CD8⁺ T Cells Expressing an HLA-DR1 Chimeric Antigen Receptor Target Autoimmune CD4⁺ T Cells in an Antigen-Specific Manner and Inhibit the Development of Autoimmune Arthritis

Autoimmune diseases are serious health problems, affecting more than 4% of the population (1). Therapeutic modalities for autoimmune diseases have been largely based on nonspecific immunosuppression that can lead to significant long-term side effects. Whereas the biologic drugs currently used for autoimmune therapy are an improvement, immunosuppressive effects are still a health concern for many patients (2). Although understanding the origins of autoimmune pathogenesis remains a complex problem, it is clear that most are associated with the expression of specific HLA class II alleles (3). Implicit in this association is a role for CD4⁺ T cells, and the mechanism by which MHC class II molecules stimulate CD4⁺ T cells is well defined. Class II molecules bind peptides, and the resulting MHC:peptide complexes serve as ligands for the TCR expressed by CD4⁺ T cells. Peptide binding is dependent on the interaction of its amino acid side chains, with five pockets located in the cleft of the class II binding groove (4), and these pockets are the primary locations of the class II allelic polymorphisms. Thus, allele-specific binding of peptides by disease-associated HLA class II is likely a basis for the susceptibility to autoimmunity. One example is rheumatoid arthritis (RA), in which susceptibility is associated with select DR alleles, including DR1 (DRB1*01:01) and DR4 (DRB1*04:01, *04:04, and*04:05) (5), which share a stretch of amino acids at positions 70 to 74 in the DRB1 chains (6), termed the shared epitope. This observation suggests that, collectively, these RA-associated alleles bind and present the same autoantigenic peptide(s) to CD4⁺ T cells that promote the autoimmune disease.

Targeting the pathogenic CD4⁺ T cells that recognize the autoantigens presented by RA-associated HLA-DR alleles would be a highly effective means of treating the autoimmune disease and would avoid the creation of a state of general immunosuppression that is commonly associated with current therapies for RA. The difficulty in designing Ag-specific modalities for treating autoimmune diseases is determining which Ag(s) is driving the autoimmune T cell response and how to use this information to target the CD4⁺ T cells that are mediating the pathogenesis of the disease. Whereas advances have been made at identifying autoantigens for a number of autoimmune diseases, a means of therapeutically targeting the pathogenic CD4⁺ T cells that are stimulated by these autoantigens remains elusive.

Recent advancements in cancer immunotherapy have demonstrated an effective means of targeting and eliminating cancer cells in an Ag-specific manner. This immunotherapy is based on the use of CD8⁺ chimeric Ag receptor (CAR) T cells, in which the CAR molecules are genetically engineered proteins. In most cases, the CAR extracellular domains are composed of Ab single-chain variable fragments (scFv) that provide the specificity for an Ag expressed by the cancer cell, and the CAR intracellular domains are derived from TCR signaling proteins (e.g., CD28 and CD3ζ) that link the CAR with the signaling pathway used by endogenous TCR (7). cDNA constructs encoding these anti-cancer CAR are inserted into CD8⁺ T cells using replication-deficient viral vectors that have been extensively developed for gene therapy (8). Essentially, these engineered CD8⁺ T cells are retargeted CTL that use the specificity of the scFv CAR to identify the target cell and the endogenous lytic mechanism of the CTL. Once the scFv CAR binds to the target cell, the CTL is activated via the CAR intracellular activation and signaling domains, and the CTL is stimulated and lyses the cancer cell. Several of these scFv CAR T cell therapies for cancer are now FDA approved, and many more are in clinical trials.

The challenge to applying the scFv CAR approach to the treatment of autoimmune diseases is defining a cellular target that is specific for the pathogenic CD4⁺ T cells. To circumvent this issue, we have developed a novel CAR therapeutic approach for treating autoimmunity that targets only the pathogenic CD4⁺ T cells that are mediating the pathology of the autoimmune disease and targets them in a manner that is dependent on the Ag specificity of their TCR. To achieve this, we designed CAR molecules based on HLA-DRB1*01:01 (DR1), which incorporates a model autoantigen as part of its molecular structure, and linked both the HLA-DRB1 and DRA1 chains to CD28 and CD3ζ activation and signaling domains in the transmembrane and intracellular portion of each DR1 chain. Upon transduction, the resulting DR1 CAR T cells lyse CD4⁺ T cells in an Ag-specific manner (i.e., only CD4⁺ T cells expressing a TCR restricted to DR1 and specific for the antigenic peptide in the DR1 CAR are lysed). Using our HLA-DRB1*01:01 humanized mouse model of RA, in which autoimmunity is induced by immunization with type II collagen (CII), our studies demonstrate that DR1-CII CAR T cells that include the immunodominant peptide from the autoantigen CII identify and lyse CII-specific CD4⁺ T cells in vivo and effectively inhibit the autoimmune T cell response. In addition, following autoimmune disease induction, the treatment of the B6.DR1 mice with DR1-CII CAR CD8⁺ T cells significantly reduced both the B cell autoantibody response and inhibited the severity and incidence of the autoimmune arthritis. In all, these data demonstrate the potential use of MHC class II–based CAR T cells in treating autoimmune diseases in an Ag-specific manner.

