2. Cancer Stem Cells (CSCs)
Early in 1875 Julius Cohnheim introduced a theory that tumors may arise from stem cells left over from embryonic development [
1]; but the concept of CSCs was put forwarded for the first time in 1994 [
2]. Thereafter; leukemia stem cells were first isolated in 1997 [
3]; and CSCs were isolated and characterized from solid tumors in breast cancer in 2003 [
4]. From the consensus of an American Association for Cancer Research (AACR) workshop in 2006, CSCs are defined as a kind of cell that possess stem cell-like properties,
i.e., self-renewal and pluripotency [
5]. CSCs usually account for 1–4 in 100 leukemia cells [
6,
7] or 1 in 1000–5000 cells in lung; ovarian; and breast cancers [
8]; but have strong tumorigenicity; metastaticity and resistance to radio- and chemotherapy; playing critical roles in cancer progression and therapeutic response [
9,
10].
CSCs are heterogenetic. The CSCs isolated from different stages or grades of the same type of tumors are distinct, while CSCs from the primary and metastatic tumors are different [
11,
12]. Even in the same tumor, there co-exist different CSC pools, and the distinct CSC subpopulations within a tumor could interconvert. For instance, two molecularly distinct populations of leukemic stem cells (LSCs) co-exist and are hierarchically ordered in primary human CD34(+) acute myeloid leukemia; one LSC population gives rise to the other [
13]. CSC heterogeneity also exists in solid tumors. Three cell populations differing in tumorigenicity and self-renewal are identified in estrogen receptor-negative breast tumors [
14]. Two highly tumorigenic CSC populations that differ in CD34 expression but are enriched in integrins co-exist at the cancer-stroma interface and display different tumor growth properties [
15]. The similar phenomenon is observed in ovarian carcinoma [
16], colorectal cancer [
17] and PTEN-deficient glioblastoma [
18].
Understanding of CSCs has advanced in the past decade. Newer concepts of CSCs consider that CSCs are a “status”, but not a fixed, immutable and frozen cell population. CSCs and non-CSCs exist in a dynamic equilibrium and could interconvert [
19,
20]. Non-CSCs could acquire the properties and tumor formation ability of CSCs by reprogramming [
20,
21]. In fact, if CSCs are a fixed cell population, the ratio of CSCs should be progressively reduced with proliferation of cancer cells, but it is clearly not the case. The proportions of CSCs in cancer cell lines remain about 0.1% in the constant culture [
10,
22]. The microenvironment plays a critical role in CSC division and interconversion. Myofibroblast-secreted factors restore CSC phenotypes in more differentiated colon cancer cells
in vitro and
in vivo [
23]. Hypoxia-inducible factor (HIF2α) promotes the self-renewal of the stem cells and enhances a more stem-like phenotype in the non-stem population [
24–
26]. The CSCs may develop
de novo from differentiated cancer cells (
i.e., reprogramming) by the induction of microenvironment. Therefore, the hierarchical model of mammalian CSCs should be considered as bidirectional between stem and non-stem cells of the tumor [
20,
21].
3. Apoptotic Signaling in CSCs
Apoptosis is an active, strictly regulated, and energy-dependent cell death process [
27]. In mammalian cells, apoptosis is regulated via two different pathways,
i.e., the extrinsic and intrinsic pathways. Caspases play important roles in apoptosis. The activation of caspase family proteins triggered by these two signaling pathways results in a series of cellular substrate excision and changes, such as chromatin condensation, DNA fragmentation, membrane blebbing, and cell shrinkage [
28]. The extrinsic pathway is triggered through the binding of extracellular proapoptotic ligands to cell surface receptors, known as death receptors, such as CD95, nerve growth factor receptor (NGFR), and TNF-related apoptosis-inducing ligand (TRAIL) receptors (
Figure 1) [
29,
30]. After binding to the receptor, a death-inducing signaling complex (DISC) composed of the Fas associated death domain (FADD) and procaspase-8 and -10 is formed [
31–
33], and this protein complex activates procaspase-8 and -10 inside itself, and then cleaves procaspase-3 and initiates the apoptosis process [
34]. In the extrinsic pathway, the downregulation of cellular FLICE inhibitory protein long isoform (c-FLIPL) by ubiquitination at lysine residue (K) 195 occurs [
35]. The intrinsic pathway, also known as the mitochondrial pathway, is induced by a variety of stress signals that trigger cellular and DNA damage, such as ionizing radiation, cytotoxic agents, and growth factor withdrawal. They lead to mitochondrial outer membrane permeabilization (MOMP) and transcription or post-translational activation of BH3-only proapoptotic B-cell leukemia/lymphoma 2 (Bcl-2) family proteins [
29]. The mitochondrial permeability is a key step in the apoptosis cascade and mediated by Bcl-2 family proteins. The mitochondrial permeability allows the release of apoptotic proteins, such as cytochrome c and second mitochondria-derived activator of caspase (Smac), from the intermembrane space into cytosol [
36,
37]. The assembly of cytochrome c and apoptotic protease-activating factor-1 (Apaf-1) activates caspase-9 which in turn activates the effectors caspase-3, -6, and -7, leading to apoptosis [
29]. Inhibitors of apoptosis protein (IAP) prevent both intrinsic and extrinsic apoptosis by inhibiting caspase activity, which represents the last protective measure against apoptosis [
38]. Death signaling can also be activated by c-Jun
N-terminal kinase (JNK) signaling which leads to phosphorylation of Bcl-xL at Ser62, decreasing its anti-apoptotic activity in the intrinsic pathway [
35]. Intrinsic and extrinsic apoptosis pathways are both disordered in cancer cells; and apoptosis evasion is one of the hallmarks in cancer cells [
39,
40].
