Integrated clinical and single-cell profiling of BCMA CAR-T therapy in relapsed/refractory multiple myeloma

Abstract

Background

Despite therapeutic advances, multiple myeloma (MM) remains incurable, especially in relapsed/refractory (R/R) disease. B-cell maturation antigen (BCMA)-targeted CAR-T therapy, exemplified by FDA-approved agents like ide-cel and cilta-cel, offers promise, yet accessibility barriers necessitate local production.

Methods

This single-arm trial evaluated safety/efficacy of autologous BCMA CAR-T cells in six R/R MM patients.

Results

Cytokine release syndrome (CRS) occurred in 83% (grade 1), managed with tocilizumab without neurotoxicity. Toxicities were transient and resolved. Serum cytokine (IFN-γ, IL-6/8/10) peaked during CRS. Responses included 83% overall response (67% stringent complete response), unaffected by extramedullary disease or high-risk cytogenetics. Median PFS/OS were unreached, with an estimated 12-month OS rate of 83.33% and PFS rate of 66.67%. CAR-T cell persistence, with a median Cmax at 20.5 days, remained detectable in 83% at 1 month and 67% at 6 months. Single-cell RNA sequencing (scRNA-seq) and T-cell receptor sequencing (TCR-seq) demonstrated complementary roles of functionally specialized CD8⁺ Te subsets. Clonal evolution in the sustained responder (patient 2) revealed a shift from a polyclonal infusion product (IP) to oligoclonal dominance, with shared clones exhibiting enhanced cytotoxic and NK-like activity. Transcriptional adaptation of persistent clones over time indicated distinct phases of proliferation, stress response, and long-term persistence, with a concomitant shift in cellular phenotype from IP-derived Tem/circling T cells to a mixed Tem/Te_1/Te_2 population. Comparative analysis of two patients with divergent clinical outcomes (sustained remission vs. transient remission) revealed that early CAR-T exhaustion and increased regulatory T cells (Tregs) were associated with relapse.

Conclusions

Locally produced BCMA CAR-T is safe and effective in heavily pretreated R/R MM, inducing deep responses. scRNA-seq/TCR-seq findings highlight interplay of CAR-T heterogeneity, clonal adaptability, and immune regulation. CD8⁺ subset specialization and clonal persistence matter for durable responses; exhaustion and immunosuppression may cause relapse.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-07106-w.

Introduction

Multiple myeloma (MM) is a hematologic malignancy characterized by the abnormal proliferation of clonal plasma cells in the bone marrow, and it remains incurable [1]. Despite the considerable advancements provided by novel agents like proteasome inhibitors (PI), immune modulators (IMiD), monoclonal antibodies, and autologous hematopoietic stem cell transplantation (auto-HSCT), which have substantially enhanced patient outcomes, a subset of individuals develops resistance to these treatments, evolving into a state of relapsed/refractory (R/R) MM. This transition presents a formidable challenge, necessitating innovative therapeutic interventions to effectively address the clinical complexities associated with this advanced stage of the disease.

The advent of innovative immunotherapies has sparked renewed hope on the oncological horizon. Among these, chimeric antigen receptor T cell (CAR-T) therapy has emerged as a groundbreaking approach, demonstrating potential in the treatment of R/R MM. B-cell maturation antigen (BCMA), expressed on MM cells, with limited expression outside of terminally differentiated B cells and normal plasma cells [2], is a promising therapeutic target. Idecabtagene vicleucel (ide-cel) and Ciltacabtagene autoleucel (cilta-cel) are two anti-BCMA CAR-T cell therapies that recently approved by the Food and Drug Administration (FDA) and can induce deep and durable responses in R/R MM patients (73%–97.9% OR rates) [3–5].

However, production, logistics, and economic constraints significantly impede access to such cellular products. In light of this, there is an urgent need to develop locally manufactured CAR-T cell therapies. Here, we report the safety and efficacy of autologous BCMA CAR-T cells in six R/R MM adult patients in single-arm trial. Most patients achieved a favorable treatment response and tolerated CAR-T therapy well, demonstrating the safety and efficacy of BCMA CAR-T cells. To further investigate the in vivo characteristics of CAR-T cells, we performed single-cell RNA-sequencing (scRNA-seq) and T cell receptor-sequencing (TCR-seq) analysis on T cells from CAR-T infusion products and post-infusion samples. Temporal subset dynamics revealed a phased CD4⁺ Te expansion in patient 2, peaking at day 25, followed by CD8⁺ Tem/Te_1/Te_2 expansion at day 94. In contrast, patient 3 exhibited a rapid CD4⁺/CD8⁺ subset equilibrium by day 13. CD8⁺ Te_1 cells exerted direct cytotoxicity via NK-like pathways, while Te_2 cells facilitated chemokine-driven immune coordination. Clonal tracking in patient 2 demonstrated progressive oligoclonal restriction, with shared clones displaying enhanced cytotoxicity, chemotaxis, and NK-like activity. Persistent clones sequentially transitioned through proliferation, stress adaptation, and epigenetic stabilization, evolving from IP-derived Tem/cycling cells into Te_1/Te_2/Tem hybrids. Comparative analysis linked patient 3’s transient efficacy to paradoxical CAR-T states, where robust early cytotoxicity coincided with exhaustion and FOXP3^hi CAR⁺ Treg infiltration, ultimately driving clonal attrition.

