Increased NFAT activity with dual CAR stimulation in CD19xCD22 CAR T-cells is associated with decreased exhaustion and improved survival

Abstract

Background

Chimeric antigen receptor (CAR) T-cell therapy is effective in treating B-cell malignancies, however relapse due to lack of CAR persistence and antigen-modulated escape remains common. Multiple strategies to simultaneously target CD19 and CD22 have been able to reduce antigen-modulated escape but not completely eliminate relapse. A bicistronic CAR construct consisting of a CD19 CAR incorporating the CD28 costimulatory domain paired with a CD22 CAR incorporating a 4-1BB costimulatory domain (CD19xCD22) demonstrated superior preclinical activity compared with other configurations and is currently under clinical investigation (NCT05098613, NCT05442515, NCT06559189). We hypothesized that simultaneous activation of CD28-containing and 4-1BB-containing CAR molecules not only allows for targeting of both antigens but creates a unique signal which enhances CAR T-cell function and efficacy.

Methods

We tested CD19xCD22 CAR T-cells generated from primary human T-cells against NALM6 with wild-type expression of CD19 and CD22 (CD19+/CD22+) or CRISPR/Cas9 knockout of one or both antigens (CD19+/CD22−, CD19−/CD22+, CD19−/CD22−) to interrogate the effect of dual-CAR stimulation on T-cell function, signaling, and in vivo efficacy in xenograft models.

Results

In vitro proliferation and cytokine production of CD19xCD22 CAR T-cells were primarily driven by activation of the CD19-28z CAR, however the CD22-BBz CAR drove equivalent cytotoxicity. Dual-CAR stimulation of CD19xCD22 CAR T-cells decreased leukemia relapse and improved survival in xenograft models. This increase in efficacy was associated with increased signaling through the phospholipase C-gamma 1 and nuclear factor of activated T-cells pathway after dual-CAR stimulation. Dual-CAR stimulation also led to decreased expression of markers associated with T-cell exhaustion in persistent CD19xCD22 CAR T-cells.

Conclusions

Stimulation of both CAR molecules in a CD19xCD22 bicistronic CAR construct impacts downstream signaling events within the CAR T-cell and subsequently drives a more efficacious in vivo response with evidence of decreased exhaustion in persisting cells. These data suggest that bicistronic CAR platforms have the potential to not only target two antigens to prevent antigen-modulated escape but can be engineered to improve multiple facets of CAR T-cell biology, such as mitigating exhaustion, thereby overcoming multiple mechanisms known to drive relapse in current CAR T-cell therapies.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• The targeting of multiple antigens simultaneously may improve chimeric antigen receptor (CAR) T-cell function and decrease antigen-modulated relapse. CD19xCD22 bicistronic CAR T-cells incorporating both CD28 and 4-1BB costimulatory domains show superior preclinical activity.

WHAT THIS STUDY ADDS

• Bicistronic CAR T-cell models have the potential to optimize CAR T-cell signaling to improve CAR T-cell function and persistence in addition to their ability to target multiple antigens.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These studies add to the understanding of CAR T-cell activation, identifying signaling events that could direct future CAR T-cell development and could inform correlative studies of current CAR T-cell trials, including those trials using CD19xCD22 bicistronic CAR T-cells.

Introduction

Chimeric antigen receptor (CAR) T-cells induce remissions in the majority of patients with B-lineage malignancies,1,8 however treatment failures and relapses after CAR therapy limit the long-term benefit to 40–50% of patients with B-cell acute lymphoblastic leukemia (B-ALL) and non-Hodgkin’s lymphoma (NHL).^{2 3 7 9 10} CD19 antigen loss is a major mechanism of post-CAR relapse in patients with B-ALL and decreased CD19 expression is associated with treatment failure in patients with NHL.11,13 Antigen escape is not restricted to CD19, as antigen modulation is a common mechanism of relapse after CD22-directed CAR T-cells^{4 5} and likely contributes to CAR T-cell resistance in solid tumors.¹⁴

Dual-antigen targeting strategies directed at CD19 and CD22 are being tested to mitigate antigen-modulated escape.15,20 Simultaneous and sequential coadministration of separate CD19 and CD22 CAR T-cell products has demonstrated high response rates with relapse rates of 20–50%.21,24 Recurrence of antigen-positive leukemia was the predominant form of relapse in these studies, highlighting the necessity of maintaining persistence in dual-antigen targeting strategies. Alternatively, a bivalent CD19/22-41BB CAR, with two single-chain variable fragments (scFvs) in-line with a single transmembrane and intracellular signaling domain, demonstrated efficacy in patients with lymphoma but showed limited effectiveness against CD22 alone and susceptibility to CD19^−/lo relapse.^{13 17 20 25} To improve CD22 targeting, a CD19xCD22 bicistronic CAR construct encoding two independent CARs was developed and is undergoing clinical testing (NCT05098613, NCT05442515, NCT06559189).^{26 27} Unlike prior CD19xCD22 bicistronic CAR constructs,¹⁸ this CD19 CAR incorporates the FMC63 scFv and a CD28 costimulatory domain and the CD22 CAR incorporates the m971 scFv with a 4-1BB costimulatory domain. Preclinical data demonstrated that this configuration was superior to two CD28 or two 4-1BB costimulatory domains,^{25 28} suggesting a benefit of incorporating different costimulatory molecules.

Costimulation is critical to the efficacy of CAR T-cells.^{12 29 30} CD28 and 4-1BB differentially drive CAR T-cell activity, with CD28 promoting earlier and more robust signaling, glycolytic-skewed metabolism, and increased expansion, whereas 4-1BB induces lower magnitude signaling, increased use of oxidative phosphorylation, and enhanced in vivo persistence.31,34 Given the distinct actions of these costimulatory domains, we hypothesized that the signals from the CD28 and 4-1BB CAR molecules integrate, leading to the previously described benefit of this configuration.²⁵ We found that while the CD19-28z CAR predominantly drove early effector functions in vitro, the addition of the CD22-BBz signal was necessary for optimal activation of phospholipase C-gamma 1 (PLCγ1) and nuclear factor of activated T-cells (NFAT) activity. Signaling through both CAR molecules was necessary to mitigate exhaustion and for maximal in vivo efficacy.

