CAR‐T therapy: Prospects in targeting cancer stem cells

Abstract Cancer stem cells (CSCs), a group of tumour cells with stem cell characteristics, have the ability of self‐renewal, multi‐lineage differentiation and tumour formation. Since CSCs are resistant to conventional radiotherapy and chemotherapy, their existence may be one of the root causes of cancer treatment failure and tumour progression. The elimination of CSCs may be effective for eventual tumour eradication. Because of the good therapeutic effects without major histocompatibility complex (MHC) restriction and the unique characteristics of CSCs, chimeric antigen receptor T‐cell (CAR‐T) therapy is expected to be an important method to eliminate CSCs. In this review, we have discussed the feasibility of CSCs‐targeted CAR‐T therapy for cancer treatment, summarized current research and clinical trials of targeting CSCs with CAR‐T cells and forecasted the challenges and future direction from the perspectives of toxicity, persistence and potency, trafficking, infiltration, immunosuppressive tumour microenvironment, and tumour heterogeneity.


| INTRODUC TI ON
The hypothesis that tumours originated from "stem cells" was first proposed about 150 years ago. However, relevant research had been progressing slowly until a growing number of experimental data strongly supported the tumour stem cell hypothesis which has changed views on tumourigenesis and tumour cell biology. 1,2 In cancer stem cell theory, tumours originate from a small portion of cancer stem cells (CSCs), and they have the capacity of immortal proliferation and multi-lineage differentiation, which drives tumour formation, growth, recurrence, metastasis, drug resistance, chemo/ radio-resistance and other malignant phenotypic characteristics. 3 In this context, while CSCs are resistant to radiotherapy, chemotherapy and certain targeted therapies, the key to cancer treatment remains in CSCs. In addition, because of the special tumour-killing mechanism, cancer stem cell may be more sensitive to immunotherapy, and in-depth study of CSCs characteristics may also significantly promote the development of tumour immunotherapy. 4 Recently, the clinical application of chimeric antigen receptor T (CAR-T)-cell therapy has made an unprecedented breakthrough in the treatment of haematological diseases. 5 The safety and feasibility of CAR-T therapy in the treatment of solid tumours have also been confirmed. 6 For CAR-T therapy, T cells from the patients will be genetically engineered to express chimeric antigen receptors (CAR), and then be adoptively transferred back to patients. The genetically engineered CAR-T cells will recognize the surface antigen of tumour cells and selectively target and kill those tumour cells. 7 CAR-T cells can recognize the target antigen independently of MHC restrictions. 8 After the recognition, CAR-T cells fix their position specifically in the tumour site and can have sustained persistence for a while. The successful application of CAR-T cells in cancer treatment represents a milestone in anticancer therapy.
Cancer stem cells have some unique characteristics, such as slow rate of division, high expression of drug efflux pumps, 9 heightened activation of DNA repair mechanisms 10 and microenvironment characteristics: hypoxia and acidosis, 11 which is due to the expression of specific surface markers. These surface markers can be used as specific targets for CAR-T therapy to eliminate CSCs. In addition, the expression of MHC molecules on the surface of CSCs is low, which causes MHC restriction when immunotherapy is used to target CSCs. 8 However, in CAR-T therapy, CAR-T cells can recognize the target antigen with no MHC restrictions, 8 which endows some advantages for the application of CAR-T therapy to eliminate CSCs.
In this review, we analysed the feasibility of targeting CSCs by using CAR-T cells, summarized published studies on CSCs-targeted CAR-T therapy, pointed out the challenges of targeting CSCs by CAR-T cells related to toxicity, persistence and potency, trafficking, infiltration, immunosuppressive tumour microenvironment, tumour heterogeneity and purposed promising strategies, such as novel CAR containing a JAK-STAT signalling domain, modulation of chemokine signalling, directing CAR-T cell to target vascular endothelial growth factor receptor 2(VEGFR2), combining CSCs-targeted therapy with FDAapproved PD-1/PD-L1 checkpoint inhibitors, multi-target CAR-T cell therapies and transgenic modification of the CAR structure, for the future development of CSCs-targeted therapy.

