Purpose:

Targeting nonspecific, tumor-associated antigens (TAA) with chimeric antigen receptors (CAR) requires specific attention to restrict possible detrimental on-target/off-tumor effects. A reduced affinity may direct CAR-engineered T (CAR-T) cells to tumor cells expressing high TAA levels while sparing low expressing normal tissues. However, decreasing the affinity of the CAR-target binding may compromise the overall antitumor effects. Here, we demonstrate the prime importance of the type of intracellular signaling on the function of low-affinity CAR-T cells.

Experimental Design:

We used a series of single-chain variable fragments (scFv) with five different affinities targeting the same epitope of the multiple myeloma–associated CD38 antigen. The scFvs were incorporated in three different CAR costimulation designs and we evaluated the antitumor functionality and off-tumor toxicity of the generated CAR-T cells in vitro and in vivo.

Results:

We show that the inferior cytotoxicity and cytokine secretion mediated by CD38 CARs of very low–affinity (Kd < 1.9 × 10−6 mol/L) bearing a 4-1BB intracellular domain can be significantly improved when a CD28 costimulatory domain is used. Additional 4-1BB signaling mediated by the coexpression of 4-1BBL provided the CD28-based CD38 CAR-T cells with superior proliferative capacity, preservation of a central memory phenotype, and significantly improved in vivo antitumor function, while preserving their ability to discriminate target antigen density.

Conclusions:

A combinatorial costimulatory design allows the use of very low–affinity binding domains (Kd < 1 μmol/L) for the construction of safe but also optimally effective CAR-T cells. Thus, very-low-affinity scFvs empowered by selected costimulatory elements can enhance the clinical potential of TAA-targeting CARs.

Translational Relevance

The broader applicability of CAR-T-cell therapy is restricted by the lack of tumor-specific target antigens while the most promising targets are tumor-associated antigens (TAA). Decreasing the affinity of a CAR for the target is a popular strategy to avert on-target/off-tumor toxicity when targeting TAAs. Here, we evaluated whether lowering the affinity for the target would affect the costimulatory requirements of CAR design for optimal functionality and persistence. We found that the combination of CD28 and 4-1BB signals potentiates even very low–affinity CAR-T cells and improves their immunotherapeutic properties while simultaneously preserving their ability to discriminate target-antigen density. Therefore, a combinatorial costimulatory design allows the use of very low–affinity binding domains (Kd < 1 μmol/L) for the construction of safe but also optimally effective CAR-T cells. These results are relevant and inform for the future design of efficient CAR-T-cell therapies to optimally target TAAs for hematologic and solid tumors.

Adoptive immunotherapy with genetically engineered T cells bearing tumor antigen–specific chimeric antigen receptors (CAR) holds the potential for effective treatment of hematologic malignancies and solid tumors. CARs are synthetic receptors that redirect antigen recognition and mediate T-cell activation, in a single molecule, through the fusion of an extracellular antigen-binding moiety, such as a single-chain-variable region (scFv), with an intracellular signaling domain usually derived from the CD3ζ chain (1). CARs endow T cells with customizable antigen recognition as scFv domains of different specificity and antigen-binding properties can be interchangeable. These properties confer a broad applicability potential to CAR-T cells for a wide range of patients and diseases. Importantly, second- and third-generation CARs provide combined activation and costimulatory signals (1, 2). The addition of intracellular components from known costimulatory receptors/molecules produces signaling cascades similar to their normal counterparts and enhances T-cell activation, expansion, and in vivo persistence. Up to date, second-generation CAR-T cells targeting CD19 have been shown to induce impressive responses in chemotherapy-resistant B-cell leukemias and lymphomas (80%–90% complete remissions in relapsed acute lymphoblastic leukemia) and the majority of clinical studies are performed using CARs containing either CD28 or 4-1BB cytoplasmic domains (3–8).

Broadening the applicability of CAR-T-cell therapy for various types of tumors remains a challenge because most of the available targets are tumor-associated antigens (TAA), which are not entirely tumor-restricted. In some cases, the expression of the target on healthy tissues can be tolerable and clinically manageable, such as the B-cell aplasia caused by CD19 CAR-T cells (4, 9), but in cases where vital tissues are involved off-tumor toxicity can be fatal (10–12). Fine-tuning the affinity of the CAR's binding domain can be a successful and easily applicable strategy to avert “on-target/off-tumor” reactivity of CAR-T cells. CARs of lower affinity targeting ERBB2/HER-2, EGFR, CD123, or CD38 showed better discrimination between tumors and normal tissues expressing the same antigen in lower levels (13–16). However, decreasing the CAR affinity results in a higher target expression-threshold for T-cell activation and, depending on the level of antigen expression on the tumor cells, it may also hamper the efficacy of antitumor function (17–19). Therefore, it is important to identify the conditions where the best discriminative potential between tumor and healthy tissues can be achieved, by using the lowest possible affinity, in combination with an optimal antitumor effect.

The influence of specific costimulatory moieties on the biology and therapeutic efficacy of CAR-T cells has been a subject of many recent studies. CAR-T cells bearing a 4-1BB costimulatory endodomain seem to persist for longer time in patient circulation in comparison with CAR-T cells having a CD28 costimulatory domain (3, 5, 20), due to better maintenance of a memory phenotype and reliance on oxidative metabolism (21–24). On the other hand, addition of a CD28 endodomain confers a more efficient and rapid cytotoxic ability to CAR-T cells (2, 23). Moreover, the combination of synergizing signaling pathways results in even greater CAR-T-cell potency, persistence, and antitumor response (25–27). Previous studies evaluating the functionality of CAR-T cells with lower affinities performed comparisons using the same costimulatory design. It is, therefore, largely unknown whether and how lowering the affinity for the target would affect the costimulatory requirements of CAR-T cells for optimal functionality and persistence.

Here, we hypothesized that the type of CAR-mediated costimulatory design is of prime importance for the optimal function of low-affinity CAR-T cells. To this end, we investigated a series of scFvs binding to the same CD38 epitope but possessing five different affinities (15). We incorporated these scFvs in three different CAR designs and evaluated the antitumor functionality, phenotype, and persistence of the generated CAR-T cells in vitro and in vivo. We demonstrate that decreasing the affinity of CARs can, depending on the CAR design, compromise the antitumor efficacy of CAR-T cells and that the combinatorial delivery of CD28 and 4-1BB signals potentiates lower affinity CARs and improves their immunotherapeutic potential in vitro and in vivo without increasing on-target/off-tumor toxicity.

Lower affinity CAR construction

Lower affinity CARs were produced with different germline variable light chains while keeping the variable heavy chain constant (clone 028). Selection and classification of lower affinity antibodies was described previously (15). The selected variable heavy and light chains separated by a G4S linker were PCR amplified with a proofreading Q5-Hotstart polymerase (NEB) using the AfeI containing forward primer 5′ctctgctgctgcctctagcgctgctgctg3′ and the NotI containing reverse primer 5′gttgtgcggccgcgctggacacggtgaccattg3′. PCR product was purified (Bioké) and cloned into SFG retroviral vector with a T4 ligase (Roche). The scFv was followed by a CD8a transmembrane domain and the 4-1BB and CD3ζ signaling domains or a CD28 transmembrane and intracellular sequence as described in Zhao and colleagues (23). The CAR sequences were linked by a P2A sequence (28) to a truncated LNGFR, dsRed, or 4-1BBL sequence.

Cloning of 4-1BBL

The 4-1BBL sequence was obtained form EBV-LCL cell line 10850, amplified by standard RT-PCR (Thermo Fisher Scientific) using a compatible primer pair. cDNA was used as a template to replace the dsRed, separated by a P2A from the CAR-CD28z. The forward primer including a RsrI restriction site 5′atcccggaccgatggaatacgcctctgacg3′ and reverse primers with a SalI restriction site 5′ccgtcgacctattattccgacctcggtgaag3′ were used to replace the dsRed sequence with the 4-1BBL. The 4-1BBL sequence is separated by the P2A sequences and is therefore expressed separate from the CAR-28z.

Generation of retroviral particles and transduction of T cells

Phoenix-Ampho packaging cells were calcium phosphate transfected with 10 μg CAR constructs. Sixteen hours posttransfection, complete medium (DMEM + 10% FBS) was refreshed, and 2–3 days after transfection, cell-free supernatants containing retroviral particles were collected and directly used for transduction.

