Chimeric receptors containing CD137 signal transduction domains mediate enhanced survival of T cells and increased antileukemic efficacy in vivo

Abstract

Persistence of T cells engineered with chimeric antigen receptors (CARs) has been a major barrier to use of these cells for molecularly targeted adoptive immunotherapy. To address this issue, we created a series of CARs that contain the T cell receptor-zeta (TCR-zeta) signal transduction domain with the CD28 and/or CD137 (4-1BB) intracellular domains in tandem. After short-term expansion, primary human T cells were subjected to lentiviral gene transfer, resulting in large numbers of cells with >85% CAR expression. In an immunodeficient mouse xenograft model of primary human pre-B-cell acute lymphoblastic leukemia, human T cells expressing anti-CD19 CARs containing CD137 exhibited the greatest antileukemic efficacy and prolonged (>6 months) survival in vivo, and were significantly more effective than cells expressing CARs containing TCR-zeta alone or CD28-zeta signaling receptors. We uncovered a previously unrecognized, antigen-independent effect of CARs expressing the CD137 cytoplasmic domain that likely contributes to the enhanced antileukemic efficacy and survival in tumor bearing mice. Furthermore, our studies revealed significant discrepancies between in vitro and in vivo surrogate measures of CAR efficacy. Together these results suggest that incorporation of the CD137 signaling domain in CARs should improve the persistence of CARs in the hematologic malignancies and hence maximize their antitumor activity.

Figures

Figure 1

Lentiviral gene transfer combined with αCD3/αCD28 coated magnetic bead activation of T cells permits generation of large numbers of CD19-specific chimeric antigen receptor (CAR+) T cells. (a) A schematic diagram showing the CD19-specific CAR used in this study. (b) Comparison of green fluorescent protein (GFP) expression under the control of different eukaryotic promoters in primary human CD4+ and CD8+ T cells over time. GFP fluorescence was compared in the indicated T cell subset in cells that were stimulated with αCD3/αCD28 coated beads followed by lentiviral transduction at an multiplicity of infection (MOI) of 0.2 on day 1 with vector expressing enhanced GFP under the control of the promoter indicated. Flow cytometric detection of GFP fluorescence was calibrated using Rainbow Calibration Particles (Spherotech, Lake Forest, IL) to correct for day-to-day variation. (c) αCD19-specific CAR surface expression in primary human CD4+ and CD8+ T cells. Expression was examined 6 days following transduction with the indicated CAR-encoding lentiviral vector at a MOI of ~8. (d) In vitro expansion of CD4+ and CD8+ T cells following activation with αCD3/αCD28 coated magnetic beads and transduction of the indicated CAR on day 1. Data are representative of >3 independent experiments.

Figure 2

CD19-specific CAR+ T cells demonstrate antigen-specific killing of CD19+ tumor cells. (a) CAR+ T cell cytotoxic activity towards K562 cells that are engineered to express human CD19 (K19) or target antigen negative wild-type K562 cells (K wild type). Following >10 days expansion, CAR+ T cells (Effector cells) were mixed with the K562 cells (Target cells) at the indicated ratios. Results represent the mean percent of target cell lysis as described in materials. Results are representative of three-independent experiments. (b) Cytotoxic activity of T cells expressing either the aCD19-BB-ζ or the aCD19-Δζ control receptor towards primary human ALL target cells using the same method described in (a). Error bars represent the standard error of the mean for three replicates.

Figure 3

The cytokines produced by CD19-specific CAR+ T cells are dependent on the presence of costimulatory domains within the CAR. Following >10 days of expansion, 1 × 106 CAR+ T cells, as indicated, were stimulated with K562 cells expressing either human CD19 (hatched bars) or antigen negative wild-type K562 cells (open bars). Supernatant was harvested after 24 hours of incubation, and the indicated cytokines were measured by cytokine bead array (BD Biosciences, San Jose, CA). Results are representative of two-independent experiments; BLQ, below the limit of quantification.

Figure 4

CD28 and 4-1BB costimulatory domains enhance αCD19 CAR-induced T cell proliferation in vitro with both antigen-dependent and antigen-independent effects. (a) In vitro expansion of CAR+ T cells following antigen stimulation. T cells were stimulated with αCD3/αCD28 coated magnetic beads on day 0, and transduced with the indicated CAR on day 1 using a bicistronic lentiviral vector expressing CAR along with eGFP using the 2A ribosomal skipping sequence as described in Materials. Cultures were restimulated (arrow) with either CD19+ K562 cells (K562-CD19), wild-type K562 cells (K562 wild type) or K562 cells expressing hCD32 and 4-1BBL in the presence of aCD3 and aCD28 antibody (K562-BBL-3/28) following washing. Exogenous IL-2 was added to the cultures every other day at 100 IU/ml. GFP+ T cells were enumerated by flow cytometry using bead-based counting. Results are reported as the number of population doublings, and they are representative of four-independent experiments. (b) Sustained CAR+ T cell expansion in the absence of restimulation. CD4+ and CD8+ T cells were engineered with the indicated CAR, expanded and enumerated as in panel (a) In the absence of any K562 stimulator cells. Results are representative of at least three-independent experiments. (c) Histogram of mean T cell volume (fl) on day 8 of culture measured using a Coulter Multisizer III particle counter following stimulation with αCD3/αCD28 coated magnetic beads on day 0, and transduction with the indicated CAR on day 1. Results are representative of at least three-independent experiments.

