Targeting endosomal trafficking-mediated antigen escape to resensitize myeloma to CAR-T therapy

Abstract

Background

Antigen escape is one of the leading causes of relapse following chimeric antigen receptor (CAR)-T therapy, particularly in multiple myeloma. A critical gap persists in understanding the tumor-intrinsic pathways that trigger antigen loss, insight essential for devising strategies to resensitize tumors to immune attack. We identify a previously uncharacterized post-translational mechanism centered on the metabolic enzyme ribonucleotide reductase subunit M2 (RRM2), termed trafficking-mediated antigen escape, to enhance cellular therapy efficacy.

Methods

We combined single-cell RNA sequencing analysis with multiplex immunofluorescence to identify a clinically relevant RRM2⁺ myeloma subpopulation exhibiting low MICA/B abundance. Functional validation included induced pluripotent stem cell-derived myeloma organoids monitored by real-time imaging and disseminated xenograft models to assess the effect of subtoxic osalmid treatment on NKG2D CAR-T cell activity. Co-immunoprecipitation, guanosine 5′-triphosphate pulldown, and confocal microscopy were used to investigate the underlying trafficking mechanism.

Results

Single-cell analysis uncovered a clinically prevalent RRM2⁺ myeloma subpopulation with profoundly reduced MICA/B surface abundance, which established tumor-intrinsic heterogeneity as one of fundamental causes of NKG2D CAR-T resistance. We further demonstrated RRM2’s non-canonical role as a trafficking regulator that actively shuttles MICA/B toward lysosomal degradation via RAB7A activation while simultaneously blocking RAB11-mediated recycling. Therapeutic intervention using subtoxic osalmid, a clinically approved drug and previously characterized as an RRM2 inhibitor, successfully reversed this trafficking defect, restored MICA/B membrane presentation and synergized with NKG2D CAR-T cells to enhance their expansion, polyfunctional cytokine secretion, and stem-like properties. This combination strategy achieved durable tumor remission in vivo by sustaining T-cell fitness while reducing exhaustion, offering an immediately actionable solution to clinical antigen escape.

Conclusions

Our study establishes RRM2-driven trafficking as a novel and targetable mechanism of antigen escape in CAR-T therapy. By repurposing osalmid to restore MICA/B surface presentation, we provide a clinically translatable strategy that specifically potentiates NKG2D CAR-T cell efficacy in multiple myeloma and could potentially enhance the efficacy of CAR-T across diverse antigens. This work highlights the therapeutic potential of modulating intracellular trafficking to overcome resistance in cellular immunotherapy.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Antigen escape drives chimeric antigen receptor (CAR)-T relapse in multiple myeloma via genetic or transcriptional loss; however, the role of post-translational trafficking in non-genetic resistance remains poorly understood.

• Low surface density of NKG2D ligands (MICA/B) impairs CAR-T efficacy, but the regulatory mechanisms controlling their surface abundance are poorly defined.

WHAT THIS STUDY ADDS

• We identified a previously unrecognized “trafficking-mediated antigen escape” mechanism where ribonucleotide reductase subunit M2 shunts MICA/B toward lysosomal degradation while inhibiting membrane recycling.

• We demonstrate for the first time that the clinical agent osalmid reverses this process, restoring antigen surface expression and resensitizing myeloma cells to CAR-T therapy.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• By shifting the resistance paradigm from static genetic loss to dynamic protein trafficking, this study provides a clinically translatable combination strategy using osalmid to stabilize surface antigens and enhance the efficacy of CAR-T cell therapies in multiple myeloma.

Introduction

Chimeric antigen receptor (CAR) T-cell therapy has revolutionized the treatment of hematologic malignancies.1,3 However, CAR-T therapy faces a critical challenge where approximately 50% of patients experience relapse or drug resistance within 1–2 years, primarily due to antigen escape (eg, target antigen downregulation, mutation, or loss).4,7 Moreover, clinical trial results revealed that numerous patients with myeloma remain non-responsive after CAR-T treatment with inadequate antigen presentation, which could likely constrain therapeutic efficacy.^{8 9} Preclinical studies confirm that enhancing antigen expression could improve cell therapy efficacy.¹⁰ An emerging paradigm is that surface protein abundance is dynamically controlled by intracellular trafficking,^{11 12} a process whose impact extends far beyond transcriptional regulation.

Notwithstanding this broader paradigm, studies of the antigen targets have remained narrowly focused on transcriptional and post-transcriptional regulation, while the critical role of intracellular trafficking has been overlooked. Intracellular trafficking pathways, including degradation and endocytic recycling, could pivotally govern therapeutic target surface abundance.¹³ As key sorting hubs in intracellular trafficking, recycling endosomes could orchestrate cargo sorting and returning to the plasma membrane as well as determine the fate of internalized proteins including immune regulators, such as programmed death-ligand 1.13,15 Recent work also highlights CAR internalization and recycling as critical determinants of CAR-T efficacy in hematological malignancies.^{16 17} However, therapeutic strategies of modulating trafficking and recycling pathways to boost target antigen abundance remain unexplored.

Ribonucleotide reductase regulatory subunit M2 (RRM2), essential for deoxyribonucleotide synthesis, is frequently overexpressed in multiple tumors, including myeloma, and promotes tumor progression.^{18 19} Our prior work showed RRM2 drives myeloma progression via enhanced proliferation, apoptosis inhibition, and deterred autophagosome–lysosome fusion.²⁰ Notably, osalmid, a clinical choleretic drug and RRM2 inhibitor, could disrupt dNTP synthesis and DNA repair, thereby shows efficacy in myeloma.²¹ Beyond its canonical role, RRM2 inhibition exhibits immunomodulatory effects, such as enhancing programmed cell death protein-1 (PD-1) blockade efficacy and promoting M1 macrophage polarization in cancer.^{19 22 23} These immunomodulatory effects prompted us to investigate whether RRM2 could directly regulate the membrane trafficking of target antigens, thereby modulating tumor immunogenicity.

In this study, we identify RRM2 as the core of a trafficking-mediated antigen escape pathway. We demonstrate that RRM2 actively shuttles the NKG2D ligands MICA/B away from the cell surface and toward lysosomal degradation, while simultaneously inhibiting their recycling. By elucidating the complete RRM2/TBC1D15/RAB7A axis, we provide a mechanistic blueprint for this novel form of resistance. Furthermore, we show that pharmacological rewiring of this pathway with osalmid effectively reverses antigen escape and potently synergizes with NKG2D CAR-T cells. Our findings establish the therapeutic targeting of antigen trafficking as a viable and promising strategy to overcome immunotherapy resistance.

Methods

Cells and clinical samples

The ARP1 and NCI-H929 were kindly provided by Professor Wen Zhou from Central South University. U266 cell line was purchased from the American Type Culture Collection (ATCC, Manassas, USA). OCI-MY5 was kindly provided by Fenghuang Zhan (Department of Internal Medicine, University of Iowa, Iowa City, Iowa, USA). They were maintained in Roswell Park Memorial Institute (RPMI) 1640 medium (Gibco, Waltham, USA) containing 10% fetal bovine serum. The human HEK-293T embryonic kidney cell line was purchased from the ATCC and was routinely maintained in a complete growth medium, consisting of Dulbecco’s Modified Eagle Medium (Basalmedia, China) enriched with 10% fetal bovine serum (C04001-500, VivaCell, China), and antibiotics with 100 U/mL penicillin and 100 U/mL streptomycin (C0222, Beyotime, China) to support optimal cell growth and viability.

