was originally described in human haematopoietic stem cells³³ and has subsequently been used to isolate CSC in many tumour types.^14,34,35 CD44 is another transmembrane glycoprotein believed to be activated in a wide range of tumours where it plays a critical role in a variety of cancer cell behaviours including adhesion, migration, invasion and survival.³⁶

LUNG CSC MARKERS

Almost 30 years ago, Carney and colleagues described a rare (<1.5%) population of cells in SCLC and adeno-carcinoma patient samples, with the ability to form colonies in soft agar.³⁷ When inoculated into athymic nude mice, these colonies were able to recapitulate the original SCLC and adenocarcinoma tumours, suggesting ‘stem cell’ characteristic.³⁷ Over the last decade, several investigators have isolated tumourigenic cells from lung cancer using different phenotypic cancer cell characteristics (Table 1).

Ho et al. initially showed that SP cells in NSCLC cancer cell lines were resistant to multiple chemotherapy agents and overexpressed several ATP-binding cassette transporters.¹⁷ SP cells also expressed elevated levels of telomerase and showed decreased expression of proliferation marker mini-chromosome maintenance 7, suggesting a quiescent nature.¹⁷ SP cells were rich in self-renewal and differentiation potential and were highly tumourigenic in non-obese diabetic/severe combined immune deficiency mice. Subsequently, different researchers investigated several membrane-bound surface markers to identify CSC in lung cancer. Of all the markers explored, CD44 and CD133 gained the most attention.^20,38,39 In cell lines from both SCLC and NSCLC, CD133-positive cells generated long-term tumour spheres and differentiated into CD133-negative cells.²⁰In vivo, as few as 1 × 10⁴ CD133-positive cells generated tumours in non-obese diabetic/severe combined immune deficiency mice. CD133-positive cells were also resistant to chemo-therapy and expressed high levels of ATP-binding cassette G2, suggesting an SP phenotype. CD133-positive cells also expressed embryonic stem cell markers (Oct-4 and NANOG),²⁰ and knockdown of Oct-4 by small interfering RNA resulted in reduced clonogenicity and enhanced chemosensitivity of CD133-positive cells.¹⁹ Further, treatment of a NSCLC cell line with cisplatin and doxorubicin resulted in enrichment for cells with enhanced tumourigenic potential in severe combined immune deficiency mice. These cells were also enriched for CD133 expression as well as SP phenotype.¹⁸ CD133-positive cells also overexpressed adhesion molecules (very late activation antigen-5, intercellular adhesion molecule-1, vascular cell adhesion molecule-1, matrix metalloprotease-2 and matrix metalloprotease-3), suggesting enhanced metastatic potential.¹⁹ CD44 expression on the other hand enriched for CSC-like properties in NSCLC cell lines³⁸ but not in SCLC cell lines.³⁹ However, in SCLC cell lines, urokinase-type plasminogen activator positive (CD87) cells were shown to possess multidrug resistance and enhanced clonogenic activity in vitro.⁴⁰ Another recent study in SCLC cell lines showed that urokinase-type plasminogen activator positive cells are also tumourigenic in athymic nude mice.³⁹

ALDH activity has been found to correlate with stem cell characteristics in lung cancer cell lines. In a study by Jiang et al., ALDH1-positive cells from NSCLC had extensive self-renewal and proliferative potential as well as in vivo tumourigenic potential.²² ALDH1-positive cells also expressed CD133 and showed resistance to common chemotherapy agents. Moreover, high ALDH1 protein expression, unlike other markers, was associated with poor prognosis in early stage NSCLC.^22,41 Further, ALDH1-positive cells expressed Notch pathway transcripts, and inhibition of Notch with γ-secretase led to the reduction of ALDH1-positive tumour cells. Additionally, direct inhibition of ALDH protein by small interfering RNA resulted in reduced proliferation and migration in these cells.⁴²

Studies performed to explore prognostic significance of these markers in NSCLC are summarized in Table 2. These findings suggest the existence of a small tumourigenic population of cells in lung cancer, favouring a hierarchical model. Further work is needed to identify more robust markers to define a distinct population of lung CSC.

Table 2.