Methods

Study design and participants

A single-centre, single-arm phase I clinical trial was conducted at the Department of Hematology and Oncology, Shenzhen University General Hospital. The trial was initiated in September 2021. The aim of this study was to investigate the safety and efficacy of the BCMA CAR-T cell product in patients with R/R MM. This study was approved by the ethics committee of Shenzhen University General Hospital and was registered with ClinicalTrials.gov (NCT05150522).

The inclusion criteria for this study were as follows: 1) age between 18 and 75 years old, 2) histological evidence of BCMA expression, 3) refractory or relapsed after at least three lines of previous therapy, and 4) Eastern Cooperative Oncology Group (ECOG) score of 2 or less. The details of the eligibility criteria can be found in the appendix. Informed consent was obtained from all patients and their families following the Declaration of Helsinki.

CAR-T cell construction and lymphodepletion chemotherapy

The BCMA CAR was constructed by fusing the BCMA single-chain variable fragment (scFv) in frame with the human CD8α hinge and transmembrane (TM) domains, followed by the human 4-1BB and human CD3ζ (Q65K) (Supplementary Fig. S1). Autologous peripheral blood mononuclear cells (PBMCs) were isolated from leukapheresis and cryopreserved until CAR-T manufacturing initiation. For CAR-T cell product, after thawing the PBMCs, T cells were isolated by EasySep™ Human T Cell Isolation Kit (STEMCELL, Catalog # 17,951) and cultured in X-VIVO® 15 Medium (Lonza, Catalog # 02-060Q) supplemented with IL-7/15/21 and ImmunoCult™ Human CD3/CD28/CD2 T Cell Activator (STEMCELL, Catalog # 10,990). Two or three days after the activation, T cells were infected with lentiviral vectors carrying the BCMA CAR. Subsequently, the cells underwent an expansion phase for an additional 8–10 days. Quality checks are carried out at every step of the manufacturing process. The transduction efficiency was evaluated using flow cytometry with APC anti-human EGFR Antibody (BioLegend, Catalog # 352,906) and T cell markers such as APC/Cyanine7 anti-human CD3 (BioLegend, Catalog # 300,426), PerCP/Cyanine5.5 anti-human CD4 (BioLegend, Catalog # 300,530), and PE/Cyanine7 anti-human CD8 (BioLegend, Catalog # 344,712).

Before CAR-T cell infusion, all patients received lymphodepleting chemotherapy. For lymphodepleting chemotherapy, all patients received a conditioning regimen comprising fludarabine (30 mg/m²) and cyclophosphamide (300 mg/m²) for three consecutive days (days − 5 to − 3). Cells were manufactured by Shenzhen Haoshi Biotechnology Co., Ltd. The doses were ranged from 0.85 × 10⁶ to 3.5 × 10⁶ CAR-T cells per kilogram of body weight.

Safety and efficacy assessment

The primary endpoint of the study was safety. Toxicity grading was performed according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI CTCAE) version 5.0. Cytokine release syndrome (CRS) and Immune effector cell associated neurological syndrome (ICANS) were graded according to the American Society for Transplantation and Cellular Therapy (ASTCT) Consensus Criteria [6]. Secondary endpoints included overall response rate (ORR), duration of response, progression-free survival (PFS), overall survival (OS), CAR-T cell persistence in peripheral blood, and serum cytokine levels. Response assessments were conducted following the International Myeloma Working Group (IMWG) consensus criteria [7]. The assessed time points covered days 0, 5, 7, 9, 11, 14, 17, 21, and 28 following CAR-T cell infusion, extending into the long-term follow-up phase. Follow-up assessments were conducted monthly from day 28 to month 6, and subsequently, every 3 months from month 6 to month 24, until the first occurrence of an event such as progression, death, initiation of new antitumor treatment, or withdrawal from the trial.

Blood samples were obtained on days 0, 5, 7, 9, 11, 14, 17, 21, and 28 following CAR-T cell infusion. Additionally, samples were collected at months 2, 3, 4, 5, 6, 9, 12, 15, 18, 21, and 24 to assess the expansion and persistence of CAR-T cells. The in vivo CAR-T cell levels were assessed using flow cytometry.

Single-cell RNA sequencing and TCR sequencing

Cell and library preparation

T cells were isolated from both the infused products (IPs) and peripheral blood (PB) samples of two patients (patient 2 and patient 3) at multiple time points using the EasySep PE Positive Selection Kit II (STEMCELL, Catalog # 17,684) following the protocol. The isolated cells were then suspended in PBS (Gibco, Catalog # 10,010,023) as single-cell suspensions, loaded into microfluidic devices using the Singleron Matrix® Single Cell Processing System (Singleron, SGR-SRAf10). scRNA-seq libraries were generated using the GEXSCOPE® Single Cell RNA Library Kits (Singleron, Catalog # 4,180,012), and scTCR-seq libraries were constructed using the GEXSCOPE Single Cell Immuno-TCR Kit (Singleron Biotechnologies, Catalog # 41,530,251), following the manufacturer’ protocols.