The generation of DRB1*0101 transgenic mice (DR1) has been previously described (9). The DRA1 and DRB1 transgenes were established in (C57BL/6 × SJL/J) F2 mice backcrossed to the B6 background, and the I-A^b molecule expressed by the B6 mice was genetically deleted by backcrossing with the B6.129-H2^dlAb1-Ea knockout strain (stock number 003584; The Jackson Laboratory). All mice used in these studies were bred in our facility, maintained in microisolators in a pathogen-free environment, and were fed standard rodent chow (Ralston Purina) and water ad libitum. All animal studies were approved by the Institutional Animal Care and Use Committee at the VA Medical Center, Memphis.

The functional design of the DR1 CAR molecules is shown in Fig. 1A. cDNA encoding the HLA-DR1*01:01 CAR was cloned into the MSGV1-28z mut1-3 (MSGV) retroviral expression vector (gift of Dr. James Kochenderfer, National Institutes of Health) (8) using In-Fusion Cloning (Takara Bio). The DR1B and DR1A chains were separated by a ribosome-skipping T2A sequence, allowing both chains to be expressed off the same promoter in a single mRNA transcript that is translated as two separate peptide chains (Fig. 1B) (10). The endogenous DR1 transmembrane and cytoplasmic (TM/Cyto) domains were removed and replaced with the corresponding domains from CD28 and CD3ζ encoded in MSGV (Fig. 1). The DR1 cDNA used is also chimeric for mouse class II by replacement of the second domains of DRA1 and DRB1, with corresponding domains derived from mouse I-E to enable interaction of the DR1 molecule with murine CD4 (9). To load the DR1 molecule with a covalently linked antigenic peptide (Fig. 1), cDNA encoding the peptide sequences for the CII peptide (GIAGFKGEQGPKGEP; pMSGV1-DR1-CII-28z mut1-3) or the control hemagglutinin (HA) peptide (PKYVKQNTLKLAT; pMSGV1-DR1-HA-28z mut1-3) were added to the N terminus of DRB1 chain as previously described (11).

Replication-deficient retrovirus encoding the DR1 CAR molecule was produced using 293T cells cotransfected with pEQ-pam 3(-E), pCAG-Eco env (gifts from Dr. Stephen Gottschalk, St Jude Children’s Research Hospital), and pMSGV-DR1. 293T cells grown in DMEM plus 10% heat-inactivated FBS and 2 mM l-glutamine were transfected with 10 µg of total DNA in a 15:15:10 ratio (pEQ-pam 3(-E)/pMSGV-DR1/pCAG-Eco env, respectively) using GeneJuice Transfection Reagent (Novagen), following the manufacturer’s protocol. Viral supernatants were collected 48 and 72 h after transfection, filtered with a 0.45-µm syringe filter and snap frozen for future use. Seventy-two hours after transfection, 293T cells were stained with an anti-DR Ab (LB3.1) and analyzed by flow cytometry to confirm expression of DR1.

DR1 CAR T cells were produced by transduction of mouse CD8⁺ T cells using RetroNectin-coated plates (Takara Bio) as follows. CD8⁺ T cells were purified from the spleen and lymph nodes of B6.DR1 mice using anti-CD8a magnetic beads and an LS Column (Miltenyi Biotec), according to manufacturer’s protocol. After purification, the CD8⁺ T cells were stimulated overnight by incubation in 24-well plates coated with anti-CD3 and anti-CD28 (1 μg/ml) in RPMI 1640 complete media containing 10% heat-inactivated FBS (HyClone), 2 mM l-glutamine (Life Technologies), 100 µg/ml streptomycin (Life Technologies), 100 U/ml penicillin (Life Technologies), 50 µM 2-ME (Life Technologies), 1 mM Na-pyruvate (Life Technologies), 1× nonessential amino acids (Life Technologies), 1 mM HEPES (Life Technologies), and 20 U of IL-2/ml (PeproTech). On the following day, the CD8⁺ cells were washed and resuspended at 1 × 10⁶ cells/ml in RPMI 1640 complete media containing 5 µM 2-ME and 40 U of IL-2/ml.

Prior to transduction, wells of 24-well plates (Falcon, nontreated tissue culture plate) were coated overnight with RetroNectin (0.5 ml/well of 10 μg/ml) at 4°C. The RetroNectin-coated wells were then blocked with 0.5 ml of 2% BSA in sterile PBS for 30 min at room temperature and washed once with PBS. A total of 0.5 ml of virus supernatant was then added to each well, and the plate was centrifuged for 2 h at 2000 × g, 32°C. After centrifugation, 1 ml of the stimulated, washed CD8⁺ T cells was centrifuged onto the virus RetroNectin-coated wells for 2 h at 1000 × g, 32°C. After centrifugation, the plates containing the cells were cultured overnight at 37°C.

