Cancer stem cells as targets for immunotherapy

Summary

Current cancer therapies target the bulk of the tumour, while a population of highly resistant tumour cells may be able to repopulate the tumour and metastasize to new sites. Cancer cells with such stem cell‐like characteristics can be identified based on their phenotypical and/or functional features which may open up ways for their targeted elimination. In this review we discuss potential off‐target effects of inhibiting cancer stem‐cell self‐renewal pathways on immune cells, and summarize some recent immunological studies specifically targeting cancer stem cells based on their unique antigen expression.

Introduction

Cancer stem cells (CSCs), also called stem‐like cells or tumour‐initiating cells (TICs), are a distinct subpopulation of tumour cells. They may arise in a variable and unpredictable manner due to genetic and epigenetic changes during tumour development (stochastic theory), or they are cells that possess a unique intrinsic ability to initiate tumour growth and self‐renewal (hierarchical theory). The existence of CSCs had been postulated for some time before experimental evidence was first provided in 1994. A subpopulation of acute myeloid leukaemia cells, with CD34^high CD32^low phenotype, was shown to be highly capable of engrafting leukaemia in severe combined immunodeficiency (SCID) mice.1 These cells were called leukaemia‐initiating cells. Solid tumour TICs were first described in breast cancer (CD44⁺ CD24^−/low/linage⁻ cells) in 2003,2 followed by further discoveries in a variety of malignancies. Regarding using the correct designation, CSCs are perceived as immature progenitors of tumour cells, residing at the top of a hierarchical organization of tumour cell differentiation. Conversely, the definition of TICs is based on the function of these cells, as they are uniquely capable, at very low cell numbers, to initiate heterogeneous, complex tumours in vivo. In this review, we use the combined CSCs/TICs abbreviation, as suggested by Maccalli et al.3 reflecting the variety of the references.

While traditional cancer therapies, such as radio‐ or chemotherapy, may eliminate the bulk of the tumours, treatment resistance in CSCs/TICs is thought to be responsible for relapse. In order to prevent or significantly delay relapse, these cells should be specifically targeted and eliminated. There are numerous ongoing trials targeting CSCs/TICs (for the latest review see Qureshi‐Baig et al.4); however, in order to design efficient novel treatment approaches we need a clearer understanding of the biology of these cells. In this review we summarize the current state of knowledge concerning the feasibility of immune targeting CSCs/TICs in solid tumours.

We also point out how some of the biological targeting of CSCs/TICs may act as a double‐edged sword by also affecting immune responses.

Developmental pathways in CSC/TIC signalling

Developmental pathways play important roles in normal stem cell function. The core stem cell pathways, Notch, Wnt/β‐catenin, Hedgehog (discussed in more detail below) and some other crucial pathways, such as Janus kinase/signal transducer and activator of transcription/phosphoinositide 3‐kinase/phosphatase and tensin homologue (JAK/STAT, PI3K/PTEN and nuclear factor kappa B (NF‐κB), promote cell proliferation and the formation of CSC‐like colonies.5, 6 They are frequently altered in cancers, such as becoming deregulated or persistently activated, and are suggested to be responsible for CSC/TIC regulation.7, 8, 9, 10

Notch

Notch ligands, located in the plasma membrane of adjacent cells are transactivated, triggering the transcription of Notch target genes, such as the hairy and enhancer of split (HES)‐related family, c‐myc, PI3K, protein kinase B (AKT), NF‐κB, peroxisome proliferator‐activated receptor (PPAR), cyclin D1, p21 and p27. The activation of these downstream targets regulates cell fate leading to differentiation, cell‐cycle progression and survival, depending on the particular signalling context. In stem‐like cells Notch may delay differentiation and promote cell survival.11, 12 There are numerous ongoing Phase I and II clinical trials in cancer with a range of targets and mechanisms investigating the usefulness of Notch targeting, alone or in combination with other therapies.11 However, Notch signalling has also been linked to peripheral T‐cell maturation into effector cells, such as developing cytotoxic T‐cell function or cytokine production.13 T‐cell activity has been shown to be impaired by Notch‐inhibition with a γ‐secretase inhibitor.14

