A Humanized Lym-1 CAR with Novel DAP10/DAP12 Signaling Domains Demonstrates Reduced Tonic Signaling and Increased Antitumor Activity in B-Cell Lymphoma Models

Non-Hodgkin's lymphomas (NHL) derived from B-cell lineage (B-NHL) are among the most common cancers in the United States with increasing incidence (1). Autologous T cells engineered to express chimeric antigen receptors (CAR) targeting CD19 (CD19-CAR T cells) is a recently FDA-approved approach to treat patients that have relapsed and are resistant (R/R) to traditional therapy. Clinical trials of CD19-CAR T cells have achieved over 80% initial response rates falling to 30% to 40% after 6 months (2–5). Loss of CD19 epitopes is a prominent mechanism of resistance in CD19-CAR T-cell therapies (6, 7). Thus, optimal immunotherapies for B-cell malignancies will likely require targeting additional epitopes on multiple antigens. Under consideration are epitopes found on CD20 (8), CD22 (9), CD123 (10), and others (11, 12). The epitope recognized by the antibody Lym-1 is also a promising candidate for the immunotherapy of B-cell lymphomas. The murine Lym-1 antibody binds to a discontinuous epitope (Lym-1 epitope) on several subtypes of HLA-DR with a higher binding avidity for malignant B cells than normal B cells (13). Thus far, three amino acid residues have been identified as binding sites for Lym-1 but additional studies may be wanted to determine if other sites also contribute to Lym-1 binding (14). However, in an extensive screen only HLA-DR–positive cells can be recognized by Lym-1 antibodies (13). Critical amino acid sequences and sufficient antigen density are both required for Lym-1 binding (14). On the basis of previous histology data, cancer cells from 80% and 40% of patients with B-NHL and chronic lymphocytic leukemia (CLL), respectively, were positive for the Lym-1 epitope (14, 15). In addition, the epitopes recognized by Lym-1 are not shed nor internalized enabling stable binding to human lymphoma cells (13, 16). Lym-1 binding is also enhanced because antigens bound by Lym-1 are concentrated in surface lipid rafts to produce a speckled membrane pattern by indirect immunofluorescence staining compared with CD19, CD20, and CD22 which produce a ring pattern indicating that, unlike the Lym-1 antigen, these other B-cell antigens are more evenly distributed over the surface of tumor cells (17). The Lym-1 antibody also has been extensively studied in patients as an I-131 radiolabeled antibody for the imaging and treatment of Lym-1–positive tumors and shown to be safe at therapeutic doses without causing B-cell aplasia (15, 16, 18). Thus, developing CAR T cells targeting a Lym-1 epitope provides a promising strategy to treat patients with R/R B-cell lymphoma, and would potentially decrease relapse rates if targeted concurrently with other lymphoma-associated antigens.

During development of Lym-1 CAR T cells, we observed impaired ex vivo expansion and exhausted phenotypes when the intracellular signaling domain (ICD) contained 4-1BB3z (BB3z), an ICD commonly used in the construction of CAR T cells. To reduce potential immunogenicity, a humanized Lym-1 antibody was generated to construct huLym-1-B-BB3z CAR T cells which also exhibited the aberrant phenotypes. Although high doses of Lym-1-BB3z (19) and huLym-1-B-BB3z CAR T cells (this report) eradicated metastatic Raji tumors in vivo, limited ex vivo expansion of the CAR T cells would pose a challenge for the manufacture of these CAR T cells. Using huLym-1-B-BB3z CAR as a model, we now demonstrate that the adverse effects of huLym-1-B-BB3z CAR on T cells is associated with ligand-dependent CAR signaling mediated by the interaction of CAR molecules with weakly expressed Lym-1 epitopes on human T cells.

There is increasing recognition that CAR tonic signaling compromises antitumor efficacy of CAR T cells (20, 21). CAR tonic signaling can arise from sustained non-coordinated and weak CD3ζ motif activation caused by ligand-independent mechanisms which result in impaired T-cell proliferation, accelerated T-cell exhaustion, enhanced activation induced cell death (AICD), and poor efficacy in vivo (21). Ligand-independent tonic signaling can be reduced by using appropriate promoter (22), hinge (23), transmembrane domain (24), and signaling domains (20). Aberrant phenotypes caused by ligand-dependent CAR signaling have been attributed to fratricide (25). We report here that the dominant cause of impaired expansion in huLym-1-B-BB3z CAR T cells is due to suboptimal ligand-dependent signaling rather than fratricide and propose a strategy using DAP10 and DAP12 (DAP) ICD to address this issue.

