Enhanced antitumoral activity of the academic CAR-T ARI0002h against normal and low BCMA-expressing myeloma cells after incorporating a transmembrane CD28 domain

Abstract

Background

B-cell maturation antigen (BCMA) is the main target for chimeric antigen receptor (CAR)-T cells in multiple myeloma (MM), demonstrating promising outcomes. However, unlike what happens with CART19 in lymphoblastic leukemia and non-Hodgkin's lymphoma, a high proportion of patients will relapse after CAR-T BCMA therapy due to insufficient antigen expression, low CAR-T cell persistence and/or T-cell exhaustion. In other B cell malignancies, second-generation anti-CD19 4-1BB CARs with CD28-transmembrane domain (TMD) have shown high efficacy and a favorable toxicity profile. We have developed a second-generation CD8α-TM BCMA-4-1BBζ CAR-T product, ARI0002h (Cesnicabtagene-autoleucel) for patients with relapsed/refractory MM. We hypothesized that replacing the TMD of ARI0002h with a CD28-TMD could increase efficacy and reduce tumor escape while maintaining a tolerable toxicity profile.

Methods

We generated CAR-T cells using T-cells isolated from buffy coats and evaluated the efficacy and fitness of CAR-Ts at day 8–10 of expansion against several MM cell lines. In vitro analyses included cytotoxicity, proliferation, cytokine secretion, T-cell subset markers, activation and exhaustion profiling, metabolomic assays, and RNA-seq after multiple tumor challenges. In in vivo xenograft studies using NSG mice, with tumor cells expressing GFP-ffLuc, disease progression was monitored weekly via bioluminescence imaging.

Results

Despite showing similar in vitro performance regarding cytotoxicity, proliferation and cytokine production, ARI2h-TM28 outperforms ARI0002h in a low BCMA expression setting and achieves superior in vivo tumor control and survival in relapse models with antigen downregulation. Furthermore, ARI2h-TM28 showed an optimized metabolic profile, more oxidative and energetic compared with ARI0002h, with downregulation of proinflammatory genes in CD8 T cells, contributing altogether both to reduced exhaustion and increased persistence of the CARs, improving their efficacy in preclinical models.

Conclusions

Incorporating a CD28-TMD into the ARI0002h CAR enhances tumor control even in relapse models with downregulation of the target antigen, offering improved long-term disease management. This modification increases potency against MM tumor cell lines with both normal and reduced BCMA expression, demonstrating superior metabolic endurance and in vivo activity.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Despite significant advances in treatment, multiple myeloma remains an incurable disease after chimeric antigen receptor (CAR)-T cell therapy, with most patients eventually experiencing relapses and disease progression. Key factors contributing to relapses include decreased or absent expression of B-cell maturation antigen (BCMA), where tumor cells downregulate or lose BCMA expression; CAR-T cell exhaustion, resulting in diminished functionality and persistence; and the immunosuppressive tumor microenvironment, which inhibits CAR-T trafficking and cytotoxic activity.

WHAT THIS STUDY ADDS

• ARI2h-TM28 improves ARI0002h (Cesnicabtagene-autoleucel) antitumor efficacy providing higher sensitivity to low target antigen expression, a more energetic metabolic profile, and a less inflammatory profile phenotype that may reduce exhaustion and increase persistence.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Our study shows a relatively small modification of the CAR structure, in this case, the transmembrane domain, can increase the CAR efficacy and improve tumor control, demonstrating a viable strategy to improve clinically our CAR-T cell products.

Introduction

Multiple myeloma (MM) is a hematologic malignancy characterized by the uncontrolled proliferation of plasma cells within the bone marrow.¹ Despite significant advances in treatment, MM remains an incurable disease, with most patients eventually experiencing relapse and disease progression.2,4 In recent years, anti-B-cell maturation antigen (BCMA) chimeric antigen receptor (CAR)-T cell therapy has revolutionized the treatment landscape for relapsed or refractory MM, offering a novel and highly effective therapeutic approach.⁵

Two CAR-T products targeting BCMA have been approved by both the European Medicines Agency (EMA) and Food and Drug Administration: idecabtagene vicleucel (Ide-cel) and ciltacabtagene autoleucel (Cilta-cel). Both are second-generation CAR-T therapies incorporating 4-1BB costimulatory domains to enhance T-cell activation, proliferation, and persistence, and both share a CD8α transmembrane (TM) domain. In pivotal clinical trials, Ide-cel achieved a median progression-free survival (PFS) of 8.8 months and a median overall survival (OS) of 24.8 months. Its use was associated with significant toxicity, with 84% of patients experiencing cytokine release syndrome (CRS), although mostly grade 1–2 with only 5% developing grade 3 or higher CRS, and 3% developing grade 3 or 4 neurotoxic effects in the form of immune effector cell-associated neurotoxicity syndrome (ICANS).⁶ Similarly, Cilta-cel demonstrated high efficacy, with a median PFS of 34.9 months, median OS of 60.7 months and a toxicity profile including 95% CRS (though only 4% with grade 3 or higher) and 10% grade 3 or higher ICANS.7,9 In addition, our institution developed ARI0002h (Cesnicabtagene-autoleucel), an academic anti-BCMA CAR-T for relapsed/refractory MM.¹⁰ In a recent Phase I/II trial, ARI0002h demonstrated a median PFS of 20.0 months, with a 63% complete response rate and a 3-year OS of 63%. The safety profile was favorable, with 90% of patients experiencing CRS, though most cases were mild (grade 1–2), and only 3% developing grade 1 ICANS.10,13

Despite these advances, most patients eventually relapse. Key factors contributing to relapse include antigen escape, where tumor cells downregulate or lose BCMA expression14,17; CAR-T cell exhaustion, resulting in diminished functionality and persistence¹⁸; and the immunosuppressive tumor microenvironment, which inhibits CAR-T trafficking and cytotoxic activity.19,21 These challenges underscore the urgent need for further optimization of CAR-T therapy, including structural improvements, to enhance efficacy and overcome resistance mechanisms.