Methods

Cell lines

NALM6 (ATCC) was transduced with luciferase and GFP. CRISPR/Cas9 interruption of CD19 and/or CD22 was used to generate double negative (CD19⁻/CD22⁻), CD19-negative and CD22-negative variants.¹⁷ Cell lines were routinely authenticated by short tandem repeat analysis and tested for mycoplasma. Experiments were performed within 2 weeks of thawing cell lines. NALM6 was cultured in RPMI supplemented with 10% heat-inactivated fetal bovine serum (FBS), 100 U/mL penicillin/streptomycin, and 2 mM GlutaMAX.

CD19xCD22 CAR

The CD19xCD22 bicistronic CAR construct consists of CD19-28z and CD22-BBz CAR molecules separated by a P2A site. CD19-28z CAR consists of the FMC63 scFv, CD28 hinge-transmembrane-signaling domains, and the CD3-zeta signaling domain. CD22-BBz CAR consists of the m971 scFv, CD8 hinge-transmembrane domain, 4-1BB signaling domains, and CD3-zeta signaling domain. The CD19xCD22 CAR transgene was cloned into the pLenti-vector and lentivirus was produced as previously described^{17 35} through transient transfection of Lenti-X 293T cells (Takara Bio) with transfer and packaging plasmids (pRSV-Rev, pMDLg/pRRE and pMD2.G) using Lipofectamine 3000 (Life Technologies) in Opti-MEM (Thermo Fisher). Media was replaced 6 hours post-transfection and supernatant collected at 24 and 56 hours post-transfection, centrifuged at 3,000 g for 10 min, aliquoted, and frozen at −80°C.

CAR T-cell manufacturing

Leukocyte traps from deidentified, healthy platelet donors were obtained from the Children’s Hospital Colorado blood bank. T-cells were isolated by Ficoll gradient separation (Sigma Aldrich) followed by negative selection using EasySep Human T Cell Isolation Kit (STEMCELL Technologies). T cells were activated with Human T-expander CD3/CD28 Dynabeads (3:1 bead:T-cell ratio, Life Technologies) in AIM V supplemented with 5% heat-inactivated FBS, 100 U/mL penicillin/streptomycin, 5 mL GlutaMAX, 10 mM HEPES, and interleukin (IL)-2 (40 IU/mL). T-cells were transduced at a multiplicity of infection of 2–10 with 10 µg/mL protamine sulfate by spinfection at 1,000 g for 2 hours at 32°C in IL-2 (40 IU/mL). Dynabeads were removed the following day and CAR T-cells were resuspended at 0.5×10⁶ cells/mL and expanded in IL-2 (100 IU/mL) for 4 days. CAR expression was measured by staining with CD22-Fc and/or CD19-Fc proteins (R&D Biosciences) labeled with AF647 or PE Antibody Labeling Kits (Thermo Fisher). CAR T-cells were cryopreserved and thawed 1–2 days prior to experiments.

Positive selection of CAR T-cells

For proliferation, nuclear flow cytometry, western blot, and select in vivo experiments, CD19xCD22 CAR T-cells were positively selected using AF647-conjugated CD22-Fc protein and anti-AF647 Microbeads (Miltenyi) on LS columns (Miltenyi).

Cytokine assay

CAR and mock T-cells were washed three times to remove exogenous IL-2. 10⁵ CAR+ cells were cocultured with NALM6 at a 1:1 effector:target (E:T) ratio in 96-well plates at 37°C for 16 hours. Supernatant cytokine concentrations were quantified by LEGENDplex platform (BioLegend) per manufacturer’s instructions.

Cytotoxicity assay

Luciferase-expressing NALM6 cells were co-cultured with CAR T-cells at various E:T ratios in 200 μL of cytokine-free AIM V for 24 hours in flat-bottom 96-well plates. NALM6 survival was evaluated by luminescence with D-luciferin (Thermo Fisher Scientific) on a Tecan microplate reader. Cytotoxicity is calculated as the difference in luminescence in co-culture relative to NALM6 alone divided by NALM6 alone.^{36 37}

Western blots

Positively selected CAR T-cells were co-cultured with NALM6 for indicated times. Cells were lysed in RIPA buffer with cOmplete Protease Inhibitor and PhosSTOP phosphatase inhibitors (Sigma). Protein concentration was calculated by BCA. 25 µg total protein was loaded onto 4–15% static Bis-Tris gels, transferred to polyvinylidene fluoride membranes and immunoblotted with primary and horseradish peroxidase (HRP)-conjugated secondary antibodies (online supplemental table S1), detected by luminol chemiluminescent reagent (Bio-Rad) on Syngene imaging system.

Flow cytometry

Surface staining: Cells were washed in FACS buffer, stained with indicated antibodies (online supplemental table S2) at 4°C for 30 min, and washed three times before analysis. Phospho-flow cytometry: CAR T-cells were coincubated with NALM6 cells (1:3 E:T) at 37°C for 15 min. Cells were washed in cold phosphate-buffered saline (PBS)/2 mM EDTA, fixed with 1.5% paraformaldehyde (Electron Microscopy Sciences) in PBS/2 mM EDTA at room temperature for 10 min, then permeabilized with 4°C methanol (90%, Sigma) overnight at −80°C. Cells were washed in PBS/2 mM EDTA once and in fluorescence-activated cell sorting (FACS) buffer twice, incubated with cell surface antibodies, PE or AF647-conjugated CD22-Fc, and intracellular antibodies (online supplemental table S2) at 4°C for 30 min, then washed three times.

