Epitope mapping of humoral immunogenicity of orvacabtagene autoleucel shows an IgM response with minimal impact on CAR T cellular kinetics

Abstract

Orvacabtagene autoleucel (orva-cel) is a fully human B cell maturation antigen (BCMA)-targeted chimeric antigen receptor (CAR) T cell therapy evaluated in a phase 1/2 study in patients with relapsed or refractory multiple myeloma (RRMM). To assess treatment-related immunogenicity, anti-CAR therapeutic domain-specific antibodies (ATAs) were monitored in 157 treated patients. The ATAs were detected in 44.6% of patients over the course of study, with titers and incidence increasing over time. The goal of this study was to further characterize the observed immune response. The ATA status did not affect CAR T cell expansion or patient survival outcomes, though reduced persistence was observed in ATA-positive patients. Comprehensive immune profiling—including isotype analysis and B cell epitope mapping—identified five immunodominant consensus peptide sequences within the CAR domain. These epitopes were targeted by both Immunoglobulin G (IgG) and Immunoglobulin M (IgM) isotypes, with a persistent IgM response detected in most ATA-positive individuals. Despite the presence of ATAs, no adverse impact on cellular expansion was observed, potentially due to lymphodepletion and baseline immune suppression characteristic of B cell malignancies. These data suggest that the limited functional T- and B-cell capacity in RRMM may attenuate the clinical consequences of ATA development. The in vitro immunogenicity risk assessment and epitope mapping identified immunogenic hotspots within the CAR structure, which could have led to the high incidence of immune response observed in the patients. However, the analysis from this study points to a weak clinically non-relevant nature of the response that could be attributed to the patient’s immune status and diseased state.

Graphical abstract

Anti-CAR antibodies were detected in 44.6% of orva-cel-treated RRMM patients and mapped to immunodominant epitopes within the CAR domain. Despite reduced CAR T cell persistence in ATA-positive patients, no impact on CAR T cell expansion or survival outcomes was observed, suggesting limited clinical relevance in the context of immune suppression and disease-related immune-dysfunction.

Introduction

Chimeric antigen receptor (CAR) T cell therapies have recently emerged as a novel approach for treating cancers^1,2 and immune-related disorders.^3,4 Orvacabtagene autoleucel (orva-cel) is an investigational second-generation CAR T cell product composed of autologous CD4+ and CD8+ T cells transduced with a genetically engineered replication-incompetent, self-inactivating lentiviral vector to express a CAR targeting human B cell maturation antigens (BCMAs).^5,6,7 BCMA, a transmembrane glycoprotein member of the tumor necrosis factor receptor superfamily 17 (TNFRSF17), is exclusively expressed on mature B lineage cells, particularly plasma cells, and is overexpressed by malignant plasma cells in multiple myeloma (MM).⁸ As a result, BCMA has emerged as a key therapeutic target in MM, with multiple innovative modalities such as antibody-drug conjugate (ADC), CAR T cells, and bispecific T cell engagers in the market or under development.^7,9,10

Two BCMA-targeted CAR T cell therapies, idecabtagene vicleucel (ide-cel) and ciltacabtagene autoleucel (cilta-cel), have been approved for patients with relapsed or refractory multiple myeloma (RRMM) due to their remarkable efficacy in clinical trials.^11,12 Real-world data have further supported their deep and durable responses with manageable safety profiles.^13,14 Compared to ide-cel and cilta-cel, orva-cel incorporates distinct design features. Ide-cel uses a murine-derived single-chain variable fragment (scFv), whereas orva-cel employs a fully human scFv, which is intended to reduce immunogenicity and enhance persistence. Cilta-cel, on the other hand, differs more substantially by integrating a biparatopic heavy-chain-only variable (VHH)-based binding domain that engages two non-overlapping BCMA epitopes.⁷

While all three constructs share the CD8 hinge/transmembrane, 4-1BB costimulatory, and CD3ζ signaling domains were associated with immunodominant epitopes, these variations in the extracellular antigen-binding components represent the key molecular distinction that may influence immune recognition, CAR T cell expansion kinetics, and long-term durability of clinical responses (Figure 1).

Figure 1.

Structural comparison of BCMA-targeted CAR T constructs

Schematic representation of the CAR architectures of orva-cel, ide-cel, and cilta-cel. All constructs incorporate CD8 transmembrane, 4-1BB costimulatory, and CD3ζ signaling domains but differ in their antigen-binding regions: orva-cel uses a fully human scFv, ide-cel uses a murine scFv, and cilta-cel uses a biparatopic VHH-based domain.