Clinical studies conducted in NSCLC to study prognostic significance of putative CSC marker

Study	Marker	Method	Population	Number of patients	Cut-off point†	Number positive (%)	Main results
Jiang et al.22	ALDH	IHC	Stage 1 NSCLC	148	>10%	29	ALDH-positive patients have worse overall survival
Sullivan et al.41	ALDH1A1 ALDH1A3 CD133	IHC	Stage 1 NSCLC	200	NM	NM	ALDH1A1 but not CD133 and ALDH1A3 are associated with poor clinical outcome
Woo et al.43	CD133	IHC	Stage 1 NSCLC	177	17.5%	45	CD133-positive patients have shorter disease-free survival
Salnikov et al.44	CD133	IHC	NSCLC all stages	88	Any expression	63	No difference between CD133-positive and negative patients
Li et al.45	ALDH1A1	IHC	NSCLC all stages	179	Any expression	45	ALDH1A1 + patients had shorter survival
Dimou et al.46	ALDH	IF	NSCLC all stages	430	—	—	ALDH-positive patients have better prognosis

Variations exist in terms of methodology, patient selection, cut-off point determination as well as in results.

^†For statistical analysis, patient population is dichotomized into positive and negative cohorts based on appropriate cut-off point on immunohistochemical scoring.

ALDH, aldehyde dehydrogenase; CSC, cancer stem cell; IF, immunofluorescence; IHC, immunohistochemistry; NSCLC, non-small cell lung cancer; —, not reported.

EMBRYONIC STEM CELL PATHWAYS IN LUNG CANCER

Embryonic signalling pathways are fundamental in organogenesis. These pathways include Hedgehog (Hh), Notch, WNT and B lymphoma MLV insertion region 1, together with growth signals from the stem cell niche. All are involved in the renewal of normal stem cells and in maintaining tissue homeostasis.⁴⁷ Dysregulation of these pathways is believed to be involved in driving CSC activity in a variety of cancers, including lung cancer.⁴⁸ Therapeutic strategies to target these pathways are summarized in Table 3.

Table 3.

Summary of various agents in clinical development to target embryonic stem cell pathways

Target	Agent	Population	Results	Phase of development	Reference
Hedgehog (SMO)	GDC-0449	Lung cancer cell lines	Inhibition of SP+ cells in cell lines	Phase II	Tian et al.49
Notch (γ-secretase)	RO4929097	Advanced NSCLC	Safety demonstrated in phase I study Encouraging activity in some cancers	Phase II	Richter et al.63(ASCO abstract)
Hedgehog (SMO)	LDE225		Exposure dependent GLI-1 inhibition and disease stabilization in lung adenocarcinoma in phase I study	Phase II	Tawbi et al.64(ASCO abstract)
Notch	DAPT†	NSCLC cell lines	Reduction in ALDH+ tumour cells with cell cycle arrest	Preclinical	Sullivan et al.41
Notch	MRK-003	NSCLC cell lines	In vitro and in vivo inhibition of Notch signalling, reduction in proliferation and induction of apoptosis	—	Konishi et al.50
WNT 1	Monoclonal antibody and RNAi	NSCLC cell line	Induction of apoptosis and growth inhibition in NSCLC cells	—	He et al.51

^†DAPT = N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester (a γ-secretase inhibitor).

ALDH, aldehyde dehydrogenase; NSCLC, non-small cell lung cancer; SMO, Smoothened; SP, side population; —, not reported.

Hh pathway

The Hh pathway is involved in regulating the proliferation, migration and differentiation of progenitor cells⁵² and is silenced in most healthy adult tissues. The pathway is activated through the transmembrane domain protein smoothened and subsequent activation of the transcription factors (GLI1, GLI2 and GLI3), which regulate downstream target genes.⁵³ The role of Hh signalling in CSC regulation has been described in many cancers.^54,55 There is emerging evidence for Hh–GLI signalling in tumour maintenance in melanoma,⁵⁶ breast⁵⁷ and lung cancer.⁵⁸ Activation of the Hh pathway in SCLC was first demonstrated in 2003.⁵⁹ Inhibition of Hh signalling with the Veratrum alkaloid cyclopamine, a naturally occurring smoothened antagonist, leads to the loss of tumourigenicity in SCLC.⁵⁹ These data suggest that deregulation of Hh signalling might result in SCLC through acquisition of neuroendocrine fate of airway progenitor cells in response to paracrine signals from adjacent epithelial cells. In recent data from Park et al., genetically engineered mice revealed that Hh signalling plays an important role in SCLC initiation. Inhibition of Hh signalling, using second-generation smoothened antagonists such as LDE-225 (Novartis), can supplement treatment with chemotherapy to inhibit tumour recurrence.⁶⁰