Nuclear flow cytometry was performed as previously described.³⁸ Briefly, CAR T-cells were positively selected and labeled with CellTrace Violet (CTV) (Thermo Fisher). CAR T-cells were co-cultured with NALM6 (1:1 E:T) at 37°C for 6–24 hours. Nuclear fractions were isolated by incubation in 320 mM sucrose, 10 mM HEPES, 8 mM MgCl₂, cOmplete Protease Inhibitor (Sigma), 0.0% (v/v) Triton-X buffer for 15 min on ice, and centrifuged at 2,000 g for 5 min at 4°C. Nuclei were washed twice in above sucrose buffer without Triton-X, then fixed with 4% paraformaldehyde in Triton-X-free sucrose buffer for 30 min on ice, washed once in 2% FBS/PBS with 8 mM MgCl₂ and permeabilized in the same buffer containing 0.3% (v/v) Triton-X. Nuclei were stained with indicated antibodies in this buffer at 4°C for 30 min, then washed three times in FACS+MgCl₂.

Data acquired on 5-laser, LSRFortessa X-20 (BD Biosciences) and analyzed with FlowJo V.10.8.1 (BD Biosciences).

CAR T-cell in vitro restimulation

Positively selected CAR T-cells were co-cultured with NALM6 cells at an E:T of 1:1 in 4 mL of cytokine-free AIM V. Cells were split 1:4 or 1:2 4 days after plating, and 300 µl was taken for cell counts and flow cytometric analysis every 7 days. Cells were resuspended in fresh AIM V media and NALM6 cells were added back to an E:T ratio of 1:1. 10⁶ cells were taken for intracellular phospho-flow assays at indicated time points.

RNA sequencing and analysis

CD19xCD22 primary human CAR T-cells were manufactured and CAR+ cells selected as described above to >95% purity. CD19xCD22 CAR T-cells were stimulated with NALM6 cells expressing CD19⁻/ CD22⁻, CD22⁺, CD19⁺, CD19⁺/CD22⁺ at an E:T of 1:2 for 6 or 24 hours in triplicates. At each time point, cells were stained with viability eFluor780 and anti-hCD45. CAR T-cells were sorted by FACS to obtain 80K–100K cells per coincubation. RNA was extracted using Qiagen RNeasy micro kit per manufacturer’s protocol. Processed RNA was sent to the University of Colorado Cancer Center (UCCC) Genomics Shared Resource for complementary DNA library preparation. Paired-end sequencing reads of 150 base pairs were generated on a NovaSeq 6000 (Illumina) sequencer at a target depth of 40 million clusters/80 million paired-end reads per sample. The quality of the fastq files was assessed using FastQC (V.0.11.8) (http://www.bioinformatics.babraham.ac.uk/projects/fastqc), FastQ Screen (V.0.13.0) and MultiQC (V.1.8).^{39 40} Illumina adapters and low-quality reads were filtered out using Trim Galore (V.0.6.7). Trimmed fastqc files were aligned to the hg38 human reference genome using STAR (V.2.7.9a) and aligned counts per gene were quantified using Salmon (V.1.10.1).^{41 42} Differential gene expression analysis was performed using the DESeq2 R package (V.1.48.1).⁴³ Transcription factor activity analysis was performed using Ingenuity Pathway Analysis (IPA).⁴⁴ Heatmaps were generated using the pheatmap R package (V.1.0.12).

Mouse in vivo studies

NOD/scid/IL-2rγ^−/− (NSG; Jackson Laboratory) mice were bred and housed in the University of Colorado, Anschutz Medical Campus vivarium. Animal studies approved by the University of Colorado Institutional Animal Care and Use Committee.

NSG mice were engrafted with 10⁶ luciferase-positive NALM6 by tail vein injection on day −4. Mice were treated with 2–5×10⁶ CAR+ T cells on day 0. Mock T-cell doses were equivalent to the total T-cell number in CAR T-cell dose. Cages were randomly assigned treatments. Group sizes were based on prior studies.^{25 28} Each cage underwent bioluminescence imaging 5 min after intraperitoneal injection of IVISbrite D-Luciferin (Revvity), using a Xenogen In Vivo Imaging System 200. Survival was determined using predefined endpoints of leukemia progression (hind limb paralysis, impaired ambulation, severe weakness and/or wasting). Mice were monitored for xenogeneic graft-versus-host disease (xGVHD) and survival studies were terminated when >50% of surviving mice of any group developed xGVHD. In non-survival studies, mice were euthanized at predefined time points and bone marrow was collected from femurs and tibias. Marrow was processed to single-cell suspensions, RBC-lysed with ACK Lysing Buffer (Gibco), stained, and analyzed by flow cytometry as described above. A single tibia from each mouse was completely flushed and counted to calculate absolute cell numbers per tibia.

Statistics

Statistical tests were performed using GraphPad Prism software and represented by the following: ns, not significant, p≥0.05; *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. Statistical significance was assessed by one-way analysis of variance (ANOVA) with Sidak post hoc tests or two-way ANOVA with Tukey post hoc tests based on experimental design. Error bars represent SEM. Survival was analyzed by Wilcoxon signed-rank test. Multiparameter flow data visualized using SPICE (Simplified Presentation of Incredibly Complex Evaluations) analysis.⁴⁵

Results

Activation of the CD19-28z CAR predominantly drives proliferation and cytokine production of CD19xCD22 CAR T-cells

