A longitudinal single-cell atlas to predict outcome and toxicity after BCMA-directed CAR T cell therapy in multiple myeloma

A longitudinal single-cell atlas to predict outcome and toxicity after BCMA-directed CAR T cell therapy in multiple myeloma | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article A longitudinal single-cell atlas to predict outcome and toxicity after BCMA-directed CAR T cell therapy in multiple myeloma Michael Rade, David Fandrei, Markus Kreuz, Sabine Seiffert, Thomas Wiemers, and 27 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6165798/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Chimeric Antigen Receptor (CAR) T-cell therapies targeting B-cell maturation antigen (BCMA) have transformed the treatment landscape for relapsed/refractory multiple myeloma (RRMM). In this study, we present a real world cohort of 61 RRMM patients treated with idecabtagene vicleucel (Ide-cel, n=34) and ciltacabtagene autoleucel (Cilta-cel, n=27). Cilta-cel demonstrated superior complete response (CR) rates (CR: 78% vs. 38%, p < 0.001) and longer progression-free survival (PFS), with a distinct CAR-T expansion profile marked by increased CD4+CAR+/CD8+CAR+ ratio. To gain insights into immune dynamics encompassing CAR-T cell infusion with either product, we developed a longitudinal multi-omics single-cell atlas using 135 peripheral blood samples from 57 of the 61 patients. There was a strong association between CD4+ cytotoxic T cells and treatment with Cilta-cel, CR and CRS occurrence. Analysis of T cell receptor repertoires showed higher clonality in CD4 T cells in CR patients at all time points. CD8 T cells of non-CR patients showed transcriptomic changes in line with impaired effector function after CAR-T infusion. The BCMA expressing circulating plasma cells, B-cells and plasmacytoid dendritic cells were depleted after infusion in a response-dependent manner, with Cilta-cel leading to significantly slower B-cell recovery (p=0.03). Increased soluble BCMA reduction between day 0 and 30 was linked to stronger CAR-T expansion and higher CRP levels, suggesting an association of tumor debulking and systemic inflammation (p < 0.01, respectively). Our analyses provide a comprehensive resource for understanding longitudinal cellular kinetics in RRMM patients treated with BCMA-directed CAR-T cells. Health sciences/Medical research/Translational research Health sciences/Diseases/Cancer/Haematological cancer/Myeloma Biological sciences/Cancer/Cancer therapy Multiple myeloma single-cell sequencing chimeric antigen receptor T cells B-cell maturation antigen Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Main Chimeric Antigen Receptor (CAR) T-cell therapies targeting the B-cell maturation antigen (BCMA) on malignant plasma cells have revolutionized the treatment landscape for relapsed and refractory multiple myeloma (RRMM) 1 . For patients who have become refractory to the three major classes of anti-myeloma drugs, i.e. immunomodulatory drugs (IMiDs), proteasome inhibitors (PIs) as well as anti-CD38 antibodies, idecabtagene vicleucel (Ide-cel) and ciltacabtagene autoleucel (Cilta-cel) were approved based on the pivotal KarMMa-1 2 and CARTITUDE-1 3 trials. Notable differences in responses and side effects between Ide-cel and Cilta-cel have been observed: Patients treated with Cilta-cel exhibited more durable responses compared to those receiving Ide-cel in real-world, retrospective analyses. However, there is conflicting evidence on a possible association of Cilta-cel with an increased risk of severe adverse events such as cytokine release syndrome (CRS), neurotoxicity and infectious complications 4,5 . While both therapies target BCMA, Cilta-cel is unique in having two BCMA-binding domains 6 , which is widely believed to be the primary factor behind its higher efficacy. Despite this key structural difference, the reasons behind the improved efficacy of Cilta-cel and the distinct toxicity profiles between these two therapies remain poorly understood. Elucidating the biological mechanisms underlying efficacy as well as the side effects of BCMA-directed CAR T-cell therapies would therefore be crucial for optimizing their clinical use and improving patient outcomes. Translational single cell transcriptomic and multi-omic studies generated new insights into the factors correlating with both success and failure of autologous CAR T-cell therapies in MM 7 , B cell neoplasia 8 and auto-immune disorders 9,10 . Nevertheless, a comprehensive longitudinal characterization of the immune cell alterations in MM patients receiving BCMA-directed CAR-T cell therapies has not been performed so far. In this study, we provide a comprehensive multi-omics single-cell dataset of longitudinal immune dynamics following autologous CAR-T cell therapy that will serve as a resource to identify clinically actionable biomarkers