Although highly and uniformly expressed on B cell malignancies, CD19 may be downregulated or mutated in tumor cells (31), and antigen loss variants have been reported as responsible of the majority (70%) of relapses in ALL (32). Alternative surface molecules, including CD20, CD22 and the immunoglobulin light chain are also frequently expressed in B cell tumors and CARs targeting these alternative lymphoma-associated antigens are currently under development for the treatment of B-cell malignancies (Table 2). Successful results obtained with anti-CD20 monoclonal antibodies have formally supported the development and clinical testing of anti-CD20 CAR-T cell therapy. CD20 has been targeted with a third generation CD28/4-1BB/CD3ζ CAR and transient responses were observed in 3 of 4 patients (33). Inclusion of dual costimulatory domains (CD28 and 4-1BB) enhanced anti-CD20 CAR T cells persistence in patients with indolent B cell and mantle cell lymphoma (34). In this study, anti-CD20 CAR T cells could be detected up to one year post transfer, and 2/3 patients treated had a PFS at 24-month follow-up. Based on preclinical results showing their antitumor efficacy, anti-CD22 CAR T cells are currently under evaluation in early-phase clinical trials. Results from a phase I trial testing anti-CD22 CAR T cells in 21 children and adults with B-cell ALL, including 15 patients previously treated with anti-CD19 CAR T cells, showed a dose-dependent anti-leukemic activity, with CR obtained in 73% (11/15) of patients - including 5 of 5 patients with CD19^dim or CD19- B-ALL - and a median remission duration of 6 months (35). The immunoglobulin kappa (k) light chain antigen is another attractive target because its expression on k-restricted B-cell lymphomas, and not on non-malignant B cells, may avoid complete B-cell aplasia and minimize humoral immunity impairment. In a phase I clinical trial, 16 patients with relapsed or refractory k+ NHL/CLL or MM were treated with autologous T cells genetically modified to express a CAR specific for the k light chain (36). Overall, of 9 patients with relapsed NHL or CLL, 2 achieved a CR remission and one a PR. Of 7 patients with MM, 4 had stable disease (SD) lasting 2–17 months. No toxicities attributable to anti-k light chain CAR T cells were observed. Another potential tumor antigen is the receptor tyrosine kinase-like orphan receptor (ROR1), a transmembrane glycoprotein expressed on embryonal tissue and aberrantly on many adult malignant tissues, such as B-cell tumors (e.g. CLL, mantle cell lymphoma, B-cell ALL) and numerous types of solid tumors (37-40). Due to its high-level surface expression as well as to its crucial role in tumor cell proliferation, survival, and metastasis, a number of pharmacological agents targeting ROR1 are under development, such as humanized monoclonal antibodies, small molecule inhibitors, bispecific T-cell engagers (BiTE) and anti-ROR1 CAR T cells (41-43). ROR1-targeted T cells have demonstrated to generate cytotoxicity against human ROR1 positive B cell malignancies in preclinical studies (39), without causing overt cytotoxicity in non-human primates (44).