Orva-cel was investigated in a phase 1/2 study in subjects with RRMM who received ≥3 prior regimens, including autologous stem cell transplant, a proteasome inhibitor, an immunomodulatory drug, and an anti-CD38 monoclonal antibody.

CAR T cell immunogenicity, driven by innate and adaptive immune mechanisms, is well-documented.^15,16,17 Innate immunity may be triggered by residual impurities in CAR T cell products, leading to cytokine release and cross-activation of adaptive immunity. Adaptive humoral immunity can arise via major histocompatibility complex (MHC) class II-mediated CD4+ T cell responses, promoting B cell activation and anti-CAR antibody production. Similarly, adaptive cellular immunity may also involve MHC class I-mediated CD8+ T cell responses to intracellular antigens, resulting in cytotoxic clearance of CAR T cells. Risk factors for CAR T cell immunogenicity include non-human sequences in CAR domains, chimeric construct, and residual viral and gene-editing artifacts.^15,18 While in silico tools and in vitro assays, such as EpiMatrix and MHC-associated peptide proteomics (MAPPs), can predict CAR T cell immunogenicity risk,¹⁸ the clinical significance of such responses varies widely and must be evaluated in patients.

Orva-cel with a fully human engineered scFv, was designed to reduce risk of anti-CAR therapeutic domain-specific antibodies (ATAs) formation and potentially improve CAR T cell persistence relative to murine- or VHH-based constructs. However, a high incidence of immunogenicity was observed in the phase 1/2 clinical trial. This study characterizes orva-cel immunogenicity using clinical samples to understand the relevance of the high incidence of observed immunogenicity. An intrinsic and extrinsic risk assessment of orva-cel by in silico and in vitro platforms like EpiMatrix algorithm and mass spectrometric peptide proteomics assays (MAPPs) confirmed presence of T cell epitopes. A B cell epitope mapping using peptide microarray analysis identified low affinity (IgM isotype) ATAs, as well as conformational epitopes that bind to both extracellular and intracellular CAR domains.

Results

Characteristics of orva-cel humoral immunogenicity

A total of 856 samples from 159 subjects were analyzed as part of the clinical strategy. Two subjects were excluded from ATA data analysis due to missing pre-treatment samples. The incidence of treatment-emergent ATAs, ATA onset time (day) and titers for each dose level are summarized in Table 1. Overall (across all doses) 70 (44.6%) subjects were ATA-positive at one or more time points at 3 months or later and 2 (1.3%) subjects were baseline (pre-treatment) ATA-positive. The CAR T cell dose level at 50 million (M) showed the highest ATA-incidence (71.4%), with low titers. However, the ATA incidence did not appear to have a strong correlation with the CAR T cell dose. The overall median titer of ATA-positive subjects was 1,536, and the highest median titer was 12,288 at doses of 300 and 600 M. The median onset time for treatment-emergent ATAs was 190 days with earliest and latest detections at 29 days and 752 days respectively (Figure 2A). The incidence of ATA-positive and negative subjects as shown in Figure 2B is based on the total number of samples assessed at each time point.

Table 1.

Summary of orva-cel ATA incidence, onset time, and titers by dose level

Dose (M)	ATA status, N (%)		Onset	Titersa	Follow-up duration
	Negative	Positive	Median (range), days	Median (range)	Median (range), days
50 (N = 14)	4 (28.6)	10 (71.4)	184 (82, 365)	60 (6, 12,288)	288 (73, 622)
150 (N = 29)	21 (72.4)	8 (27.6)	183 (91, 378)	1,152 (3, 12,288)	241 (28, 729)
300 (N = 26)	17 (65.4)	9 (34.6)	275 (92, 590)	12,288 (12, 98,304)	285 (29, 736)
450 (N = 21)	11 (52.4)	10 (47.6)	230 (100, 372)	432 (3, 98,304)	320 (29, 726)
600 (N = 67)	34 (50.7)	33 (49.3)	256 (29, 752)	12,288 (3, 1,572,864)	393 (29, 792)
Total (N = 157)	87 (55.4)	70 (44.6)	190 (29, 752)	1,536 (3, 1,572,864)	314 (29, 792)

ATA, anti-CAR therapeutic antibodies; M, million; N, patient number.