Wnt

Two pathways have been identified: the canonical pathway is β‐catenin‐dependent and involved in cell fate determination. The non‐canonical pathway is β‐catenin independent and involved in cell movement and polarity. Wnt signalling is initiated by soluble ligands released by neighbouring cells. In cancer, gain of function and loss of function mutations, regulation by methylation and histone modification of this pathway have all been observed.9, 15, 16 Its inhibition in CSCs is a main area of cancer therapy research.17 However, the Wnt/β‐catenin pathway is also a key regulator of T‐cell development and activation. Wnt/β‐catenin signalling is crucial for CD8⁺ memory T‐cell development,18 while agonists of the pathway improve immunotherapy outcomes.19

The Hedgehog (Hh) signalling pathway has been implicated in tissue homeostasis and repair and epithelial to mesenchymal transition (EMT) in normal tissues. In cancer, aberrant Hh‐signalling, such as over‐expression of its ligands, loss of function of the receptor and dysregulation of transcription factors, promote tumorigenesis and tumour progression. Hh signalling can also be canonical and non‐canonical, and triggered by a variety of factors in the tumour microenvironment, such as transforming growth factor (TGF)‐β, tumour necrosis factor (TNF)‐α and interleukin (IL)‐6. Inhibition of Hh signalling is also undergoing intense investigations for cancer treatment.20 Hh signalling is relevant in immune cell development and function, although its effect on peripheral T‐cell function is controversial.21, 22, 23, 24 Because it is also involved in myeloid‐derived suppressor cell (MDSC) function,25 Hh inhibitors may deliver additional benefits.

As there is a considerable overlap between these pathways, single targeting is unlikely to achieve a physiologically relevant level of inhibition. Furthermore, the fact that they are also involved in normal tissue homeostasis and development, including immune cell behaviour and peripheral effector function, makes their targeting a difficult challenge.

Identification and isolation of CSCs/TICs

Surface marker‐based identification

CSCs/TICs are typically isolated based on their expression of proteins shared in common with healthy stem cells. The markers most commonly used in solid tumours to identify CSCs/TICs are CD133, CD44, IL‐6R, CD24, epithelial cell adhesion molecule (EpCAM), leucine‐rich repeat‐containing G‐protein coupled receptor 5 (Lgr5), CD166 and CD29, alone or in combination. The use of these markers is relatively conserved across the spectrum of solid cancers. However, there are technical considerations which may give rise to false positives or inconsistencies in the results, including subjectivity in flow cytometry gating, the use of cell lines versus primary cells, confirmation of function in clonogenic cultures and animal models. For some of these markers there is evidence for direct stem cell‐like function, while recently the validity of some, as bona fide CSC/TIC markers, has been called into question, as discussed later. A few common markers are discussed below.

CD133

CD133 (Prominin‐1) is a five‐transmembrane glycoprotein used to identify CSCs/TICs in prostate, pancreatic, colon and liver cancer and glioblastoma.5 Although the precise function of CD133 has not been elucidated, it is known to bind cholesterol and is localized in protrusions of the membrane, e.g. in villi and cilia. Despite its initial acceptance as a CSC/TIC marker, in some instances cells expressing this marker have not demonstrated exclusive tumour‐initiating ability.26, 27 CD133 is also present in a number of adult tissues, including the kidneys, pancreas and colon28, 29 and is used as a marker for haematopoietic stem cells. Thus it is important to acknowledge that it is not a universal CSC marker, nor is it a cancer cell‐specific antigen.

Some of the inconsistencies observed in the application of CD133 as a CSC/TIC marker may be associated with its pattern of expression and the antibodies used to detect it.30 The most commonly used antibodies for CD133 detection are mouse monoclonal antibodies CD133/1 and CD133/2, which detect the epitopes AC133 and AC141, respectively. These epitopes are distinct from each other and both are glycosylated. The different glycosylation status of CD133 across different tissues may give rise to false negatives. Glycosylation status is also suggested to change as a result of differentiation in some lineages,31, 32 although this may be advantageous in the specific detection of early progenitor cells. However, a number of studies have shown that AC133 epitope expression (as detected by the CD133/1 antibody) does not correlate with CD133 protein or mRNA levels.32 The functional outcome of the loss of this epitope upon differentiation is unclear.