Novel cell therapies and several potential targets for CAR T cells are under investigation in MM (45, 46) (Table 2). Anti-CD38 and anti-SLAMF7 CAR T cells showed anti-MM effects in preclinical models (47, 48), but the expression of both target antigens on other normal tissues including hematopoietic lineages and immune effector cells may cause off-tumor toxic effects limiting their long-term clinical use. Being expressed on plasma cells of all MM patients but not on normal tissues, BCMA - a member of the TNF receptor family - is at present the most promising target in this disease setting. A first in-human phase I clinical trial conducted at the National Cancer Institute in heavily pre-treated MM patients demonstrated the safety and efficacy of adoptively transferred autologous T cells transduced with a retroviral vector incorporating an anti-BCMA scFv, a CD28 costimulatory domain, and the CD3ζ T-cell activation domain. Updated results from this study showed an ORR of 81%, with 63% very good partial response or CR, and a median EFS of 31 weeks (49). Another phase I study, conducted at the University of Pennsylvania, is currently exploring a different CAR, developed in collaboration with Novartis and consisting of a fully human anti-BCMA scFv with a 4-1BB costimulatory domain that is packaged in a lentiviral vector (50). Three cohorts have been enrolled sequentially, with the aim of collecting preliminary data about safety, efficacy, and kinetics of expansion - both with and without lymphodepleting chemotherapy. Preliminary results presented at 2017 American Society of Hematology meeting showed that, based on IMWG criteria, in cohort 1 (CAR T cells alone at a dose of 1-5x10⁸ cells, n=9) 6/9 patients responded, with 1 ongoing CR at 21 months, and other responses lasting 1.5 to 5 months. In cohort 2 (cyclophosphamide 1.5 g/m² with 1-5x10⁷ CAR T cells, n=5), 2/5 patients responded but progressed at 4 and 2 months, respectively. In cohort 3 (cyclophosphamide 1.5 g/m² with 1-5x10⁸ CAR T cells, n=10), the incorporation of cyclophosphamide with the higher dose of CAR T cells led to a disease response in 5/6 patients, and one was not yet evaluable at the time of data submission. Berdeja et al. reported updated data about the dose-escalation portion of a third BCMA CAR T-cell trial (NCT02658929) at the 2017 American Society of Hematology meeting (51). This was a multicenter study sponsored by Bluebird Bio, which used a second-generation CAR called bb2121 that contained a murine anti-BCMA scFv (the same one used in the National Cancer Institute trial) and a 4-1BB costimulatory domain, transduced by a lentiviral vector. Preliminary data from this phase I study showed that 89% of patients treated with bb2121 responded to treatment with 8 ongoing clinical responses at 6 months, and one patient demonstrating a sustained response beyond one year. Based on these promising results, two phase II multicenter studies evaluating the efficacy and safety of bb2121 in subjects with relapsed/refractory MM are currently ongoing in the US and in Europe (NCT03361748 and NCT03601078). Of note, BCMA is also expressed in a substantial number of lymphoma samples, as well as primary CLL B cells. Preclinical data have shown that treatment with bb2121 results in rapid and sustained elimination of the tumors and 100% survival in NOD/SCID gamma mouse models of human lymphomas, thus supporting the further development of anti-BCMA CAR T cells as a potential treatment for not only MM but also some lymphomas (52). Preliminary results from a fourth anti-BCMA CAR T cells trial, conducted by Nanjing Legend Biotech in China in relapsed/refractory MM patients, were reported at the 2017 American Society of Clinical Oncology Annual Meeting (53). This CAR, called LCAR-B38M, uses a novel antigen-binding domain that binds BCMA at two separate epitopes. At the time of presentation, 100% (19/19) of patients experienced a response, including 74% with a CR. During a median follow-up time of 6 months, no patient with CR experienced relapse. Toxicity was modest: 83% developed CRS, but only 6% had grade 3 CRS and no grade 3-4 neurotoxicity was seen. Several additional antigens are currently under investigation in MM as potential targets for CAR T cell therapies, including CD44v6 (54), Lewis Y (55), NKG2D ligands (56), CD229 (57), and integrin β7 (58).

Broadening of CAR T cells to treat T-cell malignancies has proven challenging, mainly because many targetable antigens are also expressed on normal T lymphocytes. This shared antigenicity can cause fratricide in CAR T cells, inhibiting their proliferation and viability, and leading to normal T cells depletion and deep impairment of host immunity. To avoid this drawback, an anti-CD5 CAR transduced into a human Natural Killer (NK) cell line was tested (59). This approach showed potent anti-tumor activity against a variety of T-cell leukemia and lymphoma cell lines as well as primary tumor cells, and was able to demonstrate significant inhibition of disease progression in xenograft mouse models of T-cell ALL. Recent data reported by Gomes-Silva et al. showed that targeted disruption of the CD7 gene using clustered, regularly interspaced short palindromic repeats (CRISPR)/Cas9 prior to CAR expression minimizes fratricide in T cells and allows the expansion of the CD7-knockout (CD7KO) anti-CD7 CAR T cells with robust antitumor properties in preclinical models (60). An alternative strategy to limit fratricide is to employ T cells transduced with a CAR specific for the T-cell receptor (TCR) β chain constant region expressed by the malignant cells (i.e. TCRBC1 or TCRBC2), thus sparing the normal T cells that express the alternative TCRBC chain. Recent data demonstrated that ex vivo selected TCRBC2-positive T cells transfected with a CD28/OX40/CD3ζ anti-TCRBC1 CAR specifically recognized TCRBC1-positive T cell leukemia and lymphoma cells in vitro and in xenograft murine models (61).

