Research ArticleCancer

The tyrosine kinase inhibitor dasatinib acts as a pharmacologic on/off switch for CAR T cells

See allHide authors and affiliations

Science Translational Medicine  03 Jul 2019:
Vol. 11, Issue 499, eaau5907
DOI: 10.1126/scitranslmed.aau5907

Putting CAR T cells in idle

Chimeric antigen receptor, or CAR, T cells can be an effective cell therapy for cancer. Unfortunately, this immunotherapy has its risks, and excessive activation of CAR T cells can occasionally cause severe, even lethal, toxicity. There are some existing approaches to suppressing overactive CAR T cells, but these are generally single-use methods that kill the CAR T cells and thereby abrogate both their toxicity and their antitumor effects. In contrast, Mestermann et al. identified dasatinib as a drug that can temporarily inactivate CAR T cells to help reduce acute toxicity, allowing the T cells to recover their antitumor effects after the drug is withdrawn.

Abstract

Immunotherapy with chimeric antigen receptor (CAR)–engineered T cells can be effective against advanced malignancies. CAR T cells are “living drugs” that require technologies to enable physicians (and patients) to maintain control over the infused cell product. Here, we demonstrate that the tyrosine kinase inhibitor dasatinib interferes with the lymphocyte-specific protein tyrosine kinase (LCK) and thereby inhibits phosphorylation of CD3ζ and ζ-chain of T cell receptor–associated protein kinase 70 kDa (ZAP70), ablating signaling in CAR constructs containing either CD28_CD3ζ or 4-1BB_CD3ζ activation modules. As a consequence, dasatinib induces a function-off state in CD8+ and CD4+ CAR T cells that is of immediate onset and can be sustained for several days without affecting T cell viability. We show that treatment with dasatinib halts cytolytic activity, cytokine production, and proliferation of CAR T cells in vitro and in vivo. The dose of dasatinib can be titrated to achieve partial or complete inhibition of CAR T cell function. Upon discontinuation of dasatinib, the inhibitory effect is rapidly and completely reversed, and CAR T cells resume their antitumor function. The favorable pharmacodynamic attributes of dasatinib can be exploited to steer the activity of CAR T cells in “function-on-off-on” sequences in real time. In a mouse model of cytokine release syndrome (CRS), we demonstrated that a short treatment course of dasatinib, administered early after CAR T cell infusion, protects a proportion of mice from otherwise fatal CRS. Our data introduce dasatinib as a broadly applicable pharmacologic on/off switch for CAR T cells.

INTRODUCTION

Adoptive immunotherapy with genetically engineered chimeric antigen receptor (CAR)–T cells is a transformative treatment against cancer. CD19-specific CAR T cells have recently been approved for the treatment of B cell malignancies in children and adults (1, 2), and numerous CAR T cell products that target alternative cancer antigens are under clinical investigation. CAR T cells are administered as a single-shot “living drug” treatment. They are able to persist in patients for several years and to undergo sequential expansion, contraction, and re-expansion in vivo upon (re-)exposure to antigen (3, 4). With these attributes, CAR T cells are fundamentally different from conventional pharmacologic compounds that decay with predictable half-life and that have to be administered repeatedly to sustain the therapeutic effect.

The pharmacokinetics of CAR T cells in vivo depend on several intrinsic and extrinsic factors, for example, product phenotype and composition, tumor burden, and lymphodepleting treatment before CAR T cell infusion. In addition, the functional activity of CAR T cells varies between patients, and defining dosing regimens that have consistent efficacy with acceptable toxicity remains challenging (3, 4). As a consequence, CAR T cell therapy has been associated with substantial acute and chronic side effects that have restricted clinical utilization to medically fit patients at highly specialized cancer centers (5, 6). The most common acute toxicity associated with CAR T cell therapy is cytokine release syndrome (CRS), which is triggered by release of inflammatory cytokines from CAR T cells and, subsequently, from innate immune cells that produce the key CRS cytokine interleukin 6 (IL-6) (68). The clinical management of CRS includes attempts at neutralizing IL-6 using receptor antagonists and systemic immunosuppression with steroids (5). However, there is currently no technology that enables physicians (and patients) to exert direct control over the activity and function of CAR T cells in vivo.

CARs are synthetic receptors, and several distinct designs have been confirmed to confer clinical activity. All of these designs use CD3ζ as part of their signaling module to induce T cell activation (1, 2). The signaling cascade after CAR engagement is similar to the signaling induced by the endogenous T cell receptor (TCR) and involves autophosphorylation of the SRC family kinase lymphocyte-specific protein tyrosine kinase (LCK), LCK-mediated phosphorylation of CD3ζ and ζ-chain of TCR-associated protein kinase 70 kDa (ZAP70), and eventually, the induction of transcription factors such as nuclear factor of activated T cells (NFAT) and nuclear factor κ light-chain enhancer of activated B cells (NF-κB) (9, 10). We hypothesized that interfering with signaling events in this cascade would provide an effective means to control the function of CAR T cells. In search for a pharmacologic on/off switch for CAR T cells, we focused on the tyrosine kinase inhibitor (TKI) dasatinib as the lead compound. Dasatinib has been developed as an inhibitor of the breakpoint cluster region– Abelson murine leukemia viral oncogene homolog 1 (BCR-ABL) fusion protein (11), and is clinically approved for the first-line treatment of Philadelphia chromosome–positive chronic myelogenous leukemia and acute lymphoblastic leukemia. In addition, dasatinib blocks the adenosine triphosphate binding sites of LCK (11, 12), which we demonstrate ablates CAR signaling and causes an immediate blockade of CD8+ and CD4+ CAR T cell function. This blockade does not affect CAR T cell viability and is rapidly and completely reversible upon dasatinib removal. In this study, we demonstrate that dasatinib can be applied to steer the function of CAR T cells in vitro and in vivo and introduce dasatinib as a clinically available pharmacologic on/off switch for CAR T cells.

