Наномедицина - мощная терапия CAR-T для множественной миеломы: к программируемой, долговременной и точной иммунотерапии.

Introduction

Multiple myeloma (MM) is a plasma cell–derived hematological malignancy characterized by clonal plasma cell expansion in the bone marrow, monoclonal immunoglobulin production, bone destruction, anemia, renal dysfunction, and immune dysregulation. 1–3 MM has entered an era in which deep remission is increasingly achievable; however, durable disease control remains elusive for many patients. 1 , 4 In 2022, MM accounted for approximately 187,952 new cases and 121,388 deaths worldwide, underscoring its persistent global burden despite substantial therapeutic advances. 5 In recent years, rapid advances in proteasome inhibitors, immunomodulatory drugs, anti-CD38 monoclonal antibodies, bispecific antibodies, and cellular immunotherapies have markedly improved survival outcomes in patients with MM. 6 , 7 Nevertheless, MM remains largely incurable, and most patients eventually develop relapsed or treatment-resistant disease after multiple lines of therapy. In patients with relapsed or refractory multiple myeloma (RRMM), durable disease control remains challenging because of greater tumor heterogeneity, progressive immune dysfunction, and the profoundly immunosuppressive bone marrow microenvironment. 7 , 8

Chimeric antigen receptor T-cell (CAR-T) therapy has emerged as a breakthrough therapeutic strategy for RRMM. 9 , 10 In particular, B-cell maturation antigen (BCMA)-targeted CAR-T products have demonstrated high response rates and deep minimal residual disease (MRD) negativity in multiple clinical studies, establishing BCMA as one of the most representative targets in MM immunotherapy. 11–13 CAR-T therapy involves ex vivo modification of a patient’s own T cells, enabling them to specifically recognize and eliminate myeloma cells and thereby induce robust antitumor responses even after conventional therapies have failed. This therapeutic model not only reshapes the clinical management of RRMM but also promotes a shift in MM treatment from pharmacological inhibition toward precise immune-mediated clearance. 14 , 15 However, clinical challenges have rapidly shifted from whether CAR-T cells can induce remission to how these responses can be maintained, safely reproduced, and made accessible across diverse healthcare settings. 16–18 First, although BCMA-targeted CAR-T therapy can induce deep remission, a substantial proportion of patients still relapse after treatment, indicating limited durability of therapeutic efficacy. 16 Relapse is not driven by a single mechanism but rather by multilevel processes, including low BCMA expression or antigen loss, tumor clonal evolution, CAR-T-cell exhaustion, insufficient memory-like phenotypes, failed metabolic adaptation, and bone marrow microenvironment–mediated immunosuppression. 16 , 19 Second, conventional CAR-T manufacturing relies on leukapheresis, viral transduction, ex vivo expansion, and stringent quality control, resulting in long production cycles, high costs, substantial batch-to-batch variability, and the risk of manufacturing failure. 17 , 20 For patients with RRMM who have rapidly progressive disease, poor T-cell quality, or prior exposure to multiple lines of therapy, these manufacturing constraints and delays may cause missed therapeutic windows. 17 , 20 In addition, CAR-T therapy may cause cytokine release syndrome, immune effector cell–associated neurotoxicity syndrome, prolonged immunosuppression, and an increased risk of infection. Therefore, its safety, controllability, and reproducibility require further optimization. 19 , 21

Nanomaterials and nanotechnology now offer new engineering strategies to address these challenges. 22–26 Lipid nanoparticles, polymeric nanocarriers, biomimetic nanoplatforms, targeted nanodelivery systems, and nanoscale viral or nonviral carriers can deliver mRNA, DNA, small interfering RNA, gene-editing components, small-molecule drugs, and immunoregulatory factors, thereby contributing to multiple stages of CAR-T therapy. 27 , 28 During ex vivo manufacturing, nanodelivery strategies may reduce dependence on viral vectors, minimize electroporation-induced damage, improve T-cell engineering efficiency, and help preserve memory-like and functional phenotypes. 29 During in vivo treatment, nanotechnology can be used to directly generate CAR-T cells, improve the controllability of CAR expression, modulate BCMA antigen density, alleviate T-cell exhaustion, enhance metabolic fitness, and precisely deliver immunoregulatory or microenvironment-remodeling agents to the bone marrow niche. 30 , 31 More importantly, nanoplatforms can be integrated with soluble BCMA detection, circulating tumor DNA analysis, extracellular vesicle profiling, molecular imaging, and spatial genomics to enable dynamic monitoring and early intervention for relapse risk. 32

Research on relapse mechanisms following MM CAR-T therapy, nanodelivery systems, and bone marrow microenvironment remodeling is rapidly expanding. However, these topics are often addressed in isolation, and systematic reviews focusing on relapse control and nanotechnology-enabled optimization across the entire CAR-T therapeutic continuum remain scarce. Existing reviews primarily examine the clinical efficacy of BCMA-targeted CAR-T therapy, selected resistance mechanisms, or the general application of nanomaterials in cancer drug delivery. Few integrate the MM-specific bone marrow niche with CAR-T manufacturing constraints, in vivo immune programming, and therapeutic strategies after BCMA-directed treatment. A comprehensive framework linking tumor biology, T-cell engineering, nanomedicine, and clinical translation is therefore needed to improve the durability, safety, and accessibility of MM CAR-T therapy. To provide a concise overview of the current cellular immunotherapy landscape in MM, Table 1 summarizes the principal antigen-directed platforms, representative targets and applications, stages of development, major advantages and limitations, and opportunities for integration with nanomedicine. This overview positions established BCMA-directed CAR-T therapy alongside emerging non-BCMA, dual-target, in vivo-programmed, innate immune-cell, and cell-free CAR-based strategies. It thereby establishes the clinical and technological context for the subsequent discussion of relapse mechanisms and nanotechnology-enabled therapeutic optimization. Accordingly, this review systematically examines recent advances and future directions in nanotechnology-enabled CAR-T therapy for MM. It first delineates the key biological mechanisms underlying relapse following MM CAR-T therapy. It then evaluates how nanomaterials can improve ex vivo CAR-T manufacturing and enable in vivo CAR-T generation. Finally, it analyzes key translational challenges, including delivery specificity, safety, manufacturing scalability, long-term monitoring, and clinical evaluation frameworks, and proposes strategies for integrating nanomedicine with MM CAR-T therapy in the post-BCMA era.
Table 1 Current Antigen-Directed Cellular Therapy Platforms and Potential Nanomedicine-Enabled Applications in MM Cellular Therapy Platform Representative Targets or Examples Current Application or Developmental Position Principal Therapeutic Rationale Major Limitations Potential Role of Nanomedicine Autologous BCMA-directed CAR-T cells Ide-cel and cilta-cel Established clinical treatment for RRMM and the current core cellular therapy platform for inducing deep remission Direct elimination of BCMA-positive plasma cells and induction of high response and MRD-negativity rates Prolonged manufacturing, variable starting T-cell quality, CRS, ICANS, cytopenias, infection, limited persistence, and BCMA-low or BCMA-negative relapse Low-damage ex vivo engineering, preservation of memory-like phenotypes, BCMA-density modulation, metabolic support, and relapse monitoring In vivo-generated BCMA CAR-T cells Targeted lentiviral or nanoparticle-based in vivo CAR programming; ESO-T01 Early clinical development Generates CAR-T cells directly within the patient and may eliminate leukapheresis, centralized ex vivo expansion, and prolonged vein-to-vein time Off-target transduction, uncontrolled CAR expression, CRS, long-term genotoxicity, viral shedding, and limited clinical evidence T-cell-selective carriers, transient mRNA or circRNA programming, expression-control modules, repeat dosing, and safety switches Dual- or multiantigen CAR-T cells BCMA/GPRC5D, BCMA/CD19, BCMA/CD24, and BCMA/B7-H3 strategies Early clinical or preclinical development Expands antigen coverage and reduces relapse caused by heterogeneous or low target expression More complex manufacturing, potential additive on-target/off-tumor toxicity, cytokine-related toxicity, and uncertain long-term persistence Co-delivery of multiple CAR-encoding RNAs, tunable expression ratios, localized cytokine support, and biomarker-guided target selection Non-BCMA-directed CAR-T cells GPRC5D and emerging CD38-, SLAMF7-, B7-H3-, or FcRH5-directed strategies Investigational, particularly relevant after failure of BCMA-directed therapy Provides an alternative treatment route for BCMA-low, BCMA-negative, or post-BCMA relapsed disease Target heterogeneity, antigen loss, potential expression in non-myeloma tissues, and limited clinical validation for several targets Nanodiagnostic assessment of target expression, spatial imaging, multitarget programming, and targeted delivery of CAR constructs CAR-NK cells BCMA-directed CAR-NK cells and FT576 Preclinical and early clinical development Offers potential off-the-shelf availability, natural cytotoxicity, ADCC, and a lower risk of GVHD than allogeneic T-cell platforms Limited persistence, inefficient homing, variable expansion, and insufficient randomized clinical evidence Nonviral mRNA-LNP engineering, IL-15 delivery, CXCR4-mediated bone marrow homing, and repeatable transient programming CAR macrophages or monocytes Exploratory BCMA-, CD24-, or CD38-directed CAR-M concepts MM-specific applications remain predominantly preclinical Promotes phagocytosis, antigen presentation, myeloid reprogramming, and conversion of an immunosuppressive bone marrow niche into a pro-inflammatory environment Difficult cell manufacturing, phenotypic plasticity, limited persistence, and scarce MM-specific clinical evidence In situ CAR monocyte programming, macrophage-targeted nanoparticles, mRNA delivery, and local microenvironment remodeling CAR-derived extracellular vesicles or exosomes BCMA CAR-T-derived exosomes Early preclinical development Provides a cell-free and potentially repeatable approach retaining CAR-mediated targeting and cytotoxic cargo Heterogeneous vesicle composition, uncertain pharmacokinetics, limited production standards, and lack of clinical validation Nanoengineered cargo loading, improved storage, surface modification, combination delivery of siRNA or small molecules, and maintenance or bridging therapy

Table 1 Current Antigen-Directed Cellular Therapy Platforms and Potential Nanomedicine-Enabled Applications in MM Cellular Therapy Platform Representative Targets or Examples Current Application or Developmental Position Principal Therapeutic Rationale Major Limitations Potential Role of Nanomedicine Autologous BCMA-directed CAR-T cells Ide-cel and cilta-cel Established clinical treatment for RRMM and the current core cellular therapy platform for inducing deep remission Direct elimination of BCMA-positive plasma cells and induction of high response and MRD-negativity rates Prolonged manufacturing, variable starting T-cell quality, CRS, ICANS, cytopenias, infection, limited persistence, and BCMA-low or BCMA-negative relapse Low-damage ex vivo engineering, preservation of memory-like phenotypes, BCMA-density modulation, metabolic support, and relapse monitoring In vivo-generated BCMA CAR-T cells Targeted lentiviral or nanoparticle-based in vivo CAR programming; ESO-T01 Early clinical development Generates CAR-T cells directly within the patient and may eliminate leukapheresis, centralized ex vivo expansion, and prolonged vein-to-vein time Off-target transduction, uncontrolled CAR expression, CRS, long-term genotoxicity, viral shedding, and limited clinical evidence T-cell-selective carriers, transient mRNA or circRNA programming, expression-control modules, repeat dosing, and safety switches Dual- or multiantigen CAR-T cells BCMA/GPRC5D, BCMA/CD19, BCMA/CD24, and BCMA/B7-H3 strategies Early clinical or preclinical development Expands antigen coverage and reduces relapse caused by heterogeneous or low target expression More complex manufacturing, potential additive on-target/off-tumor toxicity, cytokine-related toxicity, and uncertain long-term persistence Co-delivery of multiple CAR-encoding RNAs, tunable expression ratios, localized cytokine support, and biomarker-guided target selection Non-BCMA-directed CAR-T cells GPRC5D and emerging CD38-, SLAMF7-, B7-H3-, or FcRH5-directed strategies Investigational, particularly relevant after failure of BCMA-directed therapy Provides an alternative treatment route for BCMA-low, BCMA-negative, or post-BCMA relapsed disease Target heterogeneity, antigen loss, potential expression in non-myeloma tissues, and limited clinical validation for several targets Nanodiagnostic assessment of target expression, spatial imaging, multitarget programming, and targeted delivery of CAR constructs CAR-NK cells BCMA-directed CAR-NK cells and FT576 Preclinical and early clinical development Offers potential off-the-shelf availability, natural cytotoxicity, ADCC, and a lower risk of GVHD than allogeneic T-cell platforms Limited persistence, inefficient homing, variable expansion, and insufficient randomized clinical evidence Nonviral mRNA-LNP engineering, IL-15 delivery, CXCR4-mediated bone marrow homing, and repeatable transient programming CAR macrophages or monocytes Exploratory BCMA-, CD24-, or CD38-directed CAR-M concepts MM-specific applications remain predominantly preclinical Promotes phagocytosis, antigen presentation, myeloid reprogramming, and conversion of an immunosuppressive bone marrow niche into a pro-inflammatory environment Difficult cell manufacturing, phenotypic plasticity, limited persistence, and scarce MM-specific clinical evidence In situ CAR monocyte programming, macrophage-targeted nanoparticles, mRNA delivery, and local microenvironment remodeling CAR-derived extracellular vesicles or exosomes BCMA CAR-T-derived exosomes Early preclinical development Provides a cell-free and potentially repeatable approach retaining CAR-mediated targeting and cytotoxic cargo Heterogeneous vesicle composition, uncertain pharmacokinetics, limited production standards, and lack of clinical validation Nanoengineered cargo loading, improved storage, surface modification, combination delivery of siRNA or small molecules, and maintenance or bridging therapy

Current Antigen-Directed Cellular Therapy Platforms and Potential Nanomedicine-Enabled Applications in MM

Biological Classification of MM CAR-T Relapse

Relapse after chimeric antigen receptor T-cell therapy in multiple myeloma cannot be explained by a single resistance pathway. Current evidence supports a three-level biological classification comprising tumor-intrinsic mechanisms, CAR-T-cell-intrinsic dysfunction, and bone marrow microenvironment-mediated resistance. These categories are not mutually exclusive. In many patients, reduced antigen visibility, insufficient CAR-T-cell persistence, and an immunosuppressive or spatially inaccessible tumor niche may coexist and jointly drive disease recurrence. This framework is therefore more informative than classifying relapse solely as BCMA-positive or BCMA-negative ( Table 2 ).
Table 2 Biological Classification of Relapse After CAR-T Therapy in MM Biological Category Specific Mechanism Representative Evidence From Recent Studies Biological Consequence Potential Clinical Implication Ref Tumor-intrinsic BCMA downregulation and heterogeneous antigen density In a 76-patient cohort, BCMA expression at relapse was generally lower than at baseline; 50% of paired flow-cytometric samples showed ≥25% downregulation, whereas complete loss was uncommon Reduced CAR triggering and incomplete elimination of antigen-low clones Quantitative flow cytometry and paired baseline–relapse testing; consider antigen-enhancing or multitarget strategies [ 33 ] Tumor-intrinsic Biallelic TNFRSF17 inactivation Deletion of one allele plus a nonsense mutation in the second allele caused BCMA-negative relapse and resistance to CAR-T retreatment in a patient Complete loss of target recognition Avoid empirical retreatment with the same BCMA product without confirming antigen status [ 34 ] Tumor-intrinsic Homozygous BCMA deletion Selection of a clone with homozygous TNFRSF17 deletion was identified after ide-cel exposure Irreversible BCMA-negative immune escape Switch to non-BCMA targets or multitarget cellular therapy [ 35 ] Tumor-intrinsic Pre-existing resistant subclones and therapy-driven clonal selection Rare pre-existing TNFRSF17 -deficient cells expanded under BCMA-directed immune pressure; acquired focal deletions were also observed Conversion of minor resistant populations into dominant relapse clones Sensitive genomic testing and longitudinal sampling may identify emerging escape clones [ 36 , 37 ] Tumor-intrinsic Functional epitope escape BCMA extracellular-domain mutations or in-frame deletions preserved detectable protein but disrupted binding by selected BCMA-directed agents; strongest evidence is currently from T-cell engagers Apparent BCMA positivity despite loss of therapeutic recognition Use epitope-aware assays and avoid assuming cross-sensitivity among all BCMA-directed products [ 36 , 38 ] CAR-T intrinsic Insufficient expansion and persistence Greater ide-cel expansion in blood and bone marrow correlated with response; limited cellular expansion was associated with poorer outcomes Failure to maintain sufficient effector-cell numbers Monitor CAR transgene kinetics and optimize cell dose, lymphodepletion, and product fitness [ 39 ] CAR-T intrinsic Poor starting-cell quality and loss of memory phenotype Durable responders had more naïve and early-memory T cells and fewer senescent or exhausted cells in leukapheresis material Reduced self-renewal and shortened functional persistence Earlier T-cell collection, memory-preserving manufacturing, and reduced ex vivo overstimulation [ 39 ] CAR-T intrinsic CD4/CD8 exhaustion and inhibitory receptor expression Early relapse was associated with exhausted CD4-positive CAR-T states; relapse samples also showed exhausted CD8-positive T cells and TIGIT-positive NK cells Progressive loss of proliferation, cytokine production, and repeated killing capacity Transient checkpoint modulation or exhaustion-resistant CAR design [ 40 , 41 ] CAR-T intrinsic Metabolic maladaptation Glutamine limitation impaired BCMA CAR-T proliferation and cytokine production; ASCT2/SLC1A5 expression improved metabolic fitness and antitumor activity Inadequate function in nutrient-poor marrow niches Metabolic engineering or localized delivery of metabolic support [ 42 ] Microenvironment-mediated Cancer-associated fibroblast and stromal protection Bone marrow fibroblasts inhibited BCMA CAR-T activity; dual targeting of myeloma and stromal components improved control in experimental models Physical and paracrine protection of residual disease Stromal targeting, CXCL12–CXCR4 modulation, or niche-directed combination therapy [ 43 ] Microenvironment-mediated TGF-β-mediated immunosuppression Dominant-negative TGFBRII-armored BCMA CAR-T cells retained proliferation and cytotoxicity under TGF-β-rich conditions Reduced CAR-T activation, degranulation, cytokine release, and memory maintenance TGF-β-resistant CARs or controlled local TGF-β blockade [ 44 ] Microenvironment-mediated Myeloid-cell enrichment and suppressive communication Relapse samples showed increased macrophage populations and enhanced MIF/APRIL signaling together with exhausted lymphoid states Myeloid support of residual tumor and inhibition of effector cells Macrophage reprogramming and blockade of suppressive ligand–receptor networks [ 40 , 45 ] Microenvironment-mediated Hypoxia and extracellular-vesicle-mediated suppression Hypoxia reduced BCMA, impaired CAR-T killing and cytokine release, increased small extracellular-vesicle secretion, and reduced memory-like T-cell states Simultaneous antigen loss and CAR-T functional inhibition Hypoxia-responsive delivery, γ-secretase inhibition, and extracellular-vesicle targeting [ 46 ] Spatial resistance Extramedullary or extraosseous disease Clinical cohorts reported inferior outcomes and distinct immune or microenvironmental features in patients with extramedullary disease Limited trafficking, spatial antigen heterogeneity, and protection outside sampled marrow sites Whole-body imaging, lesion-specific antigen assessment, and trafficking-enhanced CAR design [ 47 , 48 ]

Table 2 Biological Classification of Relapse After CAR-T Therapy in MM Biological Category Specific Mechanism Representative Evidence From Recent Studies Biological Consequence Potential Clinical Implication Ref Tumor-intrinsic BCMA downregulation and heterogeneous antigen density In a 76-patient cohort, BCMA expression at relapse was generally lower than at baseline; 50% of paired flow-cytometric samples showed ≥25% downregulation, whereas complete loss was uncommon Reduced CAR triggering and incomplete elimination of antigen-low clones Quantitative flow cytometry and paired baseline–relapse testing; consider antigen-enhancing or multitarget strategies [ 33 ] Tumor-intrinsic Biallelic TNFRSF17 inactivation Deletion of one allele plus a nonsense mutation in the second allele caused BCMA-negative relapse and resistance to CAR-T retreatment in a patient Complete loss of target recognition Avoid empirical retreatment with the same BCMA product without confirming antigen status [ 34 ] Tumor-intrinsic Homozygous BCMA deletion Selection of a clone with homozygous TNFRSF17 deletion was identified after ide-cel exposure Irreversible BCMA-negative immune escape Switch to non-BCMA targets or multitarget cellular therapy [ 35 ] Tumor-intrinsic Pre-existing resistant subclones and therapy-driven clonal selection Rare pre-existing TNFRSF17 -deficient cells expanded under BCMA-directed immune pressure; acquired focal deletions were also observed Conversion of minor resistant populations into dominant relapse clones Sensitive genomic testing and longitudinal sampling may identify emerging escape clones [ 36 , 37 ] Tumor-intrinsic Functional epitope escape BCMA extracellular-domain mutations or in-frame deletions preserved detectable protein but disrupted binding by selected BCMA-directed agents; strongest evidence is currently from T-cell engagers Apparent BCMA positivity despite loss of therapeutic recognition Use epitope-aware assays and avoid assuming cross-sensitivity among all BCMA-directed products [ 36 , 38 ] CAR-T intrinsic Insufficient expansion and persistence Greater ide-cel expansion in blood and bone marrow correlated with response; limited cellular expansion was associated with poorer outcomes Failure to maintain sufficient effector-cell numbers Monitor CAR transgene kinetics and optimize cell dose, lymphodepletion, and product fitness [ 39 ] CAR-T intrinsic Poor starting-cell quality and loss of memory phenotype Durable responders had more naïve and early-memory T cells and fewer senescent or exhausted cells in leukapheresis material Reduced self-renewal and shortened functional persistence Earlier T-cell collection, memory-preserving manufacturing, and reduced ex vivo overstimulation [ 39 ] CAR-T intrinsic CD4/CD8 exhaustion and inhibitory receptor expression Early relapse was associated with exhausted CD4-positive CAR-T states; relapse samples also showed exhausted CD8-positive T cells and TIGIT-positive NK cells Progressive loss of proliferation, cytokine production, and repeated killing capacity Transient checkpoint modulation or exhaustion-resistant CAR design [ 40 , 41 ] CAR-T intrinsic Metabolic maladaptation Glutamine limitation impaired BCMA CAR-T proliferation and cytokine production; ASCT2/SLC1A5 expression improved metabolic fitness and antitumor activity Inadequate function in nutrient-poor marrow niches Metabolic engineering or localized delivery of metabolic support [ 42 ] Microenvironment-mediated Cancer-associated fibroblast and stromal protection Bone marrow fibroblasts inhibited BCMA CAR-T activity; dual targeting of myeloma and stromal components improved control in experimental models Physical and paracrine protection of residual disease Stromal targeting, CXCL12–CXCR4 modulation, or niche-directed combination therapy [ 43 ] Microenvironment-mediated TGF-β-mediated immunosuppression Dominant-negative TGFBRII-armored BCMA CAR-T cells retained proliferation and cytotoxicity under TGF-β-rich conditions Reduced CAR-T activation, degranulation, cytokine release, and memory maintenance TGF-β-resistant CARs or controlled local TGF-β blockade [ 44 ] Microenvironment-mediated Myeloid-cell enrichment and suppressive communication Relapse samples showed increased macrophage populations and enhanced MIF/APRIL signaling together with exhausted lymphoid states Myeloid support of residual tumor and inhibition of effector cells Macrophage reprogramming and blockade of suppressive ligand–receptor networks [ 40 , 45 ] Microenvironment-mediated Hypoxia and extracellular-vesicle-mediated suppression Hypoxia reduced BCMA, impaired CAR-T killing and cytokine release, increased small extracellular-vesicle secretion, and reduced memory-like T-cell states Simultaneous antigen loss and CAR-T functional inhibition Hypoxia-responsive delivery, γ-secretase inhibition, and extracellular-vesicle targeting [ 46 ] Spatial resistance Extramedullary or extraosseous disease Clinical cohorts reported inferior outcomes and distinct immune or microenvironmental features in patients with extramedullary disease Limited trafficking, spatial antigen heterogeneity, and protection outside sampled marrow sites Whole-body imaging, lesion-specific antigen assessment, and trafficking-enhanced CAR design [ 47 , 48 ]

Biological Classification of Relapse After CAR-T Therapy in MM

Tumor Intrinsic Mechanisms: BCMA Downregulation, Antigen Escape, and Clonal Evolution

In the biological classification of MM CAR-T relapse, tumor intrinsic mechanisms should be considered separately because residual plasma cell clones are reselected by immune pressure at three levels, target antigen density, antigen structure, and subclone composition. This framework helps explain the clinical paradox that, despite the high response rates reported in studies such as KarMMa and CARTITUDE-1, a substantial proportion of patients still experience disease progression after BCMA CAR-T therapy. 11 , 12 , 49 , 50