To address these challenges, numerous strategies are being explored, including the development of combinatorial approaches and structural modifications to CAR design. While most optimization efforts have focused on the costimulatory domain, less attention has been given to the TM domain, which may hold significant potential for improving CAR functionality.^{19 22}

The TM domain is a key structural component that anchors the CAR in the T-cell membrane, ensuring proper receptor stability and clustering.²³ Emerging evidence suggests that the TM domain can influence CAR signaling efficiency, persistence, and overall durability, providing a promising but underexplored avenue for CAR optimization.²⁴

In the lymphoma setting, three CAR-T products have been approved in different clinical trials: axicabtagene ciloleucel (Axi-cel), tisagenlecleucel (Tisa-cel), and lisocabtagene maraleucel (Liso-cel), each with distinct structural features, particularly in their TM and costimulatory domains despite sharing the same single-chain variable fragment (scFv).²⁵ Axi-cel uses a CD28 TM domain combined with a CD28 costimulatory domain, offering high efficacy but at the cost of increased toxicity, with a high incidence of severe CRS and ICANS.²⁶ Tisa-cel, on the other hand, incorporates a CD8α TM domain with a 4-1BB costimulatory domain, resulting in a better safety profile, including reduced rates of CRS and ICANS.²⁷ However, its efficacy data are less robust compared with the one observed in clinical trials with Axi-cel. Liso-cel takes a hybrid approach between Axi-cel and Tisa-cel by combining the CD28 TM domain with a 4-1BB costimulatory domain, achieving efficacy comparable to Axi-cel but with significantly lower toxicity, resembling the safety profile of Tisa-cel.²⁸ Although only Tisa-cel and Axi-cel have been compared head-to-head in retrospective studies,²⁹ and thus under the same conditions, the results obtained on their respective clinical trials highlight their structural diversity and the importance of the TM domain in balancing efficacy and safety in CAR-T therapies.

In this context, we hypothesized that replacing the CD8α TM domain of ARI0002h with the CD28 TM domain could enhance CAR activation and functionality, particularly under conditions of low BCMA expression.

In this study, we present the development and validation of a novel CAR-T product, ARI2h-TM28, which incorporates a CD28 TM domain to enhance its functionality. Our findings demonstrate that ARI2h-TM28 exhibits comparable CAR expression to the original ARI0002h and maintains similar killing capacity, cytokine production, and proliferation potential against tumor cells with standard BCMA expression. Importantly, ARI2h-TM28 outperforms ARI0002h in tumor control and significantly improves survival in preclinical models with both standard and reduced BCMA expression, addressing a critical challenge of antigen escape. Furthermore, ARI2h-TM28 displays superior metabolic fitness, lower exhaustion profiles post-rechallenge, and sustained activation, which collectively support its enhanced functional persistence and cytotoxicity. Our transcriptomic analysis reveals that ARI2h-TM28 is optimized for proliferation, metabolic activity, and improved immunologic synapse signaling while downregulating inflammatory and immune signaling pathways, contributing to its robust antitumor response. Taken together, these findings position ARI2h-TM28 as a promising therapeutic option for overcoming resistance mechanisms in MM, offering improved efficacy and durability even under challenging conditions of low antigen expression.

Materials and methods

Generation of CAR-expressing primary human T cells

Peripheral blood mononuclear cells (PBMCs) were isolated from buffy coats obtained from healthy donors through the Blood and Tissue Bank of Barcelona (Banc de Sang i Teixits) using Ficoll density gradient centrifugation. T cells were purified from PBMCs by magnetic-activated cell sorting using the Pan T-cell Isolation Kit (Miltenyi Biotec) in accordance with the manufacturer’s protocol. For activation, the isolated T cells were cultured with CD3/CD28 Dynabeads (Life Technologies). Lentiviral transduction of T cells was performed 48 hours after activation. At day 7 after activation, Dynabeads were removed, and T cells were expanded in complete T-cell medium comprising RPMI and Click’s medium (50%/50%), supplemented with 5% AB human serum (Sigma, H4522), penicillin-streptomycin (100 µg/mL), and 100 IU of interleukin-2 (IL-2) added every 2–3 days. CAR expression on T cells was analyzed via flow cytometry following 8–10 days of expansion.

Multiparameter flow cytometry

CAR expression was assessed by labeling T cells with BCMA-Fc recombinant protein (Enzo) and anti-human Fc BV421 antibodies (BioLegend). Additional phenotypic analyses were conducted for CAR-T cells and CRISPR/Cas9-modified tumor cell lines using panels targeting markers for T-cell subsets (eg, CD4, CD8), activation (eg, CD25, CD71), exhaustion (eg, programmed cell death protein-1 (PD-1), TIM3, LAG3, TIGIT), and memory compartment differentiation (eg, CCR7, CD45RA). Dead cells were excluded by forward/side scatter gating and staining with dyes such as propidium iodide or LIVE/DEAD Aqua (Thermo Fisher Scientific). Samples were analyzed using BD FACSCanto II (BD Biosciences) or BD LSRFortessa 5 L (BD Biosciences) cytometers, and FlowJo software (V.10) was used for data analysis.

CAR-T cell expansion

CAR-T cells were cultured in Click’s medium supplemented with 5% AB human serum, 1% glutamine, and 1% penicillin-streptomycin. IL-2 (100 IU/mL) was added every 2–3 days. Cell concentrations were maintained below 1×10⁶ cells/mL, and expansion continued for 7–8 days post-transduction.

Functional assays

In vitro assays involved challenging ARI0002h and ARI2h-TM28 CAR-T cells with MM cell lines (U266, MM.1S, ARP-1, and knockout cell lines) and a negative control with K562 lymphoblastic cell line. Experiments were performed normalizing to CAR+viable T cells.

Cytotoxicity assay

Cytolytic activity was evaluated using a bioluminescence-based method. CAR-T cells were co-cultured with 5×10³ ffLuc-expressing target cells at various effector-to-target (E:T) ratios for 4 to 24 hours. Before measurement, 0.15 mg/mL D-luciferin (Biosynth) was added to the cultures, and bioluminescence was quantified using an Infinite 200 Pro microplate reader (Tecan). Specific lysis was calculated as:

% specific lysis = [(mean bioluminescence of control − mean bioluminescence of sample)/mean bioluminescence of control] × 100.