RESULTS

Dasatinib locks resting CD8+ and CD4+ CAR T cells into a function-off state

To evaluate the effect of dasatinib on CAR T cell function, we first used a CD19-CAR with 4-1BB costimulation (CD19-CAR/4-1BB; fig. S1A) that conferred high rates of complete remissions in recent clinical trials (3). We enriched CAR-expressing T cells to ≥90% purity (Fig. 1A) and performed coculture assays with CD19+ target cells (K562/CD19) in the presence or absence of dasatinib. With CD8+ CAR T cells, complete blockade of target cell lysis was accomplished when dasatinib was used at a concentration of 100 nM (Fig. 1B and fig. S1B), which can be readily obtained in human serum (13). Treatment with 100 nM dasatinib also conferred complete inhibition of interferon-γ (IFN-γ) and IL-2 secretion (Fig. 1, C and D, and fig. S1C) and CAR T cell proliferation (Fig. 1E). The blockade was sustained for several hours after a single dasatinib administration at the time of assay setup, and accordingly, we did not observe induction of the activation markers CD25 and CD69 during the assay period of 24 hours (Fig. 1F and fig. S1D). We observed partial inhibition of CAR T cell cytolytic activity, cytokine secretion, and proliferation when dasatinib was used at concentrations ≤25 nM (Fig. 1, B to E).

Fig. 1 Dasatinib locks resting CD8+CAR T cells into a function-off state.

(A) Phenotype of CD8+ CAR T cells. Truncated epidermal growth factor receptor (EGFRt) is the transduction marker encoded with the CAR-transgene. Numbers indicate percentages of parental population. (B to F) CD8+ CD19-CAR/4-1BB T cells were cocultured with K562/CD19 in the presence of indicated amounts of dasatinib, which was added at start of the assay. (B) The percentage of lysed target cells was determined in 1-hour intervals over a period of 12 hours [effector–to–target cell (E:T) ratio, 5:1]. (C and D) Enzyme-linked immunosorbent assay for IFN-γ (C) and IL-2 (D). Left diagrams show data obtained in n = 3 experiments with T cells from different donors, normalized to the amount of cytokines secreted by CAR T cells in the absence of dasatinib (100%). Right diagrams show optical density (OD) from one representative experiment. (E) Proliferation of CAR T cells 72 hours after stimulation with K562/CD19 target cells. Graph shows data from quantitative analysis of n = 3 independent experiments. Histograms show data from one representative experiment. The remaining proliferation was calculated using the proliferation index and normalized to the proliferation of CD19-CAR/4-1BB T cells in the absence of dasatinib. CFSE, carboxyfluorescein diacetate succinimidyl ester; FITC, fluorescein isothiocyanate. (F) Surface expression of activation markers on CAR T cells without stimulation (no stim) or 20 hours after stimulation with K562/CD19 in the absence (0 nM) and presence (100 nM) of dasatinib. Data shown are mean values + SD (B) or ± SD (C to E) with *P ≤ 0.05, **P ≤ 0.01 by two-way analysis of variance (ANOVA) (B) or Kruskal-Wallis test (C to E).

CAR T cell products in clinical application commonly contain CD8+ cytotoxic and CD4+ helper T cells. We confirmed that dasatinib was also effective with CD4+ CD19-CAR/4-1BB T cells (fig. S2) and found that 100 nM dasatinib completely abrogated the production of cytokines including granulocyte-macrophage colony-stimulating factor (GM-CSF), IFN-γ, and IL-2 after stimulation with target cells (fig. S2D). Collectively, these data show that dasatinib can block the activation of resting CD8+ and CD4+ CAR T cells after stimulation with target cells. The dose of dasatinib can be titrated to achieve partial and complete inhibition of CAR T cell function.

Dasatinib is superior to dexamethasone in inhibiting CAR T cell function

We compared the inhibitory effect of dasatinib with that of dexamethasone, which is commonly used with the intention of exerting control over CAR T cells (5). First, we analyzed the cytolytic activity of CD8+ CAR T cells (CD19-CAR/4-1BB) in the presence of dexamethasone. When dexamethasone was added at assay setup, there was no inhibitory effect (fig. S3A). However, after a 24-hour pretreatment with dexamethasone, we observed partial inhibition of cytolytic activity (fig. S3A). Neither simultaneous treatment nor pretreatment of CAR T cells with dexamethasone resulted in reduced IFN-γ secretion (fig. S3B), whereas IL-2 secretion was reduced in both settings (fig. S3B). These data correlated with reduced proliferation of CAR T cells after exposure to dexamethasone compared to untreated CAR T cells (fig. S3C). In contrast, treatment with dasatinib mediated complete inhibition of cytolytic activity, cytokine secretion, and proliferation (fig. S3). Together, these data show that dasatinib is superior to dexamethasone in inhibiting CAR T cell function. Dasatinib is able to confer immediate onset and complete control over CAR T cell function, whereas the inhibitory effect of dexamethasone is of delayed onset and incomplete, even when dexamethasone is used at very high doses of 100 μM.