Three days after transduction, DR1 expression by the transduced CAR T cells was analyzed by flow cytometry using a fluorescent-labeled anti-DR Ab (LB3.1) to determine transduction efficiency. For maintenance in culture, the CAR T cells were seeded at 1 × 10⁶ cells/ml in complete RPMI 1640 with 20 U/ml of IL-2, and the cell concentration was adjusted daily. The cells were restimulated weekly on anti-CD3 and anti-CD28 Ab-coated (1 μg/ml) plates, and in some cases, the first restimulation was done with anti-DR–coated (LB3.1, 2 µg/ml) plates. For some experiments, purified, nontransduced (NT) CD8⁺ T cells were used as a control, and these cells were maintained in culture using the same anti-CD3/CD28 stimulation protocol.

For arthritis induction and T cell proliferation assays, mice were immunized s.c. at the base of the tail with 100 µg of peptide or bovine CII emulsified in equal volumes of CFA consisting of 85% heavy paraffin oil (Thermo Fisher Scientific), 15% mannide monooleate (Sigma-Aldrich), and 4 mg/ml heat-killed Mycobacterium (H37Ra; Difco) (12). For arthritis studies, mice were examined three times per week starting at day 18 after immunization, and the presence of arthritis, number of affected limbs, and the severity were assessed. Severity of disease was evaluated by visual inspection and assigned a score using a scale of 0 to 4, based on the degree of inflammation, as previously described (12).

Eleven or twelve days after immunization, T cells were recovered from draining lymph nodes of CII-immunized mice. T cell proliferation assays were performed in 96-well microtiter plates in a total volume of 300 μl containing 4 × 10⁵ lymph node cells and various concentrations of peptide in complete HL-1 medium. Cell cultures were maintained at 37°C in 5% humidified CO₂ for 4 d. On day 3 of culture, 1 µCi of [³H]thymidine was added, and on day 4, the plates were harvested onto filter plates. After the filter plates were dried, scintillation fluid was added, and [³H]thymidine incorporation was measured using a Hidex Sense microplate reader (LabLogic).

T cell hybridomas were established by polyethylene glycol fusion (Boehringer Mannheim) of lymph node cells with BW5147 thymoma cells (TCRα/β negative) as previously described (13). Lymph node cells were obtained from DR1 (DRB1*0101) transgenic mice (9) immunized 10 d previously with Ag/CFA. Prior to fusion, lymph node T cells were stimulated with antigenic peptide for 4 d, followed by IL-2 stimulation for 3 d. The resulting hybridomas were screened for their ability to recognize the antigenic peptide presented by DR1.

Ag presentation assays were performed in 96-well microtiter plates in a total volume of 0.3 ml containing 5 × 10⁴ APCs, 5 × 10⁴ T hybridoma cells, and 10 µg of synthetic peptide (RS Synthesis). Assays were performed in HL-1 medium (BioWhittaker) supplemented to 2 mM l-glutamine (Life Technologies), 50 U/ml penicillin, 50 µg/ml streptomycin (Life Technologies), and 50 µM 2-ME (Life Technologies). Assay cultures were incubated at 37°C in 5% humidified CO₂ for 20 to 24 h. After this time, culture supernatants were harvested, and IL-2 production by the T cell hybridomas was measured in a bioassay using the IL-2–addicted cell line HT-2 (9). HT-2 cell viability was assessed by cleavage of MTT (Sigma-Aldrich) (14, 15). IL-2 titers were quantified by the reciprocal of the highest 2-fold serial dilution maintaining HT-2 cell viability greater than 2-fold over negative control cultures. Results are presented as units of IL-2/ml, as described by Kappler et al. (16).

CII-specific Ab levels in plasma of mice immunized with CII/CFA were quantitated using a bead-based ELISA and a MAGPIX instrument (Bio-Rad Laboratories). The CII beads for the ELISA were produced by incubation of purified bovine CII with Bio-Plex Pro magnetic COOH beads (Bio-Rad Laboratories). The magnetic beads were activated with EDAC/S-NHS and the collagen was coupled using an Amine Coupling Kit (Bio-Rad Laboratories) according to manufacturer’s protocol. For the ELISA, the CII-coupled beads were incubated with various dilutions of mouse plasma starting at 1:4000, and the quantity of mouse IgG bound was detected by the addition of a biotinylated anti-mouse IgG (Sigma-Aldrich) followed by PE-streptavidin (Invitrogen). The quantity of anti-CII Ab was calculated using a curve generated using purified CII-specific Ab as the standard.

