Generation of Potent T-cell Immunotherapy for Cancer Using DAP12-Based, Multichain, Chimeric Immunoreceptors

“First-generation” chimeric antigen receptors (CAR) incorporate a cytoplasmic domain containing one or more immunotyrosine-based activation motifs (ITAM) into a single chimeric receptor that uses a single-chain variable fragment (scFv) for antigen recognition (1). Cytoplasmic domains from receptors, such as CD28, ICOS, 4-1BB, and OX-40, incorporated into these receptors enhance the proliferation, survival, and function of T cells (2–5). T cells expressing “second-generation” (one costimulatory domain) and “third-generation” (two costimulatory domains) CARs exhibit enhanced function in preclinical animal models of cancer, and T cells bearing several costimulation-enhanced CARs are in human clinical trials for cancer (reviewed in ref. 6).

Although CARs based upon a single chimeric molecule design trigger antigen-specific T-cell responses, natural immunoreceptors are typically structured as multichain complexes composed of separate ligand-binding and ITAM-containing signaling chains, such as the T-cell receptor (TCR)–CD3 complex (7). The potential benefits of a multichain immunoreceptor complex are manifold, including greater diversity of signals available through the multiple interactions between ligand binding and signaling molecules, and sustained ITAM signaling that is separable from the internalization of the ligand-binding chain (8). The consequence of combining several receptor components normally found in heterologous molecules into a single CAR has not been fully elucidated; however, antigen-independent signaling and induction of anergy have been observed with existing CARs (5, 9, 10).

We hypothesized that a CAR constructed using a natural multichain immunoreceptor design would have greater activity in T cells due to the naturally selected interactions between subunits within the receptor complex and other receptors within immune cells. We chose the killer immunoglobulin-like receptor (KIR) and DAP12 multichain immunoreceptor complex as the foundation for a CAR (11). KIR expression is observed in both CD4⁺ and CD8⁺ T cells, in addition to natural killer (NK) cells, in which these receptors contribute to natural cytotoxicity (12–14). Activating KIRs, such as KIR2DS2, possess a short cytoplasmic domain with no known signaling activity. However, KIRs form a noncovalent complex with DAP12, an ITAM-containing molecule capable of binding Syk and Zap70 kinases, through transmembrane-mediated interactions (15). In addition to stimulating cytotoxicity upon ligand binding, KIRs have been reported to costimulate T cells in the absence of DAP12 expression, suggesting that these molecules may provide both primary triggering activity and costimulation in T cells (16).

We show that a KIR-based CAR (KIR-CAR) targeting tumor-associated antigens can potently activate T cells. Adoptively transferred T cells engineered to express these KIR-CARs and DAP12 can also induce regression of tumor xenografts, including a mesothelioma xenograft resistant to treatment with T cells bearing CD3ζ-based CARs with 4-1BB or CD28 costimulatory domains.

Mesothelin and CD19-specific CD3ζ-based CARs, SS1ζ, SS1-28ζ, SS1-BBζ, CD19ζ, and CD19-BBζ, used for all studies were those previously described (4, 5). The lentiviral vector expressing Dap12-T2A-SS1-KIRS2 was constructed using standard molecular biology techniques. First, Dap12 cDNA cloned from human peripheral blood mononuclear cells was used as a template for PCR amplification of the 351-bp complete coding sequence for Dap12 using the following primers: 5_-TCTAGAATGGGGGGACTTGAAC-_3 (XbaI/ is underlined), 5_-GTCGACTTTGTAATACGGCCTC_-3 (SalI/ is underlined). The resulting PCR product was cloned in-frame 3′ to a dsRed-Thoseasigna virus 2A (T2A) fusion sequence downstream of the EF-1α promoter in the previously described third-generation self-inactivating lentiviral vector (5) to generate pELNS Dap12-T2A-dsRed. The mesothelin scFv (SS1), previously described (4), was used as a template for PCR amplification of the 801-bp SS1 fragment using the following primers: 5_-CCTAGGATGGCCTTACCAGTG-_3 (AvrII/ is underlined), 5_-GCTAGCTTTGATTTCCAACTTTGTCC-_3 (NheI/ is underlined). The resulting PCR product containing the SS1 scFv coding sequence was ligated to a 270-bp PCR product from KIR2DS2 generated by PCR from cDNA using the following primers: 5_-GCTAGCGGTGGCGGAGGTTCTGGAGGTGGGGGTTCCTCACCCACTGAACCAAGC_-3 (NheI/ is underlined) and 5′_-GTCGACTTATGCGTATGACACC_-3 (SalI/ is underlined). The resulting chimeric SS1 scFv-KIR2DS2 fragment (termed SS1-KIRS2) was subsequently cloned in-frame 5′ to the Dap12-T2A sequence in pELNS Dap12-T2A-dsRed to generate pELNS Dap12-T2A-SS1-KIRS2. CD19-KIRS2/Dap12 and FAP-KIRS2/Dap12 vector inserts were made by exchanging the SS1 scFv with a CD19-specific scFv sequence derived from FMC63 previously described (5) and FAP-specific scFv previously described (17) at BamHI and NheI sites, respectively. High-titer replication-defective lentiviral vectors were produced and concentrated as previously described (5).