We generated a CD19xCD22 bicistronic CAR composed of a CD19-28z CAR and a CD22-BBz CAR, which is undergoing testing in clinical trials NCT05098613, NCT05442515, and NCT06559189 (figure 1A).²⁸ Using the B-ALL cell line NALM6 (CD19+/CD22+) and subclones engineered to express either CD19 alone, CD22 alone, or neither antigen, we tested the relative contribution of CD19-28z or CD22-BBz CAR on the activation of CD19xCD22 CAR T-cells (figure 1A). Transduction of primary human T cells with the CD19xCD22 CAR construct resulted in concomitant surface expression of each CAR on the majority of transduced T cells (transduction efficiency was 30–65% across T-cell donors; figure 1B). To determine the contribution of each CAR to proliferation, CD19xCD22 CAR T-cells were enriched to >95% CAR+, labeled with the membrane dye CTV, and co-cultured with NALM6 cells expressing CD19, CD22, or both antigens. CAR T-cells stimulated through CD19-28z CAR alone showed comparable expansion to those stimulated through both CD19-28z and CD22-BBz CARs (figure 1C). Stimulation of CD22-BBz CAR resulted in significantly lower proliferation and reduced CTV dilution compared with stimulation through CD19-28z or both CARs (figure 1C). Similarly, we observed significantly higher secretion of IL-2, tumor necrosis factor alpha (TNFα), and interferon-gamma (IFNγ) on stimulation of CD19-28z relative to CD22-BBz (figure 1D). There was no additional TNFα or IFNγ secretion when both CD19-28z and CD22-BBz CARs were engaged compared with CD19-28z alone; however, we noted a slight increase in IL-2 production on activation of both CARs, suggesting that dual CAR stimulation could generate additive effects in certain T-cell effector functions (figure 1D).

Figure 1. Early functional responses of bicistronic CD19xCD22 CAR T-cells are predominantly driven by the CD19-28z CAR. (A) Schematic of the bicistronic CD19xCD22 CAR construct and experimental design for testing single-CAR versus dual-CAR activation. (B) Representative flow cytometry plot of surface expression of CD19-28z and CD22-BBz CARs on CD19xCD22 transduced primary human T-cells. (C) In vitro proliferation after 72-hour co-cultures of 5×10⁴ purified CAR+T cells with 5×10⁴ NALM6 cells expressing CD19 and CD22, CD22 alone, CD19 alone, or neither antigen (DN). Representative flow cytometry histograms of CTV membrane dye dilution with CAR T-cell division. Data from two T-cell donors, three replicates each. (D–E) CD19xCD22 CAR T-cells were co-cultured with antigen-variant NALM6 clones (E:T 1:1) for 16 hours. Supernatants were analyzed for IL-2, TNFα, and IFNγ concentrations (D) or concentrations of granzyme A, granzyme B, and perforin (E) by LEGENDPlex assay. Data compiled from three independent T-cell donors. (F) Luciferase-based killing assay of antigen-variant NALM6 clones after 24-hour co-culture with CD19xCD22 CAR T-cells. Data representative of three independent T-cell donors. Data graphed as means±SEM. Statistical analysis by one-way ANOVA (C–E) or two-way ANOVA (F). *p≤0.05; **p≤0.01; ***p≤0.001; ****p≤0.0001. ANOVA, analysis of variance; APC, allophyocyanin; CAR, chimeric antigen receptor; CTV, CellTrace Violet; DN, double negative; E:T, effector:target; IFNγ, interferon-gamma; IL, interleukin; ns, not significant; PE, phycoerythrin; scFv, single-chain variable fragment; TNFα, tumor necrosis factor alpha.

As with proliferation and cytokine secretion, we observed that the CD19-28z CAR drove significantly more granzyme A secretion without an additive increase with dual CAR stimulation (figure 1E). This dominant influence of the CD19-28z CAR was not seen with other cytotoxic proteins, as stimulation of either CD19-28z or CD22-BBz resulted in similar release of perforin and granzyme B (figure 1E). Furthermore, there was no difference in CD19-28z-mediated or CD22-BBz-mediated killing of NALM6 in an overnight co-culture. The cytotoxicity driven by each individual CAR was comparable to dual CAR stimulation at higher E:T ratios; however dual CAR stimulation demonstrated reduced killing at low E:T ratios (figure 1F). Differences in T-cell function did not correlate with changes in the surface expression of either CAR molecule in response to stimulation. Similar to prior studies, stimulation of the CD19-28z CAR resulted in its temporary downregulation during the first 6 hours,⁴⁶ which was not seen with stimulation of the CD22-BBz CAR, which drove less short-term effector functions (online supplemental figure S1). This downregulation of the CD19-28z CAR was greater than that of antigen-activated CD22-BBz CARs, potentially due to endogenous mechanisms for endocytosing CD28 molecules,⁴⁷ and even occurred on stimulation of the CD22-BBz CAR alone, suggesting an antigen-independent endocytosis similar to that of bystander T-cell receptor complexes.48,50 Collectively, these data suggest multiple effector functions are predominantly driven by the CD19-28z CAR with little additive effect of coincident CD22-BBz activation, except for cytotoxicity which was equally driven by each CAR in the CD19xCD22 construct.

Dual CAR stimulation improved leukemia clearance and prolonged survival

To evaluate the role of each CAR on in vivo efficacy, NSG mice engrafted with CD19+/CD22+, CD19+, or CD22+ NALM6 subclones were treated with CD19xCD22 CAR T-cells. At a dose of 2×10⁶ CAR T-cells per mouse, stimulation of CD19-28z, CD22-BBz, or both CARs led to initial leukemia clearance and prolonged survival relative to mock T-cells (figure 2A,B). However, mice engrafted with CD19+ or CD22+ NALM6 eventually relapsed and succumbed to leukemia progression. Conversely, 40% of mice engrafted with CD19+/CD22+ NALM6 remained leukemia-free with significantly prolonged survival relative to mice with CD19+ or CD22+ NALM6 (figure 2A,B). The improved survival with dual-antigen stimulation of CD19xCD22 CAR T-cells was not attributable to increased expansion of CAR T-cells in vivo, as there was no significant difference in the number of total CAR T-cells in the bone marrow of mice engrafted with CD19+/CD22+ NALM6 versus those engrafted with CD19+ NALM6 7 days after CAR T-cell infusion (figure 2C,D). Like in vitro proliferation (figure 1C), stimulation of the CD19-28z CAR (whether alone or coincident with CD22-BBz) led to increased expansion relative to CD22-BBz stimulation alone at this time point (figure 2D). By day 14, T-cells stimulated through the CD19-28z CAR had undergone contraction, whereas those stimulated through the CD22-BBz CAR alone had continued to expand (figure 2E), potentially explaining how stimulation through CD22-BBz alone provided a similar survival advantage to stimulation through CD19-28z alone; however this delayed expansion was insufficient to match the efficacy of dual CAR stimulation (figure 2A,B).