Sequencing data processing

All libraries were sequenced on an Illumina NovaSeq 6000 with 150 bp paired-end reads. For sc-RNA data, pre-processing, genome alignment, feature counting and expression matrix generation were carried out using CeleScope (https://github.com/singleron-RD/CeleScope). Cell barcode and UMI were extracted, and reads were mapped to the GRCh38 reference genome with STAR v2.6.1a. FeatureCounts (version 2.0.1) was used to obtain UMI counts and gene counts of each cell, generating expression matrix files for subsequent analysis. TCR clonotype assignment was performed using Cell Ranger (version 4.0.0) VDJ pipeline with GRCh38 as the reference. Cells with less than 200 detected genes or with a percentage of mitochondrial reads over 30% were filtered out from subsequent analysis. DoubletFinder identified and removed doublet. Seruat (v4.3.0) was used for normalized gene-cell matrix, highly variable genes identification and cell clustering.

The TCR V(D)J rearrangements and CDR3 sequence were utilized to identify the clonal composition of the CAR-T cells, ranking top TCR clonotypes by their abundance.

Differentially expressed genes identification and pathway enrichment analysis

Differentially expressed genes (DEGs) were identified using the Seurat FindMarkers function with Wilcox likelihood-ratio test (default parameters). The criteria for identifying DEGs with significant differences were as follows: 1) The absolute value of the average log2 fold change (|avg_log2FC|) must be equal to or greater than 1.5. 2) The adjusted p-value must be less than 0.05. 3) The absolute difference between the two groups must be equal to or greater than 0.1, while the percentage of genes in the less-expressing group was equal to or less than 0.3. The DEGs were up-regulated or down-regulated according to the avg_log2FC value > 1 or < − 1.

Data from the previous Seurat FindMarkers function analysis were used for Gene Ontology (GO) analysis. Gene Set Variation Analysis (GSVA v1.50.5) was employed to assess pathway enrichment in distinct CAR-T cell subsets, using hallmark gene sets from the Molecular Signatures Database (MSigDB).

Statistical analysis

All statistical analyses were conducted using GraphPad Prism 8. Continuous variables are expressed as medians and ranges, while categorical variables are presented as frequencies and percentages. The Kaplan–Meier method was employed to analyze the probability ratio of PFS and OS.

Results

Patients’ characteristics

Between September 2021 and July 2023, a total of eight patients were screened for eligibility, enrolled, and underwent leukapheresis. Two patients were excluded, either due to rapid disease progression or death. Ultimately, six patients in our study received anti-BCMA CAR-T cell therapy (Fig. 1). The characteristics of the enrolled patients are outlined in Table 1. The median age was 65.5 years (range 49–67), and the median time since diagnosis was 28.5 months. All patients had stage II or III disease, and 5 (83%) had extramedullary disease. 2 (33%) patients had high-risk cytogenetic abnormalities. 5 (83%) of patients were refractory to both a proteasome inhibitor and an immunomodulatory drug. 5 (83%) patients were refractory to anti-CD38 antibody (daratumumab). Half (50%) of patients had previously undergone autologous hematopoietic stem cell transplantation. 2 (33%) of patients had high lactate dehydrogenase (LDH) levels.

Fig. 1.

Study flow diagram

Table 1.

Patients’ baseline characteristics

	Patients (N = 6)
Median age, years	65.5 (range 49–67)
Gender
Male	4 (67%)
Female	2 (33%)
Median time since diagnosis, months	28.5 (range 6.7–37.6)
International staging system stage
I	0
II	3 (50%)
III	3 (50%)
Type of myeloma
IgG	3 (50%)
IgA	1 (17%)
Light chain only	2 (33%)
Extramedullary disease	5 (83%)
ECOG performance status score
0	0
1	5 (83%)
2	1 (17%)
Cytogenetic abnormalities
High-risk* cytogenetics	2 (33%)
Median lines of previous therapies	7 (range 3–10)
Previous therapies
Immunomodulatory drugs
Lenalidomide	5 (83%)
Thalidomide	1 (17%)
Pomalidomide	2 (33%)
Proteasome inhibitors
Bortezomib	6 (100%)
Ixazomib	4 (67%)
Carfilzomib	2 (33%)
Anti-CD38 monoclonal antibodies	5 (83%)
Previous auto HSCT	3 (50%)

Data are median (range) or n (%)

^*High‑risk cytogenetics included: t(4;14), t(14;16) and del(17p)

Safety

Six patients received BCMA CAR-T cell products. 5 patients (83%) developed CRS of grade 1 with median onset of 14 days after infusion (range 5–20) and a median duration of 1 day (range 1–5) (Table 2). No patients developed grade 3 or over CRS. None patients developed ICANS or other neurotoxicity. CRS was treated according to institutional guidelines. Tocilizumab was used in 33% of patients with a median of 2.5 administered dose (range 1–4). In terms of hematologic toxicities, the most common grade 3 or worse adverse events included anemia (n = 3, 50%), leukopenia (n = 6, 100%), neutropenia (n = 5, 83%), lymphopenia (n = 6, 100%) and thrombocytopenia (n = 4, 67%). The median duration of grade 3 or worse neutropenia was 14 days (range 7–21), and for thrombocytopenia, it was 26 days (range 17–52). All patients achieved an absolute neutrophil count of at least 1000 cells per μL within 3 weeks. Additionally, 5 out of 6 patients (83%) attained a platelet count of at least 50,000 cells per μL within 4 weeks. The remaining 1 patient (17%) exhibited platelet counts of at least 20,000 cells per μL on day 28, which subsequently normalized within the second month. Regarding non-hematologic toxicities, reported grade 3 or higher adverse events included the following: fatigue (n = 1, 17%), increased creatinine (n = 1, 17%), hypofibrinogenaemia (n = 2, 33%) and upper respiratory infection (n = 1, 17%). Importantly, all of these issues were successfully resolved.