Animals and animal models

NOD.Cg-PrkdcscidIl2rgem1Hcem3(c.645InsTA)Smoc (M-NSG) mice (female, 6 weeks old) were purchased from Shanghai Model Organisms Center (China) and used for the in vivo xenograft studies. All animal experiments were approved by the Tongji University Institutional Animal Care and Use Committee and conducted in accordance with the relevant ethical guidelines for animal research. The sample size for each experimental group was determined based on preliminary data to ensure statistically reliable results. Mice were randomly assigned to the treatment or control groups using a computer-generated randomization sequence. Investigators responsible for tumor measurement and outcome assessment were blinded to the group allocation throughout the experiment.

Tumor model

To establish the intravenous xenograft model, ARP1 cells stably expressing luciferase (ARP1-luc) were harvested and resuspended in phosphate-buffered saline. A total of 1×10⁶ ARP1-luc cells in a 100 µL volume were intravenously injected via the tail vein of each mouse. Tumor engraftment was monitored weekly by bioluminescence imaging (BLT PHOTON TECHNOLOGY, China). The minimum sample size was calculated as n=5 per group. Inclusion: female M-NSG mice aged 6–8 weeks; tumor engraftment verified by bioluminescence ≥5×10⁵ p/s/cm/sr at day 12 post-injection; no animals or data points were excluded. Body weight ≥18 g. Once the tumors were established, mice were randomized into the following treatment groups by an investigator not involved in tumor measurement (five mice per cage): (1) T-cell group; (2) osalmid-treated group; (3) NKG2D CAR-T group; (4) osalmid combined with NKG2D CAR-T cells group. The T-cell products comprised both CAR-transduced and non-transduced human T cells due to submaximal transduction efficiency. For the combination therapy, mice received daily intraperitoneal injections of 15 mg/kg osalmid (526-18-1, MedChemExpress) starting from the day of tumor cell injection. A single dose of 5×10⁶ NKG2D CAR-T cells per mouse was administered intravenously on the designated day. Tumor burden was quantified weekly by measuring bioluminescence flux. For endpoint analysis, peripheral blood was collected for flow cytometric analysis of CAR-T cell expansion (using anti-human CD45 antibody), immune checkpoint markers including programmed cell death protein 1 (PD-1) and T-cell immunoglobulin and mucin-domain containing-3 (TIM-3), memory T-cell subsets, and MICA/B surface abundance on residual tumor cells. Serum was isolated for further analysis. Major organs (liver, kidney and spleen) were harvested for histopathological examination (H&E staining) and bone marrow was used for immunohistochemical analysis of CD138+ myeloma cell infiltration. We used an a priori power analysis to determine the number of mice required to detect a biologically meaningful difference in tumor bioluminescence (photon flux) between the NKG2D CAR-T monotherapy and the osalmid+CAR T combination groups. All intraperitoneal injections were given between 09:00 and 10:00; the order of cages was rotated daily (Latin-square design) so that no group was always first or last.

Organoid experiments

Human induced pluripotent stem cell (iPSC)-derived bone marrow organoids were generated as previously described.²⁴ Briefly, mesodermal aggregates were formed under 5% O_₂ in APEL2 medium (05275, STEMCELL Technologies) supplemented with BMP4 (HZ-1045, Proteintech), FGF2 (HZ-1285, Proteintech), and VEGFA (HZ-1038, Proteintech) (days 0–3). Aggregates were then committed to hematopoietic and vascular lineages (days 3–5) before embedding in hydrogels composed of 60% collagen I+IV mix and 40% Matrigel. Vascular sprouting was induced with VEGFA+VEGFC (HZ-1336, Proteintech) (25 ng/mL each). At day 12, individual vascularized organoids were transferred into 96-well ultra-low-attachment plates for downstream use. ARP1 cells were transduced with green fluorescent protein (GFP) plasmid and seeded at 5×10³ cells per organoid in 200 µL StemPro-34 medium supplemented with 10% human serum, IL-6 (HZ-1019, Proteintech) (25 ng/mL), SDF-1α (HY-P70469, MCE) (50 ng/mL), EPO (HZ-1168, Proteintech) (25 ng/mL), TPO (HZ-1248, Proteintech) (25 ng/mL), and G-CSF (300-23-50UG, Thermo Fisher) (25 ng/mL). Engraftment was confirmed by confocal microscopy 72 hours post-seeding. Subtoxic concentrations of osalmid were determined by 48 hours. CAR-T cells were labeled with CellTracker (MX4109-100UG, Maokangbio, China). Real-time imaging was performed on an Aicell Zero (Gnano) with 10×objective. Images were acquired every 30 min for 48 hours. For flow cytometry, organoids were dissociated with 200 U/mL collagenase I and IV (17104–019, Gibco) at 37°C for 15 min followed by gentle pipetting. Single-cell suspensions were stained with anti-CD4-AF700 antibody (300526, BioLegend), anti-CD45RA-PB antibody (304123, BioLegend) and anti-CCR7-APC (353214, BioLegend). Intracellular cytokines were detected using BD Cytofix/Cytoperm kit (554714, BD) and stained with anti-IFN-γ-PE antibody (383303, BioLegend) or anti-granzyme B-PE (372207, BioLegend). Data were acquired on a BD LSRFortessa X-20 and analyzed with FlowJo V.10.8.

GTP pulldown

The guanosine 5′-triphosphate (GTP) pull-down assay was conducted following a previously established method.²⁵ Specifically, 1×10⁷ ARP1 cells were seeded into culture vessels and treated with 6 µM osalmid or 12 µM osalmid. Subsequently, the cells were lysed using three cycles of freezing and thawing to ensure complete disruption of cellular structures. The lysate was then clarified by centrifugation at 12,000 rpm for 20 min at 4°C to remove insoluble debris. The resulting supernatant was transferred to a new 1.5 mL tube, and 100 µL of GTP-agarose resin (G9768, Sigma-Aldrich, Missouri, USA) was added to selectively bind GTP-bound proteins. After centrifugation, the beads were washed three times in binding buffer and resuspended in 40 µL of sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer. Proteins were boiled and immunoblotted.

CAR-T proliferation

To evaluate the proliferative capacity of NKG2D CAR-T cells against osalmid-pretreated myeloma cells, we established a co-culture system with optimized conditions. ARP1 cells were pretreated with 12 µM osalmid or or 0.1% dimethyl sulfoxide (DMSO) as a vehicle control for 48 hours in RPMI-1640 supplemented with 10% fetal bovine serum (FBS). Following treatment, NKG2D CAR-T cells were washed, counted, and labeled with CellTrace Violet (C34557, Thermo Fisher). NKG2D CAR-T cells were then co-cultured with pretreated myeloma targets at an effector-to-target (E:T) ratio of 1.25:1, 2.5:1 and 5:1 in 6-well plates containing X-VIVO 15 medium (02-060Q, Lonza, Switzerland) supplemented with 100 IU/mL interleukin (IL)-2. After 96 hours of co-culture, cells were harvested and stained with anti-CD3-PE antibody, anti-CD8-APC-Cy7 antibody (344714, BioLegend) and anti-CD4-AF700 antibody (300526, BioLegend) for flow cytometric analysis. CAR-T cell proliferation was assessed.