BCMA downregulation is one of the most common and easily underestimated forms of tumor intrinsic resistance because BCMA CAR-T cells require a sufficient antigen density threshold to recognize and eliminate MM cells. When surface BCMA expression on myeloma cells decreases, even without complete antigen loss, CAR-T activation may become inefficient or subthreshold, resulting in residual MRD, gradual biochemical relapse, or focal progression in the bone marrow or at extramedullary sites. 33 , 51 Real-world evidence further supports this interpretation. Rana et al analyzed patients with RRMM who received standard-of-care BCMA CAR-T therapy and found that higher pretreatment BCMA expression was associated with deeper remission and longer time to progression. During relapse, BCMA remained detectable in most cases, but its overall expression was reduced, suggesting that antigen downregulation is more common than complete antigen loss. 33 BCMA downregulation is also biologically plastic and potentially targetable. Gamma secretase can cleave membrane BCMA and generate soluble BCMA, leading to reduced BCMA expression on tumor cells, increased serum soluble BCMA, and insufficient effective contact with CAR-T cells. 51–53 For instance, Cowan et al combined gamma secretase inhibitors with BCMA CAR-T therapy and showed that these agents can increase BCMA density on target cells and support CAR-T function. Thus, restoration of antigen density may represent an actionable step for nanodelivery of small molecules or local drug delivery systems. 51 In contrast to BCMA downregulation, antigen escape represents a more specific form of target mismatch. Typical mechanisms include biallelic deletion, truncating mutations, or structural variations affecting antibody or CAR recognition epitopes at the TNFRSF17 or BCMA locus. Samur et al reported relapse after BCMA CAR-T therapy caused by biallelic BCMA inactivation. Da Via et al described TNFRSF17 homozygous deletion clones selected after BCMA CAR-T therapy in KarMMa-related cases. Lee et al showed through whole-genome sequencing that target antigen gene deletion and extracellular domain mutations can occur after BCMA or GPRC5D-targeted therapy. 34–36 Antigen escape is not limited to a single CAR-T intervention and may be amplified across sequential BCMA-directed therapies. Firestone et al reported that patients previously treated with belantamab mafodotin or BCMA CAR-T therapy may fail to respond to subsequent teclistamab because of treatment-related BCMA loss. This finding suggests a risk of shared antigen escape among antibody drug conjugates, CAR-T cells, and bispecific antibodies targeting BCMA. 38

Trogocytosis represents a contact-dependent and potentially reversible form of antigen remodeling that is mechanistically distinct from genomic BCMA loss. Following CAR engagement, formation of a stable immunological synapse promotes CAR clustering, close membrane apposition, and actin-dependent mechanical pulling, allowing antigen-containing fragments of the myeloma-cell membrane to be transferred to the CAR-T-cell surface. This process can simultaneously decrease antigen density on surviving tumor cells and generate antigen-positive CAR-T cells that become susceptible to fratricide and repeated-stimulation-associated exhaustion. 54 , 55 From a nanomedicine perspective, membrane-camouflaged nanoparticles could be designed to intervene directly at this bio-interface. Tumor-cell-membrane-coated or hybrid-membrane-coated nanoparticles can retain native membrane proteins and homotypic adhesion properties, thereby promoting their localization at the tumor-cell surface. 56 , 57 When combined with a nonfusogenic membrane coating and a mechanically constrained nanoparticle core, these particles could, in principle, act as transient interfacial spacers around antigen-rich membrane domains, increasing the effective distance between the CAR-T-cell and tumor-cell membranes and reducing the sustained CAR clustering, membrane deformation, and membrane-fragment extraction required for trogocytosis. In addition, stimulus-responsive nanoparticles could locally release short-acting regulators of integrin-dependent adhesion or actin remodeling after CAR engagement, thereby limiting prolonged mechanical pulling at the immunological synapse. 56 , 57 However, the use of membrane-camouflaged nanoparticles to prevent CAR-T-cell trogocytosis remains a mechanistic design hypothesis and has not yet been directly demonstrated in BCMA-directed CAR-T therapy for MM. Because excessive interfacial shielding may also impair productive immune-synapse formation and tumor-cell killing, future studies should optimize membrane fusogenicity, particle rigidity, ligand density, and release duration, while simultaneously quantifying tumor-cell BCMA loss, BCMA acquisition by CAR-T cells, CAR-T-cell fratricide, and antimyeloma cytotoxicity. Beyond this reversible, contact-dependent form of antigen redistribution, clonal evolution can generate more stable forms of antigen escape, and BCMA loss is not always a newly acquired random event after treatment. Some patients already harbor monoallelic TNFRSF17 deletion or very low frequency antigen negative subclones before treatment. Strong selection pressure after CAR-T therapy can drive the expansion of these subclones into dominant clones. Large scale genomic studies have shown that monoallelic BCMA deletion is not uncommon in MM. 36–38 Recurrent MM also exhibits substantial genetic heterogeneity. Abnormalities in RAS or MAPK, NF-κB, cell cycle regulation, and DNA damage repair pathways can coexist with antigenic changes. Therefore, relapse should not be regarded as a single BCMA-negative event, but rather as the combined result of target antigen remodeling, enhanced proliferative fitness, and coevolution of drug tolerant subclones. 58 Recent studies using whole-genome sequencing and treatment-related mutational signatures for temporal mapping further suggest that some antigen gene loss is acquired under the pressure of T-cell-redirecting therapy rather than being entirely preexisting. These findings indicate that relapse monitoring requires dynamic, longitudinal, and multiregional sampling rather than reliance on a single pretreatment bone marrow BCMA assessment. 59 Therefore, in the context of nanomedicine enabled MM CAR-T therapy, tumor intrinsic relapse mechanisms indicate that future strategies should move beyond generating more potent CAR-T cells ex vivo. A comprehensive platform is needed to combine bio-interface engineering designed to limit trogocytosis with antigen density imaging, soluble BCMA capture and monitoring, precise delivery of gamma secretase inhibitors or epigenetic modulators, delivery of multitarget CAR components, and real-time tracking of circulating tumor DNA or nanosensor signals within the bone marrow niche. Dynamic monitoring of circulating tumor DNA after CAR-T therapy can reflect bone marrow MRD and progression risk, providing a more continuous window for intervention before clinical relapse than conventional bone marrow aspiration. 60

Taken together, BCMA downregulation, antigen escape, and clonal evolution constitute the tumor side triad of MM CAR-T relapse. Accordingly, the diagnostic approach should shift from empiric retreatment with BCMA-directed therapies after relapse to reassessment of antigen status, reconstruction of genomic profiles, and reevaluation of clonal architecture at relapse. These results can then guide the selection of antigen enhancement strategies, dual or multitarget CAR-T therapy, therapies targeting non-BCMA antigens, or nanomedicine assisted intervention in the bone marrow niche. 33 , 34 , 58 , 60

CAR-T Intrinsic Mechanisms: Exhaustion, Impaired Memory Phenotype, and Metabolic Adaptation Failure

Antigen escape explains how myeloma cells evade CAR-T cells, whereas CAR-T intrinsic mechanisms address another key question: why some CAR-T cells gradually lose function in the chronically inflamed, nutrient limited, and highly immunosuppressive bone marrow niche, even after recognizing their target cells. Therefore, analysis of exhaustion, impaired memory phenotype, and metabolic adaptation failure not only improves our understanding of MM CAR-T relapse, but also identifies actionable targets for nanomedicine based interventions during ex vivo manufacturing, in vivo reprogramming, and intervention before relapse. 11 , 12 , 49 Clinically, ide-cel and cilta-cel have shown that BCMA CAR-T therapy can induce deep remission in a high proportion of patients with relapsed or refractory MM. However, long-term follow-up indicates that some patients still experience disease progression, suggesting that relapse is not always driven solely by BCMA loss. Instead, relapse often occurs together with insufficient CAR-T expansion, poor persistence, and functional decline. 11 , 12 , 49 For instance, Lin et al conducted an 18 month follow-up of early clinical studies of ide-cel and found that patients with PFS of at least 18 months had a higher proportion of naive T cells at leukapheresis, fewer exhausted T cells, and better functional phenotypes in the final CAR-T products. These findings indicate that the quality of input T cells, even before treatment and during cell collection, partly determines the depth and duration of subsequent remission. 39 Ledergor et al further showed through single-cell analysis that patients with sustained remission had stronger CAR-T persistence and a higher proportion of CD8 effector memory CAR-T cells. In contrast, patients with early relapse showed early expansion of cytotoxic CD4 CAR-T cells, which also expressed exhaustion related molecules such as HAVCR2, TIM-3, and TIGIT. These findings suggest that transient cytotoxic activity may evolve into rapid functional exhaustion when stable memory and long-term self renewal are insufficient ( Figure 1 ). 41 Consistent with these observations, Li et al reported a reduced proportion of T cells and increased frequencies of exhausted T cells and TIGIT positive NK cells in MM bone marrow samples after BCMA CAR-T relapse. They also observed a marked increase in exhausted CD8 effector T cells at relapse compared with baseline. These results suggest that CAR-T relapse is not simply caused by insufficient cell numbers, but by continuous pressure from the bone marrow microenvironment that drives T and NK cells toward inhibitory and hyporesponsive states. 40
Figure 1 Single-cell profiling identifies CAR-T cell states associated with response durability in multiple myeloma. Serial PBMC and bone marrow samples from healthy donors and patients with multiple myeloma receiving CAR-T therapy were analyzed using FACS and 5′ single-cell RNA sequencing. Patients with sustained remission exhibited greater CAR-T persistence and enrichment of CD8+ effector memory CAR-T cells. In contrast, patients with early relapse showed early expansion of cytotoxic CD4+ CAR-T cells expressing exhaustion-associated markers, including HAVCR2/TIM-3 and TIGIT. These findings suggest that transient cytotoxic activation, in the absence of stable memory formation and long-term self-renewal, may contribute to rapid CAR-T functional exhaustion and relapse. Adapted from Ledergor, Guy et al CD4+ CAR-T cell exhaustion associated with early relapse of multiple myeloma after BCMA CAR-T cell therapy.  Blood advances . Copyright © 2024 by authors. 41 The diagram illustrates the process of analyzing CAR-T cell states in myeloma patients and healthy donors. It includes age-matched healthy donors (n = 5) and myeloma patients receiving CAR-T therapy (n = 15). Serial samples from PBMC and bone marrow are processed using FACS and 5′ single-cell RNA sequencing (GEX, TCR, TotalSeq). The data is analyzed using UMAP1 and UMAP2. The lower section shows the myeloma niche with myeloid cells, myeloma cells and CAR-T cells. TGF-beta is indicated near myeloid cells. CD4+ CAR-T cells in transient responders exhibit high levels of exhaustion markers, high levels of cytotoxic genes and are detected at an early stage (Day 14-30 post-infusion). Diagram of CAR-T cell analysis in myeloma patients and healthy donors, showing cell states and exhaustion markers.

Single-cell profiling identifies CAR-T cell states associated with response durability in multiple myeloma. Serial PBMC and bone marrow samples from healthy donors and patients with multiple myeloma receiving CAR-T therapy were analyzed using FACS and 5′ single-cell RNA sequencing. Patients with sustained remission exhibited greater CAR-T persistence and enrichment of CD8+ effector memory CAR-T cells. In contrast, patients with early relapse showed early expansion of cytotoxic CD4+ CAR-T cells expressing exhaustion-associated markers, including HAVCR2/TIM-3 and TIGIT. These findings suggest that transient cytotoxic activation, in the absence of stable memory formation and long-term self-renewal, may contribute to rapid CAR-T functional exhaustion and relapse. Adapted from Ledergor, Guy et al CD4+ CAR-T cell exhaustion associated with early relapse of multiple myeloma after BCMA CAR-T cell therapy.  Blood advances . Copyright © 2024 by authors. 41

Moreover, longitudinal single-cell transcriptomic and TCR sequencing studies by Fang et al also showed that, in patients with long-term remission, CAR-T clones gradually shifted from multiclonal expansion to persistent dominant clones with cytotoxic and NK like features. In contrast, patients with short-term remission showed earlier CAR-T exhaustion, increased Treg abundance, and loss of detectable CAR-T cells during progression. These findings further support a continuous axis linking clonal persistence, effector differentiation, and resistance to immunosuppression as determinants of relapse risk. 61 An impaired memory phenotype is a central component of this axis. Central memory T cells, stem cell like memory T cells, and some early effector memory T cells generally have greater proliferative potential, long-term survival capacity, and secondary response capacity after antigen reencounter. In contrast, exhausted CAR-T cells generated by excessive terminal differentiation or prolonged antigen stimulation are prone to reduced cytokine secretion, impaired sustained cytotoxicity, and attenuation in vivo. 39 , 41 , 61 The longitudinal single cell atlas of BCMA CAR-T therapy constructed by Rade et al suggests that the expansion and functional programs of distinct CAR-T cell populations after cilta-cel and ide-cel treatment are associated with complete remission and toxicity. In patients who did not achieve complete remission, CD8 T cells showed impaired response programs. This indicates that CAR-T efficacy is not determined solely by peak expansion, but also depends on whether expanded cells can maintain an appropriate effector memory balance. 62 Insufficient T-cell reserves before manufacturing may also increase the risk of relapse. In a Japanese national cohort, Jo et al found that the manufacturing failure rate for ide-cel was 8.4%. Low platelet count and a low CD4 to CD8 ratio were significantly associated with manufacturing failure, while recent exposure to alkylating agents was associated with poorer platelet count and CD4 to CD8 status. These findings suggest that patients with MM after multiple lines of therapy often enter the CAR-T process with depleted T-cell starting material. 63

In addition to phenotypic aging, immune checkpoint driven functional exhaustion is another targetable intrinsic limitation of CAR-T cells. Ouyang et al generated PD-1 downregulated BCMA CAR-T cells and found that they preserved the memory phenotype, reduced exhaustion, and enhanced antitumor activity. 64 Yuti et al reported that fifth generation anti-BCMA CAR-T cells secreting anti-PD-L1 scFv maintained cytotoxicity, proliferation, and cytokine release, while alleviating T-cell exhaustion in PD-L1 overexpressing myeloma models. 65 Metabolic adaptation failure is a third CAR-T intrinsic mechanism that has gained increasing attention in recent years. In the myeloma bone marrow niche, tumor cells, myeloid cells, and stromal cells compete for glucose, glutamine, and oxygen. They also impair mitochondrial homeostasis and glycolytic capacity in CAR-T cells through lactate, inflammatory mediators, and oxidative stress. 40 , 42 Navarro et al demonstrated that glutamine deprivation can weaken antigen specific T-cell proliferation and IFN-γ production. BCMA CAR-T cells are particularly sensitive to low glutamine conditions. Overexpression of the glutamine transporter ASCT2 or SLC1A5 can enhance mTORC1 signaling, basal oxygen consumption, glycolysis, in vivo persistence, and antitumor efficacy in MM models. These findings indicate that insufficient metabolic plasticity may itself serve as an intrinsic biological basis for CAR-T relapse. 42 From this perspective, exhaustion, impaired memory phenotype, and metabolic adaptation failure are not separate phenomena. Together, they constitute a CAR-T dysfunction triad. Poor quality of initial T cells biases the final product toward terminal differentiation. Bone marrow mediated immunosuppression and repeated antigen stimulation increase checkpoint signaling. Nutrient stress and mitochondrial pressure further weaken expansion, cytokine release, and sustained cytotoxicity. These processes ultimately manifest as transient CAR-T expansion followed by disappearance, MRD repositivity, and clinical relapse. 39–41 For nanomedicine enabled MM CAR-T therapy, this classification has direct translational significance. Nanodelivery systems can be designed to deliver siRNA, mRNA, metabolic regulators, or epigenetic modulators during ex vivo manufacturing to promote early memory like CAR-T phenotypes. They may also deliver modules for checkpoint regulation, mitochondrial protection, and nutrient sensing reprogramming to CAR-T cells or the bone marrow niche in vivo. In this way, relapse management may shift from treatment after relapse to prevention before CAR-T functional decline. 64 , 65

Microenvironmental Mechanisms: Bone Marrow Stroma, Endothelial Barrier, Myeloid Cells, Extracellular Vesicles, and Hypoxia

MM CAR-T relapse is not explained solely by insufficient CAR-T potency or loss of BCMA expression on myeloma cells. It can also be viewed as a spatial immune failure within the bone marrow niche. CAR-T cells must cross the vascular barrier, enter tumor niches, resist stromal and myeloid cell mediated inhibition, and maintain cytotoxicity in a hypoxic and extracellular vesicle enriched environment. Disruption of any of these steps by the bone marrow microenvironment may shift an initial deep remission toward residual MRD and clinical relapse. 43 , 66 Key clinical studies of BCMA CAR-T therapy have shown that ide-cel and cilta-cel can induce remission in a high proportion of patients. However, relapse during long-term follow-up remains common. This suggests that relapse cannot be explained solely by tumor intrinsic antigen loss and that extrinsic constraints imposed by the bone marrow niche on CAR-T migration, expansion, and effector function must also be considered. 11 , 12 Sakemura et al demonstrated in MM models that bone marrow–derived cancer associated fibroblasts, also known as BM CAFs, can promote myeloma progression and inhibit the antitumor activity of BCMA CAR-T cells. Dual target CAR-T cells directed against malignant plasma cells and FAP or SLAMF7 positive CAFs improved effector function and prolonged survival in mice. These findings provide experimental evidence that stromal cells themselves can serve as relapse associated therapeutic targets. 43 Bone marrow stromal cells, CAFs, and the extracellular matrix can also form protective niches through adhesion and chemotactic pathways involving TGF-β, IL-6, APRIL, CXCL12 and CXCR4, and VCAM-1 and VLA-4. These signals allow myeloma cells to acquire antiapoptotic and immune escape advantages and may keep CAR-T cells in a state of reduced migration, low activation, or early exhaustion before tumor contact. 40 , 43 , 44 For example, Alabanza et al generated TGF-β resistant armed BCMA CAR-T cells and showed that conventional BCMA CAR-T cytotoxicity and multifunctionality were inhibited in a TGF-β overexpression model. In contrast, CAR-T cells expressing dominant negative TGFBRII maintained proliferation, degranulation, cytokine secretion, and in vivo antitumor activity. These findings indicate that the bone marrow microenvironment can directly contribute to CAR-T failure through defined soluble inhibitory factors. 44 The endothelial barrier represents another often overlooked extrinsic mechanism. MM cells do not distribute uniformly within the bone marrow cavity. Instead, they depend on the bone marrow vascular endothelium, perivascular matrix, and chemokine axes for homing, extravasation, nutrient acquisition, and focal expansion. 67 Guimarães et al developed lipid polymer nanoparticles to deliver siRNA to the bone marrow endothelium and target cyclophilin A secreted by bone marrow endothelial cells, thereby inhibiting MM cell extravasation and bone marrow colonization. This indicates that the endothelial barrier and vascular niche are not only physical routes for CAR-T entry into the bone marrow, but also therapeutic compartments that can be reshaped by nanomedicine ( Figure 2 ). 67 The clinical relevance of this abnormal vascular niche is also reflected in extramedullary disease. Real world studies by Dima et al showed that patients with RRMM and active extramedullary lesions had worse outcomes after standard-of-care CAR-T therapy. From the perspective of the immune microenvironment of extramedullary lesions, Qi et al further suggested that local T-cell infiltration, myeloid suppression, and immune checkpoint status may influence the efficacy of anti-BCMA CAR-T therapy. 47 , 48
Figure 2 Lipid–polymer nanoparticle-mediated siRNA delivery to the bone marrow vascular niche for multiple myeloma treatment.( A ) Schematic representation of the CyPA–CD147 signaling axis involved in multiple myeloma cell trafficking, extravasation, and colonization within the bone marrow. Created with BioRender.com. ( B ) Workflow of an in vivo high-throughput screening strategy used to identify nanoparticle formulations with efficient bone marrow delivery. Fifteen lipid–polymer nanoparticle formulations carrying distinct DNA barcodes were pooled and intravenously administered, followed by tissue isolation, barcode sequencing, and comparison of normalized delivery across different organs. The top-performing bone marrow-targeting formulation was subsequently selected and loaded with CyPA siRNA to suppress CyPA-dependent myeloma progression within the bone marrow microenvironment. ( C ) Cryo-transmission electron microscopy image showing the morphology, nanoscale size, and multilamellar structure of the selected nanoparticle formulation; scale bars, 100 nm in the main image and 25 nm in the inset. ( D ) Dynamic light scattering-derived hydrodynamic size distribution of the lead B1 nanoparticle formulation, plotted as normalized intensity on the y-axis against particle diameter (nm) on the x-axis. ( E ) Zeta potential of the B1 nanoparticle formulation, with the formulation indicated on the x-axis and zeta potential (mV) displayed on the y-axis. Data are presented as mean ± SD. Adapted from Guimarães, Pedro P G et al In vivo bone marrow microenvironment siRNA delivery using lipid-polymer nanoparticles for multiple myeloma therapy.  Proceedings of the National Academy of Sciences of the United States of America . Copyright © 2023 by authors. 67 The schematic illustrates the process of lipid to polymer nanoparticle-mediated siRNA delivery to the bone marrow for multiple myeloma treatment. A shows the migration of multiple myeloma cells from the bloodstream to the bone marrow, involving BMEC, eCyPA, MM cells and CD147. B details the workflow of nanoparticle formulation, starting with polymer-lipid hybrid, PEG-lipid and DNA barcode, followed by microfluidic mixing to create 15 nanoparticle formulations. These are pooled and administered to mice, with tissue isolation and deep sequencing to identify top-performing formulations. C displays a cryo-transmission electron microscopy image of the nanoparticle morphology, with scale bars at 100 nm and 25 nm. D presents a graph of the hydrodynamic size distribution of the B1 nanoparticle formulation, with intensity on the y-axis and diameter in nanometers on the x-axis. E shows a graph of the zeta potential of the B1 formulation, with zeta potential in millivolts on the y-axis. A schematic of nanoparticle delivery to bone marrow for multiple myeloma treatment.