DAP12, an adaptor protein with one immunoreceptor tyrosine-based activation motif (ITAM) in its cytoplasmic domain, is expressed in a broad range of hematopoietic cells including macrophages, natural killer cells, and some subsets of T cells (26). A unique feature of DAP12's ITAM is that the domain also contains an immunoreceptor tyrosine-based inhibitory motif (ITIM), which is not present in any of the ITAMs from CD3ζ (27). In osteoclasts, weak stimulation of DAP12-associated receptor-TREM2 recruits Src homology 2 (SH2)-containing inositol phosphatases (SHIP-1) to the ITIM motif to blunt tonic signaling, whereas a strong receptor stimulation prevents SHIP-1 recruitment (26). In addition, antigen-specific activation of ectopically expressed DAP12 is sufficient to drive T-cell cytotoxicity (28–30), suggesting that the signaling domain from DAP12 can be used for CAR construction. DAP10, another signaling subunit, contributes to DAP12-dependent PI3K activation (26), a pathway that is critical for T-cell expansion and differentiation (31). Thus, we hypothesized a CAR constructed with a DAP10-DAP12 signaling domain would possess a threshold that blunts weak epitope recognition from causing the deleterious effects of suboptimal CAR signaling but allows stronger signals to promote cytotoxic functions. Here, we report that the replacement of ICD in huLym-1-B-BB3z CAR with the cytoplasmic domains from DAP addresses the CAR T-cell expansion issue ex vivo and leads to significantly better tumor control in vivo than its BB3z counterpart. In addition, we also demonstrate that the epitope recognized by Lym-1 on human B lymphoma cell lines does not significantly downregulate in the presence of huLym-1-B-DAP CAR T cells, whereas detection of CD19 antigen diminishes in response to CD19-CAR T cells both in vitro and in vivo. Our studies suggest that the DAP signaling domain described here offers a new strategy to circumvent adverse effects of suboptimal signaling from BB3z-based CARs and thus huLym-1-B-DAP CAR T cells provide a promising modality for patients with lymphomas and leukemias recognized by Lym-1.

Mouse experiments were approved by the USC Animal Care and Use Committee (IACUC 20585) and involved 8- to 13-week-old NSG mice (female and male) purchased from Jackson Laboratories or bred in the USC animal facility under IACUC 20697.

Chimeric Lym-1 (chLym-1; a chimeric analog of the original Lym-1 antibody), IL7-Fc, IL15-Fc, and Dylight 650 anti-261tag antibodies were developed and prepared in our laboratory. Commercial antibodies used were: Alexa Fluro 488 Goat anti-human IgG(H+L) (Thermo Fisher Scientific, Catalog No. A-11013), Alexa Fluor (AF) 647 anti-human IgG Fc (Biolegend, Catalog No. 409320), phycoerythrin (PE) anti-human phosphorylated CD3ζ (pY142; BD Biosciences, Catalog No. 558448), PE mouse anti-human CD22 (Biolegend, Catalog No. 302506), PE mouse anti-human CD19 (Biolegend, Catalog No. 302254), PE mouse anti-human PD-1 (Biolegend, Catalog No. 329906), PE mouse anti-human LAG-3 (Biolegend, Catalog No. 369306), mouse anti-CD247(pY142) (BD Biosciences, Catalog No. 558402), mouse anti-human CD247 (BD Biosciences, Catalog No. 551033).

Reagents used to conduct these studies included: RPMI-1640 (Genesee Scientific, Lot#0519108), DMEM (Genesee Scientific, Lot#05191016), Dialysed FCS (dFCS; Hyclone, Catalog No. SH30079.03), GlutaMAX (Thermo Fisher Scientific, Catalog No. 35050-061), penicillin/streptomycin (Corning, Catalog No. 30-002-CI), nonessential amino acids (Genesee Scientific, Catalog No. 25-536), Click's medium (SIGMA, Catalog No. C5572-500ML), EcoRI (NEB, Catalog No. R3101M), MluI (NEB, Catalog No. R3198L), psPAX2 (Addgene, Catalog No. 12260), pMD2.G (Addgene, Catalog No. 12259), Xfect (Clontech, Catalog No. 631418), Ficoll-Paque (Life Technologies, Catalog No. GE17-1440-02), EasySep Human T-Cell Isolation Kit (STEMCELL, Catalog No. 19051), D-(+)-Trehalose dihydrate (SIGMA, Catalog No. 90210-50G), Lentiblast (OZBiosciences, Catalog No. LB01500), 24-well G-Rex plates (Wilson Wolf, Catalog No. 80240M), Dynabeads human T-activator CD3/CD28 (Thermo Fisher Scientific, Catalog No. 11131D), ImmunoCult human CD3/CD28/CD2 T-cell activator (STEM CELL, Catalog No. 10970), IL2 ELISA Kit (Thermo Fisher Scientific, Catalog No. EH2IL2), GM-CSF ELISA Kit (Thermo Fisher Scientific, Catalog No. EHGMCSF), and INFγ ELISA Kit (Thermo Fisher Scientific, Catalog No. EHIFN), intracellular staining permeabilization wash buffer (Biolegend, Catalog No. 421002), FluoroFix Buffer (Biolegend, Catalog No. 422101), Sytox Green (Thermo Fisher Scientific, Catalog No. S7020), and CountBright Absolute counting beads (Thermo Fisher Scientific, Catalog No.C36950).