CD44

CD44 is used to identify CSCs/TICs in breast, prostate, colon, head and neck and pancreatic cancer. CD44 is a transmembrane glycoprotein that functions as a receptor for hyaluronic acid. It has a multitude of physiological and pathological functions, including adhesion and migration, proliferation, growth and survival. However, CD44 is widely expressed in healthy tissues and in multiple cell types in the cancer microenvironment, making it difficult to apply as a specific CSC/TIC marker. CD44 is subject to alternate splicing and it has been suggested that CD44 splice variants (CD44v) specifically identify cells with greater tumorigenic potential compared to cells expressing CD44s, the standard isoform.33, 34 Additionally, certain splice variants have been suggested to have pathological functions in colon and pancreatic cancer35 and have prognostic utility in other cancers, including non‐small‐cell lung carcinoma (NSCLC), acute myeloid leukaemia (AML) and gastric cancer.34 CD44^high cells were shown to be less immunogenic than CD44^low tumour cells in head and neck cancer, partially via enhanced expression of the programmed cell death ligand‐1 (PD‐L1).36 A Phase I trial of a humanized IgG1 antibody, targeting the extracellular hyaluronic acid binding domain of all CD44 isoforms37 in heavily pretreated cancer patients, proved to be safe, but had only modest clinical effects. Conversely, however, CD44 also regulates T helper type 1 (Th1) cell survival, memory function,38 T‐cell IL‐17 and IFN‐γ production, thus its targeting may impair anti‐tumour immune responses.39

IL‐6R

IL‐6 has been shown to enhance stemness markers [Notch, Lgr5 and octamer‐binding transcription factor 4 (Oct‐4)] in colon cancer40 and the survival and tumorigenicity of cancer stem cells identified as aldehyde dehydrogenases (ALDH)^high/CD44^high in head and neck carcinoma.41 Inhibition of IL‐6 signalling with an IL‐6R inhibitor antibody (tocilizumab) prevents human CSC‐mediated tumour initiation.41 Similar observations were made in an NSCLC cell line, where inhibition of IL‐6 or IL‐6R, separately or in combination, significantly inhibited CSC proliferation and growth.42 In colon cancer, blocking the IL‐6 receptor with a monoclonal antibody reduced spheroid formation, stem‐cell‐related gene expression and enhanced resistance to chemotherapy by 5‐fluorouracil.40 However, IL‐6R also plays an important role in naive and central memory T‐cells, regulating their survival, proliferation and effector function while also blocking regulatory T‐cell (T_reg) function.43

Further markers

There is a huge array of further markers that have been used alone or in combination in a variety of cancers to identify CSCs/TICs. They include CD24, EpCAM, Lgr5, CD90, CD117, CD166, CD29, CD177, epithelial‐specific antigen (ESA), alpha‐fetoprotein (AFP), nestin, chemokine receptor 4 (CXCR4) and stage‐specific embryonic antigen 4 (SSEA4). However, the plasticity of these markers means that they can be up‐regulated on cells originally not expressing them.44 Their heterogeneous expression throughout the tumour tissue45 has also been observed, making their isolation more difficult due to being a moving target. From our review viewpoint, some of these markers, when unique or over‐expressed, can serve as CSC/TIC‐associated antigens for T‐cell recognition, as discussed later.

Metabolism‐based identification

Aldehyde dehydrogenase

The aldehyde dehydrogenase (ALDH) family consists of 19 genes (in humans) that express enzymes which catalyse the oxidation of aldehyde.46 Aldehyde oxidation is required to metabolize many physiological substrates, such as vitamins, lipids and amino acids.46 These enzymes also catabolize aldehydes derived from pharmacological substrates; overall, they have a protective detoxifying effect. High ALDH activity is associated with both normal stem cells and CSCs. ALDH activity has been used to identify CSCs in breast, colon, head and neck, pancreatic and liver cancer. It is measured using the flow cytometry‐based ALDEFLUOR assay. An advantage this assay has over phenotypical detection methods is the direct readout of a fluorescent signal, based on enzyme activity. This may be less susceptible to inconsistencies encountered by antibody‐based staining, such as epitope down‐regulation or masking, or expression of splice variants. The ALDEFLUOR assay mainly measures the activity of the ALDH1 family, as the inhibitor N,N‐diethylaminobenzaldehyde (DEAB) used in the assay is a specific inhibitor of ALDH1.47 However, this may mean that the frequency of ALDH high cells is underestimated in tissues in which the predominant ALDH isozyme is not of the ALDH1 family.