Primary human T (CD4 and CD8) cells were isolated from healthy volunteer donors following leukapheresis by negative selection using RosetteSep kits (Stem Cell Technologies). All specimens were collected under a University Institutional Review Board–approved protocol, and written informed consent was obtained from each donor. T cells were cultured in RPMI 1640 supplemented with 10% FCS, 100 U/mL penicillin, 100 g/mL streptomycin sulfate, and 10 mmol/L Hepes and stimulated with magnetic beads coated with anti-CD3/anti-CD28 at a 1:3 cell to bead ratio. Approximately 24 hours after activation, T cells were transduced with lentiviral vectors at a multiplicity of infection of 3 to 6. Cells were counted and fed every 2 days until they were either used for functional assays or cryopreserved after rest down.

Target cells, K562 (Kwt), K562.meso (Kmeso), EM parental (EMp), and EM-meso cells were stained for surface expression of mesothelin using the CAK1 antibody (clone K1; Covance) followed by phycoerythrin (PE)-labeled secondary goat-anti-mouse antibody. Expression of the various SS1 scFv fusion proteins on T cells was detected using either biotinylated goat anti-mouse F(ab)₂ (Jackson ImmunoResearch) followed by staining with streptavidin-PE (BD Biosciences), or with a mesothelin-V5-hisx12 fusion protein (kindly provided by Jennifer Brogdon, Novartis Institute of Biomedical Research) followed by staining with a V5 epitope–specific, FITC-conjugated antibody (Thermo Scientific). Samples were analyzed on either LSRII or FACSCalibur flow cytometers (BD Biosciences) and analyzed with FlowJo software (TreeStar).

Target cells were loaded with ⁵¹Cr and combined with differing amounts of transduced T cells in U-bottom plates. After a 4-hour incubation at 37°C, the release of free ⁵¹Cr was measured using a COBRA II–automated gamma-counter (Packard Instrument Company). The percent-specific lysis was calculated using the formula: % specific lysis = 100 × (experimental cpm release – spontaneous cpm release)/(total cpm release − spontaneous cpm release). All data are presented as a mean ± SD of triplicate wells.

Two-color immunohistochemical staining for human CD8 alpha (Clone C8/144B; Dako M7103; 1:100 dilution) and mesothelin (Clone 5B2; Thermo Scientific MS-1320; 1:30 dilution) was performed sequentially on a Leica Bond III using the Bond Polymer Refine Detection System and the Bond Polymer Refine Red Detection System. Heat-induced epitope retrieval was done for 20 minutes with ER2 solution (Leica Microsystems AR9640). Following dual-color immunohistochemistry, multispectral imaging was performed on the stained sections using a Vectra multispectral imaging system (Perkin Elmer), and the resulting multispectral images were analyzed using InForm Imaging software (Perkin Elmer). Ten random 20× fields were selected for analysis of each tumor. The images were segmented into tumor and stromal regions using mesothelin staining and morphologic features of the tumor cells from the hematoxylin counterstain using Informs “train by example” segmentation algorithm. CD8⁺ T-cell counting was then performed by counting CD8⁺ T cells within tumor and stromal regions.