| G ENER ATI ONOFC AR-TCELL SFOR C AN CERIMMUNOTHER APY
Chimeric antigen receptor T-cell therapy has emerged as a novel therapeutic T-cell engineering practice, in which T cells derived from patient blood were engineered in vitro to express artificial receptors to target a specific tumour antigen, then the modified T cells will be adoptively infused back to the patient's body to fight against cancer. 12 For the preparation of CAR-T cells, activated T cells were infected with retroviruses or lentiviruses loaded with CAR sequences to express receptors, and these modified T cells can recognize tumour-associated antigens and express the tandem co-stimulation molecular signal transduction fragments which were related to T-cell activation. 13 The engineered CAR-T cells were expanded in vitro and then infused into patients to fight against tumours. In 1989, Gross et al. first proposed the concept of CAR-T cell therapy. 14 At present, this therapy has made breakthroughs in clinical trials for the treatment of leukaemia, and has gradually extended to the clinical treatment of solid tumours. 15 The development of CAR-T structure has gone through four generations. Each generation of CAR-T structure is modified by adding more components in the intracellular space to make it more specific, efficient and durable ( Table 1). The first generation of CAR-T cells composed of the single-chain variable fragment (scFv) and an intracellular CD3ζ signalling domain for T-cell activation mainly solved the problem of targeting, but it lacked complete costimulatory signals and cannot fully activate T cells which limited its antitumour activity. 16 Subsequently, second-and third-generation CARs were invented, which included one or two costimulatory domains respectively. The second-generation CAR-T cells had one costimulatory domain -CD28 or 4-1BB, which effectively improves the tumour-killing effect. 17 The third-generation CAR-T cells carried two costimulatory molecular domains, which significantly increased the proliferation activity of CAR-T cells and enhanced cytokines release, which improves the in vivo persistence of CAR-T cells and results in a stronger cytotoxic activity. 18 More recently, the fourth-generation CAR-T cells, also called TRUCK-T (CAR redirected T cells that deliver a transgenic product to the targeted tumour tissue) cells, were engineered to secrete specific cytokines, such as IL-12, IL15, IL-18, CCL19 and IL-7, so as to overcome the suppression from the tumour immune microenvironment, recruit and activate the second wave of immune cells to produce an immune response. 19 In addition to the evolution of CAR designs outlined above, some elements with regulatory functions can be added with the expression of an "armour" protein, by introducing suicide-initiated, negative-regulatory and switch-initiated components into CAR-T

Generation Theevolutionofchimericantigenreceptors(CARs)
1st generation First-generation CARs contain the CD3ζ chain of the T-cell receptor complex 2nd generation Second-generation synthetic antigen receptors differ from the first generation by the addition of a costimulatory domain (either CD28 or 4-1BB).
3rd generation Third-generation CARs contain two costimulatory domains, respectively, such as CD28 and OX40.
4th generation Fourth-generation CARs, the so-called TRUCKs or armoured CARs which are additionally modified with a constitutive or inducible expression cassette for a transgenic protein, which is released by the CAR-T cell to modulate the T-cell response.
Other evolution Introducing some regulatory elements into CAR-T cells, which include suicide-initiated, negative-regulatory and switchinitiated components, or using dual antigen-targeting CARs and inhibitory CARs. CAR-T cell design  cells to have better control of the cytotoxic response of engineered   CAR-T cells on tumours. 20 In recent years, some new types of CAR-T   cells have been developed to increase the safety and therapeutic   effect of CAR-T cell therapy, such as dual antigen-targeting CARs   which improved specificity through targeting multiple antigens, and inhibitory CARs which were engineered to inhibit T-cell activation upon binding to an antigen expressed on non-malignant cells instead of tumour cells 21,22 (Table 1).