Jurkat, K562, Daudi, Karpas-299, B35M, BALL-1, Chevallier, and Raji cell lines were obtained from ATCC. The SU-DHL-6 (32), SU-DHL-10 (32), and NU-DHL-1 (33) human lymphoma cell lines were developed by Alan Epstein in-house. Raji-eGFP/Luc cells were a gift from Dr. Yvonne Y. Chen at the University of California. All lymphoma lines were cultured in RPMI-1640 supplemented with 10% dFCS, 1% GlutaMAX, and 1% penicillin/streptomycin. HEK-293 LTV cells (Cell Biolabs, Catalog No. LTV-100) were cultured in DMEM supplemented with 10% dFCS, 1% GlutaMax, 1% nonessential amino acids, and 1% penicillin/streptomycin. Primary human T cells were enriched from human buffy coats (Zen-Bio, CAR#SER-BC-SDS) and cultured in T-cell medium (43% Click's medium, 43% RPMI-1640, 10% dFCS, 2% GlutaMAX, 1% nonessential amino acids, 1% penicillin/streptomycin) supplemented with 50 ng/mL IL7-Fc and 100 ng/mL IL15-Fc. All cell lines used were routinely tested for mycoplasma contamination using MycoFluor Mycoplasma Detection Kit (Thermo Fisher Scientific, Catalog No. M7006).

Antibodies at concentrations ranging from 0.013 to 1300 nmol/L in 100 μL were incubated with 0.2 million Raji cells at 4°C for 30 minutes, followed by three washes with washing buffer (2% FBS in PBS). Bound antibodies were then incubated for 30 minutes with AF 488-conjugated goat anti-human IgG(H+L) secondary antibody at a concentration of 5 μg/mL or by AF 647 conjugated anti-human IgG Fc at 5 μL per sample. Cells were washed twice and subjected to flow cytometry analysis. Mean fluorescence intensity (MFI) was recorded and plotted to evaluate antibody binding. For staining in Figs.1C and 3C, 10 μg of the antibodies was incubated with 0.2 million cells in 100 μL at 4°C for 30 minutes. After washing as above, 5 μL AF-647-conjugated anti-human IgG Fc was added to the cells in wash buffer residuals for detection. Samples were evaluated using an Attune flow cytometer (Thermo Fisher Scientific) and analyzed using Flowjo software (BD Biosciences).

The coding genes for CAR were synthesized by Integrated DNA Technologies (IDT) and ligated into the lentiviral vector pLVX-EF1α-IRES-Zsgreen (Clontech) through EcoRI and MluI restriction sites. For all CAR constructs, a 10 amino acids epitope “AVPPQQWALS” (261-tag) derived from human placenta growth factor was inserted directly after the scFv sequence. Lentivirus was produced by transient transfection of HEK-293LTV following a Xfect protocol as described previously (19). Briefly, transfer vectors alone with psPAX2 and pMD2.G (molar ratio 2:1:1) were mixed and cotransfected to HEK-293LTV cells. Supernatants containing viral particles were collected at 24 and 48 hours after transfection and were combined, filtered, and concentrated by ultracentrifugation at 20,000 × g for 2 hours. Pelleted virus was then resuspended in PBS supplemented with 1% BSA and 7% trehalose, aliquoted, and stored at −80°C. Viral titers were measured by transducing 10⁶ Jurkat T cells with 10-fold serial dilutions of virus vector. Forty-eight hours after transduction, Jurkat cells were washed and analyzed for transgene expression by flow cytometry. Positively transduced cells at a range of 10% to 20% were used to calculate the virus transducing units (TU) via the following formula: TU/mL = (10⁶ seeded cells × % positive cells × 1,000)/μL of virus vector.

Human buffy coat preparations were purchased from Zenbio Inc., and used to obtain peripheral blood mononuclear cells (PBMCs) for T-cell enrichment. PBMCs were isolated using Ficoll–Paque followed by T-cell isolation using the EasySep Human T Cell Isolation Kit as per the manufacturers' protocols. Isolated cells were then cultured in T-cell medium. On day 0, T cells were activated by adding Dynabeads human T-activator CD3/CD28 at a 1:1 ratio, and on day 3 transduced by centrifugation at 1,200 × g for 45 minutes with lentivirus (MOI = 10) and Lentiblast. Transduction was performed once, followed by a media change after 24 hours, after which cells were transferred to 24-well G-Rex plates supplemented with fresh T-cell medium. Transduction efficiency was evaluated by flow cytometry at day 7 using Dylight 650 conjugated anti-261tag antibody. For re-stimulation, on day 7, 20 μL ImmunoCult human CD3/CD28/CD2 T-cell activator was added with one million T cells in 2 mL T-cell medium and on day 9, 5 mL T-cell medium was added. On day 12, approximately 5 mL medium above the settled cells was removed and 5 mL fresh T-cell medium added. On day 14, for some preparations, restimulation was performed with the same procedure on day 7. ImmunoCult was used for restimulation because, in our hands, Dynabeads CD3/CD28 stimulator promoted the expansion of mostly CD4⁺ T cells, whereas ImmunoCult human CD3/CD28/CD2 T-cell activator led to a balanced expansion of CD4⁺ and CD8⁺ T cells. For all cell counting, the countess automated cell counter (Invitrogen) was used. PD-1 and LAG-3 expression on Mock or CAR T cells were assessed on days 7 and 14 before restimulation. One-half million cells were incubated with Dylight 650 conjugated anti-261tag antibody and PE conjugated anti-human PD-1 or anti–LAG-3. Labeled cells were then subjected to flow analysis. “Mock T cells” refers to T cells that were carried through the above procedures except no virus was added at the transduction step on day 3.