Table 2.

Adverse events

AEs	Any	Grade 1	Grade 2	Grade 3	Grade 4
CRS	5 (83)	5 (83)	0	0	0
General disorders and administration site conditions
Fever	5 (83)	2 (33)	2 (33)	1 (17)	0
Fatigue	5 (83)	4 (67)	0	1 (17)	0
Edema	1 (17)	1 (17)	0	0	0
Rigor	2 (33)	2 (33)	0	0	0
Laboratory tests
Increased creatinine	1 (17)	0	0	1 (17)	0
Prolonged APTT	1 (17)	1 (17)	0	0	0
Hypofibrinogenaemia	2 (33)	0	0	2 (33)	0
Hypokalemia	2 (33)	1 (17)	1 (17)	0	0
Hyponatremia	3 (50)	2 (33)	1 (17)	0	0
Hypocalcemia	3 (50)	2 (33)	1 (17)	0	0
Hypophosphataemia	1 (17)	1 (17)	0	0	0
Disorders of the blood and lymphatic systems
Anemia	4 (67)	0	1 (17)	3 (50)	0
Leukopenia	6 (100)	0	0	3 (50)	3 (50)
Neutropenia	6 (100)	0	1 (17)	1 (17)	4 (67)
Lymphopenia	6 (100)	0	0	2 (33)	4 (67)
Thrombocytopenia	6 (100)	2 (33)	0	1 (17)	3 (50)
Gastrointestinal disorders
Anorexia	4 (67)	3 (50)	1 (17)	0	0
Nausea	4 (67)	3 (50)	1 (17)	0	0
Constipation	3 (50)	2 (33)	1 (17)	0	0
Abdominal distension	3 (50)	3 (50)	0	0	0
Abdominal pain	2 (33)	2 (33)	0	0	0
Ascites	2 (33)	2 (33)	0	0	0
Vomiting	2 (33)	2 (33)	0	0	0
Diarrhea	3 (50)	2 (33)	1 (17)	0	0
Infections
URI	1 (17)	0	0	1 (17)	0
Respiratory and cardiac disorders
Sinus tachycardia	1 (17)	1 (17)	0	0	0
Musculoskeletal disorders
Arthralgia	2 (33)	2 (33)	0	0	0
Back pain	1 (17)	1 (17)	0	0	0
Generalized muscle weakness	1 (17)	1 (17)	0	0	0

All data are n (%). Individual symptoms of AEs were graded using CTCAE, version 5.0. AEs, adverse events; CRS, cytokine release syndrome; ICANS, immune effector cell–associated neurotoxicity syndrome; URI, upper respiratory infection

Figure 2 depicted the serum cytokine dynamics in all patients during the first month following CAR-T infusion. Notably, patients experiencing CRS exhibited a significant increase in four cytokines: IFN-γ, IL-6, IL-8, and IL-10.

Fig. 2.

The variation of inflammatory markers. Serum cytokines, such as TNF-α, IFN-γ, IL-2, IL-6, IL-8 and IL-10 were measured in all patients. Each color in the figure corresponds to a distinct CRS grade observed in different patients

Efficacy

Of all enrolled patients, five (83%) patients had overall response, four (67%) had a stringent complete response (sCR), and one (17%) had very good partial response (VGPR) ((Table 3). Patient 2 achieved PR at first follow-up and further developed responses to sCR several months later (Fig. 3A). The median time to achieve the best response was 2.3 months (range 2.1–14.1), and the median time to attain complete response or better was 2.2 months (range 2.1–14.1) (Table 3). Among the five patients with extramedullary disease, three achieved complete remission of their extramedullary lesions, and one patient attained a VGPR with ongoing reduction in the size of the extramedullary disease. The median duration of response for the five responders had not been reached.

Table 3.

Patient best responses

	Patients (n = 6)
Overall response	5 (83%)
Stringent complete response	4 (67%)
Very good partial response	1 (17%)
Median time to best response, months	2.3 (range 2.1–14.1)
Median time to complete response or better, months	2.2 (range 2.1–14.1)

Fig. 3.

Descriptions of the clinical results. A Clinical response of the six patients. B The probability of overall survival (OS). C The probability progression-free survival (PFS). NE, no evaluation; MR, minimal response; PR, partial response; VGPR, very good partial response; sCR, stringent complete response; PD, progressive disease

The median follow-up of all patients was 19.9 months (range 10.5–35.1 months). One patient progressed after achieving sCR six months post-infusion, exhibiting BCMA-negative relapse. Patient 6 progressed after achieving a VGPR for 4.2 months. Unfortunately, patient 5 died due to disease progression. Encouragingly, three patients remained sCR. The median duration of PFS was 10.8 months. The median duration of OS has not been reached (Fig. 3B, C). The estimated 12-month OS rate was 83.33% (95% CI 43.65%–99.15%), and the estimated 12-month PFS rate was 50.00% (95% CI 18.76%–81.24%).