TA B L E 1 Architectural evolution of
The CAR-T cell therapy currently used in clinical practice is based on the second-generation CAR-T cells and mainly targets B-cell-related diseases, while the clinical application of the thirdor fourth-generation CAR-T cells is at the early stages. 23 Since the infusion of allogeneic T cells is prone to cause human immune rejection, CAR-T treatment currently uses patient's own T cells. As shown in Figure 1, the general treatment process can be

| ANALYS E SOFTHEFE A S IB ILIT YAND ADVANTAG E SOFTARG E TINGC SC SWITH C AR-TTHER APYFORC AN CERTRE ATMENT
Among many anticancer therapies, the primary problem impediment against cancer curability is tumour recurrence, which is mainly caused by the presence of CSCs. 27 Recently, cell therapy represented by CAR-T cells has shown strong curative effects in tumour treatment and has the tendency to become an essential tumour treatment strategy. 28 Based on the characteristics of CAR-T cells and CSCs, we proposed that targeting CSCs with CAR-T therapy is feasible, and analysed the advantages and feasibility of this promising therapy. Harnessing the power of the immune system to target CSCs is a promising therapeutic approach. For instance, J.C. Sun et al. (2010) applied dendritic cell-based vaccines, which were treated with antigens from CD133+ hepatocellular carcinoma cells to activate specific cytotoxic lymphocytes, and therefore destroyed hepatocellular carcinoma CSCs. 36 In the last decade, cell-based immunotherapy represented by CAR-T therapy is being considered as an efficient approach for the treatment of cancer. 6 This is a way to eliminate tumour stem cells using the power of the body's own immune system, based upon the principle of targeting the surface markers of CSCs.

| Boostthebody'sownimmunesystemto eliminateCSCs
CAR-T cells eliminate CSCs relying on body's own immune system, and the infusion of millions of exogenous modified T cells can highly enhance the body's immune function.

| TheuniquecharacteristicsofCSCsare suitableforCAR-Ttherapy
Targeting CSCs may realize tumour radical eradication, while the CSCs are protected by their unique characteristics, such as infrequent replication, enhanced drug resistance and heightened activation of DNA repair mechanisms as we mentioned in part 2. However, some of the CSCs characteristics may be harnessed to eradicate CSCs. For example, ATP-binding cassette subfamily B member 5 (ABCB5), a marker of CSCs in a number of malignancies and a drug efflux transporter which associates multidrug resistance, tumour progression and recurrence, can be used in the tumour eradication. 37,38 Treatment with anti-ABCB5 monoclonal antibodies has been shown to inhibit tumour growth in xeno-transplantation models which prove that ABCB5 could be a good target for CSCs eradication. 39 Furthermore, as ABCB5-reactive CD8+ T cells are present in the peripheral blood of melanoma patients and an ABCB5-specific response can be induced in vitro in naive donors, which implicate that ABCB5 could be a potential target for cancer immunotherapy. 40 The subcellular location for ABCB5 expression is on cell membrane, which creates conditions for CAR-T therapy targeting at ABCB5. 41 Interaction between CAR and ABCB5 helps to the formation of immune synapse, through which the contact-dependent cytotoxicity may occur.

| CAR-TisanMHC-independentadoptive cellularimmunotherapy
CAR-T therapy stands out among many CSCs-targeted therapies for it is an MHC-independent adoptive cellular immunotherapy. In 1975, Doherty and Zinkernagel first proposed the phenomenon of "MHC restriction" -viral peptides can only be recognized by T cells when combined with specific MHC molecules. 42 MHC is an important component of the immune system and plays a key role in antigen presentation, enabling specific T lymphocytes detect foreign antigen. 43 However, the expression of MHC molecules on CSCs is lower, which may prevent the body from boosting immune system to eliminate CSCs. 44 Fortunately, CAR-T therapy is a MHC-independent adoptive cellular immunotherapy for the unique structure of single-chain variable fragment (scFv), mainly formed by variable regions of heavy and light chains, which can recognize cell surface antigens directly and specifically, instead of being restricted by the down-regulation of MHC molecules. 45 Therefore, although CSCs are not easily eliminated by the immune cells from the immune system of patients, it is feasible to eliminate CSCs through CAR-T therapy.