^aThe minimum required dilution (MRD) = 1:3.

Figure 2.

Kinetics of anti-CAR therapeutic antibodies

(A) Onset of treatment-emergent ATA in dosed subjects (N = 69). (B) Incidence (percentage of ATA negative/positive samples) at different visit time points on the total number of samples assessed at each time point. (C) Titers of ATA positive subjects at different time points post dosing (MRD = 1:3).

Cellular kinetics and transgene persistence of orva-cel

The cellular kinetics profile of orva-cel, as shown in Figure 3, represents a rapid expansion phase followed by bi-exponential decline, consisting of contraction and persistence phase. No differences were observed in orva-cel expansion phase between patients with ATA-positive and ATA-negative status. However, in the contraction and persistence phase, the orva-cel transgene level (persistence) appeared lower in ATA-positive patients compared to ATA-negative ones.

Figure 3.

Relationship of ATA status with ova-cel cellular kinetics

Cellular kinetics from ATA-positive and ATA-negative subjects are compared. Error bars represent interquartile range. Dashed line denotes lower limit of detection (21 copies per microgram). LLOD, lower limit of detection; Q1, first quartile; Q3, third quartile.

Association between ATA status and PFS

To evaluate the potential relationship between ATA development and clinical efficacy, landmark progression-free survival (PFS) analyses were performed at month 6 (M6) and month 12 (M12) following CAR T cell infusion (Figure 4). Only patients with evaluable ATA samples and PFS duration longer than the corresponding landmark time point were included in each analysis. PFS was comparable between ATA-positive and ATA-negative groups at both M6 and M12, with overlapping Kaplan-Meier curves and similar median PFS estimates, indicating no apparent adverse impact of ATA development on PFS within the observation period.

Figure 4.

Landmark PFS analysis by ATA status at month 6 and month 12. Kaplan-Meier plots showing progression-free survival according to anti-therapeutic antibody (ATA) status at

(A) Month 6 and (B) month 12. Patients with ATA sample status at M6 and PFS greater than M6 were included for landmark PFS analysis at M6. Patients with ATA sample status at M12 and PFS greater than M12 were included for landmark PFS analysis at M12. ATA positivity was defined relative to baseline as an increase of ≥ 4-fold in ATA titer if ATA were detected at baseline. ATA, anti-therapeutic antibodies; CI, confidence interval; M6, month 6; M12, month 12; PFS, progression-free survival

The differences in subject numbers included in the landmark analyses relative to the overall ATA-positive and ATA-negative populations reported in Table 1 reflect the landmark methodology rather than a higher proportion of early progressors in the ATA-positive group. These differences are attributable to missing ATA measurements at the relevant landmark time points, exclusion of subjects who progressed before the landmark, and changes in ATA status over time. At M6, 100 subjects were at risk (72 ATA-negative and 28 ATA-positive based on M6 ATA status). Notably, 29 of the 72 ATA-negative subjects at M6 subsequently converted to ATA-positive status at later time points, contributing to the apparent imbalance in group sizes. Similarly, at M12, 52 subjects were at risk, with 26 subjects each in the ATA-negative and ATA-positive groups. Collectively, these findings indicate that the observed differences in starting numbers are an expected consequence of landmark analysis and longitudinal ATA status evolution, rather than enrichment of early progression events in ATA-positive patients. Taken together, these additional analyses indicate that the lower number of ATA-positive subjects at landmark time points is not driven by an excess of early progressors. Moreover, un-landmarked ATA-based PFS comparisons are confounded by follow-up imbalance and should therefore be interpreted with caution.

B cell epitope mapping by peptide microarray

Figure 5 presents the IgG and IgM response heatmaps of the 10 ATA-positive samples against conformational peptides. Samples 03 and 08 displayed the strongest IgG responses with similar patterns, while samples 03 and 06 showed the strongest IgM responses. Sample 02, with the lowest titer (12), had the weakest responses for both IgG and IgM isotypes. Table 2 highlights five consensus peptide motifs that were recognized by both IgG and IgM isotypes, and for which at least 40% of the selected ATA-positive samples showed strong responses. Linear peptide epitope mapping revealed similar response profiles, with one sequences in the variable region of the light chain (VL) and another in the variable region of the heavy chain (VH), both within the complementarity-determining regions (CDRs), eliciting consistent, strong responses in most samples.