Proliferation assay

CAR-T cells were labeled with 1 µM CellTrace Far Red (Thermo Fisher Scientific) and co-cultured with different target cells in triplicate for 72 hours in cytokine-free medium. Cell division was assessed via flow cytometry.

Cytokine production

Supernatants were collected from 24-hours co-cultures of CAR-T and target cells (E:T ratio 4:1) and analyzed for cytokine secretion using ELISA kits (BioLegend) and the Cytokine 10-Plex Human Panel (Invitrogen), according to manufacturer protocols and specifications.

Rechallenge assay

CAR-T cells were produced and expanded exactly as described in methods section 1.1 and 1.3 to maintain their characteristics as similar as possible across all different experiments. Afterward, these CAR-T cells were co-cultured with ARP-1 GFP+ target cells at a 1:2 E:T for 48 hours in a 1:1 mix of RPMI 10% and media for T cells (as described in online supplemental methods), after which the remaining CAR-T cells were rechallenged with fresh target cells under identical conditions, maintaining the 1:2 E:T ratio. This process was repeated for up to four rounds. Cytotoxic activity and T-cell expansion were analyzed by flow cytometry after each challenge.

Pre-rechallenge and post-rechallenge activation, exhaustion and T-cell subset flow cytometry staining

T-cells were analyzed both pre-rechallenge and post-rechallenge by flow cytometry using different antibody panels to characterize their phenotype. Panels focused on exhaustion markers (LAG3, TIM3, TIGIT, PD-1), activation (CD25, CD71, CD127) and their T-cell memory compartment (CCR7, CD45RA), aiming to differentiate their fitness and phenotype.

RNA extraction, sequencing and analysis

Post-rechallenge, CD8+ CAR T cells were isolated using a magnetic bead-based CD8 Positive Selection Kit (EasySep). RNA was extracted using the RNeasy Mini Kit (Qiagen). Paired-end sequencing was then performed on a NovaSeq X Plus Series instrument. FastQC (V.0.11.9) was used to analyze quality control of paired-end reads. Kallisto (V.0.46.1) was used to map reads to the human genome (GRCh38). DESeq2 (V.1.44.0) was used for differential expression analyses. The Gene Ontology, Hallmark, C5 and C7 databases from the Molecular Signature Database were used for gene set enrichment analyses using ClusterProfiler (V.4.12.6)³⁰ with a p.adj value of 0.05 for significant enrichment. Downstream analyses were performed using R (V.4.4.0, R Foundation for Statistical Computing, Vienna, Austria).

In vivo xenograft studies in NSG mice

All in vivo experiments were approved by the University of Barcelona Animal Research Committee.

Animal studies were conducted in accordance with the ARRIVE (Animal Research: Reporting of In Vivo Experiments) Guidelines 2.0 ³¹ .

Animal studies were conducted in 8–12 weeks old male and female NSG mice (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ). MM.1S or MM.1S BCMAdim cells expressing GFP-ffLuc were intravenously injected at day 0 into mice irradiated (1 Gy) the previous day (day –1). On day 14 post-tumor inoculation, mice were randomized to receive either untransduced (UTD) T cells or CAR-T cells intravenously. Tumor burden was monitored weekly by bioluminescence imaging using the Perkin Elmer IVIS Spectrum In Vivo Imaging System, with the first bioluminescence image taking place less than a week after tumor infusion. After mice reached either a humanitarian endpoint condition or bioluminescence measured reached the bioluminescence sacrifice threshold (≥5×10⁸ p/sec/cm2/sr), the mice were humanely euthanized.

Criteria for application of corrective measures or humane mice euthanasia, depending on the degree of severeness, included weight loss, bad posture, bad aspect of fur coat and feces, nasal and ocular secretions, pain vocalizations, abnormal posture (hunched), self-mutilations, immobile or lethargic behavior, abnormal aggressiveness or low/reduced response to stimuli. If corrective measures (when applicable) did not cause an improvement, mice were humanely sacrificed by cervical dislocation regardless of reaching the bioluminescence threshold of the experiment.

For rechallenge experiments, a second tumor dose of MM.1S BCMA dim cells was administered on day 28 to all mice that had been previously administered CAR-T cells (ARI0002h and ARI2h-TM28) and to a further group that acted as control of the newly infused line, but not to the UTD group. The second dose was administered once the mice previously infused with CAR-T cells showed no tumor signal by bioluminescence, to mimic relapse of a clone with reduced BCMA antigen expression. CAR-T efficacy was evaluated post-rechallenge through weekly bioluminescence imaging using the Perkin Elmer IVIS Spectrum In Vivo Imaging System.

Bioenergetics assay

Both CAR-T cells were bioenergetically characterized on day 11 of manufacturing. T cells were derived from four different donors. Bioenergetic measurements were performed according to the manufacturer’s protocol. Briefly, cells were resuspended in XF Assay Medium (Agilent) (RPMI supplemented with 0.01% glucose, glutamine, and pyruvate). The cells, ranging from 140,000 to 180,000 per well (approx. 3.5 Mio cells/ml, or 175,000 cells per condition), were plated onto the plate in a monolayer and subjected to the T-cell persistence assay protocol (Agilent) using the Seahorse XF HS Mini Analyzer (Agilent) with Oligomycin, BAM15 and Rotenone C inhibitors.

Statistical analysis

Data visualization and statistical analyses were performed using GraphPad Prism (V.8.2.1). No prespecified effect size was used to determine sample sizes. Results are presented as mean±SD unless stated otherwise. The statistical tests that were used to calculate the p values are described in the relevant figure legends. Statistical significance was set at p<0.05 and identified as follows; p≥0.05, not significant, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Further details of the methods are provided in the online supplemental methods.