Lipid–polymer nanoparticle-mediated siRNA delivery to the bone marrow vascular niche for multiple myeloma treatment.( A ) Schematic representation of the CyPA–CD147 signaling axis involved in multiple myeloma cell trafficking, extravasation, and colonization within the bone marrow. Created with BioRender.com. ( B ) Workflow of an in vivo high-throughput screening strategy used to identify nanoparticle formulations with efficient bone marrow delivery. Fifteen lipid–polymer nanoparticle formulations carrying distinct DNA barcodes were pooled and intravenously administered, followed by tissue isolation, barcode sequencing, and comparison of normalized delivery across different organs. The top-performing bone marrow-targeting formulation was subsequently selected and loaded with CyPA siRNA to suppress CyPA-dependent myeloma progression within the bone marrow microenvironment. ( C ) Cryo-transmission electron microscopy image showing the morphology, nanoscale size, and multilamellar structure of the selected nanoparticle formulation; scale bars, 100 nm in the main image and 25 nm in the inset. ( D ) Dynamic light scattering-derived hydrodynamic size distribution of the lead B1 nanoparticle formulation, plotted as normalized intensity on the y-axis against particle diameter (nm) on the x-axis. ( E ) Zeta potential of the B1 nanoparticle formulation, with the formulation indicated on the x-axis and zeta potential (mV) displayed on the y-axis. Data are presented as mean ± SD. Adapted from Guimarães, Pedro P G et al In vivo bone marrow microenvironment siRNA delivery using lipid-polymer nanoparticles for multiple myeloma therapy.  Proceedings of the National Academy of Sciences of the United States of America . Copyright © 2023 by authors. 67

Myeloid cells constitute an immune braking system within the microenvironment of MM CAR-T relapse. Using high dimensional single-cell analysis, Dhodapkar et al found that sustained remission after BCMA CAR-T therapy was associated with the composition of non-CAR-T immune cells in the bone marrow. Patients with higher T-cell diversity and fewer myeloid cells had longer PFS, indicating that CAR-T efficacy is shaped by interactions among CAR-T cells, endogenous T cells, myeloid cells, dendritic cells, and tumor cells. 66 Li et al performed scRNA-seq analysis in patients with relapse after BCMA CAR-T therapy and found that the proportion of bone marrow T cells decreased during relapse, whereas the proportion of monocytes and macrophages increased. IL1β high, S100A9 high, MARCO high, and interferon responsive macrophages were increased. Cell communication analysis suggested that MIF and APRIL signaling may participate in the post relapse immunosuppressive network. 40 Yang et al further showed that, after BCMA CAR-T therapy combined with lymphodepletion, treatment response was positively correlated with the abundance of CD8 positive effector CAR-T cells and target-specific cytotoxic activity on day 28. A decrease in Treg abundance was also associated with efficacy. These findings indicate that, beyond peak CAR-T expansion, the quality of bone marrow immune reconstitution after treatment also influences relapse risk. 45 Exosomes and extracellular vesicles can be regarded as a nanoscale communication system within the myeloma microenvironment. Menu and Vanderkerken summarized that MM derived exosomes carry proteins, miRNAs, and immunoregulatory molecules that can promote angiogenesis, bone disease, drug tolerance, and immunosuppression. These vesicles may also serve as liquid biopsy markers and delivery carriers. 68 Lopes et al demonstrated that MM derived extracellular vesicles can reshape the bone marrow immune microenvironment. 69 Tian et al found that MSC derived extracellular vesicle miR let-7c can enhance bone marrow angiogenesis by inducing M2 macrophage polarization. These findings connect extracellular vesicles, myeloid polarization, and endothelial abnormalities within a shared pro relapse pathway. 70 More direct immunologic evidence has been provided by Hagiwara et al showed that MM derived extracellular vesicles can upregulate immune checkpoints related to T-cell exhaustion. 71 Soncini et al found that bone marrow–derived extracellular vesicles from patients with MM were enriched in HLA-G, PD-1, and PD-L1 and may promote adaptive immune dysfunction. These findings suggest that CAR-T cells may be chronically exposed to vesicle-mediated inhibitory signals before relapse. 72 Hypoxia is a key metabolic and ecological factor that links antigen escape, secretome changes, and immunosuppression. The MM bone marrow is relatively hypoxic, and increasing tumor burden further strengthens HIF related programs, vascular abnormalities, and metabolic stress in immune cells. 46 In a 1% O 2 hypoxia model, Tu et al found that hypoxia reduced surface and total BCMA expression in MM cells, weakened BCMA CAR mediated signaling, and induced hypoxia related secretory programs and small extracellular vesicle release. These changes significantly reduced CAR-T cytotoxicity and the secretion of IFN-γ, IL-2, and TNF-α. Gamma secretase inhibitors partially reversed this inhibition, indicating that hypoxia is not merely background variation. Instead, hypoxia is a complex relapse promoting factor that can simultaneously drive BCMA downregulation, extracellular vesicle release, and CAR-T functional inhibition. 46

Therefore, future nanoplatforms can be designed to support bone marrow–targeted delivery, endothelial barrier regulation, CAF and myeloid cell reprogramming, extracellular vesicle capture or blockade, hypoxia-responsive drug release, and local CAR-T reactivation. Such strategies could shift relapse intervention from simply changing targets to reshaping the bone marrow ecosystem in which CAR-T cells can function. 46 , 67 , 70

Nanotechnology Reshapes MM CAR-T Manufacturing: From ex vivo Engineering to in vivo Generation

BCMA CAR-T therapy has changed the therapeutic ceiling of multiple myeloma, but its manufacturing model still belongs to a slow and individualized era. In relapsed or refractory MM, the waiting time between cell collection and infusion can become a biological risk window in which tumor burden increases, T-cell fitness declines, and bridging therapy reshapes the immune landscape. Nanotechnology offers a new manufacturing grammar for CAR-T therapy: not only improving ex vivo transfection, activation, and expansion, but also enabling CAR genes to be delivered directly to endogenous T cells in vivo. Through mRNA-LNPs, nanoengineered activation platforms, targeted polymeric particles, and surface-modified viral vectors, CAR-T therapy may gradually shift from a patient-specific cell product to an injectable and programmable immune intervention. Next, we summarize this transition from ex vivo engineering to in vivo CAR-T generation and highlight its implications for faster, more accessible, and more controllable MM treatment ( Table 3 ).
Table 3 The Application of Nanotechnology in Reshaping CAR-T Manufacturing for MM Manufacturing Direction Nanotechnology/ Platform Target or Model Key Findings Relevance to MM CAR-T Manufacturing Current Limitations Ref. Conventional autologous ex vivo CAR-T as the manufacturing baseline Viral-vector-based ex vivo T-cell collection, activation, transduction, expansion, release testing and reinfusion BCMA CAR-T in RRMM; ide-cel and cilta-cel Ide-cel and cilta-cel established BCMA CAR-T as a deep-remission strategy in RRMM, but their production depends on individualized cell manufacturing and lymphodepletion. Defines the bottleneck that nanotechnology aims to solve: long vein-to-vein time, high cost, manufacturing failure risk, and reduced T-cell fitness in heavily pretreated MM patients. Still requires leukapheresis, centralized GMP manufacturing, bridging therapy, and complex logistics. [ 11 ] Nanobody/single-domain antibody CAR design Compact VHH or nanobody recognition domains incorporated into CAR constructs BCMA or alternative-antigen CAR-T in MM Cilta-cel uses two BCMA-targeting single-domain antibodies; newer nanoCAR designs such as B7-H3 nanoCARs show activity against MM cells and may help address antigen escape. Nanobody-based CARs may improve construct compactness, stability, antigen-binding geometry, and dual-target engineering, supporting more flexible CAR design for BCMA-low or BCMA-negative relapse. Nanobody design alone does not solve manufacturing time; off-tumor antigen expression and tonic signaling still require validation. [ 12 ] Ex vivo mRNA-LNP CAR-T engineering Lipid nanoparticles delivering CAR mRNA to T cells outside the body Mainly CD19 CAR-T models; generalizable to BCMA/GPRC5D/FcRH5 designs LNP delivery of CAR mRNA can outperform electroporation, with prolonged CAR-mRNA persistence, less cytotoxicity, slower proliferation, and lower exhaustion while maintaining in vitro function. Offers a transient, non-integrating, potentially faster and safer engineering route for MM CAR-T, especially for short-term bridging, repeat dosing, or safety-sensitive settings. Transient CAR expression may be insufficient for high-burden RRMM unless repeat dosing or persistence-enhancing strategies are added. [ 73 ] One-step activation and transfection Antigen-presenting-cell-mimetic activating LNPs carrying CD3/CD28 signals and CAR mRNA Anti-CD19 mRNA CAR-T model Activating LNPs combine T-cell activation and CAR mRNA delivery, allowing bead-free, one-step generation of functional mRNA CAR-T cells; the workflow can be shortened from about 48 h to 24 h. Provides a manufacturing concept for rapid MM CAR-T preparation without separate bead activation/removal; potentially useful for urgent RRMM cases needing fast cell products. Evidence remains preclinical and mostly CD19-based; translation to BCMA or GPRC5D MM CAR-T requires disease-specific validation. [ 74 ] Nanoparticle-assisted T-cell expansion Dextran-based nano-aAPC “T-Expand” conjugated with anti-CD3 and anti-CD28 CD19 CAR-T manufacturing model T-Expand nanoparticles expand T cells with efficiency comparable to commercial microbeads, reduce prolonged stimulation, avoid extra separation steps, and generate CAR-T cells with improved persistence and antitumor efficacy. Could improve the quality of MM CAR-T starting material by reducing exhaustion during expansion, a key issue in heavily pretreated RRMM patients. Not yet MM-specific; large-scale GMP reproducibility and compatibility with BCMA/GPRC5D CAR manufacturing need testing. [ 75 ] Transient in vivo CAR-T generation by targeted LNPs CD5-targeted LNPs delivering modified CAR mRNA FAP-CAR T cells in cardiac fibrosis model CD5-targeted LNPs generated transient CAR-T cells in vivo without ex vivo manufacturing, with CAR expression detected rapidly and declining over time. Provides the foundational proof that endogenous T cells can be programmed in vivo by LNPs; for MM, this supports future injectable BCMA/GPRC5D/FcRH5 CAR mRNA platforms. Disease model was fibrosis, not MM; transient expression may require repeated dosing for plasma-cell malignancies. [ 30 ] Antibody-conjugated LNPs for in vivo CAR-T Extrahepatic ionizable LNPs modified with anti-CD3, anti-CD5, or anti-CD7 antibodies CD19 CAR mRNA delivery in mice Antibody-conjugated LNPs redirected delivery toward T cells and generated functional CAR-T cells in vivo; CD3- and CD7-targeted LNPs induced CAR expression and B-cell depletion. Suggests that T-cell-selective LNPs could replace ex vivo BCMA CAR-T manufacturing with injectable, tunable CAR mRNA dosing. Requires optimization of T-cell subset targeting, repeat dosing, biodistribution, cytokine release, and non-T-cell off-target delivery. [ 76 ] Subset-specific in vivo CAR-T generation Targeted LNPs delivering mRNA to defined T-cell subsets CD19/CD20 CAR-T models; human samples, humanized mice, cynomolgus monkeys Targeted LNPs reprogrammed CD8-positive T cells in healthy donor and autoimmune patient samples; in vivo dosing-controlled tumors in humanized mice and depleted B cells in cynomolgus monkeys. Moves LNP-mediated in vivo CAR-T from mouse proof-of-concept toward translational large-animal evidence; relevant to scalable MM CAR-T access. Not yet BCMA/MM-specific; durability, repeatability, immune clearance, and CRS risk require clinical testing. [ 77 ] Stable nonviral in vivo CAR-T generation Targeted LNP system delivering minicircle DNA plus transposase mRNA; NCtx platform CD19 or CD19/CD22 CAR-T in humanized leukemia models NCtx enabled T-cell-specific nonviral DNA delivery and stable CAR integration in vivo, generating durable functional CAR-T cells and antitumor responses. Addresses the key MM tension between transient safety and durable tumor clearance; conceptually suitable for BCMA, GPRC5D, FcRH5, or dual-antigen CAR designs. Stable integration raises concerns about insertional risk, long-term clonal tracking, and controllable termination. [ 78 ] Biodegradable polymeric in vivo CAR-T generation Biodegradable targeted polymeric mRNA nanoparticles carrying CAR mRNA Anti-CD19 CAR mRNA in vivo Targeted biodegradable polymeric nanoparticles delivered anti-CD19 CAR mRNA in vivo and induced approximately 95% peripheral B-cell depletion and 50% splenic B-cell depletion in mice. Expands in vivo CAR-T manufacturing beyond LNPs, offering a potentially lower-cost and biodegradable polymer platform for future MM CAR programming. Early preclinical evidence; polymer biodistribution, inflammatory response, and repeated dosing safety remain unresolved. [ 79 ] Surface-engineered lentiviral in vivo CAR-T generation VivoVec: a T-cell-targeted lentiviral vector encoding a CAR and a rapamycin-responsive expansion module Anti-CD19 CAR-T; preclinical models Systemic administration directly transduced endogenous T cells and generated stably integrated CAR-T cells in vivo. Rapamycin activated the engineered signaling module, enabling pharmacologically controlled CAR-T-cell expansion May bypass leukapheresis and ex vivo manufacturing and could be adapted to BCMA or other MM targets requiring durable CAR expression. Not yet validated in MM; off-target transduction, biodistribution, immunogenicity, insertional genotoxicity, and long-term control require further evaluation. [ 80 ] MM-specific in vivo BCMA CAR-T clinical translation Nanobody-directed, immune-shielded lentiviral vector; ESO-T01 RRMM patients receiving in vivo anti-BCMA CAR-T ESO-T01 generated anti-BCMA CAR-T cells after single intravenous infusion without leukapheresis, ex vivo manufacturing, or lymphodepletion; in the Phase 1 report, 4 of 5 patients achieved objective responses, including 3 stringent complete remissions, but grade ≥3 adverse events and CRS occurred. Represents the most direct proof that in vivo CAR-T manufacturing can enter MM clinical practice; it reframes CAR-T from a personalized cell product into an injectable gene/cell-programming therapy. Very early clinical stage; small sample size, CRS, cytopenias, infection risk, off-target transduction, viral shedding, and long-term genotoxicity require close follow-up. [ 81 ]

Table 3 The Application of Nanotechnology in Reshaping CAR-T Manufacturing for MM Manufacturing Direction Nanotechnology/ Platform Target or Model Key Findings Relevance to MM CAR-T Manufacturing Current Limitations Ref. Conventional autologous ex vivo CAR-T as the manufacturing baseline Viral-vector-based ex vivo T-cell collection, activation, transduction, expansion, release testing and reinfusion BCMA CAR-T in RRMM; ide-cel and cilta-cel Ide-cel and cilta-cel established BCMA CAR-T as a deep-remission strategy in RRMM, but their production depends on individualized cell manufacturing and lymphodepletion. Defines the bottleneck that nanotechnology aims to solve: long vein-to-vein time, high cost, manufacturing failure risk, and reduced T-cell fitness in heavily pretreated MM patients. Still requires leukapheresis, centralized GMP manufacturing, bridging therapy, and complex logistics. [ 11 ] Nanobody/single-domain antibody CAR design Compact VHH or nanobody recognition domains incorporated into CAR constructs BCMA or alternative-antigen CAR-T in MM Cilta-cel uses two BCMA-targeting single-domain antibodies; newer nanoCAR designs such as B7-H3 nanoCARs show activity against MM cells and may help address antigen escape. Nanobody-based CARs may improve construct compactness, stability, antigen-binding geometry, and dual-target engineering, supporting more flexible CAR design for BCMA-low or BCMA-negative relapse. Nanobody design alone does not solve manufacturing time; off-tumor antigen expression and tonic signaling still require validation. [ 12 ] Ex vivo mRNA-LNP CAR-T engineering Lipid nanoparticles delivering CAR mRNA to T cells outside the body Mainly CD19 CAR-T models; generalizable to BCMA/GPRC5D/FcRH5 designs LNP delivery of CAR mRNA can outperform electroporation, with prolonged CAR-mRNA persistence, less cytotoxicity, slower proliferation, and lower exhaustion while maintaining in vitro function. Offers a transient, non-integrating, potentially faster and safer engineering route for MM CAR-T, especially for short-term bridging, repeat dosing, or safety-sensitive settings. Transient CAR expression may be insufficient for high-burden RRMM unless repeat dosing or persistence-enhancing strategies are added. [ 73 ] One-step activation and transfection Antigen-presenting-cell-mimetic activating LNPs carrying CD3/CD28 signals and CAR mRNA Anti-CD19 mRNA CAR-T model Activating LNPs combine T-cell activation and CAR mRNA delivery, allowing bead-free, one-step generation of functional mRNA CAR-T cells; the workflow can be shortened from about 48 h to 24 h. Provides a manufacturing concept for rapid MM CAR-T preparation without separate bead activation/removal; potentially useful for urgent RRMM cases needing fast cell products. Evidence remains preclinical and mostly CD19-based; translation to BCMA or GPRC5D MM CAR-T requires disease-specific validation. [ 74 ] Nanoparticle-assisted T-cell expansion Dextran-based nano-aAPC “T-Expand” conjugated with anti-CD3 and anti-CD28 CD19 CAR-T manufacturing model T-Expand nanoparticles expand T cells with efficiency comparable to commercial microbeads, reduce prolonged stimulation, avoid extra separation steps, and generate CAR-T cells with improved persistence and antitumor efficacy. Could improve the quality of MM CAR-T starting material by reducing exhaustion during expansion, a key issue in heavily pretreated RRMM patients. Not yet MM-specific; large-scale GMP reproducibility and compatibility with BCMA/GPRC5D CAR manufacturing need testing. [ 75 ] Transient in vivo CAR-T generation by targeted LNPs CD5-targeted LNPs delivering modified CAR mRNA FAP-CAR T cells in cardiac fibrosis model CD5-targeted LNPs generated transient CAR-T cells in vivo without ex vivo manufacturing, with CAR expression detected rapidly and declining over time. Provides the foundational proof that endogenous T cells can be programmed in vivo by LNPs; for MM, this supports future injectable BCMA/GPRC5D/FcRH5 CAR mRNA platforms. Disease model was fibrosis, not MM; transient expression may require repeated dosing for plasma-cell malignancies. [ 30 ] Antibody-conjugated LNPs for in vivo CAR-T Extrahepatic ionizable LNPs modified with anti-CD3, anti-CD5, or anti-CD7 antibodies CD19 CAR mRNA delivery in mice Antibody-conjugated LNPs redirected delivery toward T cells and generated functional CAR-T cells in vivo; CD3- and CD7-targeted LNPs induced CAR expression and B-cell depletion. Suggests that T-cell-selective LNPs could replace ex vivo BCMA CAR-T manufacturing with injectable, tunable CAR mRNA dosing. Requires optimization of T-cell subset targeting, repeat dosing, biodistribution, cytokine release, and non-T-cell off-target delivery. [ 76 ] Subset-specific in vivo CAR-T generation Targeted LNPs delivering mRNA to defined T-cell subsets CD19/CD20 CAR-T models; human samples, humanized mice, cynomolgus monkeys Targeted LNPs reprogrammed CD8-positive T cells in healthy donor and autoimmune patient samples; in vivo dosing-controlled tumors in humanized mice and depleted B cells in cynomolgus monkeys. Moves LNP-mediated in vivo CAR-T from mouse proof-of-concept toward translational large-animal evidence; relevant to scalable MM CAR-T access. Not yet BCMA/MM-specific; durability, repeatability, immune clearance, and CRS risk require clinical testing. [ 77 ] Stable nonviral in vivo CAR-T generation Targeted LNP system delivering minicircle DNA plus transposase mRNA; NCtx platform CD19 or CD19/CD22 CAR-T in humanized leukemia models NCtx enabled T-cell-specific nonviral DNA delivery and stable CAR integration in vivo, generating durable functional CAR-T cells and antitumor responses. Addresses the key MM tension between transient safety and durable tumor clearance; conceptually suitable for BCMA, GPRC5D, FcRH5, or dual-antigen CAR designs. Stable integration raises concerns about insertional risk, long-term clonal tracking, and controllable termination. [ 78 ] Biodegradable polymeric in vivo CAR-T generation Biodegradable targeted polymeric mRNA nanoparticles carrying CAR mRNA Anti-CD19 CAR mRNA in vivo Targeted biodegradable polymeric nanoparticles delivered anti-CD19 CAR mRNA in vivo and induced approximately 95% peripheral B-cell depletion and 50% splenic B-cell depletion in mice. Expands in vivo CAR-T manufacturing beyond LNPs, offering a potentially lower-cost and biodegradable polymer platform for future MM CAR programming. Early preclinical evidence; polymer biodistribution, inflammatory response, and repeated dosing safety remain unresolved. [ 79 ] Surface-engineered lentiviral in vivo CAR-T generation VivoVec: a T-cell-targeted lentiviral vector encoding a CAR and a rapamycin-responsive expansion module Anti-CD19 CAR-T; preclinical models Systemic administration directly transduced endogenous T cells and generated stably integrated CAR-T cells in vivo. Rapamycin activated the engineered signaling module, enabling pharmacologically controlled CAR-T-cell expansion May bypass leukapheresis and ex vivo manufacturing and could be adapted to BCMA or other MM targets requiring durable CAR expression. Not yet validated in MM; off-target transduction, biodistribution, immunogenicity, insertional genotoxicity, and long-term control require further evaluation. [ 80 ] MM-specific in vivo BCMA CAR-T clinical translation Nanobody-directed, immune-shielded lentiviral vector; ESO-T01 RRMM patients receiving in vivo anti-BCMA CAR-T ESO-T01 generated anti-BCMA CAR-T cells after single intravenous infusion without leukapheresis, ex vivo manufacturing, or lymphodepletion; in the Phase 1 report, 4 of 5 patients achieved objective responses, including 3 stringent complete remissions, but grade ≥3 adverse events and CRS occurred. Represents the most direct proof that in vivo CAR-T manufacturing can enter MM clinical practice; it reframes CAR-T from a personalized cell product into an injectable gene/cell-programming therapy. Very early clinical stage; small sample size, CRS, cytopenias, infection risk, off-target transduction, viral shedding, and long-term genotoxicity require close follow-up. [ 81 ]

The Application of Nanotechnology in Reshaping CAR-T Manufacturing for MM

Nanodelivery in ex vivo CAR-T Manufacturing

Ex vivo CAR-T manufacturing essentially involves introducing a new immune program into patient derived T cells. Nanodelivery can shift this process from viral transduction and high damage electroporation toward synthetic delivery systems with tunable design, scalable production, and transient or stable expression. This provides a new process entry point for low cost, rapid, and functionally optimized BCMA CAR-T manufacturing in multiple myeloma. 82 , 83

BCMA CAR-T products such as ide-cel and cilta-cel have demonstrated the ability to induce deep remission in relapsed or refractory MM. However, their manufacturing still relies on complex procedures, including leukapheresis, T-cell activation, viral transduction, ex vivo expansion, and stringent quality control. These procedures lead to prolonged production timelines, substantial batch-to-batch variability, high costs, and limited suitability for patients with MM who have rapidly progressive disease or poor T-cell quality. 82 , 83 Pinto et al noted that conventional viral vectors pose safety, cost, and production challenges. Although electroporation can introduce mRNA or gene editing components, it may cause substantial cytotoxicity. Therefore, lipid and polymer nanosystems are considered important directions for next generation CAR-T gene delivery platforms. 83 Mechanistically, lipid nanoparticles are typically composed of ionizable lipids, cholesterol or cholesterol derivatives, helper phospholipids, and PEG lipids. In the acidic endosomal environment, these nanoparticles become protonated and promote endosomal escape, allowing CAR mRNA to be rapidly translated into CAR protein after entering the T-cell cytoplasm. This nonintegrating and transient expression profile is particularly suitable for controllable CAR-T strategies, bridging therapy, or settings in which long-term depletion of B cells or plasma cells should be avoided. 73 , 83 , 84 Billingsley et al used an orthogonal experimental design to screen the C14-4 LNP formulation and engineered human T cells with mRNA to express CAR. The optimized LNP achieved effective CAR expression and tumor killing while maintaining T-cell activity. This study provides a representative example of how CAR-T manufacturing can shift from empirical formulation to data driven nanoscale design. 29

In addition, Patel et al further found that partial replacement of cholesterol with 7α-hydroxycholesterol improved mRNA delivery efficiency in primary human T cells by increasing late endosome formation and reducing endosomal recycling. This indicates that nanoparticles are not merely lipid carriers for mRNA encapsulation. Instead, they can reshape T-cell engineering efficiency by regulating endocytic trafficking and endosomal escape. 84 Recently, Kitte et al performed a head-to-head comparison between LNP delivery and electroporation. They found that LNP engineered CAR-T cells showed longer CAR mRNA persistence, more durable CAR surface expression, lower cytotoxicity, and slower but more stable proliferation than electroporation engineered CAR-T cells. Their ex vivo function was comparable to that of stably transduced CAR-T cells, with a lower degree of exhaustion. These results suggest that the manufacturing process itself may influence the functional fate of the final CAR-T product. 73 Hamilton et al integrated immune checkpoint regulation into LNP design by constructing an ionizable LNP capable of delivering CAR mRNA while simultaneously modulating the PD-1 related inhibitory axis. They demonstrated that nanodelivery can combine CAR expression, checkpoint regulation, and functional enhancement within a single engineering step. 85 A more process oriented strategy is the antigen presenting cell mimicking lipid nanoparticle system proposed by Metzloff et al This system presents CD3 and CD28 stimulation signals on the LNP surface while simultaneously delivering CAR mRNA. It compresses the traditional manufacturing steps of magnetic bead activation and gene transfection into a one step activation and transfection process. The resulting mRNA CAR-T cells reduced tumor burden in mouse tumor models, indicating that nanomaterials can replace selected cell manufacturing units rather than only viral vectors ( Figure 3 ). 74
Figure 3 Antigen-presenting cell-mimicking activating lipid nanoparticles for streamlined ex vivo mRNA CAR-T cell engineering. ( a ) Schematic illustration of physiological T-cell activation in vivo, where antigen-presenting cells provide coordinated TCR/CD3 and CD28 costimulatory signals through peptide–MHC and CD80/CD86 interactions. This native immune synapse provides the biological rationale for designing bio-inspired ex vivo engineering platforms for CAR-T cell generation. ( b ) Comparison between the conventional magnetic bead-based manufacturing workflow and the activating lipid nanoparticle (LNP)-based workflow. In the bead-based process, T cells are first stimulated with CD3/CD28-coated activating beads, followed by bead removal and subsequent transfection with CAR mRNA-loaded LNPs, resulting in a multistep procedure requiring approximately 48 h. In contrast, activating LNPs integrate T-cell stimulation and CAR mRNA delivery into a single platform by presenting CD3 and CD28 antibody fragments on the LNP surface while encapsulating CAR mRNA. This one-step strategy enables simultaneous T-cell activation and CAR expression within approximately 24 h, thereby replacing selected manufacturing unit operations and supporting the generation of functional mRNA CAR-T cells with antitumor activity in mouse tumor models. Adapted from Metzloff AE, Padilla MS, Gong Net al Antigen Presenting Cell Mimetic Lipid Nanoparticles for Rapid mRNA CAR-T Cell Cancer Immunotherapy. Adv Mater . Copyright © 2024 by authors. 74 The illustration consists of two diagrams comparing workflows for CAR-T cell engineering. The first diagram, labeled ′In the body′, depicts physiological T-cell activation with antigen-presenting cells (APC) interacting with inactive T cells through CD80/CD86 and peptide-MHC. The second part shows two workflows. The ′Bead workflow (current gold standard)′ involves three steps: adding activating beads, removing beads and adding mRNA LNPs, taking 48 hours. The ′Activating LNP workflow (new workflow)′ simplifies the process to one step: adding activating LNPs, taking 24 hours. The legend identifies APC, inactive T cell, active T cell, CAR T cell, CD80/CD86, peptide-MHC, CD3 Ab, CD28 Ab, CD3 Ab frag, CD28 Ab frag, CD28 and TCR/CD3. The diagrams highlight the efficiency of the new LNP workflow compared to the traditional bead method. Two diagrams compare bead and LNP workflows for CAR-T cell engineering, showing steps and time differences.