CAR-T cell expansion and persistence

Flow cytometry was employed to track the presence of CAR-T cells in peripheral blood at different time points following CAR-T cell infusion. The CAR-T cell expansion data for all patients were organized and displayed as dynamic curves (Fig. 4A, B). The changing percentage of CAR-T cells within the lymphocyte population corresponded with the CAR-T cell count. The median time to reach peak CAR-T cell concentration (Cmax) was 20.5 days (range 9–28), with a median Cmax of 173 cells per μL (range 2–3213). On day 26, patient 4 experienced a second peak CAR-T cell with a blood count of 3041 cells/μL, coinciding with observed CRS symptoms. CAR-T cell persistence in vivo was robust, as 83% of patients retained detectable CAR-T cells at 1 month, and 67% at 6 months. Notably, three progressing patients had undetectable CAR-T cell at the time of progression.

Fig. 4.

CAR-T cell pharmacokinetics. A CAR-T cell number in peripheral blood mononuclear cells (PBMCs). Absolute counts of CAR-T cells in peripheral blood were measured using flow cytometry. B Presence of CAR-expressing T cells among lymphocytes as quantified by flow cytometry

Temporal dynamics of CAR-T cell subsets and functional specialization of CD8⁺ effector subsets

To investigate the in vivo functionality of CAR-T cells in R/R MM and their interaction with the host immune environment, we performed scRNA-seq and TCR-seq on IPs and serial post-infusion samples from patient 2 and 3. Unsupervised clustering identified seven distinct T cell subsets: CD4 naive/central memory T cells (Tn/Tcm), CD4 effector T cells (Te), regulatory T cells (Treg), CD8 effector memory T cells (Tem), CD8 effector type 1 (Te_1), CD8 effector type 2 (Te_2) and cycling T cells (Fig. 5A, B). Circling T cells were predominant in IPs, consistent with ex vivo activation during manufacturing.

Fig. 5.

Temporal dynamics of CAR-T cell subsets and functional heterogeneity in effector CD8 T Cells. A Left: uniform manifold approximation and projection (UMAP) visualization was performed on 40,122 single-cell transcriptomes of T cells from infusion products and serial PBMCs in two MM patients (patient 2 and patient 3). Distinct clusters, represented by different colors, are annotated with inferred cell types, including three CD4⁺ T cell clusters, three CD8⁺ T cell clusters, cycling T cells, and other cell populations. Right: UMAP colored by CAR detection status (CAR⁺ or CAR^–). B Dot plot showing expression levels (color) and proportion of expressing cells (size) for key markers. C Proportion of cell types in the CAR-T product and CAR-T cells integrated from two patients with MM at different time points post-infusion. D Left: UMAP plot of all cells categorized by clone size, showing the distribution of cells with clone size ≥ 5 over time in two patients. Clonotypes with size ≥ 5 are defined as those observed more than four times. Right: Proportion of cell types within the CAR-T product and among CAR-T cells with clone size ≥ 5 from two multiple myeloma patients at various time points post-infusion. E Differentially expressed genes (DEGs) between CD8 Te_1 and CD8 Te_2 CAR-T cells. x-axis: difference between the percentage of cells highly expressing one certain gene in the designated cluster; y-axis: log2 fold-change (FC) of the average expression between the two groups. F Gene set variation analysis (GSVA) plots comparing CD8⁺ Te_1 and CD8⁺ Te_2 CAR-T cells in patient 2 and patient 3. GSVA based on a Kolmogorov Smirnov test. The adjusted P value < 0.05 was considered statistically significant. P values between the two groups are shown

In patient 2, CD4⁺ and CD8⁺ CAR-T cells were initially balanced in IPs but exhibited phased expansion: CD4⁺ Te subsets peaked at day 25, whereas CD8⁺ Tem/Te_1/Te_2 subsets expanded later (day 94; Fig. 5C). In patient 3, despite an initial CD4⁺ skewing in IPs, CD8⁺ subsets progressively expanded, reaching a balanced ratio by day 13 (Fig. 5C). TCR clonotype analysis in patient 2 revealed divergent trajectories: high-frequency CD4⁺ Te clonotypes (clone size ≥ 5) expanded continuously, whereas CD8⁺ Te_1/Te_2 clonotypes transiently declined at day 25 before rebounding at day 94 (Fig. 5D).

Further differential gene expression analysis and GSVA revealed subset-specific functional programs. CD8⁺ Te_2 cells, characterized by high expression of CCL3, CCL4, CCL4L2, and GZMK (Fig. 5E), were enriched in pathways related to chemotaxis, immune activation, and intercellular communication (Fig. 5F). In contrast, CD8⁺ Te_1 cells were associated with pathways governing natural killer cell-mediated cytotoxicity (Fig. 5F). These findings suggest that CD8⁺ Te_1 cells mediate direct tumor killing, while Te_2 cells facilitate immune cell recruitment and microenvironment remodeling.

Clonal restriction and transcriptional adaptation

To explore clonal evolution and intra-clonal differentiation of CAR-T cells, clonal analysis was performed in patient 2. Clonal tracking revealed a progressive shift from polyclonal IPs to oligoclonal dominance, with the top 10 clones increasing from 36.3% (IPs) to 82.5% (day 94), accompanied by contraction of transient clones (Fig. 6A, B, C). Further DEGs and GO analysis indicated that shared clones (persistent or intermediate, across ≥ 2 time points, n = 545) exhibited enhanced cytotoxicity (GZMK/H), chemotaxis (CCL3/4/5), and NK-like activity (KLRD1/NKG7) compared to IP-restricted clones. Shared clones were characterized by heightened immunity and chemotaxis (Fig. 6B, D, E).