Co-immunoprecipitation

Cells were lysed in lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP-40, 0.25% sodium deoxycholate, 1 mM EDTA) supplemented with protease/phosphatase inhibitors. Lysates were pre-cleared with protein A/G agarose beads (P2055-10ml, Beyotime) for 1 hour at 4°C. Pre-cleared supernatants were incubated overnight at 4°C with primary antibodies (target-specific or species-matched IgG controls; final dilution 1:100). Immune complexes were captured using protein A/G beads at 4°C for 2 hours (36403ES, Yeasen, China), washed four times with lysis buffer, and eluted in 2×loading buffer at 95°C for 10 min. Eluates were resolved by SDS-PAGE and analyzed via western blotting.

RNA sequencing

ARP1 cells were pretreated with 12 µM osalmid for 48 hours. Total RNA was isolated from both control and treated groups using TRIzol reagent (15596018CN, Invitrogen), followed by purification with the RNeasy Mini Kit (Qiagen, Germany). RNA integrity was confirmed with an Agilent 2100 Bioanalyzer (RNA integrity number ≥8.0). Sequencing was carried out on an Illumina NovaSeq 6000 platform (USA) to generate approximately 40 million paired-end reads per sample. Reads were aligned to the GRCh38 reference genome using STAR, and gene expression was quantified with featureCounts. Differential expression analysis was performed using DESeq2, with an adjusted p value (false discovery rate) threshold of <0.05. Biological triplicates were included for all conditions.

Statistical analysis

Data were analyzed using GraphPad Prism V.9.0 (GraphPad Software) and R V.4.3.1. Comparisons between two groups were analyzed using a two-tailed unpaired Student’s t-test. Comparisons among multiple groups were performed using one-way or two-way analysis of variance, as appropriate, followed by Tukey’s post hoc test. Correlations were evaluated by Pearson’s correlation analysis. In addition, differences in survival rates were analyzed using Kaplan-Meier survival curves, and statistical significance was determined by the log-rank test. Data are presented as mean±SE unless stated otherwise. Statistical significance was defined as p<0.05. Figures were generated in Prism. Sample sizes (n) represent biological replicates, detailed in figure legends.

Results

RRM2 elevation correlates with impaired MICA/B surface abundance in myeloma

It was previously thought that NKG2D ligands (NKG2DLs), including MICA/B and ULBP1-6, are generally expressed in MM.^{26 27} Indeed, our results revealed these ligands are expressed in myeloma (figure 1A). In addition, our results revealed marked interpatient heterogeneity in MM, which potentially contributes to poor treatment outcomes (figure 1B, online supplemental figure S1A–E). Stratifying samples by Δmean fluorescence intensity (MFI of specific antibody minus matched isotype control) showed most patients belonged to the low MICA/B group (figure 1B). However, this intrapatient heterogeneity of NKG2DL is often overlooked, particularly the existence of subclones with low antigen levels. We identified that a significant proportion of patient samples harbor myeloma subclones with MICA/B abundance as low as isotype controls, which likely serve as a reservoir for resistance to NKG2DL-targeted therapies. Furthermore, single-cell analysis of myeloma samples proved marked intrapatient heterogeneity in patients with myeloma, with substantial subclones of myeloma cells exhibiting low expression of MICA and MICB (online supplemental figure S1F). Building on previous research implicating RRM2 in myeloma pathogenesis, we observed an inverse correlation between RRM2 expression and transcript levels of MICA, MICB, and ULBP3 from RNA microarray data (figure 1C and online supplemental figure S1G-I). Furthermore, flow cytometry of primary samples confirmed high intracellular RRM2 was negatively associated with MICA/B surface abundance, while showing no significant association with other therapeutic targets (eg, CD138, BCMA), underscoring the specificity of the RRM2-MICA/B axis (online supplemental figure S1J,K). Multiplex immunohistochemical staining of bone-marrow biopsies further validated a significant negative correlation between the mean RRM2 fluorescence intensity and the mean MICA/B fluorescence intensity (figure 1E,F).

Figure 1. RRM2 elevation correlates with impaired MICA/B surface localization in myeloma. (A) Single-cell sequencing analysis of NKG2DL expression in plasma cells from patients with multiple myeloma and healthy individuals. (B) Flow cytometry analysis of membrane MICA/B expression on the surface of patients with multiple myeloma cells. High MICA/B abundance was defined as ΔMFI>twofold and low MICA/B abundance defined as ΔMFI≤twofold relative to isotype control, respectively. (C) Box plots comparing MICA and MICB mRNA expression levels between RRM2-low and RRM2-high groups in microarray data from patients with multiple myeloma. (D) Flow cytometry detection of intracellular RRM2 and membrane MICA/B expression in patient multiple myeloma cells. Correlation analysis was performed to assess the relationship between RRM2 expression and membrane expression of MICA/B. (E) Multiplex immunofluorescence shows the abundance of RRM2 and MICA/B in myeloma cells within bone marrow samples from patients. Scale bar: 2 µm. (F) Correlation analysis of multiplex immunofluorescence between mean fluorescence intensity of RRM2 and MICA/B signals in CD138⁺ plasma cells across n=15 patients with newly diagnosed MM samples in E. Each dot represents one sample. (G) UMAP plots showing distribution of RRM2⁺ myeloma cells in the myeloma subset. (H) Bubble plot showing increased signaling communications between RRM2⁺/RRM2⁻ myeloma cells and NK/T cells. (I) Heatmap showing outgoing signaling patterns of RRM2⁺ myeloma cells and RRM2⁻ myeloma cells. (J) Cell–cell communication interaction among RRM2⁺ myeloma cells, RRM2⁻ myeloma cells and immune cells. (K) Hierarchical clustering of different immune cell proportions in RRM-high and RRM2-low groups. (L) Gene expression profiling of T cells (upper panel) and NK cells (lower panel) identified between RRM-high and RRM2-low groups. (M) Flow cytometry analysis of MICA/B abundance on the cell membrane of ARP1 cells from MM xenograft. (N) Statistical analysis of MICA/B MFI after osalmid treatment in M. Data are representative of three independent experiments and are given as mean±SE. A two-tailed unpaired t-test or a non-parametric Mann-Whitney U test was used for comparisons between two groups; ****p<0.0001, **p<0.01. MFI, mean fluorescence intensity; MM, multiple myeloma; mRNA, messenger RNA; NKG2DL, NKG2D ligands; NK, natural killer; RRM2, ribonucleotide reductase subunit M2; UMAP, uniform manifold approximation and projection.