Nalm6 leukemic cells transduced with Click Beetle Luciferase emitting in the green spectrum (CBG, λ_peak 550 nm, Chrom-Luc Green; Promega) were injected into adult NSG mice on day 0 as previously described (18). T cells with or without CARs were injected on day 5, and the tumor burden was assessed by imaging anesthetized mice using a Xenogen Spectrum system and Living Image v4.2 software following intraperitoneal injection of 150 mg/kg D-luciferin (Caliper Life Sciences). Each animal was imaged alone (for photon quantification) or in groups of up to five mice (for display purposes) in the anterior–posterior-prone position at the same relative time point after luciferin injection (6 minutes). Data were collected until the midrange of the linear scale was reached (600–60,000 counts) or maximal exposure settings reached (f-stop 1, large binning, and 120 seconds), and then converted to photons/sec/cm²/steradian to normalize each image for exposure time, f-stop, binning, and animal size.

Tumors were harvested at various time points and processed as previously described (19). In brief, tumors were harvested, microdissected into small fragments, and then digested for 1 hour at 37°C with a cocktail of collagenase type I (1 U/10 mL; Worthington), II (1 U/10 mL; Worthington), and IV (1 U/10 mL; Worthington), DNase I (0.5 U/10 mL; Worthington), and elastase (0.5 U/10 mL; Worthington), in L15-medium (10 mL per tumor; Leibovitz). Digested tumors were then filtered through 70-μm nylon mesh cell strainers, and red blood cells were lysed before analysis by flow cytometry (BD Pharm Lyse; BD Biosciences). For functional analysis and Western blotting experiments, further purification of the TILs was performed. Dead cells were first removed by labeling with anti-phosphotidylserine microbeads (#130-090-101; Miltenyi Biotec) to avoid nonspecific binding of antibody to dead cells during the tumor-infiltrating lymphocyte (TIL) enrichment process. The live single-cell suspension was then blocked with anti-mouse CD16/CD32 antibody for 15 minutes, followed by staining with PE-conjugated anti-human CD45 (3 μg/mL as the final concentration; BD Biosciences) for an additional 20 minutes. Human T cells were then targeted with tetrameric antibody complexes (EasySep PE selection kit, #18551; STEMCELL Technologies) recognizing PE and dextran-coated microbeads. Labeled cells were positively selected using an EasySep magnet. The purity of isolated TILs was then confirmed by flow cytometry, and >90% purity was achieved throughout the entire study. Cryopreserved T cells expressing CD3ζ-based and KIR-based CARs were processed using the same TIL isolation procedure described with no effect on CAR expression as assessed by either flow cytometry for surface expression (data not shown) or immunoblotting for CD3ζ (Supplementary Fig. S7).

TILs were cocultured with firefly luciferase-expressing tumor cells at different effector TIL to target (E:T) ratios for a specified period of time. At the end of the coculture incubation period, supernatant was saved for IFNγ concentration measurement by ELISA (Biolegend; #430106), wells were washed, and remaining adherent tumor cells were lysed with a 1× cell lysis buffer for 30 minutes. Luciferase activity in the lysates was analyzed using the Luciferase Assay System on a GloMax Multi Detection System (Promega). Results are reported as percent killing based on luciferase activity in wells with tumor, but no T cells. {% killing = 100 − [Relative Light Units (RLU) from well with effector and target cell coculture]/[RLU from well with target cells] × 100}. E:T ratios represent total T cells per target cell.

For immunoblot analysis, 1.5 million T cells or isolated TILs were lysed with RIPA buffer supplemented with a protease inhibitor cocktail (Roche). Whole cell lysates (20 μg per sample) were separated by SDS-PAGE, transferred to nitrocellulose membranes, and probed with antibodies specific for target proteins as indicated followed by goat-anti-rabbit or goat-anti-mouse secondary antibodies conjugated to alkaline phosphatase. Binding was detected by chemiluminescence using Westerj Lighting Plus-Enhanced Chemiluminescence Substrate (NEL104001EA; Perkin Elmer).