for optimizing patient selection and therapy management. Results Study overview and clinical characteristics We recruited a real-world cohort of 61 RRMM patients treated with BCMA-directed CAR-T cells at our center. Samples of 57 patients were available for single-cell multi-omics profiling: A total of 135 peripheral blood mononuclear cell (PBMC) samples at 3 different time points were collected for single cell RNA, B-cell receptor (BCR), T-cell receptor (TCR) and surface protein profiling. These included samples on the day of leukapheresis (LP), at the late (day 21–60, average of 31) and very late (day 79–169, average of 101) time points after infusion. In addition, flow cytometry (FC) was performed at 6 different time points using 2 different antibody panels. Soluble BCMA (sBCMA) levels were measured by ELISA at 4 time points ( Fig. 1a summarizes the study design and sample collection workflow). In total, 27 (44%) patients received Cilta-cel ( Fig. 1b ) and 34 (56%) Ide-cel ( Fig. 1c ). For further analyses, patients were categorized based on best response within 6 months after infusion into complete responders (CR) and non-CR. In line with recent observations 5,11 , we observed a higher overall response rate (93% vs 67%) and CR rate (78% vs 38%) with Cilta-cel compared to Ide-cel. There were no significant differences in age, sex, R-ISS stage, refractoriness, type of bridging therapy, remission status at CAR-T cell infusion and percentages of high-risk cytogenetic aberrations at baseline. However, patients who received Cilta-cel had fewer prior lines of treatment (median 6 vs 8, p=0.04) ( Extended Data Fig 1a and Supplementary Table 1 ). After a median follow-up of 13 months (range, 2-34), median progression-free survival (PFS) was not reached (n.r.) in the Cilta-cel group vs 6 months (95% CI, 3-n.r.) in the Ide-cel group (p=0.001) ( Fig. 1d ). Median PFS was not reached in CR patients vs 3 months (range 2-9) in non-CR patients (p<0.001) ( Fig. 1e ). Triple-class/penta-drug refractory patients had a significantly shorter median PFS with 10 months (4–n.r.) vs not reached in exposed, but non-refractory patients (p=0.007) ( Fig. 1f ). Age greater than 65 years at infusion, type of lymphodepletion or BCMA-directed therapy in an earlier treatment line (excluding bridging) did not influence PFS ( Extended Data Fig. 1b-d ). Analysis of conversion from baseline to post-infusion response confirmed that deeper remission before CAR-T infusion was associated with superior responses with both products, emphasizing our recent findings on the importance of bridging therapy 12 . Cilta-cel demonstrated superior conversion rates (p=0.02, Wilcoxon rank sum test, see methods), with 70% of patients upstaged from stable or progressive disease (SD/PD) to complete response (CR) and 80% from very good partial response or partial response (VGPR/PR) to CR. In contrast, Ide-cel led to an 18% conversion from SD/PD to CR and a 46% conversion from VGPR/PR to CR, respectively ( Fig. 1g ). Concurrently, cox proportional hazard regression adjusted for CAR-T product showed that deeper baseline remission status was associated with a trend towards improved PFS (p=0.07) ( Fig. 1h ). The presence of extramedullary disease (p=0.004) and elevated sBCMA (>107 ng/mL, p=0.017) were independent negative baseline predictors, whereas high-risk cytogenetics did not show a strong association with PFS (p=0.29). Analyzing CAR-T cell expansion by flow cytometry on days 0, 7, 14, 30 and 100 after infusion revealed an increased peak expansion (c max ) in CR patients on day 14 compared to non-CR patients, however not statistically significant (p=0.135). The proportion of CAR-T cells was significantly higher in CR patients at days 30 (p=0.007) and 100 (p=0.027), indicating slower contraction compared to non-CR patients. ( Fig. 1i and Supplementary Table 2 ). Furthermore, CD4+CAR+/CD8+CAR+ ratio was significantly higher in CR vs non-CR patients on days 7 (p<0.001), 14 (p=0.022) and 30 (p=0.004), emphasizing the critical role of CD4+ CAR-T cells in tumor eradication ( Fig. 1j ). Longitudinal single cell landscape of patients treated with anti-BCMA CAR T cells Following the identification of key clinical factors associated with patient outcomes, we explored the longitudinal peripheral blood immune dynamics encompassing CAR-T cell infusion. Fig. 2a provides an overview of sample availability for single cell multi-omics analyses. After rigorous quality filtering and cell type annotation, approx. 