Chimeric antigen receptor (CAR) T-cells targeting CD19 demonstrated remarkable efficacy for the treatment of B-cell malignancies. The development of CAR T-cells against T-cell malignancies appears more challenging due to the similarities between the therapeutic, normal and malignant T-cells. The obstacles include CAR T-cell fratricide, T-cell aplasia, and contamination of CAR T-cell products with malignant T-cells. Here, we review these challenges and propose solutions to overcome these limitations.
Chimeric antigen receptor (CAR) T-cells demonstrated remarkable efficacy for the treatment of B-cell malignancies and have been approved by the US Food and Drug Administration (FDA) for the treatment of relapsed/refractory B-cell acute lymphoblastic leukemia (B-ALL) and diffuse large B-cell lymphoma (DLBCL) [1,2,3,4,5]. This proof of concept generated great enthusiasm for the development of CAR T-cells directed against other types of cancer, including T-cell malignancies [6, 7].
T-cell malignancies encompass immature (i.e., T-cell acute lymphoblastic leukemias (T-ALL)) and mature (i.e., T-cell lymphomas (TCL)) lymphoid neoplasms and are often associated with a dismal prognosis [8,9,10].
Despite great interest, the development of CAR T-cells against T-cell malignancies has been limited so far due to the difficulties to distinguish between therapeutic, normal and malignant T-cells. Here, we review the challenges raised by such development and describe the solutions that have been proposed to address these limitations.
CAR T-cells, directed against antigens shared with normal T-cells, may recognize and kill three types of cells: tumor T-cells, normal T-cells, and CAR T-cells (Fig. 1). Mutual killing of CAR T-cells, also called fratricide, may prevent the generation, expansion and persistence of CAR T-cells. Prolonged and profound T-cell aplasia induced by the destruction of normal T-cells exposes patients to severe opportunistic infections [11, 12]. Thus, developing CAR T-cells for T-cell malignancies requires targeting of malignant T-cells while sparing normal and CAR T-cells, or at least some subsets of them.
Furthermore, CAR T-cell products may be contaminated with malignant T-cells. Indeed, circulating tumor T-cells are often found in the peripheral blood of patients with T-ALL [9, 13,14,15] and, although less frequently, with TCL [8, 16]. Because tumor T-cells may harbor the same phenotypic and functional properties as normal T-cells , they may be harvested, transduced, expanded, and infused concomitantly with normal T-cells. This process may lead to the generation of “CAR tumor T-cells”. Ruella et al. recently described accidental transduction of CAR construct in leukemic B-ALL cells leading to CAR expressing blasts (so called “CARB”) . In these patients, CD19 CAR at the blast surface bound to CD19, thus preventing their recognition by CAR T-cells. A similar mechanism may be anticipated with malignant T-cells if transduced with the CAR construct. Furthermore, contaminating tumor T-cells may also be genetically edited to prevent the expression of a T-cell target along with normal T-cells and thereby escape CAR T-cells recognition and eradication. Thus, developing CAR T-cells for T-cell malignancies also requires to avoid contamination of the CAR T-cell product with malignant transduced T-cells.
To prevent fratricide, CAR cells should be directed against a tumor antigen that is not shared (or not completely shared) between malignant and therapeutic T-cells. This can be achieved in two ways: (i) either by targeting a tumor antigen that is not naturally expressed by the CAR T-cells, (ii) or by using CAR cells that do not express the T-cell target which can be achieved by using CAR T-cells that have been genetically edited ex vivo to prevent expression of the T-cell target or by using non-T CAR cells such as NK-cells.
CAR T-cells directed against antigens that spare CAR T-cells (Table 1)
Most target antigens are shared between normal and malignant T-cells [8, 19, 20] rendering specific targeting of tumor T-cells challenging. Strategies that target tumor antigens while sparing CAR T-cells include targeting of pan-T antigens which are downregulated during CAR T-cell expansion (e.g., CD5) thereby preventing or minimizing CAR T-cell fratricide , or targeting of antigens which are expressed only by a subset of normal T-cells (e.g., CD4, CD30, or CCR4) thereby sparing subsets of CAR T-cells (Table 1) [22,23,24].
CD5 CAR T-cells
CD5 is expressed by normal T-cells and by most T-ALL and TCL [8, 25]. In preclinical models, CD5 is downregulated through internalization upon ligation with its ligand or an antibody thus preventing fratricide [21, 26]. Interestingly, fratricide effect is enhanced when a 4-1BB co-stimulatory domain is used instead of a CD28 co-stimulatory domain [27,28,29]. Indeed, 4-1BB stabilizes the fratricide immunologic synapse through TRAF signaling by upregulating the adhesion molecule ICAM1 . Nevertheless, this fratricide effect can be counteracted with conditional CAR expression through a Tet-OFF-inducible expression system .
CD4 CAR T-cells
CD4 is expressed by two thirds of normal T-cells, by most TCL and a subset of T-ALL [8, 17, 25]. Preclinical data testing CD4 CAR T-cells have shown a highly enriched CD8+ CAR T-cells product which efficiently killed lymphoma cells in vitro and in vivo . However, such CD4 CAR T-cells will induce a CD4 T-cell aplasia which would result in an HIV/AIDS-like syndrome. Thus, we view this approach as a temporary or bridging strategy.