Results

Development and validation of an anti-BCMA CAR-T cell with a CD28 TM domain

The ARI2h-TM28 was generated using the anti-BCMA ARI0002h CAR structure as base, maintaining the scFv, hinge, 4-1BB costimulatory domain and CD3ζ activation domain, but replacing the CD8α TM domain with the CD28 TM domain (figure 1A). The transduction efficiency of the new ARI2h-TM28 CAR was similar to that of the ARI0002h CAR, ranging from 65–75% efficiency (figure 1B).

Figure 1. Development of a second-generation anti-BCMA CAR-T T cell with a CD28 transmembrane domain. (A) The sequence of the second-generation ARI0002h CAR was used to generate the new ARI2h-TM28 CAR replacing the CD8α transmembrane domain with the CD28 transmembrane domain. The rest of the CAR structure was left unaltered. (B) Detection of CAR protein on the cell surface on BV421 channel. Both ARI0002h and ARI2h-TM28 CARs were stained using antibodies conjugated with BV421 fluorochromes and showed similar expression on the surface of T-cells. Analyses were performed by flow cytometry. (C) Specific cytolytic activity of ARI0002h and ARI2h-TM28 CAR-T cells co-cultured during 24 hours with K562 (BCMA−), U266 (BCMA+) and MM.1S (BCMA+) tumor cells with different E:T ratios (5:1 to 0.6:1) obtained in n=6 experiments with CAR-T cells manufactured from n=6 different donors. The killing capacity was evaluated in a bioluminescence-based cytotoxicity assay. (D) Cytokine production (IFNγ, IL-2 (n=6) and GM-CSF (n=3)) of CAR-T cells in co-culture with K562, MM.1S and U266 cells at the 24 hours time point measured by ELISA. Supernatants obtained from triplicate wells and T-cells obtained from different donors. Statistical differences were determined by Kruskal-Wallis test with Dunn’s test for multiple comparisons. (E) CAR-T cell proliferation when co-cultured for 72 hours with BCMA-positive MM.1S cells, only medium (negative control), and medium+IL-2 (positive control) as determined by dilution of FAR RED Cell Trace, measured in the APC channel. For the analysis triplicate wells were pooled and the proliferation of live (7-AAD⁻) T cells was analyzed. Panels show representative flow cytometry images. (F) In vivo experimental design using NSG NOD mice. Mice were irradiated on day 1, then injected with 4×10⁶ GFP-ffLuc-expressing MM.1S cells. On day 14, mice were randomized to treatment with 1.0×10⁶ effector cells. (G and H) Tumor progression was measured by taking weekly bioluminescence images. Quantification of tumor burden by bioluminescence imaging over time of tumor cell infusion. Lines show the median and the 95% CI from all mice in that group. (I) Kaplan-Meier survival curve. Differences in survival were determined by log-rank test. (J) Plot represents average radiance and SD measured at week 5. Statistical significance was calculated using the Kruskal-Wallis test. P values for all panels were calculated by Mann-Whitney two-tailed test unless specified; *p<0.05; **p<0.01, ***p<0.001, ****p<0.0001. ARI0002h, Cesnicabtagene-autoleucel; BCMA, B-cell maturation antigen; CAR, chimeric antigen receptor; E:T, effector-to-target; GM-CSF, granulocyte-macrophage colony-stimulating factor; IFN-γ, interferon-gamma; IL-2, interleukin-2; n.s., not statistically significant; UTD, untransduced.

We evaluated the in vitro efficacy of the new ARI2h-TM28 CAR by comparing it with the ARI0002h against myeloma cell lines expressing BCMA antigen (U266 and MM.1S) and a cell line not expressing BCMA (K562). Both anti-BCMA CAR-T cells showed comparable cytotoxic activity against U266 and MM.1S cell lines with a dose–response relationship observed across different E:T almost identical, while showing no killing of the BCMA-negative K562 cell line (figure 1C). Both CAR-T cells secreted comparable concentrations of IL-2, interferon-gamma (IFN-γ), IL-6 as well as several other cytokines (figure 1D, online supplemental figure 1) and showed a similar proliferation capacity (figure 1E, online supplemental figure 2), with no statistical differences between ARI0002h and ARI2h-TM28 when co-cultured with tumor cells positive for BCMA (U266 and MM.1S).

We conducted a xenograft model using NSG mice with the MM.1S cell line to assess the efficacy of both CAR-T cells in an in vivo model. Mice were infused with 4×10⁶ tumor cells and randomized to receive either 1×10⁶ UTD T lymphocytes, ARI0002h or ARI2h-TM28 CAR-T cells 14 days later (figure 1F, online supplemental figure 3). We did not observe significant weight loss or other toxicities in most treated mice that was not attributable to tumor progression-induced cachexia, except in one case in which the mouse was humanely euthanized. Mice were monitored weekly using bioluminescence imaging (figure 1G), which showed a faster disease progression in the UTD group while ARI0002h and ARI2h-TM28 showed early control of the tumor (figure 1G,H). However, the most effective disease control and survival advantage were observed in the ARI2h-TM28 group, which showed a significant superior survival at the end of the experiment (figure 1I) and a significantly superior tumor control in some early weeks of the experiment (figure 1J). These findings prompted our group to investigate which mechanisms were responsible for this better performance in an in vivo setting.

ARI2h-TM28 maintained cytotoxic activity and did not show an increased exhaustion or reduced activation profile after repeated challenges with target cells expressing BCMA

In order to study the consistency of the cytotoxic capacity, the persistence, exhaustion, and T-cell subsets of ARI0002h and ARI2h-TM28, both anti-BCMA CAR-T cells were challenged with BCMA expressing ARP-1 cells at a 1:2 E:T ratio for 48 hours, repeating the challenge up to four times in total after assessing a killing capacity superior to 95% of the target cells throughflow cytometry (figure 2A).