Fig. 6.

Longitudinal clonal architecture, transcriptional diversification, and lineage tracking of CAR-T cells in relapsed/refractory multiple myeloma. A Proportions of CAR-T clones stratified by size ranks (Top10/11-100/101-1000/1001-10000) at manufacturing (infusion product, IP) and post-infusion timepoints (D17/25/53/94). B Venn diagram illustrating the number of shared and transient CAR-T cell clonotypes across different time points, with overlaps representing shared clonotypes. C Alluvial plot showing the temporal flow of shared clonotypes between adjacent samples. “Transient” (red) for clonotypes unique to single timepoint, “Persistent” (blue) for clonotypes consistently found across all five samples, and “Intermediate” (green) for clonotypes shared between ≥ 2 but not all samples. The flows between bars indicate the movement and persistence of clonotypes across time points. D Differentially expressed genes (DEGs) between shared (persistent + intermediate) and transient clones in the infusion product. x-axis: difference between the percentage of cells highly expressing one certain gene in the designated cluster; y-axis: log2 fold-change (FC) of the average expression between the two groups. E Top Gene Ontology (GO) biological process terms for DEGs enriched in shared clones. F Heatmap showing longitudinal expression patterns of differentially expressed genes from persistent TCR clones over time (product to D94). G The top five most abundant persistent TCR clones identified at D94 post-treatment with corresponding clones at infusion product or other time points are presented for CAR-T subsets. For each, circles denote the clone presence at each time point, sized corresponding to the clone frequency in its sample. Pie charts displaying cell-type distribution in each subset

Longitudinal transcriptomic profiling of persistent clones revealed phase-specific adaptations. Early post-infusion (from IPs to day 17) was marked by high expression of proliferation-related genes (MCM4/5/7, PCNA, MKI67) and transcription factor (JUN, JUNB, FOS, FOSB), indicating an expansion phase. At day 25 and day 53, stress-related genes (HSPA1, HSPH, HSPD, DNAJB1) were upregulated, suggesting stress response and cellular adaptation. By day 94, chromatin stability genes (H1-2, H1-3, H1-4) were highly expressed, implying long-term persistence (Fig. 6F, Supplementary Fig. S2). Consistent with these transcriptional changes, subsequent clone tracking indicated that the top five persistent TCR clones at day 94 were predominantly CD8⁺ subsets. In vivo, these dominant clones mainly originated from IP-derived Tem and cycling T cells, differentiating into Te_1 phenotype at days 17/25, into Te_1/Te_2 cells at day 53, and into Te_1/Te_2/Tem cell by day 94 (Fig. 6G).

Correlation between CAR-T heterogeneity and clinical outcomes

Given the contrasting clinical outcome between patient 2 (ongoing remission) and 3 (relapse after transient CR), we conducted DEGs analysis on CAR-T cells from patient 2 (day 17) and patient 3 (day 13). CAR-T cells from patient 3 exhibited high expression of cytotoxicity-related genes (GZMM) and NK-like genes (KLRD1, KLRC4-KLRK1), potentially explaining the initial CR despite lower in vivo CAR-T cell concentrations (Fig. 7A). GVSA analysis corroborated these findings, showing enhanced chemotaxis, immunity, and cell interaction in patient 3’s CAR-T cells (Fig. 7B).

Fig. 7.

Inter-patient heterogeneity in functional states of CAR-T clones. A Differentially expressed genes (DEGs) between CAR-T cells from patient 2 (P2, Day 17) and patient 3 (P3, Day 13). x-axis: difference between the percentage of cells highly expressing one certain gene in the designated cluster; y-axis: log2 fold-change (FC) of the average expression between the two groups. B Violin plots showing pathway enrichment in CAR-T cells from P2 (Day 17) vs. P3 (Day 13), estimated by gene set variation analysis (GSVA). An adjusted P value of < 0.05 was considered statistically significant. P values between the two groups are shown. C Ridge plots visualizing the longitudinal expression levels of ID2, KLRG1, and CTSW across sequential time points in P2 and P3. D Temporal expression trajectories of FOXP3 in CAR⁺ regulatory T (Treg) cells across sequential time points in P2 and P3

Notably, CAR-T cells from patient 3 also displayed high expression of exhaustion/senescence-related genes (ID2, KLRG1, and CTSW) and a greater proportion of Treg cells at day 13 (Figs. 5C, D, 7A, C). To further examine the role of Treg cells, we analyzed CAR⁺ Treg cells from patient 2 (day 17) and patient 3 (day 13). CAR⁺ Treg cells from patient 3 exhibited high FOXP3 expression (Fig. 7D, Supplementary Fig. S3). These results potentially contributed to patient 3’s subsequent relapse.

Overall, in patient 2 (Fig. 8A), our results revealed a model of dynamic changes in persistent CAR-T cell clones with stage-specific transcriptional programs that enabled sustained antitumor efficacy and durable remission by optimizing functional persistence. In contrast, patient 3 (Fig. 8B) experienced early relapse through dual mechanisms: upregulation of T cell exhaustion/senescence genes that reduced CAR-T cell cytotoxicity, and infiltration of immunosuppressive CAR⁺ Treg cells, both of which compromised tumor control and accelerated progression.

Fig. 8.