Apoptotic signaling pathways, including extrinsic and intrinsic pathways, are also deregulated in CSCs. In glioblastoma and lung CSCs, the death receptors (DR) mediating the extrinsic pathway are expressed at a high level [
41], and the upregulation of DR4 in colon CSCs leads to chemo-resistance [
42]. The FLICE-like inhibitory proteins (cFLIP) are a negative modulator of death receptor-induced apoptosis, consisting of two subtypes: long cFLIP (cFLIPL) and short cFLIP (cFLIPS) [
43,
44]. In CD133+ glioblastoma, breast cancer, and T-cell acute leukemia cells, the cFLIPs are upregulated [
45,
46]. Silencing of cFLIPs by siRNA restores cell sensitivity to death stimuli, suppressing CSC self-renewal and tumor metastasis [
47,
48]. It was reported that insufficient expression of death receptors and upexpression of c-FLIPs leads to CSCs-enriched neurosphere resistance to TRAIL [
49].
Survivin is an anti-apoptotic protein, belonging to the inhibitors of apoptosis protein (IAP) family that regulates cell division, apoptosis and pluripotency [
38,
50,
51]. Survivin is enriched in hematopoietic stem cells, neuronal precursor cells, CD34(+)/38(−) AML stem cells and glioblastoma and astrocytoma CSCs [
52–
54]. Other IAP proteins upregulated in CSCs include XIAP, c-IAP1, and Livin [
45,
54].
Dysregulation of the intrinsic pathway in CSCs is mainly reflected in Bcl-2 family proteins and the DNA damage response. Bcl-2 family proteins are composed of anti-apoptotic proteins (Bcl-2, Bcl-X
L and Mcl-1) and pro-apoptotic molecules (Bax, Bak, Bid, Bim, Bik, Noxa and Puma [
55]. It is the imbalance of anti- to pro-apoptotic protein ratio rather than a specific molecule expression level that tips the balance to cell survival and regulates sensitivity to apoptotic stimuli [
55]. In most tumors, anti-apoptotic Bcl-2 family proteins are overexpressed in CSCs [
56]. For instance, CD133+ glioma CSCs express a high level of anti-apoptotic proteins Bcl-2 and Bcl-X
L [
45,
57], and high expression of Mcl-1 correlates with resistance to the Bcl-2 inhibitor ABT-737 in glioma CSCs [
57]. In colon CSCs, Bcl-2 is increased and inhibits apoptosis and autophagy [
58]. Downregulation of Bcl-2 or upregulation of Bax induces apoptosis of CSCs [
37,
59]. Therefore, inhibition of the mitochondrial death cascade has been attractive for CSC-targeted therapeutic intervention of cancers.
DNA damage response (DDR) is tumor suppressor p53-mediated cell-cycle arrest, DNA repair, and apoptosis in response to DNA damage [
60]. Glioma CSCs isolated from human glioma xenografts and primary glioblastoma produce radio-resistance by preferential activation of the DNA damage response, and the radio-resistance of CSCs could be reversed by specific inhibition of Chk1 and Chk2 checkpoint kinases, upstream activators of p53 [
61]. In addition, nuclear factor-kappa B (NF-κB) signaling regulates apoptosis of CSCs through affecting the expression of pro and anti-apoptotic proteins. In leukemic stem cells, NF-κB is activated [
62] and increases the quiescent LSC number [
63]. Breast CSCs exhibit sensitivity and apoptosis to NF-κB pathway inhibitors, such as parthenolide, pyrrolidinedithiocarbamate, and diethyldithiocarbamate, and the expression of CD24 in CD44+ breast CSCs potentiates DNA damage-induced apoptosis by suppressing NF-κB signaling [
64–
66].
MiRNAs are small non-coding RNAs (ncRNAs) that regulate protein translation by binding to target mRNAs [
67]. MiRNAs are widely involved in cancer cell growth, migration, invasion, and drug sensitivity [
68,
69]. In tumor cells, miRNA expression is dysregulated, and function as tumor suppressors or oncogenes. For instance, miR-223, miR-122, and miR-26 function as tumor suppressors of liver carcinogenesis whereas miR-130b, miR-221, and miR-222 are oncogenic factors [
70–
74]. Emerging evidence indicates that miRNAs are key regulators of stemness. For instance, miRNAs let-7, miR-21, miR-22, miR-34, miR-101, miR-146a, and miR-200 affect CSC phenotypes and functions through targeting oncogenic signaling pathways [
75]. Other miRNAs appear to affect the fate of CSCs by controlling their self-renewal. For example, MiR-34 inhibits human pancreatic CSCs by regulating Notch and bcl-2 gene expression [
76], and miRNA-34a suppresses glioma CSC growth by targeting several oncogenes [
77]. There is also a difference in miRNA expression levels between CSCs and non-CSC cancer cells [
78]. For example, mir-21 and mir-302 expression is increased in CSCs whereas let-7a is downregulated. In addition, mir-372, mir-373, and mir-520c-5p are expressed at higher levels in non-CSC than in CSCs [
78].