| TheexistencesurfaceantigensofCSCsthat canbetargetedbyCAR-Tcells
Chimeric   Compared with targeting the specific surface antigens, engineering the corresponding CAR-T cells to target "general" CSCs markers is another way to eliminate CSCs with CAR-T therapy. General antigentargeted CAR-T cells may not be initially designed to specifically kill

| REP ORTEDL ABOR ATORYRE S E ARCH OFC SC S -TARG E TEDC AR-TTHER APY
CSCs. Due to the expression of these markers was also detected on the surface of CSCs, when such CAR-T cells were co-cultured with corresponding CSCs, they also have a cancer-killing effect ( Table 2).
HER2 is a tumour-associated antigen that is expressed by up to 80% of glioblastomas (GBMs) but not by normal postnatal neurons or glia. 53  is also expressed on GSC and demonstrate that anti-CSPG4 CARtransduced T cells recognize and kill these GSC. 55 Similarly, the expression of NKG2DLs is usually expressed in most epithelial-derived tumour cells, such as ovarian cancer, colon cancer and leukaemia, while it is rarely detected in healthy adult tissues. 56  did not exhibit significant toxicity in the mice model. 47 The above experimental results indicated that adoptive cellular immunotherapy with CSCs-targeted CAR-T cells is expected to become a promising cancer treatment method. For other common markers of CSCs, such as CD90 and ALDH that could be theoretically ideal targets for CAR-T therapy, unfortunately, there is no reported studies using CD90 or ALDH-specific CAR-T cell for cancer treatment.

| CLINI C ALAPPLI C ATI ONOFC SC S -TARG E TEDC AR-TTHER APY
To investigate the clinical application of CSCs-targeted CAR-T therapy, we have searched the ClinicalTrials.gov website and summarized the latest registered clinical trials of CAR-T therapy using surface markers of CSCs (Table 3)

| CHALLENG E SANDFUTURE DIREC TIONS
Although some successful animal experiments conducted with

CSCs-targeted CAR-T cells have been reported and some ongoing
CSCs-targeted CAR-T therapy clinical trials have shown good tumour treatment prospects, many challenges in clinical application exist. (Table 4).

| Toxicity
One of the limitations in CAR-T therapy is on-target off-tumour toxicity, caused by the direct attack on normal cells which have the shared expression of the targeted antigen. 61 On-target/off-tumour toxicity becomes a major hindrance of CSCs-targeted CAR-T therapy, because in normal cells, some CSCs markers are found, such as CD133 expressed in normal neural stem cells or ALDH expressed in hematopoietic stem cells. 62,63 To reduce the toxicity, selecting a safer antigen of CAR-T cells is needed or designing dual-targeted CARs to enhance the tumour specificity of CAR-T cells may work. 64 Toxicity can also be minimized by local (intratumoural) delivery of CSCs-targeted CAR-T cells. 65 Moreover, introducing suicidal genes as a "safety switch" in CAR-T cells when adverse reactions are uncontrollable may limit on-target, off-tumour toxicities. 66 Similarly, modifying CAR-T cells and enabling them to express an inhibitory chimeric antigen receptor, such as CTLA-4 or PD-1, can achieve antigen-specific suppression of T-cell cytotoxicity, cytokine release and proliferation. 21 Adverse events other than on-target/off-tumour toxicity include cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS), which were common toxicities of CAR-T cells in treating tumours. 67 In the first several days after CAR-T cell infusion, CRS is mostly found and patients may have fever, hypotension and tachycardia which might lead to haemodynamic instability, causing end-organ injury; after the onset of CRS, neurotoxicity syndrome, which manifests as subtle cognitive decline, may occur. 12 CRS is associated with elevated IL-6 levels in patients receiving CAR-T therapy and anti-IL-6 receptor antagonist tocilizumab is, thus, used to treat CRS. For instance, FDA sanctified the use of the drug in the treatment of CRS when the first CAR-T cell product was approved. 68 Corticosteroids, which suppress immune responses, are also commonly used in the management of the toxicity once the patient does not have a rapid response to IL-6 receptor blockade. 69 Alternatively, therapeutic options for ICANS are corticosteroids, antiepileptics and care measures with intensive care unit (ICU) monitoring. 12