Statistical analyses were performed using STATA version 12 (STATAcorp). Unless indicated, a one-way ANOVA was performed in experiments involving multiple comparisons of groups with a threshold P value of <0.05 before performance of post hoc analysis by the Scheffe F-test.

A KIR-CAR was constructed by splicing a mesothelin-specific SS1 scFv antibody onto the transmembrane and short cytoplasmic domain of KIR2DS2 (SS1-KIRS2) as illustrated schematically in Fig. 1A. DAP12 is constitutively expressed in NK cells, but it is only expressed in a subset of human T cells (12). We therefore generated a bicistronic, lentiviral vector encoding SS1-KIRS2 and DAP12 separated by the Thoseasigna virus 2A (T2A) sequence in order to achieve coexpression of both molecules (Fig. 1B). Transduction of primary human T cells with this bicistronic lentivirus following anti-CD3 and anti-CD28 activation yields surface expression of SS1-KIRS2 that is comparable with the CD3ζ-based SS1ζ CAR (Fig. 1C). SS1-KIRS2/DAP12–transduced T cells expand following anti-CD3 and anti-CD28 agonist antibody stimulation with kinetics comparable with mock-transduced T cells (data not shown). SS1-KIRS2/DAP12 triggers potent in vitro T-cell cytotoxic activity toward K562 cells that express human mesothelin (Kmeso) with similar magnitude to SS1ζ. None of the engineered T cells demonstrate lytic activity toward wild-type K562 (Kwt), supporting the specific activation of the SS1-KIRS2/Dap12 receptor by the cognate mesothelin target antigen (Fig. 1D). Because T cells express KIR2DS2 in the absence of detectable DAP12 expression, we evaluated SS1-KIRS2 receptor expression and function with and without codelivery of DAP12. As expected, SS1-KIRS2 is expressed on the surface of T cells without the addition of DAP12; however, T cells expressing SS1-KIRS2 without cotransduction of DAP12 fail to lyse mesothelin-expressing target cells demonstrating a requirement for codelivery of DAP12 for SS1-KIRS2–triggered T-cell cytotoxic activity (Supplementary Fig. S1). These data do not preclude the possibility that a chimeric KIR provides signals independently of its association with DAP12 as previously reported for the natural KIR2DS2 receptor in T-cell clones (16).

The noncovalent association of KIR2DS2 and DAP12 depends upon the electrostatic interactions between an aspartic acid residue in the KIR transmembrane (TM) domain and a lysine residue in the DAP12 TM domain. Although the position of the ionizable residues within the hydrophobic TM domains of multichain immunoreceptors appears to direct the specificity of these receptor interactions, we explored the possibility that SS1-KIRS2 expression in the absence of DAP12 codelivery might be due to interactions with components of the endogenous CD3 complex expressed by T cells, which shares a similar mechanism of TM interaction for complex assembly and surface expression (20). The introduction of an ectopic Vβ chain into T cells has been shown to reduce the surface expression of endogenous TCR Vβ due to competition during complex assembly (21, 22). We observed a similar frequency and intensity of TCR Vβ 13.1+ in SS1-KIRS2–transduced polyclonal T cells compared with mock-transduced, control T cells, supporting the absence of a significant interaction between SS1-KIRS2 and members of the endogenous CD3 complex (Supplementary Fig. S2).

Cytokine secretion and T-cell proliferation are important characteristics that are generally dependent upon costimulation and correlate with robust antitumor activity in vivo. We compared the antigen-triggered secretion of IFNγ and IL2 by SS1-KIRS2/DAP12–modified T cells with T cells bearing a CD3ζ-based CAR without a costimulatory domain (SS1-ζ) or with CD28 or 4-1BB costimulatory domains (SS1-28ζ and SS1-BBζ, respectively; ref. 4). As expected, SS1-ζ stimulates the lowest secretion of IFNγ and IL2 (Fig. 2A and B). IFNγ production is increased and comparable in T cells expressing SS1-KIRS2/DAP12 or SS1-BBζ, whereas T cells expressing the SS1-28ζ CAR show significantly greater IL2 and IFNγ production (Fig. 2A). Analysis of a larger panel of cytokines and chemokines demonstrates that SS1-KIRS2/DAP12 stimulates a pattern of expression that is qualitatively similar across CD3ζ-based CARs with a magnitude of antigen-induced cytokine and chemokine secretion that is comparable with SS1-ζ and SS1-BBζ CARs (Fig. 2B).

