3.2. Innate Immune Effectors Targeting CSCs
3.2.1. NK Cells
Natural killer (NK) cells are large granular lymphocytes, constitute 5–15% of circulating lymphocytes, and represent the main effectors belonging to the innate immunity . NK cells recognize tumor cells and infected cells, in a HLA independent manner without prior sensitization  and release pre-formed granules known as perforin and granzyme B, which can induce necrotic as well as apoptotic or programmed cell death in target cells [123,124,125]. NK cells simultaneously express activating and inhibitory receptors that encounter target cells by the subtle balance of transmitted signals for activation or inhibition . NK cells mediate direct and antibody-dependent cellular cytotoxicity (ADCC) against tumors and regulate the function of other cells through the secretion of cytokines and chemokines .
Adoptive NK cell therapy has been explored with either autologous or allogenic NK cells [128,129]. Ex vivo NK cell culture is demanding because of their limited life span and expansion potential . Established human NK cell lines (e.g., NK92, KHYG-1, NKL and NKG) can also be explored as a valuable alternative to primary NK cells. NK92 cell line in particular is approved by the US Food and Drug Administration (FDA) for phase I and II clinical trials, allowing a “off-the-shelf” CAR.NK92 production [131,132]. NK92 can easily be expanded to high numbers and maintained for therapeutic use in the presence of IL2, while retaining consistent phenotypic and functional features .
Increasing data demonstrate that NK cells can selectively identify and lyse CSCs [134,135], as they have low or no MHC-I molecules but up-regulate the ligands for NKG2D, DNAM1 and NKp30 NK-activating receptors [134,136,137]. Several studies highlighted NK cell ability to recognize and kill poorly differentiated tumors [138,139,140] and that cytokine-activated (IL2 and/or IL15-activated) NK cells were effective against human breast, colon, melanoma and glioblastoma CSCs [141,142,143]. Studies on oral squamous carcinoma reported higher levels of NK cell activating ligands on CSCs as compared to non-CSCs, resulting in their higher sensitivities to NK cell killing . In particular, Tallerico and colleagues observed lower levels of MHC class I expression on colorectal cancer CSCs compared to non-CSCs. They demonstrated that CSCs showed increased susceptibility to NK killing , linked to upregulation of the activating natural cytotoxicity receptors, particularly NKp30 and NKp44. Castriconi and colleagues demonstrated low or absent expression of MHC class I molecules on GBM-derived CSCs and their high susceptibility to both allogeneic and autologous NK cells in co-culture models after pre-treatment with IL-2 and IL-15 . In melanoma, Pietra and colleagues reported that both CSCs and non-CSCs showed sensitivity to activated allogeneic NK cells, possibly mediated by the DNAM-1 ligands Nestin-2 and PVR . In breast cancer, Yin and colleagues reported that CSCs showed to be lysed by IL-2 and IL-15 activated NK cells, and such cytotoxicity was likely mediated by the increased expression of NKG2D ligands ULBP1, ULBP2 and MICA on CD44+CD24− CSCs . On the other side, CSCs can evade NK cell killing by shedding MICA and MICB and also by recruiting immunosuppressive Treg cells [147,148]. In glioma, CD133+ brain CSCs do not express either detectable MHC-I and NK cell activating ligands, escaping NK cell-mediated-immune surveillance . IFN-γ stimulated the expression of these molecules on CD133+ CSCs, restoring their sensitivity to NK cell-mediated lysis in vitro . In a preclinical study high-grade non-muscle invasive bladder cancer (NMIBC), NK cells from healthy donors, but not from NMIBC patients, upon activation with IL-2 and IL-15, could kill both CSCs and bulk tumor cells, promoting their differentiation and enhancing the efficacy of a possible combined chemotherapy . In oral squamous carcinoma, a preclinical study proposed a novel NK ex vivo expansion and activation strategy, based on co-culture with osteoclasts to generate “super-charged” NK cells endowed with higher secretion of IL12 and IL15, increasing killing capability against CSCs .