Peripheral blood mononuclear cells (PBMC) from healthy donors (3 × 106/well) were stimulated with lectin-like phytohemagglutinin (PHA-L) in a 6-well plate (Greiner Bio-One) in culture medium (RPMI1640, 10% FBS, penicillin; 100 U/mL, streptomycin; 100 μg/mL). After 48 hours, 1 mL 3 × 106/mL of cells were transferred to retronectin-coated (15 μg/mL) (Takara) 6-well plates (Falcon). Retroviral transduction was performed by addition of 2 mL virus per well followed by spinoculation (1,500 × g, 1 hour at room temperature) in the presence of 4 μg/mL Polybrene. A second transduction was conducted after 16 hours, replacing 2/3 of the cell supernatant with freshly obtained virus (2 mL). Six to 8 hours after the second hit, half of the cell supernatant was replaced by fresh culture RPMI1640 + 10% FBS and 50 IE/mL rhIL2 (Proleukin, Novartis) was added once. Seventy-two hours posttransduction LNGFR, dsRed or 4-1BBL and CD38 expression were measured by flow cytometry to determine transduction efficiency.

Primary cells from patients with multiple myeloma and healthy individuals

Healthy donor peripheral blood mononuclear cells (PBMC) from buffy coats (Sanquin blood-bank) or bone marrow mononuclear cells (BM-MNCs) from multiple myeloma patient's bone marrow aspirates [∼10%–40% malignant cells, determined by flow cytometry (CD138+/CD38+)], were isolated through Ficoll-Paque (GE Healthcare Life Sciences) density centrifugation. Primary apheresis material was thawed and subsequently sorted by EasySep (StemCell Technologies) with CD34+ magnetic beads according to the manufacturer's protocol. Isolated cells were directly used in cytotoxicity assays or cryopreserved in liquid nitrogen until use. All primary samples were obtained after informed consent and approval by the institutional medical ethical committee.

Cell lines

The human multiple myeloma cell line, UM9 [unmodified or luciferase (Luc-GFP)-transduced], was cultured in RPMI1640 (Thermo Fisher Scientific) + 10% FBS (Invitrogen) + antibiotics (penicillin; 100 U/mL, streptomycin; 100 μg/mL). The mouse fibroblast cell line NIH/3T3 cell line was obtained from ATCC and transduced with a lentivirus to express human CD38. NIH/3T3 (modified or CD38-transduced) cells and Phoenix Ampho cells were cultured in DMEM (Thermo Fisher Scientific) + GlutaMAX.10% FBS (Invitrogen) and penicillin (100 U/mL) and streptomycin (100 μg/mL).

Flow cytometry

Flow cytometry assays were performed on BD LSRFortessa. Viable cells were determined with live/dead cell marker (LIVE/DEAD Fixable Near-IR; Life Technologies L10119). Transduction efficiency and associated CAR expression was measured with an APC-conjugated antibody toward NGFR (CD271; BioLegend) for CAR-4-1BBz-LNGFR and APC antibody 4-1BBL (CD137L; BioLegend) for CAR-28z-41BBL. CAR-28z-dsRed were measured in the PE-CF594 channel to detect dsRed. Additional antibodies were used for weekly differentiation phenotype: CD3, CD4, CD8, CD38, (BD Biosciences), CD45RA and CD62L (BioLegend), for exhaustion assays antibodies: PD-1, LAG-3, and TIM-3 (BioLegend) and for cytotoxicity assays: CD3, CD14, CD19, CD38, CD56, and CD138 (BD Biosciences). To distinguish Mock/CAR-T cells from target cells, target cell were stained with 0.5 μmol/L Violet tracer (Thermo Fisher Scientific) for 25 minutes and washed before cytotoxicity assay cocultures. Flow cytometry data analysis was performed with FACS Diva 6.1 software.

Proliferation assays

CAR-T cells were counted and stimulated weekly with irradiated (50 Gy) CD38+ UM9 cells or (80 Gy) 3T3-CD38. Starting 7 days posttransduction, 1 × 106 CAR+ T cells were seeded in a 24-well plate containing 3 × 105 UM9 or 3T3-CD38 cells, to a volume of 1–1.5 mL. No additional cytokines were added and when a cell count of approximately 2 × 106 cells/mL was exceeded the cell culture was split to a density of 1 × 106 per well.

Flow cytometry–based cytotoxicity assay

Seven to 10 days after transduction, serial dilutions (effector:target 3:1, 1:1, or 1:3) of CAR-T cells were incubated with Violet tracer (Thermo Fisher Scientific) labeled BM-MNCs or PBMCs for 16–24 hours. After addition of Flow-Count Fluorospheres (Beckman 7547053), cells were harvested and stained for different CD markers (see section flow cytometry) to distinguish different subsets. Viable cells were then quantitatively analyzed through Flow-Count–equalized measurements. Percentage cell lysis was calculated as followed and only if the analyzed target cell population contained >500 viable cells in the untreated samples. % lysis cells = 1 − ((# viable target cells in treated wells/# of beads)/(#viable target cells in untreated wells/# of beads)) × 100%.

Bioluminescent imaging–based cytotoxicity assay

Seven to 10 days after transduction, serial dilutions (effector:target 10:1, 3:1, 1:1, 1:3, or 1:10) of CAR-T cells were incubated with Luc-GFP–transduced human multiple myeloma cell line UM9. The luciferase signal produced by surviving UM9 cells was determined after 16–24 hours with a GloMax 96 Microplate Luminometer (Promega) within 15 minutes after the addition of 125 μg/mL beetle luciferin (Promega). % lysis cells = 1 − (BLI signal in treated wells/BLI signal in untreated wells) × 100%.

Cytokine measurements

To determine cytokine production by CAR-T cells, cell supernatants were harvested 24 hours after coculture with target cells. To measure cytokines, we used Cytokine Bead Array (CBA) Human Th1/Th2/Th17 cytokine Kit (BD Biosciences) according to the manufacturer's protocol. In brief, a mixture of capture beads (IL2, IL4, IL6, IL10, IL17A, TNF, and IFNγ), PE-detection reagent, and cell supernatant were incubated for 3 hours. Beads were washed and analyzed by a BD standardized flow cytometry assay.

In vivo xenograft studies

H2d-Rag2−/−γc−/− female mice used in this study were bred and maintained at the Amsterdam Animal Research Center. The animal experiments were performed under the approval of the central authority for scientific procedures on animals (protocol number AVD114002015345). We used an in vivo model, in which a humanized bone marrow–like environment is created in mice to allow the growth of human multiple myeloma tumors or normal CD34+ cells in their natural niche. Briefly, hybrid scaffolds consisting of three 2- to 3-mm3 triphasic calcium phosphate particles were coated in vitro with human bone marrow mesenchymal stromal cells (BM-MSC; 2 × 105 cells/scaffold). The scaffolds were implanted subcutaneously into the mice (29). Eight to 12 weeks after implantation, for the antitumor model, 1 × 107 luciferase-transduced multiple myeloma cells (UM9) were injected intravenously or in the separate off-tumor experiment, 1 × 106 fluorescent (FarRed) labeled healthy CD34+ cells were injected transcutaneously into the scaffold. After 1 week, when the tumor or CD34+ cells became detectable by bioluminescence imaging (BLI) or fluorescence live imaging (FLI) respectively, mice were divided in equal groups. Mice received CD38-CAR-BBz, 28z, 28z-BBL, or mock-transduced T cells (5 × 106 cells/mice), by intravenous injection in the tail vein. Tumor growth or CD34+ cell persistence was monitored by weekly BLI/FLI measurements. Postmortem, bone marrow, spleen, and scaffolds were harvested from each mouse, bone marrow was flushed, and spleen and scaffolds were dissociated. Flushed or dissociated tissues were filtered through a 70-μm filter and single-cell suspensions were counted, stained, and measured by flow cytometry.

Hematopoietic progenitor cell growth inhibition assay

A total of 2,000 CD34+ EasySep sorted (StemCell Technologies) cells from multiple myeloma patient apheresis material were mixed with effector CD38-CAR-T cells at a CART:BM cell ratio of 1:1 in 0.2 mL of RPMI + FBS culture medium. After culturing for 4 hours in this small volume, the cells were resuspended to a final volume of 2 mL with semisolid Methocult (StemCell Technologies, H4534) and then plated in 6-cm dishes and incubated at 37°C in 5% CO2. Between 14 and 21 days, the number of colony-forming unit-granulocytes (CFU-G), and CFU-monocytes (CFU-M), were scored under a microscope.

Statistical analysis

Statistical analyses were performed using GraphPad Prism software version 7.0. For normal distributions, parametric Student t tests were used. In analyses where multiple groups were compared, either a parametric ANOVA with Bonferroni post hoc test or nonparametric Kruskal–Wallis test were used with subsequent multiple comparison. Two-tailed statistical tests were always used. A P value <0.05 was considered significant.