Antigen-presenting cell-mimicking activating lipid nanoparticles for streamlined ex vivo mRNA CAR-T cell engineering. ( a ) Schematic illustration of physiological T-cell activation in vivo, where antigen-presenting cells provide coordinated TCR/CD3 and CD28 costimulatory signals through peptide–MHC and CD80/CD86 interactions. This native immune synapse provides the biological rationale for designing bio-inspired ex vivo engineering platforms for CAR-T cell generation. ( b ) Comparison between the conventional magnetic bead-based manufacturing workflow and the activating lipid nanoparticle (LNP)-based workflow. In the bead-based process, T cells are first stimulated with CD3/CD28-coated activating beads, followed by bead removal and subsequent transfection with CAR mRNA-loaded LNPs, resulting in a multistep procedure requiring approximately 48 h. In contrast, activating LNPs integrate T-cell stimulation and CAR mRNA delivery into a single platform by presenting CD3 and CD28 antibody fragments on the LNP surface while encapsulating CAR mRNA. This one-step strategy enables simultaneous T-cell activation and CAR expression within approximately 24 h, thereby replacing selected manufacturing unit operations and supporting the generation of functional mRNA CAR-T cells with antitumor activity in mouse tumor models. Adapted from Metzloff AE, Padilla MS, Gong Net al Antigen Presenting Cell Mimetic Lipid Nanoparticles for Rapid mRNA CAR-T Cell Cancer Immunotherapy. Adv Mater . Copyright © 2024 by authors. 74

For MM CAR-T therapy, this strategy is particularly valuable because patients often receive multiple lines of proteasome inhibitors, immunomodulatory drugs, anti-CD38 antibodies, and alkylating agents. Their T-cell starting material may be exhausted, aged, or poorly proliferative. Low toxicity and high efficiency LNP delivery may reduce additional manufacturing stress and preserve more early memory like and functionally adaptable T cells. 73 , 82 , 83 In addition to CAR mRNA delivery, nanodelivery can also support ex vivo gene editing and multimodule programming. Vavassori et al reported in Blood that LNP delivery of nuclease RNA significantly reduced T-cell death, cell cycle delay, metabolic disturbance, and inflammatory responses caused by electroporation. It also reduced p53 related damage signals in hematopoietic cells, suggesting that LNPs may enable safer CRISPR edited or universal CAR-T manufacturing. 86 The nonviral and nonintegrating DNA nanocarrier platform developed by Bozza et al demonstrated that minimized DNA vectors can rapidly and persistently generate recombinant T cells in a closed system while maintaining CAR-T cytotoxicity in several tumor models. This platform provides another translatable manufacturing route for stable expression while avoiding the risk of viral integration. 87 Recently, Pinto et al reported that polymer DNA nanoplatforms can deliver anti CD19 CAR genes to T cells and generate CAR-T cells with leukemia cell killing activity. Although this model is not specific to MM, its low cost, nonviral, and polymer regulated approach could be adapted for constructing CAR platforms targeting bone marrow tumor antigens such as BCMA, GPRC5D, or FcRH5. 88

From a clinical translation perspective, ex vivo nanodelivery for CAR-T manufacturing still needs to address several issues, including GMP grade consistency of LNP and polymer materials, T-cell subpopulation selectivity, mRNA expression kinetics, residual nanomaterials, endotoxin and complement activation, large-scale mixing, and release testing. However, these challenges are primarily engineering issues related to drug formulation and cell processing rather than insurmountable biological barriers. 74 , 86 Importantly, the safety requirements for ex vivo nanodelivery differ from those for systemic LNP administration: ex vivo exposure can be terminated through cell washing, purification, and product release testing, whereas in vivo administration additionally requires control of whole-body biodistribution, hepatic exposure, innate immune reactogenicity, complement activation, and unintended transfection of non-T cells. More importantly, ex vivo nanodelivery is not an isolated endpoint but a transitional platform toward in vivo CAR-T generation. A phase 1 study of in vivo anti-BCMA CAR-T generation has entered clinical validation in relapsed or refractory MM. This suggests that MM CAR-T manufacturing may evolve toward a parallel model that includes rapid ex vivo nanoprogramming, bedside short process manufacturing, and in vivo directed generation. 89 Accordingly, the transition from ex vivo nanoprogramming to systemic in vivo CAR-T generation should prioritize biodegradable, extrahepatic, and T-cell-selective formulations rather than directly repurposing conventional liver-tropic LNPs. Therefore, future material design should focus on controlling the intensity and duration of CAR expression, protecting T-cell adaptability through low damage delivery, and integrating CAR programming, exhaustion inhibition, and metabolic or homing optimization through multicargo delivery. These functions should be accompanied by predefined biodistribution, clearance, immunogenicity, and dose-stopping criteria to ensure that improved manufacturing convenience does not come at the cost of uncontrolled systemic exposure. These strategies may ultimately enable faster, more stable, and more personalized CAR-T product generation for patients with MM. 90 , 91

Nanomedicine-Enabled in vivo CAR-T Generation in MM

The disruptive potential of in vivo CAR-T generation lies in its attempt to transform CAR-T therapy from an individualized cell product requiring ex vivo collection, transduction, expansion, quality control, and reinfusion into an injectable gene or nanomedicine product that can be scaled and programmed within the patient. This strategy is particularly important for patients with multiple myeloma, because patients with RRMM may miss the optimal CAR-T treatment window because of rapid disease progression, poor T-cell quality, a high need for bridging therapy, and prolonged manufacturing wait times. 11 , 12 Conventional BCMA CAR-T products have demonstrated the ability to induce deep remission, but their manufacturing relies on leukapheresis, viral transduction, ex vivo expansion, and lymphodepletion. Therefore, in vivo CAR-T generation is considered a next generation approach to reducing time costs, manufacturing costs, and barriers to treatment accessibility. 11 , 12 , 82 At the nanomedicine level, a representative strategy is to encapsulate CAR mRNA, circular RNA, DNA, or gene editing components into lipid nanoparticles or polymer lipid hybrid particles. Particle tropism can then be redirected using T-cell targeting ligands such as CD3, CD5, CD7, or CD8, allowing preferential entry into T cells after intravenous injection and subsequent generation of CAR expressing cells through cytoplasmic translation or gene integration. 30 , 76 , 77 , 92 However, the choice of targeting ligand should be matched to the intended CAR-T-cell composition and its corresponding off-target profile. Anti-CD5 ligands provide relatively broad access to circulating T cells and have supported in vivo mRNA CAR-T generation, although CD5 is not absolutely restricted to T cells; anti-CD8 ligands can preferentially program cytotoxic T-cell subsets but may exclude the CD4-positive helper compartment; anti-CD7 ligands can facilitate efficient lymphocyte uptake but may also recognize natural killer cells; and anti-CD3 ligands provide pan-T-cell coverage but require careful engineering because excessive CD3 cross-linking may induce nonspecific T-cell activation and systemic cytokine release. 30 , 93 , 94 For this reason, nonagonistic and Fc-free targeting formats, including scFvs, Fab fragments, or nanobodies, should be favored over intact antibodies to reduce Fc receptor-mediated uptake by myeloid cells and unintended immune activation. These ligands can be displayed at controlled densities on distal PEG–lipid spacers to preserve receptor accessibility while minimizing multivalent receptor clustering; dual-ligand designs, such as CD3/CD7 targeting, may further increase cellular avidity but must be evaluated for effects on T-cell activation and delivery to non-T lymphocyte populations. 76 , 77 , 95 Before clinical translation, candidate formulations should undergo cell-resolved biodistribution and dose-escalation studies in humanized mouse models and nonhuman primates. The evaluation should quantify functional CAR expression in CD4-positive and CD8-positive T cells as well as unintended expression in B cells, natural killer cells, monocytes, hepatocytes, and bone marrow progenitor cells. Safety testing should include serum alanine aminotransferase, aspartate aminotransferase, bilirubin, inflammatory cytokines, type I interferon responses, complement activation products, hematological parameters, and liver and spleen histopathology. Empty LNPs, noncoding mRNA-loaded LNPs, and CAR mRNA-loaded LNPs should be assessed in parallel to distinguish material-related hepatotoxicity and innate reactogenicity from CAR-dependent immune toxicity. Repeat-dose studies should additionally evaluate accelerated clearance, anti-PEG or anti-ligand antibodies, changes in T-cell-subset distribution, and whether repeated administration produces cumulative inflammatory or hepatic injury. 78 , 96 Rurik et al first provided compelling proof of concept evidence by using CD5 targeted LNPs to deliver modified mRNA encoding FAP-CAR, thereby generating transient antifibrotic CAR-T cells in mice. CAR expression was detected within approximately 48 hours and then declined, ultimately reducing myocardial fibrosis and improving cardiac function. This indicates that mRNA LNPs can achieve transient T-cell programming in vivo without ex vivo cell manufacturing. 30

This transient expression profile has dual implications for MM CAR-T therapy. On one hand, it may reduce the risks of persistent plasma cell depletion, delayed toxicity, and uncontrolled expansion caused by long-term CAR expression. On the other hand, BCMA-positive myeloma typically requires strong and sustained tumor clearance pressure. Therefore, mRNA LNP platforms must further optimize T-cell targeting efficiency, the CAR expression window, and repeat administration strategies. 76 , 77 , 92 Billingsley et al reported that antibody conjugated, extrahepatic oriented ionizable LNPs can deliver CAR mRNA and generate functional CAR-T cells in vivo while inducing dose-dependent B-cell clearance. This indicates that, by regulating lipid composition, antibody conjugation, and in vivo distribution, LNPs can be advanced from liver directed delivery tools to immune cell engineering tools. 76 Furthermore, Tilsed et al further found that IL-7 can enhance protein translation capacity in T cells and increase mRNA expression mediated by CD5 targeted LNPs. These findings indicate that in vivo CAR-T generation is not only a carrier problem, but also involves pharmacologic regulation of T-cell metabolic and translational states, which are closely related to T-cell exhaustion and insufficient starting material in patients with MM. 92 In addition, the targeted LNP study by Hunter et al further advanced this field toward translation. Their tLNPs delivered mRNA to specific T-cell subpopulations, reprogrammed CD8 positive T cells in samples from healthy donors and patients with autoimmune disease, controlled tumors in humanized mice after in vivo administration, and induced B-cell clearance in cynomolgus monkeys. These findings suggest that LNP mediated CAR-T generation has progressed from mouse proof of concept studies to safety and efficacy evaluation in nonhuman primates. 77 In addition to transient mRNA delivery, the NCtx targeted LNP developed by Bimbo et al coencapsulates minicircle DNA encoding the CAR component and SB100x transposase mRNA, achieving T-cell specific delivery through anti CD7 and anti CD3 ligands. A single intravenous administration produced stable CAR-T cells, controlled tumors, and prolonged survival in leukemia humanized mice. These findings indicate that nonviral nanosystems may also achieve more persistent in vivo CAR expression, although integration sites, genotoxicity, and long-term follow-up requirements must be carefully evaluated. 78

In parallel with nonviral nanoparticles, nanoscale viral vectors represent another pathway for in vivo CAR-T generation. The VivoVec platform developed by Michels et al transduces T cells in vivo through surface engineered lentiviral particles and generates CAR-T cells with antitumor activity in preclinical models. This demonstrates that stably integrated in vivo CAR-T cells may have strong expansion and sustained cytotoxic potential ( Figure 4 ). 80 Nicolai et al further used a lentiviral vector displaying multidomain fusion ligands to generate CAR-T cells in nonhuman primates. This system integrates CD3 targeting with CD80 and CD58 costimulatory signals, enhances T-cell binding, activation, and transduction, and induces deep B-cell clearance. These results provide key large animal evidence supporting the clinical application of viral vector based in vivo CAR-T generation. 97
Figure 4 Surface-engineered lentiviral VivoVec platform for in vivo CAR-T cell generation and pharmacologically controlled expansion. ( A ) Schematic representation of the VivoVec lentiviral particle designed for selective T-cell transduction in vivo. The vector incorporates Cocal envelope glycoprotein and surface-displayed anti-CD3 scFv to promote T-cell targeting, while the lentiviral nucleocapsid carries the therapeutic transgene cassette. ( B ) Structure of the integrated transgene cassette encoding an anti-CD19 CAR and rapamycin-regulated signaling components. The CD19 CAR contains an FMC63-derived antigen-binding domain, IgG4 hinge, transmembrane region, 4–1BB costimulatory domain, and CD3ζ signaling domain. P2A sequences enable co-expression of additional modules, including FRB and rapamycin-activated cytokine receptor subunits RACRγ and RACRβ. ( C ) Functional mechanism of VivoVec-generated CAR-T cells. The CD19 CAR mediates tumor recognition, whereas the FRB module sequesters rapamycin/FKBP12 complexes and limits canonical mTOR inhibition. In parallel, rapamycin promotes assembly of the engineered RACRγ/RACRβ receptor complex, activating cytokine-like survival and proliferative signaling. Together, this platform enables stable in vivo CAR-T cell generation with antigen-directed cytotoxicity and drug-controllable expansion potential. Adapted from Michels KR, Sheih A, Hernandez SA, et al Preclinical proof of concept for VivoVec, a lentiviral-based platform for in vivo CAR-T cell engineering. J Immunother Cancer . Copyright © 2023 by authors. 80 Image A depicts a stylized lentiviral particle with surface spikes and an internal strand, labeled with icons: Cocal glycoprotein, anti-CD3 scFv, Nucleocapsid, Transgene. Image B shows a block map of an integrated transgene cassette with a left arrow labeled MND. A bracket labeled CD19 CAR spans blocks: FMC63 binder, IgG4 hinge, TM domain, 4-1BB, CD3z, separated by P2A connectors. FRB, RACRg and RACRb brackets sit above blocks labeled FRB, FKBP, IL-2Rg and IL-2Rb. Image C illustrates a membrane line with receptor cartoons and arrows. Labels indicate CD19 and CD19 CAR targeting tumors, with a receptor labeled CD19 CAR. Below are Rapamycin, Native FKBP12, FRB and mTOR labels. A sentence states FRB scavenges rapamycin/FKBP12 complexes, preventing mTOR association, marked by a red X. Right text explains Rapamycin engages RACR, with RACRg and RACRb parts leading to ′Survival and proliferation.′. Three-part schematic of VivoVec: lentiviral particle, CD19 CAR cassette map, rapamycin-controlled RACR signaling.

Surface-engineered lentiviral VivoVec platform for in vivo CAR-T cell generation and pharmacologically controlled expansion. ( A ) Schematic representation of the VivoVec lentiviral particle designed for selective T-cell transduction in vivo. The vector incorporates Cocal envelope glycoprotein and surface-displayed anti-CD3 scFv to promote T-cell targeting, while the lentiviral nucleocapsid carries the therapeutic transgene cassette. ( B ) Structure of the integrated transgene cassette encoding an anti-CD19 CAR and rapamycin-regulated signaling components. The CD19 CAR contains an FMC63-derived antigen-binding domain, IgG4 hinge, transmembrane region, 4–1BB costimulatory domain, and CD3ζ signaling domain. P2A sequences enable co-expression of additional modules, including FRB and rapamycin-activated cytokine receptor subunits RACRγ and RACRβ. ( C ) Functional mechanism of VivoVec-generated CAR-T cells. The CD19 CAR mediates tumor recognition, whereas the FRB module sequesters rapamycin/FKBP12 complexes and limits canonical mTOR inhibition. In parallel, rapamycin promotes assembly of the engineered RACRγ/RACRβ receptor complex, activating cytokine-like survival and proliferative signaling. Together, this platform enables stable in vivo CAR-T cell generation with antigen-directed cytotoxicity and drug-controllable expansion potential. Adapted from Michels KR, Sheih A, Hernandez SA, et al Preclinical proof of concept for VivoVec, a lentiviral-based platform for in vivo CAR-T cell engineering. J Immunother Cancer . Copyright © 2023 by authors. 80

For MM, a true milestone is the entry of BCMA-targeted in vivo CAR-T therapy into patient studies. Xu et al reported in Lancet an early case series of in vivo BCMA CAR-T treatment for relapsed or refractory MM, suggesting that a single in vivo administration can generate anti-BCMA CAR-T cells and induce antimyeloma responses in patients. However, immunotoxicities such as CRS were also observed, indicating that eliminating ex vivo manufacturing does not eliminate CAR-T related toxicity. 81 Subsequently, An et al reported the phase 1 study of ESO-T01 in Nature Medicine. ESO-T01 is a nanobody-directed and immune-shielded lentiviral vector encoding a humanized anti-BCMA CAR. It can generate anti-BCMA CAR-T cells in patients through a single intravenous infusion without leukapheresis or lymphodepletion. No dose-limiting toxicities were observed in the five heavily pretreated male patients with RRMM. However, all patients experienced grade 3 or higher adverse events. Initial efficacy included objective responses in four of five patients and stringent complete remission in three patients. These results demonstrate the feasibility of in vivo CAR-T generation in MM while also indicating that this technology remains in a high risk early clinical stage. 89 Mechanistically, ESO-T01 confers T-cell targeting through anti TCR nanobodies, reduces broad tropism through mutant VSVG, decreases mononuclear phagocyte system clearance through CD47 overexpression, and restricts expression using T-cell-specific promoters. In this way, delivery, immune evasion, cell selectivity, and CAR expression control are integrated into a single injectable carrier. This represents a core feature of the integration of nanomedicine and cell therapy. 89

However, the clinical translation of in vivo-generated CAR-T cells still requires solutions to four key issues. First, CAR genes must be prevented from entering non-T cells or tumor cells and causing aberrant expression. Because ligand-mediated cellular uptake alone cannot guarantee absolute specificity, strict T-cell restriction should be established through multiple sequential safeguards. These include the use of nonagonistic and Fc-free targeting ligands against T-cell-associated receptors such as CD3, CD5, CD7, or CD8, reduction of broad carrier tropism, and incorporation of T-cell-restricted promoters or enhancer elements to limit transgene expression when a vector is inadvertently internalized by a non-T cell. Second, the safety advantage of transient mRNA platforms must be balanced with the persistence required for MM clearance. Nonintegrating mRNA or circRNA delivery is preferable when transient CAR expression is sufficient, because these cargos remain predominantly cytoplasmic and avoid permanent genomic insertion. In contrast, DNA-, transposon-, and lentiviral-vector-based systems require additional safeguards, including transient transposase expression, self-inactivating vector design, testing for replication-competent vectors, integration-site analysis, and long-term monitoring for abnormal clonal expansion. Third, host immune clearance and bone marrow immunosuppression must be overcome without lymphodepletion. Fourth, systems are needed for pharmacokinetic and efficacy monitoring, controllable stopping switches, and long-term insertional mutagenesis surveillance. 89 , 98 In the future, faster treatment initiation, fewer manufacturing failures, and more controllable relapse interventions may be achieved through multitarget mRNA or circRNA nanodelivery against BCMA, GPRC5D, or FcRH5, enhanced bone marrow homing, co delivery of IL-7 or metabolic regulators, repeatable administration, and safety switch design. 81 , 89 , 98

Nanomedicine Enhances the Efficacy of MM CAR-T Therapy

Improving the efficacy of MM CAR-T therapy requires more than increasing CAR-T cell dose or cytotoxic potency. Relapse after BCMA CAR-T therapy often emerges from a layered biological network, including insufficient antigen visibility, progressive T-cell dysfunction, and protective bone marrow niches that shield residual myeloma cells from immune attack. Nanomedicine provides a way to intervene across these layers rather than targeting a single resistance mechanism. At the antigen level, nanoenabled delivery and antigen-modulating strategies may increase target density and reduce antigen-low escape. At the T-cell level, nanoscale cytokine delivery, RNA therapeutics, and immune-regulatory platforms may enhance CAR-T expansion, persistence, and exhaustion resistance. At the bone marrow microenvironment level, targeted nanoparticles can remodel stromal, endothelial, myeloid, and extracellular-vesicle-mediated barriers that sustain residual disease. Therefore, Table 4 discusses how nanomedicine may enhance MM CAR-T efficacy by synchronously improving tumor recognition, CAR-T fitness, and the bone marrow niche.
Table 4 Multilevel Nanomedicine Strategies to Enhance MM CAR-T Efficacy Level Strategy/Platform Biological Target Key Evidence How it May Enhance MM CAR-T Efficacy Translational Notes Ref Antigen level γ-secretase inhibition to increase BCMA surface density BCMA shedding, sBCMA, BCMA-low MM cells The phase 1 trial combining crenigacestat with BCMA CAR-T showed that γ-secretase inhibition can be clinically combined with BCMA CAR-T in RRMM. The rationale is to prevent BCMA cleavage, increase surface BCMA, and reduce soluble BCMA interference. Converts weakly visible BCMA-low cells into more recognizable CAR-T targets; may delay antigen-low escape and improve tumor-cell engagement. This is not yet a nanodrug itself, but GSI is a strong candidate payload for targeted nano-delivery to reduce systemic Notch-related toxicity. [ 51 ] Antigen level Single-cell-guided BCMA modulation after GSI and CAR-T BCMA expression and bone marrow immune response Single-cell analysis showed that γ-secretase inhibition plus CAR-T does not only affect BCMA shedding, but also changes transcriptional and epigenetic states in the MM microenvironment. Supports a precision strategy in which antigen densification is monitored together with immune-cell remodeling, rather than treated as a purely tumor-cell event. Provides mechanistic rationale for combining antigen-modulating drugs with nano-delivery and immune monitoring. [ 53 ] Antigen level Nanobody-based alternative antigen CAR-T B7-H3, BCMA antigen escape B7-H3 nanobody-based CAR-T controlled MM growth in vitro and in vivo, while dual BCMA/B7-H3 CAR-T or pooled CAR-T strategies overcame BCMA antigen loss in preclinical models. Expands target coverage beyond BCMA and provides a strategy for BCMA-negative or BCMA-low relapse after CAR-T. B7-H3 expression on non-myeloma tissues and myeloid compartments requires careful safety evaluation. [ 99 ] Antigen level Multiantigen CAR-T with IL-18 secretion Antigen-low MM; BCMA/BAFF-R-related targeting logic IL-18-secreting multiantigen CAR-T cells eliminated antigen-low myeloma in an immunocompetent mouse model and enhanced antitumor activity through stronger T-cell activation and microenvironment remodeling. Addresses antigen heterogeneity by increasing target avidity and coupling antigen recognition with local immune activation. Promising for antigen-low relapse, but cytokine-related toxicity and dosing control need further evaluation. [ 100 ] T-cell level CAR enhancer protein, CAR-E CAR-selective cytokine signaling; BCMA–IL-2 fusion A BCMA–low-affinity IL-2 fusion protein selectively stimulated CAR-T cells through antigen–CAR binding, increasing CAR-T activation, proliferation, memory formation, and tumor clearance while reducing nonspecific IL-2 toxicity. Functions as an “on-demand amplifier” for CAR-T cells, potentially improving expansion and persistence without re-engineering the original CAR-T product. Could be useful after weak expansion or early functional decline, but cytokine exposure must be carefully controlled. [ 101 ] T-cell level CAR-E optimization to preserve stemness and resist exhaustion CAR-T fate, memory, exhaustion resistance Follow-up work showed that IL-2Rα-sparing CAR-E maintained potency and drove CAR-T expansion, persistence, tumor clearance, and exhaustion resistance across models, including CAR-T cells from previously treated MM patients. May rescue suboptimal CAR-T products from heavily treated MM patients and support lower-dose or repeatable CAR-T strategies. Still requires prospective validation in MM CAR-T clinical settings. [ 102 ] T-cell level Microenvironment-informed anti-exhaustion strategies Exhausted CD4+ CAR-T cells, TIGIT/HAVCR2, TGF-β-related suppression Early relapse after BCMA CAR-T has been associated with CD4+ CAR-T exhaustion, reduced persistence, and suppressive interactions with nonclassical monocytes. Provides a rationale for nano-delivered checkpoint modulators, cytokines, metabolic regulators, or RNA therapeutics that protect CAR-T fitness after infusion. More of a mechanistic foundation than a mature nanomedicine intervention; needs targeted delivery to avoid systemic immune toxicity. [ 41 ] Bone marrow microenvironment level Bone marrow immune atlas after BCMA CAR-T BAFF+PD-L1+ myeloid cells, exhausted T-cell clones, marrow immune composition High-dimensional profiling showed that remission durability after BCMA CAR-T correlates with bone marrow tumor and immune-cell states; shorter PFS was linked to exhausted T-cell clones and suppressive myeloid features. Indicates that CAR-T efficacy depends on the marrow ecosystem, not only CAR-T dose or antigen expression. Supports combining CAR-T with niche-remodeling nanomedicine rather than simply escalating CAR-T potency. [ 66 ] Bone marrow microenvironment level CD38-targeted lipid nanoparticles delivering therapeutic RNA CD38-positive MM cells in bone marrow CD38-targeted LNPs delivered siRNA to MM cells in bone marrow; CKAP5 siRNA markedly reduced MM-cell viability and showed therapeutic effects in MM models. Could reduce tumor burden before CAR-T, provide low-toxicity bridging, or deliver RNA payloads that sensitize MM cells to CAR-T killing. Strong preclinical MM relevance; combination with CAR-T still needs direct testing. [ 103 ] Bone marrow microenvironment level Lipid–polymer nanoparticles targeting marrow endothelium Cyclophilin A, E-selectin, MM extravasation and niche colonization Bone marrow endothelial siRNA delivery against CyPA or E-selectin/CyPA reduced MM extravasation, decreased tumor burden, extended survival, and improved combination therapy with bortezomib in mouse models. Remodels the vascular niche that supports MM homing and relapse, potentially making residual disease less protected after CAR-T. Particularly relevant for relapse prevention and post-CAR-T niche control; not yet clinically validated with CAR-T. [ 67 ] Bone marrow microenvironment level Antibody-mediated nanodrug delivery of carfilzomib CD38-positive MM cells; proteasome inhibition Daratumumab-decorated polypeptide micelles delivered carfilzomib to CD38-positive MM cells, achieved stronger MM-cell inhibition than nontargeted controls, prolonged survival in an orthotopic MM model, and reduced osteolytic lesions. May serve as tumor-debulking or bridging nanotherapy before CAR-T while reducing systemic proteasome-inhibitor toxicity. Direct CAR-T combination data are still lacking, but the platform fits CAR-T bridging and relapse-intervention logic. [ 104 ] Bone marrow microenvironment level Phagocytic niche remodeling by BCMA/CD24 targeting CD24–Siglec-10 “do-not-eat-me” axis; macrophage polarization BCMA/CD24 bispecific CAR-T blocked the CD24–Siglec-10 axis, enhanced macrophage-mediated phagocytosis, promoted M1-like polarization, and improved MM control compared with monospecific BCMA CAR-T. Nanotechnology has the potential to permanently maintain anti-tumor M1 status in vivo. Shows that CAR-T efficacy can be enhanced by converting macrophages from passive niche components into active clearance partners. This provides a clear target axis for future macrophage guided nanoparticle or CAR-M/nanocell strategies. [ 105 ]