Promising results and limited toxicities obtained with toxin-conjugated CD30-specific monoclonal antibodies (such as brentuximab vedotin) (62, 63) have encouraged the development and clinical testing of CD30-directed CAR T cell therapies in Hodgkin lymphoma (HL) and anaplastic large cell lymphoma (ALCL) (Table 2). The clinical efficacy of this approach in relapsed or refractory HL patients was first demonstrated by results obtained in a phase I study conducted in China, which tested the adoptive transfer of autologous T cells transfected with a lentiviral vector encoding for a second generation CAR containing an anti-CD30 scFv combined with a 4-1BB costimulatory domain (64). Within this study, CART-30 cell infusion was well tolerated, with grade ≥3 toxicities occurring only in 2 of 18 patients, and effective, with 7 patients achieving PR and 6 achieving SD. Promising results were obtained by a second phase I dose-escalation study evaluating the safety of autologous T cells gene-modified with a retroviral vector to express an anti-CD30 CAR combined with the CD28 costimulatory domain (65). In this study, of 7 patients with relapsed HL, one entered CR lasting more than 2.5 years after the second infusion of anti-CD30 CAR T cells, one remained in continued CR for almost 2 years, and 3 had transient SD. Of 2 patients with ALCL, one had a CR that persisted 9 months after the fourth infusion of anti-CD30 CAR T cells. No toxicities were observed and even though CD30 may also be expressed by normal activated T cells, no patients developed impaired virus-specific immunity. These studies demonstrate that targeting CD30 with CAR T cells is safe and can have antitumor activity in CD30+ malignancies.

Despite the increasing availability of new targeted molecules, including IDH, FLT3 and BCL2 inhibitors, the prognosis of relapsed and refractory acute myeloid leukemia (AML) remains dismal. AlloHSCT is the only realistic curative option but it is not feasible in many patients and results are unsatisfactory in this setting, especially if a good quality remission is not achieved before transplantation (66). Despite considerable efforts, the development of CAR T cells in AML represents a major challenge mainly due to the absence of a leukemia-specific surface antigen (67) (Table 2). First, the co-expression of AML antigens on extra-hematopoietic tissues (e.g. CD33 in hepatic Kupffer cells, CD123 on endothelial cells) poses important safety concerns. Besides, currently known AML surface targets are also expressed on normal hematopoietic stem and progenitor cells (HPSCs), leading to a very high risk of severe hematologic toxicities. Given that a long-term pan-myeloablation following AML-directed CAR T cells infusion may not be tolerable, unlike B cell aplasia due to CD19-targeted therapy, different solutions are being tested (68). AlloHSCT represents the logical rescue strategy, but the persistence of even a few CAR T cells after transplantation could lead to graft rejection (67). Therefore, some groups explored the possibility of developing short-term living CAR T cells as a “bridge to transplant”. The University of Pennsylvania employed for this purpose CAR mRNA electroporation, in order to allow transient expression of their anti-CD123 CAR as opposed to the stability obtained by viral delivery (69). Unfortunately, this strategy was unsuccessful, since none of the 7 patients enrolled in the clinical trial showed a response, likely due to the lack of CAR T cells persistence (70). As a matter of fact, a prolonged CAR T cells expansion is essential to obtain a sustained remission (15), and finding the right balance between the benefits of CAR persistence and the risks of long-term myelosuppresion is one of the major challenge in AML-directed CAR T cells development. Alternative ways of CAR T cells termination, including the use of antibodies against both artificially expressed and constitutively present surface antigens (71), inducible suicide genes (72) and on-off switch strategies (73) are being explored and could permit to promptly induce CAR T cells exhaustion when needed.

The University of Pennsylvania group recently presented another innovative way to prevent the risk of prolonged myelotoxicity and graft rejection due the lack of specificity of CARs in AML. The investigators generated CD33-deficient human HSPCs and demonstrated prolonged engraftment and normal differentiation in mouse and macaque models. CD33-deficient cells were unaffected by anti-CD33 CAR T cells, thus allowing for effective anti-leukemic activity without myelotoxicity (74). With the aim of identifying new specific antigens, Perna et al. performed an exploratory study by combining proteomics and transcriptomics analysis from malignant and normal tissues. Albeit no antigen with characteristics as favorable as CD19 was found, some promising targets, namely ADGRE2, CCR1, CD70, and LILRB2, were identified (75). Besides, a recent proof-of-concept publication showed that also intracellular peptides can be targeted and that the efficacy of CAR T cells can be boosted by vaccination. The authors developed a CAR specific to a WT1/human leukocyte antigen (HLA)-A complex, and vaccination with dendritic cells loaded with the corresponding antigen led to CAR T cells expansion and activation, thus enhancing anti-leukemic activity in a xenograft model (76).