Dasatinib prevents LCK phosphorylation and NFAT induction after CAR engagement

We sought to elucidate the molecular mechanism behind the inhibitory effect of dasatinib on CAR T cell function. We performed Western blots to determine the phosphorylation state of LCK (Y394), CAR-CD3ζ (Y142), and ZAP70 (Y319). In the presence of dasatinib, phosphorylation of all three proteins was inhibited (Fig. 2A and fig. S4). Quantitative analyses revealed that phosphorylation relative to non–dasatinib-treated CAR T cells was reduced to 22 (LCK), 13 (CD3ζ), and 12% (ZAP70) (Fig. 2B).

Fig. 2 Dasatinib prevents LCK phosphorylation and NFAT induction after CAR engagement.

For Western blot analysis, CD19-CAR/4-1BB T cells were stimulated with CD19+ RCH-ACV and were either treated with 100 nM dasatinib (+) or not (−). (A) Phosphorylation of LCK (Y394), CAR-CD3ζ (Y142), and ZAP70 (Y319) in one representative experiment. (B) Quantitative analysis of Western blot data obtained in n = 3 experiments, normalized to total protein expression. (C) CD19-CAR/4-1BB T cells were transduced with an NFAT/GFP reporter gene and stimulated with CD19+ Raji or CD19 K562 at an E:T ratio of 5:1. Diagram shows reporter gene expression obtained in CD8+ (left) and CD4+ (right) in the presence (+) or absence (−) of dasatinib (100 nM). (B and C) Data are mean values ± SD analyzed by Kruskal-Wallis test (B) or ordinary one-way ANOVA (C) with *P ≤ 0.05, ***P ≤ 0.001. MFI, mean fluorescence intensity.

We hypothesized that inhibition of CAR signaling by dasatinib would prevent induction of NFAT, a key transcription factor in activated CAR T cells (10). Therefore, we cotransduced CAR T cells with a reporter gene encoding a short-lived green fluorescent protein (GFP) mutant variant under control of an NFAT-inducible promotor (NFAT/GFP; fig. S5). Subsequently, we challenged CD8+ and CD4+ CAR T cells with CD19+ Raji lymphoma cells and found that induction of the GFP reporter signal was indeed significantly reduced after dasatinib treatment [P = 0.0495 (CD8+) and P = 0.0006 (CD4+); Fig. 2C].

These data suggested that dasatinib may be broadly effective with CAR constructs that contain a CD3ζ signaling module. Accordingly, we confirmed that dasatinib was also able to inhibit the effector functions of CD8+ and CD4+ T cells expressing a CD19-CAR with CD28 costimulation (CD19-CAR/CD28) (figs. S6 and S7), as well as receptor tyrosine kinase–like orphan receptor 1 (ROR1)–specific (fig. S8) and signaling lymphocytic activation molecule family member 7 (SLAMF7)–specific CARs (fig. S9). We found that the inhibitory effect of dasatinib was equally potent with CAR constructs that contained a CD28_CD3ζ versus 4-1BB_CD3ζ signaling domain and identified 40 nM of dasatinib as the benchmark dose at which complete inhibition of CAR T cell function was accomplished with each CAR construct (figs. S1, S2, and S6 to S9). Together, these data show that dasatinib reduces the phosphorylation of LCK, CD3ζ, and ZAP70, resulting in limited induction of NFAT. This mechanism of action suggests that dasatinib is broadly applicable as a pharmacologic function-off switch for CAR constructs that use this signaling machinery.

The dasatinib-induced function-off state in CAR T cells is not compromised by TCR stimulation

The potent inhibitory effect of dasatinib on CD3ζ phosphorylation suggested that the function-off state in CAR T cells could be sustained even when in-trans stimulation was provided through the endogenous TCR. We generated polyclonal CD8+ CAR T cell lines (CD19-CAR/4-1BB) that recognized the cytomegalovirus (CMV) pp65/HLA-A2 NLVPMVATV (NLV) epitope through their endogenous TCR. We confirmed coexpression of CD19-CAR and CMV-specific TCR by flow cytometry (fig. S10A). Then, we stimulated T cells with K562 cells that expressed HLA-A2 either alone (pulsed with pp65 NLV peptide) or in combination with CD19 and assessed function over time. In both cases, dasatinib completely blocked T cell activation and effector function even when CD19-CAR and CMV-specific TCR were engaged simultaneously (fig. S10, B to E). Together, these data show that the dasatinib-induced function-off in CAR T cells is not compromised by stimulation through the endogenous TCR.

Dasatinib pauses activated CAR T cells in a function-off state

We were interested in determining whether dasatinib was able to inhibit CAR T cells that are already in an activated state, meaning in the process of executing their effector functions. To assess this, we activated CAR T cells with antigen-positive target cells to turn them on and then added dasatinib to turn them off. CD8+ CD19-CAR/4-1BB T cells rapidly started lysing target cells in the function-on phase. However, as soon as dasatinib was added to the culture (1 hour after assay setup), target cell lysis stagnated, indicating that CAR T cell function was switched off (Fig. 3A). Similarly, addition of dasatinib to activated CD8+ CAR T cells interfered with subsequent cytokine secretion and proliferation (Fig. 3, B and C). Consistent with the mechanism of action revealed above, the inhibitory effect of dasatinib was less strong when CAR T cells had progressed beyond a certain point of activation. If added later than 1 hour after turning CAR T cells on, then the inhibitory effect of dasatinib on subsequent proliferation was low (Fig. 3C). We reasoned that activated CAR T cells that were not inhibited by dasatinib during a primary encounter with target cells would eventually be captured upon a sequential encounter with target cells. Accordingly, we performed a sequential stimulation study where Raji lymphoma cells were added every 24 hours to CD8+ and CD4+ CD19-CAR T cells that coexpressed the NFAT/GFP reporter gene. The data show that dasatinib was able to block T cell activation upon secondary and tertiary stimulation with target cells (fig. S11). In summary, these data demonstrate that dasatinib can pause activated CAR T cells and induce a function-off state that is stable even when CAR T cells encounter target cells sequentially.