Figure 2. Analyzing proliferation, activation, exhaustion and T-cell subset after repeated challenges with target cells. (A) Rechallenge experiment. ARI0002h and ARI2h-TM28 CAR-T cells were co-incubated at a 1:2 E:T ratio for 48 hours with ARP-1 target cells, before assessing the killing of the target cells by flow cytometry. Afterward, the experiment was set up again with the same conditions up to three more times. (B) T-cell expansion measured by population doublings after repeated challenges against target cells expressing BCMA (ARP-1). (C) Specific cytolytic activity of ARI0002h and ARI2h-TM28 CAR-T cells before rechallenges, co-cultured during 24 hours with K562 (BCMA−), MM.1S (BCMA+) and ARP-1 (BCMA+) tumor cells with different E:T (5:1 to 0.6:1) obtained in n=8 experiments with CAR-T cells manufactured from n=8 different donors. The killing capacity was evaluated in a bioluminescence-based cytotoxicity assay. (D) Specific cytolytic activity of ARI0002h and ARI2h-TM28 CAR-T cells after rechallenges, co-cultured during 24 hours with K562 (BCMA−), MM.1S (BCMA+) and ARP-1 (BCMA+) tumor cells with different E:T ratios (5:1 to 0.6:1) obtained in n=7 experiments with CAR-T cells manufactured from n=7 different donors. The killing capacity was evaluated in a bioluminescence-based cytotoxicity assay. (E) Expression of several exhaustion markers (LAG3, TIM3, TIGIT, PD-1) in ARI0002h and ARI2h-TM28 before (left) and after rechallenging (right) with tumor cells (n=8). (F) Expression of several activation markers (CD127, CD71, CD25) in ARI0002h and ARI2h-TM28 before (left) and after rechallenging (right) with tumor cells (n=7). (G) Mean of each T-cell subset before (n=6) and after rechallenges (n=7) according to CCR7 and CD45RA expression, where naive (CCR7+, CD45RA+), effector (CCR7−, CD45RA+), effector memory (CCR7−, CD45RA−) and central memory (CCR7+, CD45RA−). P values for all panels were calculated by Mann-Whitney two-tailed test unless specified; *p<0.05; **p<0.01, ***p<0.001, ****p<0.0001. ARI0002h, Cesnicabtagene-autoleucel; BCMA, B-cell maturation antigen; CAR, chimeric antigen receptor; E:T, effector-to-target; n.s., not statistically significant; PD-1, programmed cell death protein-1.

The repeated challenges showed no significant differences in CAR-T cell expansion between both CAR-Ts (figure 2B, online supplemental figure 4) after the four challenges. The cytotoxicity assays performed pre-rechallenge and post-rechallenge showed no differences between ARI0002h and ARI2h-TM28, both in target cells with (ARP-1, MM.1S) or without (K562) BCMA expression and followed a similar dose–response relationship across different E:T (figure 2C,D). At the same time, no differences were observed between ARI0002h and ARI2h-TM28 in cytokine secretion (IFNγ, IL-2 and granulocyte-macrophage colony-stimulating factor (GM-CSF)) after the challenges (online supplemental figure 5). The percentage of CAR+ T cells post-rechallenge stayed constant or even slightly increased after the rechallenges, oscillating between 77–85% for ARI0002h and 65–80% for ARI2h-TM28 across all different donors.

CAR-T cells were stained for exhaustion, activation, and T-cell subset markers before and after rechallenges. Pre-rechallenge, PD-1, TIGIT, and LAG-3 observed expression appeared slightly higher in ARI2h-TM28 CD4 and CD8 T cells, but these differences were not statistically significant. TIM-3 was slightly more expressed in ARI0002h, also without significance. Post-rechallenge, expression levels of PD-1 and TIGIT in ARI0002h matched those of ARI2h-TM28, and LAG-3 was slightly higher in ARI0002h CD8 cells. TIM-3 remained comparable. These trends are shown in figure 2E and online supplemental figure 6, and we note that they should be interpreted cautiously and validated in future experiments.

Regarding activation markers, before the rechallenge, CD25 and CD71 were higher in ARI2h-TM28, with CD25 reaching statistical significance. CD127 was slightly higher in ARI0002h compared with ARI2h-TM28, but not significantly. After rechallenge, activation marker levels were similar between both CARs, including CD25, with minimal variations (figure 2F, online supplemental figure 7).

T-cell subsets were analyzed according to their CCR7 and CD45RA expressions. Pre-rechallenge, both CARs showed a predominance of effector memory cells in CD4, and effector cells in CD8. Central memory was slightly higher in ARI0002h compared with ARI2h-TM28, but not significantly. Post-rechallenge, naive and effector populations decreased in favor of effector memory in both CARs, both in CD4 and CD8, with slightly higher percentages in ARI0002h while central memory populations were maintained after the challenges in both CARs. Effector populations were slightly more abundant in ARI2h-TM28 CD8 cells (up to 35%), but again, none of these differences were statistically significant (figure 2G).

ARI2h-TM28 shows a more energetic and oxidative phenotype compared with ARI0002h

To study the metabolic profile of both CAR-T cells, we used the Persistence Assay from SeaHorse (Agilent). Regarding the oxidative metabolism, the oxygen consumption ratio (OCR) was significantly increased in ARI2h-TM28 compared with ARI0002h during rest and after culture with Oligomycin A and BAM15 inhibitors, showing a more oxidative phenotype (figure 3A). The anaerobic metabolism was also consistently increased in ARI2h-TM28, although not significantly, as measured by the extracellular acidification rate, showing a more energetic metabolism of ARI2h-TM28, both oxidative and anaerobic (figure 3B). The assay showed that in line with the higher energy production of ARI2h-TM28, it also retained a significantly higher spare respiratory capacity compared with ARI0002h (figure 3C).

Figure 3. Metabolomics study demonstrates ARI2h-TM28 possesses a more energetic phenotype compared with ARI0002h. (A) Measurement of OCR, indicating the activation of the oxidative phosphorylation pathway, after the administration of Oligomycin A, BAM15 and Rotenone inhibitors for both ARI0002h and ARI2h-TM28. (B) Measurement of ECAR, indicating activation of glycolysis, after the administration of Oligomycin A, BAM15 and Rotenone inhibitors for both ARI0002h and ARI2h-TM28. (C) SRC of both CARs during the persistence assay, according to the OCR of each CAR. (D) % ATP production obtained from glycolysis. All experiments were performed using SeaHorse Persistence protocol, by Agilent. (E) Energetic map showing the energetic phenotype of ARI0002h and ARI2h-TM28 according to their production of ATP from oxidative phosphorylation and glycolysis. P values for all panels were calculated by Mann-Whitney test unless specified; *p<0.05; **p<0.01, ***p<0.001, ****p<0.0001. ARI0002h, Cesnicabtagene-autoleucel; CAR, chimeric antigen receptor; n.s., not statistically significant.