CD38-CAR constructs combining different affinities and costimulation design

We used CD38 as a model for a TAA because it is highly and uniformly expressed on multiple myeloma cells but is also present at lower levels on subsets of healthy hematopoietic cells. We generated 15 different CAR constructs through the combination of five scFv domains of variable affinity targeting the same CD38 epitope with three described CAR structural designs (Fig. 1A). Affinity was ranked from high to very low as depicted in Fig. 1B. The scFvs were cloned into vectors encoding the components of the two most popular second-generation CAR structures providing the 4-1BB (BBz) or the CD28 (28z) modalities (Fig. 1A). To deliver both 4-1BB and CD28 signaling, we chose to use a design recently shown to provide the optimal combination of these two signals, where a second-generation CD28z CAR is coexpressed with 4-1BBL (28zBBL; Fig. 1A; refs. 23, 30). The CAR transgenes were linked to functionally irrelevant markers whose expression levels was used as surrogate marker of CAR expression on transduced T cells, as previously validated for these constructs (16). All CAR constructs were well expressed upon transduction on human T lymphocytes without significant difference in expression levels (Fig. 1C and D; Supplementary Fig. S1A and S1B). In contrast to high-affinity CAR028 cells, lower affinity CD38-CAR-T cells showed no significant decrease of CD38 expression compared with mock-transduced T cells, indicating reduced CD38-directed cytotoxicity against targets with intermediate CD38 expression (Fig. 1E; Supplementary Fig. S1C). After acquiring a decreased CD38 expression on the cell surface, high-affinity CD38-CAR-T cells were eventually able to expand (see also Drent and colleagues, Haematologica). Therefore, CAR-T cells were used at least after one week of culture so that fratricide would not bias the absolute number of CAR-T cells.

Figure 1.

CAR constructs and expression. A, Left, a graphic overview of CAR construct design. The scFv light (VL) and heavy (VH) chains are followed by: (1) the CD8a transmembrane domain, the 4-1BB and CD3ζ signaling domains (BBz) linked by a P2A sequence (28) to a truncated LNGFR or (2) the CD28 transmembrane and intracellular sequence and CD3ζ signaling domain (28z) linked to dsRed fluorescent marker or (3) the 28z construct linked to the 4-1BBL coding sequence. B, Summary of the characteristics of scFvs and the parental antibodies, the surface plasmon resonance–determined Kd value (nmol/L) and half-effective concentration (EC50) when titrated on CHO-CD38 cells (μg/mL), described in ref. 15. C, Flow cytometry histogram plots of expression of LNGFR, dsRed, and 4-1BBL after transduction of T cells with BBz, 28z, or 28z-BBL. D, Mean fluorescent intensity (MFI) of marker expression. E, % CD38+ CAR-T cells when transduced with different affinity CD38CARs. NA, not available; *, P <0.05; **, P < 0.01 compared with mock, using standard Student t test analysis.

Figure 1.

CAR constructs and expression. A, Left, a graphic overview of CAR construct design. The scFv light (VL) and heavy (VH) chains are followed by: (1) the CD8a transmembrane domain, the 4-1BB and CD3ζ signaling domains (BBz) linked by a P2A sequence (28) to a truncated LNGFR or (2) the CD28 transmembrane and intracellular sequence and CD3ζ signaling domain (28z) linked to dsRed fluorescent marker or (3) the 28z construct linked to the 4-1BBL coding sequence. B, Summary of the characteristics of scFvs and the parental antibodies, the surface plasmon resonance–determined Kd value (nmol/L) and half-effective concentration (EC50) when titrated on CHO-CD38 cells (μg/mL), described in ref. 15. C, Flow cytometry histogram plots of expression of LNGFR, dsRed, and 4-1BBL after transduction of T cells with BBz, 28z, or 28z-BBL. D, Mean fluorescent intensity (MFI) of marker expression. E, % CD38+ CAR-T cells when transduced with different affinity CD38CARs. NA, not available; *, P <0.05; **, P < 0.01 compared with mock, using standard Student t test analysis.

Close modal

CD28 costimulation lowers the affinity threshold for efficient cytotoxicity and cytokine production

To elucidate the potential functional aberrancies caused by lowering the affinity of CD38-CAR-T cells in relation to their costimulatory design, we first studied their lytic capacity against an multiple myeloma cell line with CD38 expression similar to primary multiple myeloma cells (>20,000 molecules/cell; Supplementary Fig. S2). For CD38-CARs carrying the 4-1BB costimulatory domain, lowering the affinity further than Kd = 1.9 μmol/L (CARB1 and CARB3) substantially diminished antitumor cytotoxicity (Fig. 2A). In striking contrast, when a CD28 domain was used in the CAR design, alone or in combination with 4-1BBL, the cytotoxic potential of CAR-T cells was not impacted at all by affinity change. Even CARB1 and CARB3 cells, with >1,000-fold lower affinity for CD38 as compared with the high-affinity CAR028, displayed no significant decrease in cytotoxicity (Fig. 2A).

Figure 2.

CD28 signaling improves lytic capacity and cytokine production of affinity-tuned CD38-CAR-T cells. A, Functional experiments were performed 7 to 10 days after transduction. The series of high- and low-affinity CAR-T cells were incubated with Firefly-Luciferase–transduced human multiple myeloma cell line UM9 (n = 3 for each condition). The luciferase signal produced by surviving UM9 cells was determined after 16–24 hours within 15 minutes after the addition of 125 μg/mL beetle luciferin. % lysis cells = 1 − (BLI signal in treated wells/BLI signal in untreated wells) × 100%. B, Twenty-four hours after coincubation with UM9 (E:T ratio 1:1), cell supernatants were harvested to measure cytokine secretion with a flow cytometry–based assay. Graph shows the secretion of IFNγ, TNF, and IL2. C, Comparison of cytokine production of CARB1 with different costimulatory designs. Mean values (± SEM) are shown (n = 3 each condition). *, P < 0.05; **, P < 0.01; ***, P <0.001 using one-way ANOVA and subsequent multiple comparison; ns, nonsignificant.

Figure 2.

CD28 signaling improves lytic capacity and cytokine production of affinity-tuned CD38-CAR-T cells. A, Functional experiments were performed 7 to 10 days after transduction. The series of high- and low-affinity CAR-T cells were incubated with Firefly-Luciferase–transduced human multiple myeloma cell line UM9 (n = 3 for each condition). The luciferase signal produced by surviving UM9 cells was determined after 16–24 hours within 15 minutes after the addition of 125 μg/mL beetle luciferin. % lysis cells = 1 − (BLI signal in treated wells/BLI signal in untreated wells) × 100%. B, Twenty-four hours after coincubation with UM9 (E:T ratio 1:1), cell supernatants were harvested to measure cytokine secretion with a flow cytometry–based assay. Graph shows the secretion of IFNγ, TNF, and IL2. C, Comparison of cytokine production of CARB1 with different costimulatory designs. Mean values (± SEM) are shown (n = 3 each condition). *, P < 0.05; **, P < 0.01; ***, P <0.001 using one-way ANOVA and subsequent multiple comparison; ns, nonsignificant.

Close modal

We further analyzed the impact of an affinity decrease on a later effector function, such as the cytokine secretion. All CAR constructs in our study showed no significant unspecific cytokine production against a CD38 multiple myeloma cell line (Supplementary Fig. S3). A remarkable decrease of all cytokine production was observed for BBz CARs bearing very low–affinity scFvs (Fig. 2B). Although 28z and 28z-BBL CAR-T cells also showed an affinity-dependent decrease in cytokine secretion, this effect was not significant or significant only for CARB3 (Fig. 2B). Specifically for CARB1, 28z and 28z-BBL designs maintained approximately 4 times higher TNFα and up to 15 times higher IL2 levels than BBz design (Fig. 2C). Therefore, the affinity threshold for effective antitumor cytotoxicity and efficient cytokine production seems to be lower for CARs that incorporate a CD28 intracellular domain.

Because cytotoxic function and cytokine secretion are the consequence of antigen engagement and CAR-T-cell activation, we investigated whether the decrease of affinity would lead to a reduced intensity of downstream signaling in BBz compared with 28z CD38CARs. However, we found no impact of affinity on the phosphorylation of Zap70, the main mediator of signaling downstream of CD3ζ, on the BBz or the 28zBBL CARs at rest and 24 hours after antigen stimulation (Supplementary Fig. S4A and S4B). For 28z CAR designs, we observed a slight decrease of pZap70 when decreasing the affinity, which was statistically significant only for the very low–affinity CARs B1 and B3 (Supplementary Fig. S4B). Therefore, the reduction of the cytotoxic potential of lower affinity BBz CD38CARs could not be explained by the intensity of the downstream CD3ζ-mediated CAR signaling in our system.