Table 4 Multilevel Nanomedicine Strategies to Enhance MM CAR-T Efficacy Level Strategy/Platform Biological Target Key Evidence How it May Enhance MM CAR-T Efficacy Translational Notes Ref Antigen level γ-secretase inhibition to increase BCMA surface density BCMA shedding, sBCMA, BCMA-low MM cells The phase 1 trial combining crenigacestat with BCMA CAR-T showed that γ-secretase inhibition can be clinically combined with BCMA CAR-T in RRMM. The rationale is to prevent BCMA cleavage, increase surface BCMA, and reduce soluble BCMA interference. Converts weakly visible BCMA-low cells into more recognizable CAR-T targets; may delay antigen-low escape and improve tumor-cell engagement. This is not yet a nanodrug itself, but GSI is a strong candidate payload for targeted nano-delivery to reduce systemic Notch-related toxicity. [ 51 ] Antigen level Single-cell-guided BCMA modulation after GSI and CAR-T BCMA expression and bone marrow immune response Single-cell analysis showed that γ-secretase inhibition plus CAR-T does not only affect BCMA shedding, but also changes transcriptional and epigenetic states in the MM microenvironment. Supports a precision strategy in which antigen densification is monitored together with immune-cell remodeling, rather than treated as a purely tumor-cell event. Provides mechanistic rationale for combining antigen-modulating drugs with nano-delivery and immune monitoring. [ 53 ] Antigen level Nanobody-based alternative antigen CAR-T B7-H3, BCMA antigen escape B7-H3 nanobody-based CAR-T controlled MM growth in vitro and in vivo, while dual BCMA/B7-H3 CAR-T or pooled CAR-T strategies overcame BCMA antigen loss in preclinical models. Expands target coverage beyond BCMA and provides a strategy for BCMA-negative or BCMA-low relapse after CAR-T. B7-H3 expression on non-myeloma tissues and myeloid compartments requires careful safety evaluation. [ 99 ] Antigen level Multiantigen CAR-T with IL-18 secretion Antigen-low MM; BCMA/BAFF-R-related targeting logic IL-18-secreting multiantigen CAR-T cells eliminated antigen-low myeloma in an immunocompetent mouse model and enhanced antitumor activity through stronger T-cell activation and microenvironment remodeling. Addresses antigen heterogeneity by increasing target avidity and coupling antigen recognition with local immune activation. Promising for antigen-low relapse, but cytokine-related toxicity and dosing control need further evaluation. [ 100 ] T-cell level CAR enhancer protein, CAR-E CAR-selective cytokine signaling; BCMA–IL-2 fusion A BCMA–low-affinity IL-2 fusion protein selectively stimulated CAR-T cells through antigen–CAR binding, increasing CAR-T activation, proliferation, memory formation, and tumor clearance while reducing nonspecific IL-2 toxicity. Functions as an “on-demand amplifier” for CAR-T cells, potentially improving expansion and persistence without re-engineering the original CAR-T product. Could be useful after weak expansion or early functional decline, but cytokine exposure must be carefully controlled. [ 101 ] T-cell level CAR-E optimization to preserve stemness and resist exhaustion CAR-T fate, memory, exhaustion resistance Follow-up work showed that IL-2Rα-sparing CAR-E maintained potency and drove CAR-T expansion, persistence, tumor clearance, and exhaustion resistance across models, including CAR-T cells from previously treated MM patients. May rescue suboptimal CAR-T products from heavily treated MM patients and support lower-dose or repeatable CAR-T strategies. Still requires prospective validation in MM CAR-T clinical settings. [ 102 ] T-cell level Microenvironment-informed anti-exhaustion strategies Exhausted CD4+ CAR-T cells, TIGIT/HAVCR2, TGF-β-related suppression Early relapse after BCMA CAR-T has been associated with CD4+ CAR-T exhaustion, reduced persistence, and suppressive interactions with nonclassical monocytes. Provides a rationale for nano-delivered checkpoint modulators, cytokines, metabolic regulators, or RNA therapeutics that protect CAR-T fitness after infusion. More of a mechanistic foundation than a mature nanomedicine intervention; needs targeted delivery to avoid systemic immune toxicity. [ 41 ] Bone marrow microenvironment level Bone marrow immune atlas after BCMA CAR-T BAFF+PD-L1+ myeloid cells, exhausted T-cell clones, marrow immune composition High-dimensional profiling showed that remission durability after BCMA CAR-T correlates with bone marrow tumor and immune-cell states; shorter PFS was linked to exhausted T-cell clones and suppressive myeloid features. Indicates that CAR-T efficacy depends on the marrow ecosystem, not only CAR-T dose or antigen expression. Supports combining CAR-T with niche-remodeling nanomedicine rather than simply escalating CAR-T potency. [ 66 ] Bone marrow microenvironment level CD38-targeted lipid nanoparticles delivering therapeutic RNA CD38-positive MM cells in bone marrow CD38-targeted LNPs delivered siRNA to MM cells in bone marrow; CKAP5 siRNA markedly reduced MM-cell viability and showed therapeutic effects in MM models. Could reduce tumor burden before CAR-T, provide low-toxicity bridging, or deliver RNA payloads that sensitize MM cells to CAR-T killing. Strong preclinical MM relevance; combination with CAR-T still needs direct testing. [ 103 ] Bone marrow microenvironment level Lipid–polymer nanoparticles targeting marrow endothelium Cyclophilin A, E-selectin, MM extravasation and niche colonization Bone marrow endothelial siRNA delivery against CyPA or E-selectin/CyPA reduced MM extravasation, decreased tumor burden, extended survival, and improved combination therapy with bortezomib in mouse models. Remodels the vascular niche that supports MM homing and relapse, potentially making residual disease less protected after CAR-T. Particularly relevant for relapse prevention and post-CAR-T niche control; not yet clinically validated with CAR-T. [ 67 ] Bone marrow microenvironment level Antibody-mediated nanodrug delivery of carfilzomib CD38-positive MM cells; proteasome inhibition Daratumumab-decorated polypeptide micelles delivered carfilzomib to CD38-positive MM cells, achieved stronger MM-cell inhibition than nontargeted controls, prolonged survival in an orthotopic MM model, and reduced osteolytic lesions. May serve as tumor-debulking or bridging nanotherapy before CAR-T while reducing systemic proteasome-inhibitor toxicity. Direct CAR-T combination data are still lacking, but the platform fits CAR-T bridging and relapse-intervention logic. [ 104 ] Bone marrow microenvironment level Phagocytic niche remodeling by BCMA/CD24 targeting CD24–Siglec-10 “do-not-eat-me” axis; macrophage polarization BCMA/CD24 bispecific CAR-T blocked the CD24–Siglec-10 axis, enhanced macrophage-mediated phagocytosis, promoted M1-like polarization, and improved MM control compared with monospecific BCMA CAR-T. Nanotechnology has the potential to permanently maintain anti-tumor M1 status in vivo. Shows that CAR-T efficacy can be enhanced by converting macrophages from passive niche components into active clearance partners. This provides a clear target axis for future macrophage guided nanoparticle or CAR-M/nanocell strategies. [ 105 ]

Multilevel Nanomedicine Strategies to Enhance MM CAR-T Efficacy

Antigen Level: Increasing Target Density and Preventing Antigen Escape

In CAR-T therapy, antigen expression is not static but represents a dynamic interface continuously reshaped by shedding, degradation, epigenetic regulation, and treatment pressure. Therefore, in MM CAR-T therapy, increasing the surface density of targets such as BCMA and establishing multitarget redundancy represent direct and translatable entry points for nanomedicine to enhance efficacy and delay relapse. 33 , 51 BCMA CAR-T therapy has been shown to induce deep remission in studies of ide-cel and cilta-cel. However, relapse samples and longitudinal analyses suggest that decreased BCMA expression, increased antigen heterogeneity, or expansion of rare BCMA-negative clones can weaken CAR-T recognition and sustained cytotoxicity. Therefore, insufficient antigen density should be considered an early biomarker of efficacy attenuation rather than a terminal event detected only after relapse. 33 , 51 Rana et al conducted dynamic analyses of BCMA in patients with RRMM receiving BCMA directed CAR-T therapy and found that complete BCMA loss during relapse was not always the dominant pattern. Instead, more patients showed decreased or heterogeneous BCMA expression, indicating that a biological window may remain for restoring target density through pharmacologic or delivery based strategies. 33 Gamma secretase inhibitors are currently among the most mature antigen densification strategies. Gamma secretase can cleave membrane bound BCMA and generate soluble BCMA, which reduces tumor cell surface targets and may act as antigen bait that interferes with CAR-T recognition. 51 , 52 In a first-in-human phase 1 study, Cowan et al combined crenigacestat with BCMA CAR-T therapy in patients with RRMM and showed that GSI could increase BCMA target density on tumor cells. Deep remission was observed in heavily pretreated patients, regardless of prior BCMA-directed therapy, suggesting that increasing target density before CAR-T infusion is clinically feasible. 51 Moreover, Chen et al further demonstrated in a BCMA-targeted bispecific antibody model that GSI can enhance BCMA directed immunotherapy without substantially impairing T-cell activation or differentiation. These findings indicate that antigen density regulation may not only support CAR-T therapy but also serve as a broadly applicable enhancement module across the BCMA-targeted treatment sequence. 52

ATRA provides another antigen enhancement strategy through transcriptional regulation. García-Guerrero et al found that ATRA can increase BCMA transcript and membrane protein expression in MM cells, enhance BCMA CAR-T mediated ex vivo lysis, cytokine release, and proliferation, and improve CAR-T antimyeloma activity in the NSG MM.1S mouse model. More importantly, the combination of ATRA and crenigacestat can synergistically increase BCMA expression at both the gene transcription and membrane protein cleavage levels and further enhance the ex vivo and in vivo efficacy of BCMA CAR-T therapy. These findings provide pharmacologic support for nanodelivery systems co loaded with ATRA and GSI to achieve local antigen densification in the bone marrow. 105 Regulation of BCMA protein homeostasis is also an emerging direction worthy of attention. Rieger et al found that BCMA is a short-lived membrane protein whose surface level is controlled by the ubiquitin proteasome system and p97 dependent degradation. Carfilzomib can block BCMA degradation, increase membrane BCMA, and restore BCMA expression and antitumor activity in some patients receiving CAR-T therapy. 106 Li et al further demonstrated that low dose bortezomib can upregulate BCMA expression in MM1.S cells and enhance the ex vivo cytotoxicity and in vivo anti-MM effects of BCMA CAR-T therapy. These findings suggest that proteasome inhibitors are not only conventional antimyeloma drugs but may also serve as pharmacologic switches for antigen reexposure before and after CAR-T therapy. 107 Some studies have shown that bortezomib pretreatment can upregulate BCMA and enhance BCMA CAR-T cytotoxicity and antimyeloma effects in mice.

However, simply increasing BCMA density cannot fully overcome antigen escape. Samur et al and Da Via et al reported BCMA-negative relapse caused by loss or homozygous deletion of TNFRSF17 or BCMA alleles after BCMA CAR-T therapy, indicating a transition from a low antigen state to a target absent state. 34 , 35 Lee et al confirmed through whole-genome sequencing and single cell copy number analysis that target antigen gene deletion, structural variation, or extracellular domain mutation can occur after BCMA or GPRC5D targeted immunotherapy. Low frequency subclones present before treatment can expand into relapse dominant clones under treatment pressure. 36 Meanwhile, Firestone et al also demonstrated that loss of BCMA expression can become a shared resistance mechanism among BCMA antibody drug conjugates, CAR-T therapy, and bispecific antibodies, thereby limiting responses to subsequent BCMA-targeted therapies. Therefore, antigen level interventions should expand from enhancing BCMA alone to monitoring BCMA, incorporating backup non-BCMA targets, and designing multitarget strategies in advance. 38 Dual target CAR-T therapy is a direct engineering strategy for preventing antigen escape. The BCMA and GPRC5D bispecific CAR-T therapy reported by Zhou et al achieved an overall response rate of 86% and a complete response or better rate of 62% in a phase 1 study of RRMM. The researchers also suggested that this strategy may have value for patients with BCMA or GPRC5D negative or low expression disease. 108 Then, Yao et al further conducted a phase 1 study of BCMA and GPRC5D bispecific CAR-T therapy in patients with RRMM and extramedullary lesions. This study showed that the strategy is clinically feasible in high risk extramedullary disease, while also highlighting the need for larger cohorts and longer follow-up to determine whether it reduces antigen escape mediated relapse. 109 BCMA and CD19 dual target CAR-T therapy represents another antigen redundancy approach. Qiang et al reported that GC012F showed a high MRD negative rate and manageable safety in patients with high risk newly diagnosed MM. These findings suggest that early use of multitarget CAR-T therapy may suppress residual disease before clonal complexity increases. 110

From the perspective of nanomedicine, antigen level enhancement should not be limited to systemic administration of GSI or ATRA. Instead, bone marrow homing or plasma cell targeted nanosystems should be developed to deliver GSI, ATRA, proteasome modulators, siRNA, or mRNA regulatory modules to myeloma cells and their ecological niches. These approaches may enhance local antigen densification and reduce systemic toxicities related to Notch inhibition, retinoic acid signaling, or proteasome inhibition. 67 , 103 , 107 The CD38 targeted lipid nanoparticles developed by Tarab Ravski et al can deliver siRNA to MM cells in the bone marrow and improve therapeutic outcomes in mice. 103 The lipid polymer nanoparticles developed by Guimarães et al can deliver siRNA to the bone marrow microenvironment in vivo. Although these studies did not directly deliver BCMA densification agents, they demonstrated the feasibility of bone marrow or plasma cell directed nucleic acid and drug delivery in MM. 67 Some research shows that CD38 targeting LNPs can deliver therapeutic RNA to bone marrow MM cells, while lipid polymer nanoparticles can overcome the bone marrow delivery barrier and deliver siRNA therapy to MM lesions. Chen et al developed daratumumab modified carfilzomib polymeric micelles that can target CD38-positive MM cells and enhance the anti-MM effect of carfilzomib. This provides a practical materials science example of antibody guided nanomedicine for delivering proteasome inhibitors near myeloma cells and indirectly stabilizing BCMA. 104

The antigen level strategy for enhancing MM CAR-T efficacy through nanomedicine should be upgraded from increasing BCMA density alone to an integrated system combining pharmacologic antigen density enhancement, multitarget escape prevention, bone marrow–targeted delivery, and dynamic antigen monitoring. The translational pathway may begin with short-term GSI, ATRA, or proteasome inhibitor pretreatment before CAR-T therapy and gradually evolve into personalized relapse prevention guided by nanocarrier based antigen enhancers, multitarget CAR design, and combined monitoring using ctDNA, flow cytometry, and imaging.

T-Cell Level: Enhancing Expansion, Persistence, and Resistance to Exhaustion

In MM CAR-T therapy, the therapeutic effect is determined not by nanoparticles themselves, but by the T-cell state reshaped by nanotechnology. Therefore, T-cell level enhancement should shift from simply increasing CAR transduction efficiency to synchronously optimizing expansion kinetics, memory and stem cell like phenotypes, metabolic adaptability, and resistance to exhaustion. 39 , 41 For instance, Lin et al conducted an 18 month follow-up study of ide-cel CRB-401 and found that patients with PFS of at least 18 months had more naive T cells and fewer exhausted T cells in their leukapheresis starting material, and that the final product had a better functional phenotype. This suggests that nanodelivery may improve CAR-T persistence from the source by protecting early memory like T cells and reducing stress induced differentiation during manufacturing. 39 In addition, Ledergor et al further showed through single-cell analysis that patients with sustained remission after BCMA CAR-T therapy had greater CAR-T persistence and a higher proportion of CD8 effector memory CAR-T cells. In contrast, patients with early relapse showed an exhausted CD4 CAR-T population with high expression of HAVCR2, TIM-3, and TIGIT. These findings suggest that the composition and exhaustion trajectory of T-cell subsets can predict relapse risk. 41

Currently, several mainstream nanomedicine-based strategies exist to enhance the efficacy of CAR-T therapy for MM. One enhancement pathway is low damage engineering, which uses LNPs to deliver CAR mRNA, siRNA, or metabolic regulatory RNA and thereby reduce cell damage, inflammatory responses, and terminal differentiation caused by electroporation or viral transduction. 73 , 83 , 85 Kitte et al compared mRNA LNP delivery with electroporation and found that LNP engineered CAR-T cells had better CAR mRNA persistence and CAR surface expression, lower cytotoxicity, more stable proliferation, and fewer exhaustion features. These findings indicate that the delivery method itself can alter the fate of CAR-T cells, not only the efficiency of CAR expression. 73 Hamilton et al further developed an ionizable LNP platform that can simultaneously deliver CAR mRNA and PD-1 siRNA, enabling T cells to transiently inhibit checkpoint pathways while acquiring CAR expression. This co delivery strategy, which combines CAR programming with transient exhaustion control, may be suitable for reversible intervention of PD-1, TIGIT, and TIM-3 related exhaustion in MM CAR-T therapy. 85 The LNP platform is designed to achieve both therapeutic mRNA expression and RNAi mediated transient gene silencing in primary T cells. Records indicate that this platform aims to integrate CAR-T engineering and immune checkpoint inhibition into a single delivery step.

The second pathway is nanobiomimetic activation and expansion, which integrates conventional CD3 and CD28 magnetic bead activation, cytokine driven expansion, and nucleic acid delivery into a programmable nanointerface. This approach may reduce excessive differentiation while increasing expansion efficiency. 74 , 75 For example, Metzloff et al constructed antigen presenting cell mimicking LNPs that present T-cell activation signals on the particle surface while delivering CAR mRNA, thereby achieving one step T-cell activation and transfection. The resulting mRNA CAR-T cells reduced tumor burden in a mouse tumor model, suggesting that nanomaterials can replace selected ex vivo manufacturing units and shorten the manufacturing process. 74 The T-Expand glucan nanoparticles developed by Zheng et al trigger polyclonal expansion of human T cells by coupling anti CD3 and anti CD28 antibodies, with efficiency comparable to commercial Dynabeads. CD19 CAR-T cells manufactured with T-Expand showed stronger proliferation, cytotoxicity, and ex vivo persistence, and achieved complete tumor clearance at one quarter of the standard dose in a mouse model. This provides a practical example for future BCMA CAR-T manufacturing using a degradable, magnetic bead free nanoexpansion platform that preserves T-cell fitness. 75 The third pathway involves nanoscale regulation of cytokine and translational programs. Gamma chain cytokines such as IL-7 and IL-15 can affect CAR-T expansion, memory maintenance, and antiapoptotic capacity, but systemic administration is often limited by toxicity and nonspecific immune activation. 92 , 111 In a BCMA CAR-T model, Battram et al found that, compared with IL-2 or IL-15 plus IL-7, ARI2h BCMA CAR-T cells expanded with IL-15 showed less functional impairment, stronger persistence, reduced differentiation, and better anti-MM efficacy in mice. These findings indicate that the IL-15 axis is a directly translatable target for enhancing MM CAR-T persistence. 111 The study showed that BCMA CAR-T cells cultured with IL-15 exhibited superior in vivo efficacy and persistence in the MM mouse model, accompanied by fewer signs of senescence, DNA damage, and functional impairment. Then, Tilsed et al found that IL-7 can enhance protein translation capacity in T cells and increase targeted LNP mediated mRNA expression. This suggests that cytokine signaling may act as a translation sensitizer to improve the expression efficiency of mRNA CAR or function enhancing RNA during CAR-T nanoprogramming in vivo or ex vivo. 92

In addition, anti-exhaustion and microenvironment-resistance programming represents a key strategy for enhancing CAR-T cell function. In this strategy, PD-1 downregulation, TGF-β resistance, and local checkpoint blockade can be transformed into short range, controllable, and localized T-cell enhancement modules through nanodelivery. 44 , 64 , 85 Ouyang and Yu et al reported that PD-1 downregulation in BCMA CAR-T cells can preserve the memory phenotype, reduce exhaustion, and enhance anti-MM activity. This is consistent with the LNP strategy reported by Hamilton et al, which co delivers CAR mRNA and PD-1 siRNA. This suggests that LNPs may be used to transiently knock down PD-1 during manufacturing, potentially avoiding the autoimmune risk associated with long-term systemic PD-1 blockade. 64 , 85 The TGF-β resistant armed BCMA CAR-T cells generated by Alabanza et al maintained proliferation, degranulation, cytokine secretion, multifunctionality, and in vivo antitumor activity under high TGF-β expression conditions. These findings indicate that anti exhaustion engineering targeting bone marrow immunosuppressive signals can improve the functional stability of MM CAR-T cells. 44 This study demonstrated that B2ARM CAR-T cells retained Ki67, granzyme B, CD107a, and multicytokine secretion in the presence of TGF-β compared with conventional BCMA CAR-T cells, and achieved stronger tumor control in the RPMI-8226 model with high TGF-β expression.