Figure 5

CD19-specific CAR+ human T cells display significant antileukemic activity in an immunodeficient mouse model of human pre-B ALL. (a) Xenograft model using human CD19-specific CAR+ T cells to treat a primary human pre-B ALL in immunodeficient mice. After establishment of ALL, mice were randomized to treatment with the indicated T cell populations. (b and c) Hematoxylin & Eosin stained section of a NOD-SCID-β2−/− mouse calvarium (b) and spleen (c) 7–8 weeks following injection of pre-B ALL. Arrows indicate areas of bone marrow infiltrated by malignant ALL cells. The asterisk indicates leukemic cells that have invaded the leptomeningeal membranes. (d) Dose dependent antileukemic effect of CAR+ T cells given 2 weeks after establishing leukemia. Peripheral blood CD19+ B-ALL blast cell counts were measured at weekly intervals in mice (>4 mice/group) that were injected with the indicated numbers of αCD19-ζ CAR+ T cells or mock-transduced T cells The blast count in the 5 × 106 CAR+ T cell group is significantly lower than the count in the Mock and no T cell groups (ANOVA on the log-transformed blast counts, F-test P = 0.008) (e) Leukemia-free survival over time in animals described in panel D receiving no T cells, mock-transduced T cells, or αCD19 -ζ CAR+ T cells (5 × 106). Animals were assessed for leukemia at weekly intervals. Survival curves for the indicated groups were compared using the log-rank test. The group receiving αCD19 -ζ CAR+ T cells show a significantly increased median survival (log-rank test, P < 0.001) compared to animals receiving mock-transduced or no T cells. (f) Increased antileukemic effect of CAR+ T cells expressing costimulatory domains given 2 weeks after establishing leukemia. Peripheral blood CD19+ ALL blast cell counts were measured in leukemic mice (≥7 mice/group) that were injected with 2 × 106 CAR T cells or mock-transduced T cells. At week 5, animals in all CAR+ T cell groups expressing costimulatory domains have significantly lower blast counts than Mock and no T cell groups by ANOVA (using Scheffe, P < 0.01). Results represent mean and standard error of mean in d and f.

Figure 6

The 4-1BB costimulatory domain enhances CAR+ T cell survival and antileukemic efficacy in vivo. (a) Absolute peripheral blood CD4+ and CD8+ T cell counts 4 weeks following T cell injection in NOD-SCID-γ−/− mice. 8 × 106 T cells engineered to express the indicated CAR by a bicistronic lentiviral vector that encodes the CAR linked to eGFP were injected 3 weeks after injection of 2 × 106 leukemic cells. T cells were normalized to 45–50% input GFP+ T cells by mixing with mock-transduced cells prior to injection, and confirmed by flow cytometry (data not shown). Results represent the mean and SEM of the absolute number of cells per ml of whole blood (measured by TruCount assay) in at least 4 mice/group. (b) The mean and SEM of the % of CAR+GFP+CD4+ and CD8+ T cells in the same samples shown in a. The overall F-test in a one-way ANOVA comparing the mean across groups was significant (P < 0.01). The asterisk indicates results that are significantly different from the other two groups by a post hoc pairwise comparison of the means (Scheffe F-test at P = 0.05). (c) Enhanced antigen-independent survival of 4-1BB CARs. The mean and SEM of the % of CAR+GFP+ CD4+ and CD8+ T cells in animals injected with T cells is shown at the same time as in a, but in nonleukemic NOD-SCID-γ−/− mice. The overall F-test in a one-way ANOVA comparing the mean across groups was significant (P < 0.01). The asterisk indicates results that are significantly different from the other two groups by a post hoc pairwise comparison of means (Scheffe F-test at P = 0.05). (d) Leukemia-free survival over time in animals described in a and b. Animals were assessed for leukemia at 1-week intervals. Survival curves for the indicated CAR+ T cell groups were compared using the log-rank test. The αCD19-BB-ζ group shows a significantly increased median survival (log-rank test, P = 0.009) compared to either the αCD19-ζ or αCD19-28-ζ groups.

Figure 7

Long-term expression and survival of T cells engineered to express a CD19-specific CAR with the 4-1BB costimulatory domain in vivo. (a) Schematic diagram of a competitive experiment in which different numbers of αCD19-ζ and αCD19-BB-ζ engineered T cells are coinjected at a 1:1 ratio into NOD-SCID-γ−/− mice bearing B-ALL; see Supplementary Materials and Methods for detailed description of experimental design. (b) The number of copies of αCD19-ζ and αCD19-BB-ζ vector in spleen DNA from 27 mice evaluated at various times following T cell injection as described in a and Supplementary Table S2. The dotted line represents the expected copy number for each receptor based on the injected T cells containing an average of 2.4 copies of αCD19-ζ and 1.7 copies of αCD19-BB-ζ per cell. The mean of the ratio of αCD19-BB-ζ: αCD19-ζ following injection was compared to the ratio in injected T cells following log transformation, and was found to be significantly different (Student's t-test, P = 0.0001). The ratio of αCD19-BB-ζ: αCD19-ζ at baseline (0.708) was significantly different from the geometric mean ratio of 1.756 after engraftment with 95% confidence interval (1.178, 3.696). (c) Dose dependent CAR treatment response. Peripheral blood was obtained 35–70 days after establishing leukemia in mice injected on day 21 with either 1, 5 or 20 × 106 CAR T cells, an equivalent number of mock-transduced T cells, or no T cells. There were 16 mice per T cell group and four animals in the no T cell group; four mice from each group were randomly bled for determination of peripheral blood CD19+ ALL blast counts and then killed on days 35 and 49. The remainder of the animals were evaluated on days 57 and 70. The ALL blast counts were significantly lower on days 57 and 70 in the CAR mice compared to mice engrafted with mock-transduced T cells (P < 0.01 by Kruskal-Wallis test). (d) Surface expression of the CD19-specific CAR on T cells isolated from the spleen of a mouse in the mock-transduced T cell group and the CAR+ T cell group 198 days following injection of the T cells into NOD-SCID-γ−/− mice bearing B-ALL.