Combination of CD28 and 4-1BB signaling improves in vitro proliferative capacity of low-affinity CAR-T cells

We, further, weekly stimulated the CD38CAR-T cells in vitro with irradiated NIH-3T3 cells expressing high levels of CD38 (Supplementary Fig. S2A). In this system, activation and costimulation signals depend solely on the CAR binding to CD38. In all affinity groups, BBz CARs showed a better proliferation potential in comparison with 28z CARs confirming previous reports (22–24), although the difference was not statistically significant (Fig. 3A). Notably, 28z-BBL CARs showed a stable growth response to repetitive antigen stimulations, which was always consistently higher, compared with the growth rates of both BBz and 28z CARs (Fig. 3A). Reduction of this stable proliferative response was only seen with the lowest affinity CARB3 (Fig. 3A). Hence, CARB1 sets the threshold below which CD38 CARs of all designs showed a reduced proliferative capacity and most importantly, CD28 and 4-1BB signaling synergize for an optimal and persistent proliferative response irrespective of CAR affinity.

Figure 3.

Combination of CD28 and 4-1BB signaling improves in vitro proliferation and delays differentiation of affinity-tuned CD38-CAR-T cells. A, Starting one week after transduction, CD38-CAR-T cells were cocultured with mouse fibroblast cells NIH-3T3 transduced with human CD38 at E:T ratio 3:1, and restimulated weekly. No cytokines were added to culture. Cells were counted and corrected for % of CAR+ cells as determined by flow cytometry. B, Flow cytometry density plots of phenotypic profile of each CD38-CAR affinity and costimulation type cell type at week 0, before expansion. Cells have either a naïve (N; CD45RA+/CD62L+) central memory (CM; CD45RA/CD62L+), effector memory (EM; CD45RA/CD62L), or effector phenotype (CD45RA+/CD62L). C, Percentage of CAR+ cells that have a CM phenotype. Graphs depict mean ± SEM (n = 3 for each condition). Statistical analysis was done using one-way ANOVA and subsequent multiple comparison. * indicates P <0.05, between BBz and 28z.

Figure 3.

Combination of CD28 and 4-1BB signaling improves in vitro proliferation and delays differentiation of affinity-tuned CD38-CAR-T cells. A, Starting one week after transduction, CD38-CAR-T cells were cocultured with mouse fibroblast cells NIH-3T3 transduced with human CD38 at E:T ratio 3:1, and restimulated weekly. No cytokines were added to culture. Cells were counted and corrected for % of CAR+ cells as determined by flow cytometry. B, Flow cytometry density plots of phenotypic profile of each CD38-CAR affinity and costimulation type cell type at week 0, before expansion. Cells have either a naïve (N; CD45RA+/CD62L+) central memory (CM; CD45RA/CD62L+), effector memory (EM; CD45RA/CD62L), or effector phenotype (CD45RA+/CD62L). C, Percentage of CAR+ cells that have a CM phenotype. Graphs depict mean ± SEM (n = 3 for each condition). Statistical analysis was done using one-way ANOVA and subsequent multiple comparison. * indicates P <0.05, between BBz and 28z.

Close modal

4-1BB signaling endows low-affinity CAR-T cells with a less exhausted memory phenotype

We further analyzed the differentiation status and the exhaustion level of affinity-tuned CD38-CAR-T cells during expansion on 3T3-CD38 cells. At the end of production (week 0) CD38-CAR-T cells equipped with 4-1BB signaling, either BBz or 28z-BBL, showed a higher percentage of central memory (TCM) cells compared with 28z CAR-T cells, which was significant for low and very low–affinity groups (Fig. 3B and C). The same pattern was observed even after 2 weeks of expansion (Fig. 3C). We then analyzed the expression pattern of inhibitory receptors PD-1, TIM-3, and LAG-3 (Fig. 4A) and determined the percentage of cells expressing none, one, two, or three of the markers (Fig. 4B). Overall, compared with 28z CD38-CAR-T cells, BBz and 28z-BBL CD38-CAR-T cells showed a lower percentage of triple positive cells (expressing all three markers) for the higher affinity CARs and higher percentage of triple negative cells (expressing none of the three markers) in all affinity groups (Fig. 4B; Supplementary Fig. S5). Thus, including 4-1BB signaling moieties in the CAR-T-cell design results in longer preservation of a central memory phenotype and a delay of the induction of immune-inhibitory receptors on the cell surface. The effect of 4-1BB costimulation in ameliorating exhaustion of CAR-T cells has been previously attributed to modulation of the CAR signaling (22). We found that indeed BBz and especially 28zBBL CD38CARs showed less phosphorylation of ZAP70 than CARB1-28z at a basal level as well as after antigenic stimulation across all affinity levels (Supplementary Fig. S4C and S4D), indicating a possible role of the 4-1BB pathway in regulating the CAR signaling strength.

Figure 4.

Exhaustion of affinity-tuned CD38-CAR-T cells. A, Flow cytometry density plots illustrating expression of inhibitory receptors PD-1, LAG3, and TIM3, 2 weeks after stimulation with 3T3-CD38. Representative figure of three independent experiments. B, Pie charts illustrating the % of cells expressing either 0, 1 (PD-1+, LAG3+ or TIM3+), 2 (PD-1+/LAG3+ or PD-1+/TIM3+ or LAG3+/TIM3+), or 3 (PD-1+, LAG3+, and TIM3+) exhaustion markers gated on live CD3+CAR+ cells. Mean values of n = 3 are shown.

Figure 4.

Exhaustion of affinity-tuned CD38-CAR-T cells. A, Flow cytometry density plots illustrating expression of inhibitory receptors PD-1, LAG3, and TIM3, 2 weeks after stimulation with 3T3-CD38. Representative figure of three independent experiments. B, Pie charts illustrating the % of cells expressing either 0, 1 (PD-1+, LAG3+ or TIM3+), 2 (PD-1+/LAG3+ or PD-1+/TIM3+ or LAG3+/TIM3+), or 3 (PD-1+, LAG3+, and TIM3+) exhaustion markers gated on live CD3+CAR+ cells. Mean values of n = 3 are shown.

Close modal

Combined costimulation better supports very low–affinity CD38 CAR-T cells to control tumor growth in vivo

CARB1 was the most optimal low-affinity CAR used in our in vitro studies, showing an effective cytotoxic capacity similar to high-affinity CAR028 when coupled with a CD28 domain and the best proliferative response when including a 4-1BB-signaling moiety. Thus, we further investigated the anti-multiple myeloma effects of CARB1 with either a BBz, 28z, or 28zBBL design using a previously described xenograft murine model, in which tumor cells are grown in humanized bone marrow–like niches (Fig. 5A; refs. 29, 31). In this model, multiple myeloma tumor cells preferentially populate the humanized bone marrow scaffolds as revealed by BLI 1 week after injection (Supplementary Fig. S6, week 1). Treatment with CARB1-28z or CARB1-28zBBL T cells resulted in a significant delay of tumor progression compared with mock-treated mice while the CARB1-BBz cells failed to control tumor growth (Fig. 5B–D; Supplementary Fig. S6). Postmortem analysis of the scaffold material, 7 weeks after T-cell injection, showed significantly lower numbers of (GFP+/CD38+/CD138+) tumor cells in all CAR-treated groups compared with mock-treated group (Fig. 5E). The CARB1-28zBBL group showed the lowest median of tumor cell numbers in the scaffolds compared with both CARB1-BBz and CARB1-28z. When looking specifically at CAR-T-cell numbers, CARB1-28z cells initially expanded, reached higher effector:tumor (E:T) ratio than CARB1-BBz T cells and rapidly reduced tumor burden at 3 weeks after injection (Supplementary Fig. S7), but they did not persist as their numbers were significantly reduced after 7 weeks (Fig. 5F). On the other hand, BBz CAR-T cells showed a more delayed pattern of expansion at week 3 (Supplementary Fig. S7), but they persisted for longer and eventually achieved control of tumor growth in the scaffolds (Fig. 5E–G). These data indicate that CARB1-BBz T cells failed to control tumor spread to secondary sites (e.g., murine bone marrow, skull; Supplementary Fig. S6). Indeed, although CARB1-BBz T cells infiltrated the murine bone marrow and persisted longer than CARB1-28z cells, they failed to restrict the local tumor cell growth (Supplementary Fig. S8). Importantly, CARB1-28zBBL T cells displayed the best expansion and persistence features compared with the other groups and eventually achieved the best antitumor effect as they reached the highest E:T ratios within the scaffolds and murine bone marrow at both an early (3 weeks) and a later (7 weeks) time point (Fig. 5G; Supplementary Fig. S7). Previously, using a BBz CAR design, we found that CARA4-BBz could elicit a significant anti-multiple myeloma effect in vivo (15). Interestingly, when displaying data from both experiments (Fig. 5H), we found that the anti-multiple myeloma activity of CD38-CAR-T cells is sensitive to affinity reduction from A4 to B1 level when a BBz costimulatory design is used. Most notably, providing both CD28 and 4-1BB signaling can potentiate even very low–affinity CARB1-T cells to cause significant reduction of multiple myeloma tumor growth, comparable with that obtained after treatment with a high-affinity CAR028-BBz CAR (Fig. 5H).