CAR T cells from day 9, not restimulated on day 7, were adjusted to 50% positive for CAR T cells by the addition of Mock T cells. These adjusted preparations were incubated with 0.1 million target cells at various ratios of CAR T cells in flat-bottom 96-well plates for 24 hours without the addition of cytokines. Luminescent reads from target cells without effector cells were used as controls. Two-fold serial dilutions of 0.2 million target cells were used to generate a standard curve to correlate live cells to luminescent reads. Live target cell number after 24-hour incubations was calculated by correlating the luminescent signal reads to the standard curve. The percentage of cell lysis was calculated by the following formula:

Flow cytometry-based cytotoxicity assays: Two × 10⁵ CAR T cells were incubated with target cells at a 1:1 ratio in 24-well plates. The percentage of live target cells at 1 and 48 hours after mixing was recorded and used to calculate Lysis% by the following formula:

Mock or CAR T cells from day 9, not restimulated on day 7, were used for cytokine secretion assays. Two × 10⁵ effector cells and target cells were cocultured in the absence of added cytokines at a ratio of 1:1 in 96-well plates for 24 hours. Supernatants were collected and subjected to ELISA measurement per manufacturer's instructions.

One-half million Mock or CAR T cells from day 9, not stimulated on day 7, were fixed and then stained with 2 μg Dylight-conjugated anti-261 tag antibody. Labeled cells were permeabilizated per manufacturer's instructions. Permeabilized cells were then stained with PE conjugated anti-human phosphorylated CD3ζ antibody at 5 μL per sample. Samples were washed and subjected to flow cytometry analysis.

Mock or CAR T cells from day 9, not restimulated on day 7, were used for these experiments. Tumor cells (4 × 10⁵) were cocultured with 2 × 10⁵ CAR T cells at an E:T ratio of 1:2 in 2 mL T-cell medium without cytokine supplement for 24 hours. One hundred microliters of medium with cells were collected and stained for CD22, Lym-1, or CD19. For Raji-eGFP/Luc, instead of using CD22, GFP was used to identify tumor cells. After incubation at room temperature for 20 minutes, 400 μL PBS and 25 μL counting beads were added to each tube. Samples were then subject to flow cytometry analysis. MFI was quantified by Flowjo software.

One million Raji-eGFP/Luc cells in 100 μL PBS were injected intravenously via the lateral tail vein (designated as day 0). Luciferase activity was measured on day 6 via bioluminescence imaging (BLI) to assess tumor burden. On the same day, five million Mock T or CAR T cells were prepared in 100 μL PBS and injected intravenously using insulin syringes. Tumor progression was monitored by bioluminescence at indicated days by using an Xenogen IVIS 200 at the USC Molecular Imaging Center or IVIS Lumina Series III at the USC Translational Research Laboratory. Mice were anesthetized with vaporized isoflurane and administrated D-luciferin (50 mg/kg) via intraperitoneal injection before imaging. In some experiments, as specified in figure legends, the first BLI measurement was performed on day 7 followed by injection with variable amount of Mock or CAR T cells on day 8. In all studies, Mock or CAR T cells from day 9 (no restimulation on day 7) were used and hind-leg paralysis was used as endpoint for euthanasia. Survival data are shown by Kaplan–Meier plots and analyzed by the log-rank test.

Graphs were plotted using GraphPad Prism Software. Data were analyzed using SPSS software (IBM). The statistical analysis method is indicated in the figure legends. Unless otherwise stated, data are presented as mean ±SD, and P < 0.05 was considered as significant.

In previous studies, we demonstrated that Lym-1-B-BB3z CAR T cells induced complete remission in mice with disseminated Raji tumors (19). Because of these promising results, antibody humanization was performed to reduce potential immunogenicity of the Lym-1 ScFv when used in the CAR T cells for future patient studies. Humanization was outsourced to Oak BioScience and a panel of 12 humanized antibodies were supplied to us (Fig. 1A). The binding ability of those antibodies on Raji cells was measured by flow cytometry and the huLym-1-B antibody, which showed the highest binding at each tested concentration, was chosen for the development of CAR T cells (Fig. 1A). We then produced the huLym-1-B antibody in-house using transient expression with the provided sequence for subsequent evaluations. Compared with chLym-1, huLym-1-B binds to Raji cells with reduced MFI and an approximately 2-fold higher ED50 (Fig. 1B). Similar to its binding to Raji cells, huLym-1-B shows slightly lower MFI than chLym-1 in most Lym-1–positive cell lines (Fig. 1C). Both chLym-1 and huLym-1-B did not bind to K562 cells indicating that huLym-1-B retains specificity similar to that of the parent Lym-1 antibody (Fig. 1C).