Proposed model illustrating that the clonal dynamics and transcriptional adaptability of CAR-T cells underpin the divergent clinical outcomes of BCMA CAR-T Therapy in multiple myeloma. A Stage-specific clonal evolution in patient 2 with durable remission. B Transcriptional signature of CAR-T cell in patient 3 with disease relapse

Discussion

In this study, a phase 1 clinical trial was conducted to assess the safety, response, and duration of locally produced anti-BCMA CAR-T cells in six patients with R/R MM. No serious adverse events were noted. 83% of patients experienced mild CRS without any neurotoxicity. Positive outcomes were achieved in 83% of patients, with 67% reaching sCR and 17% achieving VGPR. While these were preliminary findings with a small sample size, the efficacy of this locally produced BCMA CAR-T product appears comparable to that reported in the literature for BCMA-targeted CAR-T cell therapy [4, 8, 9].

In our study, CRS manifested in 5 patients, constituting 83% of cases. Significantly, all cases were categorized as mild (grade 1) and promptly resolved with tocilizumab treatment. None of the patients progressed to grade 2 or higher CRS, and no instances of neurotoxic effects were observed during the median 19.9-month follow-up. In contrast, other BCMA-targeted CAR-T trials reported CAR-T-related neurotoxicity in 18%–21% of patients [3, 4]. Further studies are required to validate the incidence of this toxicity. The predominant grade 3 or higher adverse events in our study were hematologic, likely stemming from the interplay of lymphodepleting chemotherapy and CAR-T cells. Fortunately, recovery from these hematological events was swift, and no bleeding events was detected. The recovery time from cytopenia in our study appeared comparable to other CAR-T cell therapies targeting BCMA [8]. These studies identified mild gastrointestinal toxic effects, which were alleviated through supportive therapy. The findings suggest that BCMA CAR-T demonstrates a favorable safety profile. Earlier research found that in patients with MM, BCMA was significantly more abundant in malignant plasma cells (PCs) than in normal bone marrow cells from healthy donors [10]. The heightened and selective expression of BCMA in PCs could be a reason behind the safety profile of BCMA CAR-T treatment.

The efficacy data presented in this study showed comparability to the results documented in the literature [8, 11–14]. In our study, five out of six patients (83%) demonstrated a clinical response, with all achieving a best response of VGPR or better. Notably, 67% of these responders reached sCR. These positive responses were observed even in patients with relatively unfavorable prognostic indicators, including 80% (4/5) of those with extramedullary disease and 100% (2/2) of those with high-risk cytogenetics among the evaluable cases. The median time to first response was 2.1 month (range 1.5–2.3); the median time to a complete response or better was 2.2 months (range 2.1–14.1). Following a median follow-up of 19.9 months, the median duration of PFS was 10.8 months, the median OS have not yet been reached. More significantly, the responses to BCMA CAR-T therapy persisted over time, with three out of four subjects with sCR remaining relapse-free beyond the 12-month evaluation point, underscoring the durability of the treatment’s efficacy.

Despite the 83% overall response rate observed in this study, disease recurrence was inevitable. By the time of the last follow-up, disease progression was noted in two patients: one experienced a BCMA-negative relapse, while in the other, pathological confirmation could not be obtained. In clinical studies done on BCMA-targeted CAR-T cells, BCMA-negative relapse was rare. In the KarMMa study, loss of tumor BCMA expression was suspected in only 4% of (3/71) patients who could be assessed at progression [3]. This rate appeared to be lower than what was observed in our study. However, due to the limited samples in our study, the frequency of antigen escape at relapse in BCMA CAR-T cell therapy still needs to be validated in large samples. Data is available on the mechanism of BCMA-negative relapse, which includes biallelic deletions, missense mutations, or in-frame deletions on chromosome 16 that encompass the BCMA locus [15, 16].

Regarding BCMA CAR-T cell kinetics, the peak time of CAR-T cells in patients in this study occurred slightly later than in patients receiving other CAR-T cell therapy, like ide-cel or cilta-cel. Additionally, the proportion of patients with detectable CAR-T cells at 6 months was slightly higher compared to other studies [3, 4]. The data indicated a decrease in peripheral CAR-T cell levels within 60 days, with the presence of CAR-T cells in peripheral blood becoming undetectable within 120 days for the majority of subjects [17–19]. In our study, however, 4 out of 6 patients (67%) exhibited detectable CAR-T cells beyond 120 days, with a range of 180 to 568 days, reaching a maximum duration of 568 days. This finding may elucidate the sustained response observed in these four patients.

We conducted longitudinal scRNA-seq and TCR-seq to evaluate functional and clonal dynamics of BCMA CAR-T cells in R/R MM. Identification of two CD8⁺ effector subsets—Te_1 and Te_2—suggests distinct roles in antitumor activity. Te_1 cells exhibit cytotoxicity via NK-like pathways, while Te_2 cells secrete chemokines for immune cell recruitment and microenvironment modulation. This T cell functional specialization corroborates recent findings on heterogeneity in CAR-T therapy and anti-tumor immunity, highlighting the importance of coordinated immune responses for tumor control [20, 21].