Figure 5.

Very low–affinity CD38-CAR-T cells with 28z and 4-1BBL costimulation show better tumor control and persistence in vivo. A, Schematic of in vivo experimental set-up. Scaffolds consisting of three 2- to 3-mm3 triphasic calcium phosphate particles coated in vitro with human mesenchymal stromal cells were implanted subcutaneously in H2d-Rag2−/−γc−/− mice. Eight to 12 weeks later, mice were intravenously injected with 1 × 107 cells of luciferase-transduced UM9 cells. One week after tumor injection, mice were treated with intravenous injections of 5 × 106 mock, CD38-CARB1 BBz, -28z, or -28z-BBL T cells. B, Representative bioluminescence images (BLI) are shown n = 4 per group for week 1 and 7 (all BLI images in Supplementary Fig. S4). C, Quantification of BLI measurements, normalized to week 1. Each group contained 4 mice and each mouse harbored 4 scaffolds (n = 4). Indicated are the median values of normalized tumor load per group ± range; *, P <0.05; **, P <0.01 using Kruskal–Wallis ANOVA. D, Kaplan–Meier curve based on the set up of a maximum BLI measurement. In our model, BLI measurement reaches a plateau after between 8 weeks without the tumor being lethal for the mice. Therefore, we have chosen week 7 as the end point of our experiments. This plateau roughly corresponds with 500 cpm/m2. Thus, we considered this BLI value as the potential human endpoint of the experiment to obtain a Kaplan–Meier curve. E–G, Postmortem analysis of scaffolds harboring UM9 tumor cells and infiltrated T cells. Scaffolds were dissociated and filtered through a 70-μm filter. Single cells were stained for mouse and human CD45, counted and analyzed by flow cytometry. E, Absolute numbers of multiple cells (GFP+/CD138+). F, Number of CAR-T cells (CD45+/CD3+ and LNGFR+ or dsRed+ or 4-1BBL+). Mock <103 T cells. G, Ratio between CAR-T cells and MM cells. N = 4, results are median cell number of 16 scaffolds (4 per mouse) ± range. H, Because tumor growth was similar for the mock-treated group between experiments and tumor load between treatment groups was equal before CAR-T-cell injection, we normalized BLI measurements to mock control values and compared data with previous experiment (15). Displayed is the median of relative tumor load (quantified BLI measurements) at week 7, relative to mock (set at 100%), n = 4 mice. *, P <0.05; **, P <0.01; ***, P <0.001 using Kruskal–Wallis ANOVA.

Figure 5.

Very low–affinity CD38-CAR-T cells with 28z and 4-1BBL costimulation show better tumor control and persistence in vivo. A, Schematic of in vivo experimental set-up. Scaffolds consisting of three 2- to 3-mm3 triphasic calcium phosphate particles coated in vitro with human mesenchymal stromal cells were implanted subcutaneously in H2d-Rag2−/−γc−/− mice. Eight to 12 weeks later, mice were intravenously injected with 1 × 107 cells of luciferase-transduced UM9 cells. One week after tumor injection, mice were treated with intravenous injections of 5 × 106 mock, CD38-CARB1 BBz, -28z, or -28z-BBL T cells. B, Representative bioluminescence images (BLI) are shown n = 4 per group for week 1 and 7 (all BLI images in Supplementary Fig. S4). C, Quantification of BLI measurements, normalized to week 1. Each group contained 4 mice and each mouse harbored 4 scaffolds (n = 4). Indicated are the median values of normalized tumor load per group ± range; *, P <0.05; **, P <0.01 using Kruskal–Wallis ANOVA. D, Kaplan–Meier curve based on the set up of a maximum BLI measurement. In our model, BLI measurement reaches a plateau after between 8 weeks without the tumor being lethal for the mice. Therefore, we have chosen week 7 as the end point of our experiments. This plateau roughly corresponds with 500 cpm/m2. Thus, we considered this BLI value as the potential human endpoint of the experiment to obtain a Kaplan–Meier curve. E–G, Postmortem analysis of scaffolds harboring UM9 tumor cells and infiltrated T cells. Scaffolds were dissociated and filtered through a 70-μm filter. Single cells were stained for mouse and human CD45, counted and analyzed by flow cytometry. E, Absolute numbers of multiple cells (GFP+/CD138+). F, Number of CAR-T cells (CD45+/CD3+ and LNGFR+ or dsRed+ or 4-1BBL+). Mock <103 T cells. G, Ratio between CAR-T cells and MM cells. N = 4, results are median cell number of 16 scaffolds (4 per mouse) ± range. H, Because tumor growth was similar for the mock-treated group between experiments and tumor load between treatment groups was equal before CAR-T-cell injection, we normalized BLI measurements to mock control values and compared data with previous experiment (15). Displayed is the median of relative tumor load (quantified BLI measurements) at week 7, relative to mock (set at 100%), n = 4 mice. *, P <0.05; **, P <0.01; ***, P <0.001 using Kruskal–Wallis ANOVA.

Close modal

Combined costimulation does not increase off-tumor toxicity of low-affinity CAR-T cells

Because very low–affinity CARB1-28zBBL T cells elicited a similar anti-multiple myeloma response to that of high-affinity CAR028-28zBBL in vitro, we further investigated, in a whole bone marrow cytotoxicity assay, if that would be at the expense of increased on-target/off-tumor cytotoxicity against normal hematopoietic cells, which are known to express intermediate-to-low levels of CD38 (200–3,000 molecules/cell; Supplementary Fig. S2). As observed when using UM9 cells, there was a significant decrease in anti-multiple myeloma cytotoxicity for CARB1- and CARB3-BBz T cells compared with CAR028 (Supplementary Fig. S9A). When the identical scFvs were coupled to the 28z costimulatory domain (± 4-1BBL), the overall anti-multiple myeloma cytotoxicity was improved and was not significantly different from that of CAR028 (Supplementary Fig. S9A). CARA4-, CARB1-, and CARB3-BBz T cells resulted in a very limited, nonsignificant lysis of healthy mononuclear cells (MNC; Supplementary Figs. S9B and S10) expressing 200–3,000 CD38 molecules/cell (Supplementary Fig. S2). Importantly, even in a 28z or 28zBBL format, the very low–affinity CARB1- and CARB3-T cells did not elicit significant MNC lysis despite their effective anti-multiple myeloma cytotoxic response (Supplementary Figs. S9B and S10).

We next evaluated the on-target/off-tumor toxicity of CARB1 T cells against human FarRed-labeled CD34+ normal hematopoietic progenitor cells in vivo, in a modified version of our murine xenograft model (ref. 15; Fig. 6A). The FarRed-labeled cells were still detected 14 days after CAR-T-cell injection in all treatment groups. Thus, in our system treatment with either high- or low-affinity CD38-CARs leaves CD34+/CD38 cells intact and does not inhibit overall hematopoiesis (Fig. 6A and B). This was also demonstrated in a colony-forming assay where colony-forming capacity of CD34+ cells was not affected whether the CD34+/CD38+ cell subset was eliminated by CAR028-28zBBL cells or left intact by CARA4-BBz and CARB1-28zBBL cells (Supplementary Fig. S11A–S11D). Postmortem analysis revealed that, low affinity CARB1-28zBBL T cells caused no significant decrease of the percentage and absolute numbers of CD34+/CD38+ progenitors (2.680 CD38 molecules/cell) and total CD38+ cells (Fig. 6C and D; Supplementary Fig. S11E) compared with mock-treated controls, in contrast to high-affinity CAR028-28zBBL T cells and similar to CARA4-BBz cells, which were previously shown to have reduced off-tumor toxicity (15). Therefore, we conclude that although equipping very low–affinity CD38-CARs with combined CD28 and 4-1BB costimulatory signaling moieties can significantly improve their antitumor cytotoxic function, this does not induce off-tumor cytotoxicity of healthy cells expressing lower levels of the target antigen.

Figure 6.

Very low–affinity CD38-CAR-T cells with 28zBBL design do not lyse healthy CD38+ hematopoietic cells in vivo. A, Mice were injected intrascaffold with 1 × 106 fluorescently (FarRed) labeled CD34+ hematopoietic progenitors and treated 6 days later with intravenous injections of 5 × 106 mock, high-affinity CAR028-28z-BBL, or low-affinity CARA4-BBz or CARB1-28z-BBL T cells. B, Fluorescence images (FLI) are shown per group at week 1, 2, and 3. Percentages of cells expressing CD34 and/or CD38 (C) and percentage of total CD38+ cells in the total population of FarRed+ cells within the scaffolds (D). Median values (± range) of two scaffolds per mouse and n = 4 mice per group (*, P < 0.05, using Mann–Whitney test).