Fig. 3 Dasatinib pauses activated CAR T cells in a function-off state.

CAR T cells were treated with 100 nM dasatinib added either at assay setup or 1, 2, 3, or 48 hours after stimulation with K562/CD19 target cells. (A) Lysis of target cells measured in 1-hour intervals over a 10-hour period. Dasatinib (100 nM) was added 1 hour after assay setup (orange squares) to switch CAR T cells off after the initial on phase. For comparison, the diagram shows lysis of target cells by untreated CAR T cells (blue circles) and by T cells that were treated with dasatinib at assay setup (red circles). (B) Diagram shows the concentrations of IFN-γ (left) and IL-2 (right). Data were normalized to the amount of cytokines produced by untreated CAR T cells (−, represented by the dotted line). CAR T cells were treated either at the beginning (0) or 2 hours (2) after assay setup. (C) Proliferation was calculated on the basis of the cell proliferation index and normalized to proliferation of untreated CAR T cells (−, represented by the dotted line). (A to C) Data shown are mean values + SD (A) or ± SD (B and C) with *P ≤ 0.05, ***P ≤ 0.001 by two-way ANOVA (A), or Kruskal-Wallis test (B and C).

Removal of dasatinib rapidly releases CAR T cells from their function-off state

Next, we were curious whether the function-off state could be reversed and CAR T cells could be shifted back into a function-on state through removal of dasatinib. We coincubated CD19-CAR/4-1BB T cells with target cells in the presence of dasatinib (function off) and did not observe cytolytic activity, whereas in parallel cultures without dasatinib, target cell lysis started immediately (Fig. 4A). After 2 hours, dasatinib was removed from the cocultures through a thorough medium change. The data show that after dasatinib removal, CAR T cells rapidly went into a function-on state and commenced target cell lysis. Within 2 hours of dasatinib removal, CAR T cells had achieved half of the specific lysis and within 7 hours had achieved the same extent of specific lysis as untreated CAR T cells.

Fig. 4 Removal of dasatinib rapidly releases CAR T cells from their function-off state.

(A) Cytolytic activity of CD19-CAR/4-1BB T cells against K562/CD19. Dasatinib (100 nM) was present in the first 2 hours after assay setup (function-off phase) and was removed at t = 0 (function-on phase). Cytolytic activity without dasatinib (0 nM) is shown for comparison. (B to E) T cells were pretreated for 1 or 7 days [dasa pre (+)/1 or 7] with 100 nM dasatinib or kept in culture without dasatinib [dasa pre (−)]. For functional analysis, T cells were washed and cocultured with target cells in the presence [dasa during (+)] or absence [dasa during (−)] of 100 nM dasatinib. (B) Cytolytic capacity of CAR T cells treated with dasatinib at defined schedules. (C and D) Concentrations of IFN-γ (C) and IL-2 (D) after 1 and 7 days of treatment (“Dasa pre”) normalized to the amount of cytokines produced by untreated T cells (−, represented by the dotted line). Dasatinib (100 nM) was added at assay setup in the indicated groups (“Dasa during +”). (E) The proliferation was calculated on the basis of the cell proliferation index and normalized to proliferation of untreated CAR T cells (represented by the dotted line). (F) T cells were treated with 100 nM dasatinib for 7 days (+) or left untreated (−). Viability was analyzed on day 0 (d0; baseline) and on days 2, 4, and 8 by flow cytometry to distinguish alive, early-, and late-stage apoptotic cells (n = 1). (A to E) Data shown are mean values + SD (A) or ± SD (B to E) with **P ≤ 0.01, ***P ≤ 0.001 by two-way (A and B) or one-way ANOVA (C to E). n.s., not significant.

We were interested in determining how long the function-off state could be maintained in CAR T cells. We found that complete blockade of CAR T cell function was sustained after 1 day and also after 7 days of continuous exposure to dasatinib, with no signs of the inhibitory effect wearing off (Fig. 4, B to E). Even after a 7-day exposure to dasatinib, removal of the drug reignited CAR T cell function, as evidenced by specific high-level cytolytic activity, IFN-γ, and IL-2 secretion and productive proliferation, which occurred with similar kinetics as in CAR T cells that had never been exposed to dasatinib (Fig. 4, B to E). We determined the percentage of live, early apoptotic, and late apoptotic T cells at several time points during and after the 7-day exposure to dasatinib and did not observe a negative effect of dasatinib treatment on CAR T cell viability (Fig. 4F). In summary, these data show that dasatinib induces a function-off state in CAR T cells that can be sustained for several days without affecting CAR T cell viability and rapidly reversed into a function-on state after removal of the drug.