Cytokines produced by CAR T cells after stimulation with anti-DR or anti-CD3/CD28 Abs were measured using a multiplexed bead assay (Bio-Rad Laboratories) and analyzed using a Bio-Rad MAGPIX. A total of 1 × 10⁶ CAR T cells were cultured in a 48-well plate in which the wells had been coated with 250 µl of an anti-DR Ab (LB 3.1, 2 µg/ml) or anti-CD3 and anti-CD28 Abs (1 µg/ml each) or PBS. After 18 h of stimulation, 500 μl of supernatant was collected and frozen at 20°C prior to the cytokine ELISA. Supernatants were assayed for expression of IFN-γ, TNF-α, IL-10, IL-17A, and IL-6. Standard curves were used to calculate cytokine concentration, and data are based on duplicate samples and representative of two independent experiments.

A flow cytometry–based assay was used to measure the cytolytic activity of the CAR T cells. A total of 1 × 10⁵ DR1-CII or DR1-HA CAR T cells were cocultured in 300 µl RPMI medium in a 96-well microtiter plate with 1 × 10⁴ cloned DR1-restricted T cells for a 10:1 E:T ratio. Targets consisted of T cell hybridomas specific for either the CII peptide (GIAGFKGEQGPKGEP) or the HA peptide (PKYVKQNTLKLAT). After 4 h at 37°C, cells were collected from the wells and stained with anti-CD8- allophycocyanin-Cy7 (clone 53-6.7, BD Biosciences), anti-BV14-FITC (clone 14.2, BD Biosciences), anti-CD4-PerCP-Cy5.5 (clone RM4-5, Tonbo Biosciences), anti-BV8-PE (clone F23.1, BD Biosciences), anti-HLA-DR-Pacific Orange (clone LB 3.1), and DAPI (Sigma-Aldrich) was added just prior to analysis for exclusion of dead cells. Data analyses were focused on live cells (DAPI-negative), HLA-DR1–negative, CD8-negative, and CD4⁺/BV14⁺ (HA-specific) or CD4⁺/BV8⁺ (CII-specific) T cells.

A chimeric receptor for expression in CD8⁺ T cells was designed using HLA-DRB1*01:01 and HLA-DRA1:01:01 cDNA in which the TM/Cyto domains were replaced with corresponding domains from CD28 and CD3ζ, respectively, to provide activation and signaling to the CD8⁺ T cell (Fig. 1A). In addition, the second domains of the HLA-DR chains were replaced with corresponding domains from the murine I-E α- and β-chains to enable interaction with murine CD4 (9), and cDNA encoding an antigenic peptide was inserted near the N terminus of the DRB1 chain (Fig. 1B) (11). This design was selected to provide the highest potential affinity between the DR1 CAR molecule and the TCR expressed by CD4⁺ T cells and to provide Ag specificity to the CAR T cell targeting and subsequent lysis of the CD4⁺ T cells.

Using replication-deficient retrovirus produced from the MSGV expression vector containing the DR1 CAR, mouse CD8⁺ T cells from B6.DR1 mice were transduced, and expression of the DR1-CII CAR molecule was assessed by flow cytometry (Fig. 1C). Transduction results were similar for the DR1-HA and the DR1-empty CAR constructs (Fig. 1B and data not shown), and overall efficiencies ranged from 25 to 50%. Given the dependency of the anti-DR Ab on tertiary structure, these data indicated that the DR1 CAR molecule was folded in a native state. Although anti-CD3/CD28 stimulation of the DR1 CAR T cells allowed us to maintain the CAR T cells and to expand their number (Fig. 1C, lower panels), we found that stimulation directly through the CAR molecule with an anti-DR Ab selectively enriched for the CAR T cells. These data support the concept that intracellular signaling is occurring through the CAR to the CD8⁺ T cell; however, subsequent restimulation of the CAR T cells with anti-DR Ab alone was insufficient to maintain the CAR T cell line over prolonged periods (data not shown). Consequently, for the remainder of our experiments, the first stimulation posttransduction was done with anti-DR Ab to enrich for the CAR T cells, and all subsequent stimulations were done with anti-CD3/CD28 Abs to expand the cells for experimental use.

To demonstrate that the DR1 CAR molecule was recognized by DR1-restricted TCR, Ag presentation assays were performed with T cell hybridomas specific for the immunodominant determinant of CII (259-273), or an HA peptide (306-318), presented by DR1. 293T cells that had been transfected with the DR1 CAR molecules were mixed with the T cell hybridomas, and IL-2 production by the T cells was measured as an indication of TCR recognition of the CAR molecule. As shown in (Fig. 2, DR1 CAR molecules that have the CII peptide covalently attached to the DRB1 chain (DR1-CII CAR) are recognized by DR1-restricted CII-specific T cells but not HA-specific T cells in the absence of exogenous added Ag. CAR molecules that are “empty” of an antigenic peptide (DR1-CAR) are not recognized by these T cells; however, the addition of exogenous CII peptide to the cells expressing the DR1-CAR enabled stimulation of the CII-specific T cells. Thus, these data demonstrate that the DR1 CAR molecule is in a native conformation that is recognized by CD4⁺ T cells in an Ag-specific manner.