450,000 cells were retained for downstream analyses (median of 3,053 cells per sample, range 675– 8,478, Fig. 2a ). Coarse level cell type annotation, expression of the corresponding canonical marker genes, quality control metrics, and cell type composition for each sample are shown in Fig. 2b-g and Extended Data Fig. 2a-k (see Supplementary Tables 3, 4 for further quality control metrics). In addition, we used CD4 and CD8 reference atlases 13 to refine the annotation of T cell identities ( Fig. 2h and Extended Data Fig. 3d,e for marker genes). Two distinct clusters of differentiated effector memory T cells re-expressing CD45RA (EMRA) were identified. The EMRA 1 subset exhibited high expression of THEMIS and CD5 , whereas the EMRA 2 subset was characterized by elevated levels of IKZF2 and KLRC2 . ( Extended Data Fig. 3 a-c ). We observed a significant (p <0.05) Spearman correlation between the estimated proportions of T cell subtypes by flow cytometry and scRNA-seq ( Fig. 2i and Supplementary Table 5 ), which was also confirmed for CAR + T cells ( Fig. 2j ). However, we noted an underestimation of the expression of the Cilta-cel CAR construct by scRNA-seq, which may be due to drop-out events 14 caused by lower expression of the Cilta-cel compared to the Ide-cel construct ( Extended Data Fig. 3f ). We analyzed the influence of cell type composition at each time point on survival outcomes: Using Kaplan-Meier analysis, elevated levels of circulating plasma cells (cPC), CD56-bright NK cells, CD4+ T follicular helper cells (Tfh), and plasmacytoid dendritic cells (pDC) were significantly associated with reduced PFS, while increased proportions of both CD8+ EMRA subsets, central memory CD8+ (CM), and granulysin-expressing cytotoxic (CTL) CD4+ T cells (CD4 GNLY) at late time points were associated with improved PFS. ( Fig. 2k , see Extended Data Fig. 3g for all time points). Univariate cox proportional hazard regression using the scaled cell type fractions confirmed these results (estimated log hazard ratios (logHRs) on a continuous scale for each time point are shown in Fig. 2l). Importantly, logHRs remained consistent after adjustment for the CAR-T product ( Extended Data Fig. 3h ). Although γδ T cells did not show a significant association with PFS, we noted that a median of 82% of γδ T cells expressed the variable δ variant Vδ1 and only 3% expressed Vδ2 or Vδ3. This observation is consistent with previous findings that Vδ1 is more abundant in mucosal tissues and cancer specimens than γδ T cells from healthy individuals 15 . Next, we hypothesized that cell types linked to unfavorable outcomes might express BCMA, making their reduction a surrogate marker of effective tumor cell killing or, alternatively, a factor that disrupts CAR-T cell interactions with malignant PCs. Querying a scRNA-seq atlas 16 spanning tumor and normal cells across 26 tissues revealed that pDC, which had the strongest negative impact on PFS after cPC, exhibited the highest TNFRSF17 (BCMA) specificity and expression among non-plasma and B cell populations ( Supplementary Fig. 1 ). Temporal evolution of the T-cell repertoire We analyzed significantly differentially expressed (DE) genes between non-CR and CR in T cell subtypes ( Fig. 3a ). DE genes (adjusted p-value < 0.05) at the late time point included the CAR construct, which was significantly lower expressed in eomesodermin ( EOMES ) expressing CD4 CTL. Gene ontology (GO) enrichment analysis of DE genes revealed upregulation of GO terms associated with regulation of T cell mediated immunity and activation in CD4 CTL subsets, which were enriched with DE genes such as TCF7 and IL7R . The metabolic pathways OXPHOS and aerobic respiration enriched with mitochondrial genes were decreased in CD8 effector subsets. Recently, mitochondrial dysfunction has been described as a hallmark of T cell dysfunction by driving metabolic reprogramming triggering terminal T cell exhaustion 17,18 . TIGIT was the only significantly upregulated checkpoint marker in CD8 effector populations ( Fig. 3b, DE genes and enrichment analysis for all time points are shown in Extended Data Fig. 4 ), and integration with surface protein data revealed higher expression of CD38 on CD8 effector subsets , which was recently identified to drive terminal CAR-T cell exhaustion 19 (see Supplementary Table 6 ). T cells were subjected to trajectory inference modeling and subsequently partitioned into nodes across the trajectory. The root node is depicted by a dot in Fig 3c . Cells were grouped into clusters according to the nearest node along the trajectories ( Fig. 3c-e ). The T cell compartment captured the differentiation trajectory spanning from naïve like cells to memory and effector cells, providing a framework for analyzing dynamic changes in cellular proportions. Naïve like CD4 cells underwent clonal expansion along trajectory 2 into T cells exhibiting increased activation, and cytotoxic effector function ( Fig. 3c-d and Extended Data Fig. 5 ). When comparing the cell abundances of non-CR with CR patients for CD4+ T cells along the trajectory 2, we observed a significantly (p <0.05) lower abundance of cytotoxic cells in non-responders ( Fig. 3e ). Overall, non-CR showed an increased proportion