Evasion mechanisms induced by TME can block NK cell mediated-CSC lysis through an increase in IL6 and IL8 secretion and decrease in IFNγ secretion. NK cells can reach a functional state, known as “split anergy”, characterized by a reduced NK cell cytotoxicity maintaining cytokine and chemokine secretion. This NK functional state decrease NK cytotoxic activity against CSCs but induce CSCs differentiation trough cytokine production, especially IFNγ and TNFα . This phenomenon was found to be associated with an increase in MHC class I, PD-L1, and CD54 expressions and a reduction in CD44 levels on tumor cells .
NK cells are a promising approach to target both CSCs and non-CSCs, leading to prolonged therapeutic results. NK cells targeting of CSCs may initiate and amplify adaptive T cell-mediated responses. Previous clinical trials using NK cells as monotherapy in solid tumors obtained modest results, new trials point to a new application of NK cells in combination with traditional treatments in order to overcome the therapeutic resistance which CSCs may contribute .
Different promising strategies are based on antibody anti-CSC markers, such as CD44, CD24, CD133 and ALDH-1  and bispecific antibody concomitantly binding CD16 on NK cells and CD133 on colorectal CSCs, that significantly improved CSC targeting ability by NK cells . Finally, NK cells administration and concomitant inhibitory killer immunoglobulin receptor (KIR)-blockade, with or without other cancer drugs, may represent new opportunities for cancer patients . CAR engineering of NK cells can enhance their specific recognition and elimination of tumor cells, providing an opportunity to generate NK-cell therapeutics of defined specificity. Numerous preclinical studies demonstrated the successfully generation of CAR-NK cells [157,158]. Furthermore, several groups improved CAR-NK activity including in the receptor construction one or more signaling domains derived from CD244 (2B4), NKG2D, DAP10 or DAP12 [157,159,160]. CAR.NK cell therapy presents several advantages compared to CAR-T cells: a) reduced on-target/off-tumor toxicity and low cytokine storm risk as they have limited in vivo persistence b) CAR-NK cells, endowed with innate killing activity, can attack tumors with heterogeneous expression of the CAR target antigen . Preclinical study demonstrated that CAR-NK cells targeting specific antigens linked to CSCs (e.g., GD2, EGFRvIII, ErbB2, CD133, PSCA) displayed superior anti-tumor activity compared to parallel-unmodified NK cells. CAR-NK cells against prostate stem cell antigen (PSCA) displayed in vitro anti-tumor efficacy against PSCA+ CSCs, and GD2-CAR.NK showed cytotoxic activity against neuroblastoma and Ewing sarcoma cells . GD2-CAR.NK92 cells have been tested in preclinical assays against neuroblastoma, melanoma, breast carcinoma and Ewing sarcoma, demonstrating selective anti-tumor activity [160,162,163,164]. Different preclinical approaches employ CAR-NK cells for GBM immunotherapy . Stem-like GBM cells seem to be more sensitive to natural cytotoxicity of NK cells, as CSCs showed increased expression of ligands for activating NK cell receptors and down-regulated class I HLA ligands for NK cell inhibitory receptors  Preclinical data with NK92 cells showed that ErbB2-CAR.NK92 cells lysed ErbB2-positive stem-like GBM cells growing as neurospheres quite rapidly. EGFRvIII-CAR.NK92 inhibited tumor growth and extended survival of GBM xenograft. Other studies highlighted that ErbB2-CAR.NK92 prolonged survival and induced reduction of primary tumors and metastasis in breast cancer and GBM xenografts, while parallel-unmodified NK92 cells were unable to inhibit tumor progression [159,167,168]. CD133-CAR.NK92 have been explored in vitro against GBM and ovarian cancer in combination with cisplatin, demonstrating efficient anti-tumor activity . A phase I clinical trial (CAR2BRAIN, protocol number NCT03383978) based on intracranial injection of ErbB2-CAR.NK92 in patients with recurrent ErbB2+ GBM is currently ongoing [168,170,171].