When the DR1 CAR CD8⁺ T cells were tested for cytolytic function, the CAR T cells were found to be active CTL and lysed CD4⁺ T cells in an Ag-specific manner. Using a flow cytometric assay, CII-specific (TCR-BV8⁺) and HA-specific (TCR-BV14⁺) T cells were incubated with either DR1-CII or DR1-HA CAR T cells at a 10:1 E:T ratio for 4 h, and the cultures were then analyzed for CD8-negative, DR1-negative, and BV14⁺ and BV8⁺ cells (Fig. 3). The CII-specific and HA-specific target cells were mixed at a 1:1 ratio, and after 4 h in the absence of CAR T cells, they were still found to be present in roughly equal numbers (Fig. 3A). Whereas the addition of NT CD8⁺ T cells to the cultures had no effect on either the BV8⁺ (CII-specific) or BV14⁺ (HA-specific) target cell populations (Fig. 3B), the addition of the CAR T cells resulted in the elimination of the target cells in an Ag-specific manner. The DR1-CII CAR T cells effectively lysed the BV8⁺ CII-specific T cells and not the BV14⁺ HA-specific T cells (Fig. 3C), whereas the DR1-HA CAR T cells lysed the HA-specific T cell but not the CII-specific T cells (Fig. 3D). Similar results were observed using ⁵¹Cr-release cytolytic assays (data not shown). Summaries of these data and their statistical analyses are shown in (Figs. 3E and 3F. In all, these data demonstrate the lytic functionality of the DR1 CAR T cells and the specificity of the DR1 CAR in targeting CD4⁺ T cells in a manner that reflects the TCR specificity of the CD4⁺ target cells.

Given that the in vitro CTL data indicated the therapeutic potential of these CAR T cells, we tested their ability to inhibit an Ag-specific T cell response in vivo. B6.DR1 mice were immunized with either CII or the HA peptide, treated with 2 × 10⁶ DR1-CII, DR1-HA CAR T cells, or NT CD8⁺ T cells as control on day 4 and 7 after immunization. On day 11, T cells from draining lymph node were recovered and tested for their ability to be restimulated by Ag in an in vitro proliferation assay. As shown in (Fig. 4A, mice immunized with CII and treated with DR1-CII CAR T cells had a significantly reduced T cell proliferative response (p < 0.03). Whereas the effect varied among the CAR T cell–treated mice, with some mice having a 50% reduction and others a near total reduction, all showed a definitive therapeutic effect on the autoimmune CII T cell response. Similar data were obtained from DR1 CAR T cells that had been in culture for 6 d or 76 d (data not shown). The efficacy of this CAR T cell approach was not limited to autoimmune T cell responses. Treatment of HA-immunized B6.DR1 mice with DR1-HA CAR T cells also effectively reduced the T cell response to this foreign Ag (Fig. 4B, p < 0.004). Whereas in both of these experiments some T cell responses were still observed in the DR1 CAR T cell–treated mice (Fig. 4A, 4B), increasing the number of DR1 CAR T cells administered did not enhance the overall effect (Fig. 4C). Although all three doses of DR1-CII CAR T cells significantly reduced the in vivo T cell response to CII in comparison with the NT control cells (p < 0.05), treating mice twice within 10 d with the highest number of DR1-CII CAR T cells (6 × 10⁶) did not result in a complete elimination of the CII-specific T cell response in the DR1 mice. The small proliferative response still detectable indicated that not all CD4⁺ CII-specific T cells had been eliminated in vivo by the DR1 CAR T cells.

Given the ability of the DR1-CII CAR T cells to downregulate an autoimmune T cell response, we tested their ability to prevent the development of autoimmune arthritis in the B6.DR1 mouse model of RA (9). Mice were immunized with CII, and on days 7, 14, and 21 after disease induction, they were treated with either 2 × 10⁶ DR1-CII CAR T cells, NT CD8⁺ T cells as a control, or left untreated. Mice were then examined three times per week for the presence and severity of arthritis in both fore and hind limbs. In comparison with the ≥80% disease incidence in the two control groups (NT CD8⁺ T cells and untreated, Fig. 5A), mice receiving the DR1-CAR T cells had a 50% reduction in the incidence of autoimmune arthritis that was statistically significant (p < 0.05 by Fisher Exact test) at nearly every time point (Fig. 5). For the DR1-CII CAR-treated mice that did develop disease, the onset was delayed in comparison with controls. Although the effect of the DR1-CII CAR T cells in this model was statistically significant, 40% of the mice still developed disease by the later time point (Fig. 5A). However, the arthritis severity in these DR1-CII CAR T cell–treated mice was generally lower than that of the arthritic mice in the untreated and NT CD8⁺ T cell control groups (Fig. 5B). Given that the CAR T cell treatment ended as the arthritis first began to develop in the CAR-treated mice, these data imply that the DR1-CII CAR T cells may have a short half-life in vivo. As noted for the T cell proliferation studies, similar arthritis incidence data were obtained using CAR T cells that were maintained in culture for short periods (23 d) or long periods (>70 d).