of naïve like CD4 and CD8 cells along the trajectories, with the highest percentage of CAR+ cells in nodes with cytotoxic/effector functions (depicted as numbers in Fig 3e ). Analysis of the TCR repertoire showed that non-CR vs CR patients exhibited a higher proportion of CD4 clones consisting of only one cell (singletons) at all time points ( Fig 3f ). Conversely, clonotype richness was higher in CR patients ( Fig 3g ). Significant differences in clonality for CD8 subsets were only observed at leukapheresis. Composition of the T cell compartment, grouped by CAR- and CAR+ cells, showed that CD4+CAR+ singletons are predominantly of the cytotoxic phenotype expressing EOMES ( Fig 3h) . This phenotype was also more pronounced in the expanded CAR+ clones compared to the CAR- clones. The comparison of gene expression of CAR+ with CAR- (61 DE genes) or pre-infusional (314 DE genes) T cells revealed that CAR+ cells expressed significantly higher levels of genes associated with cytolysis and exhaustion, where this effect was less pronounced when compared to CAR-. Naïve-like and memory gene sets, on the other hand, were downregulated in CAR+ cells ( Fig 3i and Supplementary Table 6) . To characterize potentially tumor-reactive T cells (pTRT), we used an in vitro validated transcriptional T cell signature for anti-tumor reactivity in MM, as described recently 20 . The highest median enrichment score of pTRTs was observed for CD4 GNLY and CD8 subtypes, especially for EMRA identities ( Fig 3j) . Cox-regression analysis using the fractions of pTRTs for each subtype on a continuous scale showed a significant association with PFS for late time points ( Fig 3k) . Here, a higher proportion of pTRTs in EMRA subsets showed the strongest association with improved PFS. Single cell profiling of cytokine release syndrome during CAR therapy Following CAR-T infusion, 67% of the patients experienced grade I or II CRS, 37% of whom required tocilizumab application (no CRS: n = 20 CRS/without tocilizumab: n = 26, CRS/with tocilizumab: n = 15). By product, 79% of patients treated with Ide-cel and 52% of patients treated with Cilta-cel developed any grade CRS (p=0.03, Fisher’s exact test). Only 3 patients who received Cilta-cel (11%) and one patient (3%) treated with Ide-cel experienced grade I ICANS (p=0.4). As expected, patients with grade I and II CRS had significantly increased peak levels of serum C-reactive protein (CRP) (p = 0.002) ( Fig. 4a ). There were no significant differences in maximum CRP values between response groups (p = 0.28). In addition, we observed a significant Spearman correlation (p = 0.011) between the estimated proportions of CAR+ cells seven days after infusion and the maximum CRP values ( Fig. 4b ), with proportions of CAR+ cells increasing significantly with CRS grade (p = 0.002) ( Fig. 4c ). As only 6 patients with available single-cell data developed grade II CRS, we divided the cohort into CRS (grade I and II) and non-CRS groups for subsequent analyses. We observed significant changes in cell type composition of patients who developed any grade CRS: B cell subsets, naïve-like CD4 and CD8 T cells were significantly more abundant at the time of leukapheresis, whereas regulatory T cells were enriched at the late time point ( Fig. 4d ). Comparison of CD4 and CD8 T cell clones between CRS and non-CRS patients revealed a stronger increase in the cytotoxicity enrichment score of CD4 clones compared to CD8 clones at the late time point ( Fig. 4e ). In previous studies, transcriptional changes involving genes encoding cytokines and the number of polyfunctional CAR+ cells correlated with CRS 15 . Here, we found a significant increase in the proportion of polyfunctional CD8 T cells at the late time point in patients with CRS ( Fig. 4f ), but no significant differences in fractions of polyfunctional CD4 T cells at any time point. Differential gene expression analysis (DGEA) comparing CRS with non-CRS patients showed upregulation of genes involved in TNF-α signaling via NF-κB (e.g. NFKBIA , RELB , JUN, PIM3 ) in most lymphoid and myeloid populations at the late time point, reflecting systemic inflammation ADDIN