Figure 2. Dual-CAR stimulation improved survival over activation through either the CD19-28z or CD22-BBz CARs alone. (A–B) NSG mice injected with 10⁶ antigen-variant NALM6 clones on day −4, followed by 2×10⁶ CAR or mock T-cells on day 0. (A) BLI of mice treated with mock or CD19xCD22 CAR T-cells. Representative of three independent T-cell donors/experiments. (B) Survival of mice compiled from three independent experiments. Analysis by log-rank (Mantel-Cox) test. (C–E) NSG mice injected with 10⁶ NALM6 clones on day −4 followed by 3–5×10⁶ CAR T-cells on day 0. Bone marrow analyzed on Day 7 (D) and Day 14 (E) for absolute counts of total CAR T-cells compiled from 2 to 3 independent T-cell donors. (F) Day 7 phenotype of CAR T-cells from representative donor. Data graphed as means±SEM. One-way ANOVA or two-way ANOVA with Tukey’s multiple comparisons test used for D–E and F, respectively. *p≤0.05; **p≤0.01; ***p≤0.001; ****p≤0.0001. ANOVA, analysis of variance; BLI, bioluminescent imaging; CAR, chimeric antigen receptor; DN, double negative; ns, not significant; Tcm, T central memory; Tem, T effector memory; Temra, T effector memory with re-expression of CD45RA; Tscm, T stem cell memory.

T-cell differentiation during expansion also did not explain the survival difference seen between dual and single CAR stimulation. CD19-28z, CD22-BBz, or dual CAR stimulation did not differentially impact the expansion of CD4+ CAR T-cells (CAR4s) or CD8+ CAR T-cells (CAR8s) (online supplemental figure S2). Stimulation through CD22-BBz alone drove significantly higher proportions of T-cells with a T stem cell memory phenotype and CAR4s with a T central memory phenotype than did stimulation through CD19-28z. Conversely, stimulation through CD19-28z alone resulted in a greater proportion of CAR8s with a T effector memory with re-expression of CD45RA (TEMRA) phenotype and increased proportions of the T effector memory phenotype across both CD4 and CD8 compartments (figure 2F). Furthermore, there were no significant differences in these populations between CAR T-cells stimulated through both receptors or through CD19-28z alone, suggesting that while each CAR can differentially impact T-cell differentiation, the CD19-28z dominantly drives greater effector differentiation (figure 2F). While this dominant effect of the CD19-28z CAR over the CD22-BBz CAR is evident in the expansion and differentiation of CD19xCD22 CAR T-cells in vivo, these data do not explain the benefit of dual CAR activation on efficacy and survival.

Dual CAR stimulation increases NFAT activity in CD19xCD22 bicistronic CAR T-cells

We hypothesized that the survival advantage provided by dual CAR stimulation of CD19xCD22 CAR T-cells was related to the integration of signals from the CD19-28z and CD22-BBz CARs. Phosphorylation of the ZAP70 activating residue, Y319, was strongly driven by the CD19-28z CAR, without a detectable increase with additional activation of CD22-BBz (figure 3A). Conversely, phosphorylation of the activating residue, Y783, of PLCγ1 was increased with dual CAR stimulation relative to stimulation through either CAR alone (figure 3B). The impact of dual stimulation on PLCγ1 phosphorylation was not limited to initial responses, as this trend was observable on restimulation 7 and 28 days later (online supplemental figure S3A,B). Downstream of PLCγ1, we found that phosphorylation of NF-κB was primarily driven by CD19-28z CAR stimulation and was not significantly increased with the addition of CD22-BBz activation (figure 3C).

Figure 3. Dual-CAR stimulation leads to increased NFAT activity in CD19xCD22 bicistronic CAR T-cells: (A) Western blot of phosphorylated ZAP70 in CD19xCD22 CAR T-cells after 5 min co-culture with antigen-variant NALM6 clones. (B–C) Intracellular phospho-flow analysis of phospho-PLCγ1 (B) and phospho-NF-κB (C) in CAR T-cells after a 15 min co-culture with NALM6 subclones. Data compiled from three independent T-cell donors and experiments. (D-F) Nuclear flow cytometry analysis of NFAT in total (D), CAR4 (E), and CAR8 (F) CAR T-cells co-cultured with NALM6 subclones for 6 hours. Geometric MFI of one representative donor of three independent donors. Data graphed as means±SEM. Statistical analysis by one-way ANOVA with Tukey’s multiple comparisons test. *p≤0.05; **p≤0.01; ***p≤0.001; ****p≤0.0001. ANOVA, analysis of variance; CAR, chimeric antigen receptor; DN, double negative; MFI, mean fluorescence intensity; NFAT, nuclear factor of activated T-cells; ns, not significant; PLCγ1, phospholipase C-gamma 1.