Figure 5.

Heatmaps of the IgG and IgM responses of ten ATA-positive serum samples against conformational peptides

(A) An IgG isotype heatmap. (B) Sample IDs, visits, and titers of ten ATA positive samples. (C) IgM isotype heatmap. ^aMRD = 1:3.

Table 2.

B cell epitope mapping result summary

Peptide sequence	CAR region	% subjects with IgG isotype response (n = 10)	% subjects with IgM isotype response (n = 10)
QAEDEADY	VL	60	70
ARSQRDGYMDY	VH CDR3	60	90
FYPSDIAVEW	Hinge CH3	40	50
DGCSCRFPEE	4-1BB (CD137)	50	50
LNLGRREEY	CD3z	50	60

CAR, chimeric antigen receptor; VH, variable region of the heavy chain; VL, variable region of the light chain; CDR, complementarity-determining region; CH3, constant heavy chain 3. Letters in italic indicating it was not clear if a certain amino acid contributed to antibody binding.

Immunopeptidomic evaluation

MHC class I mediated cellular immunogenicity is a possibility for CAR T therapeutics due to the intracellular expression of the CAR. To investigate the MHC classes I and II presented peptides from orva-cel, we developed a novel MAPPs assay utilizing the CAR-transduced CD8 and CD4 therapeutic cells as antigen-presenting cells (APCs) versus the traditional monocyte-derived dendritic cells and separate anti-MHC classes I and II immunoprecipitate (IP) antibodies.¹⁹ MAPPs data from 2 donor CAR T cells showed that MHC class I and class II peptides spanning the full length of the CAR (Figure 6). While most of the peptides were determined to be fully human sequence, several sequence peptides from the ScFv CDRs were presented on MHC classes I and II. These hypermutated, non-germline CDR epitopes may drive immunogenicity in monoclonal antibodies¹⁹ and therefore represent potential T cell epitopes for cellular immune responses, which were not measured in patients, in addition to the observed humoral immune responses to orva-cel.

Figure 6.

Comparison of in silico predicted MHC classes I and II T cell epitopes, in vitro MAPPs-assessed T cell epitopes and B cell epitopes by B cell epitope mapping

MHC, major histocompatibility complex; MAPPs, MHC-associated peptide proteomics; VH, variable region of the heavy chain; VL, variable region of the light chain; CDR: complementarity-determining region; TM, transmembrane. ^a: These are not the only class I epitopes but a cluster of epitopes validated with the MAPPs assay.

In silico immunogenicity prediction using EpiMatrix showed that orva-cel sequences had a low immunogenicity risk, scoring −35.19, which is below the median score of human secreted proteins (Figure S1). Despite the low overall score, some MHC classes I and II epitopes were identified by the in-silico tool as strong predicted binders across a broad range of HLA supertypes (Figure 6).

Discussion

The immune response to CAR T cell therapies is possible due to the presence of non-self-regions in the extracellular domains (ECDs), viral vector-mediated transduction, and CRISPR/CAS-related elements, or other foreign components needed for expansion of CAR T cell therapies.^16,17,20,21 Unlike conventional biologics, the immune response to CAR T therapies in hematological malignancies tends to be delayed because CAR T cells need time to expand and be presented in the context of MHC class II molecules, necessary to drive a T cell-mediated humoral response. This typically occurs when there is sufficient expansion of CAR T cells. Additionally, since current CAR T therapies primarily treat malignancies-targeting B cells, the ability to induce a robust, affinity-matured antibody response may be compromised.

ATA development occurs when the immune system recognizes the CAR as foreign, through both T cell-dependent (Td) and T cell-independent (Ti) adaptive humoral immune mechanisms. The immunogenicity of the CAR T domain is influenced by specific epitopes within the scFv targeting the CAR’s antigen and additional domains in the CAR construct design. Currently, six FDA-approved autologous CAR T therapies use murine-derived scFv, except cilta-cel, which incorporates two camelid VHH regions. The incidence of treatment-emergent ATAs across these therapies ranges from 0% to 53%.^17,22,23 Among the three BCMA-directed CAR T cell therapies, cilta-cel (CARVYKTI) exhibited the lowest incidence of treatment-emergent ATAs (approximately 21%),²³ whereas ide-cel (ABECMA), which uses a murine-derived binder, showed the highest rate (about 53%).²² The overall ATA incidence observed for orva-cel (44.6%) was like that of ide-cel but higher than expected for a construct utilizing a fully human binder.