An alternative approach to prevent antigen escape-mediated relapses and off-target side effects is the development of CAR T cells targeting two different antigens on the leukemic cells (77). These constructs appear of particular interest in AML, due to the impressive clonal heterogeneity and the lack of highly specific antigens, and in this context leukemia stem cells markers are particularly attractive as potential co-targets (78). Some groups are also testing small molecules with the aim of improving CAR T cells efficacy. Jetani and colleagues recently reported that the FLT3 inhibitor crenolanib had a synergist activity with FLT3-directed CAR T cells in a FLT3-ITD AML model, at least in part by increasing FLT3 surface expression (79), and similar results were also presented for midostaurin (80). Besides, it was recently reported that PI3K inhibition can enhance CD33-directed CAR T cells durability, thus improving antitumor activity (81).

Despite many in vitro and in vivo preclinical studies and the significant number of clinical trials currently enrolling worldwide, only a few data of AML patients treated with CAR T cells have been so far reported. The first published clinical trial employed a second generation CAR targeting the Lewis Y antigen, an oligosaccharide overexpressed by some epithelial and hematological malignancies: signs of clinical activity and good tolerability were shown in the 4 treated patients, but responses were short-lived and CAR T cells persistence was limited (82). CD33 is an obvious target in AML, due to its very frequent expression on leukemic blasts, and several efforts are ongoing to increase anti-leukemia efficacy of CD33-directed constructs, including the optimization of co-stimulatory molecules (83). So far, Chinese investigators reported only on one heavily pretreated patient who experienced a short-lasting partial response and moderate toxicity after CD33-CAR T cell treatment (84). Despite the disappointing results obtained with the adoptive transfer of T cells transiently modified by mRNA electroporation to express a CD123 specific CAR (70), preliminary clinical data on lentiviral transduced anti-CD123 CARs are promising. One case report from China showed some clinical activity (85), and more recently, the City of Hope group reported encouraging results on 6 AML patients, 3 of whom experienced a clinical meaningful response and, interestingly, no significant myelosuppression after anti-CD123 CAR T cell infusion. One patient with refractory blastic plasmocytoid dendritic cell neoplasm responded as well, obtaining a CR. CRS was observed, but it was manageable and always of grade 1 or 2 (86). Preclinical results of this construct were also reported in high risk MDS (87).

Universal allogeneic CAR T cells are of particular interest in AML, giving the increased manufacturing problems often faced with autologous CARs in this disease (88). Unfortunately, a phase I study of UCAR123 (i.e. allogenic anti-CD123 CART T cells) was temporarily halted due to unacceptable toxicity in the first two patients treated, including one fatal case, and data on further development of this construct are awaited (68). NKG2D-expressing CAR T cells (89) were tested in a phase I trial including 6 patients with MDS or AML, but poor CAR T cell persistence, no CRS and very limited clinical activity were shown (56). However, a relapsed AML patient enrolled in a trial testing multiple infusions of NKG2D-bearing CAR T cells obtained marrow leukemia free status at day 28 post infusion and was later successfully allografted (90). Recently, preliminary results of a compound CAR simultaneously targeting two different AML antigens, CLL1 and CD33, have been reported. After demonstrating significant activity of the construct in vitro and in a mouse model, and after proving the efficacy of alemtuzumab as safety switch, Liu and colleagues presented the results of the first patient treated, who achieved a minimal residual disease (MRD)-negative CR and could be successfully bridged to alloHSCT (91). Other AML targets have been studied in preclinical models with encouraging results, including folate receptor ß (92), CD7 (93), CD44v6 (54), CD38 (94), CLL1 (95), FLT3 (96), and LILRB4 (97), but CAR T cells directed toward these antigens have not entered clinical development or are tested in early trials whose results are not available yet.