Figure 6.

Very low–affinity CD38-CAR-T cells with 28zBBL design do not lyse healthy CD38+ hematopoietic cells in vivo. A, Mice were injected intrascaffold with 1 × 106 fluorescently (FarRed) labeled CD34+ hematopoietic progenitors and treated 6 days later with intravenous injections of 5 × 106 mock, high-affinity CAR028-28z-BBL, or low-affinity CARA4-BBz or CARB1-28z-BBL T cells. B, Fluorescence images (FLI) are shown per group at week 1, 2, and 3. Percentages of cells expressing CD34 and/or CD38 (C) and percentage of total CD38+ cells in the total population of FarRed+ cells within the scaffolds (D). Median values (± range) of two scaffolds per mouse and n = 4 mice per group (*, P < 0.05, using Mann–Whitney test).

Close modal

The applicability of CAR-T-cell therapy beyond B-cell malignancies is, in part, impeded by safety concerns about the on-target/off-tumor effect when targeting TAAs. Adjusting the affinity of CAR-T cells to target antigen density is one popular strategy to avert “on-target/off-tumor” toxicity and confer specificity to the tumor (13–15, 32). For a given affinity, CAR-T-cell activation and cytotoxic response is relative to the level of expression of the target and lowering the affinity increases the target-expression threshold for effective CAR-T-cell activation (16–19, 33). This effect favors the ability of CAR-T cells to distinguish high and low target antigen expression but on the other hand may impede the antitumor effectiveness, depending on the level of target expression on tumor cells. However, CAR-T-cell activation and function is a multifactorial process, which is affected not only by affinity but also by costimulation. In this study, we assessed the antitumor function of CAR-T cells of different affinities and costimulatory designs and evaluated whether efficient tumor-selectivity achieved by a low-affinity CAR could be combined with optimal immunotherapeutic properties delivered by a specific costimulatory design.

CD38 was used as a paradigm of tumor-associated antigen (TAA) and a panel of five CD38-targeting scFv domains spanning a wide range of affinities for the target was used (15). Importantly, our scFv panel not only covered high- and low-affinity levels similar to the ones used in other studies (refs. 14, 18, 32, 33; Kd ranging between 1.8 × 10−6–1.9 × 10−9 mol/L), but also included scFvs from antibodies with unmeasurable Kd, whose binding to the target was only measurable in cell-binding assays. It is well demonstrated that costimulatory moieties derived from CD28 or 4-1BB or a combination of the two provide different biological and metabolic characteristics to CAR-T cells leading to differences in antitumor lytic capacity, differentiation, and persistence (2, 23, 24, 34). The CD38-scFv domains were coupled to CD28 or 4-1BB intracellular signaling domains to construct the most commonly used second-generation CARs. In addition, based on a study by Zhao and colleagues (23), we coexpressed a 28z-based second-generation CAR with 4-1BBL as an effective conformation to provide CAR-T cells with both CD28 and 4-1BB signaling.

Our data revealed that the different cell programming, mediated by costimulation, affects also the efficiency of both early and later CAR-T-cell responses when scFvs of lower affinity are used. When using affinities in the micromolar range (Kd < 1.9 × 10−6 mol/L), 4-1BB–based CD38 CAR-T cells began to lose their capacity for anti-multiple myeloma lysis and cytokine production (such as IL2 and TNFα). Previous studies evaluating targeting of ERBB2 and EGFRvIII-positive tumors using low-affinity CARs with a 4-1BB intracellular domain reported no significant loss of cytotoxicity or cytokine production against tumor cells. However, Kd values of 1.1 × 10−9 and 1.01 × 10−7 mol/L for ERBB2 and EGFRvIII, respectively, were the lowest affinities tested in these studies (14, 32). Interestingly, we found that inclusion of CD28 signaling reduced the affinity threshold for efficient activation after antigen encounter and rescued cytotoxicity and cytokine production in very low–affinity CD38-CAR-T cells. This indicated that 4-1BB–based CAR designs are more sensitive to lowering the affinity to the antigen than 28z-based CARs. However, the decreased cytotoxicity of very low–affinity BBz CD38 CAR-T cells could not be attributed to a weaker intracellular CAR signaling because we observed no significant difference in ZAP70 phosphorylation between BBz CARs of all affinities before or after antigenic stimulation and there was no difference of pZAP70 between BBz and 28zBBL CARs. CD28 can lower the threshold for effective TCR activation and can enhance responses in cases of low antigen availability (low avidity; refs. 35, 36) but its effect in cases of lower TCR affinity is not known. Our results are in contrast with Chmielewski and colleagues (33), reporting an affinity threshold of 10−8, same affinity level as CARA1, below which CD28 signaling did not improve activation of ErbB2-CAR-T cells. This discrepancy could be explained by the different dissociation rate described in Chmielewski and colleagues (33), which was 10-fold lower than that of our low-affinity CARs. It seems that affinity fine-tuning by keeping a low dissociation rate results in longer interaction of the CAR with the antigen and ensures efficient and potent CAR activation.

Irrespective to the CAR affinity, our data confirmed previous studies showing different biological properties between CAR-T cells bearing CD28 or 4-1BB costimulatory signaling domains (22–24, 27). For all different affinities in our study, 4-1BB–based CAR-T cells showed less rapid differentiation, less exhaustion, and better proliferative capacity in vitro than CD28-based CAR-T cells. We found that both BBz and 28z CD38-CAR-T cells restricted similarly multiple myeloma cell growth within the scaffolds in vivo, although they did so by following different kinetics. However, CD28-based CAR-T cells resulted in better control of total tumor growth than BBz CARs and they showed more rapid and efficient tumor elimination even in lower E:T ratio. It should be noted that the 28z and BBz CAR designs of our study contain different transmembrane domains (CD28 and CD8α respectively) which, although not expected, could be a cofounder influencing their functional differences. Recently, Salter and colleagues (37) showed, using high-affinity CARs, that 28z design results in significantly stronger intracellular signaling that is responsible for the effector phenotype, decreased persistence, and worse antitumor performance of 28z compared with BBz CAR-T cells. Our data, using also a high-affinity CAR, are in line with this observation, showing significantly higher pZap70 in 28z than in BBz CAR028. Interpreting the results from Salter and colleagues and our study, we have reason to believe that for high-affinity CARs, where cytotoxicity between 28z and BBz CARs does not differ, the signaling strength plays a crucial role in determining the persistence and therefore the long-term antitumor function of CAR-T cells. On the other hand, when using low affinity, the BBz CARs lose their cytotoxic capacity, but this is not connected to a decrease or increase of signaling strength, as discussed above. It seems that the cytotoxicity and cytokine production mechanism is not related to the CAR activation level but possibly on another mechanism.

Supplying full CAR costimulation triggering both pathways results in a more balanced T-cell stimulation (23, 27, 38, 39). We demonstrate that the CD28z + 4-1BBL configuration is the optimal design to provide low-affinity CD38 CAR-T cells with enhanced antitumor cytotoxic potential (through CD28 signaling) and ameliorated proliferative capacity, retention of a memory phenotype, and reduced exhaustion (through 4-1BB/4-1BBL signaling) in vitro. Along with other proposed mechanisms (23, 27) for this combinatorial effect, we found that addition of 4-1BBL expression to a 28z-based CAR restricted the downstream Zap70 phosphorylation in our system. We observed this effect not only at rest but also after antigenic stimulation and for all affinities used.

Furthermore, in our in vivo xenograft model, very low–affinity CAR-T cells (CARB1) having a 28zBBL design resulted in superior restriction of tumor growth and better expansion and persistence as compared with any of the second-generation CD38-CAR designs. Most notably, comparative analysis with previous in vivo data revealed that a 28zBBL design could potentiate even very low–affinity CAR-T cells to elicit antitumor responses comparable with that obtained by CAR-T cells with >1,000 times higher affinity for the target. Clinically most relevant is the fact that this potentiation of antitumor function did not compromise the safety of very low–affinity CARs when tested against healthy CD38+ hematopoietic cells in vitro as well as in vivo. Especially in the whole primary bone marrow cytotoxicity assay, the CARB1-28zBBL cells were able to successfully eliminate primary multiple myeloma tumor cells while sparing surrounding healthy hematopoietic cells. Although optimal affinity levels for different individual target molecules cannot be compared, to our knowledge, this is the lowest affinity for a TAA-targeting CAR that has been shown to be both efficient and safe. Our findings are informative for the design of CAR constructs and suggest that even scFvs with practically unmeasurable Kd values can be effective when used for CAR-T-cell therapy if optimal costimulation is provided.