We next interrogated other pathways downstream of PLCγ1 for evidence of an additive effect of dual CAR signaling. Jurkat cells were transduced with constructs encoding the promoters for the main transcriptional regulators of T-cell activity (AP-1, NF-κB, and NFAT) upstream of fluorescent reporters. Stimulation of Jurkat cells with phorbol myristate acetate (PMA)/ionomycin, which mimics PLCγ1 activity, demonstrated transcriptional activity of AP-1 and NF-κB and weak NFAT activity at 24 hours, with higher NFAT reporter activity observed at 8 hours poststimulation (online supplemental figure S4A). As translocation of NFAT from the cytoplasm to the nucleus is necessary for its transcriptional regulatory activity, Jurkat cells were activated with PMA/ionomycin and whole cells or isolated nuclei were fixed, permeabilized and stained for NFAT.³⁸ As expected, the level of NFAT in whole Jurkat cells was unchanged with activation, however nuclear NFAT was increased with stimulation, consistent with activation-induced nuclear translocation (online supplemental figure S4B). A time course demonstrated the peak of nuclear NFAT staining around 6 hours, similar to the NFAT reporter assay (online supplemental figure S4C). Stimulation of Jurkat cells expressing the CD19xCD22 CAR with NALM6 cells also resulted in increased nuclear translocation of NFAT, driven most potently by CD19-28z (online supplemental figure S4D). Interestingly, in CD19xCD22 CAR T-cells generated from healthy donor T-cells, nuclear translocation of NFAT was highest on dual CAR stimulation (figure 3D). This was largely driven by CAR4s, as nuclear translocation of NFAT in CAR8s was primarily driven by CD19-28z, although at reduced levels compared with CAR4s (figure 3E,F). This discrepancy between primary T-cells and Jurkat cells is likely related to the deficiency of PTEN in Jurkat cells,⁵¹ which could otherwise antagonize the activity of PI3K which associates with the CD28 intracellular domain.⁵² These data demonstrate that simultaneous activation of CD19-28z and CD22-BBz CARs differentially increases select signal transduction events within CD19xCD22 CAR T-cell, and this signal integration may underlie the differences in in vivo efficacy seen with dual CAR stimulation by CD19+/CD22+ NALM6.

MAPK signaling is predominantly driven through the CD19-28z CAR

Imbalances between NFAT and AP-1 activity have been associated with T-cell dysfunction.53,55 Therefore, we assessed MAPK signaling in CD19xCD22 CAR T-cells. Phosphorylation of p38 was primarily driven by CD19-28z and was not significantly increased with coincident stimulation of CD22-BBz (figure 4A). Similarly, phosphorylation of ERK1/2 was driven by CD19-28z without an additive effect from CD22-BBz (figure 4B). c-Jun is a canonical component of AP-1 and has been shown to improve CAR T-cell function.⁵⁶ We optimized nuclear flow staining for c-Jun in Jurkat cells and found optimal expression after dual CAR stimulation (online supplemental figure S4D). However, nuclear expression of c-Jun in primary T-cells was driven by CD19-28z in both CAR4s and CAR8s, like p38 and ERK1/2 phosphorylation (figure 4C–E). These data suggest that optimal AP-1 activity in CD19xCD22 CAR T-cells requires stimulation of CD19-28z, correlating with the diminished cytokine production, proliferation, and in vivo expansion seen when bicistronic CAR T-cells are activated through CD22-BBz only (figures1C,D 2D).

Figure 4. MAPK pathway signaling is predominantly driven through stimulation of the CD19-28z CAR in CD19xCD22 bicistronic CAR T-cells: (A–B) Intracellular phospho-flow analysis of phospho-p38 (A) and phospho-ERK (B) in CAR T-cells after 15 min co-culture with antigen-variant NALM6 clones. Data compiled from three independent T-cell donors. (C–E) Nuclear flow cytometry analysis of c-Jun in total (C), CAR4 (D) and CAR8 (E) CAR T-cells co-cultured with NALM6 subclones for 24 hours. Geometric MFI of one representative donor of three independent donors. Data graphed as means±SEM. Statistical analysis by one-way ANOVA with Tukey’s multiple comparisons test. *p≤0.05; **p≤0.01; ***p≤0.001; ****p≤0.0001. ANOVA, analysis of variance; CAR, chimeric antigen receptor; DN, double negative; MFI, mean fluorescence intensity; ns, not significant; PE, phycoerythrin.

Dual CAR stimulation produces a unique global transcriptomic profile driven by increased activity of canonical T-cell transcriptional regulators

Canonical T-cell signaling results in widespread transcriptional changes mediated through transcriptional activators, such as the NFAT, NF-κB, and AP-1 families. Stimulation of CD19xCD22 CAR T-cells resulted in global transcriptional changes at 6 and 24 hours. Consistent with our findings of increased NFAT nuclear translocation (figure 3D–F), CD19xCD22 CAR T-cells undergoing dual stimulation demonstrated a transcriptomic profile distinct from stimulation of either CAR alone by 24 hours (figure 5A). IPA of transcription factor activity demonstrated higher activity of AP-1 (JUN, JUNB, FOS), NF-κB (NFKB1, REL) and NFAT (NFATC2, NFATC3) family members at 6 hours. This trend continued at the 24-hour time point, with further increases of NFAT family member activity in dual-stimulated CAR T-cells (figure 5B). Consistent with NFAT’s known roles in T-cell activity, we observed increased expression of transcription factors related to T-cell differentiation, growth, and survival (EGR2, EGR3, MYC, ATF3, IRF8), transcripts related to cytokine responsiveness (IL2RA, STAT5A), cytokines (IL2, IFNG, TNFA, IL4, IL9, IL21), trafficking molecules (CXCR5, ITGB3, VCAM1), and negative regulators of T-cell activation (CTLA4, CISH, SOCS3, HAVCR2) (figure 5C). Dual CAR stimulation resulted in dramatic changes in a variety of transcriptional regulators beyond the NFAT, AP-1, and NF-κB families (online supplemental figure S5A,B). Notably, dual CAR stimulation increased activity of transcriptional regulators which promote T-cell survival and persistence (STAT5A, STAT5B, STAT3) while decreasing the activity of regulators associated with limited persistence (IRF7, TOX; figure 5B, online supplemental figure S5A,B).57,60 Indeed, we observed increased accumulation of CD19xCD22 CAR T-cells undergoing weekly stimulation through both CAR molecules relative to stimulation through either the CD19-28z or CD22-BBz CAR alone (figure 5D), suggesting dual CAR stimulation could promote persistence while reducing T-cell exhaustion.