The development of CAR T cells in solid tumors has yielded limited results, since unique challenges come in addition to those faced in hematological malignancies (98) (Table 2). Likewise, in AML, important safety concerns derive from the lack of highly specific target antigens, and severe and sometimes fatal toxicities due to on-target off-tumor effects on normal tissues occurred in some trials (99-102). Besides, the tumor microenvironment represents a unique obstacle, which significantly limits CAR T cells efficacy in solid cancers, due to the presence of environmental barriers (e.g. extracellular matrix), chronic inflammation and the presence of immunosuppressive molecules and immune cells. Furthermore, high tumor heterogeneity can facilitate antigen escape and in vivo CAR T cells expansion is significantly less pronounced than in patients with hematologic malignancies (103, 104). In order to prevent potentially lethal toxicities, safety switch strategies are being tested (98) and the University of Pennsylvania group demonstrated the feasibility, although with limited efficacy, of administering mRNA electroporated mesothelin-specific CAR T cells in pancreatic adenocarcinoma patients (105). Other groups are exploring the possibility of local delivery, to enhance tumor killing and to reduce toxicity. Indeed, it has been shown that regional administration of anti-AFP or anti-CEA CAR T cells was equally effective or even superior to a systemic administration in xenograft models of peritoneal carcinomatosis (106, 107), and the delivery of CAR T cells through percutaneous hepatic artery infusions has shown to be safe and clinically active in patients with CEA-expressing liver metastases (108). A different approach to favor the localization into the tumor site consists in the administration of CAR T cells overexpressing tumor-specific chemokine receptors enhancing their capacity of reaching the tumor tissues (109).

In order to overcome antigen escape risk and inter-patient variability, CARs co-targeting two or even three antigens are being developed in several solid malignancies, including breast, pancreatic and brain tumors (110-113). Conversely, other investigators are testing dual constructs requiring both targets to be expressed on tumor cells in order to exert their cytotoxic activity (114) and inhibitory CARs able to redirect T cells activity from healthy tissues (115), with the aim of increasing CAR specificity and of preventing off-tumor side effects. Furthermore, many efforts are being made to optimize co-stimulatory molecules and to improve the structure of CARs (116), and new constructs able to locally deliver immune-modulating cytokines (117-120), bearing modified cytokine receptors (121) or targeting tumor matrix components (122, 123) are being developed to overcome environmental barriers. Some groups are also exploring the possibility of combing CAR T cells with oncolytic viruses (124) or with checkpoint inhibitors (125-127). Indeed, targeting the PD-1/PD-L1 axis is of particular interest for CAR-T cells development in solid tumors (128), and investigators are testing constructs which are able to deliver anti-PD-1 molecules (129, 130) and strategies of cell-intrinsic PD-1 inhibition and inactivation (126, 131, 132), with the aim of improving the efficacy of CAR T cells also reducing the systemic side effects of checkpoint inhibitors.

Despite these challenges, basic and clinical research on CAR T cells in solid malignancies is constantly expanding, with more and more clinical trials opening worldwide (133) and many new data being presented. In sarcoma, anti-HER2 CAR T cells showed a reassuring safety profile (134) and led to CR in one case (135), while anti-GD2 constructs demonstrated significant clinical activity in 5 out of 11 patients with active neuroblastoma, including 3 CR (136). Signs of activity were also shown in recent studies involving epithelial tumors (137-140). The concept of the central nervous system (CNS) as an immune privileged site has been overturned in recent years, since T cells can penetrate the blood-brain barrier and infiltrate the brain in a diffuse manner (141); as a matter of fact, one of the most promising field of CAR T cells development in solid malignancies are CNS tumors. In glioblastoma patients, the University of Pennsylvania group showed that EGFRvIII-directed CAR T cells effectively trafficked to regions of active tumor after systemic infusion (142), and Ahmed and colleagues reported clinically significant activity of HER2-specific CAR T cells in 8 out of 17 cases, including one PR and 7 persistent SD (143). Besides, local delivery showed encouraging results in this setting: in xenograft models, investigators demonstrated robust antitumor efficacy of anti-HER2 CAR T cells against medulloblastoma (144) and brain metastases of breast cancer (145) and anti-GD2 constructs were effective against midline gliomas (146). In addition, Brown and colleagues reported that IL-13Rα2–directed CAR T cells delivered via an implanted reservoir/catheter system exert a significant clinical activity in glioblastoma patients (147), including one CR (148).