In prostate cancer (PCa), ALDH^high population frequency varies from 1·2 to 8·3% in the classic prostate cancer cell lines (DU145, PC3, LNCaP and 22RV.1), while cell lines derived from metastatic PCa cells had an even higher ALDH^high frequency; up to 30%.48, 49 ALDH^high PCa CSCs have demonstrated enhanced clonogenicity and migration in vitro, expression of stemness‐associated genes, in‐vivo tumour initiation49, 50, 51 and induction of metastasis.48 ALDH activity has been correlated with the expression of other CSC or clinical markers; CD44, EpCaM and integrin expression were significantly higher in ALDH^high compared to ALDH^low cells in one report,48 while CD44 and α ₂‐integrin levels did not differ in another PCa cell line.50 CD133 was not detectable in both these studies; in primary samples it was detected in freshly isolated cells but was greatly reduced upon passaging of the cells.52 In ovarian cancer, both ALDH activity and CD133 expression were detected in primary specimens. Upon passage of the primary cells, CD133 expression was reduced but could be rescued by CSC culture conditions, i.e. sphere culture and serum‐free conditions.53 ALDH^high ovarian cancer cells, which range in frequency from 0·1 to 7·9% in cell lines and 1–7% in primary samples, demonstrated tumorigenicity in vivo.53, 54 High ALDH activity has been used to identify both stem and cancer stem cells in breast cancer with frequencies of 8 and 4% ALDH^high cells, respectively.55 ALDH^high CD44⁺ CD24⁻ and ALDH^high CD44⁺ CD133⁺ breast cancer cells demonstrated greater tumorigenicity than the corresponding low/negative populations.56 Conflicting evidence exists for the utility of ALDH activity as a CSC marker in lung cancer.57 In one study, both ALDH^high and ALDH^low cells (from a single cell line) were capable of initiating tumours in vivo,58 while in another, ALDH^high but not ALDH^low cells from two different cell lines demonstrated tumorigenicity in vivo.59 One further study using eight lung cancer cell lines identified STAT‐3 signalling as a mediator of ALDH3A1 activity and demonstrated tumorigenicity in two cell lines.60 This suggests that variability in cell lines could lead to rejection of a potentially applicable CSC marker and highlights the importance of testing in multiple cell lines/primary tissues.

ALDH expression has also been correlated with clinical outcome in a number of cancers. High ALDH expression (as well as the presence of other CSC markers) correlates with poor prognosis in pancreatic cancer61 and in both serous (n = 62) and clear cell (n = 37) ovarian carcinoma cases54 and in 112 serous carcinoma cases in a further study.62 In prostate cancer, ALDH expression correlates with more advanced stage, compared to localized cancer and BPH,63 while there seems to be no consensus on the utility of ALDH activity for the selection of tumorigenic melanoma cells.64

ALDH has been shown to contribute to resistance mechanisms to radiation therapy and chemotherapy, thus a different, more specific, therapeutic approach must be undertaken. Some groups have investigated chemical inhibitors47 such drug development requires the design of highly specific inhibitors, as the widespread expression of ALDH in healthy stem cells may result in off‐target effects. It is also possible that functional redundancy within the large isozyme family could compensate for inhibition of one ALDH target. In the immunotherapy setting, ALDH^high CSC‐loaded dendritic cells (DC) have been used successfully in two in‐vivo melanoma models.65, 66 ALDH activity has been demonstrated in the T_reg cell immune subset in the transplantation setting67; thus, ALDH targeting may also have an anti‐tumour effect via affecting the T_reg subset.

Glycolytic activity

CSCs/TICs have fewer mitochondria, thus they are more glycolytic than differentiated tumour cells, as shown in melanoma, breast, lung and liver carcinomas.68, 69, 70, 71 This can be detected by reduced mitochondrial activity, perinuclear mitochondrial distribution, lower intracellular concentrations of ROS and ATP and lower amounts of mitochondrial DNA in CSCs/TICs.72

CSC/TIC inhibitors targeting either self‐renewal pathways, surface markers or enzymes have potential off‐target effects on other cell types. Figure 1 illustrates the potential side effects of some of these inhibitors on immune cells.

Figure 1.

Targeting stem cell pathways, functions or markers may have off‐target effects on immune cells. The red blocking symbols on the right represent negative effects by inhibiting immune responses, while the green symbols indicate positive effects by blocking immunosuppressive cells.