Notch pathway

Notch signalling is a highly conserved embryonic pathway in mammals and is frequently involved in cellular fate determination. Activation of Notch signalling through cell-to-cell interaction retains stem cell viability via asymmetric cell division.⁶¹ The Notch pathway is involved in the development of normal lung and in determining proximal and distal epithelial cell fate in studies involving knockout mice.⁶² The role of Notch pathways in normal lung stem cells has been demonstrated in mouse studies involving knockout of Hes1 (hairy and enhancer of split 1), a key downstream target in Notch pathway. Here, suppression of Notch signalling resulted in increased neuroendocrine differentiation but reduced club cells (formerly known as Clara cells).⁶⁵ Constitutive activation of Notch signalling resulted in delayed differentiation and accumulation of distal airway stem cells.⁶⁶ The evidence for Notch pathway involvement in lung tumourigenesis comes from studies that showed elevated levels of Notch transcripts in NSCLC cell lines.^67,68 Notch activity promoted pro-apoptotic signals through Bim and inhibited anti-apoptotic signals through induction of Survivin.^50,67 Blockade of Notch pathway with inhibitors of γ-secretase (a protein complex required for proteolytic cleavage of Notch receptors) resulted in growth inhibition and induction of apoptosis both in vitro and in vivo.⁵⁰ Further evidence comes from observations that deregulation in Notch signalling is involved in maintenance of oncogenic-Ras-driven cancers.⁶⁹ Recently, active notch pathway was reported to be required for the clonogenic activity of ALDH-positive cells in NSCLC.⁴¹ Blockade of Notch3 by γ-secretase inhibition leading to suppression of clonogenic survival was demonstrated in NSCLC cell lines.⁶⁸ Restoration of clonogenic survival was achieved after reintroduction of the Notch3 receptor domain.

In human glioma xenografts, inhibition of Notch using γ-secretase inhibitors resulted in diminished tumour growth as well as reduced numbers of the CD133-positive glioma stem cells.⁷⁰ Some reports have also suggested possible oncogenic mutations in Notch1 receptor in NSCLC.⁷¹ However, despite several reports of a possible role for Notch signalling in lung carcinogenesis, there is no clear evidence that this pathway is involved in the maintenance of CSC phenotype. Notch may also have a tumour suppressor effect in squamous epithelia and myeloid malignancies^72,73 and an antiangiogenic effect in blood vessels.⁷⁴ Overall, the Notch signalling pathway is much more complex than currently understood.

In small cell lung cancer, a malignancy characterized by neuroendocrine features, the expression of Notch receptors is rarely seen. Notch signalling, however, is believed to be an attractive anticancer target, not only due to involvement in carcinogenesis and stem cell maintenance but also in tumour angiogenesis⁷⁵ and immune function.⁷⁶ Due to positive outcomes in preclinical models, γ-secretase inhibitors were developed, but due to lack of specificity, excessive toxicities especially in the gastrointestinal tract were observed in rodent models.⁷⁷ More specific inhibitors of Notch transcriptional complex are currently under development and investigation.

WNT/β-CATENIN pathway

Another less understood pathway is WNT/β-CATENIN pathway. Aberrancies in WNT pathway have been linked to lung tumourigenesis.⁷⁸ Targeting the WNT pathway with monoclonal antibodies resulted in induction of apoptosis in lung cancer cell lines overexpressing WNT-2 protein.⁷⁹ Further investigation of this pathway in lung CSC is needed in order to determine the therapeutic potential of targeting the WNT pathway.