3.2.2. CIK and NKT Cells
CIK cells are heterogeneous ex vivo lymphocytes featuring a mixed T- and NK cell phenotype, generated and expanded in vitro from peripheral blood mononuclear cells (PBMC) and endowed with MHC-independent antitumor activity [172,173,174,175,176,177]. CIK cells can be easily and efficiently expanded with the timed addition of IFN-γ, antibody (Ab) anti-CD3 and IL2 [178,179]. At the end of the expansion, the CD3+CD56+ cells represent the subset with the most potent cytotoxic activity against multiple tumor types [180,181]. The cytotoxic activity is primarily mediated by the interaction between the activating natural killer cell receptors of CIK cells, in particular NKG2D, and the corresponding stress-inducible ligands, including MIC A/B and ULBPs [182,183,184]. The clinical activity and safety profile of CIK cells was demonstrated in several clinical trials in both hematological and solid settings [175,185,186,187,188,189].
Immunotherapy based on CIK cells may overcome limitations caused by tumor downregulation of MHC molecules on CSCs and may be advantageous over T cells endowed with MHC-dependent activity. Further, CIK cells may be a valuable therapeutic strategy applicable to all patients, regardless their HLA-haplotypes.
Patient-derived CIK cells showed a potent killing ability against CSCs in preclinical in vitro and in vivo studies against putative melanoma, sarcoma, hepatocellular carcinoma (HCC) and nasopharyngeal carcinoma (NPC) CSCs , with possible relevant clinical implications. In these studies, CIK cells were equally effective against both putative CSCs and non-CSCs. In distinct studies putative CSCs were visualized with a promoter-fluorescence reporter gene strategy, where cancer cells were transduced with a lentiviral “CSC-detector” with GFP gene under control of the stem cell-specific Oct4 or Nanog promoter [191,192,193]. Based on this methodology, CIK cells showed to be effective to kill CSCs surviving to chemotherapy or targeted therapy either in in vitro and in vivo preclinical studies.
In another investigation, CIK cells sensitized by EpCAM and CD44 peptide DCs were effective in vitro against Prostate Cancer Stem like Cell (PCSC)-enriched prostate-spheroids and in vivo against PCSC-enriched prostate-spheroid xenografts .
Recent evidences showed the success of CIK cell redirection by CAR strategy to enhance CIK cell anti-tumor efficacy [26,195,196,197,198,199]. CAR.CIK are an appealing platform for CAR engineering, as they may generate bipotential killers combining the specificity of CAR with their intrinsic tumor killing capacity [200,201].
In recent years, preclinical studies reported first evidences of enhanced CAR.CIK activity in high grade soft tissue sarcoma (STS) thanks to the expression of a CAR specific for CD44v6 antigen or CSPG4 [26,27], and in NPC thanks to a CAR specific for 5T4 antigen . These CAR.CIK cells could efficiently eliminate tumor cells and also stem cell-like cells in vitro, as these tumor antigens are shared with CSCs and involved in tumor initiating process, EMT and clinical aggressiveness. In addition, a first evidence showed in vitro the anti-tumor efficacy of CIK cells expressing a CAR specific for CSPG4 antigen in high grade STS . It has been widely described that CSPG4 has a key role in several oncogenic pathways required for malignant progression and metastatization and is overexpressed by tumor cells and CSCs . Overall CAR-CIK cells have gradually become a realistic new option of cancer immunotherapy and are studying in vitro and in vivo as a potential effective platform against a wide variety of cancers. Further preclinical and clinical investigation are needed to evaluate the potential of targeting putative CSCs with CIK cells, also in synergism with other therapeutic strategies.