This analysis focuses on the well-described and most frequently used 4-1BB and CD28 signaling domains. Nevertheless, there are other costimulatory domains that have been used in the design of CARs. The OX40, like 4-1BB, belongs to the TNFR family and primarily signals through the NF-κB pathway (40, 41). On the other hand, CD40L and ICOS are CD28-like and signal through PI3K pathway (41–43). Although 4-1BB and CD28 can be considered representatives of the two receptor families, a different combination of costimulatory moieties may result in similar findings (44). Finally, in our study, CD38 is also expressed on T cells resulting in significant fratricide when high-affinity CARs were used. Taking this into account, we performed functional assays after one week of culture when we assured that no further fratricide takes place. Although, a first antigen encounter of high-affinity CARs could influence comparisons (especially proliferation and phenotype) between CARs of different affinities, we would not expect any influence on the comparisons between same affinity CARs with different costimulatory designs.

In conclusion, we demonstrate here that the selection of the costimulatory design of CARs is of critical importance when using scFv domains with very low–affinity for the target. We show that, if equipped with an optimal combination of CD28 and 4-1BB costimulatory moieties, CARs bearing antigen-binding domains with Kd values even lower than 10−6 mol/L can elicit significant antitumor cytotoxic and proliferative response without compromising their safety. The results of this study unlock the use of very low–affinity CARs to target TAAs, potentially increasing the capacity to discriminate between tumor and healthy cells and suggest that careful construction of TAA-targeting CARs will enhance their clinical potential.

N.W.C.J. van de Donk is a consultant/advisory board member for Janssen, Amgen, Celgene, Bristol-Myers Squibb, Novartis, Takeda, Bayer and Servier, and reports receiving commercial research support from Janssen, Amgen, Celgene, Novartis, and Bristol-Myers Squibb. S. Zweegman is a consultant/advisory board member for Janssen, Celgene, and Takeda. J. de Bruijn is an employee of and has ownership interests (including patents) at Kuros Biosciences. M. Sadelain is a consultant/advisory board member for Berkeley Lights and reports receiving commercial research grants from Juno Therapeutics, Fate Therapeutics, and Atara Biotherapeutics. R. Groen reports receiving commercial research support from Johnson & Johnson. T. Mutis reports receiving commercial research grants from Genmab, Novartis, Janssen, Takeda, and Onkimmune. M. Themeli is a consultant/advisory board member for Covagen AG. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E. Drent, R. Poels, H. Yuan, M. Sadelain, H.M. Lokhorst, T. Mutis, M. Themeli

Development of methodology: E. Drent, R. Poels, H. Yuan, R. Groen, T. Mutis, M. Themeli

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Ruiter, N.W.C.J. van de Donk, S. Zweegman, R. Groen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Drent, R. Poels, N.W.C.J. van de Donk, T. Mutis, M. Themeli

Writing, review, and/or revision of the manuscript: E. Drent, N.W.C.J. van de Donk, S. Zweegman, H. Yuan, H.M. Lokhorst, R. Groen, T. Mutis, M. Themeli

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Poels, J. de Bruijn

Study supervision: T. Mutis, M. Themeli

The H2d-Rag2−/−γc−/− mice used in this study were originally obtained from the Amsterdam Medical Center (AMC, Amsterdam, the Netherlands). This work is in part financially supported by ‘Fonds Stimulans’ the Netherlands (to E. Drent), the European Commission (Marie Curie Individual Fellowship, to M. Themeli), and Stichting VUmc CCA (to M. Themeli).

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.