Function‐based identification

Slow cell division

Slow cell division of CSCs/TICs, that exist mainly in G0 phase, can be detected by their characteristic of retaining dyes that normally become diluted during proliferation, such as PKH, bromodeoxyuridine (BrdU) or carboxyfluorescein succinimidyl ester (CFSE). Dye‐retaining cells that give progeny to xenotransplants have been observed in glioma, melanoma, breast and pancreatic cancer.73, 74, 75, 76, 77

Enhanced drug resistance and detoxifying pathways

The ATP‐binding cassette (ABC) transporter family of proteins, especially ABCB1, ABCC1, ABCG2 and ABCB5, are active in cancer stem cells, but switched off during differentiation.78 These transporter proteins are extremely efficient in pumping out complex molecules from the cytoplasm, thus protecting the cells from exogenous toxins, including numerous chemotherapeutic drugs. They have a wide range of substrates, such as peptides, lipids, hydrophobic drugs, polysaccharides and proteins.79 Their targeting with specific inhibitors is one of the active areas of drug development in cancer.80 As CSCs/TICs also exclude hydrophobic Hoescht dyes via this mechanism, they can be identified by forming a side population (SP) based on low dye levels.81 However, ABCB proteins, such as transporter associated with antigen processing (TAP), also have an important role in intracellular peptide trafficking across membranes, with crucial involvement in major histocompatibility complex (MHC) class I antigen presentation and DC function.82 Thus, the off‐target effect of tumour ABCB targeting may be deleterious for generating efficient anti‐tumour T‐cell responses.

Enhanced resistance to radiotherapy

It has been speculated that the failure of radiation therapy correlates with the survival of at least some of the CSCs/TICs in the tumour, although the underlying protective mechanisms are incompletely understood. Radiation causes direct damage to the DNA, such as single‐ or double‐strand breaks or acts indirectly via reactive oxygen species (ROS). The consequence of radiation damage can be temporary or permanent cell cycle arrest, mainly via checkpoint kinase activity (Chk‐1 and Chk‐2), which allows time for DNA damage repair (DDR) to take place. If the damage is irreparable, DDR mediates senescence or cell death.83 ROS mediates oxidative stress which, if beyond the capacity of the cell's anti‐oxidant defence mechanism, may also lead to cell death. Both DDR and the ROS scavenging system have been reported to be highly efficient in CSCs/TICs in numerous solid cancers,84 contributing to their intrinsic radioresistance. Extrinsic radioresistance may be supported by the localization of CSCs. It has been suggested that CSCs/TICs residing in hypoxic areas are more resistant to radiation‐induced damage than cells in normoxic areas. Hypoxia‐inducing factors (HIF)1α and HIF2α may lead to the activation of the Notch, Wnt and Hedgehog signalling pathways, which are essential for CSC/TIC maintenance. There is clinical evidence that radiotherapy is generally less successful in CSC‐rich laryngeal cancer where CSCs were identified as a residing subset within the CD44^high population.85

However, the effects of hypoxia are far from inhibitory on immune cell function. Activated DC become glycolytic and their long‐term survival is regulated by HIF1α.86 Effector T‐cells are also resistant to hypoxia – even more, their cytolytic machinery is turned on by HIF2α.87 As effector T‐cells also display glycolytic characteristics, they can exert their effector function even in the depth of tissues with poor vasculature, maybe where CSCs/TICs reside. Furthermore, we have shown that low dose radiation spares effector and memory T‐cells compared to naive T‐cells.88 Thus, combination of radiation with T‐cell targeting of stem cells sounds like a two‐pronged attack with potential synergistic effects.

High‐dose radiation alone may eliminate CSCs/TICs, as early prostate and lung cancers can be cured by radiotherapy alone. It may happen partly via generating immunogenic cell death, which then initiates tumour antigen uptake and antigen cross‐presentation by DC.89 One could speculate whether radiation‐induced CSC death is also immunogenic and weather radiation would generate CSC/TIC‐specific effector and memory T‐cells, significantly contributing to an abscopal effect and subsequent protection from relapse. Despite its obvious clinical importance, this question has not been studied previously.

Immunological characteristics of CIC/CSC

Immunosuppression

Proliferative T‐cell responses and IL‐2 production were inhibited by CSCs/TICs in gliobastoma90 and melanoma91 in vitro. T_reg frequencies were also increased in melanoma91 but not in glioblastoma CSC–T‐cell co‐cultures.90 Secretion of TGF‐β, IL‐10, IL‐4 and IL‐13 by CSCs/TICs has been shown to have immunosuppressive effects on NK‐cells, T‐cells and antigen‐presenting cells.90, 92, 93 Cell surface molecules, such as CD200, expressed on CSCs/TICs,94 can also dampen immune responses. CD200 over‐expression has been associated with the suppression of Th1 responses, decreased neutrophil infiltration and increased IL‐10 production induced by the tumour, as shown in a breast carcinoma model.95 PD‐L1 is often over‐expressed on tumour cells, with a function of promoting tumour glycolysis. Up‐regulation on CSCs/TICs is probably tumour‐type or localization‐dependent as hypoxia, for example, is one of the triggers that can up‐regulate PD‐L1.96 High expression on CSCs/TICs has been reported on head and neck carcinoma,36 on CD133⁺ colorectal97 and gastric,98 but not on melanoma CSCs/TICs.91