Autoimmune arthritis in the B6.DR1 mouse model is dependent on both a CD4⁺ T cell response to the autoantigen CII as well as a B cell response in the production of pathogenic autoantibody to CII (9, 17). Consistent with the inhibition of both the autoimmune T cell response (Fig. 4) and the inhibition of autoimmune arthritis (Fig. 5), the autoantibody levels in the mice treated with the DR1-CII CAR T cells were also significantly reduced in comparison with the control groups (Fig. 6). At all three time points measured, the anti-CII autoantibody response was statistically reduced (p ≤ 0.02) by more than 50% in comparison with both the untreated mice and those treated with the control NT CD8⁺ T cells (Fig. 6). In all, these data demonstrate that CAR T cells expressing HLA-DR as a chimeric receptor have the potential to be a highly specific therapeutic modality for the treatment of autoimmune disorders.

Whereas the DR1-CAR T cells were effective in reducing autoimmune disease manifestations, the CAR T cells did not appear to undergo significant expansion in vivo. Enumeration of the CAR T cells by flow cytometry after therapeutic administration indicated that CAR T cells were only detectable a short time after passive transfer. As shown in (Fig. 7, 4 d after adoptive transfer, the DR1-CII CAR T cells could be detected in lymph nodes (4.9% DR1⁺/CD8⁺ T cells) and in the spleen (2.1%) and possibly in the blood at a very small percentage. However, by day 10, they were no longer detectable by this approach in any of these tissues. These data imply that the DR1-CII CAR T cells are not strongly stimulated to expand in vivo, as is necessary for successful therapy with CAR T cells used to treat cancer (18, 19). This observation is consistent with the concept that the CD4⁺ CII-specific T cells that are the targets for the DR1-CII CAR T cells are in low frequency (11, 20); thus, it is likely that only a small number of these CAR T cells undergo expansion in vivo.

The structure of our CAR differs substantially from those used in cancer therapy in that the DR1-CAR is a two-chain molecule, each with its own activation and costimulatory domains (Fig. 1). To analyze how this affects the cell signaling through the DR1-CAR in these T cells, we analyzed cytokine production following stimulation of these cells through their CAR molecule compared with pan stimulation with anti-CD3 and anti-CD28 Abs. As shown in Table I, cytokine production via stimulation through DR1-CII CAR was very similar to anti-CD3/CD28 stimulation in terms of the production of TNF-α, IFN-γ, IL-6, IL-10, and IL-17A, although some differences were observed with the stimulation of DR1-HA CAR T cells using these two stimulations, particularly TNF-α production. It is unclear why these two CARs would differ in cytokine production, given that both are identical, with the exception of the nature of the peptide bound by the DR first domains. In all, these data indicate that the signaling domains are functional and that stimulation of these T cells through their CAR does not appear to differ from that generated through CD3/CD28 stimulation, at least when Ab is used to stimulate. However, given that the DR1-CAR cell lines could not be maintained by repetitive stimulation through the CAR with an anti-DR Ab, this indicates that, despite the cytokine production, some signaling aspect is different when initiated through the CAR molecule in comparison with signaling through the endogenous CD3 and CD28 cell membrane proteins.

To determine whether the expression of exhaustion markers was playing a negative role in the function of the DR1-CAR T cells (21), CAR T cell expression of PD-1, Tim-3, and CD69 was measured at several time points after these cell lines were established. Like most other CAR T cells, the expression of these exhaustion markers by our DR1-CII CAR cells increases over time in culture (Fig. 8). By day 10 posttransduction, 34% of the cells have acquired Tim-3⁺ expression, and by day 23, nearly all of the cells express Tim-3 and continue to express this marker through at least day 65 (Fig. 8A). Only a small percentage of these cells express PD-1 through day 23 and reach 50% by day 37 and maintain this level through day 65. Similarly, more than 50% of these CAR T cells express CD69 by day 10 and stay at that level until day 37, after which nearly all of the cells are CD25⁺/CD69⁺. The memory phenotype of these CAR T cells initially included CD62L⁺ stem cell memory (T_SCM) phenotype, but this was quickly lost in culture, as the phenotype became predominantly CD44⁺/CD62L-negative T effector by day 17.

Despite the expression of these exhaustion and CTL phenotypic markers, there was no loss of their cytolytic function using either CD4⁺ CII-specific T cell lines (Fig. 9) or naive CII-specific T cells from a TCR transgenic mouse (data not shown) (22) as targets. Short-term cultured cells lysed CD4⁺ target T cells with similar efficiency as long-term cultured CAR T cells that had now acquired exhaustion marker expression, and similar data were observed when comparing efficacy of short-term and long-term cultured cells in preventing the development of autoimmune arthritis. Our analysis of these target cells revealed that they do not express PD-L1, the ligand for PD-1, indicating that this exhaustion marker is likely not inhibiting the activity of our CAR T cells in vivo, unlike cancer therapy, in which PD-1/PD-L1 interactions play a significant inhibitory role in CAR T cell efficacy in vivo (23).