Dasatinib pauses activated CAR T cells in a function-off state in vivo

To obtain proof of concept for dasatinib as a pharmacologic on/off switch for CAR T cells in vivo, we used a lymphoma xenograft model in immunodeficient NOD.Cg-Prkdcscid IL-2rgtm1Wjl/SzJ (NSG) mice. Mice were inoculated with firefly luciferase (FFLUC)/GFP-expressing Raji lymphoma cells on day −7 and received CD19-CAR/4-1BB–modified or control T cells on day 0. In this set of experiments, we administered dasatinib between days 3 and 5 after T cell transfer to create a “function-on-off-on” sequence (Fig. 5A). In the first phase (days 0 to 3), CD19-CAR T cells commenced exerting their anti-lymphoma activity and were strongly activated, as demonstrated by bioluminescence imaging (BLI) (Fig. 5, B to D) and serum cytokine analysis (function on) (Fig. 5E). In the second phase after CAR T cell transfer (days 3 to 5), we administered dasatinib to a subgroup of mice. Dasatinib rapidly induced a function-off state and halted anti-lymphoma reactivity, as evidenced by increasing BLI signal (Fig. 5C) and decreasing IFN-γ in serum from each of the mice in this subgroup (fig. S12). In contrast, the BLI signal did not increase, and IFN-γ remained stable during this phase in mice that had received CD19-CAR T cells but no dasatinib. In the third phase (after day 5), administration of dasatinib was discontinued to allow CAR T cells to revert back into their function-on state. CAR T cells rapidly resumed their anti-lymphoma function as revealed by decreasing BLI signal (Fig. 5C) and increasing serum IFN-γ in each of the mice (fig. S12). Similar data were obtained in analogous experiments with T cells expressing CD19-CAR with CD28 costimulation (Fig. 5F).

Fig. 5 Dasatinib pauses activated CAR T cells in a function-off state in vivo.

(A) Treatment schedule and experimental setup. NSG mice received either 5 × 106 CD19-CAR/4-1BB T cells or untransduced T cells (ctrl) on day 0 (function-on phase). Dasatinib (10 mg/kg) was administered to indicated groups from days 3 to 5 to create a function-off phase. Afterward, dasatinib was discontinued (function-on phase). (B) Diagram shows average tumor burden (mean) in each treatment group in the function-on-off-on sequence. (C) The ventral BLI signal is displayed for individual mice for up to 94 days after T cell infusion. (D) Bar diagrams show tumor progression/regression in percent during on (d0 to d3), off (d3 to d5), and on phase (d5 to d7) with data for individual mice in each treatment group using CD19-CAR/4-1BB T cells. (E) IFN-γ serum concentration during on (days 3, 7, and 10) and off phases (day 5). Data shown are mean and individual values. (F) Bar diagrams show lymphoma progression/regression in percent during the on-off-on sequence for individual mice in each treatment group using CD19-CAR/CD28 T cells. (G) The presence of human T cells (CD3+/CD45+) in peripheral blood of mice is depicted as percentage of living cells (7AAD). Blood samples were analyzed by flow cytometry. Shaded areas (B, C, E, and G) indicate the function-off phase. (D to G) Statistical analysis was performed to compare ctrl (on) and CAR (on) or CAR (on) and CAR (on/off/on) by Mann-Whitney test (D and F) or two-way ANOVA (E and G) with *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.

We analyzed peripheral blood at several time points throughout the experiment and, on day 7, found a lower frequency of CAR T cells in the subgroup of mice that had received dasatinib, suggesting that their proliferation had been effectively controlled during the function-off phase (Fig. 5G). However, with subsequent follow-up, we found a potent anti-lymphoma effect in each of the mice that had gone through the function-on-off-on sequence, demonstrating that the inhibitory effect of dasatinib was indeed reversible and that the therapeutic effect of CD19-CAR T cells had not been compromised by this intervention (Fig. 5C).

Collectively, these data demonstrate that dasatinib can be used as a pharmacologic control switch to steer CAR T cells in vivo. The data show that administration of dasatinib for a short-term interval does not compromise the anti-lymphoma efficacy of CD19-CAR T cells in this in vivo model and confirm the favorable pharmacodynamic attributes of dasatinib, including rapid onset and complete reversion of the inhibitory effect that we established in our in vitro experiments.

Dasatinib can be used as an emergency drug to prevent CRS

We sought to determine whether dasatinib is suitable as an emergency drug to prevent fatal CRS in a murine model that we recently reported (7). In this model, C.B.Igh-1b/GbmsTac-PrkdcscidLystbg N7 [severe combined immunodeficient (SCID)/beige] mice are engrafted with Raji lymphoma cells and then treated with CD19-CAR/CD28 T cells, which cause rapid-onset CRS with a high rate of mortality within the first 48 hours (Fig. 6A). In this model, we created a function-on-off-on sequence and, because of the aggressive nature of CRS, commenced administering dasatinib at 3 hours after CAR T cell transfer for a 30-hour period. We measured human serum cytokines at 8 hours after CAR T cell transfer and found significantly lower concentrations of IFN-γ (P < 0.0001), IL-2 (P = 0.0002), GM-CSF (P < 0.0001), and tumor necrosis factor–α (TNFα) (P = 0.0006) in each of the mice that had received dasatinib compared to mice that had not (Fig. 6B). We repeated this analysis at 16 hours after CAR T cell transfer and detected no increase in serum cytokines, demonstrating that cytokine production from CAR T cells was under complete control (Fig. 6B). We also found lower concentrations of murine IL-6 that is produced by endogenous innate immune cells as part of CRS pathophysiology (Fig. 6C). During acute CRS in the first 48 hours after CAR T cell transfer, we observed a high rate of mortality, with only 25% survival in the group of mice that had not received dasatinib (Fig. 6D). By contrast, in the dasatinib treatment group, 70% of mice were alive at the 48-hour interim analysis. With subsequent follow-up until 168 hours, survival rates remained stable in both groups (Fig. 6D).