To delineate the cellular basis, single-cell RNA sequencing analysis of bone marrow aspirates uncovered distinct RRM2⁺ malignant clones (figure 1G, online supplemental figure S1L,M). Complementarily, CellChat-based mapping revealed that these RRM2⁺ myeloma cells establish predominantly inhibitory interactions with natural killer (NK)/T cells via MIF-(CD74+CXCR4), MIF-(CD74+CD44) and ADGRE5-CD55 (figure 1H and J). In addition, the CellChat-derived outgoing signaling map indicated that RRM2⁺ myeloma cells actively secrete multiple suppressive ligands that engage cognate receptors on NK, CD8⁺ cytotoxic T, CD4⁺ helper T and other immune cells (figure 1I). Specifically, the inhibitory axis is mediated by major histocompatibility complex-I (HLA-E/G) that could silence NK-cell cytotoxicity. The suppression is driven by MIF binding to CD74-CXCR4 and CD74-CD44 on T cells and macrophages, and by ADGRE5–CD55 interactions,28,30 which could expand regulatory T cells (figure 1I). Both NK and T cells exhibit enhanced PPIA–BSG interactions with RRM2⁺ tumor cells, associated with tumor progression and poor clinical outcomes (online supplemental figure S1N). In contrast, NK cells engage in IFN-IFNG signaling with RRM2⁻ cells, suggesting that RRM2⁻ tumor cells are more susceptible to NK cell-mediated cytotoxicity (online supplemental figure S1N). Transcriptomic profiling showed RRM2⁺ cells enrich pathways in DNA replication and protein transport regulation (online supplemental figure S1O). In addition, when patients were dichotomized based on RRM2 expression, the RRM2-high group exhibited decreased infiltration of NK cells and memory CD4⁺ and CD8⁺ T cells (figure 1K). Notably, within the RRM2-high group, NK cells displayed suppressed cytotoxicity signatures, including granzyme A (GZMA), granzyme B (GZMB) and granzyme H (GZMH), whereas T cells showed elevated exhaustion markers (PDCD1 and CTLA4) with concomitant reduction in memory-associated transcripts (CCR7 and SELL) (figure 1L, online supplemental figure S2A,B). Collectively, RRM2 drives immune escape by downregulating NKG2DLs and activating immunosuppressive pathways.

To validate the roles of RRM2, subcutaneous myeloma xenografts were treated with the RRM2 inhibitor osalmid. Notably, osalmid-treated tumors exhibited significantly increased total and membrane-localized MICA/B (figure 1M,N). Taken together, these results indicate that RRM2 regulates MICA/B membrane abundance, thereby influencing the efficacy of cell therapy.

Low-dose RRM2 inhibitor osalmid enhances MICA/B surface abundance in myeloma cells

To determine whether RRM2 inhibition could selectively enhance NKG2DL (MICA/B) expression in myeloma cells without inducing cytotoxicity, we first investigated the dose-dependent effects of the RRM2 inhibitor osalmid. Initial cytotoxicity screening revealed that high concentrations of osalmid (≥50 µM) induced significant myeloma cell death.¹⁹ Building on our previous work, 12.5 µM osalmid was first confirmed to leave myeloma cell apoptosis and proliferation unaffected (online supplemental figure S3A–C). To isolate the effect on MICA/B modulation independent of cytotoxicity, we applied subtoxic doses of osalmid (1.5–12 µM). Treatment of several multiple myeloma cell lines (ARP1, NCI-H929, U266 and OCI-MY5) with osalmid for 48 hours resulted in a dose-dependent upregulation of membrane MICA/B abundance, as quantified via flow cytometry. Following osalmid treatment for 48 hours, flow cytometry revealed a significant increase in membrane MICA/B localization (figure 2A–D and online supplemental figure S4A–H). Notably, osalmid also enhanced MICA/B membrane abundance in primary myeloma cells from patients with both low membrane abundance and high membrane abundance groups (figure 2E). Immunofluorescence staining further confirmed increased MICA/B abundance and localization on the cell membrane (figure 2F and G). These findings demonstrate that low-dose osalmid enhances MICA/B membrane surface localization without leading to cytotoxicity. Analysis of membrane and cytoplasmic protein fractions revealed that low-dose osalmid markedly increased MICA and MICB levels in both cellular compartments of myeloma cells (figure 2H and I). Collectively, these results demonstrate that subtoxic RRM2 inhibition selectively restores MICA/B on membrane and cytoplasm, functionally distinct from the established cytotoxic effects of osalmid observed at higher doses.

Figure 2. Low-dose RRM2 inhibitor osalmid induces upregulation of MICA/B on the cell membrane in multiple myeloma cells. (A) Flow cytometry analysis of MICA/B expression on the cell membrane of ARP1 cells induced by different concentrations of osalmid. Upper panel: representative flow cytometry plots. Lower panel: corresponding bar chart. (B) Flow cytometry analysis of MICA/B expression on the cell membrane of NCI-H929 cells induced by different concentrations of osalmid. Upper panel: representative flow cytometry plots. Lower panel: corresponding bar chart. (C) Flow cytometry analysis of MICA/B expression on the cell membrane of U266 cells induced by different concentrations of osalmid. Upper panel: representative flow cytometry plots. Lower panel: corresponding bar chart. (D) Flow cytometry analysis of MICA/B expression on the cell membrane of OCI-MY5 cells induced by different concentrations of osalmid. Upper panel: representative flow cytometry plots. Lower panel: corresponding bar chart. (E) Membrane expression of MICA/B in patient-derived multiple myeloma cells with high and low baseline MICA/B expression after in vitro induction with osalmid. (F) Immunofluorescence staining demonstrating significant upregulation of MICA/B on the cell membrane of multiple myeloma cells after 48 hours treatment with DMSO or osalmid. Scale bar: 2 µm. (G) Quantification of (left) MFI of MICA/B and (right) membrane-localized MICA/B fraction in DMSO-treated versus osalmid-treated ARP1 cells (n=10) in F. (H–I) Western blotting analysis of MICA and MICB abundance on the cell membrane as well as cytoplasm of ARP1 cells induced by osalmid at a low dose. Data are representative of three independent experiments and are given as mean±SE. A one-way ANOVA with Tukey’s post hoc test was used for multiple group comparisons, while a two-tailed unpaired t-test was used for comparisons between two groups; ****p<0.0001, **p<0.01. ANOVA, analysis of variance; DMSO, dimethyl sulfoxide; MFI, mean fluorescence intensity; RRM2, ribonucleotide reductase subunit M2.

Low-dose osalmid potentiates NKG2D CAR-T cell cytotoxicity against myeloma

Early clinical trials evaluating NKG2D CAR-T cells in patients with MM revealed limited antitumor activity.⁸ Consistent with these findings, bone marrow samples of enrolled patients with MM exhibited low surface abundance of MICA and MICB on the membrane of MM cells, providing a plausible mechanistic basis for the suboptimal therapeutic efficacy.⁹ Given the mechanistic relation between MICA/B surface density and CAR-T efficacy, we hypothesized that osalmid-induced ligand upregulation could overcome antigen escape barriers. To verify this, a second-generation NKG2D CAR (4-1BB/CD3ζ) was engineered for optimized effector function, followed by assessment of osalmid-enhanced cytotoxicity against myeloma cells (figure 3A). Osalmid-pretreated myeloma cells markedly improved NKG2D CAR-T cytotoxicity against ARP1 and NCI-H929 cells, which stably expressed GFP-Luciferase, across multiple E:T ratios, as assessed by flow cytometry (figure 3B–E) and luciferase-based killing assays (figure 3F–G). Besides, this concentration does not deter CAR-T cell functionality (online supplemental figure S5A,B). Osalmid pretreatment also enhanced susceptibility of primary patient myeloma cells to CAR-T killing (figure 3H). To determine whether osalmid enhances cytotoxicity solely by increasing MICA/B membrane abundance, we established MICA, MICB, and MICA/B knockdown ARP1 cells (online supplemental figure S5C,D). Genetic ablation of MICA and MICB in myeloma cells could completely abrogate osalmid’s enhancement of NKG2D CAR-T cytotoxicity, proving that the potentiation is exclusively dependent on osalmid-induced MICA/B membrane upregulation (figure 3I).