To generate CAR against Lym-1 epitope, the ScFvs derived from Lym-1 or huLym-1-B were fused to a conventional second-generation CAR framework with 4-1BB and CD3ζ signaling domains (Fig. 2A). A 10–amino acid epitope “AVPPQQWALS” (261-tag) derived from human placenta growth factor was inserted between the ScFv and CD8a hinge to enable CAR detection by using an in-house antibody (Dylight 650 conjugated anti-261tag antibody; Fig. 2A). Both constructs were successfully expressed in human primary T cells with comparable transduction efficiency (Fig. 2C). T cells expressing Lym-1- or huLym-1-B-BB3z CAR T lysed Lym-1–positive Raji cells efficiently after overnight coculture, whereas increased cytotoxicity was not observed when cocultured with Lym-1–negative K562 cells, indicating similar epitope specific cytotoxicity of the two CAR T cells (Fig. 2D). In addition, a dose of 5 million Lym-1- and huLym-1-B-BB3z CAR T cells eradicated disseminated Raji-eGFP/Luc tumors in NSG mice and led to tumor-free survival for at least 60 days (Supplementary Fig. S1). However, in response to α-CD2/CD3/CD28 stimulation, both Lym-1- and huLym-1-B-BB3z CAR T cells exhibited impaired expansion (Fig. 2E). When restimulated on day 7, Mock T cells increased approximately 30-fold by day 14, compared with less than an average of 4-fold increase for Lym-1- and huLym-1-B-BB3z CAR T cells (Fig. 2E). Furthermore, increased CD3ζ phosphorylation was observed in Lym-1- and huLym-1-B-BB3z CAR–positive cells, but not in untransduced T cells in the same preparation indicating sustained activation of CAR transduced cells (Fig. 2F and G). Consistent with these results, CAR-positive T cells also manifested increased expression of inhibitory receptors PD-1 and LAG-3 (Fig. 2H and I). Enhanced CD3ζ phosphorylation and inhibitory receptor expression in CAR T cells suggested that the impaired proliferation arose from increased basal level activation. Although huLym-1-B-BB3z CAR T cells showed strong activity in vitro and in vivo, limited proliferation would challenge the production of huLym-1-B-BB3z CAR T cells for clinical application because extensive ex vivo expansion is required to generate optimal therapeutic doses.

CD19-BB3z CAR T cells consisting of the anti-CD19 FMC63 ScFv generated in our laboratory, a construct with the same CAR framework as huLym-1-B-BB3z, did not show the impaired expansion seen with huLym-1-B-BB3z CAR T-cell preparations (19). This difference suggested that the cause involves the huLym-1-B ScFv. The Lym-1 antibody recognizes a conformational epitope in several subtypes of HLA-DR (14), but Lym-1 binding on human T cells has not been reported. Thus, previously unreported sparse Lym-1 epitope expression on activated T cells might be sufficient to induce ligand-dependent suboptimal CAR signaling or lead to CAR-mediated fratricide, either of which could cause impaired expansion of huLym-1-B-BB3z CAR T cells. To test this hypothesis, Lym-1 and huLym-1-B binding to activated T cells was carefully assessed and a small but real amount of binding was detected (Fig. 3C). To further test this hypothesis, two CAR constructs were generated. In one construct, two point-mutations were introduced in the CDR3 region of the variable heavy chain to disrupt the binding ability of huLym-1-B and to construct huLym-1-Bmut-BB3z CAR (Fig. 3A). The antibody version of huLym-1-B with these two mutations in CDR3 (huLym-1-Bmut) was also produced. The huLym-1-Bmut antibody has approximately a 25-fold greater ED50 than huLym-1-B (ED50, 1.4 × 10⁻⁷ vs. 5.9 × 10⁻⁹; Fig. 3B) when measured against Raji cells and showed no enhanced binding to T cells compared with isotype (Fig. 3C). Although CD3ζ phosphorylation was found in about 5% of huLym-1-Bmut-BB3z CAR T cells and PD-1 and LAG-3 were transiently upregulated on day 7 (Fig. 3E–H), huLym-1-Bmut-BB3z CAR T-cell preparations did not exhibit impaired expansion suggesting epitope recognition was required for the impairment (Fig. 3D–H).

To determine if CAR signaling was required to impair expansion, a second construct, huLym-1-B-BB3zY-F, was generated wherein all six tyrosines in the three ITAMs of the CAR-CD3ζ domain were converted to phenylalanines (Fig. 3A). HuLym-1-B-BB3zY-F CAR T cells had neither increased CD3ζ phosphorylation nor upregulation of PD-1 and LAG-3 and expanded as efficiently as Mock T cells, indicating signaling involving CAR-CD3ζ was required for the impairment (Fig. 3D–H).

We next investigated whether fratricide is a substantial cause of the diminished expansion of huLym-1-B-BB3z CAR T cells. huLym-1-B-BB3z CAR T cells showed dramatically enhanced spontaneous apoptosis (∼56%) compared with the CAR-negative population (∼10%) in the same preparation (Supplementary Fig. S2A). In addition, we did not observe markedly increased apoptosis of CD19-BB3z CAR T cells when they were cocultured overnight with huLym-1-B-BB3z CAR T cells (Supplementary Fig. S2B). Taken together, these results suggest that in this time frame the dominant cause of impaired ex vivo proliferation of huLym-1-B-BB3z CAR T cell is ligand-dependent suboptimal CAR signaling and not fratricide. It is possible that fratricide could occur later and was not detected in this experiment.