Fig. 6 Dasatinib mitigates CRS in vivo.

(A) Treatment schedule and experimental setup. SCID/beige mice received CD19-CAR/CD28 T cells at t = 0. Five doses of dasatinib were administered between t = 3 hours and t = 33 hours. Afterward, dasatinib was discontinued. i.p., intraperitoneally. Human (B) and mouse (C) cytokines were measured in serum 8 and 16 hours after T cell infusion. Data shown are mean values from n = 3 (ctrl) and n = 4 (dasa) treated animals. Statistical analysis was performed using unpaired t test (B and C) with *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001. (D) Kaplan-Meyer survival plot for mice receiving CAR and vehicle only (ctrl, n = 12) or CAR and dasatinib (dasa, n = 13) (statistical analysis by Mantel-Cox test, P = 0.0951).

Collectively, these data show that a short treatment course with dasatinib can prevent fatal CRS after CAR T cell transfer in mice. These data highlight the potential of dasatinib as a pharmacologic on/off switch for CAR T cells to steer their function and to reduce toxicity in clinical settings.

DISCUSSION

CAR T cells are living anticancer “drugs” that are essentially out of control after administration to the patient. Because of their potency, there is an unmet desire for technologies that provide control over CAR T cells and their effector functions. Here, we demonstrate that the TKI dasatinib can be used as a pharmacologic on/off switch for CAR T cells. Our data show that dasatinib confers titratable partial to complete control over cytolytic activity, cytokine secretion, and proliferation in CD8+ and CD4+ CAR T cells. The data also show that treatment with dasatinib retains resting CAR T cells in an inactive function-off state, pauses CAR T cells that are captured early during their activation phase, and prevents their subsequent activation if they continue to encounter antigen-expressing target cells. Intriguingly, dasatinib treatment does not affect the viability of CAR T cells, and after discontinuation, the function-off state is rapidly and completely reversible. These features distinguish this pharmacologic on/off switch from conventional safety switches that have been developed with the intention to terminate CAR T cells in case of severe toxicity. Depletion markers such as EGFRt or CD20 can be applied to remove CAR T cells through administration of the cognate mAb (14, 15), and suicide switches such as inducible caspase 9 or herpes simplex virus–derived thymidine kinase can be triggered to remove CAR T cells within several hours or a few days after administering an inducer drug (16, 17). Unfortunately, these conventional safety switches can only be triggered once and terminate the antitumor effect (14). As a consequence, physicians and patients have been reluctant to use these safety switches, even when side effects of CAR T cells were severe (3, 5).

The concept underlying the pharmacologic control switch with dasatinib is that in the absence of the drug, CAR T cells are able to respond to antigen stimulation and to exert their antitumor function, and in the presence of the drug, they are in an inactive state. We demonstrate that treatment of CAR T cells with dasatinib results in reduced phosphorylation of LCK, CD3ζ, and ZAP70. After CAR engagement, LCK undergoes autophosphorylation and confers phosphorylation of CD3ζ and ZAP70 (9). At the dose range used in this study (up to 100 nM), dasatinib inhibits LCK but not ZAP70 (18), and thus, inhibition of LCK is the likely mechanism underlying the inhibitory effect of dasatinib on CAR signaling.

Structural analyses have revealed that the interaction between dasatinib and LCK is reversible (11). Accordingly, we found that the blockade of CAR T cell function is reversible after removal of dasatinib in our in vitro assays and discontinuation of dasatinib administration in our in vivo models. To induce and sustain the function-off state in CAR T cells, a steady concentration of ≥40 nM dasatinib is required. Analyses of dasatinib’s pharmacokinetics in humans have shown that serum concentrations of ≥40 nM can be readily achieved and sustained even through an oral route of administration (13). In humans, dasatinib has a serum half-life of about 4 hours (13), which is longer than in mice (19). In our mouse model, we administered dasatinib every 6 hours to maintain the serum concentration above the required threshold. The intended time frame for the clinical use of dasatinib as a pharmacologic control switch for CAR T cells is on the order of a few hours up to several consecutive days, and within this short time frame, no considerable dasatinib-related toxicity is anticipated.

There are multiple clinical scenarios where CAR T cell–induced toxicity is of transient nature, for example, in patients who experience tumor lysis syndrome, CRS, or neurotoxicity after CD19-CAR T cell therapy (3, 5, 6). In these settings, it would be sufficient to place CAR T cells into a temporary function-off phase and to reignite their function once the adverse event is over. Our data show that dasatinib can exert temporary control over CAR T cell function with rapid onset upon exposure and rapid release upon removal of the drug. In our experiments, we have retained CAR T cells in a function-off state for time periods between 2 hours and 7 days. Even after 7 days, the inhibitory effect of dasatinib on CAR T cell function did not diminish, and CAR T cells reignited their antitumor function after withdrawal of the drug.

A potential limitation for the clinical use of dasatinib as an emergency drug is that the inhibitory effect on CAR T cells that are already activated is weaker with the consequence that a proportion of CAR T cells may only be blocked upon subsequent antigen encounter. Encouragingly, we found that dasatinib was capable of effectively blocking cytokine secretion from a polyclonal, nonsynchronized CAR T cell population in vivo and of preventing morbidity and mortality from CRS. In our in vivo CRS model, we administered dasatinib within 3 hours after CAR T cell administration when mice presented with imminent CRS. Further studies are warranted to determine whether dasatinib is also effective in clinical situations with established CRS.