Natural killer T (NKT) lymphocytes respond rapidly to a wide variety of glycolipids and stress-related proteins and share properties of both T and NK cells, such as CD56, CD16 expression and granzyme and perforin productions [204,205]. In contrast to CIK cells, they are already present in small percentage in blood circulation; NKT cells express an invariant αβTCR that recognizes antigens presented by MHC class I CD1d molecule, as glycolipids. NKT cells are involved in anti-tumor immunity acting as recruiter of adaptive immune cells trough their rapid cytokine secretion . Due to their restriction to the monomorphic HLA-like molecule CD1d, but not to HLA, NKT CAR cells show potential for enabling off-the-shelf cancer immunotherapy, even if dedicated clinical trials have not yet been reported. NKT cells may be isolated from patients or allogenic donor and are most commonly expanded with the glycolipid α-GalCer, transduced to express a tumor-specific CAR and reinfused in cancer patients with a favourable safety profile based on absence of alloreactivity and limited in vivo persistence [207,208].
Few preclinical studies reporting the cytotoxic activity of NKT cells against CSCs are based on redirection with CARs. In neuroblastoma and B cell lymphoma, GD2-CAR.NKT cells efficiently localized at tumor site, reduced tumor growth and prolonged survival of xenograft models, targeting also GD2+ CSCs . CSPG4-CAR.NKT cells displayed similar efficient cytotoxicity compared to conventional CAR.T cells redirected by CSPG4-CAR .
Currently, an ongoing phase I clinical trial is exploring efficacy and persistence of autologous GD2-CAR.NKT cells in neuroblastoma patients (GINAKIT, NCT03294954).
3.2.3. γδ T Cells
γδ T lymphocytes are unconventional non-MHC-restricted T cells, characterized by an invariant γδTCR. They represent 1–5% of circulating lymphocytes and are a significant subset of resident T cells in lymphoid organs, epidermis, gastrointestinal mucosa, and reproductive system . The Vγ9Vδ2 phenotype is the most represented of peripheral blood γδ T lymphocytes, while δ1 e δ3 are mostly tissue-located [211,212,213]. Γδ T cells recognize stress inducible molecules and are characterized by their ability to recognize early metabolic changes, recognizing non-peptide metabolites like phosphoantigens or aminobisphosphoantigens and the cholesterol precursor isopentyl pyrophosphate, that differentiate healthy cells from transforming ones . γδ T cell protection against cancer occurs mainly by the production of pro-inflammatory cytokines such as IFNγ, TNFα, and IL-17 and through their cytotoxic activity .
In preclinical studies, Vγ9/Vδ2 T cells efficiently killed CSCs derived from colon cancer , ovarian cancer , and neuroblastoma  but were less effective against prostatic CSCs  and breast cancer . CSCs derived by breast cancer showed to be hypo-responsive to γδ T-cell targeting. Breast CSCs are characterized by increased levels of PD-L1, anti-apoptotic MCL-1 and MICA shedding compared with non-stem counterpart. In vitro either PD-1 blockade or treatment with MCL-1 degrader or proteolytic cleavage inhibitor (ADAMi, GW280264X) were able to restore breast CSCs sensitivity to γδ T-cell cytotoxicity . In breast cancer, γδ T cells could kill CSCs, expressing relatively low levels of MHC-I and CD54, following pre-treatment with γδ T-cell agonist zoledronate. Zoledronate exposure increased γδ T cells proliferation rate, TNFα and IFNγ secretion, granzymes production, expression of CD69 molecule and tissue-homing chemokine receptors CCR5 and CXCR3, while decreased lymphoid-homing chemokine receptors CCR7 and CXCR5 [216,218]. Zoledronate-activated γδ T-cells enhanced the killing activity of CD8+ T cells through the IFN-γ-mediated upregulation of MHC-I and ICAM-1 molecules . Combination therapy with γδ T cells and zoledronate is feasible in patients with different advanced solid tumors .
However, clinical trials stimulating γδ T cells or even transferring γδ T cells with or without activating stimuli into cancer patients show very low efficiency and very limited success [222,223,224]. This might be due to the lack of knowledge regarding the specificity and diversity of these cells. In breast cancer, synergism between CD8+ T cells and γδ T cells has been described in the eradication of tumor cells including CSCs: γδ T cells induced upregulation of MHC class I and CD54/ICAM-1 on CSCs, enhancing their susceptibility to CD8+ T cells . γδ T cells are expected to be associated with the same level of safety reported for CAR-NK cells  and may represent an intriguing T cell subset to exploit CAR redirection as a possible strategy to target CSCs.