Immune resistance

Some extent of MHC class I down‐regulation has been shown on glioblastoma CSCs/TICs compared to non‐CSCs. Nevertheless, these CSCs/TICs were still able to induce autologous T‐cell responses in vitro. Furthermore, IFN‐γ‐treatment enhanced the susceptibility of CSCs to T‐cell‐mediated immune responses.90

Cancer vaccines that generate T‐cell and antibody‐responses against tumour‐associated antigens (TAA) work well in preclinical preventative models; however, the results obtained with therapeutic vaccines in the clinic are somewhat disappointing. Focusing on CSCs, it is feasible that stem‐like cells are inherently resistant to T‐cell attack, although lack of vaccine specificity for CSC/TIC‐antigens is also a possibility. Development of stem‐like features, such as Nanog, in surviving tumour cells following vaccination has been observed.99 Resistance to T‐cell killing was successfully abolished by silencing Nanog in these cells. This work suggests that stem‐like features may develop in a population of cells under immunological pressure; however, it does not show inherent stem cell resistance.

The argument that cancer stem‐like cells may not be inherently resistant to immune attack is also supported by using purified CSCs/TICs as a vaccine.66 This treatment generated T‐cells in immunocompetent mice that were highly efficient at killing CSCs/TICs and provided greater protection in the D5 melanoma model against pulmonary metastasis than vaccination with unseparated cells. As a further example to prove T‐cell susceptibility, CD33⁺ brain tumour CSCs/TICs, transfected with the pp65 antigen of human cytomegalovirus, were efficiently killed by virus‐specific memory T‐cells,100 indicating the feasibility of CSC/TIC targeting by T‐cells.

Potential immune targeting of CSCs/TICs

Antigen non‐specific immune targeting

NK cells

As MHC class I molecule expression is often lower on CSCs/TICs than on the bulk of tumour cells, CSCs/TICs are more likely to be susceptible to NK cell‐mediated killing. However, they also often lack NK‐activating ligands, such as NKG2D, as shown in brain and breast cancers.101, 102 Conversely, glioma, oral squamous cell carcinoma and colorectal cancer CSCs/TICs 103, 104, 105 have been reported to express various ligands for NK cells [most frequently poliovirus receptor (PVR), which is recognized by DNAX accessory molecule‐1 (DNAM‐1)] and are highly susceptible to NK cell killing. However, while cytokine‐activated NK cells have efficient CSC/TIC killing ability, freshly isolated NK cells from the same patient do not kill CSCs.103 This points towards microenvironmental regulation of NK cell activity in cancer patients.

γδ T‐cells

γδ T‐cells also exhibit MHC‐unrestricted lysis of targets, including tumour cells. Vγ9Vδ2 T‐cells can be activated with phosphoantigens or aminobisphosphonates and have been shown to efficiently kill CSCs/TICs in colon, ovarian and breast cancer models, especially after treatment with Zoledronate.106, 107, 108 They have also been observed infiltrating the tumour tissue, emphasizing their physiological relevance.

Antigen‐specific targeting by T‐cells

Solid cancer cells and their CSCs/TICs express human leucocyte antigen (HLA) class I but not HLA class II molecules. Their efficient targeting by CD8⁺ T‐cells depends upon a sufficient level of HLA class‐I molecule expression and intact antigen presenting machinery in these cells. CSC/TIC resistance or susceptibility to T‐cell killing has both been reported, depending on tumour type, origin of cells and culture conditions. Tissue‐derived CSCs/TICs from colon cancer were shown to express lower levels of MHC class I molecules than non‐stem‐like cells.105

Some TAA, such as mucin 1 (MUC‐1) or centrosomal protein 55 (CEP55), are expressed equally in both CSCs/TICs and non‐CSCs/TICs,109, 110 while others, such as the olfactory receptor family 7 subfamily C member 1 (OR7C1) are dominantly expressed in CSCs/TICs.111 The former TAA are classified as shared antigens and the latter are functionally linked to cancer stemness, because their expression ceases after CSC differentiation into non‐CSCs. T‐cell responses to shared TAA could achieve temporary tumour control, but it may lead to tumour escape by inducing loss of antigens that are not necessary for cellular fitness (immunoediting).112 Multiple antigen targeting, incorporating those TAA that are specifically expressed in CSCs/TICs, is much more likely to result in therapeutic success in clinical settings,113 as indicated in Fig. 2. Although it may be crucial to have high frequencies of CSC/TIC‐specific T‐cells present in the tumour tissue, breaking the localized immunosuppressive milieu is likely be helped by activated infiltrating T‐cells of multiple specificity.