We compared dasatinib to dexamethasone, which is used clinically in attempts at gaining control over CAR T cells in case of toxicity. Our data show that dexamethasone exerts inferior control over CAR T cell function and acts more slowly than dasatinib. This is consistent with dexamethasone’s mode of action, which is not through direct interference with CAR signaling but through impairment of NF-κB induction at the transcriptional level. The pattern and extent of functional inhibition by dexamethasone in our study are comparable with earlier work in non–CAR-modified T cells (20).

Previous studies have shown that dasatinib affects the function of T cells that recognize antigen through their endogenous TCR (12). In several studies, dasatinib with a concentration of 10 to 50 nM (dependent on the respective antigen) completely inhibited TCR signaling in virus-specific and tumor-reactive T cells, consistent with our data in CMV-specific CAR-modified T cells that we stimulated through their endogenous TCR. A practical consequence is that, in the presence of dasatinib, not only CAR T cells but also endogenous (non–CAR-modified, TCR-restricted) T cells will be inhibited; however, as a short-term intervention to prevent or mitigate toxicity, this seems acceptable.

In this study, we focused our efforts on evaluating dasatinib as a control drug for CAR T cells. However, dasatinib may also have merit as an enhancer of antitumor potency. We and others have demonstrated that poorly calibrated or sustained CAR signaling in T cells may result in a state of exhaustion (21) or activation-induced cell death that limits the antitumor efficacy of T cell therapy (2224). The metronomic use of dasatinib, with short treatment courses that are administered repeatedly at defined intervals, to protect CAR T cells from chronic signaling may be able to prevent exhaustion. The concept of dasatinib-mediated reversible control over CAR T cell function presented in this study is attractive because it provides effective and complete control over CAR T cells and conserves their therapeutic potential. However, during a function-off phase, the antitumor effect of CAR T cells is on hold, which may limit the ability to use dasatinib in some patients with rapidly progressing tumors. Nevertheless, in our lymphoma xenograft model, a 48-hour function-off phase did not compromise the overall therapeutic outcome and benefit of CAR T cell therapy, and we anticipate that this would also not be the case in a clinical setting. There is more than a decade-long experience with the clinical use of dasatinib in hematology, and thus, the evaluation and implementation of dasatinib as an on/off control drug in CAR T cell immunotherapy should be feasible and straightforward.

MATERIALS AND METHODS

Study design

In vitro experiments and experiments in NSG mice were done three times with T cells from different healthy donors. Experiments in SCID/beige mice were done once. In mouse experiments, the group size was determined by biostatistical calculation using G*Power based on expected effect size, which is the mean difference in SD units detectable with 80% power at a two-sided 0.05 level of significance with two-sample t test, and the SD fraction, which is SD as a fraction of biologically meaningful difference. Mice were randomly allocated to treatment groups before tumor injection (NSG model) or before T cell infusion (SCID/beige model). Data analysis was based on objectively measurable data; the investigators were not blinded to group allocation during data collection and analysis.

Human participants

Blood samples were obtained from healthy donors who provided written informed consent to participate in research protocols approved by the Institutional Review Board (IRB) of the University of Würzburg [Universitätsklinikum Würzburg (UKW)] or were purchased from the New York Blood Center (IRB-exempted) and handled following all required ethical and safety procedures set forth by the Memorial Sloan Kettering Cancer Center (MSKCC). Peripheral blood mononuclear cells were isolated by centrifugation over Ficoll-Hypaque (Sigma).

Experiments in NSG mice

The UKW Institutional Animal Care and Use Committee approved all experiments in NSG mice. NSG mice (female, 6 to 8 weeks old) were purchased from Charles River Laboratories. Mice were inoculated with 1 × 106 FFLUC/GFP-expressing Raji cells via intravenous injections (tail vein). Native Raji cells were obtained from the German collection of microorganisms and cell cultures (DSMZ) and transduced with FFLUC/GFP encoding lentivirus. Tumor burden and distribution were analyzed by serial BLI on an IVIS Lumina imager (PerkinElmer): Mice received luciferin (0.3 mg/g) intraperitoneally, and images were acquired 10 min after luciferin injection in small binning mode at an acquisition time of 1 s to 1 min to obtain unsaturated images. Data were analyzed using Living Image Software (Caliper), and the average radiance was analyzed in regions of interest that encompassed the entire body of each mouse. Mice were treated with 5 × 106 CAR-modified or control untransduced T cells (CD4:CD8 ratio = 1:1) intravenously after 7 days of tumor engraftment. Dasatinib was administered intraperitoneally at a dose of 10 mg/kg. The presence of human cytokines in serum was measured using multiplex cytokine analysis (Thermo Fisher Scientific).