In silico analysis of orva-cel immunogenicity indicated a low aggregate sequence risk, although some neoepitopes were predicted to be presented. MAPPs assessment confirmed the processing and presentation of CAR-derived epitopes, which could provide the Signal 1 required to initiate ATA responses. Interestingly, B cell epitope mapping analysis identified sequences within the ECD of orva-cel that overlapped with some of the predicted T cell epitopes and the in vitro MAPPs findings (Figure 6). Even with the sequence liability associated with the ECD domain, orva-cel demonstrated a delayed ATA onset, with a median onset of approximately 6 months. This delay may be partially explained by the lymphodepletion regimen administered before CAR T cell infusion, which eliminates most lymphocytes and plasma cells responsible for antibody production. Additionally, given that BCMA is expressed on both malignant B cells and normal plasma cells, orva-cel’s expansion and persistence could lead to the destruction of normal plasma cells and result in hypogammaglobulinemia through an “on-target, off-tumor” effect.²⁴ Indeed, B cell aplasia has been shown to occur shortly after lymphodepleting chemotherapy and can persist for months following anti-BCMA CAR T therapy, resulting in prolonged hypogammaglobulinemia.²⁵

In terms of B cell epitope mapping, linear epitopes consist of sequential amino acid residues, while conformational epitopes are composed of spatially proximate, solvent-exposed residues that are discontinuous in sequence.²⁶ The conformational epitope mapping of orva-cel identified five consensus peptide motifs recognized by both IgG and IgM antibodies in ATA-positive patient samples. IgM isotype ATA production is triggered by a T-cell-independent mechanism, where multivalent crosslinking of B cell receptors (BCRs) leads to rapid antibody secretion, although these antibodies are usually of low binding affinity and short-lived.²⁷ In contrast, IgG-type ATA are produced through a T cell-dependent process, which involves the uptake of therapeutic proteins by APCs, engagement of peptide-MHC II complex with T cell receptors (TCRs), and B cell maturation toward plasma cells.²⁸

No CAR T-specific T cell response was detected (see Figure S2), which may indicate that such a response was absent, possibly due to a lack of strong intrinsic antigenicity within the CAR T construct domains (e.g., the scFv or linker region).^29,30 Nevertheless, the detection of shared peptide motifs targeted by IgG antibodies that overlap with MAPPs-identified MHC II-restricted peptides may still suggest limited or subthreshold T cell engagement contributing to the humoral response.¹⁹

A CAR construct redesign can help to eliminate the epitopes identified by both in silico and by peptide microarray to avoid making ATA responses to those regions; even though no impact was observed in RRMM; it may be good to optimize the sequence to reduce an ATA response, which could impact safety. The ATA does impact CAR T persistence to a certain extent; we feel that removing any sequence-based liability would be beneficial especially for next-generation CAR T design to allow a re-administration. An interesting observation in ATA-positive patients was the persistence of IgM isotype antibodies (Figure S3). This could be due to the potential inhibition of isotype switching, particularly from IgM to IgA, possibly because anti-BCMA CAR T cells may interfere with the mechanisms that typically drive this switch by disrupting BCMA-mediated signaling required for plasma-cell differentiation and class-switch recombination.^31,32 These clinical findings support the concept that IgM antibody responses are typically of low affinity and short-lived, and therefore less likely to exert meaningful pharmacologic or immunogenic impact unless they undergo class-switching to IgG.^33,34,35 Similar to orva-cel, preliminary B cell epitope mapping analyses in selected ATA-positive subjects dosed with ide-cel revealed immunogenic peptide sequence motifs recognized by both anti-human IgG and IgM antibodies. In addition, distinct IgG and IgM response profiles were observed, both reaching moderate signal intensities (preliminary, unpublished observations).