Figure 3. Low-dose osalmid enhances NKG2D CAR-T cytotoxicity in multiple myeloma cells. (A) Schematic graph of the second generation NKG2D CAR structure with NKG2D/CD8-TM/41BB/CD3ζ. (B–E) Flow cytometry analysis of the relative proportion of NKG2D CAR-T cells co-cultured with osalmid-induced ARP1 and NCI-H929 cells at different E:T ratios. (F–G) Luciferase-expressing tumor cells were generated to determine the relative cytotoxicity of NKG2D CAR-T cells co-cultured with osalmid-induced ARP1 and NCI-H929 cells at different E:T ratios by measuring luciferase activity. (H) Flow cytometry analysis of the relative proportion of NKG2D CAR-T cells co-cultured with patient-derived samples pretreated with osalmid for 1 day. (I) Luciferase-expressing tumor cells were generated to determine the relative cytotoxicity of NKG2D CAR-T cells co-cultured with ARP1 cells or osalmid-induced ARP1 cells, MICA knock-down ARP1 cells, MICB knock-down ARP1 cells, or MICA/B double knock-down ARP1 cells at E:T ratio (1.25:1) by measuring luciferase activity. (J–K) Flow cytometry analysis of CD4⁺ and CD8⁺ T-cell proliferation in co-cultures of NKG2D CAR-T cells with osalmid-induced ARP1 cells at different E:T ratios. (L) Cytokine release in the supernatant of co-cultures of NKG2D CAR-T cells with osalmid-induced ARP1 cells at different E:T ratios. (M) Cytokine release in the supernatant of co-cultures of NKG2D CAR-T cells with osalmid-induced NCI-H929 cells at different E:T ratios. Data are representative of three independent experiments and are given as mean±SE. A one-way ANOVA with Tukey’s post hoc test was used for multiple group comparisons, while a two-tailed unpaired t-test or a two-way ANOVA was used for comparisons between two groups. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05. ANOVA, analysis of variance; CAR, chimeric antigen receptor; E:T, effector-to-target.

Given that CAR sensitivity to low antigen density is profoundly dictated by its structural architecture.³¹ We investigated whether enhancing receptor sensitivity could circumvent the necessity for osalmid-mediated MICA/B upregulation. Following the engineering of two CD28-based constructs (NKG2D-CD28ζ and NKG2D-CD28-41BBζ), their efficacy was evaluated against OCI-MY5 cells characterized by low baseline MICA/B expression (online supplemental figure S6A). Although the CD28-based CARs demonstrated superior basal recognition compared with the 4-1BB-based design, osalmid treatment provided a further significant boost to their cytotoxic efficacy (online supplemental figure S6B,C). These observations underscore that increasing surface antigen density yields a functional advantage that is not fully redundant with receptor engineering. Instead, the findings suggest a potent synergism between pharmacological antigen modulation and high-sensitivity CAR designs in overcoming immune evasion.

We further investigated the effect of osalmid-induced MM cell surface MICA/B abundance on CAR-T cell function. Osalmid-pretreated targets enhanced the proliferation of both CD4⁺ and CD8⁺ NKG2D CAR-T cell subsets (figure 3J and K). In addition, osalmid pretreatment of MM cells enhanced secretion of functionally distinct effector cytokines: cytolytic (perforin), pro-inflammatory (tumor necrosis factor (TNF)-α, interferon (IFN)-γ), and persistence-associated (IL-6) cytokines, collectively potentiating antitumor efficacy (figure 3L–M). To determine whether the elevated cytokine levels observed in the co-culture supernatants were a result of an enhancement of individual cell effector function, we performed intracellular cytokine staining. Flow cytometric analysis revealed that osalmid pretreatment significantly increased the frequencies of IFN-γ, TNF-α, IL-6 and perforin positive populations within the CAR-T cell compartment (online supplemental figure S7A–F). Significantly, the upregulation of MICA/B on tumor cells correlated with enhanced recognition and killing by NKG2D CAR-T cells. Moreover, the proliferation and cytokine secretion profiles indicated that osalmid not only enhances the cytotoxic potential but also supports the sustained activity of CAR-T cells. These results collectively demonstrate that osalmid can effectively enhance the antitumor efficacy of NKG2D CAR-T cells by upregulating MICA/B membrane abundance in MM.

Osalmid enhances CAR-T efficacy in myeloma organoids

To further model the tumor microenvironment and evaluate the immunomodulatory potential of osalmid, we established three-dimensional iPSC-induced myeloma organoids (figure 4A and online supplemental figure S8A). Consistent with our previous findings, low-dose osalmid (12 µM) treatment significantly increased the surface abundance of MICA/B on myeloma organoid MM cells, as confirmed by flow cytometry (figure 4B and C). Real-time live-cell imaging further demonstrated enhanced NKG2D CAR-T cell function in osalmid-pretreated organoids. At an E:T ratio of 1:1, osalmid pretreatment promoted CAR-T infiltration and markedly increased tumor-killing capacity, leading to rapid infiltration (figure 4D and E, online supplemental figure S8B–D and online supplemental video S1, online supplemental video S2). Even at a lower E:T ratio of 1:3, where CAR-T-mediated cytotoxicity was limited, osalmid still facilitated T-cell recruitment and sustained persistence within the tumor microenvironment (figure 4D and E, online supplemental figure S8B,C and online supplemental video S3, online supplemental video S4). These findings suggest that osalmid not only enhances immediate cytotoxicity under sufficient CAR-T availability but also improves T-cell infiltration and maintenance under suboptimal conditions. This enhanced response was associated with increased activation of CAR-T cells, indicated by a marked expansion of IFN-γ–positive and granzyme B–positive populations (figure 4F, online supplemental figure S8E,F). Furthermore, after extended co-culture, osalmid promoted the expansion of a stem-like T-cell population characterized by CD45RA⁺CCR7⁺ expression, suggesting a potential role in sustaining long-term T-cell activity and antitumor immunity (figure 4G).

Figure 4. Low dose osalmid enhances NKG2D CAR-T efficacy in a myeloma organoid model. (A) Schematic of bone marrow organoid generation and NKG2D CAR-T antitumor activity assessment. (B) Flow cytometry analysis of MICA/B surface expression on myeloma cells after osalmid treatment. (C) Quantification of MICA/B MFI from B. (D) Left panel: live imaging of organoids co-cultured with NKG2D CAR-T cells at E:T=1:1 over time. Right panel: a representative image from the 36-hour time point. Red: CellTrace-stained CAR-T cells; green: GFP-transfected ARP1 cells. Scale bars: 100 µm. (E) Left panel: live imaging of organoids co-cultured with NKG2D CAR-T cells at E:T=1:3 over time. Right panel: a representative image from the 36-hour time point. Red: CellTrace-stained CAR-T cells; green: GFP-transfected ARP1 cells. Scale bars: 100 µm. (F) Flow cytometry analysis of granzyme B and IFN-γ in CD8⁺ NKG2D CAR-T cells of myeloma organoid models at E:T=1:3. (G) Flow cytometry analysis of T_SCM, T_CM, T_EM, and T_E subsets within CD4⁺ and CD8⁺ NKG2D CAR-T cells after 7-day co-culture with organoids. Data are representative of three independent experiments and are given as mean±SE. A two-tailed unpaired t-test was used for two-group comparisons, and a two-way ANOVA was used for analyzing data with two independent factors; ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05. ANOVA, analysis of variance; CAR, chimeric antigen receptor; E:T, effector-to-target; GFP, green fluorescent protein; IFN, interferon; iPSC, induced pluripotent stem cell; MFI, mean fluorescence intensity; T_CM, central memory T cell; T_E, effector T cell; T_EM, effector memory T cell; T_SCM, stem cell-like T cell.