An overview of the main cellular immunotherapy clinical trials directed against CSC-relevant targets is reported in Table 1.
4. Conclusions and Challenging Perspective
Biological and immunological characterization of CSCs as long as definition of their interaction with immune cells in the TME are crucial to set up more efficacious strategies and innovative anticancer therapies. New emerging methodologies, as single cell molecular analysis, may provide new insights in understanding the relationship among immune characters, stromal, cancer cells and CSCs in the TME and elucidate their heterogeneous contribution in tumor progression [226,227].
Important challenges need to be faced to develop new and more effective immunotherapy strategies capable to involve CSCs. The first is represented by the heterogeneity in CSC populations. Distinct CSC subpopulations expressing different phenotypic markers have been reported inside the same cancer type . This means that a specific immunological treatment could eliminate only a subset of CSC. Furthermore, CSCs may escape from antigen-dependent immunotherapies by lacking or decreasing the target density on their surfaces. In addition, the majority of the reported CSC markers and TAAs are not CSC-exclusive, and therefore identification of CSC-specific antigens is critical for the success of antigen-dependent immunotherapies, avoiding potential toxicities and achieving treatment specificity. For instance, in CAR based immunotherapies, one of the major difficulty is the possible development of on-target/off-tumor toxicity caused by CAR cell killing activity against normal cells .
The second important challenge is tumor cell plasticity. CSCs constantly evolve as well as tumor evolves and progresses, further CSCs evolve upon treatment. Tumor plasticity represents a huge hurdle in the development of durable targeted cancer therapies, as eradication of existing CSC populations might be followed by their regeneration, under treatment pressure, from non-CSC counterpart within the tumor . Promising results may derive from MHC-unrestricted approaches (e.g., NK and CIK cells), as they kill without HLA restrictions and might recognize ligands whose expression is induced by heterogeneous stimuli as stress, chemotherapies and other agents, overcoming issues arising from tumor plasticity and heterogeneity.
Another challenge is the CSC low immunogenicity and negative immunomodulating effects. CSCs are mainly resistant to conventional cancer therapies, as they can escape from antitumor immunity through lower expression of antigens and HLA recognized by immune cells.
Lower immunogenicity of CSCs may be enhanced by inhibiting negative immunoregulatory pathways and by upregulating HLA I and APM components through combination therapies with IFNs, chemotherapy, radiotherapy, and/or epigenetic treatments . A new intriguing possibility is represented by epigenetic therapies combined with immunotherapy, as epigenetic drugs modulate the expression of immune-related genes either on tumor cells and on tumor-associated immune cells .
All the reported CSC features may have contributed to the disappointing outcomes of current adoptive immunotherapies in solid tumors. Strategies that combine conventional anti-tumor therapies and CSC-specific immunotherapies would be desirable to eradicate cancer.
In the future, CAR effector cells specific to CSCs combined with chemotherapy, radiotherapy or immune checkpoint inhibitors will hopefully be more effective, helping to achieve better outcomes as compared to monotherapies.
Immunotherapy strategies based on NK and CIK cells have the advantage over other types of autologous T cell therapies, including CAR T cells, of an intrinsic tumor-killing ability by recognizing HLA-independent inducible stress ligands . These properties extend their therapeutic value to numerous types of solid tumors. Applying CARs or bispecific antibodies to NK and CIK cells, we could hopefully add specificity to their tumor killing capabilities [155,198,199,230,231].
In near future, rigorous evaluation of the different cell therapy strategies alone or in combination with other treatments (e.g., chemotherapy and/or radiotherapy) is advisable to provide insights into the optimization and development of novel anti-cancer immunotherapy protocols capable of involving CSCs.