CitaviPlaceholder{{"$id":"1","$type":"SwissAcademic.Citavi.Citations.WordPlaceholder, SwissAcademic.Citavi","Entries":[{"$id":"2","$type":"SwissAcademic.Citavi.Citations.WordPlaceholderEntry, SwissAcademic.Citavi","Id":"80fab19a-ebb7-4482-a86e-370473ce2769","RangeLength":2,"ReferenceId":"a2057701-522b-4bdf-9545-789c9c372b35","PageRange":{"$id":"3","$type":"SwissAcademic.PageRange, SwissAcademic","EndPage":{"$id":"4","$type":"SwissAcademic.PageNumber, SwissAcademic","IsFullyNumeric":false,"NumberingType":0,"NumeralSystem":0},"NumberingType":0,"NumeralSystem":0,"StartPage":{"$id":"5","$type":"SwissAcademic.PageNumber, SwissAcademic","IsFullyNumeric":false,"NumberingType":0,"NumeralSystem":0}},"Reference":{"$id":"6","$type":"SwissAcademic.Citavi.Reference, SwissAcademic.Citavi","AbstractComplexity":0,"AbstractSourceTextFormat":0,"Authors":[{"$id":"7","$type":"SwissAcademic.Citavi.Person, SwissAcademic.Citavi","FirstName":"Ting","LastName":"Liu","Protected":false,"Sex":0,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"0c527c05-0d9c-4e1f-ad08-29f85fc3bd6d","ModifiedOn":"2025-01-26T22:25:52","Project":{"$id":"8","$type":"SwissAcademic.Citavi.Project, SwissAcademic.Citavi"}},{"$id":"9","$type":"SwissAcademic.Citavi.Person, SwissAcademic.Citavi","FirstName":"Lingyun","LastName":"Zhang","Protected":false,"Sex":0,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"6f74842b-b038-42c5-85c7-970a6eade8ad","ModifiedOn":"2025-01-26T22:25:52","Project":{"$ref":"8"}},{"$id":"10","$type":"SwissAcademic.Citavi.Person, SwissAcademic.Citavi","FirstName":"Donghyun","LastName":"Joo","Protected":false,"Sex":0,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"24615fa6-e415-4a37-8555-2f2549a42879","ModifiedOn":"2025-01-26T22:25:52","Project":{"$ref":"8"}},{"$id":"11","$type":"SwissAcademic.Citavi.Person, SwissAcademic.Citavi","FirstName":"Shao-Cong","LastName":"Sun","Protected":false,"Sex":0,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"ca56fd92-2aeb-4c23-9c43-a7dfaaaaac55","ModifiedOn":"2025-01-26T22:25:52","Project":{"$ref":"8"}}],"CitationKeyUpdateType":0,"Collaborators":[],"Date2":"14.07.2017","Doi":"10.1038/sigtrans.2017.23","Editors":[],"EvaluationComplexity":0,"EvaluationSourceTextFormat":0,"Groups":[],"HasLabel1":false,"HasLabel2":false,"Keywords":[],"Language":"eng","LanguageCode":"en","Locations":[{"$id":"12","$type":"SwissAcademic.Citavi.Location, SwissAcademic.Citavi","Address":{"$id":"13","$type":"SwissAcademic.Citavi.LinkedResource, SwissAcademic.Citavi","LinkedResourceType":5,"OriginalString":"PMC5661633","UriString":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5661633","LinkedResourceStatus":8,"Properties":{"$id":"14","$type":"SwissAcademic.Citavi.LinkedResourceProperties, SwissAcademic.Citavi"},"SyncFolderType":0,"IsLocalCloudProjectFileLink":false,"IsCloudRestore":false,"IsCloudCopy":false,"AttachmentFolderWasInFallbackMode":false},"Annotations":[],"LocationType":0,"MirrorsReferencePropertyId":208,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"bf0dc913-de16-4f10-be8c-06403ba9a05b","ModifiedOn":"2025-01-26T22:25:52","Project":{"$ref":"8"}},{"$id":"15","$type":"SwissAcademic.Citavi.Location, SwissAcademic.Citavi","Address":{"$id":"16","$type":"SwissAcademic.Citavi.LinkedResource, SwissAcademic.Citavi","LinkedResourceType":5,"OriginalString":"10.1038/sigtrans.2017.23","UriString":"https://doi.org/10.1038/sigtrans.2017.23","LinkedResourceStatus":8,"Properties":{"$id":"17","$type":"SwissAcademic.Citavi.LinkedResourceProperties, SwissAcademic.Citavi"},"SyncFolderType":0,"IsLocalCloudProjectFileLink":false,"IsCloudRestore":false,"IsCloudCopy":false,"AttachmentFolderWasInFallbackMode":false},"Annotations":[],"LocationType":0,"MirrorsReferencePropertyId":128,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"eb890f24-6b76-421e-8c55-10f413fcef14","ModifiedOn":"2025-01-26T22:25:52","Project":{"$ref":"8"}},{"$id":"18","$type":"SwissAcademic.Citavi.Location, SwissAcademic.Citavi","Address":{"$id":"19","$type":"SwissAcademic.Citavi.LinkedResource, SwissAcademic.Citavi","LinkedResourceType":5,"OriginalString":"29158945","UriString":"http://www.ncbi.nlm.nih.gov/pubmed/29158945","LinkedResourceStatus":8,"Properties":{"$id":"20","$type":"SwissAcademic.Citavi.LinkedResourceProperties, SwissAcademic.Citavi"},"SyncFolderType":0,"IsLocalCloudProjectFileLink":false,"IsCloudRestore":false,"IsCloudCopy":false,"AttachmentFolderWasInFallbackMode":false},"Annotations":[],"LocationType":0,"MirrorsReferencePropertyId":164,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"75c43b9f-15a0-4faf-b8d6-226d24a7dc3a","ModifiedOn":"2025-01-26T22:25:52","Project":{"$ref":"8"}}],"Organizations":[],"OthersInvolved":[],"PageRange":"<sp>\r\n  <n>17023</n>\r\n  <ns>Omit</ns>\r\n  <os>17023-</os>\r\n  <ps>17023-</ps>\r\n</sp>\r\n<os>17023-</os>","Periodical":{"$id":"21","$type":"SwissAcademic.Citavi.Periodical, SwissAcademic.Citavi","Eissn":"2059-3635","Name":"Signal transduction and targeted therapy","Pagination":0,"Protected":false,"UserAbbreviation1":"Signal Transduct Target Ther","CreatedBy":"_Michael Rade","CreatedOn":"2025-01-15T22:00:15","ModifiedBy":"_Michael Rade","Id":"f30ef87b-d02e-47dd-9bfa-63a163580d1c","ModifiedOn":"2025-01-15T22:00:15","Project":{"$ref":"8"}},"PmcId":"PMC5661633","Publishers":[],"PubMedId":"29158945","Quotations":[],"Rating":0,"ReferenceType":"JournalArticle","ShortTitle":"Liu, Zhang et al. 2017 – NF-κB signaling in inflammation","ShortTitleUpdateType":0,"SourceOfBibliographicInformation":"PubMed","StaticIds":["3d612eab-59f6-45a9-8c44-30426e1d57de"],"TableOfContentsComplexity":0,"TableOfContentsSourceTextFormat":0,"Tasks":[],"Title":"NF-κB