The results of these studies demonstrate the potential use of CAR T cells in treating autoimmune disorders. Using a humanized mouse model of RA, we have shown that chimeric MHC class II molecules can be used to construct CAR T cells that target and lyse CD4⁺ T cells in an Ag-specific manner in vivo. By encoding the immunodominant peptide for CII in the HLA-DR1 CAR molecule, the DR1 CAR T cells downregulated both the CII-specific CD4⁺ T cell response and the CII-specific autoantibody response, resulting in a significantly reduced incidence of autoimmune arthritis in the B6.DR1 mouse model of RA. The Ag specificity of the CAR T cell was devised by encoding an antigenic peptide as part of the DRB1 chain construct (24), and as was demonstrated using a second antigenic peptide HA(306-318), the specificity of the CAR T cell was easily reprogramed by changing this peptide sequence. Based on the cytokines produced by ligation of the DR1 CAR molecule with an anti-DR Ab, our CAR construct with CD28–CD3ζ TM/Cyto costimulatory domains is capable of generating TCR-like signaling events in CD8⁺ T cells. Although cytokines characteristically expressed by CD8⁺ T cells were produced by this approach, periodic anti-CD3/CD28 stimulation in vitro was required to maintain the DR1 CAR T cells in culture, as repeated signaling through the DR1 CAR was insufficient to maintain the DR1 CAR T cells. In vitro stimulation of the CAR T cells with a combination of anti-CD3/CD28 did drive the expression of Tim-3, PD-1, and CD69, markers associated with CD8 T cell exhaustion, although we saw little to no effect of the expression of these markers on the function of the DR1 CAR T cells. DR1 CAR T cells maintained in culture for >90 days were as effective in inhibiting both the autoimmune response and the development of arthritis in vivo as CAR T cells maintained for only 6 to 25 days, and we have seen no loss of their cytolytic function in vitro with a variety of target cells after their expression of these markers. In all, these data demonstrate the therapeutic potential for an Ag-specific therapy using engineered HLA-based CAR to target and lyse autoimmune CD4⁺ T cells.

Our HLA-DR CAR T cell therapy differs in a number of ways from the anti-CD19 CAR T cells used in cancer therapy. CAR molecules designed for cancer therapy are primarily based on genetically engineered Abs, scFv single-chain molecules composed of the H and L chain variable regions (8). These scFv have only one TM/Cyto domain, hence a single CD28/CD3ζ activation and costimulatory complex in the CAR molecule. HLA-DR molecules are composed of two chains, an α-chain and a β-chain, and each chain has its own TM/Cyto domains. The constructs used in our studies replaced the endogenous TM/Cyto in each of these chains with corresponding CD28–CD3ζ signaling domains. How this affects the function of our DR1 CAR in comparison with the scFv CARs is currently unknown. Although it might be anticipated that such a design would double the signaling strength upon engagement of the DR1 CAR by the TCR, we have seen no evidence of this in our preliminary signaling studies measuring ZAP-70 phosphorylation (K. Whittington and E. Rosloniec, unpublished observation). Although we have CD28/CD3 signaling domains on both the DR1 chains, it is not clear if both are participating in the generation of ZAP-70 phosphorylation. It is also possible that inclusion of signaling domains on both chains generates steric hindrance for the phosphorylating enzymes and limits the CD28–CD3ζ signaling events in the DR1 CAR CD8⁺ T cells. Assessing the signaling events that occur through the DR1 CAR will be paramount in optimizing therapeutic use of these CAR T cells. Given several differences between the DR1 CAR and the scFv CAR, including affinity and membrane density, it is possible that the TM/Cyto signaling domains required for optimal CD8⁺ T cell activation through the DR1 CAR will be different from the CD28–CD3ζ signaling domains often used with scFv CAR.

CAR T cell therapy in cancer is dependent on three primary outcomes in vivo: activation, expansion, and persistence via maturation of the CD⁺8 T_SCM phenotype (25, 26). Although our DR1 CAR T cells demonstrated function in vivo, we were unable to detect evidence of expansion, implying that persistence to a memory phenotype did not occur. Two factors that are likely playing a role in this outcome are frequency of the target cells and affinity of the DR1 CAR for the TCR on the target cells. In terms of frequency, Ag-stimulated CD4⁺ T cells comprise a very small percentage of total CD4⁺ T cells in the mouse, even during an active autoimmune disease. Our previous studies using a DR1-CII tetramer to analyze the T cell response in B6.DR1 mice with autoimmune arthritis indicated that only 1 to 4% of CD4⁺ T cells are specific for the immunodominant determinant of the CII autoantigen, and these comprise 1% or less of the total lymphoid population (11). In terms of the number of target cells for the CAR T cells in vivo, this is a target frequency that is orders of magnitude less than the target frequency of the anti-CD19 CAR T cells used for cancer therapy. Second, the affinity of the DR1 CAR for its target cells is likely significantly lower than the affinity of the Ab-based scFv CARs for their ligand. The affinity of MHC class II and class I binding by TCR is considered to be 10- to 100-fold lower than the affinity of Ab for Ag (27, 28), and how this difference in affinity affects the ability of the CAR to initiate T cell stimulation and function is unknown. Recent studies of CAR signaling focused on protein phosphorylation, by Salter et al. (29) have indicated that signal strength is a key determinant in the fate of CAR T cells after adoptive transfer. Whereas sufficient stimulation of the CAR T cell is required for its cytolytic activity, too much stimulation tends to drive the CAR T cell to an exhaustion phenotype with less than optimal persistence for long-term efficacy. To address this issue, new second- and third-generation signaling domains have been developed for CARs to moderate signal strength for enhanced clinical efficacy (30), and some have been found to improve clinical CAR T cell expansion and persistence (31). Investigation of these new domains with our DR1 CAR may prove beneficial for its function, although given the lower affinity of our CAR with its receptor, it is likely that enhanced signaling through these new domains will be the goal. This concept is supported by our observation that continuous stimulation of the DR1 CAR T cell through the DR1 CAR was insufficient to maintain these cells in culture over extended periods.