Another promising strategy involves metabolic adaptive nanoreprogramming. In the MM bone marrow niche, tumor cells compete with immune cells for glutamine, glucose, and oxygen. If CAR-T cells cannot maintain mTORC1 signaling, mitochondrial respiration, and glycolysis in a nutrient limited environment, they may more rapidly enter a state of low proliferation and reduced cytotoxicity. 42 Navarro et al used MM as a glutamine dependent model and demonstrated that BCMA CAR-T cells are functionally restricted under low glutamine conditions. Overexpression of the glutamine transporter ASCT2 or SLC1A5 enhanced mTORC1 signaling, basal oxygen consumption, glycolytic function, in vivo persistence, and anti-MM efficacy. These findings provide a clear target for delivering metabolically enhancing mRNA or gene editing components using LNPs, polymer nanoparticles, or nonviral carriers. 42 In terms of clinical translation, the closest example of patient validation for nanomedicine enhanced T-cell efficacy is not an MM specific platform, but the BNT211 paradigm of mRNA vaccine mediated CAR-T expansion. Mackensen et al reported a phase 1 and 2 study of CLDN6 CAR-T combined with CARVac, demonstrating that RNA nanovaccines can expand CAR-T cells in patients and produce observable antitumor activity. This provides a valuable clinical pathway for future nanovaccine based re expansion of BCMA, GPRC5D, or FcRH5 CAR-T cells. 112

Therefore, T-cell level nanoenhancement strategies for MM CAR-T should be designed as integrated systems. Ex vivo, low damage LNPs or APC like nanoparticles can be used to protect the memory phenotype and complete co delivery of CAR, siRNA, and metabolic modules. In vivo, cytokine or mRNA nanosystems can be used to maintain CAR-T re expansion, while transient PD-1, TGF-β, or metabolic reprogramming can enhance resistance to exhaustion. From a safety perspective, these strategies must also address translational barriers, including CAR-T overexpansion, CRS, ICANS, long-term checkpoint inhibition, insertional mutagenesis, nanomaterial residues, and batch to batch consistency. A more rational clinical pathway may be to first evaluate these approaches in patients with MM who have a high risk of relapse or poor T-cell fitness, and then dynamically adjust the dose and administration window of nanoenhancers using CAR-T pharmacokinetics, single cell exhaustion profiling, ctDNA, MRD, and cytokine monitoring. 42 , 75 , 85 , 112

Bone Marrow Microenvironment Level: Transforming MM from an Immune Cold Niche to a CAR-T Friendly Niche

Failure of CAR-T therapy in multiple myeloma is often not caused by a lack of CAR-T cytotoxic potential, but by deployment into an immune cold bone marrow niche characterized by hypoxia, adhesion mediated protection, myeloid suppression, extracellular vesicle interference, and vascular barriers. Therefore, the key to enhancing MM CAR-T efficacy through nanomedicine is not only to strengthen CAR-T cells, but also to transform the bone marrow microenvironment into a supportive niche in which CAR-T cells can enter, expand, sustain cytotoxicity, and form immune memory. 40 , 45 , 47 , 66

For instance, Dhodapkar et al performed high dimensional analysis of bone marrow tumor and immune cells after BCMA CAR-T therapy and found that remission persistence was closely associated with the composition of non-CAR-T immune cells. These findings suggest that CAR-T efficacy is not solely an intrinsic attribute of the cellular product, but an ecological outcome shaped by interactions among CAR-T cells, endogenous T cells, myeloid cells, dendritic cells, and tumor cells. 66 This study reported that bone marrow tumor and immune cell changes are associated with BCMA CAR-T response persistence, providing clinical sample evidence for interpreting relapse from a microenvironmental perspective. In addition, Li et al further analyzed bone marrow samples from patients with relapse after BCMA CAR-T therapy using single-cell sequencing and found that T cells, NK cells, dendritic cells, neutrophils, monocytes, and macrophages exhibited protumor phenotypes. MIF and APRIL related cell communication was prominent in the relapse microenvironment, indicating that myeloid cells are not merely bystanders but may serve as core nodes of immunosuppression after CAR-T failure. 40 This study states that the bone marrow immune microenvironment in patients with relapse may explain the short response duration and high relapse rate after BCMA CAR-T therapy, and identifies monocytes, macrophages, MIF, and APRIL signaling as key participants. Yang et al found that, after BCMA CAR-T therapy combined with lymphodepletion, the abundance and target-specific cytotoxic activity of CD8 positive effector CAR-T cells on day 28 were associated with treatment response. A decrease in Treg abundance was also associated with efficacy. These findings suggest that resetting the immune niche before tumor clearance may be an important prerequisite for successful CAR-T expansion. 45

In clinical practice, patients with extramedullary lesions who receive CAR-T therapy often have poorer outcomes, supporting the relevance of spatial immune ecology. Extramedullary and extraosseous lesions often show poorer T-cell infiltration, stronger myeloid suppression, and more complex barrier structures, making it more difficult for CAR-T cells to establish persistent immune pressure. 47 , 48 Qi et al analyzed the clinical outcomes and microenvironment of patients with RRMM and extramedullary lesions after anti-BCMA CAR-T therapy and showed that CAR-T cells have activity against extramedullary lesions, although long-term survival benefits remain limited. This suggests that cold niche remodeling should be considered before and after CAR-T therapy in patients with high risk MM. 48

Bone marrow stroma and CAFs represent the first barriers that form the cold niche. Sakemura et al demonstrated that bone marrow cancer associated fibroblasts can promote MM progression and inhibit the antitumor activity of BCMA CAR-T cells. Dual target CAR-T cells directed against malignant plasma cells and FAP or SLAMF7 positive CAFs overcame CAF mediated inhibition and prolonged survival in mice. 43 From the perspective of nanomedicine, CAF remodeling does not necessarily require additional CAR structures. It can also be achieved through bone marrow–targeted nanoparticle delivery of siRNA, small molecule inhibitors, or immunomodulators related to TGF-β, CXCL12 and CXCR4, VCAM-1 and VLA-4, or other CAF activation pathways. These approaches may reduce matrix adhesion mediated protection and improve the spatial accessibility of CAR-T cells to tumor niches. 43 , 44 , 67 TGF-β is a representative soluble factor that links CAFs, myeloid suppression, and CAR-T exhaustion. Alabanza et al generated TGF-β resistant armed BCMA CAR-T cells that maintained proliferation, degranulation, cytokine release, and in vivo anti-MM activity in the presence of TGF-β. These findings indicate that microenvironmental inhibitory signals can be selectively counteracted through engineering or nanodelivery strategies. 44 The endothelial barrier and vascular niche represent the second layer of barriers. Guimarães et al used combinatorial chemistry and high throughput in vivo screening to develop lipid polymer nanoparticles that delivered siRNA to bone marrow endothelial cells and silenced cyclophilin A, thereby inhibiting transendothelial extravasation of MM cells and bone marrow colonization. This indicates that nanoparticles can transform bone marrow blood vessels from tumor homing channels into antitumor therapeutic targets. 67 Figueroa Espada et al further used siRNA lipid polymer nanoparticles to target E-selectin and CyPA and proposed that targeting the vascular microenvironment could represent a paradigm shift in MM therapy. This provides a translatable nanomedicine strategy for reducing tumor cell homing, adhesion, and vascular protection before CAR-T therapy. 113

Tumor cells themselves can also be displaced from their supportive niche through nanodelivery. Tarab Ravski et al developed anti CD38 modified LNPs to deliver siRNA to bone marrow MM cells. Targeting CKAP5 reduced MM cell viability by approximately 90% ex vivo and demonstrated bone marrow delivery potential in a mouse model. These findings indicate that plasma cell targeting LNPs can serve as a technical platform for tumor debulking before CAR-T therapy, reducing the immunosuppressive burden, and delivering microenvironment regulating RNA. 103

Extracellular vesicles represent the third type of nanoscale cold niche signal. Lopes et al demonstrated that MM derived extracellular vesicles can reshape the bone marrow immune microenvironment. Hagiwara et al found that MM derived extracellular vesicles can promote T-cell exhaustion by upregulating immune checkpoint markers. Soncini et al showed that bone marrow extracellular vesicles from patients with MM are enriched in HLA-G, PD-1, and PD-L1 and can promote adaptive immune dysfunction. 69 , 71 , 72 These studies suggest that nanomedicine should consider extracellular vesicles not only as delivery vehicles but also as immunosuppressive nanonetworks that may need to be captured, blocked, or reverse engineered. 69 , 71 , 72 Hypoxia is the fourth driver of the cold niche. In a hypoxic model, Tu et al found that hypoxia can reduce surface and total BCMA expression in MM cells, weaken BCMA CAR mediated signaling, and enhance the release of immunosuppressive factors and small extracellular vesicles. These effects simultaneously impair CAR-T recognition, activation, and effector function. 46 Therefore, hypoxia-responsive nanoparticles, ROS regulated nanoenzymes, local oxygen releasing materials, and HIF pathway inhibitor delivery systems may complement BCMA CAR-T therapy by restoring antigen accessibility and reducing hypoxia induced immunosuppressive secretion. 46 Coffey et al analyzed the single cell profile of gamma secretase inhibitors combined with BCMA CAR-T therapy and found that GSI not only affects BCMA cleavage but also triggers transcriptional and epigenetic changes in tumor cells and the immune microenvironment. These findings indicate that antigen enhancers may also have niche remodeling functions. 53 This finding suggests that clinically translatable nanocombination strategies can start from established human evidence for GSI combined with BCMA CAR-T therapy, lymphodepletion combined with CAR-T therapy, and antigen enhancement protocols. Additional modules, such as bone marrow–targeted LNPs, vascular barrier directed siRNA, extracellular vesicle interception, and hypoxia-responsive drug delivery, can then be added gradually rather than designing overly complex multidrug nanosystems at the outset. 53 , 103

From the perspective of translational pathways, the most practical first step is to use bone marrow–targeted nanomedicine for niche pretreatment within a short window before and after CAR-T infusion. Potential goals include reducing tumor burden, loosening matrix adhesion, inhibiting protumor myeloid signaling, reducing extracellular vesicle associated immune checkpoint load, and improving hypoxia related antigen downregulation. 67 , 103 , 113 The second step is niche reignition during peak CAR-T expansion or early MRD repositivity. Reusable LNPs or polymer nanoparticles could deliver IL-7 and IL-15 axis modulators, RNA inhibitors targeting TGF-β, PD-L1, and TIGIT pathways, or metabolic support modules, allowing residual CAR-T cells to regain expansion and cytotoxic capacity. 53 , 103

Overall, transforming MM from an immune cold niche to a CAR-T friendly niche is not simply a matter of increasing immune stimulation. Rather, it requires nanomedicine-based regulation of the bone marrow matrix, vascular endothelium, myeloid cells, extracellular vesicle networks, and hypoxic metabolism across space, time, and cell types. With such remodeling, CAR-T cells may no longer need to penetrate the bone marrow as isolated effectors but may instead achieve durable clearance within a reorganized therapeutic landscape.

Multimodal Nanomedicine and Precision Medicine Strategies in MM CAR-T Treatment

As MM CAR-T therapy enters the era of deep remission, its clinical frontier is shifting from achieving response to understanding how response is maintained, lost, and restored. Relapse is rarely a single event; it is often preceded by molecular drift, antigen remodeling, spatial immune exclusion, and exhaustion of available therapeutic routes. Multimodal nanomedicine offers a way to connect these dispersed signals into a precision medicine framework, integrating nanoenabled liquid biopsy, spatial multiomics, nanoimaging, rational treatment sequencing, and emerging nanocell platforms. This section therefore outlines how these technologies may transform MM CAR-T care from endpoint-based assessment into a dynamic strategy for early relapse warning, target selection, niche remodeling, and individualized post-CAR-T intervention.

Nanoenabled Liquid Biopsy for Molecular Relapse Monitoring

As MM CAR-T therapy enters the deep remission era, efficacy monitoring is shifting from a single bone marrow MRD negative assessment toward liquid biopsy approaches that are more continuous, less invasive, and better able to capture spatial heterogeneity. This transition is important because relapse often does not occur suddenly. Instead, it may first leave detectable molecular trajectories, including elevated sBCMA, altered extracellular vesicle cargo, reappearance of ctDNA, or expansion of antigen positive residual cells. 114–117

Soluble BCMA is the blood biomarker most closely aligned with the biological logic of BCMA CAR-T therapy. Membrane bound BCMA can be cleaved by gamma secretase to form sBCMA, which reflects tumor burden and antigen shedding status and may also act as antigen bait that interferes with the binding of BCMA-targeted agents to tumor cell surface BCMA. 114–117 Costa et al systematically summarized the value of sBCMA detection in MM, noting that sBCMA can be used to evaluate disease burden, treatment response, and prognosis. 114 While Meinl and Krumbholz further emphasized that sBCMA and sTACI are not only biomarkers but may also act as immunoregulatory factors and potential barriers to BCMA-targeted therapy. 115 As for the study by Shen et al on post CAR-T monitoring has more direct relevance to CAR-T response surveillance. This study showed that serum sBCMA can be used to monitor relapse in patients with MM after CAR-T immunotherapy, suggesting that dynamic changes in sBCMA may serve as an early warning signal beyond bone marrow aspiration. 116 Some studies indicate that, compared with conventional ELISA, nanodiagnostics can introduce antibody or adaptor functionalized gold nanoparticles, magnetic nanoparticle enrichment, surface plasmon resonance, SERS, or electrochemical nanointerfaces for sBCMA detection. These approaches may improve the sensitivity of low abundance protein detection and support multiplexed readouts of sBCMA, free light chains, inflammatory factors, and CAR-T copy number. 60 , 117 , 118 Shen et al used LC-MS to analyze the precise sequence of endogenous sBCMA and achieved unbiased quantification, indicating that sBCMA detection is moving from a single immunoassay readout toward precise measurement at the molecular structural level. If this approach is combined with nanoimmune enrichment or magnetic bead capture, it may reduce interference from therapeutic antibodies, APRIL and BAFF ligands, and other factors in detection results. 118 In addition, Lee et al found in a study of BCMA T-cell engagers that sBCMA and non T-cell factors can affect sensitivity to BCMA-targeted therapy. This suggests that post CAR-T monitoring should not only determine whether relapse has occurred but also evaluate whether circulating sBCMA is reshaping the therapeutic environment for subsequent BCMA-directed therapy. 60

Extracellular vesicles are the second class of biomarkers particularly suitable for nanodetection because they are natural nanoparticles of approximately 30 to 150 nm that carry proteins, miRNAs, mRNA, DNA fragments, lipids, and immune checkpoint molecules. They can capture ecological communication among myeloma cells, stromal cells, myeloid cells, and T cells in peripheral blood or bone marrow fluid samples. 68 , 119–121 For example, Menu and Vanderkerken noted in a review that MM exosomes are involved in tumor growth, angiogenesis, bone disease, drug tolerance, and immune escape and have dual potential as liquid biopsy markers and therapeutic carriers. 68 In the study of Laurenzana et al, they proposed that circulating extracellular vesicles can support the discovery of novel MM biomarkers and enable real time noninvasive monitoring through analysis of vesicle quantity, size, protein phenotype, and molecular content. 119 While Ferreira et al compared the proteomes of peripheral blood and bone marrow extracellular vesicles in real world studies and found high proteomic similarity between the two sources. They identified candidate extracellular vesicle proteins, including PDIA3, C4BPA, BTN1A1, and TNFSF13, that are associated with prognosis, immune dysfunction, or treatment response. 120 Bergantim et al further provided proof of concept evidence that MRD related biomarkers can be detected in liquid biopsy derived extracellular vesicles from patients with MM. This indicates that extracellular vesicles can serve as a complementary window for bone marrow MRD and may be particularly useful for capturing spatial residual disease affected by sampling bias in conventional bone marrow aspiration. 121 Technologically, nanodetection of extracellular vesicles can use immunomagnetic beads to capture CD9, CD63, CD81, BCMA, CD38, or SLAMF7 positive vesicles. These approaches can be combined with nanoflow cytometry, SERS, plasmonic chips, digital ELISA, electrochemical sensing, and microfluidic sorting to achieve multidimensional readouts of vesicle quantity, particle size, surface proteins, RNA, and protein content. 69 , 72 , 119 For CAR-T therapy monitoring, an optimal extracellular vesicle panel should include not only tumor derived biomarkers but also PD-L1, HLA-G, TIGIT related ligands, TGF-β pathway proteins, and inflammatory factors. These signals may indicate a shift of the bone marrow niche toward immunosuppression earlier than M protein changes. 69 , 72 Specifically, Soncini et al’s study showed that bone marrow–derived extracellular vesicles from patients with MM were enriched in HLA-G, PD-1, and PD-L1 and promoted adaptive immune dysfunction. These findings provide experimental evidence supporting extracellular vesicle monitoring to detect reformation of immune cold niches after CAR-T therapy. 72

Circulating nucleic acids, especially ctDNA, are the third class of monitoring targets that can be deeply integrated with nanodiagnostics. They can simultaneously reflect tumor burden, clonal evolution, copy number abnormalities, BCMA and TNFRSF17 related events, and drug resistant mutations, which are often not fully captured by sBCMA or M protein. 58 , 122 , 123 Dhakal et al used ctDNA to evaluate molecular residual disease in MM and demonstrated that ctDNA can identify patients at high risk of relapse, at the same time, Mithraprabhu et al further demonstrated that ctDNA negativity combined with bone marrow MRD negativity can define a population with better prognosis, suggesting that ctDNA can compensate for the spatial sampling limitations of bone marrow MRD. 122 , 124 Furthermore, Hosoya et al developed a ctDNA method for genomic characterization, treatment response monitoring, and early relapse detection in MM. They sequenced 412 samples from 64 patients and demonstrated that ctDNA was associated with key clinical biomarkers and patient outcomes. 58 Chen et al analyzed pretreatment ctDNA in patients with RRMM receiving anti-BCMA CAR-T therapy and found that high ctDNA levels were associated with shorter PFS. These findings indicate that ctDNA can be used not only for post CAR-T follow-up but also for pre CAR-T risk stratification and decisions regarding the intensity of bridging therapy. 123 The value of nanotechnology in circulating nucleic acid detection lies mainly in three aspects. First, magnetic nanoparticles or nanostructured surfaces can enrich short cfDNA fragments. Second, gold nanoparticles, graphene, nanopores, and electrochemical nanointerfaces can amplify low frequency mutation signals. Third, microfluidic nanochips can integrate cfDNA extraction, amplification, probe hybridization, and signal readout into a workflow close to bedside detection. 125–127 In recent years, nanotechnology for ctDNA detection has indicated that nanoplatforms can improve low abundance ctDNA detection through optical, electrochemical, magnetic, and plasmonic signal amplification. Reviews of electrochemical ctDNA detection have also emphasized its potential for clinical translation because of its speed, low cost, and miniaturization. 125–127

Therefore, nanodiagnostics for MM CAR-T should not be understood as a single more sensitive detector. Instead, they should be designed as a longitudinal relapse warning system. In this system, sBCMA reflects antigen shedding and the BCMA therapeutic environment, BCMA, CD38, or PD-L1 positive extracellular vesicles reflect bone marrow niche communication and rebound immunosuppression, and ctDNA reflects clonal residual disease, antigen escape, and spatial lesion dissemination. 72 , 121 , 123 From a clinical translation perspective, the most feasible pathway is to establish multicomponent liquid biopsy curves before CAR-T infusion, at peak expansion, at 1, 3, and 6 months, and when MRD becomes positive again. These curves should be modeled together with bone marrow NGF and NGS MRD, PET and CT findings, CAR-T copy number, and cytokine profiles, thereby advancing relapse intervention to the molecular relapse stage. 122–124 If sBCMA nanoimmunosensing, extracellular vesicle nanoflow or microfluidic capture, and ctDNA nanoenhanced sequencing can be integrated into the same monitoring platform, efficacy evaluation of MM CAR-T may shift from the static conclusion of whether MRD negativity has been achieved to a dynamic spatial immune map showing when, where, and through which biological pathway relapse occurs. 68 , 72 , 121 , 123

Spatial Multiomics and Nanoimaging

If MRD, sBCMA, and ctDNA indicate whether disease remains, spatial multiomics and nanoimaging further clarify where residual lesions are located, which cells surround them, whether the target antigen remains detectable, and whether CAR-T cells have reached the lesion and exerted function. This issue is particularly important for MM CAR-T therapy because myeloma relapse often shows spatial features, including patchy bone marrow involvement, extramedullary lesions, antigen heterogeneity, and immune exclusion. 66 , 128–130

Conventional bone marrow aspiration, flow cytometry MRD, and NGS MRD are affected by sampling site, hemodilution, and missed detection of extramedullary lesions. In contrast, spatial transcriptomics, spatial proteomics, imaging mass spectrometry, cyclic staining imaging, and AI pathology can analyze the in situ spatial relationships among plasma cells, CAR-T cells, endogenous T cells, myeloid cells, stromal matrix, blood vessels, and bone remodeling areas. These approaches advance MM efficacy monitoring from cell proportion analysis to spatial immune atlas construction. 45 , 131–133 For example, Dhodapkar et al performed high dimensional analysis of bone marrow tumor and immune cells after BCMA CAR-T therapy and found that remission persistence was associated with the non-CAR-T immune composition of the bone marrow. These findings suggest that CAR-T efficacy is not determined solely by the number of infused cells but is jointly shaped by CAR-T cells, endogenous immune cells, and the tumor niche. 66 In addition, Robinson et al further studied MM and MGUS using high dimensional spatial analysis combined with ex vivo and in vivo models and identified regions of T-cell enrichment, T-cell sparsity, and T-cell exclusion in MM biopsies. They also demonstrated that CD2 and C558 interactions and CLEC9A positive dendritic cells can regulate antigen specific T-cell entry into tumor clusters, providing spatial mechanistic evidence for failed CAR-T entry into myeloma niches. 129 John et al combined tomo seq, 10x Visium, and scRNA-seq in 14 extramedullary MM biopsies and found substantial copy number subclonal heterogeneity, differential spatial expression of GPRC5D and TNFRSF17 or BCMA, and diffuse colocalization of TIM3 positive and PD-1 positive exhausted T cells with MM cells in extramedullary lesions. Functional CD8 positive T cells and M1 macrophages were mainly distributed in relatively tumor free areas, indicating that CAR-T recognizable targets and CAR-T accessible lesions do not always spatially overlap. 128 Furthermore, Desantis et al used MACSima imaging cyclic staining to perform high dimensional spatial imaging of 10 matched pairs of bone marrow and extramedullary lesion samples. They found increased BCL-2 and EZH2 expression in extramedullary plasma cells, decreased CD38 expression, and significantly increased distances from CD3 positive T cells. These findings suggest that spatial protein imaging can simultaneously reveal therapeutic target loss, antiapoptotic vulnerability, and impaired immune surveillance, these differences may serve individualized diagnosis and targeted therapy design. 130

Recently, in the study of Yip et al, they used subcellular transcriptomics to analyze 5,001 genes in 21 bone marrow biopsy samples and generated a human bone marrow spatial map. They found that malignant plasma cells can form spatially separated microenvironments and that plasma cell subpopulations and stromal or immune compositions differ substantially across patients and even across different regions of the same tissue. 131 This study indicates that spatial transcriptomics can decipher major bone marrow components in healthy, precursor, and active MM samples and identify spatially restricted plasma cell subpopulations and their distinct microenvironments. In addition, Muiños Lopez et al applied Visium spatial transcriptomics to FFPE bone marrow samples and demonstrated its ability to identify hotspots, boundary regions, and distal regions of MM plasma cells in mineralized bone tissue. They also observed spatial gradients of NETosis and IL-17 pathways, as well as a transition from effector T cells to exhausted T cells. These findings indicate that CAR-T monitoring should focus not only on the presence of T cells but also on whether T cells are located in the tumor core, boundary, or distal niche. 132 The study also emphasized that decalcification, RNA integrity, the non-single cell resolution of Visium, and low bone marrow cell viability remain key technical limitations for translating bone marrow spatial omics. Hagos et al used deep learning to spatially map bone marrow trephine biopsies, showing that AI pathology can reconstruct the mosaic cell topology of the myeloma microenvironment from conventional tissue sections and correlate it with disease progression and structural changes after treatment. This provides a basis for transferring spatial omics models to more affordable and accessible digital pathology platforms in the future ( Figure 5 ). 133 In CAR-T efficacy monitoring, such spatial maps can support clinical decision making at three levels. Before treatment, they can identify spatial heterogeneity of targets such as BCMA, GPRC5D, and CD38 in bone marrow and extramedullary lesions. During treatment, they can determine whether CAR-T cells have entered plasma cell enriched areas. After treatment, they can identify whether MRD positive cells are protected by CAF, endothelial, myeloid, or hypoxic niches. 128 , 129 , 132 , 133
Figure 5 Deep learning-assisted spatial profiling of the multiple myeloma bone marrow microenvironment from multiplex immunohistochemistry images. ( A ) Workflow for multiplex immunohistochemical analysis of myeloma bone marrow trephine biopsies. Formalin-fixed paraffin-embedded bone marrow biopsy sections are processed using established automated staining procedures with three markers, providing conventional tissue images suitable for downstream digital pathology analysis. ( B ) Development of deep learning models for automated tissue segmentation and cellular phenotyping. MoSaicNet is used to segment bone marrow histological compartments, including bone, fat, and hematopoietic tissue regions, whereas AwareNet performs cell detection and classification while accounting for rare cell populations within the complex marrow niche. ( C ) Quantitative and spatial readouts enabled by the deep learning pipeline. The platform supports measurement of cell-type abundance, density, and ratios, as well as spatial organization features such as cellular clustering and cell-to-cell proximity. In parallel, bone structural parameters, including texture and thickness, can be extracted to assess bone physiology. Together, this AI-based pathology workflow reconstructs the spatial architecture of the myeloma bone marrow niche from routine biopsy sections and provides a scalable approach for linking microenvironmental topology with disease progression and treatment-associated structural changes. Adapted from Hagos YB, Lecat CSY, Patel Det al Deep Learning Enables Spatial Mapping of the Mosaic Microenvironment of Myeloma Bone Marrow Trephine Biopsies.  Cancer Res . Copyright © 2024 by authors. 133 A) Multiplex immunohistochemistry of myeloma bone marrow biopsy using well-established methods. It shows FFPE trephine biopsy and automated staining with three markers. B) Development of a novel deep learning model for tissue segmentation, cell detection and classification. Tissue segmentation (MoSaicNet) identifies bone, fat and tissue. Cell detection and classification (AwareNet) accounts for rare cell types, showing input and output stages. C) New capability in quantitative and spatial analyses of tumor niche in bone marrow. It includes abundance of cell types with density and ratio, spatial analysis with clustering and cell proximity and bone physiology with texture and thickness. Infographic on deep learning for myeloma bone marrow biopsy analysis.