signaling in inflammation","Translators":[],"Volume":"2","Year":"2017","YearResolved":"2017","CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"a2057701-522b-4bdf-9545-789c9c372b35","ModifiedOn":"2025-03-05T18:34:08","Project":{"$ref":"8"}},"UseNumberingTypeOfParentDocument":false}],"FormattedText":{"$id":"22","Count":1,"TextUnits":[{"$id":"23","FontStyle":{"$id":"24","Superscript":true},"ReadingOrder":1,"Text":"21"}]},"Tag":"CitaviPlaceholder#8285fb0c-58a8-41d0-8a25-0c634287b511","Text":"21","WAIVersion":"6.15.2.0"}} 21 ( Fig. 4g , see Extended Data Fig. 6 and Supplementary Table 6 for DE genes of all cell types and time points). In addition, we observed a higher expression of the CAR construct in CD4 EOMES, along with a higher expression of cytotoxic-related genes GNLY and GZMB in CD4 CTL subsets. Furthermore, the costimulatory receptor CD27, which governs an essential signaling cascade stimulating differentiation and long-term survival in activated T cells 22,23 , was upregulated in various CD4 and CD8 populations. In line with these findings, increased TCR signaling was confirmed in both CD8 and CD4 effector subsets at late time points ( Fig. 4h ). Gene set enrichment analysis further revealed upregulation of the hypoxia signature in post-infusional CD4 and CD8 effector T cells, cDCs and CD14 + monocytes ( Extended Data, Fig. 7 ). At the time of LP, naive B cells showed increased MHC-I presentation to CD8 effector subsets in patients with CRS ( Fig. 4i) . Potential receptor-ligand interactions in post-infusional T cells were mostly incoming signals to the CD69 receptor, which was upregulated on CD8 effector and γδ T populations and downregulated on CD8 naive cells. The T cell activation marker CD69 has been identified to play a key role in tissue retention and might therefore play a crucial role for CRS pathophysiology 24 . To detect gene programs within our single-cell dataset independent of cell clustering and annotation, we performed supervised pathway deconvolution analysis using the Spectra method 25 . Spectra decomposes the scRNA-seq into a set of interpretable factors or gene programs (sets of co-expressed genes) using a curated list of global and T cell specific gene sets as prior information for modeling. Fig. 4j depicts the differences in enrichment of estimated gene programs between patients with CRS and without CRS. Gene programs associated with tumor reactivity, exhaustion, and TNF were significantly (adjusted p-value < 0.05) upregulated in patients with CRS while TNK cytotoxicity was upregulated CD4 CTL GNLY and CD8 EMRA subtypes. Genes associated with the programs and the cell factor loadings are shown in F ig. 4k . Comparison between patients treated with Cilta-cel and Ide-cel We observed distinct expansion profiles between the two CAR-T products: Patients who received Cilta-cel showed increased c max and slower expansion kinetics with c max reached on day 14 vs day 7. The CD4+CAR+/CD8+CAR+ ratio was significantly elevated with Cilta-cel compared to Ide-cel on days 7, 14 and 30 ( Fig. 5a ). Single-cell analysis did not show differentially abundant cell populations at LP, but significant changes in post-infusional cell type composition between patients treated with Cilta-cel and Ide-cel. CD4 CTL and memory/effector/naïve-like CD8 cells were more abundant in Cilta-cel-treated patients ( Fig. 5b ). In addition, we performed paired analyses of cell type fractions between the time points ( Fig. 5c) . In Cilta-cel-treated patients, CD4 EOMES and CD8 EM cells increased significantly (p <0.05) comparing the late time point to LP and decreased significantly comparing the very late to the late time point (see Extended Data Fig. 8 for all cell types). Next, we performed DGEA for T cell subtypes from Cilta-cel and Ide-cel patients comparing Late with LP time points. In Cilta-cel patients, we observed significant upregulation of CD27 and downregulation of CXCR4 , GZMB and GZMK in CD4 CTL subsets. Both products showed similar changes in CD8 populations, with upregulation of T-cell activation and exhaustion genes and downregulation of naive and stem-like genes ( Fig. 5d ). Fig. 5e depicts DE genes (adjusted p-value <0.05) with opposite direction of log2 fold change between Cilta-cel and Ide-cel. Unexpectedly, cytotoxic markers GNLY and GZMB were upregulated in T cells of Ide-cel patients and downregulated in Cilta-cel patients. Analysis of longitudinal dynamics for the most abundant clones with CAR+ cells revealed a distinct CD4 EOMES phenotype in patients treated with Cilta-cel ( Fig. 5f ). DGEA for the late time point, comparing CD8+CAR+ Cilta-cel with CD8+CAR+ Ide-cel cells showed upregulation of memory-associated ( SELL , IL7R ) 26 and downregulation of effector-associated genes ( KLRD1 , GZMB , GNLY ). Accordingly, enrichment analysis revealed a higher