The SS1-KIRS2/DAP12 receptor is also a potent stimulator of T-cell proliferation in response to cognate antigen (Fig. 2C). Stimulation of SS1-ζ CART cells with mesothelin-expressing cells resulted in minimal T-cell proliferation as previously demonstrated with first-generation CD3ζ-based CARs (5, 23–26). Proliferation could be enhanced in the presence of soluble anti-CD28 agonist antibody. In contrast, SS1-KIRS2/DAP12 induced proliferation in response to mesothelin antigen that was comparable with SS1ζ with addition of anti-CD28 agonist antibody with no evidence of increased proliferation with the addition of anti-CD28 agonist.

CD28 and 4-1BB have been incorporated into CARs to enhance CAR T-cell activity in vivo (4); however, costimulation is not always able to overcome the immunosuppressive tumor microenvironment (TME). We recently reported that SS1-BBζ CAR T cells injected into immunodeficient NOD-SCID-γ_c^−/− (NSG) mice bearing a xenograft of EM-meso cells (a cell line derived from the pleural effusion of a patient with malignant mesothelioma) expand in vivo, but become hypofunctional within the TME associated with failure to clear tumors (19). We evaluated the activity of SS1-KIRS2/DAP12–modified T cells in this T-cell immunotherapy–resistant model of mesothelioma. SS1-KIRS2/DAP12– and CD3ζ-based CARs with or without costimulation lyse EM-meso target cells in vitro with comparable efficacy. Mock-transduced and DAP12-dsRed–transduced T cells show minimal lytic activity toward EM-meso cells (Supplementary Fig. S3). A single intravenous injection of 5 million mock, SS1ζ-, or SS1BBζ-transduced T cells had no observable antitumor effect on established EM-meso xenografts (Fig. 3A). Tumor growth was slightly, but statistically significantly, delayed by T cells expressing SS1-28ζ; however, only SS1-KIRS2/DAP12–modified T cells induced regression of tumors with significant suppression of EM-meso tumor growth at 52 days (P < 0.001; ANOVA with post hoc Scheffe F-test). A second experiment comparing T cells expressing SS1-KIRS2/DAP12 to DAP12 alone or SS1-28ζ engineered T cells showed similar enhanced antitumor activity of SS1-KIRS2/DAP12 T cells (Supplementary Fig. S4).

To determine if the potent antitumor activity of T cells bearing a KIR-CAR/Dap12 is unique to the mesothelin specificity, we generated additional KIR-based CARs targeting CD19 and fibroblast activation protein (FAP). T cells expressing a CD19-specific KIRS2-based CAR with DAP12 (CD19-KIRS2/DAP12) showed in vitro cytotoxic activity that was comparable with previously described CD3ζ-based CARs (Supplementary Fig. S5; ref. 5). In a NALM-6 B-cell leukemia xenograft model, T cells expressing CD19-KIRS2/DAP12 achieve superior control of leukemia when compared with a first-generation CD19-specific CAR bearing only a CD3ζ cytoplasmic domain (19ζ). CD19-KIRS2/DAP12–modified T cells exhibited control of leukemia that was comparable with a second-generation CD19-specific CAR with the 4-1BB costimulatory domain (CD19-BBζ) that has been shown to have potent antileukemic activity in humans (Fig. 3B; refs. 27, 28). The similar in vivo activity of CD19-KIRS2/DAP12 CAR T cells and CD19-BBζ CAR T cells in this model confirms the enhanced activity of KIR-CARs/Dap12 without the incorporation of additional costimulatory signals.