Deep learning-assisted spatial profiling of the multiple myeloma bone marrow microenvironment from multiplex immunohistochemistry images. ( A ) Workflow for multiplex immunohistochemical analysis of myeloma bone marrow trephine biopsies. Formalin-fixed paraffin-embedded bone marrow biopsy sections are processed using established automated staining procedures with three markers, providing conventional tissue images suitable for downstream digital pathology analysis. ( B ) Development of deep learning models for automated tissue segmentation and cellular phenotyping. MoSaicNet is used to segment bone marrow histological compartments, including bone, fat, and hematopoietic tissue regions, whereas AwareNet performs cell detection and classification while accounting for rare cell populations within the complex marrow niche. ( C ) Quantitative and spatial readouts enabled by the deep learning pipeline. The platform supports measurement of cell-type abundance, density, and ratios, as well as spatial organization features such as cellular clustering and cell-to-cell proximity. In parallel, bone structural parameters, including texture and thickness, can be extracted to assess bone physiology. Together, this AI-based pathology workflow reconstructs the spatial architecture of the myeloma bone marrow niche from routine biopsy sections and provides a scalable approach for linking microenvironmental topology with disease progression and treatment-associated structural changes. Adapted from Hagos YB, Lecat CSY, Patel Det al Deep Learning Enables Spatial Mapping of the Mosaic Microenvironment of Myeloma Bone Marrow Trephine Biopsies.  Cancer Res . Copyright © 2024 by authors. 133

Nanoimaging provides dynamic in vivo complementarity to spatial multiomics because tissue spatial omics typically represents a static post sampling snapshot. Nanotracers, nanobody PET, and iron oxide nanoparticles can track target expression, lesion distribution, and CAR-T migration at the systemic scale. 134–136 For instance, Kiru et al labeled CAR-T cells with ferumoxytol, a clinically translatable iron oxide nanoparticle, and achieved efficient nanoparticle uptake through microfluidic mechanoporation while preserving T-cell viability, proliferation, and cytotoxic function. MRI, photoacoustic imaging, and magnetic particle imaging subsequently demonstrated CAR-T homing to tumors and non-target organs in animals. These findings suggest that nanolabeled CAR-T cells can help determine whether treatment failure results from failure to reach the lesion, insufficient expansion, or functional impairment after arrival. 134

PET imaging has higher sensitivity and deeper tissue penetration for CAR-T tracking. Glennan et al concluded that PET can evaluate CAR-T migration, expansion, and tissue retention in patients through direct cell labeling or reporter gene strategies, thereby optimizing administration, toxicity warning, and efficacy assessment. 135 For MM, target-specific imaging is more closely aligned with CAR-T biology than conventional FDG PET. For example, Gu et al conducted the first-in-human BCMA-targeted PET CT study using 68Ga labeled nanobody tracers to visualize intramedullary, paramedullary, and extramedullary lesions. Compared with 18F FDG PET CT, this approach showed better lesion visualization and dynamic assessment of BCMA expression, suggesting that BCMA PET may become an imaging companion diagnostic tool for pretreatment target screening, post CAR-T relapse localization, and decisions regarding subsequent BCMA-directed therapy. 137

In addition, Huang et al developed the CD38 specific 68Ga nanobody PET probe 68Ga Ga-TOHP-CD38 and completed preclinical and first-in-human studies, demonstrating its advantages of rapid blood clearance, rapid enrichment in target lesions, and same day imaging. This provides a nanoimaging tool for evaluating plasma cell distribution and CD38 target status during anti CD38 therapy, CAR-T bridging therapy, or BCMA-negative relapse. 138 Ghaemi et al further developed 18F labeled BCMA nanobodies for PET imaging and explored the therapeutic potential of 131I labeled radioligands. They achieved high contrast lesion imaging and proof of concept validation of radiotherapy in a disseminated MM model. These findings suggest that the nanobody platform may integrate visualization of BCMA-positive residual lesions with targeted clearance of residual lesions in a future diagnostic and therapeutic strategy. 139

From a translational perspective, the current priority is not to simply combine spatial omics and nanoimaging as advanced detection technologies, but to establish an operational monitoring loop. Before treatment, BCMA, CD38, and GPRC5D nano PET and bone marrow spatial mapping can guide target selection and extramedullary risk assessment. During early treatment, CAR-T nanotracing or PET reporter genes can determine whether cells have reached the lesion. After treatment, spatial omics can verify whether antigen negativity, T-cell exclusion, myeloid barriers, or hypoxic and matrix protection exist at relapse sites. 128 , 134 , 135 , 139

This closed loop can also be integrated with MRD, ctDNA, sBCMA, and extracellular vesicle detection, enabling cross validation among molecular residual disease, spatial residual disease, and imaging residual disease. It may be particularly useful for patients who are bone marrow MRD negative but PET positive, have elevated sBCMA, or are at risk of extramedullary relapse. 135 , 137–139 At present, clinical translation still needs to address several issues, including decalcification and standardized preservation of spatial samples, standardization of antibody and probe panels, the effects of CAR-T tracers on cell function and label dilution, renal uptake and radiation dose of nanobodies, and prospective validation linking imaging readouts with CAR-T efficacy endpoints. 132 , 135 , 137–139

Overall, spatial multiomics enables visualization of the CAR-T treatment landscape at the tissue level, whereas nanoimaging enables tracking of CAR-T dynamics at the in vivo scale. By integrating these two approaches, MM CAR-T efficacy monitoring may shift from single time point MRD assessment to a dynamic spatial immune map that can guide target selection, niche remodeling, and intervention before relapse.

Nanomedicine-Guided Therapeutic Sequencing in the Post-BCMA Era

In the post BCMA era, MM treatment is no longer defined by whether new drugs are available, but by how CAR-T therapy, bispecific antibodies, ADCs, and CELMoDs can be sequenced for the same patient under limited T-cell reserves, limited BCMA antigen availability, and sustained relapse pressure. Therefore, nanomedicine is not merely to provide new carriers, but to support antigen monitoring, immune remodeling, bridging therapy, and low toxicity combination delivery. These functions may shift sequential therapy from empirical decision making toward measurable and modifiable dynamic strategies. 136 , 140–142

BCMA and CD3 bispecific antibodies, such as teclistamab and elranatamab, can form immune synapses by simultaneously binding CD3 on T cells and BCMA on bone marrow tumor cells. They can be used as ready to use T-cell redirecting therapies. They have practical value during the CAR-T waiting period, when CAR-T therapy is not accessible, or after relapse following CAR-T therapy. Both MajesTEC-1 and MagnetisMM-3 demonstrated that teclistamab and elranatamab, respectively, achieved high response rates and manageable toxicity in triple class exposed RRMM. In MagnetisMM-3 cohort A, the reported ORR was 61.0% and the complete response or better rate was 35.0%. However, infection, cytopenia, and sustained T-cell stimulation remain important risks when these agents are used sequentially. 136 , 140

Compared with CAR-T therapy, bispecific antibodies do not require individualized manufacturing and have greater treatment accessibility, but continuous administration may chronically impair endogenous T-cell function. Therefore, when patients can wait and CAR-T manufacturing is feasible, existing real world studies tend to support preserving T-cell fitness for CAR-T therapy and using bispecific antibodies for bridging, relapse after CAR-T therapy, or conversion to non-BCMA targets. 50 , 143 , 144 For example, Ferreri et al’s 11 center real world study supports this cautious approach. Patients previously exposed to BCMA-targeted therapy who subsequently received ide-cel still achieved an ORR of 74%, but their ORR, DOR, and PFS were lower than those of BCMA naive patients, especially among patients previously exposed to BCMA bispecific antibodies. This suggests that sequential use of therapies against the same target under sustained T-cell pressure may lead to combined loss of antigen availability and T-cell fitness. 143 This study showed that the ORR was 74% in the BCMA-targeted therapy exposed group and 88% in the unexposed group, while median PFS was 3.2 months and 9.0 months, respectively. It also suggests that longer intervals after BCMA-targeted therapy may be associated with better responses. 143 Meanwhile, Fandrei et al’s study on bispecific antibody bridging before BCMA CAR-T therapy suggests that this approach is feasible. However, bridging protocols should be short, stoppable, and closely monitored, with priority given to avoiding excessive T-cell exhaustion or severe infection before CAR-T collection. 144

The sequential logic of ADCs differs from that of bispecific antibodies. Belantamab mafodotin delivers MMAF to BCMA-positive plasma cells through an anti BCMA antibody, and its cytotoxic mechanism is closer to targeted chemotherapy. Theoretically, it may consume less T-cell function and can be considered for tumor reduction during the CAR-T waiting period or for relapse after CAR-T therapy. However, it also exerts selective pressure on BCMA and may cause corneal toxicity, low BCMA expression, or selection of BCMA loss clones. 38 , 145 , 146 Firestone et al reported that loss of BCMA expression can become a shared resistance mechanism among BCMA CAR-T therapy, belantamab, and BCMA bispecific antibodies. This indicates that, although ADCs, CAR-T therapy, and bispecific antibodies are different therapeutic modalities, they may share the risk of antigen escape when they target BCMA. Therefore, before rechallenging with BCMA-directed therapies after relapse, BCMA expression and TNFRSF17 status should be reassessed. 38 As DREAMM-7 and DREAMM-8 demonstrate the clinical value of belantamab combined with bortezomib and dexamethasone or pomalidomide and dexamethasone in RRMM, ADCs may no longer be limited to last line monotherapy in the post BCMA era. Instead, they may enter earlier line combination regimens or CAR-T sequential scenarios. However, this forward shift will make the question of whether to use ADCs first or preserve BCMA targeting for CAR-T therapy more clinically urgent. 145 , 146

Non BCMA-targeted therapies provide the possibility of changing treatment tracks in sequential therapy. The GPRC5D and CD3 bispecific antibody talquetamab induces T-cell cytotoxicity in MonumenTAL-1 by linking CD3 and GPRC5D and can be used in patients who relapse after BCMA-targeted therapy. The phase 1 updated analysis of the GPRC5D CAR-T product MCARH109 also showed that 12 of 17 patients responded, suggesting that switching to the GPRC5D axis after BCMA failure is biologically and clinically feasible. 147 , 148 The specific evidence comes from Chari et al’s study, their study indicated that talquetamab kills myeloma cells by redirecting T cells through GPRC5D and CD3. 147 At the same time, Jurgens et al’s JCO updated analysis reported that MCARH109 has acceptable safety at a maximum tolerated dose of 150 million CAR-T cells. 148 Furthermore, RedirecTT-1 combined the BCMA-targeted agent teclistamab with the GPRC5D targeted agent talquetamab and demonstrated that dual antigen T-cell redirection can generate higher responses in heavily pretreated RRMM, especially in extramedullary lesions. However, the burden of infection and immune toxicity suggests that dual target combinations should not be understood simply as two effective drugs added together. They require more precise dosing, treatment intervals, infection prevention, and immune recovery monitoring. 136 CELMoDs may function as immune retuning agents in the post BCMA treatment sequence. Iberdomide and mezigdomide enhance CRBN mediated degradation of Ikaros and Iolos, which can directly inhibit myeloma cells and enhance T-cell and NK-cell immune responses. Therefore, they may be suitable as non-antigen targeted therapies for bridging before CAR-T therapy, immune recovery after CAR-T therapy, T-cell reactivation before bispecific antibody therapy, or treatment intervals between BCMA and GPRC5D directed therapies. 149–151 For example, Lonial et al demonstrated in CC-220-MM-001 that iberdomide combined with dexamethasone had clinical activity and manageable safety in heavily pretreated RRMM. 149 Furthermore, Richardson et al reported in NEJM that mezigdomide combined with dexamethasone showed promising efficacy in multiline relapsed or refractory MM. However, myelosuppression remains a limiting factor that requires caution when these agents are sequenced with CAR-T therapy or bispecific antibodies. 150 More mechanistically, Van Oekelen et al used patient bone marrow samples to demonstrate that iberdomide can increase innate and adaptive immune cell subpopulations. This provides sample level support for positioning CELMoDs as immune ecological remodeling agents before and after T-cell redirecting therapy. 151

Nanomedicine can play three roles in this complex sequencing process. The first is delivery guided sequencing, in which bone marrow or plasma cell targeted nanoparticles deliver small molecules, siRNA, or RNA therapeutics to MM cells and bone marrow niches. This approach may enable low toxicity tumor reduction, antigen densification, or microenvironment mobilization before CAR-T therapy rather than relying on high dose chemotherapy bridging. 67 , 103 , 104 Tarab Ravski et al developed CD38 targeted LNPs to deliver therapeutic RNA to bone marrow MM cells. The results demonstrate the potential of RNA therapy in the treatment of multiple myeloma, as well as the importance of developing effective targeted delivery systems and reliable preclinical models. 103 Furthermore, Guimarães et al used lipid polymer nanoparticles to deliver siRNA to the bone marrow endothelium, silence cyclophilin A, and inhibit MM extravasation. 67 Chen et al used antibody mediated nanomedicine delivery of carfilzomib to enhance anti-MM effects. 104 Together, these studies demonstrate that nanodelivery can shift bridging therapy from systemic tumor suppression toward a bone marrow–targeted remodeling window. 67 , 103 , 104 Another key strategy involves monitoring-guided sequencing, in which sBCMA, BCMA- or CD38-positive extracellular vesicles, ctDNA, and spatial or nanoimaging are used to assess whether patients should continue BCMA-directed therapy or transition to GPRC5D-, FcRH5-, CELMoD-, or ADC-based combinations. This strategy is particularly suitable for relapse after CAR-T therapy because antigen escape, T-cell exhaustion, and extramedullary lesions often require different next line regimens. 38 , 136 , 152

In addition, combination-guided sequencing enables the integration of CELMoDs, GSIs, proteasome inhibitors, checkpoint-regulating RNAs, or cytokine modules into short-range, low-dose, and controllable nanocombination regimens. These regimens may enhance antigen density and T-cell fitness before CAR-T therapy, prevent exhaustion after peak CAR-T expansion, or reduce T-cell failure caused by sustained antigen stimulation during bispecific antibody therapy. 67 , 103 , 104 , 151 Based on existing evidence, a rational framework for the post BCMA era is that, if CAR-T manufacturing is feasible and disease can be controlled with short-term bridging, BCMA CAR-T therapy should be considered before long-term exposure to BCMA bispecific antibodies. If relapse occurs after CAR-T therapy and BCMA remains positive with sufficient T-cell reserve, BCMA bispecific antibodies or ADCs may be considered, although limited response durability should be considered. If BCMA expression is low or lost, recent BCMA-directed therapy has failed, or high-risk extramedullary lesions are present, priority should be given to non BCMA strategies such as GPRC5D or FcRH5 targeting, CELMoD mediated immune remodeling, or dual target combinations. Nanodiagnostics and nanodelivery should then be used to narrow the toxicity window, prolong treatment intervals, and reduce cross resistance. 38 , 136 , 152

Nanocell Therapies Beyond CAR-T: CAR-NK, CAR-M, and CAR Exosomes

As BCMA CAR-T therapy moves MM treatment into the era of deep remission, the next step is not simply to manufacture more potent CAR-T cells. Instead, it is to establish a nanocell therapy ecosystem composed of CAR-NK cells, CAR-M cells, and CAR exosomes. This ecosystem could expand treatment from T-cell centered cytotoxicity to multidimensional strategies involving innate immunity, phagocytic clearance, extracellular vesicle delivery, and off the shelf platforms. 153–155 CAR-NK cells represent the closest clinical alternative to CAR-T cells because NK cells can kill MM cells through CAR recognition, natural killer receptors, CD16 mediated ADCC, perforin and granzyme release, and TRAIL and FasL pathways. In addition, allogeneic NK cells have a lower risk of inducing GVHD and are suitable as off the shelf cell products. 153–155 FT576, developed by Cichocki et al, is one of the most representative MM examples in this field. This product is constructed from iPSC derived NK cells and integrates an anti-BCMA CAR, high affinity noncleavable CD16, an IL-15 and IL-15Rα fusion protein, and CD38 knockout. These features provide multiple functions, including direct BCMA recognition, ADCC coordination with daratumumab, enhanced persistence, and avoidance of CD38 antibody mediated fratricide. 153 In MM models, this study demonstrated that FT576 can achieve multiantigen targeting and durable antitumor activity. A ClinicalTrials.gov record indicates that FT576 has entered a phase 1 dose exploration study in RRMM to evaluate monotherapy and combination regimens with daratumumab. These findings suggest that CAR-NK therapy has moved from an engineering concept toward clinical translation. 153 , 156

The single domain antibody or VHH directed BCMA CAR-NK cells developed by Ren et al further demonstrated that BCMA CD28 IL-15 CAR-NK cells had stronger ex vivo cytotoxicity and cytokine secretion than BCMA 2B4 IL-15 CAR-NK cells. These findings indicate that the CAR intracellular domain and IL-15 autocrine module directly influence NK-cell expansion, activation, and anti-MM activity. 154 Ng et al modified CAR-NK cells from the perspective of bone marrow homing by co introducing CXCR4 and an anti-BCMA CAR into NK cells, thereby enhancing bone marrow localization and inhibiting disease progression in a xenograft MM model. This design closely matches the spatial biological characteristics of MM as a bone marrow niche disease. 155 In addition, Kaczmarek et al further demonstrated that coexpression of anti-BCMA CAR and IL-15 is critical for sustained CAR-NK mediated control of MM in blood and bone marrow. However, their model also showed BCMA-positive extramedullary plasma cell tumors, suggesting that CAR-NK therapy should be combined with extramedullary lesion imaging, local delivery, or multitarget strategies in the future rather than relying solely on bone marrow clearance as the efficacy endpoint. 157 The value of nanotechnology for CAR-NK therapy is first reflected in nonviral engineering. Golubovskaya et al efficiently transfected and expanded NK cells using CAR mRNA LNPs of approximately 90 to 100 nm, enabling BCMA CAR-NK cells to achieve strong MM cell cytotoxicity, IFN-γ release, and granzyme B release. 158 Douka et al also demonstrated that LNP mRNA delivery can achieve high transfection rates and good activity in NK cells, indicating that mRNA LNPs may reduce viral manufacturing costs and support short-term, reproducible CAR-NK programming. 159

Compared with CAR-NK cells, CAR-M cells can be viewed as engineered phagocytic cells that activate the cold bone marrow niche. Macrophages can not only mediate CAR directed phagocytosis but also secrete inflammatory factors, present antigens, reshape the myeloid network, and activate secondary T-cell responses. 160 , 161 At present, clinical research on MM specific CAR-M remains limited. However, Reiss et al’s phase 1 study of the anti HER2 CAR-M product CT-0508 showed that autologous CAR-M cells can be manufactured and infused in patients with solid tumors and can mediate phagocytosis, cytokine release, antigen presentation, and tumor microenvironment activation. This provides clinical evidence supporting the manufacturability, infusibility, and monitoring of CAR-M cells, which may facilitate translation of the CAR-M concept to the MM bone marrow niche. 160 Pierini et al further demonstrated that CAR-M cells can reshape the tumor microenvironment, induce T-cell antitumor immunity, and sensitize solid tumors to PD-1 and PD-L1 blockade. This suggests that the most valuable application of CAR-M in MM may not be replacement of CAR-T therapy, but use as a niche remodeling platform to clear residual plasma cells, relieve myeloid suppression, and enhance subsequent responses to CAR-T therapy or bispecific antibodies. 161 In MM, the BCMA and CD24 bispecific CAR-T cells developed by Sun et al showed that residual MM cells were enriched for CD24 after BCMA CAR-T therapy. CD24 CAR-T cells could block the CD24 and Siglec-10 do not eat me signal and enhance macrophage mediated phagocytic clearance while inducing M1 like polarization. This provides a direct phagocytic immunologic rationale for developing CD24, BCMA, or CD38 directed CAR-M cells in MM ( Figure 6 ). 162 However, macrophage polarization is highly plastic, and hypoxia and immunosuppressive signals in the bone marrow may drive macrophages back toward an M2-like state. 162 , 163 Therefore, nanotechnology is unlikely to permanently lock macrophages into an M1 phenotype, but may help sustain antitumor activity through bone marrow-targeted or hypoxia-responsive delivery of mRNA, siRNA, or modulators of pathways such as HIF, STAT3, PI3Kγ, CSF1R, IL-10, and TGF-β. 67 , 164 Repeatable, nonintegrating mRNA delivery may be particularly suitable for periodically reinforcing CAR-M function while limiting permanent genetic modification. At present, nanomedicine-enhanced CAR-M platforms remain preclinical in MM. 160 , 165 Oxidized or targeted mRNA LNPs have enabled in situ CAR monocyte or macrophage programming in animal models, whereas the available phase 1 CAR-M clinical evidence is limited to the anti-HER2 product CT-0508 in solid tumors rather than MM. Thus, these findings provide a rationale for future CD24-, BCMA-, or CD38-directed CAR-M strategies, although durable activity will likely require repeated reprogramming and simultaneous control of hypoxia-driven M2 polarization. 103 , 160
Figure 6 BCMA/CD24 bispecific CAR-T cells remodel the phagocytic immune niche in multiple myeloma. Pre-CAR-T treatment: Multiple myeloma (MM) cells, including drug-resistant MM cells, proliferate within the bone marrow niche and exploit immune-evasive interactions with macrophages. CD24 expressed on MM cells engages Siglec-10 on macrophages, activating ITIM-associated inhibitory signaling, suppressing cytoskeletal rearrangement and inflammatory activation, and thereby limiting macrophage-mediated phagocytosis. In parallel, CXCL12/CXCR4 signaling and tumor-derived cues favor macrophage polarization toward an M2-like, pro-tumoral, anti-inflammatory, and immunosuppressive phenotype. Post-CAR-T treatment: BCMA/CD24 bispecific CAR-T cells target MM cells through BCMA-directed cytotoxicity while also disrupting the CD24–Siglec-10 “do-not-eat-me” signal. CAR-T-cell-derived inflammatory mediators, including IFN-γ and TNF-α, together with blockade of CD24-mediated anti-phagocytic signaling, promote macrophage cytoskeletal activation, restore phagocytic clearance of residual MM cells, and shift macrophage polarization toward an M1-like anti-tumoral phenotype. This mechanism provides an immunologic rationale for targeting CD24, BCMA, or related myeloma antigens to enhance macrophage-mediated tumor elimination in MM. Adapted from Sun F, Cheng Y, Wanchai Vet al Bispecific BCMA/CD24 CAR-T cells control multiple myeloma growth.  Nat Commun . Copyright © 2024 by authors. 162 The diagram shows the impact of BCMA/CD24 bispecific CAR-T cells on multiple myeloma cells before and after treatment. Before CAR-T treatment, myeloma cells grow in the bone marrow, engaging Siglec-10 on macrophages via CD24, triggering inhibitory signaling that prevents cytoskeletal changes and inflammation. This results in anti-phagocytosis and macrophage polarization to a pro-tumoral M2-like state. After CAR-T treatment, BCMA/CD24 CAR-T cells target myeloma cells, disrupting CD24 to Siglec-10 signaling. CAR-T cells release inflammatory mediators like IFN-gamma and TNF-alpha, activating macrophage cytoskeleton and promoting phagocytosis of remaining myeloma cells, shifting macrophages to an anti-tumoral M1-like state. The diagram labels myeloma cells, drug-resistant cells, macrophages and signaling molecules. BCMA/CD24 CAR-T cells impact myeloma, altering macrophage polarization and signaling pre/post-treatment.