The development of CAR T cells for the treatment of other autoimmune diseases has been described by several investigators, with a variety of results (32). For the treatment of pemphigus vulgaris, a CAR T cell expressing a construct encoding the autoantigen desmoglein fused to CD137–CD3ζ signaling domains was designed to target B cells that produce the autoantibody for desmoglein (33). A high degree of success was achieved with the desmoglein CAR T cells in treating a mouse model of pemphigus, although the model is based on CAR T cell targeting of passively transferred B cell hybridomas and not endogenous autoimmune B cells. Two CAR T cell studies also have been described for the treatment of autoimmune diabetes. In one study, a CAR based on an Ab that recognizes an MHC class II:peptide complex linked to CD28–CD137–CD247 signaling domains was used to target the APC involved in stimulating autoimmune CD4⁺ T cells in diabetes (34). In a second approach, Fishman et al. (35) developed CAR T cells that target autoimmune CD8⁺ T cells in diabetes using an MHC class I CAR linked to antigenic peptide/β2m/CD3-ζ domains. Similar to our studies, both of these CAR T cells significantly reduced the incidence of autoimmune disease, but neither generated a durable or curative therapy. As with our CAR T cells used to treat autoimmune arthritis, the target cells for the diabetes CAR T cells are very small in number and reside predominantly within lymphoid tissue. The combination of the low frequency of the target cells and their location in solid tissue may result in the inability of CAR T cells to achieve the activation, expansion, and persistence necessary for high therapeutic efficacy. This problem is similar to the use of CAR T cells in cancer therapy, when the target frequency is low and/or the targets are in solid tissue (36, 37). Both in our CAR studies and in the diabetes CAR studies, expansion and persistence appear to be lacking, suggesting that different costimulatory domains for these CARs may be needed to achieve these functional goals. Recent studies of costimulatory domains used in CAR T cells for cancer therapy suggest that the use of ICOS and 4-1BB domains (38) or a mutation in the CD28 domain (39) might solve these problems.

Although immunodepletion prior to CAR T cell infusion enhances the efficacy of CAR T cell cancer therapy (40), we were not able to include this approach in our studies. Immunodepletion appears to create space for the CAR T cells to take up residence within the lymphoid organs, and promotes activation, expansion, and persistence through the development of a T_SCM phenotype (41–43). Because immunodepletion in our studies would also interfere with the development of the autoimmune arthritis, it would confound the interpretation of the CAR T cell treatment data. The inability to create this lymphoid space may also explain why infusing more CAR T cells did not enhance the efficacy of the CAR T cell treatment in our autoimmune arthritis model. However, for treatment of patients with autoimmune disorders, immunodepletion could be used prior to CAR T cell treatment, as an acute immunosuppression in combination with CAR T cell therapy would likely provide significant benefit in treating the disease.

In summary, the ability to target T cells in an Ag-specific manner would represent a significant advancement in immunotherapy of autoimmune diseases, and the data presented in this study demonstrate the feasibility of achieving this goal using HLA-DR–engineered CAR T cells. By targeting the CD4⁺ T cells that drive the autoimmune response in our model of RA, the T cell and B cell components of this autoimmune disease as well as the incidence of disease were significantly suppressed, clearly indicating the therapeutic potential of this approach. Additionally, these data also demonstrate the flexibility of CAR T cell design and the potential use in treating a variety of diseases. Whereas design of CAR for producing engineered T cells is highly attractive for therapeutic use, the optimal signaling domains to be used are likely to be highly variable, based on several factors, including target frequency, affinity between the CAR and its ligand on the target cells, and the necessity for long- or short-term CAR T cell function in vivo. Combining the knowledge of these functional components with recent advancements in the regulation of CAR T cells after adoptive transfer, both in terms of turning them on and off (44–46) or activating them via exogenous stimuli (47, 48), has significant potential in the generation of new, highly effective immunotherapies.

The authors have no financial conflicts of interest.