Although a substantial incidence of ATAs was observed for orva-cel and ATA-positive subjects exhibited shorter cellular persistence, no definitive evidence was found for the impact of treatment-emergent ATAs on orva-cel expansion, efficacy, or safety.⁶ Furthermore, landmark PFS analyses at month 6 and month 12 revealed no significant differences between ATA positive and ATA negative groups, indicating ATA status did not affect survival outcomes, and these findings are consistent with observations for ide-cel,³⁶ cilta-cel,^17,23 and other approved CAR T cell therapies.¹⁷ The lack of association between ATA development and clinical outcomes across the six approved CAR T cell therapies may be due to lymphodepleting chemotherapy before CAR T cell infusion and the B cell targeting nature of CAR T cell therapy, both of which suppress humoral immunogenicity.^17,20 Furthermore, the low-affinity IgM antibodies observed in ATA-positive patients are unlikely to neutralize the CAR’s function by interfering with target binding or cause CAR T cell elimination via antibody-dependent cellular cytotoxicity or complement activation.

In conclusion, while humoral immunogenicity was observed in orva-cel-treated patients, the delayed onset of immune responses, weak overall immune responses, and the low-affinity nature of IgM antibodies may have reduced the impact of the immunogenicity on the CAR T cellular kinetics and safety. The rates of immunogenicity incidence between the orva-cel with a humanized CAR T domain and ide-cel with a murine CAR T product were similar (53% vs. 44%), and the non-human content did not change the immunogenicity rates or its impact on expansion and persistence. The ATA development may differ among CAR T cell products, and it may be difficult to draw general conclusions about the immunogenicity of CAR T cell therapies . Further research is needed to better understand the long-term implications of ATA development in CAR T therapies and its potential effects on treatment outcomes especially in the context of repeat administration and new disease indications like patients in solid tumor setting and autoimmunity.

Materials and methods

Clinical study design and samples for analysis

Whole blood and serum samples from the phase 1/2 study were used for immunogenicity and cellular kinetics analyses. A total of 159 subjects received orva-cel, including 125 subjects in the phase 1/1a dose-escalation cohort (dose range: 50–600 × 10⁶ CAR T cells) and 34 subjects in the phase 2/2a dose-expansion cohort (600 × 10⁶ CAR T cells). The study (NCT03430011) was approved by the Institutional Review Boards (IRBs) at participating institutions and conducted in accordance with the Declaration of Helsinki, International Council for Harmonization (ICH) Good Clinical Practice guidelines, and applicable regulatory requirements. Written informed consent was obtained from all participants prior to enrollment.

Humoral immunogenicity assessment

An electrochemiluminescent (ECL)-based bridging immunogenicity assay was developed to detect, confirm, and titer the presence of antibodies targeting the ECD of orva-cel’s CAR in human serum. Serum samples were diluted and incubated with a mixture of biotinylated orva-cel ECD and sulfonate (SULFO)-TAG-labeled orva-cel ECD, forming a bridging immuno-complex to enable capture and detection of anti-orva-cel antibodies. Following overnight incubation, samples were transferred to a pre-blocked, washed streptavidin-coated Meso Scale Discovery (MSD) plate. The biotinylated component of the complex bound to the streptavidin-coated plate, and unbound material was removed through washing steps. Upon addition of MSD read buffer T, ECL signals were generated by the SULFO-TAG label under applied voltage. Only anti-orva-cel antibodies capable of bridging the biotinylated and SULFO-TAG-labeled ECD reagents bound to the plate, thereby producing a detectable ECL signal. An anti-orva-cel ECD monoclonal antibody was used as positive control and the orva-cel ECD protein was used for confirmatory assay. The minimum required dilution (MRD) was 1:3. The sensitivity was 4.13 ng/mL for the screening assay and 4.78 ng/mL for the confirmatory assay. The immunogenicity method was fully validated in accordance with FDA immunogenicity testing guidance.³⁷ Serum was collected at pre-infusion (baseline), day 29, and months 3, 6, 9, 12, 18, and 24. Patients with ≥ one ATA positive-sample relative to baseline (an increase of ≥ 4-fold in ATA titer if ATA are detected at baseline) at any time after initiation of treatment were considered ATA-positive or treatment-emergent ATA-positive.

Cellular kinetics analysis

Orva-cel levels were quantified using a validated real-time qPCR assay³⁸ targeting the woodchuck hepatitis virus post-transcriptional regulatory element (WPRE) sequence. Genomic DNA (gDNA) was isolated from whole blood following plasma removal, and WPRE copies were amplified to determine transgene levels. Whole blood samples were collected on days 1 (preinfusion), 2, 4/5, 8, 11, 15, 22, 29 and months: 2, 3, 6, 9, 12, 18, and 24. Results were expressed as transgene copies/μg gDNA, calculated by normalizing WPRE copies to human albumin (housekeeping gene) copies. The assay’s lower limit of detection (LLOD) is 21 copies/μg gDNA.