Next, we used DAP signaling domains to construct huLym-1-B-DAP CAR (Fig. 4A and B). huLym-1-B-DAP CARs were expressed on human primary T cells with equivalent transduction efficiency as huLym-1-BB3z CAR (Fig. 4D). Importantly, ex vivo expansion of huLym-1-B-DAP CAR T cells was not impaired (Fig. 4C). In addition, compared with huLym-1-B-BB3z CAR T cells, huLym-1-B-DAP CAR T cells showed no enhanced spontaneous Annexin V staining in culture (Supplementary Fig. S2A) and exhibited less AICD when cultured with Raji cells (Supplementary Fig. S3).

Interestingly, phosphorylation of the endogenous CD3ζ was increased in huLym-1-B-DAP CAR T cells compared with mock T cells (Fig. 4D and E; Supplementary Fig. S4). Consistent with the basal levels of CD3ζ phosphorylation in huLym-1-B-DAP CAR T cells, PD-1 and LAG-3 expression were also higher than Mock T cells but were significantly lower than huLym-1-B-BB3z CAR T cells on day 14 (Fig. 4F and G). Together, these data demonstrate that DAP signaling domains circumvent adverse effects caused by suboptimal CAR-CD3ζ signaling and enable nonimpaired production of huLym-1-B CAR T cells.

To evaluate the effector function of huLym-1-B-DAP CAR T cells in vitro, cytotoxicity and cytokine release in response to Lym-1 epitope-negative (K562) and -positive (Raji) cell lines were assessed. huLym-1-B-DAP CAR T cells lysed Raji cells in proportion to increased effector to target ratio, reaching about 80% killing at 2:1 ratio after overnight coculture (Fig. 5A). No enhanced cytotoxicity was evident when K562 cells were used as target cells (Fig. 5A). Consistent with these findings, huLym-1-B-DAP CAR T cells also secreted multiple cytokines when cocultured with Raji but not K562 cells (Fig. 5B). In addition, we assessed the function of huLym-1-B-DAP CAR T cells against a panel of human lymphoma and leukemia B-cell lines with variable Lym-1 epitope expression. Despite highly variable cytokine release, huLym-1-B-DAP CAR T cells exhibited equivalent cytotoxicity (Supplementary Fig. S5). These data demonstrated huLym-1-B-DAP CAR T cells retained the specificity of the parent Lym-1 antibody.

We next examined the in vivo antitumor efficacy of huLym-1-B-DAP CAR T cells against disseminated Raji tumors in NSG mice. To better reveal if the DAP signaling domains confer improved function in vivo, only 1 million CAR T cells were injected instead of 5 million cells and, to increase the tumor burden challenge treatment, was given on day 8 after injecting Raji cells rather than on day 6. Using this modified protocol, 1 million huLym-1-B-BB3z CAR T cells were unable to eliminate Raji tumors and all mice succumbed to tumor progression by day 51 (Fig. 5D–F). In contrast, treatment with 1 million huLym-1-B-DAP CAR T cells led to durable tumor control and significantly better survival (Fig. 5F). Moreover, surviving mice rechallenged with tumor cells showed delayed tumor progression and prolonged survival, indicating huLym-1-B-DAP CAR T cells persisted and conferred resistance in those NSG mice (Fig. 5G and H).

Relapse is often observed in the treatment of B-cell malignancies by CAR T cells targeting CD19 (CD19-CAR) and this downregulation represents an important mechanism enabling resistance to CD19-CAR therapy (34). CAR T cells directed against antigens that are less prone to downregulate could reduce antigen escape and improve therapeutic efficacy. To determine if huLym-1-B-DAP CAR T cells promote downregulation of the Lym-1 epitopes, we cocultured Raji cells with Mock, huLym-1-B-DAP CAR, or CD19-BB3z CAR T cells. Within 24 hours, both huLym-1-B-DAP CAR and CD19-BB3z CAR T cells inhibited Raji expansion (Fig. 6C and D). However, there was marked CD19 antigen downregulation when Raji cells were cocultured with CD19-BB3z CAR T cells, whereas neither CD19 nor Lym-1 epitope downregulation was evident when cocultured with huLym-1-B-DAP CAR T cells (Fig. 6A and B). Similar results were obtained from a panel of human B lymphoma cell lines (Supplementary Fig. S6). To determine if the difference was due to the use of the DAP signaling domain instead of the BB3z domain, CD19-DAP CAR T cells were generated and cocultured with Raji cells. Downregulation of CD19 antigen on Raji cells still occurred (Supplementary Fig. S7). In addition, significant Lym-1 epitope downregulation was not observed on Raji cells when they were cocultured with either huLym-1-B-BB3z or huLym-1-B-DAP CAR T cells (Supplementary Fig. S7). These results indicate that target downregulation is attributed to a property of the antigen rather than the signaling domain of the CAR construct.