Experiments in SCID/beige mice

The MSKCC Institutional Animal Care and Use Committee approved all experiments in SCID/beige mice. Experiments were performed as previously described (7). In brief, primary human T cells were purified from buffy coats by negative bead selection (Pan T cell Isolation Kit, Miltenyi Biotec) and activated with CD3/CD28 CTS Dynabeads (Thermo Fisher Scientific) at a ratio of one bead per cell. T cells were cultured in X-VIVO 15 (Lonza) supplemented with 5% human serum (Gemini), 10 mM HEPES (Invitrogen), 2 mM GlutaMAX (Invitrogen), 1× minimum essential medium vitamin solution (Invitrogen), and 1 mM sodium pyruvate (Invitrogen). In addition, the medium was supplemented with penicillin-streptomycin (50 IU/ml; Invitrogen) and recombinant IL-2 (60 IU/ml; Novartis). SCID/beige mice (female, 6 to 8 weeks old, Taconic Biosciences) were inoculated with 3 × 106 FFLUC/GFP-expressing Raji tumor cells (American Type Culture Collection) via intraperitoneal injection, and tumors were left to grow for 20 days. Tumor burden was evaluated by BLI using IVIS Imaging System (PerkinElmer) and Living Image Software (Caliper) 2 days before CAR T cell transfer. Mice were treated with 3 × 107 CD19-CAR/CD28 (7) CAR-modified T cells via intraperitoneal injection. Dasatinib was administered intraperitoneally in five doses (every 6 ± 1 hours) at 10 mg/kg starting 3 hours after CAR T cell injection. Blood was collected from mice by tail clipping. Serum cytokines were measured by cytometric bead array (BD Biosciences).

Statistical analysis

Data from in vitro experiments were analyzed and plotted with GraphPad Prism (GraphPad Software Inc). The mean value + SD, unless indicated otherwise, is shown. Specific lysis was analyzed using two-way ANOVA with Sidak’s multiple comparison test. Cytokine secretion and proliferation were analyzed using ordinary one-way ANOVA with Holm-Sidak’s multiple comparison test or Kruskal-Wallis test with Dunn’s multiple comparison test as indicated in the figure legends. Data on quantitative Western blots were analyzed by Kruskal-Wallis test with Dunn’s multiple comparison test. In NFAT/GFP reporter experiments, GFP expression was analyzed by one-way ANOVA with Holm-Sidak’s multiple comparison test. In NSG mouse experiments, datasets were analyzed by Mann-Whitney test (tumor progression/regression) and two-way ANOVA with Sidak’s multiple comparison test (cytokine secretion and flow cytometry). In SCID/beige experiments, datasets were analyzed by unpaired t test (cytokines) and Mantel-Cox test (survival). Original data are included in data file S1.

SUPPLEMENTARY MATERIALS

stm.sciencemag.org/cgi/content/full/11/499/eaau5907/DC1

Materials and Methods

Fig. S1. Dasatinib inhibits CD8+ CD19-CAR/4-1BB cells in a dose-dependent manner.

Fig. S2. Dasatinib locks resting CD4+ CD19-CAR/4-1BB T cells into a function-off state.

Fig. S3. Dasatinib exerts greater control over CAR T cells than dexamethasone.

Fig. S4. Dasatinib prevents LCK phosphorylation.

Fig. S5. Dasatinib prevents NFAT induction after CAR engagement.

Fig. S6. Dasatinib locks resting CD8+ CD19-CAR/CD28 T cells into a function-off state.

Fig. S7. Dasatinib locks resting CD4+ CD19-CAR/CD28 T cells into a function-off state.

Fig. S8. Dasatinib inhibits the effector function of CD8+ and CD4+ T cells expressing a ROR1-specific CAR.

Fig. S9. Dasatinib inhibits the effector function of CD8+ and CD4+ T cells expressing a SLAMF7-specific CAR.

Fig. S10. The dasatinib-induced function-off state in CAR T cells is not compromised by TCR stimulation.

Fig. S11. Dasatinib blocks CAR T cell activation upon repeated antigen encounter.

Fig. S12. Dasatinib pauses activated CAR T cells in a function-off state in vivo.

Data file S1. Original data (provided as an Excel file).

References (25, 26)

REFERENCES AND NOTES

Acknowledgments: We thank E. Spirk for technical assistance and L. Wallstabe and T. Gogishvili for discussions in the course of the project. Funding: This work was supported by the m4 Award in Personalized Medicine of the Cluster Biotechnologie Bayern and Free State of Bavaria (grant no. BIO-1601-0002 to M.H.), a grant from the German Cancer Aid (Deutsche Krebshilfe, Max Eder Program grant no. 70110313 to M.H.), and intramural grants from the Interdisziplinäres Zentrum für Klinische Forschung (IZKF) at the University of Würzburg (grant no. Z4/109 and D-244 to M.H.). M.H. is a member of the Young Scholar Program (Junges Kolleg) and Extraordinary Member of the Bavarian Academy of Sciences (Bayerische Akademie der Wissenschaften). Cytokine Release Syndrome experiments were performed with the support of the MSKCC Support Grant/Core Grant (P30 CA008748). T.G. was supported by the Alexander S. Onassis public benefit foundation. Author contributions: K.M. designed and performed experiments, analyzed data, and wrote the manuscript. T.G., J.R., and S.F. designed and performed experiments and analyzed data. T.N. and J.W. designed and performed experiments. A.M., M.S., and H.E. analyzed data and wrote the manuscript. M.H. designed experiments, analyzed data, wrote the manuscript and supervised the project. Competing interests: K.M. and M.H. are co-inventors on a patent application related to the use of dasatinib in the context of CAR T cell immunotherapy that has been filed by the University of Würzburg (PCT/EP2018/084018, “Control and modulation of the function of gene-modified chimeric antigen receptor T cells”). M.H. is a member of the speakers bureau of Celgene and Gilead. H.E. has received research support, has worked as a consultant or member of the scientific advisory board for Janssen, Celgene, Amgen, Novartis, and Bristol-Myers Squibb, and has also worked as a consultant or member of the scientific advisory board for Takeda. M.S. has research collaborations with Takeda, Atara Biotherapeutics, and Fate Therapeutics. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article