CD30 CAR T-cells
CD30 is expressed by a subset of activated B and T-cells, virtually all Hodgkin lymphomas (HL) and anaplastic large cell lymphomas (ALCL), subsets of peripheral TCL, and about one third of T-ALL [30, 31]. Following demonstration of a preclinical activity , CD30 CAR T-cells have been evaluated in two phase I clinical trials, mostly in patients with HL [23, 24]. No decrease in B or T-cell counts were described in these studies. Furthermore, T-cell immunity to common viral pathogens did not seem to be impaired . This is in line with what has been observed in patients treated with brentuximab-vedotin, an antibody-drug conjugate directed against CD30, where no unexpected opportunistic infections have been observed . In the study by Ramos et al. two patients with ALCL (one systemic ALK+ and one cutaneous ALK-) were infused with CD30 CAR T-cells [23, 32]. One of the two patients achieved a complete remission which lasted for 9 months following 4 infusions of CD30 CAR T-cells at the highest dose (2 × 108 cells/m2) . Wang and colleagues observed a partial response following infusion without conditioning regimen in one (and only) ALCL (cutaneous) patient . These encouraging results may be further improved by adding a lymphodepleting conditioning regimen prior to the CD30 CAR T-cell infusion [34, 35].
CCR4 CAR T-cells
CCR4 is a chemokine receptor expressed by subsets of T-cells (regulatory T-cells (Tregs), Th2, and Th17), virtually all adult T-cell leukemia (ATL), most cutaneous TCL and a significant number of peripheral TCL [8, 36]. Mogamulizumab is an anti-CCR4 antibody which has been approved in Japan for the treatment of relapsed/refractory ATL . In preclinical models, CCR4 CAR T-cells efficiently killed ATL and TCL cell lines, both in vitro and in vivo . No extensive fratricide has been observed in preclinical models but the CD4:CD8 ratio among CAR T-cells was reversed after transduction, suggesting that a subset of CD4 T-cells underwent mutual killing . Furthermore, CCR4 CAR T-cells are expected to eradicate Tregs in the tumor microenvironment. Consistent with this hypothesis, cases of severe cutaneous toxicities associated with Treg depletion have been reported in patients treated with mogamulizumab [39,40,41].
Genetically-edited CAR T-cells to prevent target expression
Fratricide may also be avoided by knocking-out the target gene using gene editing (such as TALEN or CRISPR system). This approach has been evaluated preclinically with CD7 CAR T-cells. CD7 is expressed by normal T and NK-cells, by most T-ALL and a subset of TCL [8, 25]. Unlike CD5, CD7 is poorly downregulated upon CAR T-cell expansion/activation. Thus, prevention of fratricide requires genomic disruption of CD7 prior to CAR transduction. In preclinical models, CRISPR/Cas9-mediated editing of CD7 abrogates fratricide and enables the expansion of CAR T-cells [42, 43]. Similar results have been achieved using PEBL technology , a method that prevents CD7 surface expression by anchoring newly synthesized CD7 in the endoplasmic reticulum and/or Golgi [45,46,47]. Gomes-Silva et al. suggested that the infused CD7 CAR T-cells may retain antiviral activity through their native receptor and therefore counteract the profound immunodeficiency induced by on-target/off-tumor effects of CD7 CAR T-cells .
CARs are commonly transduced into T-cells but the use of NK-cells is emerging . Using NK-cells is a promising strategy to avoid fratricide. The typical cell surface phenotype of NK-cells shows lack of T-cell receptor (TCR), CD3 and CD5 expression [49, 50]. Conversely, NK-cells are characterized by CD56 and CD7 expression . NK-cells are part of the innate immune system and have natural cytotoxic properties against tumors which can be further improved by CAR engineering [48, 52, 53]. NK-cells present several advantages for CAR engineering: (i) their phenotype (different from T-cells) can be used to prevent fratricide and avoid contamination, (ii) due to their lack of a TCR, they do not naturally induce graft-versus-host disease (GVHD)  and thus can be used in allogeneic conditions, (iii) their short lifespan may prevent prolonged T-cell aplasia . However, using blood NK-cells to manufacture CAR cells is challenging because the collection, expansion and transduction of these cells is difficult . For these reasons, CAR NK-cells directed against CD3, CD4 and CD5 have been engineered using the NK-92 cell line, a human cell line derived from a patient with a NK-cell lymphoma, rather than natural NK-cells [50, 55,56,57,58,59]. No fratricide is expected since NK-cells do not express these targets. In preclinical models however, CAR T-cells seem to outperform CAR NK-92 cells [21, 22, 29, 50, 57]. Although CAR NK-92 cells can induce significant reduction of tumor burden, they lack persistence in xenograft mouse models, consistent with the short lifespan of NK-cells [48, 60]. Moreover, some concerns may be raised regarding the potential tumorigenicity of CAR NK-92 cells since they originate from a transformed cell line. To prevent this risk, NK-92 cells are irradiated before injection to patients. NK-92 cells (not genetically engineered) have been evaluated in phase I clinical trials in patients with metastatic solid tumors [61, 62]. Another safety concern is the advent of neurotoxicity (strokes) after infusion of CD3 and CD5 CAR NK-92 cells in mice [50, 56].