CRS is an acute inflammatory process that commonly occurs within the first days after CAR-T cell infusion. Closely related to the macrophage activation syndrome, it is characterized by non-specific constitutional symptoms with transient elevations of serum cytokines and other biomarkers. The delayed onset of clinical symptoms suggests an on-target, antigen-driven T-cell activation process. Mild CRS is a self-limiting clinical syndrome of low-grade fever, arthralgias and myalgias and is usually managed with minimal intervention. However, in a small number of patients it can progress to a dramatic syndrome of multi-organ dysfunction, requiring aggressive intervention and intensive life-support. Severe CRS is characterized by the development of a capillary leak syndrome with third spacing and hypotension, insufficient renal blood flow and pulmonary edema leading to hypoxia. Neurological symptoms may also occur during CRS though they generally do not follow the same time course as systemic CRS symptoms.

Though the pivotal role played by IL-6 in severe CRS is well known (9, 14), new intriguing observations correlate severe CRS with the host macrophage/monocyte system rather than the CAR-T cells induced cytokine milieu (149). This observation may reveal a novel scenario where the cross-talk between CAR T cells and host myeloid cells is essential for the recruitment of macrophages which, in turn, are responsible for the release of IL-6. Furthermore, the demonstration of the involvement of inducible nitric oxide synthase (iNOS) in CRS patho-physiology recognizes a role for IL-1, a strong iNOS inducer. In a mouse model, the administration of an IL-1 receptor antagonist prevented severe CRS while sparing anti-tumor efficacy (150).

Severe CRS treatment is based upon direct cytokine inhibition, corticosteroids and intensive life support. The IL-6 receptor antagonist tocilizumab has often been used and has resulted in rapid improvement of clinical symptoms without affecting potential anti-tumor activity (9, 14, 151). The use of tocilizumab was approved by the FDA in August 2017. A currently ongoing clinical trial (NCT02906371) may contribute to establish the role of tocilizumab as prophylaxis of CRS in anti-CD19 CAR-T cells recipients. Corticosteroids have also been used as first-line treatment for severe CRS due to their well-known ability to blunt activated T-cells. However, given their potential risk of compromising CAR-T cells efficacy in vivo, their use is commonly limited to life-threatening CRS unresponsive to cytokine inhibition. Intensive care treatment may be required in case of refractory severe CRS with unstable blood pressure and/or ventilator-requiring hypoxia.

CAR-T cell related neurologic adverse events include a wide spectrum of clinical symptoms that may develop with or without CRS. They are generally of mild-moderate intensity and reversible while only a minority of patients require specific treatment. Reported incidence ranges from 7% to 63% (15, 23, 24, 152).

The initial hypothesis of a brain-located CD19 acting as a powerful stimulus for an antigen-driven CAR-T cell infiltration appeared not consistent. The recent development of a pre-clinical model of neurotoxicity through the transfer of autologous CD20-specific CAR T cells showed the association of neurotoxicity with pro-inflammatory cytokines in the cerebrospinal fluid (CSF) determined by a CAR T cell mediated meningeal inflammation (153). It was demonstrated that CAR T cells accumulate in CSF and brain parenchyma along with measurable CSF levels of multiple molecules such as IL-6, IL-2, granulocyte-macrophage colony stimulating factor, and vascular endothelial growth factor. Disruption of central nervous system vascular integrity and disregulation of the platelet-von Willebrand factor axis along with high serum levels of endothelial activation biomarkers appear to play a major role during acute neurologic toxicity. The blood-brain barrier of patients with acute neurotoxicity does not prevent massive cytokine penetration into CSF, that, in turn, induces brain vascular pericyte stress and secretion of cytokines further increasing blood-brain barrier permeability in a cyclic process (154). Common neurologic symptoms include delirium, confusion, disorientation, headache, decreased level of consciousness, hallucinations and tremor, while cranial nerve damage, seizures and focal deficits have been infrequently observed. The occurrence of fatal cerebral edema has also been reported in the Kite Pharma trial and by Gust et al. (154, 155). Heavy disease burden, fludarabine-based preparative regimens and higher infused CAR-T cell doses have been associated with higher incidence of neurologic events. Higher serum levels of IL-6 and of monocyte chemo-attractant protein-1 within the first 36 hours after infusion appear to predict a higher risk of developing grade ≥4 neurotoxicity (27). Most patients with severe neurotoxic events had previously been diagnosed with grade ≥2 CRS but the physio-pathologic relationship between the two morbid processes is currently unknown.