Figure 5. Dual CAR stimulation produces a unique global transcriptomic profile driven by increased activity of canonical T-cell transcriptional regulators: (A) PC analysis of RNA-Seq data from CD19xCD22 CAR T-cells 6 and 24 hours after stimulation with CD22+, CD19+, CD19+/CD22+ NALM6 clones, or without stimulation (CD19−/CD22− NALM6, DN). (B) IPA predicted transcription factor activity of CD19xCD22 CAR T-cells stimulated with antigen-variant NALM6 cells at 6 and 24 hours. (C) Heat maps of NFATC1 and NFATC2 target genes 6 and 24 hours poststimulation. (D) Expansion of CD19xCD22 CAR T-cells stimulated with antigen-variant NALM6 clones on day 0 and every 7 days thereafter (E:T 1:1). Statistical analysis by two-way ANOVA with Tukey’s multiple comparisons test. *p≤0.05; **p≤0.01; ***p≤0.001; ****p≤0.0001. ANOVA, analysis of variance; CAR, chimeric antigen receptor; DN, double negative; E:T, effector:target; IFNG, interferon-gamma; IL, interleukin; IPA, Ingenuity Pathway Analysis; NFAT, nuclear factor of activated T-cells; PC, principal component; TNF, tumor necrosis factor.

Dual CAR stimulation does not drive exhaustion in CD19xCD22 CAR T-cells

To test the impact of dual CAR stimulation on T-cell persistence in vivo, NSG mice engrafted with CD19+/CD22+, CD19+, or CD22+ NALM6 were treated with a curative dose of CD19xCD22 CAR T-cells and CAR T-cell persistence was evaluated 28 days later (figure 6A). CAR T-cell persistence in the bone marrow was not different based on which CAR was stimulated and persistent CAR T-cells stimulated through CD19-28z alone or through both receptors demonstrated similar phenotypes, whereas CD22-BBz stimulation resulted in fewer T-cells with a TEMRA phenotype (figure 6B,C). Dual CAR stimulation also increased expression of IL-7 receptor-alpha over single CAR stimulation in CAR4s and in CAR8s relative to CD22-BBz stimulation (figure 6D). Unexpectedly, inhibitory receptors characteristic of T-cell exhaustion (PD1, TIM3, TIGIT, CD39, 2B4) were not increased after dual CAR stimulation relative to single CAR stimulation (figure 6E,F). Indeed, the majority of persistent CAR4s that underwent dual CAR stimulation expressed one or no inhibitory receptors (figure 6E). Similar findings were more prominent in the CAR8s that received dual stimulation, particularly in comparison to CAR T-cells stimulated through CD19-28z (figure 6F). These differences in inhibitory receptor expression developed over time, as CD19xCD22 CAR T-cells expressed similar levels of exhaustion markers during expansion regardless of which CAR(s) were stimulated (online supplemental figure S6A,B). Furthermore, the decreased exhaustion of dual-stimulated CD19xCD22 CAR T-cells was not attributable to greater downregulation of CAR surface expression. Dual stimulated CAR T-cells demonstrated a drop in both CD19-28z and CD22-BBz CAR surface expression during expansion, with a rebound in CD22-BBz surface expression 35 days postinfusion (online supplemental figure S7A,B) likely resulting from the recycling of the CD22 CAR to the surface without further antigen-mediated endocytosis after leukemia clearance. While a similar pattern of CAR surface expression was seen with repeated stimulation in vitro, there was no significant difference in CAR surface expression after dual CAR stimulation relative to CD19-28z CAR stimulation alone (online supplemental figure S7C,D). Collectively, these studies suggest that the increased signaling with dual CAR activation does not compromise CAR T-cell persistence but generates a less exhausted phenotype suggestive of enhanced potential for functional persistence.

Figure 6. Dual CAR stimulation does not compromise CD19xCD22 CAR T-cell persistence and leads to decreased expression of exhaustion markers. (A) NSG mice were injected with 10⁶ antigen-variant NALM6 clones on day −4. Engraftment was confirmed by BLI on day −1 and 3–5×10⁶ CAR T-cells were given on day 0. Bone marrow was analyzed for persistent CAR T-cells on day 28. (B) Absolute counts of total CAR T-cells, CAR4, or CAR8 CAR T-cells. (C) Phenotyping of CAR4 and CAR8 cells by flow cytometry on day 28. (D) Flow cytometric analysis of IL7Rα expression on CAR4 and CAR8 cells on day 28. (E) Flow cytometry for exhaustion marker expression by persistent CAR4 and CAR8 cells on day 28. All data compiled from three independent T-cell donors/experiments (total n=15 mice/group). Data graphed as means±SEM. Statistical analysis of data in (B,D) by one-way ANOVA with Tukey’s multiple comparisons test. Statistical analysis of data in (C,E,F) by two-way ANOVA with Tukey’s multiple comparisons test. SPICE analysis used for (E–F). Asterisks in (C) represent difference from CD19+/CD22+ group. All comparisons not significant unless specified. *p≤0.05; **p≤0.01; ***p≤0.001; ****p≤0.0001. ANOVA, analysis of variance; BLI, bioluminescent imaging; CAR, chimeric antigen receptor; IL7Rα, interleukin-7 receptor-alpha; MFI, mean fluorescence intensity; ns, not significant; SPICE, Simplified Presentation of Incredibly Complex Evaluations; Tcm, T central memory; Tem, T effector memory; Temra, T effector memory with re-expression of CD45RA; Tscm, T stem cell memory.