Unlike B-cell aplasia which is usually well tolerated and can be compensated with infusions of immunoglobulins for the lack of humoral adaptive immunity [63, 64], prolonged T-cell aplasia exposes patients to opportunistic infections [11, 12]. Prevention of prolonged T-cell aplasia may be achieved in three ways: (i) either by targeting a tumor antigen that is not expressed by all or a subset of normal T-cells, (ii) by using short-lived CAR cells and (iii) by myeloablation and subsequent bridging to allogeneic hematopoietic stem cell transplantation (HSCT).
CAR T-cells directed against antigens that spare all or subsets of normal T-cells (Table 1)
Such strategies have been previously described (paragraph “Fratricide”). Targeting of certain T-cell markers may induce profound immune suppression, either because they induce depletion of most T-cells (e.g., CD5 and CD7) or because they deplete T-cell subsets which are important for the prevention of opportunistic infections (e.g., CD4). Depletion of other T-cell subsets such as CD30 may be better tolerated [23, 24]. Another promising approach is the targeting of the T-cell receptor beta constant 1 (TRBC1) or TRBC2. Physiologically, the TCR β chain expresses either TRBC1 or the TRBC2 constant region . Maciocia et al. have shown that the proportion of TRBC1+ T-cells varies between 25 and 47% in healthy donors, regardless of the T-cell subset . T-cell leukemias and lymphomas, instead, are clonally TRBC1 positive or negative . Therefore, TRBC1 CAR T-cells kill specifically TRBC1 malignancies while sparing TRBC2+ normal T-cells . A clinical trial testing TRBC1 CAR T-cells in T-cell lymphomas is ongoing (AUTO4).
Short-lived CAR cells
Another way to prevent prolonged T-cell aplasia is to use CAR cells with limited lifespan. This can be achieved by using (i) allogeneic CAR T-cells, (ii) CAR NK-cells, (iii) non-viral mRNA transfection with electroporation , or (iv) a safety switch (such as suicide gene or a targetable surface marker) [68,69,70,71,72]. However, these strategies do not allow prolonged persistence of CAR T-cells meant to prevent disease recurrence. Thus, they may rather be used as a bridge to transplant. Nonetheless, it is yet unclear how long CAR T-cells need to persist in order to prevent tumor recurrence. In refractory DLBCL treated with CD19 CAR T-cells containing a CD28 co-stimulatory domain, durable responses were present in patients with and without detectable persisting CAR T-cells. Indeed, 29% (13/45) of patients remaining in response at 1 year had undetectable CAR T-cells . In these patients, objective responses correlated with CAR T-cells expansion.
Contamination of CAR T-cells product with malignant T-cells
Purifying the apheresis product from circulating tumor T-cells to produce CAR T-cells is challenging since it is often difficult to distinguish between normal and neoplastic T-cells. Thus, avoiding contamination can be achieved in two ways: (i) either by purifying and transfecting non-T cells, such as NK-cells (described previously), or (ii) by producing CAR T-cells from an allogeneic healthy donor.
Allogeneic CAR T-cells
CAR T-cells can be generated from allogeneic donors . Nevertheless, infusion of allogeneic CAR T-cells may cause life-threatening GVHD, even after HLA matching [73, 74]. To overcome this issue, Cooper et al. developed “off-the-shelf”, universal CD7 CAR T-cells (UCART7) . Using multiplex CRISPR/Cas9 gene editing of T-cells before CAR transduction, they deleted both CD7 and T-cell receptor alpha chain (TRAC). In preclinical models, their CD7 CAR efficiently killed T-ALL without inducing xenogeneic GVHD in a patient-derived xenograft (PDX) mouse model . These allogeneic CAR T-cells are expected to have a short lifespan because they will be eliminated upon immune reconstitution of the host. This short persistence may be seen as an advantage to prevent T-cell aplasia but as a disadvantage to prevent cancer recurrence.
The development of CAR T-cells for T-cell malignancies faces unique challenges due to the similarities between therapeutic, normal, and malignant T-cells. Many of the solutions that have been proposed do not seem optimal, either because they lack specificity (risk of fratricide, immune suppression and/or contamination) or persistence (risk of tumor recurrence). Targeting of certain subsets (e.g., CD30 or TRBC1 CAR T-cells) seems promising but is restricted to subtypes of T-cell malignancies. It is unlikely that one type of CAR T-cells will be used for all T-cell malignancies (unlike CD19 CAR T-cells for B-cell malignancies). To date, few studies evaluated CAR T-cells in patients with T-cell malignancies [23, 24] but several trials are underway or about to be launched (Table 1). Results from these clinical trials are eagerly awaited.
M.A., M.T., C.H.J., and R.H. performed the literature review, wrote the manuscript, and created the table and figure.