B cell epitope mapping by peptide microarray

B cell epitope mapping is the process of identifying the specific regions on an antigen, which are recognized and bound by antibodies produced by B cells. A subset of ATA-positive samples from 10 treatment-emergent ATA positive subjects were further characterized for epitope specificity by B cell epitope mapping using PEPperPRINT peptide microarrays. These 10 patients were representative of the patients that developed low, medium, and high ADA responses based on their onset and titers. Samples were selected for characterization based on ATA onset time and magnitude of antibody response (titer), as detailed in Figure 5B. The orva-cel sequence (697 amino acids) was flanked with inert GSGSGSG linkers to prevent peptide truncation and parsed into 13-mer peptides with 12-amino-acid overlaps. Following on-chip synthesis, peptides for conformational epitope mapping were cyclized via a disulfide bond between C-terminal cysteine residues and a compatible N-terminal modification. The resulting microarrays contained 699 unique linear and conformational peptides printed in duplicate (1,398 total spots). Fluorescence-labeled secondary antibodies (goat anti-human IgG [Fc]-DyLight680; goat anti-human IgM [μ chain]-DyLight800) were used for detection. Influenza hemagglutinin (HA) peptides and mouse monoclonal anti-HA (12CA5)-DyLight800 served as positive controls. Microarrays were scanned using an Innopsys InnoScan 710-IR scanner. Spot intensities were quantified from 16-bit grayscale TIFF files using PepSlide Analyzer software. The algorithm segmented raw, foreground, and background signals for each spot, calculating average median foreground intensities and duplicate deviations. Intensity maps were generated using a color code: red (IgG) or green (IgM) for high-intensity interactions and white for low-intensity signals. Averaged spot intensities were plotted against the orva-cel antigen sequence (N terminus to C terminus) to visualize signal-to-noise ratios. Intensity plots were correlated with peptide maps, intensity maps, and raw scan images to identify linear and conformational epitopes.

MAPPs assay

Immunopeptidomic presentation was evaluated preclinically using healthy human donor T cells transduced with the orva-cel CAR sequence. Each sample contained 100e⁶ total T cells from one of six donors. Cells were lysed and sequentially immunoprecipitated for HLA-A/B/C (W6/32; ATCC) and HLA-DR (L243; ATCC). Peptides were eluted from MHC complexes with acid, desalted with C18 reverse-phase chromatography, and prepared for proteomic analysis. Peptide samples were separated chromatographically by a nanoElute HPLC (Bruker) and data was acquired with a timsTOF Pro2 (Bruker) in data-dependent acquisition mode. Raw proteomic data were searched using the FragPipe proteomic pipeline, and identified sequences were aligned back to the designed CAR sequence. Peptide clusters are represented as a single epitope around the common binding core. In silico class II analysis was performed using EpiMatrix ISPRI platform (EpiVax Inc).

Data and code availability

Bristol Myers Squibb policy on data sharing may be found at https://www.bms.com/researchers-and-partners/clinical-trials-and-research/disclosure-commitment.html.

Acknowledgments

The authors thank all study participants, their families, and clinical study team members. This study was funded by Bristol Myers Squibb. The authors also acknowledge Thomas Evan Kelemen for his support with MAPPs LC/MS analyses; Faiza Adeel for supplemental data collection using Biacore and Elispot experiments; Julia Piasecki for assistance with translational research; and Madhan Masilamani for his insightful review comments. Finally, we thank the bioanalytical staff at Charles River Laboratories for their support with study bioanalysis.

Author contributions

X.L and V.J. contributed to conception and design; X.L., Y.D., O.S., and M.P. contributed to acquisition of data; H.H., Y.D., K.O., M.P., and O.S. contributed to data analysis; X.L., V.J., Y.D., H.H., and M.P. drafted and revised the article; X.L., V.J., H.H., Y.D., K.O., O.S., and J.G. interpreted the data; all authors reviewed and provided input during the preparation of the manuscript; and agreed to be accountable for all aspects of the work.

Declaration of interests

X.L., Y.D., H.H., M.P., J.G., K.O., J.M., and V.J., are current or former employees and shareholders of Bristol Myers Squibb, United States. O.S. and V.S. are current employees and/or shareholders of PEPperPRINT GmbH.