Low-dose osalmid synergizes with NKG2D CAR-T therapy in vivo

To further evaluate the therapeutic efficacy of low-dose osalmid in combination with NKG2D CAR-T cells in in vivo MM models, NSG mice bearing ARP1-luciferase xenografts were used and treated with low-dose osalmid each day after MM injection (15 mg/kg). 5×10⁶ cells NKG2D CAR-T cells/mice were intravenously injected 12 days after MM injection (figure 5A). The combination of low-dose osalmid with NKG2D CAR-T cells significantly reduced tumor bioluminescence and markedly prolonged survival compared with single NKG2D CAR-T treatment (figure 5B–E).

Figure 5. Validation of low-dose osalmid-mediated enhancement of antitumor efficacy in vivo. (A) ARP1-luciferase xenograft model in NSG mice. Mice received 1×10⁶ ARP1-luciferase cells subcutaneously. On day 12, mice in the CAR-T and CAR-T+osalmid groups were administered 5×10⁶ NKG2D CAR-T cells intravenously via the tail vein. Osalmid and CAR-T+osalmid groups received 15 mg/kg osalmid daily. Tumor growth was monitored via bioluminescence imaging every 7 days post-CAR-T injection. Red arrows indicate osalmid administration. (B) Bioluminescent images depicting tumor progression across the timeline. (C–D) Quantitative analysis of bioluminescence intensity of different groups. (E) Survival analysis of the animal model in different groups. (F) Quantification of serum cytokines (TNF-α, IFN-γ, perforin and IL-6) in mice 2 weeks after CAR-T cell injection. Values below the lower limit of detection of the assay kit were assigned a value of zero. (G) Schematic diagram of the collection of mouse peripheral blood and tissues/organs for detection in the ARP1-luciferase xenograft model. Red arrows indicate osalmid administration. (H) Flow cytometry analysis of the frequency of NKG2D positive CAR-T cells in peripheral blood. (I) Flow cytometry analysis quantifying MICA/B fluorescence on myeloma cell membranes in peripheral blood. (J) Flow cytometry analysis of exhaustion markers PD-1 and Tim-3 on CD4⁺ and CD8⁺ NKG2D CAR-T cells in peripheral blood. (K) Flow cytometry analysis of stem cell-like T_SCM, T_CM, T_EM, and T_E subsets within CD4⁺ and CD8⁺ NKG2D CAR-T cells. (L) Flow cytometry analysis of CD69 fluorescence on CD4⁺ and CD8⁺ NKG2D CAR-T cells. (M) H&E staining of liver and kidney tissues from different groups. Scale bar: 50 µm. (N) Immunohistochemical detection of CD138-positive cells in the bone marrow from different groups. Scale bar: 12.5 µm. Data are representative of three independent experiments and are given as mean±SE. A one-way ANOVA with Tukey’s post hoc test was used for multiple group comparisons, while a two-tailed unpaired t-test was used for comparisons between two groups; ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05. ANOVA, analysis of variance; CAR, chimeric antigen receptor; IFN, interferon; IL, interleukin; i.v., intravenous; PD-1, programmed cell death protein-1; T_CM, central memory T cell; T_E, effector T cell; T_EM, effector memory T cell; Tim-3, T cell immunoglobulin and mucin-domain containing-3; TNF, tumor necrosis factor; T_SCM, stem cell-like T cell.

By collecting serum after 1 week treatment, CAR-T expansion and cytokine release (TNF-α, IFN-γ, IL-6 and perforin) were elevated in osalmid-treated mice (figure 5F). Flow cytometry showed a significant increase in NKG2D CAR-T expansion and enhanced MICA/B abundance in MM of peripheral blood (figure 5G–I). In addition, flow cytometric analysis of peripheral blood revealed that osalmid administration significantly increased the proportion of NKG2D CAR-T cells, reduced exhaustion markers including PD-1 in both CD4⁺ and CD8⁺ NKG2D CAR-T cells, while a reduction in Tim-3 was observed but did not reach statistical significance, and markedly increased the proportion of stem-like T cells (T_SCM) within the NKG2D CAR-T cell population compared with the group receiving NKG2D CAR-T alone (figure 5J and K). Furthermore, osalmid-treated NKG2D CAR-T cells demonstrated significantly enhanced effector function by detecting MFI of CD69 (figure 5L). Compared with the NKG2D CAR-T treated group, H&E staining of the NKG2D CAR-T and osalmid treatment group demonstrated decreased tumor infiltration in liver and kidney (figure 5M, online supplemental figure S9A,B). Immunohistochemical analysis of bone marrow and spleen sections revealed a significantly greater reduction of CD138⁺ myeloma cells in mice treated with osalmid and NKG2D CAR-T cells compared with treated with NKG2D CAR-T alone (figure 5N, online supplemental figure S9C–E). These findings indicate that osalmid enhances the antitumor efficacy and fitness of NKG2D CAR-T cells in vivo, providing a promising therapeutic strategy for MM.

Osalmid promotes MICA/B membrane trafficking via RRM2 inhibition

We further elucidated the mechanism by which osalmid enhances MICA/B membrane abundance and evaluated its impact on the efficacy of NKG2D CAR-T cells. Low-dose osalmid does not affect MICA and MICB RNA transcription in MM cells (online supplemental figure S10A, online supplemental table S1), but significantly alters the total MICA and MICB protein levels in MM cells (figure 6A, online supplemental figure S10B). However, it does not affect the overall ubiquitination levels (online supplemental figure S10C), suggesting that the enhanced expression is not mediated by ubiquitination-related degradation processes. ARP1 cells were induced with low dose osalmid, and RNA sequencing analysis was performed (figure 6B). Transcriptomic profiling revealed high expression of genes compared with control were significantly enriched in regulation of protein localization, transport, and protein localization to plasma membrane (figure 6B and C, online supplemental figure S10D,E). Gene Set Enrichment Analysis results revealed that the osalmid-induced group was significantly enriched in pathways related to protein localization and protein translocation to the plasma membrane (figure 6C, online supplemental figure S10F,G). Based on the preceding results, we hypothesized that osalmid administration affects the translocation of MICA/B to the plasma membrane, potentially through mechanisms involving recycling endosomes and protein relocation. Confocal microscopy immunofluorescence results revealed that osalmid treatment significantly reduced intracellular co-localization of the lysosomal marker LAMP1 with MICA/B, suppressing lysosomal degradation of MICA/B (figure 6D–F). LAMP1 is a heavily glycosylated membrane protein that lines the lysosomal, serving as the canonical late-endosome/lysosome marker. In addition, osalmid induction markedly increased co-localization between MICA/B and RAB11,³² a small GTPase that drives receptor recycling from recycling endosomes back to the plasma membrane (figure 6G–I). Intriguingly, RAB11 exhibited significantly increased fluorescence intensity in MM cells on osalmid treatment, suggesting that osalmid may promote RAB11 expression and enhance endosomal recycling. In contrast, no significant alteration was observed in co-localization of the early endosome marker RAB5 with MICA/B (figure 6J–L). Therefore, the above results demonstrate that osalmid inhibits the lysosomal degradation of MICA/B proteins and enhances their endosomal recycling to the plasma membrane.