| Trafficking
Apart from effectively treating haematological tumours, the CSCs- • Engineering bispecific CAR-T cells by designing a single CAR molecule with two (or more) distinct binding domains 103 • Multi-target CAR-T cell therapies: creation by mixing different CAR-T cell products targeting single antigens prior to infusion, or transducing T cells with multiple CAR constructs 13 • CAR-T cells expressing bispecific T-cell engagers (BiTEs) to recruit bystander T cells against a second tumourassociated surface antigen [104][105][106] are usually surrounded by compact stroma and tumour cells. 80 We proposed the following methods currently used to improve the trafficking or homing ability of CAR-T cells so as to overcome the challenge. The most straightforward method is to directly infuse the CSCs-targeted CAR-T cells into tumour sites. In a previous report, the local infusion of CAR-T cells resulted in significant regression of glioblastoma. Nevertheless, localized therapy is not suitable for many metastatic solid tumours. 81

| Immunosuppressivetumour microenvironment
Cancer stem cells survive in an immunosuppressive tumour microenvironment (TME) composed of vascular niches, cancer-associated fibroblasts, cancer-associated mesenchymal stem cells, hypoxia, tumour-associated macrophages and extracellular matrix, which hinders the direct killing of tumour stem cells by one's own immune cells and the adoptive CAR-T cells. 90 Therefore, combining CSCs-targeted CAR-T therapy and the strategy of targeting the immunosuppressive TME of CSCs may help improve the efficiency of CSCs removal. As we mentioned in "persistence and potency" section in this review, cytokine support served as one of the important signals for optimal T-cell activation and proliferation. However, this signal was lacking in the TME of CSCs. 91 Constructing CSCstargeted CAR-T cells which overexpress IL-12, IL-18, IL-7, IL-15 and   IL-21 cytokines may be an effective way to provide support for the activation, proliferation and killing of CSCs of CAR-T cells in immunosuppressive TME. 78 Several studies have proved that CSCs have the ability to evade the immune system, because these cells secrete several substances into the TME, such as TGFβ, IL-10, IL-4 and IL-13, which exert inhibitory effects on an array of immune cells. 3

| Heterogeneity
One great challenge in cancer therapy is intratumour heterogeneity, while CSCs are one of the determining factors causing the problem. 98 Therefore, eradication of CSCs by CAR-T therapy is promising for overcoming the heterogeneity. However, accumulating evidence suggests that CSCs represent phenotypically and functionally heterogeneous populations, which has been found in colorectal, 99 colon, 100 hepatocellular 101 and breast cancer stem cells, 102 leading to antigen loss or escape when applying CSCs-targeted CAR-T therapy.
Given that CD19/CD22 bispecific CAR-T cells have demonstrated clinical efficacy in patients with B-cell malignancies, 103 bispecific CAR-T cells can be bio-engineered by designing a single CAR molecule with two (or more) distinct binding domains of CSCsspecific markers, so as to overcome antigen escape caused by CSCs heterogeneity. Furthermore, multi-target CAR-T cell therapies can be created to overcome the limitation of antigen loss by mixing different CAR-T cell products targeting single antigens prior to infusion or by transducing T cells with multiple CAR constructs. 13 Transgenic modification of the CAR structure to elicit an endogenous immune response through recruiting additional effector cells is an alternative approach to avoid the heterogeneity.
To recruit bystander T cells against a second tumour-associated surface antigen, CAR-T cell targeting can be combined with the release of bispecific T-cell engagers (BiTEs). 104  Finally, we discussed the current challenges of this therapy and the solutions that can be adopted for the future development of CSCstargeted therapy. As the first therapy mentioned in the review has the potential to cure cancers and is currently on the market, harnessing CAR-T cells to target CSCs is believed to achieve greater success in treating tumours.

ACK N OWLED G EM ENTS
We thank members in Zhao laboratory for providing the platform to conduct research related to CAR-T therapy, and we are grateful to Dr. Yang Dong and Dr. Sun Bin for their previous technical support.

CO N FLI C TO FI NTE R E S T
Conflict of interest relevant to this article was not reported.