Unlike mesothelin and CD19, FAP, while expressed on the surface of some tumors such as osteosarcoma (29), is expressed within a range of tumors by stromal cells in the TME, including tumor-associated fibroblasts and tumor-associated macrophages (30, 31). These stromal cells support tumor growth and limit immunologic reactions against the tumor (32). Targeting FAPs has therefore been proposed as an adjunct to tumor-targeted therapy to deplete these FAP⁺ stromal cells (17, 33–36). We generated a murine FAP-specific KIR-based CAR (FAP-KIRS2) using the scFv from the 73.3 hybridoma that was previously described by our group in a CD3ζ-based FAP-specific CAR bearing the 4-1BB costimulatory domain (FAP-CAR; ref. 17). T cells expressing FAP-KIRS2 and DAP12 (FAP-KIRS2/DAP12) show antigen-specific cytotoxicity in vitro that is comparable with T cells expressing CD3ζ-based CARs (data not shown). As we have seen in a mouse model of human lung cancer (17), when we have treated established human EM-meso tumors growing in NSG mice with the CD3ζ-based FAP-CAR construct, we observed a significant, but modest, slowing of tumor growth and no toxicity (Fig. 3C). However, when the same tumors were treated with the FAP-KIRS2/DAP12–modified T cells, a much more marked antitumor effect was observed, with tumor growth completely halted (Fig. 3D). Interestingly, after using these effective FAP-KIRS2/DAP12–modified T cells, we now observed anemia (Fig. 3E) along with hypocellularity of the bone marrow and weight loss (Supplementary Fig. S6). This is similar to toxicity reported in a genetic model of complete FAP ablation (37). These effects can be explained by a more complete depletion of the intratumoral CD45⁻/CD90⁺/FAP⁺ and CD45⁺/FAP⁺ cells within the treated tumors (Fig. 3F) than observed in our previous studies with CD3ζ-based FAP-specific CARs (17).

Overall, these data show that the KIR-based CAR platform is a potent CAR design for generating tumor-directed T-cell immunotherapy with significantly enhanced antitumor activity over second-generation CD3ζ-based CARs in some settings. Understanding the mechanism for this enhanced activity is therefore critical to understanding this new CAR design and perhaps improving existing CD3ζ-based CAR designs.

To explore the mechanism of the enhanced antitumor activity of the mesothelin-specific SS1-KIRS2/DAP12 T cells compared with CD3ζ-based mesothelin-specific CARs in the EM-meso xenograft model, an analysis of T-cell engraftment and TILs was performed. Few hCD45⁺ TILs were detected in mock or SS1ζ-treated mice from the experiment shown in Fig. 3A. In contrast, tumors treated with SS1-KIRS2/DAP12, SS1-28ζ, and SS1-41BBζ CAR T cells had hCD45⁺ TILs that comprised 2% to 4% of the total viable cells within the tumor with comparable frequencies for each group (Fig. 4A). Immunohistochemical staining showed abundant CD8⁺ (Fig. 4B) and CD4⁺ TILs (data not shown) within the tumors of SS1-KIRS2/DAP12, SS1-28ζ, and SS1-41BBζ CAR T-cell–treated mice confirming the flow cytometric analysis. The increased efficacy of the SS1-KIRS2/DAP12 T cells is therefore not due to increased TIL frequency within the tumors at late stages of tumor growth. Only mice receiving the SS1-BBζ CAR T cells had detectable human CD45⁺ (hCD45⁺) cells in the blood and spleen (Fig. 4C) consistent with the previously observed effect of the 4-1BB costimulatory domain on in vivo CAR⁺ T-cell persistence (5).

Because the comparison of TILs was limited by the large differences in tumor volume at the late time points in our early experiments, we sought to examine TILs isolated at earlier time points following T-cell injection. Evaluation of TILs at 10 days following T-cell injection demonstrated comparable frequencies of hCD45⁺ TILs within tumors treated with SS1-28ζ and SS1-KIRS2/DAP12 CAR T cells; however, few CD45⁺ TILs were present in the SS1-BBζ CAR T-cell group (Fig. 5A). Limited analysis of the isolated TILs showed that only SS1-KIRS2/DAP12–modified T cells were capable of in vitro lytic activity toward EM-meso cells (data not shown). These results suggest that delayed accumulation of SS1-BBζ T cells into the tumor along with tumor-induced hypofunction contributes to the poor antitumor activity of these cells despite their high frequency at later stages of tumor development. A repeat experiment was conducted comparing SS1-28ζ and SS1-KIRS2/DAP12 T cells with TIL isolation at 18 days following T-cell injection to obtain greater numbers of TILs for phenotypic and functional analysis (Fig. 5B). Isolated SS1-28ζ TILs demonstrated markedly reduced cytotoxic activity and antigen-specific IFNγ production compared with cryopreserved T cells used for treatment. In contrast, TILs from SS1-KIRS2/DAP12–modified T-cell–treated tumors showed comparable in vitro cytotoxicity and IFNγ production to cryopreserved cells (Fig. 5C and D, respectively).