The ATP production was higher in the ARI2h-TM28 CAR compared with the ARI0002h CAR, and as the OCR already indicated, there was a significant difference in the mitochondrial ATP production, where ARI2h-TM28 had a higher mitochondrial ATP output compared with ARI0002h (online supplemental figure 8). In contrast, and despite a higher ATP production tendency from glycolysis in ARI2h-TM28, there were no significant differences in ATP production with glycolysis between both CARs (figure 3D,E). To conclude, ARI2h-TM28 presents a more energetic and aerobic metabolism compared with ARI0002h, while also being less reliant on glycolysis to produce the energy it needs (figure 3E).

ARI2h-TM28 has a less inflammatory transcriptional profile compared with ARI0002h, while increasing expression of immune synapse genes

After four challenges with ARP-1 tumor cells, CD8+ T cells were isolated using magnetic separation and their RNA was extracted and sequenced by bulk RNA-Seq. The RNA-seq analysis showed 585 differentially expressed genes by p value (<0.05). ARI2h-TM28 downregulated genes like SOCS1, CSF2 (GM-CSF) and LIF, and upregulated like LIME1, CNOT6L and UHRF1. The CD28 gene was also upregulated but showed a non-significant adjusted p value (figure 4A, online supplemental figure 9). Analysis against the Gene Ontology database showed that gene sets related to cytokine activity, cytokine signaling pathways and receptor signaling activity were downregulated in ARI2h-TM28 CD8+ cells, while genes related to the T-cell receptor complex were upregulated (figure 4B). Further analysis with the Hallmark and C2 curated databases showed that several proinflammatory gene sets were downregulated in ARI2h-TM28 compared with ARI0002h, including ones related to IFN-γ, IFN-α, IL-2 and IL-6, as well as cytokine interaction pathways. In contrast, T-cell differentiation genes were upregulated in ARI2h-TM28, showing ARI2h-TM28 presents a less inflammatory transcriptomic profile compared with ARI0002h (figure 4C,D, online supplemental figures 10 and 11).

Figure 4. mRNA sequencing of CD8+ CAR T cells shows differences in expression in T-cell differentiation and inflammatory pathways between ARI2h-TM28 and ARI0002h. (A) Volcano plot showing differentially expressed genes between ARI2h-TM28 and ARI0002h, with genes UP in ARI2h-TM28 to the right and upregulated in ARI0002h to the left. Most relevant genes highlighted with arrows. (*The gene at 5 of −log₁₀P and 5 log₂ Fold Change corresponds to a lncRNA with no clear function described in gene databases). (B) Dot plot representing significantly enriched pathways, according to GO database. (C) GSEA representing significantly enriched pathways according to Hallmark in ARI0002h: IFN-γ Response (p.adj-value: 3.764e⁻⁰⁵), IL-2 STAT5 Signaling (p.adj-value: 5.502e⁻⁰⁵), IL-6 JAK-STAT3 signaling (p.adj-value: 0.00092) and MYC Target genes (p.adj-value: 0.0131). (D) GSEA representing significantly enriched pathways according to C2 curated databases in ARI0002h or in ARI2h-TM28: Differentiating T lymphocyte-related genes (p.adj-value: 0.001736), Inflammatory response LPS UP (p.adj-value: 2.455e⁻⁰⁵), Cytokine-cytokine-receptor interaction (p.adj-value: 0.0001823) and Response to IFN-γ UP (p.adj-value: 0.02019). ARI0002h, Cesnicabtagene-autoleucel; CAR, chimeric antigen receptor; GO, Gene Ontology; GSEA, gene set enrichment analysis; IFN-γ, interferon-gamma; IL-2, interleukin-2; lncRNA, Long non-coding RNA; LPS, Lipopolysaccharides;mRNA, messenger RNA; NES, Normalized Enrichment Score; UP, upregulated.

ARI2h-TM28 demonstrates better capacity compared with ARI0002h to recognize and eliminate BCMA dim tumor cells in vitro and in vivo

We generated an MM.1S cell line with diminished expression of BCMA to test if ARI2h-TM28 presented a lower activation threshold compared with ARI0002h (figure 5A). We evaluated the in vitro efficacy of the new ARI2h-TM28 CAR compared with ARI0002h against this new BCMA dim cell line and observed ARI2h-TM28 showed a significantly superior cytotoxic effect on the BCMA dim cell line, together with the expected dose-response relationship across different E:T (figure 5B).