CONTROVERSIES SURROUNDING THE CSC MODEL IN LUNG CANCER

Although experimental evidence in support of the hierarchical arrangement of cells in lung cancer has accumulated in the last few years, controversies and uncertainties remain. Isolation of tumourigenic cells both in vivo and in vitro relies on specific surface markers that facilitate sorting of cancer cells into phenotypically distinct populations. However, no universal lung CSC marker has yet been discovered, as most of the markers tend to be histotype specific.^38,41,80 For example, in many lung cancer samples, CD133 was not detected at all,^21,44 and the prognostic role of CD133 expression in lung cancer has also not been clearly established due to discrepant reports.^41,43,44 CD44 expression and ALDH activity marks CSC in NSCLC, but the protein expression of both these markers appears to be associated particularly strongly with squamous histology.^80,81 Furthermore, the combined effect of these markers in lung CSC is at the moment unclear. Apart from a report where CD133 expression partially overlapped with ALDH activity in NSCLC cell lines,²² there is no convincing data to suggest that the enrichment for one marker also enriches for the others. Instead, it may be possible that these markers play their part independently of each other to represent a distinct subpopulation.⁸² In order to isolate pure CSC in lung cancer, more robust markers need to be identified.

The sensitivity of these surface markers to accurately delineate tumourigenic population is also being questioned. For example, CD133-negative cells in human lung cancer cell lines may also be tumourigenic, just like their CD133-positive counterparts.⁸³ Similarly, discrepant results were seen with ALDH and SP phenotypes. Both ALDH-positive and ALDH-negative cells were tumourigenic in non-obese diabetic/severe combined immune deficiency mice,⁸⁴ while non-SP cells retained CSC properties in vitro as well as in xenotransplantation.⁸⁵ These findings suggest that either other tumour-initiating cells exist independently of these markers, or a phenomenon of ‘cellular plasticity’ exists. The latter, where non-cancer stem/progenitor cells convert to a stem/ progenitor cell phenotype, has recently been demonstrated in experiments conducted by Akunuru et al.⁸² Using NSCLC cell lines and primary lung adenocarcinoma samples, the authors demonstrated that non-SP, CD133-negative and ALDH-negative cells can generate SP, CD133-positive and ALDH-positive cells under adherent culture conditions.⁸² This phenomenon closely resembles the hypothesis of ‘phenotypic switching’ described in some other tumour types. Gupta et al. showed in breast cancer cells that different subpopulations of cells exist in phenotypic equilibrium.⁸⁶ Cells purified for a given phenotype can, over a period of time, switch to an alternate phenotype, thus returning the tumour into an original state of phenotypic equilibrium. These findings are more in keeping with a stochastic model, in which cells under selective pressure are genetically, and therefore phenotypically, more able to form colonies and engraft in mice. However, the phenomenon of cellular plasticity has not been observed in some other tumours. For example, in melanoma, ATP-binding cassette B5-negative cells are unable to generate ATP-binding cassette B5-positive cells.⁸⁷ In the case of solid tumours like lung cancer, where identification of CSC relies on expression of markers, it then becomes difficult to determine whether the tumour follows stochastic of hierarchical growth. As currently known markers only identify tumourigenic populations to varying extents, the overall accuracy of these markers remains doubtful.

CONCLUSION

The CSC model remains a very important model of cancer biology with major implications for lung cancer. The clinical picture in lung cancer with innate resistance to treatment and high rates of relapse fits this model well. There is considerable experimental data in support of a subpopulation of cancer cells that is capable of resistance to therapeutic strategies and tumour regrowth. Given the high rate of recurrence following therapy in all forms of lung cancer, the capacity to target minimal residual disease populations by blocking their CSC-like activity remains a tantalizing, but as yet theoretical, treatment option. The controversies largely stem from the fact that the markers studied so far are not sufficiently specific or sensitive. Nevertheless, the current identification of markers and pathways is already underpinning some novel developments in therapeutic strategies for patients with lung cancer.

Abbreviations

ALDH

aldehyde dehydrogenase

CSC

cancer stem cell

NSCLC

non-small cell lung cancer

SCLC

small cell lung cancer

SP

side population