cytotoxicity and effector function in Ide-cel CARs ( Fig. 5g and h ). sBCMA dynamics and immune reconstitution We examined changes in B-cell lineage subtypes and the B-cell receptor (BCR) landscape ( Fig. 6a and Extended Data Fig. 9a) . This confirmed monoclonal cPCs and suppressed polyclonal B-cells at each analyzed timepoint (Fig. 6b). Five patients had relevant proportions of cPCs ( Fig. 6c and Extended Data Fig. 9b ), which had retained expression of the immuno-therapy targets TNFRSF17 and CD38 and to a lesser extent of putative novel targets FCRL5 and LAMP5 ( Fig. 6d ). Interestingly, TNFRSF17 expression remained high even at the very late timepoint, suggesting there was no BCMA loss in refractory malignant PCs of these patients ( Extended Data Fig. 9c ). We observed depletion of B-cells, pDCs and cPCs at the late time point and their recovery at the very late time point in scRNA-seq data. As expected, there were significantly lower proportions of all BCMA-expressing cell types in CR compared to non-CR patients at the time points following CAR-T cell infusion ( Fig 6e-g ). B-cells showed a shift towards early stages of B-cell differentiation for the late and recovery of memory B-cells at the very late time point ( Extended Data Fig. 9d ). There was no significant association between the incidence of infections and B-cell proportions at any of the analyzed time points ( Extended Data Fig. 9e ). Next, we analyzed longitudinal B cell frequencies from routine clinical blood counts (n=55), revealing that all patients except for one (98%) had B cell aplasia following CAR-T cell infusion. B-cell recovery defined as any detectable B-cells occurred significantly earlier in non-CR vs CR patients ( Fig. 6h) , which was driven by the Ide-cel subgroup ( Extended Data Fig. 9f ). However, most patients remained below the lower normal range even 200 days after infusion. Direct comparison of B-cell recovery dynamics between patients treated with Cilta-cel and Ide-cel using the inverse Kaplan-Meier method showed significantly slower B-cell recovery after Cilta-cel infusion ( Fig. 6i ). This was confirmed using a Cox proportional hazard model adjusted for clinical response (p=0.04). Longitudinal analysis of sBCMA showed significantly lower levels in CR vs non-CR patients on days 0, 30 and 100, consistent with recent data emphasizing its role as a putative surrogate marker for MM tumor burden 27,28 ( Fig. 6j ). To determine the likely source of sBCMA, we conducted correlation analyses with BCMA-expressing cell types at LP, on day 30 and 100 after infusion ( Fig. 6k ). No correlation was observed at LP, whereas there was a significant correlation between sBCMA, pDC and cPC fractions at the late time point, confirming similar depletion of all BCMA-expressing cell types. At the very late time point, sBCMA levels and the cPC fraction showed the strongest correlation, suggesting that sBCMA after CAR-T cell infusion is mostly derived from cPCs. Finally, we investigated the relationship between sBCMA reduction between days 0 and 30 and CAR-T cell expansion ( Fig. 6l ). We observed a significant correlation between c max and sBCMA reduction (p=0.00013). Furthermore, maximal CRP levels significantly correlated with sBCMA reduction (p=0.0064; Fig. 6m and Extended Data Fig. 9g ) providing a connection between tumor debulking and the severity of system inflammation following CAR-T cell infusion. Taken together, these findings validate the role of sBCMA as a prognostic biomarker for BCMA-directed CAR-T cells in RRMM and link high initial tumor burden and effective CAR-T cell expansion with CAR-T related toxicities. Discussion Our study provides a comprehensive clinical and molecular characterization of patients treated with BCMA-directed CAR T-cell therapies. Consistent with previous real-world studies 5,29–31 , Cilta-cel exhibited superior response rates, with a significantly higher CR rate (78% vs. 38%) and PFS compared to Ide-cel. We observed distinct cellular kinetics between the two products. Cilta-cel exhibited higher and delayed peak expansion aligning with the later onset of CRS reported in recent real-world studies 5 . This delayed expansion has important implications for the ambulatory application of Cilta-cel and individualized patient monitoring 32 . Given that peak CAR-T expansion was associated with higher-grade CRS, preemptive strategies such as dexamethasone administration in patients with high CAR-T cell counts are currently being explored in clinical trials. Furthermore, Cilta-cel-treated patients had a significantly higher CD4+CAR+/CD8+CAR+ ratio on days 7, 14, and 30 compared to Ide-cel patients, suggesting an important role for CD4+ CAR-T cells in mediating durable anti-tumor responses. Recent single cell transcriptomic studies of patients who received anti-CD19 CAR-T cell therapy for B cell neoplasia emphasized