Figure 5. ARI2h-TM28 shows a higher killing capacity in vitro and better in vivo tumor control against a cell line with diminished BCMA expression. (A) BCMA antigen expression of MM.1S cell lines, analyzed by flow cytometry. (B) Specific cytolytic activity of ARI0002h and ARI2h-TM28 CAR-T cells co-cultured during 24 hours with MM.1S (BCMA+) and MM.1S BCMA dim (BCMA diminished) tumor cells with different E:T (5:1 to 0.6:1) obtained in n=3 experiments with CAR-T cells manufactured from n=3 different donors. The killing capacity was evaluated in a bioluminescence-based cytotoxicity assay. (C) BCMA antigen expression of MM.1S cell lines, including a new batch of MM.1S BCMA dim, analyzed by flow cytometry. (D) Specific cytolytic activity of ARI0002h and ARI2h-TM28 CAR-T cells co-cultured during 24 hours with MM.1S (BCMA+) and MM.1S BCMA dim (BCMA diminished) tumor cells with different E:T (5:1 to 0.6:1) obtained in n=5 experiments with CAR-T cells manufactured from n=5 different donors. The killing capacity was evaluated in a bioluminescence-based cytotoxicity assay. (E) In vivo experimental design using NSG NOD mice. Mice were irradiated on day –1, then injected with 2×10⁶ GFP-ffLuc-expressing MM.1S cells. On day 14, mice were randomized to treatment with 4.0×10⁶ UTD, ARI0002h, or ARI2h-TM28 effector cells. On day 28, after tumor control was confirmed by bioluminescence and before week 4 BLI, all mice in the ARI0002h group (10 mice), ARI2h-TM28 group (11 mice), and a new BCMA-dim control group (irradiated at 1 Gy the day before) were injected with 2×10⁶ GFP-ffLuc-expressing MM.1S BCMA Dim cells to simulate the emergence of a tumor clone with reduced BCMA expression due to immune pressure exerted by the CAR-T cells. (F) Tumor progression was monitored weekly by bioluminescence imaging. Dotted lines indicate the infusion of MM.1S tumor cells (red), CAR-T cells (blue), and MM.1S BCMA-dim cells (green). (G) Quantification of tumor burden over time. Lines show the median and 95% CI for each group. Dotted lines indicate infusion timepoints as described above. (H) Kaplan-Meier survival curve. Differences in survival were determined by log-rank test. Dotted lines indicate infusion time points as described above. (I) Plot represents average radiance and SD measured in week 8. P values for all panels were calculated by Mann-Whitney two-tailed test unless specified unless specified; *p<0.05; **p<0.01, ***p<0.001, ****p<0.0001. ARI0002h, Cesnicabtagene-autoleucel; BCMA, B-cell maturation antigen; BLI, bioluminescence imaging; CAR, chimeric antigen receptor; E:T, effector-to-target; MFI, Mean Fluorescence Intensity; n.s., not statistically significant; UTD, untransduced.

We subcloned the MM.1S BCMA dim cell line to obtain even a lower BCMA expression and tested it with both CAR-T cells as done previously (figure 5C). We observed a reduced cytotoxic effect at every E:T for both CARs but still maintained the dose–response relationship observed across different E:T in all previous experiments, while ARI2h-TM28 still showed a significant cytotoxic advantage compared with ARI0002h (figure 5D).

To evaluate CAR-T cell performance in a model that closely mimics the immune pressure exerted by the CAR-T cells that results in a downregulation of the target antigen expression, we generated another in vivo model. In this model, after the usual initial infusion of 2×10⁶ target cells expressing BCMA and the 4×10⁶ CAR-T cells 14 days after, once the CAR-T cells controlled the initial growth, 2×10⁶ MM.1S BCMA dim cells were infused to all the mice in ARI0002h (10 mice) and ARI2h-TM28 (11 mice) groups to simulate the appearance of a subclone with diminished BCMA expression (figure 5E, online supplemental figure 12), together with the infusion of the BCMA dim cells to a newly generated control group. Bioluminescence imaging showed that despite the initial tumor control demonstrated by both CARs, after the second infusion neither ARI2h-TM28 nor ARI0002h was able to control tumor progression completely, but ARI2h-TM28 showed a slower progression of the tumor measured by bioluminescence (figure 5F,G). Most importantly, ARI2h-TM28 showed a significantly superior survival compared with ARI0002h, which had a survival closely mirroring that of the BCMA Dim Control group (figure 5H), as well as a significantly superior early tumor control in week 8, 4 weeks after the BCMA dim tumor infusion (figure 5I).

Discussion

The TM domain of the CAR in CAR-T cell design has traditionally been viewed as a structural anchor for receptor stability. However, its potential role in modulating CAR signaling and enhancing functionality remains largely unexplored, particularly in the context of relapse mechanisms in MM, such as antigen escape due to a decreased expression of BCMA. Addressing this knowledge gap, our study investigated the impact of replacing the CD8α TM domain in ARI0002h with a CD28 TM domain, hypothesizing that this modification could enhance CAR-T efficacy against BCMA-diminished surface expression on malignant cells. In this study, ARI2h-TM28 demonstrated superior tumor control and a significant survival advantage in preclinical models, even under conditions of reduced BCMA expression.

Target antigen density has emerged as a critical determinant of CAR-T cell efficacy, with diminished antigen expression posing a significant challenge to sustained therapeutic responses. Only one prior study²² suggested that CARs incorporating the CD28 TM domain exhibit enhanced reactivity against low-antigen-density targets, outperforming those with CD8α TM domains in both in vitro and in vivo settings.

From previous in vivo experiments in our group,³² our data show some cases of reduction or loss of BCMA expression in mice infused with MM tumor cells and treated with ARI0002h CAR-T cells while other mice that were not able to control the tumor maintained expression of BCMA. So, we expect BCMA reduction or loss to be a distinct possibility in our experiments once the tumors escape control by the CAR-T cells.

In our study, ARI2h-TM28, which integrates a CD28 TM domain, exhibited superior tumor control and survival compared with ARI0002h, even in models with reduced BCMA expression. Although with our model we cannot determine the exact mechanism of relapse (antigen escape by loss or reduction of BCMA,^{14 15} CAR-T cell exhaustion,¹⁸ loss of CAR-T persistence^{33 34} and immunosuppressive tumor microenvironment,²⁰ among others), our data support that under identical conditions in a model that mimics BCMA downregulation after CAR exposure, ARI2h-TM28 showed a superior tumor control and survival compared with ARI0002h. All together, these results suggest that the CD28 TM domain might enhance receptor stability and clustering, allowing for effective signal transduction under suboptimal target antigen density by lowering the antigen density threshold required for CAR activation, requiring only minor structural modifications, as previous works suggest.²² By stabilizing the immunological synapse and amplifying proximal signaling, ARI2h-TM28 may overcome the limitations typically observed in conventional CAR-T cells when facing antigen escape or tumor heterogeneity.