The potential mechanisms by which T cells might lose function within a TME are many; however, loss of CAR expression is one potential explanation that could be readily examined. Immunoblotting of protein lysates from TILs and cryopreserved cells for CAR protein demonstrated loss of SS1-28ζ expression in the recovered TILs (Fig. 5E). In contrast, DAP12 expression in SS1-KIRS2/DAP12 TILs was preserved correlating with the retained function for these isolated TILs (Fig. 5F). In a separate experiment, we confirmed the similar loss of SS1-BBζ CAR expression by TILs, and further demonstrate that the loss of SS1-BBζ CAR expression is reversible with ex vivo culture of TILs for 24 hours (Fig. 5G). These data indicate that CAR loss is an important, and previously unrecognized, mechanism of acquired hypofunction for CD3ζ-based CAR T cells. The retention of CAR and DAP12 expression by TILs explains, at least in part, the enhanced antitumor activity observed with SS1-KIRS2/DAP12–modified T cells.

We have shown that a “KIR-CAR” can be simply constructed by swapping the two immunoglobulin-like domains of the KIR2DS2 ectodomain with an scFv capable of binding a desired target antigen. When delivered to T cells together with DAP12, this KIR-based CAR triggers antigen-specific cytotoxicity, cytokine production, and proliferation that is comparable with second-generation CD3ζ-based CARs in vitro without the need for additional domains from costimulatory receptors. The ability of a KIR-based CAR to activate T cells in the absence of added costimulation is interesting in light of the critical importance of costimulation for full T-cell activation and acquisition of effector function. KIR2DS2, the natural KIR upon which the presented KIR-CAR is based, has previously been reported to deliver a costimulation-like signal to T cells. In these studies, engagement of the KIR in T-cell clones lacking DAP12 expression augmented anti–CD3-induced IFNγ production (16). The mechanism of this costimulatory response has not been elucidated, but might be related to interactions between the KIR or DAP12 and other, yet to be identified, adaptor molecules or costimulatory receptors. In particular, integrins are well recognized to provide costimulatory signals to T cells (38, 39). In cellular contexts, such as macrophages and neutrophils, DAP12 appears critical to outside-in signaling by integrin receptors (40). The expression of DAP12 in T cells might confer unique signaling activity to LFA-1 and other integrins expressed by T cells, and these effects could underlie some of the enhanced functional activity of T cells expressing SS1-KIRS2/DAP12. Cytokines also deliver important secondary signals to T cells that can augment T-cell activation and differentiation. Bezbradica and colleagues found that Jak2 signaling downstream of the IFNγ receptor is directly coupled to signaling by FcγRI, a related ITAM-based signaling complex, and these interactions regulate the signaling of each pathway (41). Similar integration between DAP12 and the IFNαR signaling has been observed (42). In addition to interactions with the cytokine c-FMS in osteoclast development, DAP12 signaling also interacts with TRAF signaling downstream of the type II TNF receptor, RANKL (43, 44). In aggregate, these studies increasingly demonstrate that ITAM-based adaptor proteins like DAP12 have the potential to interact with diverse, heterologous proteins and modify signaling pathways that are likely to be important in T-cell differentiation and function. It is unknown what role, if any, these processes play in the KIR-CAR system, but exploring these potential interactions will be an important area of future study.