Patients who develop at least grade 2 neurologic toxicity after CAR T infusion should be managed by the neurology consult service. Brain magnetic resonance, electroencephalogram and examination of CSF could help to rule out other causes. Events of mild intensity are almost universally self-limited and do not need intervention. Severe neurotoxicity is treated with systemic corticosteroids with dexamethasone being the first choice given its excellent blood-brain barrier penetration, and, in absence of prompt response, with tocilizumab although clear evidence of its efficacy is lacking. Given the inability of tocilizumab to cross the blood-brain barrier, it was also suggested that it may increase IL-6 levels in CSF and worsen neurotoxicity. Thus, the IL-6 antagonist siltuximab may be a valid alternative (156). A recent published observation showed that in a mouse model of CAR T-induced neurotoxicity, the use of IL-1 receptor antagonist anakinra abolished neurologic symptoms, whereas tocilizumab failed to improve them (150).

Disease relapse post allo-HSCT is at least theoretically the easiest setting for allogeneic donor CAR T cells. Unmanipulated donor lymphocyte infusion (DLI) is commonly used to treat patients with disease recurrence after allo-HSCT (157-159) though the risks of GVHD and lack of disease response remain an issue. The use of engineered donor-derived T cells to enhance graft-versus-leukemia effects and to limit the risk of GVHD is fascinating.

Most trials evaluated donor-derived CAR T cells against the B-cell antigen CD19 as target antigen. T cells were harvested from donors (160-163) or from engrafted recipients after allo-HSCT. In the latter case, all studies reported low rate of GVHD as CAR T cells were generated from tolerized cells (14, 27, 151, 164, 165). Life-threatening viral infections with pathogens such as Epstein-Barr virus, cytomegalovirus, and adenoviruses after alloHSCT can be treated without toxicity (including GVHD) by infusing ex vivo–expanded, donor-derived, virus-specific cytotoxic T cells (VST) (166-169), which are also capable of persisting several years after infusion (170). Donor-derived VST genetically modified to express a CD19-specific CAR (CD19.CAR-VST) are therefore being investigated in patients with B-cell malignancies who have either disease relapse or are at high risk of disease relapse after alloHSCT, with the hypothesis that CAR-VSTs would be activated by endogenous viral antigens, increasing their expansion and persistence irrespective of the presence of CD19-expressing normal or malignant B cells. Eight patients received CD19.CAR-VST 3 months to 13 years after alloHSCT. There were no infusion-related toxicities. Objective antitumor activity was evident in 2 of 6 patients with relapsed disease during the period of CD19.CAR-VST persistence, whereas 2 patients who received cells while in remission remain disease free. Of note, in 2 of 3 patients with viral reactivation, donor CD19.CAR-VSTs expanded concomitantly with VSTs (160). Twenty patients with hematological malignancy that persisted after alloHSCT and standard DLI received infusion of T cells, obtained from each patient’s alloHSCT donor, genetically modified to express a CAR targeting CD19. Eight of 20 responded (6 CR and 2 PR). The response rate was highest for ALL, with 4/5 patients obtaining MRD-negative remission. Responses also occurred in CLL and NHL. The longest ongoing CR was more than 30 months in a patient with CLL. No new-onset acute GVHD after CAR-T infusion developed. Toxicities included fever, tachycardia, and hypotension. Peak blood CAR-T levels were higher in patients who obtained remissions than in those who did not (162, 171).

As stated above, the promising clinical results of engineered T cell therapy could be further amplified and broadened if potent and histocompatible T cells were readily available, instead of being collected and manufactured on demand. T cells can be easily harvested from healthy donor, but their use is compromised by the high allo-reactive potential. TCR are naturally prone to react against non-autologous tissues, recognizing either allogeneic HLA molecules or other polymorphic gene products (minor antigens) (172). Recent advances in gene editing technology allow the manufacture of CAR-T cells from healthy donor leukapheresis where quantity and quality of T cells can be pre-selected. However, the use of these CAR T cells requires gene editing technology to prevent expression of endogenous TCR in order to minimize the potential to cause GVHD in non-HLA matched recipients. TCR is a heterodimer and both α and β chains need to be present for it to be expressed. A single gene codes for the α chain (TRAC), whereas there are 2 genes coding for the β chain, and TRAC loci knocked out using nucleases is the choice strategy for removing TCR expression. Initially, procedures in two steps were used: the CAR was introduced with a viral vector and the TRAC loci disrupted using a nuclease. Zinc finger nucleases (ZFN) (173), transcription activator-like effector nucleases (TALEN) and megaTAL nucleases (174-176), and CRISPR/Cas9 systems (177) have all been applied to modify T cells (178-180). As previously mentioned, Qasim et al. reported the clinical use of UCART19s, generated by lentiviral transduction of donor cells and simultaneous TALEN-mediated gene editing of TRAC and CD52 gene loci (29). More recently, a one step only procedure was introduced: the CAR-encoding DNA was directly inserted into the TRAC locus using CRISPR/Cas9 technology together with an adeno-associated virus (AAV) vector repair matrix, simultaneously generating a TCR-negative CAR-positive T cell (181). These cells were shown to be more potent than conventional lentivirally transduced CAR T cells because of a more physiological, TCR-like regulation of CAR expression. Although TCR-negative off-the-shelf CAR T cells are able to reduce the risk of GVHD, they may still be subjected to killing by the patient’s own T cells that recognize non-self HLA if there is mismatch, causing rejection and subsequently leading to short-lasting responses. To solve this issue, it has been proposed to eliminate the HLA molecules from CAR T cells using gene-editing technologies like ZFN (182).