In the patient with ongoing remission (patient 2), clonal evolution shifted from polyclonal to oligoclonal dominance over time. Compared to IP-restricted clones, shared clones had enhanced cytotoxicity (GZMK/H), chemotaxis (CCL3/4/5), and NK-like activity (KLRD1/NKG7), along with elevated immunity and chemotactic potential. Persistent clones underwent transcriptional adaptations in three phases: early proliferation, stress response, and late chromatin stabilization. Upregulation of histone H1, associated with chromatin compaction and stability [22, 23], may represent a mechanism for maintaining CAR-T persistence. This aligns with recent studies linking epigenetic regulation to T-cell memory or persistence. Perturbation of chromatin remodeling complexes (e.g. mSWI/SNF or INO8) [24, 25] as well as disruption of DNA methylation regulators (SUV39H1, DNMT3a and TET2) [26–28] have been shown to improve T cell persistence. These results suggest that durable efficacy may depend on clonal fitness and adaptability to post-infusion stressors.

Conversely, in the patient with transient remission (patient 3), early CAR-T cytotoxicity (GZMM, KLRD1) was accompanied by exhaustion/senescence markers (ID2, KLRG1 and CTSW) [29–32] and increased Treg infiltration. The presence of CAR⁺ Tregs with high FOXP3 expression in patient 3 may have immunosuppressive effects, as Tregs can inhibit effector T-cell function and limit CAR-T expansion and anti-tumor efficacy [21, 33]. The co-expression of activation and exhaustion markers within CAR-T cells of patient 3 resembles chronic antigen exposure models, where persistent stimulation leads to T cell dysfunction [34, 35]. These findings suggest that early exhaustion and immune suppression may compromise CAR-T durability, even in patients with initial tumor clearance.

This study has several limitations. Given the small sample size (n = 6), it should be regarded as a pilot investigation. Larger, prospective studies are needed to confirm the observed therapeutic efficacy. While the limited cohort size constrains statistical power and precludes definitive conclusions, a key strength of this study lies in the acquisition of longitudinal single-cell data from both continuous responders and relapsing patients. This enables hypothesis generation regarding mechanisms of response, drug resistance, and immune dynamics. These preliminary findings warrant further validation in larger, more diverse patient cohorts encompassing a range of response depths, relapse kinetics, and baseline clinical characteristics.

In conclusion, locally produced BCMA CAR-T cell demonstrates a well-tolerated safety profile and effective outcomes in heavily pretreated R/R MM patients. It shows promise in inducing deep and enduring responses. The findings from scRNA-seq and TCR-seq emphasize the interplay among CAR-T functional heterogeneity, clonal adaptability, and immune regulation in determining clinical outcomes. CD8⁺ subset specialization and clonal persistence appear crucial for durable responses, while exhaustion and immunosuppressive signals may contribute to relapse. Further research is needed to translate these insights into strategies to optimize CAR-T cell fitness and tumor microenvironment interactions.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

MM

Multiple myeloma

R/R

Relapsed/refractory

BCMA

B-cell maturation antigen

CRS

Cytokine release syndrome

PFS

Progression-free survival

OS

Overall survival

scRNA-seq

Single-cell RNA sequencing

TCR-seq

T-cell receptor sequencing

IP

Infusion product

Tregs

Regulatory T cells

PI

Proteasome inhibitors

IMiD

Immune modulators

auto-HSCT

Autologous hematopoietic stem cell transplantation

CAR-T

Chimeric antigen receptor T cell

ide-cel

Idecabtagene vicleucel

cilta-cel

Ciltacabtagene autoleucel

FDA

Food and drug administration

ECOG

Eastern cooperative oncology group

scFv

Single-chain variable fragment

PBMCs

Peripheral blood mononuclear cells

NCI CTCAE

National cancer institute common terminology criteria for adverse events

ICANS

Immune effector cell associated neurological syndrome

ASTCT

American society for transplantation and cellular therapy

ORR

Overall response rate

IMWG

International myeloma working group

DEGs

Differentially expressed genes

GO

Gene ontology

GSVA

Gene set variation analysis

sCR

Stringent complete response

VGPR

Very good partial response

Author contributions

Conceptualization of the project: CLF, LXW, YSL, and LY. Development of methodology: CLF, LXW, YSL, and LY. Collection, analysis and interpretation of data: CLF, WQZ, LW, WFH, ZRC, YRW, KT, XG, YYX, SHW, LJW, JQQ, XYM, ZQH, CY, JHM, HXW and YSL. Writing articles and revision of the manuscript: CLF, LXW, YSL, and LY. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

Funding

Chinese National Major Project for New Drug Innovation (2019ZX09201002003), National Natural Science Foundation of China (82030076, 82070161, 81970151, 81670162 and 81870134), Shenzhen Clinical Research Center for hematologic disease (LCYSSQ20220823091401002), Shenzhen Science and Technology Foundation (JCYJ20190808163601776, JCYJ20200109113810154), Shenzhen Key Laboratory Foundation (ZDSYS20200811143757022), Sanming Project of Medicine in Shenzhen (SZSM202111004), Natural Science Foundation of Shenzhen University General Hospital (SUGH2019QD012), Shenzhen Natural Science Fund (the Stable Support Plan Program, 20200830182623001), HaiYa Young Scientist Foundation of Shenzhen University General Hospital (HY002), and Shenzhen Medical Research Fund (C2301003).

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the ethics committee of Shenzhen University General Hospital and was registered with ClinicalTrials.gov (NCT05150522). Informed consent was obtained from all enrolled patients.

Consent for publication

The authors have obtained consent to publish from the participant to report patient data.

Competing interests

The authors declare no potential conflicts of interest.