Figure 2.

Simultaneous targeting of shared and cancer stem cell/tumour‐initiating cell (CSC/TIC)‐specific antigens by antigen‐specific T‐cells can eliminate CSC/TIC‐like cells and differentiated tumour cells alike.

The main types of TAA, expressed in CSCs/TICs and can be targeted by T‐cells, are summarized below.

Cancer/testis (CT) antigens

CT antigens are expressed only in germ cells; however, they have been shown to be re‐expressed in some malignancies. This cancer‐specific expression makes these, and the onco‐fetal antigen group (discussed in the next paragraph), therapeutically highly relevant compared to other types of TAA that are discussed later. Germ cells do not express MHC class‐I molecules, thus deletional tolerance may not occur during the negative selection phase of T‐cell development against these antigens. CT antigens have oncogenic functions, such as supporting tumour growth, enhancing treatment resistance and facilitating metastasis.114 There are more than 100 gene families of CT antigens listed on the Ludwig Institute of Cancer Research's website (http://www.cta.lncc.br/), such as MAGE, BAGE, GAGE, XAGE, SPANX, NY‐ESO1, etc.115 Because of their unique expression pattern, they often serve as TAA for immunotherapy trials. Interestingly, a transcriptome analysis of SP versus main population (MP) cells from colon, breast and lung cancer cell lines revealed that 18 of 74 of these antigens are preferentially expressed on CSCs/TICs.116 Novel CT antigens have recently been identified by transcriptome analysis, such as the DNAJ Hsp40 homologue, subfamily B, member 8 (DNAJB8), in SP cells of renal cancer cells.117 OLF7C1 is also a novel cancer‐testis antigen, observed on SP cells of colorectal cancer. It has been successfully targeted by HLA A24‐restricted T‐cells.111 Another CT antigen, the brother of the regulator of the imprinted site (BORIS), subfamily 6, was found preferentially expressed on cervical cancer CSCs/TICs. It has a role in maintaining CSC function and serves as a target for BORIS‐specific cytotoxic T‐cells.118

Oncofetal antigens

These antigens are typically only expressed during embryonic development in fetal tissues, but similarly to CT antigens, they can be re‐expressed in some cancers. A typical example, carcinoembryonic antigen (CEA), is expressed in numerous solid cancers but not associated with stem‐like cells. Other antigens, such as stage‐specific embryonic antigen 3 and Globo‐H, are expressed on stem‐like and non‐stem‐like breast cancer cells at similar rates,119 while 5T4 is preferentially expressed on CSC‐like cells in lung cancer and head and neck cancer.120, 121 5T4 expression, in both malignancies, is associated with poor prognosis. Xenograft models indicated that the CSC‐fraction can be reduced, tumour progression halted and local recurrence inhibited by treatment with an antibody‐drug conjugate targeting 5T4.120, 121

Over‐expressed antigens

This group of antigens has recently been reviewed by Hirohashi et al.122 and includes apoptosis‐resistance genes, such as survivin, proto‐oncogenes, such as human epidermal growth factor receptor 2 (HER‐2), centrosomal protein 55 (CEP55), sex‐determining region Y‐box 2 (SOX‐2) and cytochrome C oxidase assembly factor 1 homologue (COA‐190, and stress‐response related genes, such as heat shock protein (Hsp) or HOX genes.123 Some of the over‐expressed antigens are general stem cell‐ and not exclusively cancer stem‐cell‐specific. In PCa, Numb has been identified as a potential target for controlling tumorigenesis.124 Numb is lost in differentiating tumour cells, due to exaggerated ubiquitination and subsequent proteasomal degradation, but its expression is maintained in stem‐like cells. Its close homologue, Numb‐like (NumbL), is able to regulate the CSC/TIC pool by inhibiting the Notch pathway.125 Numb‐1‐ and Notch‐specific T‐cells eliminated luminal CSC/TIC‐like cells in a breast cancer model.126 In a gynaecological cancer (endometrioid adenocarcinoma), ALDH^high cells preferentially expressed Hsp27, the over‐expression of which is associated with poor prognosis.127 Mitogen‐activated kinase (MAPK)13, pituitary tumour‐transforming gene protein‐binding factor (PTTG1IP), calpain 1 (CAPN1) and ubiquilin 2 (UBQLN2) were also found highly expressed in these cells compared with that in ALDH^low cells.128 SOX2, a transcription factor that regulates the Wnt/β‐catenin pathway, is amplified in numerous solid tumours and expressed predominantly in CSCs/TICs.129, 130 Interestingly, class I histone deacetylase (HDAC) inhibitors increase the frequency of SP cells with CSC/TIC markers, due to de‐differentiation of cancer cells. HOXA5 was shown to be the main transcription factor responsible for de‐differentiation and induction of SOX2 in lung cancer cells.131