Another way to protect allogeneic CAR T cells from rejection included the knockout of the β2-microglobulin gene combined with TCR knockout. In the absence of β2-microglobulin (β chain), class I HLA molecules do not form stable heterodimer on the cell surface (183). However, while the absence of HLA-I on the surface of CAR T cells significantly reduces rejection through HLA-mismatched CD8 T cells (class-I restricted), it can result in increased destruction by recipients’ NK cells, due to “missing self” recognition. To prevent activation of recipient’s NK cells through this mechanism, different way have been tested, such as the enforced expression of non-classical HLA molecules (e.g. HLA-E and HLA-G) (182, 184), and the overexpression of Siglec-7 and -9 ligands (185). Furthermore, the use of HLA homozygous donors to generate a bank of universal CAR T products may represent an additional strategy to avoid rejection. It was calculated that with limited numbers of donors homozygous for HLA-A/B/DRB1, it would be possible to generate compatible products to cover the majority of the population (186).

Additional immune cells are on studies as potential off-the-shelf CAR carriers, although the safety, efficacy and persistence in the clinical setting require more investigations. NK cells belong to the lymphoid branch of the immune system and account for up to 6% of circulating lymphocytes. NK cells do not express rearranged receptors and can be easily transferred across HLA barriers without causing GVHD, thus they became attractive as allogeneic effector cells (187, 188). At least three different sources of NK cells are currently available: NK cell line (189), peripheral blood NK cells (189), and induced pluripotent stem cells (iPSC) NK cells (190). In clinical studies of post-alloHSCT, NK cell infusion demonstrated the safety of using such an approach in an off-the-shelf fashion (191). CAR expression in NK cells could increase their specificity and enhanced their anti-tumor activity. NK cells have been utilized to evaluate several different antigens as CAR target. In pre-clinical studies, potent antitumor activity has been demonstrated using CAR NK cells generated from NK cell lines as well as NK cells derived from patients (192).

Gammadelta T cells (γδ T cells) are less prone to alloreactivity compare to other T cells. Immunotherapy with γδ T cells requires their activation and expansion as they represent only a small amount of circulating T cells. Aminobisphosphonates are the most efficient reagents to grow γδ T cells ex vivo. Afterward, the same technique have been transitioned to the clinic for investigational treatments of cancer and HIV (193, 194). In the alloHSCT setting γδ T cells exert anti-tumor activities and contribute to host defense against different pathogens, whereas their role in GVHD management remains unclear (195, 196). For this reason, γδ T cells have been evaluated in the setting of T cell-depleted alloHSCT (195, 196). Given their inherent potential for antitumor effects and their apparent lack of alloreactivity the combination of γδ T cells with CAR is appealing (197-199), however, although initial clinical trial with CAR γδ T cells are ongoing, far more data are needed to assess if preclinical findings translate into powerful and persistent anti-tumor activity with acceptable toxicity.

NKT cells constitute a relatively uncommon circulating immune cell population (0.1-0.5% of T cells) that co-expresses αβTCR and NK cell markers. In mice models, was found an inverse correlation of recovery of NKT cells and GVHD after alloHSCT while preserving a graft-versus-tumor response (200). NKT cells can also be expanded in vitro and can be genetically modified to express CAR (201, 202). Thanks to limited TCR alloreactivity and emerging technology to obtain large numbers, NKT cells represent another attractive source to generate off-the-shelf CAR+ immune cells.