To assess epitope downregulation in vivo, CD19 and Lym-1 epitope expression were measured on Raji cells obtained from bone marrow of mice undergoing CAR T-cell therapy. As seen ex vivo, treating NSG mice bearing Raji cells with CD19-BB3z CAR led to significant CD19 antigen downregulation (Fig. 6E and F). Importantly, detection of neither Lym-1 epitope nor CD19 antigen was downregulated during huLym-1-B-DAP CAR T-cell treatment (Fig. 6E and F). The in vivo antitumor efficacy of huLym-1-B-DAP, CD19-BB3z, and CD19-DAP CAR T cells in disseminated Raji-bearing NSG mice was also characterized in a protocol wherein on day 0, 10⁶ Raji cells were injected intravenously followed by various doses of Mock or CAR T cells on day 8. One dose of 2 million huLym-1-B-DAP CAR T cells led to tumor free survival for at least 90 days (Supplementary Fig. S8B and S8C). In contrast, high tumor burden existed in most of the NSG mice treated with CD19-BB3z or CD19-DAP CAR T cells and all mice succumbed to tumor progression by day 79 at the dose of 2 million cells (Supplementary Fig. S8B and S8C). Moreover, in this experimental model, increasing the CD19-CAR T cell dose to 5 million cells still failed to achieve tumor-free survival (Supplementary Fig. S8B and S8C). In summary, the Lym-1 epitope was not found to downregulate under the pressure of huLym-1-B-DAP CAR T-cell treatment and this property of Lym-1 epitope may have contributed to the superior in vivo efficacy of huLym-1-DAP CAR T cells compared with CD19-CAR T cells.

Clinical trials of CD19-CAR T cells to treat B-cell malignancies have produced high initial complete response rates. Unfortunately, a substantial fraction of treated patients relapse with CD19-negative/low tumors (34), indicating the need to identify additional effective targets. Here, we describe the design and development of human CAR T cells directed to the Lym-1 epitope, which is highly expressed on most human B-cell lymphomas and leukemias (15, 35). By substituting the conventional 4-1BB3z signaling domain with DAP, we were able to circumvent impaired ex vivo proliferation of huLym-1-B-BB3z CAR T cells induced by sustained interaction of huLym-1-B-BB3z CAR with weakly expressed Lym-1 epitope on T cells. Moreover, huLym-1-B-DAP CAR T cells exhibited epitope-driven effector functions as evidenced by increased in vitro cytotoxicity, cytokine release, as well as potent in vivo tumor control even with reduced doses of CAR T cells. Furthermore, neither the Lym-1 epitope nor CD19 antigen on B-cell lines downregulated in the presence of huLym-1-B-DAP CAR T cells. These findings indicate huLym-1-B-DAP CAR T cells appear to be a promising cell therapy product to explore in the clinic.

During the course of these studies, we observed impaired expansion of huLym-1-B-BB3z CAR T cells that targets the Lym-1 epitope, but such limited expansion was not found in huLym-1-B-BB3z CAR T cells with crippled binding ability (huLym-1-Bmut-BB3z) nor with ablated CD3ζ activity (huLym-1-B-BB3zY-F). These data suggested that limited expansion was mediated by ligand-dependent activation of the CD3ζ ITAMs signaling moiety in the huLym-1-B-BB3z CAR construct. This observation is consistent with previous reports of second-generation CAR with CD28 costimulation domain redirected against GD2 (20) and ErbB2 (25), where limited expansion, AICD, progressive exhaustion, and poor in vivo efficacy were attributed to unconstrained CAR-CD3ζ activation. In both cases, replacing the CD28 costimulation domain with the signaling domain from 4-1BB mitigated adverse effects induced by chronic CAR-CD3ζ signaling through incompletely understood mechanisms. In our hands, however, using a 4-1BB costimulation domain still resulted in a preparation with poor in vitro expansion of huLym-1-B-BB3z CAR T cells.

Qualitatively different functions of ITAMs were first documented by Combadiere and colleagues (36), who reported that phosphorylation of the first and third ITAMs in CD3ζ stimulated greater apoptosis than phosphorylation of the second ITAM in T cells. Consistent with this observation, in a murine B-cell lymphoma model, Kochenderfer and colleagues demonstrated that anti-murine CD19-CAR T cells with mutated first and third CAR-CD3ζ ITAMs are resistant to apoptosis and could mediate antilymphoma efficacy better than CD19-CAR with three functional ITAMs (37). A recent study by Feucht and colleagues (38) found ablating the function of the second and third ITAMs in CD3ζ moiety of CD19-CAR resulted in preferential central memory differentiation, decreased T-cell exhaustion, and increased persistence in vivo. These data suggest that the function of each ITAM can be qualitatively different and the selection of ITAM(s) in CAR signaling domains is an approach to improve the efficacy of CAR T cells.