Figure 6. Osalmid induces MICA/B expression on the cell membrane of multiple myeloma cells by inhibiting RRM2, reducing lysosomal degradation of MICA/B, and promoting endosomal recycling pathways. (A) Western blotting analysis of MICA/B protein levels in multiple myeloma cells induced by low-dose osalmid. (B) Bubble plot analysis of RNA sequencing data for pathway enrichment in membrane-related processes following low-dose osalmid treatment. (C) RNA sequencing data analysis of pathways involved in membrane-related processes following low-dose osalmid treatment. (D) Immunofluorescence staining showing reduced co-localization of MICA/B with the lysosomal marker LAMP1. Scale bar: 2 µm. (E) Upper panel: intensity profiles of MICA/B (red) and LAMP1 (green) in control cells. Peaks indicate co-localization at various distances. Lower panel: intensity profiles in osalmid-treated cells, showing co-localization patterns but with altered peak intensities and distributions. (F) Quantitative analysis of the co-localization between MICA/B and LAMP1. (G) Immunofluorescence staining showing increased co-localization of MICA/B with RAB11. Scale bar: 2 µm. (H) Upper panel: intensity profiles of MICA/B (red) and RAB11 (green) in control cells. Peaks indicate co-localization at various distances. Lower panel: intensity profiles in osalmid-treated cells, showing co-localization patterns but with altered peak intensities and distributions. (I) Quantitative analysis of the co-localization between MICA/B and RAB11. (J) Immunofluorescence staining showing co-localization of MICA/B with RAB5. Scale bar: 2 µm. (K) Upper panel: intensity profiles of MICA/B (red) and RAB5 (green) in control cells. Peaks indicate co-localization at various distances. Lower panel: intensity profiles in osalmid-treated cells, showing co-localization patterns but with altered peak intensities and distributions. (L) Quantitative analysis of the co-localization between MICA/B and RAB5. Data are representative of three independent experiments and are given as mean±SE. A two-tailed unpaired t-test was used for comparisons between two groups; ****p<0.0001, ***p<0.001. RRM2, ribonucleotide reductase subunit M2.

The RRM2/TBC1D15/RAB7A axis controls MICA/B membrane abundance

To elucidate the molecular mechanism underlying RRM2-mediated regulation of MICA/B abundance and its pharmacological modulation by osalmid, we conducted systematic analyses of RRM2’s interactome and downstream signaling pathways. Building on our observation that RRM2 knockdown recapitulates the MICA/B upregulation pattern induced by pharmacological RRM2 inhibition (figure 7A–C). Proteomic profiling of RRM2-interacting partners through co-immunoprecipitation (Co-IP) sequencing identified RAB7A, a core regulator of endo-lysosomal sorting, as a novel binding partner (online supplemental table S2). RAB7A is a small GTPase that orchestrates late endosome–lysosome fusion and cargo degradation, whose active GTP-bound form is essential for lysosomal trafficking.³³ Reciprocal Co-IP assays confirmed direct RRM2-RAB7A interaction in MM cells (figure 7D). Our results verified whether both osalmid treatment and RRM2 knockdown could suppress the activity of RAB7A (RAB7A-GTP). Notably, using GTP-agarose pull-down assays, we demonstrated that RRM2 sustains the GTP-bound (active) state of RAB7A, a mechanism critical for its role in endo-lysosomal trafficking regulation (figure 7E). These findings indicate that RRM2 modulates RAB7A activity by maintaining its GTP-bound conformation, while osalmid disrupts this interaction to inactivate RAB7A, thereby reducing RAB7A GTP-bound state and impairing lysosomal trafficking. Previous studies have shown that TBC1D15 can directly influence RAB7A activity and promote lysosomal degradation.³⁴ Our results showed osalmid treatment and RRM2 knockdown could promote TBC1D15-RAB7A interaction (figure 7F–H), which could impair MICA and MICB lysosomal trafficking. MICA/B knockdown abolished osalmid-enhanced CAR-T killing, while RAB7A knockdown further potentiated cytotoxicity by MICA/B upregulation (figure 7I and J, online supplemental figure S11A). Complementarily, RRM2 inhibition significantly upregulated RAB11, a small GTPase that drives endosomal recycling to the plasma membrane. Immunoblotting and GTP pull-down assays confirmed elevated RAB11 protein and increased RAB11-GTP (online supplemental figure S11B–D), collectively promoting MICA/B recycling to the cell surface. To further validate the dual regulatory roles of RAB7A and RAB11 in MICA/B trafficking, we performed complementary genetic rescue and overexpression experiments. While knockdown of RAB7A (RAB7A-sh) alone increased MICA/B surface levels by limiting lysosomal degradation, the simultaneous overexpression of RAB11 (RAB7A-sh+RAB11 OE) led to a maximal synergistic increase in surface ligand density (online supplemental figure S12). Consistent with the increased ligand density, cells with dual RAB7A/RAB11 manipulation (RAB7A-sh+RAB11 OE) exhibited significantly enhanced sensitivity to NKG2D CAR-T mediated lysis in vitro (online supplemental figure S13A). In contrast, cells with RAB7A OE+shRAB11 significantly attenuated susceptibility to CAR-T cell lysis (online supplemental figure S13B). This dual control of RAB7A-mediated degradation and RAB11-mediated recycling provides insight into how RRM2 and its inhibitor osalmid dictate MICA/B membrane abundance and tumor cell recognition by NKG2D CAR-T cells (figure 8).

Figure 7. Osalmid promotes NKG2D CAR-T cytotoxicity by upregulating MICA/B surface expression via the RRM2/TBC1D15/RAB7A pathway. (A) Representative western blotting analysis of MICA/B expression following RRM2 knockdown. (B) Flow cytometric validation of MICA/B expression after RRM2 knockdown. (C) Immunofluorescence analysis of MICA/B distribution in tumor cells post-RRM2 knockdown. Scale bar: 2 µm. (D) Cell lysates were immunoprecipitated with anti-RRM2 antibody or control IgG, followed by immunoblotting with anti-RAB7A antibody (left). Conversely, cell lysates were immunoprecipitated with anti-RAB7A antibody or control IgG, followed by immunoblotting with anti-RRM2 antibody (right). Input lanes represent 5% of total cell lysate. (E) GTP pulldown assays showing reduced RAB7A-GTP activity on osalmid induction. (F) Co-IP assays showing decreased RAB7A activity following RRM2 knockdown. (G) Co-IP assays indicating increased interaction between TBC1D15 and RAB7A on osalmid induction. (H) Co-IP assays indicating reduced interaction between TBC1D15 and RAB7A in RRM2 knockdown ARP1 cells. (I) Flow cytometry analysis of MICA/B fluorescence on ARP1 cell membranes after RAB7A knockdown treated with osalmid. (J) Cytotoxicity assessment of NKG2D CAR-T cells against RAB7A knockdown ARP1-luciferase cells at varying effector-to-target ratios. Data are representative of three independent experiments and are given as mean±SE. A one-way ANOVA with Tukey’s post hoc test was used for multiple group comparisons, while a two-tailed unpaired t-test was used for comparisons between two groups; ****p<0.0001, ***p<0.001, *p<0.05. ANOVA, analysis of variance; Co-IP, co-immunoprecipitation; GTP, guanosine 5′-triphosphate; RRM2, ribonucleotide reductase subunit M2.