One of the most important findings in this study is that, although T cells engineered to express the mesothelin-specific CD3ζ-based CARs and KIR-CAR/Dap12 showed similar functional activity in vitro, KIR-CAR/Dap12–modified T cells were much more effective in the in vivo setting (see Fig. 3A). Although interaction with heterologous signaling pathways as discussed above might contribute to the enhanced in vivo activity of KIR-CAR/Dap12–modified T cells, we have demonstrated that SS1-KIRS2/DAP12 TILs, unlike CD3ζ-based CARs, retain receptor expression in the TME that correlates with the maintenance of TIL effector function. Enhanced stability of the KIR/DAP12 complex within the plasma membrane following antigen engagement is one potential mechanism to explain the observed difference in CAR expression within TILs. The loss of endogenous CD3ζ expression in TILs has been long recognized and is associated with loss of T-cell function within the TME (45). TCR/CD3 complex internalization and degradation following antigen engagement are also well-described phenomena in T cells (46, 47). Although the precise mechanism for the loss of CD3ζ expression by TILs has not been fully elucidated, TCR/CD3 receptor complex internalization and degradation following antigen activation combined with metabolic stress within the TME that limits new synthesis of new complex components may contribute to the loss of CD3ζ by TILs (48). The surface expression and recycling of KIRs and DAP12 are less well studied. As we have confirmed with our KIR-based CAR, KIRs do not require assembly into a multichain immunoreceptor complex with DAP12 for surface expression, which is distinctly different from TCR that critically depends upon CD3ζ association for surface TCR expression (16, 49, 50). Mulrooney and colleagues made the additional, intriguing observation while studying natural KIRs that internalization and loss of surface KIR expression following receptor crosslinking are reduced in the presence of DAP12 (49). Reduced KIR-CAR internalization and degradation following antigen engagement would therefore provide a potential explanation for the observed difference in receptor expression in freshly isolated TILs. In addition to providing a mechanism for TILs to retain antigen recognition and effector function in the context of continuous ligand exposure, the duration of T-cell engagement with antigen-presenting cells also correlates with T-cell survival and effector function (51). Enhanced surface stability of a KIR-CAR might therefore contribute to the costimulation-independent stimulatory activity observed with the KIR-based CAR system by permitting more sustained interactions with antigen-positive target cells and longer duration of receptor signaling.

Eshhar and colleagues first described the single-molecule CAR design in 1993 (52). This design has shown great utility in adoptive T-cell immunotherapy; however, the vast majority of research on CARs to date has focused upon the modification of this single-chain receptor design, with introduction of domains from other costimulatory receptors, like CD28 and 4-1BB, to enhance function. The results presented here show that alternative CAR designs, especially multichain designs incorporating alternative ITAM-containing adaptor proteins beyond CD3ζ, are worthy of future exploration, especially in solid malignancies in which T cells with CD3ζ-based CARs have shown limited clinical activity to date.

E. Puré, S. Albelda, and M.C. Milone report receiving a commercial research grant from Novartis. No potential conflicts of interest were disclosed by the other authors.

Plasmids and lentiviral vectors encoding the KIR-based CARs as well as some of the cell lines described in this article are available to interested investigators, but their availability will depend upon the execution of a material transfer agreement with the University of Pennsylvania.

Conception and design: E. Wang, E. Moon, D. Barrett, E. Puré, S. Albelda, M.C. Milone

Development of methodology: E. Wang, L.-C. Wang, V. Bhoj, E. Moon, M.C. Milone

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Wang, L.-C. Wang, V. Bhoj, Z. Gershenson, E. Moon, K. Newick, A. Lo, T. Baradet, M.D. Feldman, S. Albelda

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Wang, L.-C. Wang, V. Bhoj, Z. Gershenson, E. Moon, A. Lo, M.D. Feldman, M.C. Milone

Writing, review, and/or revision of the manuscript: E. Wang, L.-C. Wang, V. Bhoj, E. Moon, M.D. Feldman, E. Puré, S. Albelda, M.C. Milone

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Wang, C.-Y. Tsai, J. Sun, T. Baradet, M.C. Milone

Study supervision: E. Wang, S. Albelda, M.C. Milone

The authors thank Carl June at the University of Pennsylvania for his helpful review of the article.

The work presented was funded through a grant from the NIH Common Fund (1PN1EY016586) with additional support from Novartis.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.