1.
Sadelain
M
,
Brentjens
R
,
Rivière
I
. 
The basic principles of chimeric antigen receptor design
.
Cancer Discov
2013
;
3
:
388
98
.
2.
van der Stegen
SJC
,
Hamieh
M
,
Sadelain
M
. 
The pharmacology of second-generation chimeric antigen receptors
.
Nat Rev Drug Discov
2015
;
14
:
499
509
.
3.
Davila
ML
,
Riviere
I
,
Wang
X
,
Bartido
S
,
Park
J
,
Curran
K
, et al
Efficacy and toxicity management of 19-28z CAR T cell therapy in B cell acute lymphoblastic leukemia
.
Sci Transl Med
2014
;
6
:
224
5
.
4.
Maude
SL
,
Frey
N
,
Shaw
PA
,
Aplenc
R
,
Barrett
DM
,
Bunin
NJ
, et al
Chimeric antigen receptor T cells for sustained remissions in leukemia
.
N Engl J Med
2014
;
371
:
1507
17
.
5.
Lee
DW
,
Kochenderfer
JN
,
Stetler-Stevenson
M
,
Cui
YK
,
Delbrook
C
,
Feldman
SA
, et al
T cells expressing CD19 chimeric antigen receptors for acute lymphoblastic leukaemia in children and young adults: a phase 1 dose-escalation trial
.
Lancet
2014
;
385
:
517
28
.
6.
Kochenderfer
JN
,
Dudley
ME
,
Kassim
SH
,
Somerville
RP
,
Carpenter
RO
,
Stetler-Stevenson
M
, et al
Chemotherapy-refractory diffuse large B-cell lymphoma and indolent B-cell malignancies can be effectively treated with autologous T cells expressing an anti-CD19 chimeric antigen receptor
.
J Clin Oncol
2015
;
33
:
540
9
.
7.
Turtle
CJ
,
Hanafi
L-A
,
Berger
C
,
Hudecek
M
,
Pender
B
,
Robinson
E
, et al
Immunotherapy of non-Hodgkin's lymphoma with a defined ratio of CD8+ and CD4+ CD19-specific chimeric antigen receptor–modified T cells
.
Sci Transl Med
2016
;
8
:
355ra116
.
8.
Schuster
SJ
,
Svoboda
J
,
Chong
EA
,
Nasta
SD
,
Mato
AR
,
Anak
Ö
, et al
Chimeric antigen receptor T cells in refractory B-cell lymphomas
.
N Engl J Med
2017
;
377
:
2545
54
.
9.
Kochenderfer
JN
,
Dudley
ME
,
Feldman
SA
,
Wilson
WH
,
Spaner
DE
,
Maric
I
, et al
B-cell depletion and remissions of malignancy along with cytokine-associated toxicity in a clinical trial of anti-CD19 chimeric-antigen-receptor-transduced T cells
.
Blood
2012
;
119
:
2709
20
.
10.
Lamers
CHJ
,
Sleijfer
S
,
Vulto
AG
,
Kruit
WH
,
Kliffen
M
,
Debets
R
, et al
Treatment of metastatic renal cell carcinoma with autologous T-lymphocytes genetically retargeted against carbonic anhydrase IX: first clinical experience
.
J Clin Oncol
2006
;
24
:
904
12
.
11.
Morgan
RA
,
Yang
JC
,
Kitano
M
,
Dudley
ME
,
Laurencot
CM
,
Rosenberg
SA
. 
Case report of a serious adverse event following the administration of T cells transduced with a chimeric antigen receptor recognizing ERBB2
.
Mol Ther
2010
;
18
:
843
51
.
12.
Parkhurst
MR
,
Yang
JC
,
Langan
RC
,
Dudley
ME
,
Nathan
DA
,
Feldman
SA
, et al
T cells targeting carcinoembryonic antigen can mediate regression of metastatic colorectal cancer but induce severe transient colitis
.
Mol Ther
2011
;
19
:
620
6
.
13.
Caruso
HG
,
Hurton
LV
,
Najjar
A
,
Rushworth
D
,
Ang
S
,
Olivares
S
, et al
Tuning sensitivity of CAR to EGFR density limits recognition of normal tissue while maintaining potent antitumor activity
.
Cancer Res
2015
;
75
:
3505
18
.
14.
Liu
X
,
Jiang
S
,
Fang
C
,
Yang
S
,
Olalere
D
,
Pequignot
EC
, et al
Affinity-tuned ErbB2 or EGFR chimeric antigen receptor T cells exhibit an increased therapeutic index against tumors in mice
.
Cancer Res
2015
;
75
:
3596
607
.
15.
Drent
E
,
Themeli
M
,
Poels
R
,
de Jong-Korlaar
R
,
Yuan
H
,
de Bruijn
J
, et al
A rational strategy for reducing on-target off-tumor effects of CD38-chimeric antigen receptors by affinity optimization
.
Mol Ther
2017
;
25
:
1946
58
.
16.
Arcangeli
S
,
Rotiroti
MC
,
Bardelli
M
,
Simonelli
L
,
Magnani
CF
,
Biondi
A
, et al
Balance of anti-CD123 chimeric antigen receptor binding affinity and density for the targeting of acute myeloid leukemia
.
Mol Ther
2017
;
25
:
1933
45
.
17.
Chmielewski
M
,
Hombach
A
,
Heuser
C
,
Adams
GP
,
Abken
H
. 
T cell activation by antibody-like immunoreceptors: increase in affinity of the single-chain fragment domain above threshold does not increase T cell activation against antigen-positive target cells but decreases selectivity
.
J Immunol
2004
;
173
:
7647
53
.
18.
Hudecek
M
,
Lupo-Stanghellini
M-T
,
Kosasih
PL
,
Sommermeyer
D
,
Jensen
MC
,
Rader
C
, et al
Receptor affinity and extracellular domain modifications affect tumor recognition by ROR1-specific chimeric antigen receptor T cells
.
Clin Cancer Res
2013
;
19
:
3153
64
.
19.
Walker
AJ
,
Majzner
RG
,
Zhang
L
,
Wanhainen
K
,
Long
AH
,
Nguyen
SM
, et al
Tumor antigen and receptor densities regulate efficacy of a chimeric antigen receptor targeting anaplastic lymphoma kinase
.
Mol Ther
2017
;
25
:
2189
201
.
20.
Porter
DL
,
Hwang
W-T
,
Frey
NV
,
Lacey
SF
,
Shaw
PA
,
Loren
AW
, et al
Chimeric antigen receptor T cells persist and induce sustained remissions in relapsed refractory chronic lymphocytic leukemia
.
Sci Transl Med
2015
;
7
:
303ra139
.
21.
Milone
MC
,
Fish
JD
,
Carpenito
C
,
Carroll
RG
,
Binder
GK
,
Teachey
D
, et al
Chimeric receptors containing CD137 signal transduction domains mediate enhanced survival of T cells and increased antileukemic efficacy in vivo
.
Mol Ther
2009
;
17
:
1453
64
.
22.
Long
AH
,
Haso
WM
,
Shern
JF
,
Wanhainen
KM
,
Murgai
M
,
Ingaramo
M
, et al
4-1BB costimulation ameliorates T cell exhaustion induced by tonic signaling of chimeric antigen receptors
.
Nat Med
2015
;
21
:
581
90
.
23.
Zhao
Z
,
Condomines
M
,
van der Stegen
SJC
,
Perna
F
,
Kloss
CC
,
Gunset
G
, et al
Structural design of engineered costimulation determines tumor rejection kinetics and persistence of CAR T cells
.
Cancer Cell
2015
;
28
:
415
28
.
24.
Kawalekar
OU
,
O’Connor
RS
,
Fraietta
JA
,
Guo
L
,
McGettigan
SE
,
Posey
AD
 Jr
, et al
Distinct signaling of coreceptors regulates specific metabolism pathways and impacts memory development in CAR T cells
.
Immunity
2016
;
44
:
380
90
.
25.
Melero
I
,
Bach
N
,
Hellström
KE
,
Aruffo
A
,
Mittler
RS
,
Chen
L
. 
Amplification of tumor immunity by gene transfer of the co-stimulatory 4-1BB ligand: synergy with the CD28 co-stimulatory pathway
.
Eur J Immunol
1998
;
28
:
1116
21
.
26.
Rudolf
D
,
Silberzahn
T
,
Walter
S
,
Maurer
D
,
Engelhard
J
,
Wernet
D
, et al
Potent costimulation of human CD8 T cells by anti-4-1BB and anti-CD28 on synthetic artificial antigen presenting cells
.
Cancer Immunol Immunother
2008
;
57
:
175
83
.
27.
Zhong
X-S
,
Matsushita
M
,
Plotkin
J
,
Riviere
I
,
Sadelain
M
. 
Chimeric antigen receptors combining 4-1BB and CD28 signaling domains augment PI3kinase/AKT/Bcl-XL activation and CD8+ T cell-mediated tumor eradication
.
Mol Ther
2010
;
18
:
413
20
.
28.
Kim
JH
,
Lee
S-R
,
Li
L-H
,
Park
HJ
,
Park
JH
,
Lee
KY
, et al
High cleavage efficiency of a 2A peptide derived from porcine teschovirus-1 in human cell lines, zebrafish and mice
.
PLoS One
2011
;
6
:
e18556
.
29.
Groen
RWJ
,
Noort
WA
,
Raymakers
RA
,
Prins
HJ
,
Aalders
L
,
Hofhuis
FM
, et al
Reconstructing the human hematopoietic niche in immunodeficient mice: Opportunities for studying primary multiple myeloma
.
Blood
2012
;
120
:
9
16
.
30.
Stephan
MT
,
Ponomarev
V
,
Brentjens
RJ
,
Chang
AH
,
Dobrenkov
KV
,
Heller
G
, et al
T cell-encoded CD80 and 4-1BBL induce auto- and transcostimulation, resulting in potent tumor rejection
.
Nat Med
2007
;
13
:
1440
9
.
31.
de Haart
SJ
,
van de Donk
NWCJ
,
Minnema
MC
,
Huang
JH
,
Aarts-Riemens
T
,
Bovenschen
N
, et al
Accessory cells of the microenvironment protect multiple myeloma from T-cell cytotoxicity through cell adhesion-mediated immune resistance
.
Clin Cancer Res
2013
;
19
:
5591
601
.
32.
Johnson
LA
,
Scholler
J
,
Ohkuri
T
,
Kosaka
A
,
Patel
PR
,
McGettigan
SE
, et al
Rational development and characterization of humanized anti-EGFR variant III chimeric antigen receptor T cells for glioblastoma
.
Sci Transl Med
2015
;
7
:
275ra22
.
33.
Chmielewski
M
,
Hombach
AA
,
Abken
H
. 
CD28 cosignalling does not affect the activation threshold in a chimeric antigen receptor-redirected T-cell attack
.
Gene Ther
2011
;
18
:
62
72
.
34.
Abken
H
. 
Costimulation engages the gear in driving CARs
.
Immunity
2016
;
44
:
214
6
.
35.
Viola
A
,
Lanzavecchia
A
. 
T cell activation determined by T cell receptor number and tunable thresholds
.
Science
1996
;
273
:
104
6
.
36.
Itoh
Y
,
Germain
RN
. 
Single cell analysis reveals regulated hierarchical T cell antigen receptor signaling thresholds and intraclonal heterogeneity for individual cytokine responses of CD4+ T cells
.
J Exp Med
1997
;
186
:
757
66
.
37.
Salter
AI
,
Ivey
RG
,
Kennedy
JJ
,
Voillet
V
,
Rajan
A
,
Alderman
EJ
, et al
Phosphoproteomic analysis of chimeric antigen receptor signaling reveals kinetic and quantitative differences that affect cell function
.
Sci Signal
2018
;
11
:
eaat6753
.
38.
Carpenito
C
,
Milone
MC
,
Hassan
R
,
Simonet
JC
,
Lakhal
M
,
Suhoski
MM
, et al
Control of large, established tumor xenografts with genetically retargeted human T cells containing CD28 and CD137 domains
.
Proc Natl Acad Sci U S A
2009
;
106
:
3360
5
.
39.
Tammana
S
,
Huang
X
,
Wong
M
,
Milone
MC
,
Ma
L
,
Levine
BL
, et al
4-1BB and CD28 signaling plays a synergistic role in redirecting umbilical cord blood T cells against B-cell malignancies
.
Hum Gene Ther
2010
;
21
:
75
86
.
40.
Croft
M
,
So
T
,
Duan
W
,
Soroosh
P
. 
The significance of OX40 and OX40L to T-cell biology and immune disease
.
Immunol Rev
2009
;
229
:
173
91
.
41.
Chen
L
,
Flies
DB
. 
Molecular mechanisms of T cell co-stimulation and co-inhibition
.
Nat Rev Immunol
2013
;
13
:
227
42
.
42.
Watanabe
M
,
Hara
Y
,
Tanabe
K
,
Toma
H
,
Abe
R
. 
A distinct role for ICOS-mediated co-stimulatory signaling in CD4+ and CD8+ T cell subsets
.
Int Immunol
2005
;
17
:
269
78
.
43.
Elgueta
R
,
Benson
MJ
,
de Vries
VC
,
Wasiuk
A
,
Guo
Y
,
Noelle
RJ
. 
Molecular mechanism and function of CD40/CD40L engagement in the immune system
.
Immunol Rev
2009
;
229
:
152
72
.
44.
Guedan
S
,
Posey
AD
,
Shaw
C
,
Wing
A
,
Da
T
,
Patel
PR
, et al
Enhancing CAR T cell persistence through ICOS and 4-1BB costimulation
.
JCI Insight
2018
;
3
.
pii: 96976. doi: 10.1172/jci.insight.96976. eCollection 2018 Jan 11
.

Supplementary data