The superior performance of ARI2h-TM28 in conditions of reduced BCMA expression may be partly explained by the dimerization of the CD28 TM domain. The TM domain is not merely a structural anchor; it actively influences CAR signaling by promoting receptor clustering and stabilizing the immunological synapse. Recent studies have identified conserved motifs within the CD28 TM domain, like GxxxA, which facilitates transmembrane helix-helix interactions and enhances receptor dimerization, and YxxxxT, which promotes dimer formation and signal transduction.^{35 36} This dimerization, which can occur with endogenous CD28 molecules, likely amplifies downstream signaling pathways, including PI3K-AKT and NF-κB, which are critical for sustained T-cell activation and functionality.³⁷ In the context of low-antigen-density tumors, the ability of the CD28 TM domain to stabilize receptor complexes may lower the activation threshold required for CAR signaling, improving CAR-T cell activation and persistence through receptor clustering and signal amplification.³⁷ This mechanism could explain the enhanced metabolic fitness, sustained activation and absence of exhaustion increase observed in ARI2h-TM28 compared with ARI0002h. Furthermore, by preventing premature internalization or degradation of the CAR,³⁸ dimerization may contribute to prolonged antigen engagement, improving the overall efficacy and durability of the therapy.

A critical factor influencing CAR-T cell efficacy is the ability to sustain activation and resist exhaustion, particularly under conditions of prolonged antigen exposure or diminished target antigen density. As seen in figure 2E,F, although most differences did not reach statistical significance, ARI2h-TM28 did not show increased exhaustion post-rechallenge and exhibited a significantly higher CD25 expression pre-rechallenge. These trends may suggest a more sustained activation profile, but further studies with larger sample sizes are needed to confirm these observations. Similarly, analysis of T-cell subsets on figure 2G indicated a possible skewing of ARI2h-TM28 cells toward an effector memory differentiation profile over central memory profile, particularly within CD8+ T cells. While these observations are descriptive, they may point to enhanced immediate cytotoxic potential and warrant further investigation in future studies with larger sample sizes.

This shift was complemented by superior metabolic reprogramming. Unlike ARI0002h, ARI2h-TM28 demonstrated a more robust glycolytic profile, facilitating energy production for rapid proliferation and effector functions, while also maintaining a superior mitochondrial activity to support oxidative phosphorylation.^{39 40} The integration of these metabolic and functional advantages enables ARI2h-TM28 to achieve better persistence and antitumor activity, even in metabolically challenging environments. These findings emphasize the importance of optimizing the CAR structure, not only for antigen recognition but also to balance the metabolic demands required for effective and sustained antitumor responses.⁴¹

The RNAseq analysis revealed significant transcriptional differences between ARI2h-TM28 and ARI0002h post-rechallenge, particularly in genes associated with inflammatory signaling and immune regulation. Notably, key genes such as SOCS1, CSF2 (GM-CSF), and LIF were downregulated in ARI2h-TM28, suggesting a less inflammatory and more functionally persistent CAR-T profile. SOCS1, a critical regulator of IL-2 and IFN-γ signaling, is known to suppress T-cell activation and effector functions. Its downregulation in ARI2h-TM28 likely enhances T-cell proliferation and survival effector function while increasing intratumoral accumulation in in vivo settings, as well as promoting the polyfunctionality of both CD4+ and CD8+ subsets, contributing to superior antitumor efficacy.^{42 43} Similarly, the downregulation of CSF2 reduces CAR-T-associated toxicities, such as CRS and neurotoxicity, without compromising antitumor activity.⁴⁴ This aligns with preclinical findings where GM-CSF disruption improved CAR-T cell proliferation, antitumoral activity and resistance to apoptosis by Fas ligand pathway.⁴⁵ Additionally, the downregulation of LIF in ARI2h-TM28 may diminish its protumorigenic effects, as LIF is implicated in promoting malignant plasma cell growth and drug resistance to IL-6 blockers through the gp130 receptor signaling, shared by both cytokines.⁴⁶ LIF deprivation also causes a lasting increase in T cell activation in rats.⁴⁷ By targeting these inflammatory pathways, ARI2h-TM28 would reduce its potential tumor-supportive interactions while also fostering a more activated and durable CAR-T cell phenotype.

In contrast, genes upregulated in ARI2h-TM28, such as CNOT6L (Ccr4b) or LIME1, were related to T-cell signaling, lymphocyte migration, and cell survival. CNOT6L mediates cell survival and inhibits cellular senescence, as well as controlling proliferation as part of the Ccr4 deadenylase complex which shortens the poly-A tail of mRNA.⁴⁸ LIME1 is expressed in B cells and T cells, where it integrates the T cell receptor signaling pathway downstream after T cell activation being essential in T cell migration and infiltration into the inflammatory site. At the same time, T cell activation in effector T cells upregulates LIME1 expression, which increases its chemokine-mediated migration to inflammatory sites in response to CCL5 and CXCL10 chemokines.⁴⁹ In preclinical studies with an anti-GPC2 CAR in neuroblastoma, the CAR construct that elicited the best results had CD28 hinge, CD28 transmembrane domain, and 4-1BB costimulatory domain. This CAR showed an upregulation of LIME1 expression that was accompanied by an enrichment of CD8+ effector T cells in the tumor microenvironment.⁵⁰ These transcriptional changes likely contribute to the reduced exhaustion and superior tumor control observed in ARI2h-TM28.

This study demonstrates that ARI2h-TM28, incorporating a CD28 TM domain, outperforms ARI0002h in preclinical models, particularly under reduced BCMA expression. ARI2h-TM28 maintains comparable cytotoxicity and cytokine production but achieves superior tumor control and survival, addressing antigen escape. Transcriptomic analysis revealed downregulation of inflammatory genes, contributing to enhanced persistence. These findings, coupled with ARI2h-TM28’s optimized metabolic profile, highlight the potential of TM domain modifications to overcome key resistance mechanisms in CAR-T cell therapy. Future studies incorporating humanized immune system in vivo models may provide a more accurate assessment of potential adverse effects in patients before moving into a clinical setting. Also, the possibility of improving efficacy with shorter expansion period with the CD28 TM motif could be further explored. In conclusion, ARI2h-TM28 represents a promising candidate for improving outcomes in relapsed/refractory MM, being very easily translated from the bench to the clinics through an academic clinical trial for patients with MM.

Data availability statement

Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.