Figure 8. Osalmid promotes NKG2D CAR-T cytotoxicity by inhibiting MICA/B surface abundance via the RRM2/TBC1D15/RAB7A pathway. The schematic illustrates the mechanism by which RRM2 regulates the intracellular trafficking of NKG2D ligands (MICA/B). In an RRM2-high myeloma subpopulation, RRM2 promotes the degradation of MICA/B via the RRM2-TBC1D15-RAB7A axis, steering MICA/B-containing vesicles toward late endosomes and lysosomes for degradation, while simultaneously inhibiting RAB11-mediated recycling to the cell membrane. This results in reduced surface MICA/B abundance and immune evasion. Treatment with the RRM2 inhibitor osalmid blocks this RRM2-mediated effect, shifting the balance to favor RAB11-dependent recycling and increasing MICA/B surface expression. Enhanced ligand availability on myeloma cells subsequently potentiates NKG2D CAR-T cell recognition, cytotoxicity, and infiltration. CAR, chimeric antigen receptor; GTP, guanosine 5′-triphosphate; RRM2, ribonucleotide reductase subunit M2.

Discussion

Despite CAR-T therapy demonstrates remarkable efficacy in hematological malignancies like MM, long-term success is often hampered by antigen escape.^{35 36} Nevertheless, clinical outcomes have fallen short of expectations, with a substantial proportion of patients failing to respond.^{8 9} Our study delineates a previously unappreciated mechanism of therapeutic resistance in cancer: trafficking-mediated antigen escape. We establish the nucleotide metabolic enzyme RRM2 as a central regulator of immune evasion. Our work initially observed heterogeneity in NKG2DLs at both the transcript levels and membrane abundance on MM cells, which was an inverse association between RRM2 overexpression and surface MICA/B abundance. Collectively, we identify RRM2 as a key regulator of MICA/B membrane trafficking, explaining the limited efficacy of NKG2D CAR-T in MM.

The regulation of immune ligand surface abundance is increasingly recognized as crucial for therapy response. Recent advances highlight how intracellular trafficking, including lysosomal degradation and endocytic recycling, modulates immune checkpoints.37,42 In cellular immunotherapy, CAR internalization and subsequent recycling substantially impact the persistence and efficacy of CAR-T cells against hematological malignancies, potentially facilitating escape and relapse in MM,¹⁷ while several groups focused on upstream proteins and compounds that modulate target antigen expression, aiming to enhance CAR-T therapy.^{43 44} However, trafficking mechanisms regulating target antigens remain poorly explored. RRM2 governs MICA/B membrane localization through a dual trafficking mechanism independent of transcriptional control or ubiquitin-mediated degradation. Notably, we elucidate an RRM2/TBC1D15/RAB7A axis where RRM2 directly binds and maintains RAB7A in its active GTP-bound state, regulating endosomal maturation and lysosomal degradation.45,47 Pharmacologic or genetic inhibition of RRM2 enhances the interaction between the GTPase-activating protein TBC1D15 and RAB7A, increasing RAB7A GTP hydrolysis and consequently suppressing MICA/B lysosomal degradation.⁴⁸ Concurrently, RRM2 inhibition elevates total RAB11 and active RAB11-GTP, promoting endosomal recycling and increasing MICA/B surface localization. Collectively, RRM2 suppresses MICA/B surface density and membrane retention by governing MICA/B intracellular trafficking.

Previous studies established RRM2 as a critical therapeutic target in MM and other malignancies,49,52 while its role in impacting NKG2D CAR-T recognition remained unexplored. Based on our findings that osalmid functions as an RRM2 inhibitor independent of dNTP synthesis,²⁰ we further identified an RRM2⁺ malignant subpopulation that exhibits enhanced immunosuppressive activity and reduced MICA/B surface abundance using single-cell RNA sequencing analysis. Moreover, we demonstrate that RRM2 suppresses MICA/B membrane retention, providing a mechanistic explanation for NKG2D-based therapy failure and expanding the functional repertoire of RRM2 beyond nucleotide metabolism. Subtoxic osalmid pretreatment reversed low MICA/B abundance in myeloma cells and enhanced NKG2D CAR-T cytotoxicity, proliferation, and cytokine secretion in vitro. In addition, osalmid synergized with CAR-T cells to eradicate xenografts while enhancing CAR-T activation, memory phenotypes, and reducing exhaustion in vivo. Current research in bone marrow organoids (BMOs) focuses on using iPSC-derived models to mimic human marrow niche for studying hematologic malignancies and drug screening.²⁴ Using this iPSC-derived myeloma BMO platform, we validated osalmid as a novel combinational agent for NKG2D CAR-T therapy. Osalmid pretreatment reinforced CAR-T function by elevating IFN-γ/granzyme B and sustaining a T_SCM phenotype (CD45RA⁺CCR7⁺), promoting prolonged immunity. Thus, RRM2 inhibition with osalmid represents a translatable strategy to potentiate NKG2D CAR-T efficacy in MM by overcoming antigen escape.

Current strategies to overcome antigen escape predominantly focus on the structural refinement of CARs to lower their activation thresholds and enhance sensitivity toward low-density targets.^{31 53} In this context, we sought to determine whether pharmacological modulation of antigen trafficking could further potentiate the efficacy of these highly sensitive receptor designs. We engineered two CD28-based constructs (NKG2D-CD28ζ and NKG2D-CD28-41BBζ), which showed superior basal recognition of MICA/B-low OCI-MY5 cells compared with 4-1BB-based CARs. Notably, osalmid treatment provided a further significant boost to their cytotoxicity, surpassing the efficacy of receptor engineering alone. This underscores that RRM2 inhibition acts synergistically with high-sensitivity CAR designs to ensure robust tumor eradication.

In summary, this work expands the oncogenic paradigm of RRM2 by unveiling its non-canonical role in immune evasion, revealing it suppresses NKG2DL surface presentation via RAB7A-mediated degradation and dysregulated RAB11-mediated recycling in MM. In addition, we present the combination of subtoxic osalmid and NKG2D CAR-T therapy as a novel regimen to overcome antigen escape validated across experimental models, highlighting its translational potential. Despite these advances, limitations remain to be addressed. The precise mechanism by which RRM2 modulates RAB11 activity requires further elucidation, and the specificity of osalmid toward RRM2 warrants additional validation. Furthermore, while the MICA/B expression levels observed in this study provide a preliminary basis for antigen stratification, establishing standardized clinical criteria requires further investigation. Future work might focus on exploring applicability to solid tumors or a wide array of therapeutic targets, thereby sensitizing them to CAR-T, antibody-based, or other targeted therapies. Collectively, by establishing the RRM2-TBC1D15-RAB7A axis as a clinically actionable target, we pioneer an RRM2-based trafficking strategy to potentiate cell therapies, positioning osalmid as a combinatory agent to overcome antigen escape and improve CAR-T efficacy in MM.

Data availability statement

Data are available upon reasonable request.