Discussion

CD19-directed CAR T-cell success in B-ALL is limited by post-CAR relapse driven by insufficient persistence or loss of CD19 expression.^{2 3 7 9 10} CD22-directed CAR T-cells can reinduce remissions after CD19 loss, however long-term efficacy is limited by relapse, primarily with CD22-low leukemia.⁴ Multiple strategies to mitigate antigen-modulated escape in B-ALL have undergone clinical testing, including sequential^{22 23} or simultaneous^{21 24} infusions of CD19-directed and CD22-directed CAR T-cells, bivalent CD19/22 CAR T-cells,^{13 25} and bicistronic CD19xCD22 CAR T-cells.^{18 19} While some of these strategies decreased relapse rates to 20%^{21 22} with little antigen-modulated escape, relapses with antigen-positive leukemia underscore the necessity of maintaining persistence in multiantigen-targeted CAR T-cell approaches.

Bicistronic constructs offer the possibility of targeting multiple antigens while engineering CAR signaling to enhance function and persistence. A BCMAxGPRC5D-targeting bicistronic construct demonstrated more efficacy than pooled BCMA-directed and GPRC5D-directed CAR T-cells at low doses.⁶¹ Similar to our findings, this increased efficacy was dependent on dual-antigen expression. The BCMAxGPRC5D bicistronic construct incorporated 4-1BB signaling domains in each CAR, thus the increased efficacy could be related to overall strength of signaling. However, we hypothesize that the benefit of stimulating both CARs in our CD19xCD22 construct was due to the integration of signals from the CD19-28z and CD22-BBz CAR molecules. In vitro T-cell functions were not augmented by dual-antigen stimulation, as proliferation and cytokine production were primarily driven by the CD19-28z CAR and there was no additive benefit to dual-CAR stimulation in cytotoxicity. IL-2 production was modestly increased with dual CAR simulation, suggesting some T-cell functions could be impacted in an additive way, but this was the exception to the otherwise predominant CD19-28z CAR effect.

Dual-antigen stimulation led to optimal leukemia clearance in vivo, however this did not correlate with CAR T-cell expansion or differentiation, which were similar between CD19-stimulated and dual-antigen-stimulated CAR T-cells. The delayed expansion and decreased differentiation driven by the CD22-BBz CAR was consistent with other 4-1BB-containing CARs, however we did not find the expected increased persistence when bicistronic CAR T-cells were stimulated through the CD22-BBz CAR only, but instead found equal persistence when activating the CD19-28z CAR only or both CARs together. We anticipated increased exhaustion in CD19-28z stimulated cells and, given its dominant impact on early T-cell function, anticipated a similar outcome in dual-antigen stimulated CAR T-cells. However, we unexpectedly found that dual-antigen stimulation generated the largest proportion of T-cells expressing 0–1 exhaustion markers. We hypothesize that the ability of 4-1BB to mitigate exhaustion in tonically signaling CARs⁶² and to recruit THEMIS to the synapse³¹ may restrain the exhaustion driven by CD28 activity and/or increased signaling.

To understand the advantage of dual-antigen stimulation on in vivo activity, we assessed signaling events that ultimately drive T-cell function, differentiation, and persistence. Most signaling pathways were strongly dependent on CD19-28z CAR activation, including phosphorylation of ZAP70, an essential kinase in proximal T-cell signaling,⁶³ and downstream pathways including MAPK signaling and nuclear expression of c-Jun. However, PLCγ1 phosphorylation was synergistically increased with dual CAR activation. PLCγ1 activity is critical for T-cell activation, generating inositol-1,4,5-triphosphate and diacylglycerol which promotes increased intracellular calcium, activation of protein kinase C and RasGRP1, and ultimately activation of the NFAT, NF-κB and MAPK pathways, respectively.64,66 Phosphorylation of PLCγ1 can be induced by monovalent CARs containing either CD28 or 4-1BB, and is sufficient to drive CAR T-cell activity.^{32 67} We observed minimal PLCγ1 phosphorylation with single CAR stimulation which could reflect lower CAR expression in our bicistronic construct or our use of cell-based stimulation, as opposed to bead-based stimulation which may be stronger, more synchronous, and is independent of synapse formation. While dual-antigen stimulation could generate stronger signaling through the engagement of more CAR molecules, this does not explain the lack of additive effect on ZAP70 phosphorylation which lies upstream of PLCγ1. The role of synapse formation is of particular interest, as PLCγ1 depends on LAT for recruitment to the synapse. Inefficiencies in LAT phosphorylation in CAR T-cells decrease PLCγ1 activation and downstream signaling.⁶⁸ Future studies will evaluate the impact of dual-antigen stimulation on synapse formation and maturation which could explain the seemingly emergent signaling properties we observe in the setting of dual-antigen stimulation.

Ultimately, increased phosphorylation of PLCγ1 was reflected by increased activity of multiple translational regulators, most prominently NFAT family members which demonstrated broad changes in the transcriptome of T-cells receiving dual-antigen stimulation. While unbalanced NFAT is associated with T-cell exhaustion, ternary complexes composed of NFAT and AP-1 transcription factors promote cytokine production, T-cell activation, and persistence.53,5569 While nuclear c-Jun was observed after activation of each CAR, its transcriptional activity was highest in T cells receiving dual-CAR activation, likely representing cooperative binding and activity with NFAT. We hypothesize this cooperative activity underlies the decreased exhaustion of dual-stimulated CD19xCD22 CAR T-cells.

In conclusion, we have shown that dual CAR stimulation in CD19xCD22 CAR T-cells enhances in vivo efficacy while decreasing CAR T-cell exhaustion and leukemia relapse in xenograft models. This was associated with increased signaling through the PLCγ1-NFAT signaling pathway with dual-activation of the CD19-28z and CD22-BBz CARs. These results suggest that signal integration downstream of bicistronic CAR constructs has the potential to enhance T-cell function and persistence, broadening the application of bicistronic constructs beyond mitigating antigen-modulated escape via OR-gating. As this construct is currently being tested in early phase clinical trials (NCT05098613, NCT05442515, NCT06559189), these data will inform correlative studies of clinical CAR T-cell products in response to a patient’s leukemia to identify prognostic patterns of signaling to guide current and future CAR T-cell therapy.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.