BCMA/CD24 bispecific CAR-T cells remodel the phagocytic immune niche in multiple myeloma. Pre-CAR-T treatment: Multiple myeloma (MM) cells, including drug-resistant MM cells, proliferate within the bone marrow niche and exploit immune-evasive interactions with macrophages. CD24 expressed on MM cells engages Siglec-10 on macrophages, activating ITIM-associated inhibitory signaling, suppressing cytoskeletal rearrangement and inflammatory activation, and thereby limiting macrophage-mediated phagocytosis. In parallel, CXCL12/CXCR4 signaling and tumor-derived cues favor macrophage polarization toward an M2-like, pro-tumoral, anti-inflammatory, and immunosuppressive phenotype. Post-CAR-T treatment: BCMA/CD24 bispecific CAR-T cells target MM cells through BCMA-directed cytotoxicity while also disrupting the CD24–Siglec-10 “do-not-eat-me” signal. CAR-T-cell-derived inflammatory mediators, including IFN-γ and TNF-α, together with blockade of CD24-mediated anti-phagocytic signaling, promote macrophage cytoskeletal activation, restore phagocytic clearance of residual MM cells, and shift macrophage polarization toward an M1-like anti-tumoral phenotype. This mechanism provides an immunologic rationale for targeting CD24, BCMA, or related myeloma antigens to enhance macrophage-mediated tumor elimination in MM. Adapted from Sun F, Cheng Y, Wanchai Vet al Bispecific BCMA/CD24 CAR-T cells control multiple myeloma growth.  Nat Commun . Copyright © 2024 by authors. 162

Nanomedicine is shifting CAR-M development from ex vivo manufacturing toward in vivo programming. For example, Mukalel et al developed oxidized lipid mRNA LNPs for in situ generation of CAR monocytes and macrophages. In response to the difficulty of traditional screening methods in obtaining LNPs with both monocyte tropism and efficient intracellular delivery capability, researchers identified oxidized LNPs through ionizable lipid design and high-throughput in vivo DNA barcode screening. Among them, C14-O2 exhibited innate monocyte tropism and could generate functional CD19-CAR monocytes in situ in healthy mice, inducing approximately 45% of B cell loss. This suggests that oxidized LNPs can serve as a promising delivery platform for CAR monocyte/macrophage engineering ( Figure 7 ). 165 Gu et al constructed customized CAR-M cells in vivo using targeted mRNA LNPs and screened different combinations of intracellular signaling domains. These studies indicate that injectable nanosystems may be used in the future to transiently program BCMA, CD38, or CD24 CARs in bone marrow myeloid cells, thereby avoiding complex cell preparation and improving spatial accessibility. 166
Figure 7 Oxidized lipid nanoparticles for monocyte-tropic mRNA delivery and in situ CAR monocyte engineering. ( A ) Design rationale for oxidation-engineered ionizable lipid nanoparticles. Ether-spaced oxidized ionizable lipids exhibit increased ionization under endosomal pH conditions compared with alkyl-spaced counterparts, thereby promoting endosomal escape and enhancing mRNA delivery to macrophages, while showing limited delivery efficiency in non-target immune cells such as T cells. ( B ) High-throughput in vivo screening strategy for identifying monocyte-tropic LNP formulations. A library of LNPs was systemically administered and evaluated in mice using barcoded nucleic acid readouts to determine formulations with preferential delivery to circulating monocytes. The lead oxidized LNP formulation, represented by C14-O2, enabled in situ engineering of CAR monocytes through mRNA delivery, supporting transient CAR expression and functional immune-cell reprogramming without conventional ex vivo cell manufacturing. This platform highlights oxidized LNPs as a promising nonviral strategy for generating CAR monocyte/macrophage therapies with improved manufacturability and controllable expression profiles. Adapted from Mukalel AJ, Hamilton AG, Billingsley MMet al Oxidized mRNA Lipid Nanoparticles for In Situ Chimeric Antigen Receptor Monocyte Engineering. Adv Funct Mater. Copyright © 2024by authors. 165 The diagram consists of two parts. A shows ionizable lipid synthesis with two pathways: ether-spaced and alkyl-spaced ionizable lipids. Ether-spaced lipids show increased ionization at endosomal pH, enhancing delivery to macrophages, depicted by directional arrows leading to macrophage cultures. B illustrates high throughput in vivo screening of a lipid nanoparticle library in mice. The process involves systemic administration, leading to in situ engineering of chimeric antigen receptor monocytes, shown by arrows indicating delivery to blood vessels and monocyte transformation. Diagram showing ionizable lipid synthesis and in vivo screening for CAR monocyte engineering in two parts.

Oxidized lipid nanoparticles for monocyte-tropic mRNA delivery and in situ CAR monocyte engineering. ( A ) Design rationale for oxidation-engineered ionizable lipid nanoparticles. Ether-spaced oxidized ionizable lipids exhibit increased ionization under endosomal pH conditions compared with alkyl-spaced counterparts, thereby promoting endosomal escape and enhancing mRNA delivery to macrophages, while showing limited delivery efficiency in non-target immune cells such as T cells. ( B ) High-throughput in vivo screening strategy for identifying monocyte-tropic LNP formulations. A library of LNPs was systemically administered and evaluated in mice using barcoded nucleic acid readouts to determine formulations with preferential delivery to circulating monocytes. The lead oxidized LNP formulation, represented by C14-O2, enabled in situ engineering of CAR monocytes through mRNA delivery, supporting transient CAR expression and functional immune-cell reprogramming without conventional ex vivo cell manufacturing. This platform highlights oxidized LNPs as a promising nonviral strategy for generating CAR monocyte/macrophage therapies with improved manufacturability and controllable expression profiles. Adapted from Mukalel AJ, Hamilton AG, Billingsley MMet al Oxidized mRNA Lipid Nanoparticles for In Situ Chimeric Antigen Receptor Monocyte Engineering. Adv Funct Mater. Copyright © 2024by authors. 165

CAR exosomes represent a cell free approach to cell therapy. The core idea is to use natural nanovesicles released by CAR-T or CAR-NK cells that retain CAR targeting structures, cytotoxic molecules such as perforin and granzyme, and selected immunoregulatory cargo, while avoiding excessive expansion of live cells, long-term retention, and some CRS risk. Recent research has shown that CAR exosomes often lack inhibitory receptors such as PD-1, and their antitumor effects are less easily weakened by PD-L1. They also have potential advantages, including low immunogenicity, easier preservation, engineerable drug loading, and better tissue penetration. 167 , 168 Preliminary evidence is emerging in MM. For example, Sun et al reported that BCMA CAR-T derived exosomes had a typical nanoscale cup shaped structure, an average diameter of approximately 109 nm, and detectable BCMA CAR on the membrane. They also produced dose-dependent lysis of BCMA-positive MM cells, including OPM2, MM.1S, and H929 cells, whereas lysis of OPM2 BCMA knockout cells was very low. This indicates that the anti-MM activity of BCMA CAR exosomes is target antigen dependent. 169 Although this result remains early and is based on a conference abstract, it provides a new cell free platform for post CAR-T maintenance therapy, bridging therapy when CAR-T cannot be performed in a timely manner, or relapse intervention using small molecules or siRNA co loaded with nanomedicine. 68 , 169 Another significance of CAR exosomes in MM is that they may counteract tumor exosome mediated signaling. MM derived exosomes can promote bone marrow immunosuppression, angiogenesis, and drug tolerance, whereas engineered CAR exosomes or immune cell derived exosomes may deliver cytotoxic and immune activating signals. In this way, a relapse promoting vesicle communication system could be transformed into a therapeutic tool. 68 , 170

Overall, CAR-NK cells may address slow manufacturing, high cost, and limited allogeneic accessibility associated with CAR-T therapy. CAR-M cells may address the cold bone marrow microenvironment, myeloid suppression, and insufficient phagocytosis of residual lesions. CAR exosomes may address storage, repeated administration, and safety controllability challenges associated with live cell therapy. Across these platforms, nanotechnology provides a common logic for redefining the manufacturing, delivery, and spatial action of cell therapy through LNPs, exosomes, nanobodies, and programmable carriers. 68 , 154 , 167 , 170 In terms of clinical translation, a realistic hierarchy may be proposed. BCMA CAR-T therapy remains the core platform for deep remission. CAR-NK cells may be used during the CAR-T waiting period, as non-cross resistant cellular therapy after relapse, or in combination with CD38 antibodies. CAR-M cells may be most suitable for patients with strong myeloid suppression, an active CD24 and Siglec-10 axis, or a cold bone marrow niche. CAR exosomes may serve as repeatable maintenance, bridging, or intervention before relapse. Decisions on when to switch to or combine these nanocell therapies should be guided by sBCMA, ctDNA, extracellular vesicles, and spatial imaging monitoring. 68 , 154 , 167 , 170

Clinical Translational Challenges and Future Directions

Clinical Translational Challenges

Although nanomedicine has substantial potential advantages in MM CAR-T therapy, several limitations remain. These challenges require further investigation and systematic clinical validation. The central translational challenge of nanomedicine enabled MM CAR-T therapy is how to integrate complex modules, including ex vivo low damage manufacturing, in vivo T-cell programming, bone marrow–targeted delivery, and relapse monitoring, into clinical strategies that are scalable, quality controlled, regulatory compliant, affordable, and truly beneficial for patients. 171–173

BCMA CAR-T products, including ide-cel and cilta-cel, have demonstrated the ability to induce deep remission in relapsed or refractory MM. However, long-term follow-up and real world studies have shown that relapse, infection, cytopenia, neurotoxicity, antigen escape, and limited post treatment salvage options still restrict claims of curative potential. 11 , 12 , 19 First, manufacturing and accessibility remain the most immediate bottlenecks because autologous CAR-T therapy relies on leukapheresis, viral or nonviral transduction, ex vivo expansion, release testing, cold chain transport, and bridging therapy. The EBMT position paper states that timely collection, manufacturing delivery, availability of qualified centers, multidisciplinary training, clinical trial accessibility, and pricing collectively determine whether patients with MM can receive CAR-T therapy in practice. 63 , 174 Second, although nanodelivery may replace viral transduction and electroporation, its clinical grade implementation remains complex. Key process parameters, raw material variability, cell activation methods, transgene delivery, expansion conditions, release testing, and regulatory requirements for autologous CAR-T therapy can all affect the safety, efficacy, and accessibility of the final product. 175 Current results remain largely limited to ex vivo engineering and preclinical functional validation, and a substantial gap remains before GMP scale production, long-term stability, residual nanomaterial limits, and cross center consistency can be achieved. 73 For instance, Billingsley et al and Metzloff et al used targeted LNPs and APC like LNPs, respectively, to achieve in vivo or rapid mRNA CAR-T engineering. These studies indicate that nanomaterials can shorten the manufacturing process and may partially replace traditional activation and transfection steps. 74 , 76 However, these platforms still need to address translational issues, including T-cell subset selectivity, the mRNA expression window, immune responses after repeated administration, and release potency testing. In vivo CAR-T generation is one of the most disruptive directions, but it also concentrates the highest level of risk because it transforms fully testable ex vivo cell products into gene delivery and immune expansion events that occur inside the patient. Therefore, off-target transduction, non-T-cell expression, insertional mutagenesis, unpredictable expansion, and acute inflammatory reactions must all be evaluated in advance. 77 , 78 , 89 The safety of nanocarriers themselves cannot be inferred solely from the success of mRNA vaccines. Ionizable lipids, PEG lipids, particle size, surface charge, protein corona formation, administration route, and repeated LNP exposure may alter complement activation, TLR related inflammation, liver and spleen uptake, and myeloid cell responses. 176 , 177

Targeted delivery to the bone marrow is another immature translational challenge because MM lesions exhibit spatial heterogeneity, including patchy bone marrow involvement, extramedullary disease, vascular barriers, matrix protection, and myeloid suppression. The actual distribution of nanoparticles among the liver, spleen, bone marrow endothelium, plasma cells, and myeloid cells can substantially affect efficacy and toxicity. 67 , 103 Efficacy monitoring and companion diagnostics also remain clinical gaps. Nanodiagnostics based on ctDNA, sBCMA, extracellular vesicles, and spatial imaging can provide earlier relapse signals, but prospective evidence is still lacking for unified sampling time points, thresholds, cross platform standardization, and direct linkage to treatment decisions. 51 , 137 For example, Hosoya et al demonstrated that ctDNA can analyze response kinetics and drug resistance in MM CAR-T therapy. 58 While Gu et al conducted a first-in-human study using BCMA-targeted PET CT to visualize BCMA-positive lesions. These technologies provide opportunities for dynamic monitoring, but the key clinical question remains whether treatment changes guided by these results truly improve survival. 137 Combination therapy also carries a risk of excessive complexity. GSI, CELMoDs, bispecific antibodies, ADCs, checkpoint regulating RNA, cytokines, and nanodelivery systems may enhance CAR-T efficacy, but they may also increase the risks of infection, cytopenia, CRS, neurotoxicity, and immunosuppression. 9 , 38 , 51 From a regulatory perspective, nano CAR-T related products often span multiple regulatory categories, including cell therapy, gene therapy, nanomedicine, and companion diagnostics. Therefore, systems must be established for carrier characterization, cargo release, cell transduction and transfection efficiency, functional potency, off target distribution, insertion site analysis, immunotoxicity, and long-term follow-up. 78 , 89 , 176 , 177

From the perspective of clinical trials, much of the evidence for nanoenhanced CAR-T therapy remains limited to mouse models, humanized models, or small phase 1 studies. Randomized controls, real world scalability, cost effectiveness analyses, and racial and regional equity evaluations are still lacking. 77 , 78 , 89 Therefore, the clinical translation of nanomedicine enabled MM CAR-T therapy should not be framed as nanodelivery solving all problems. A more rational approach is to begin with low risk, short window, and closely monitored applications, such as ex vivo low damage manufacturing, antigen densification before CAR-T therapy, bone marrow microenvironment pretreatment, and relapse monitoring, and then gradually advance toward in vivo CAR-T generation and multimodule combination therapy. 51 , 67 , 103 , 137 Whether this field can truly change the future landscape of MM treatment depends on whether material innovation can be translated into quantifiable clinical benefits, including shorter waiting times, lower manufacturing failure rates, less severe toxicity, earlier relapse warnings, more rational sequential treatment, and more affordable global accessibility.

Future Research Directions

Future research on nanomedicine enabled MM CAR-T therapy should shift from single point technological innovation toward systems engineering. The goal should not only be to produce CAR-T cells faster and with less damage, but also to ensure that they can accurately reach the bone marrow niche, expand, mediate cytotoxicity, be monitored, and be reactivated when needed. Therefore, the next stage should promote five parallel directions: ex vivo nanomanufacturing, controllable in vivo programming, bone marrow niche remodeling, spatial efficacy monitoring, and standardized clinical translation. 9 , 11 , 12 , 73

On the manufacturing side, lipid nanoparticles, polymer nanocarriers, and antigen presenting cell mimicking nanoparticles should be further developed into low damage, GMP scalable, and multicargo CAR-T manufacturing platforms. LNP delivery of CAR mRNA has been shown to reduce cytotoxicity and maintain stable CAR expression compared with electroporation, while APC mimicking LNPs can integrate T-cell activation and mRNA transfection into a one-step engineering process. 73 , 74 The key scientific question is not whether CAR can be expressed, but how nanomaterials can control the duration of CAR expression, T-cell memory phenotype, metabolic adaptability, and exhaustion trajectory. It is also important to establish release indicators linked to clinical efficacy, such as the proportion of early memory T cells, mitochondrial function, single cell exhaustion scores, and ex vivo sustained cytotoxicity. 39 , 73 , 74 In vivo CAR-T generation is one of the most disruptive frontiers because it may bypass leukapheresis and ex vivo expansion. However, it also compresses gene delivery, T-cell selectivity, in vivo expansion, and acute immune toxicity into a single administration event. Therefore, future work should focus on developing T-cell targeted nanocarriers that are degradable, repeatable, stoppable, and traceable. 30 , 77 , 81 , 89 Future in vivo programming research should prioritize comparison of the benefit risk profiles of transient mRNA expression, prolonged circRNA expression, nonintegrating DNA expression, and integrating viral or transposon systems. Non T-cell transduction, insertional mutagenesis, CAR expression silencing, CRS rescue, and long-term secondary malignancy monitoring should be mandatory design elements in early trials. 30 , 77 , 81 , 89 For bone marrow–targeted delivery, nanomedicine should evolve from systemic administration to enhance CAR-T therapy toward precise pretreatment of the bone marrow niche. This strategy should involve targeted regulation of BCMA density, matrix adhesion, the endothelial barrier, myeloid suppression, extracellular vesicle communication, and hypoxia before and after CAR-T therapy. 51 , 66 , 67 , 103 However, the major limitation of this direction is that most studies remain limited to mouse or ex vivo models and lack prospective patient data on combinations with CAR-T therapy. Therefore, future studies should prioritize window period clinical trials that include pretreatment before CAR-T infusion, niche monitoring during peak CAR-T expansion, and nano reactivation when MRD becomes positive again. 51 , 66 , 67 , 103 Regarding antigen escape, future research should shift from single target enhancement of BCMA to multitarget antigen redundancy design, because TNFRSF17 or BCMA deficiency, low antigen expression, and cross resistance among BCMA-targeted therapies have been confirmed by longitudinal sequencing and clinical samples. 36 , 38 Enhancement of T-cell function remains another important frontier because clinical samples have shown that durable CAR-T responses are closely related to the quality of starting T cells, memory phenotype, and degree of exhaustion. In the future, nanodelivery may be used for transient knockdown of PD-1, TIGIT, and TIM-3, delivery of IL-7 and IL-15 signaling modules, enhancement of glutamine metabolism, or protection of mitochondrial homeostasis. 41 , 42

The most promising future direction for diagnosis and monitoring is to integrate sBCMA, extracellular vesicles, ctDNA, CAR-T copy number, spatial transcriptomics, and nanoimaging into a dynamic spatial immune map, rather than relying on a single bone marrow MRD result as the sole criterion for efficacy assessment. 131 , 132 , 137 , 178 At present, spatial multiomics and nanoimaging still face challenges, including sample decalcification, high cost, complex data analysis, insufficient probe standardization, and lack of treatment decision thresholds. Therefore, future efforts should promote multicenter sample libraries, unified spatial antibody and probe panels, AI pathology algorithms, and prospective studies linking imaging, MRD, and survival. 131 , 132 , 137

In the post BCMA era, future research should address the sequencing of CAR-T therapy, bispecific antibodies, ADCs, CELMoDs, CAR-NK cells, CAR-M cells, and CAR exosomes. Continuous treatment against the same target may accelerate antigen pressure and T-cell exhaustion, whereas target switching or immune remodeling may preserve new treatment windows for patients with relapse. 150 , 153 CELMoDs can enhance T-cell and NK-cell immune responses through the CRBN pathway, CAR-NK cells have off-the-shelf availability and a lower risk of GVHD, CAR-M cells may reshape the myeloid cold niche, and CAR exosomes represent a direction for cell-free nanoimmunotherapy. However, these platforms still lack randomized clinical evidence in MM. Therefore, stratified validation should be prioritized in patients with relapse after CAR-T therapy, extramedullary lesions, or poor T-cell starting material. 150 , 153

From an interdisciplinary perspective, future work should integrate materials science, synthetic biology, spatial genomics, AI modeling, nuclear medicine imaging, and GMP engineering. The central challenge of MM CAR-T therapy has shifted from whether it can kill myeloma cells to whether it can be controlled and monitored over the long-term, at the right time, in the right cells, and within the right niche. 131 , 132 , 137 Regulatory innovation is also needed because nano CAR-T products may simultaneously have properties of cell therapy, gene therapy, nanomedicine, and companion diagnostics. Interdisciplinary evaluation criteria should be established for particle size, surface ligands, cargo release, tissue distribution, immunogenicity, insertion site analysis, CAR expression kinetics, and long-term follow-up. 30 , 179–181 For instance, cell membrane-coated nanoparticles and extracellular vesicles (EVs) offer attractive strategies for remodeling the multiple myeloma bone marrow niche because they can combine the tunable cargo capacity of synthetic nanocarriers with biologically derived membrane proteins, adhesion molecules, and immune-interactive signals. However, direct MM-specific evidence remains predominantly preclinical. CD38-targeted LNPs have enabled therapeutic RNA delivery to myeloma cells in the bone marrow, whereas lipid–polymer nanoparticles targeting bone marrow endothelium have delivered siRNA against vascular-niche pathways and reduced MM extravasation and colonization. 67 , 103 Cell membrane coating and EV-based delivery could further provide immune evasion, homotypic tumor recognition, myeloid-cell targeting, or carriage of niche-modulating RNA and proteins, but their clinical translation is constrained by low production yield, donor- and culture-dependent biological heterogeneity, incomplete or incorrectly oriented membrane coating, variable cargo loading, labor-intensive purification, limited storage stability, and high cost of goods. Membrane integrity is particularly important because differences in coating completeness can alter nanoparticle internalization and biological identity, while EV composition is sensitive to producer-cell source, passage number, culture medium, collection interval, and isolation method. 182–184 A practical route toward scalability is to develop these products as standardized manufacturing platforms rather than patient-specific biological preparations. Upstream production should use qualified master cell banks, serum-free or xeno-free media, predefined passage and harvesting windows, and closed perfusion, hollow-fiber, or stirred-tank bioreactors to reduce operator dependence and increase membrane or EV yield. For membrane-camouflaged nanoparticles, scalable core formation by continuous microfluidic or impingement-based mixing should be separated from a controlled membrane-fusion step, with fixed membrane-to-core ratios and validated shear conditions. Microfluidic manufacturing has already demonstrated its value for controlling nanoparticle size, composition, and batch reproducibility in RNA-delivery systems. 185 , 186 For EVs, a closed downstream process combining tangential-flow filtration for concentration with size-exclusion or ion-exchange chromatography for purification is more scalable and easier to standardize than repeated differential ultracentrifugation. Product development should follow a quality-by-design framework in which critical quality attributes include particle size and polydispersity, cargo loading, membrane coverage and orientation, source-cell protein fingerprints, EV identity and depletion markers, residual host-cell DNA and protein, endotoxin, sterility, complement activation, and storage stability. MISEV2023 provides a useful foundation for defining EV identity, purity, and reporting requirements, although therapeutic manufacturing requires additional validated potency and comparability assays. 187 , 188 For MM applications, release testing should extend beyond generic physicochemical characterization. Potency assays should determine whether each batch retains binding to BCMA-, CD38-, or other MM-associated targets; delivery to marrow endothelial, stromal, or myeloid compartments; hypoxia-responsive cargo release; macrophage-repolarizing or vascular-remodeling activity; and compatibility with CAR-T-cell viability and cytotoxicity. 67 , 103 , 189 To reduce production costs further, a “minimal biomimicry” strategy could replace highly heterogeneous whole-cell membranes with semi-synthetic vesicles containing only validated functional components, such as selected homing ligands, CD47-related immune-evasion signals, or MM-targeting proteins. Off-the-shelf erythrocyte-, platelet-, or clonal producer-cell-derived membranes may also be more scalable than autologous tumor-cell membranes, provided that donor screening, pathogen control, and batch comparability are established. 190 , 191 Thus, the most realistic translational pathway is a staged progression from standardized, off-the-shelf biomimetic carriers toward more complex patient-derived platforms only when their additional targeting benefit clearly justifies the increased manufacturing burden and cost. The central objective is not to reproduce every biological feature of the source cell, but to preserve the minimum functional membrane or EV attributes required for reproducible bone marrow targeting, niche remodeling, and improvement of CAR-T efficacy. 192 , 193

Finally, future research should follow a progressive strategy from lower risk to higher risk. The first step should promote ex vivo nanomanufacturing and short window antigen or niche pretreatment before CAR-T therapy. The second step should implement intervention before relapse guided by nanodiagnostics. The third step should advance toward in vivo CAR-T generation, multitarget nanodelivery, and combination therapy with CAR-NK cells, CAR-M cells, or CAR exosomes. At each step, real clinical benefit, toxicity reduction, cost reduction, and improved accessibility should serve as the final evaluation criteria.

Conclusion

Multiple myeloma has entered an era in which BCMA-targeted CAR-T therapy can induce deep responses, yet durable disease control remains limited by relapse, manufacturing barriers, and unequal access. Relapse is shaped by multiple interconnected mechanisms, including antigen remodeling, clonal evolution, CAR-T exhaustion, metabolic dysfunction, and bone marrow niche–mediated immunosuppression. These findings support a shift from improving CAR-T-cell potency in isolation toward coordinated optimization of target-antigen availability, engineered immune-cell fitness, and the bone marrow microenvironment. Within this framework, nanomedicine should be regarded as an enabling toolkit rather than a stand-alone solution, with potential applications in low-damage ex vivo manufacturing, controllable in vivo immune programming, antigen-density modulation, multitarget engineering, bone marrow niche remodeling, and longitudinal relapse monitoring. Nevertheless, most nano-enabled CAR-T strategies remain at the preclinical or early clinical stage, and their translation will depend on demonstrating cell-selective delivery, predictable CAR-expression kinetics, limited off-target biodistribution, manageable immunotoxicity, scalable GMP production, and durable safety. Accordingly, near-term development should prioritize lower-risk and clinically measurable applications, including ex vivo manufacturing optimization, short-window antigen or niche conditioning, and biomarker-guided relapse surveillance using MRD, ctDNA, soluble BCMA, extracellular vesicles, spatial imaging, and CAR-T pharmacokinetics. Progression toward in vivo CAR-T generation and complex multimodule combinations should be guided by comparative clinical studies, standardized product and safety criteria, long-term follow-up, and evidence of improved efficacy, reduced toxicity, shorter treatment delays, and greater cost-effectiveness and accessibility. Thus, the future value of nano-enabled CAR-T therapy in MM should be judged not by technological complexity alone, but by its ability to generate reproducible and clinically meaningful benefits across manufacturing, treatment, and relapse management in the post-BCMA era.