Differentiation antigens

Many differentiation antigens are expressed in primary cancer cells as well as in CSC, such as MUC‐1 in breast cancer, tyrosinase and gp100 in melanoma. The human telomerase reverse transcriptase (hTERT) is also a target in primary cancer cells and in stem cells. T‐cells have been generated against a HLA‐A3‐restricted epitope: peptide K973 (KLFGVLRLK).132 Successful targeting of these antigens has the potential of eliminating CSCs and non‐CSCs alike.

Neoantigens

Although CT and onco‐fetal antigens can also be considered as neoantigens, this category typically includes mutated tumour antigens. The mutations may generate entirely new T‐cell epitopes for which high‐affinity T‐cell receptors have not been deleted during development and thus these T‐cells can be efficiently expanded. High mutation index has been indicated to be the underlying mechanism behind successful immune checkpoint inhibitor therapy.133 Although it is as yet unclear whether the mutation index of CSCs is different from that of primary cancer cells, there are data available that neoantigens are present in colorectal cancer both in CSCs and non‐CSCs in a manner that is targetable with T‐cells.134 Identifying CSC‐specific neoantigens may lead to further breakthroughs in cancer immunotherapy.

Immune targeting CSCs/TICs – clinical applications

Antigen‐specific immunological targeting of CSCs/TICs requires the generation of primary T‐cell responses, reactivation of memory responses or the adoptive transfer of engineered antigen‐specific T‐cells into the host.

Generation of T‐cell responses

Most clinical trials, listed on the http://clinicaltrials.gov website that use immunotherapy and target cancer stem cells, are based on isolating CSCs/TICs from solid tumours and loading them onto DC, which are then used as a cancer vaccine. The approach is based on preclinical data showing the success of this approach in immunocompetent mice.66 Trials are listed in pancreatic, nasopharyngeal, colorectal, ovarian, lung, liver and brain tumours. One trial studied the outcome of vaccination with DC, transfected with hTert and survivin as amplified ovarian CSC mRNA (NCT01334047). The stem‐like cell‐associated antigen(s), serving as a vaccine in these trials, are undetermined and likely to be individual patient‐specific. The outcome of these trials is not yet publicly available; however, as with non‐CSCs‐DC vaccines, it is unlikely that any monotherapy alone will be hugely effective. The explosion of approved immunotherapies with checkpoint inhibitors opens up the way for designing combination therapies. However, a shift in treatment design must be considered: vaccines or redirected T‐cells should be targeting multiple antigens, including CSC‐specific ones; this treatment should then be combined either with appropriate chemotherapy based on patient stratification data or high‐dose radiation and/or immune checkpoint inhibitor treatments in order to provide the best chance for generating robust and long‐lasting T‐cell responses leading to tumour rejection.

T‐cell transfer

The success of chimeric antigen receptor (CAR) T‐cells in treating haematological malignancies (https://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm574058.htm) has generated interest in using the approach against solid tumours and also CSCs. CAR T‐cells express a CAR which consists of an extracellular binding domain of a single‐chain fragment of the antibody variable region (scFv), providing antigen‐specificity and the intracellular signalling domains of CD3‐zeta chain. The receptor can be coupled with co‐stimulatory molecules, such as CD28 and CD137. Cell surface antigens, expressed on CSCs, such as CD44, CD133, aldehyde dehydrogenases (ALDH) and EpCAM can be targeted in this way. Preclinical work using these CAR T‐cells in glioblastoma, prostate cancer and gynecological tumours have been encouraging (reviewed by Guo135). Clinical trials are currently ongoing against EGFR and CD133 (NCT01869166 and NCT02541370).

Disclosures

The authors have no competing interests to declare.