Evidence from others (20, 25) and this report supports the hypothesis that chronic suboptimal activation of CAR-CD3ζ ITAMs is a cause of aberrant phenotypes of CAR T cells. We therefore sought to mitigate these adverse effects by using other ITAM-containing motifs to substitute CAR-CD3ζ moiety while retaining T-cell activation potential. We chose DAP12 because there is an ITIM motif embedded in its ITAM sequence, this property may provide distinct signal outputs of DAP12 in response to stimuli with differential strength (26). DAP12 does not have target recognition domains in the extracellular region. Instead, DAP12 associated proteins, such as KIR2DS2 (29) and NKp44 (39), are responsible for target recognition, resulting in DAP12 activation to initiate cytotoxic functions. Teng and colleagues (28) and Wang and colleagues (29) demonstrated that ectopic coexpression of DAP12 along with its associated scFv-modified receptors in either murine or human T cells mediated antigen-specific tumor eradiation, indicating that activation of DAP12 in T cells is sufficient to drive T-cell cytotoxicity. In our CAR construct design, instead of using a multichain format, we directly substituted the CD3ζ signaling domain with DAP12 and added DAP10 for costimulation. The resulting huLym-1-B-DAP CAR addressed the in vitro expansion problem and mediated significantly better in vivo efficacy than huLym-1-B-BB3z CAR T cells, even though the two showed equivalent cytotoxicity in vitro (Fig. 5A). These results further support the finding that in vitro cytotoxicity is insufficient to predict relative in vivo efficacy (20) and highlights the impact of the signaling domain has on in vivo function of CAR T cells. Additional studies are required to identify the mechanisms by which the DAP signaling domain circumvents ligand-dependent suboptimal CAR signaling which caused huLym-1-B-BB3z CAR T cell expansion failure.

Our results demonstrate that, unlike CD19, the Lym-1 epitope does not significantly downregulate upon CAR engagement. This lack of downregulation of the Lym-1 epitope was replicated in a panel of human B-lymphoma cell lines when cocultured with huLym-1-DAP CAR T cells (Supplementary Fig. S6). Although CAR T-cell trogocytosis may play a role in antigen downregulation (40), we did not observe equivalent Lym-1 epitope downregulation, suggesting that markedly CD19 antigen downregulation may involve other mechanisms (Fig. 6; Supplementary Fig. S6). Crosslinking with anti-CD19 antibodies could induce CD19 antigen downregulation through receptor-mediated endocytosis (41), indicating that interaction between CD19-CAR and CD19 antigen interaction may contribute to surface CD19 antigen downregulation. Although CD19 downregulation in tumor cells under the pressure of CD19-CAR T cells is a reversable process (42), transient antigen downregulation could diminish CD19-CAR T cells' antitumor efficacy and allow tumor immune escape (40). Consistent with this hypothesis, neither CD19-BB3z nor CD19-DAP CAR T cells were able to induce tumor-free survival at a dose up to 5 million in the modified animal protocol, whereas huLym-1-B-DAP CAR T cells mediated a rapid and sustained tumor control leading to tumor-free survival at the lower dose of 2 million cells (Supplementary Fig. S8B and S8C). Interestingly, treatment with CD19-DAP CAR T cells induced a significantly better survival than CD19-BB3z T cells at the 1 million cell dose level (Supplementary Fig. S8D and S8E). In addition, regardless of the CAR T-cell dose, no hind-leg paralysis, the criteria for euthanasia, was observed in CD19-DAP CAR before day 41 (Supplementary Fig. S8B–S8E). In contrast, earlier development of hind-leg paralysis (between days 20 and 30) was repeatedly seen in CD19-BB3z CAR-treated mice (Fig. 8B–E). These observations suggest using the DAP ICD instead of the BB3z ICD may improve the efficacy of CAR T-cell preparations that do not exhibit impaired function due to suboptimal CAR signaling. The underlying mechanisms and the significance of this difference remain to be investigated.

In summary, our work indicates that huLym-1-B-DAP CAR T cells hold promise for treating Lym-1–positive B-cell lymphomas. The observation that the DAP signaling domain can circumvent impaired proliferation induced by ligand-dependent signaling in CD3ζ-based CAR while retaining equivalent or higher antitumor efficacy, highlights the importance of the stimulation domain selection for CAR design and identifies a new CAR structure format to address the adverse effects of suboptimal CAR signaling on T cells. Furthermore, DAP signaling domains may also improve the function of other CAR T-cell preparations even if there is no evidence of adverse CAR signaling. Finally, our report suggests that targeting an epitope that does not significantly downregulate upon CAR engagement may also contribute to sustained CAR T-cell efficacy.

P. Hu holds ownership interest (including patents) in Cell BT. A.L. Epstein is an employee/paid consultant for and reports receiving commercial research grants from Cell BT. No potential conflicts of interest were disclosed by the other authors. The University of Southern California has a financial interest in Cell BT, Inc.

Conception and design: L. Zheng, L.A. Khawli, P. Hu, H.R. Kaslow, A.L. Epstein

Development of methodology: L. Zheng, L. Ren, A. Kouhi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Zheng, L. Ren, A.L. Epstein

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Zheng, L. Ren, H.R. Kaslow, A.L. Epstein

Writing, review, and/or revision of the manuscript: L. Zheng, L. Ren, A. Kouhi, L.A. Khawli, H.R. Kaslow, A.L. Epstein

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.L. Epstein

Study supervision: L.A. Khawli, H.R. Kaslow, A.L. Epstein

We gratefully thank Ivetta Vorobyova and Ryan Park from the Molecular Imaging Center (USC) and Dr. Junji Watanabe from the Translation Research Laboratory (USC) for the mouse Bioluminescence imaging and data processing. We are also grateful to the USC animal facility for providing animal support and breeding protocols. This work was supported in part by Cell BT, Inc., and grant no. 9031415 from the Ming Hsieh Institute of University of Southern California.

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.