the role of CD4+CAR+ T cells for tumor-clearance and long-term remission 33,34 . In line with the expansion data, single cell analyses revealed that CD4 CTL subsets were more abundant in Cilta-cel vs Ide-cel and CR vs non-CR patients. Additionally, the CD4 CTL population of patients treated with Cilta-cel showed a distinct transcriptional profile with upregulation of CD27 and downregulation of CXCR4 , GZMB and GNLY . Recent analyses demonstrated superior antitumor efficacy in CD27-costimulated CAR-T cells 23 . DGEA of non-CR compared to CR patients revealed a transcriptional signature of impaired effector function, mitochondrial dysfunction and upregulation of exhaustion-related genes such as TIGIT and CD38 in CD8 T effector subsets at the late time point. Therefore, impaired T cell effector function is a main driver of suboptimal clinical responses to anti-BCMA CAR-T cell therapy in RRMM. However, recent studies have highlighted that predicting response and toxicity in CAR T-cell therapy is not solely dependent on CAR T-cell properties and host immune fitness, but also on tumor burden, with sBCMA emerging as a key predictive factor influencing both efficacy and side effects 27,28,35 . sBCMA levels correlated with initial tumor burden, with greater reductions linked to stronger CAR-T expansion. Interestingly, patients with greater sBCMA reduction also exhibited higher peak CRP levels, suggesting a connection between tumor debulking and systemic inflammation. This finding aligns with prior observations that high antigen density can drive more robust CAR-T expansion 36 and inflammatory responses, potentially contributing to CRS pathophysiology. We identified pDCs to potentially contribute to ongoing BCMA ligand availability, which might impair CAR T-cell efficacy by acting as antigen decoys. Given their high BCMA expression, pDCs might interfere with CAR-T functionality by occupying BCMA-targeted CAR constructs, reducing effective antigen engagement on malignant plasma cells. The persistence of pDCs in circulation post-infusion and their association with lower response rates suggest that they may serve as a surrogate marker for therapeutic efficacy. Our findings may also have broader implications for dendritic cell-associated diseases, including autoimmune disorders 37 and blastic plasmacytoid dendritic cell neoplasm (BPDCN), a clinically aggressive neoplasm derived from non-activated precursors of pDCs 38 . Taken together, our findings highlight the superior remission conversion potential of Cilta-cel, its distinct CAR-T expansion and transcriptional profile, and the interplay between tumor burden, T cell fitness, and systemic inflammation. Methods Study design Data and sample collection were approved by the local ethics committee (Leipzig Medical Biobank/LMB No. 311-15-24082015 and LMB-UCCL-2022_06 // 361/22-ek) and performed in accordance with the Declaration of Helsinki after obtaining written informed consent. All 61 RMM patients who received Ide-cel or Cilta-cel outside of clinical trials at University Hospital Leipzig (Leipzig, Germany) between 01/06/2022 and 01/04/2024 were included in this study. The median age at CAR-T infusion was 64 (range 31-75) years and 44% of patients were female. All patients were of Non-Hispanic European origin. The median number of prior lines of therapy before CAR-T was 7 (1-15) across all patients. Manufacturing failure occurred in 10 patients, and 4 patients received products that were out-of-specification of the manufacturer's quality criteria ( Extended Data Fig. 1a ). BCMA-directed therapies in earlier treatment lines (excluding bridging) were administered to 8 (13%) patients. All patients received bridging therapy with either chemotherapy (31%), CD38 antibody-based, (31%) SLAMF7 antibody-based (13%) regimens or bispecific T cell engaging antibodies (25%). Lymphodepletion was performed using fludarabine and cyclophosphamide (n=56, 92%) or bendamustine (n=5, 8%) in patients with severe renal insufficiency. Patients with mild/moderate renal insufficiency received a Flu/Cy dose adjusted for creatinine clearance determined by CKD-EPI (n=5, 8%). The patients were followed up in our outpatient clinic and response was assessed according to International Myeloma Working Group (IMWG) criteria 39 at the time point of LP, on the day of CAR-T infusion, and at months 1, 3, 6, and 12 thereafter. Overall response rate was defined as the proportion of patients who were in partial response or better and participants were categorized into CR and non-CR based on the best response within 6 months post CAR-T infusion. CRS and ICANS were assessed according to the American Society for Transplantation and Cellular Therapy (ASTCT) recommendations 40 . Cytogenetic analyses were performed using fluorescence in-situ hybridization. The presence of gain(1q) (3 copies) or amp(1q) (4 or more copies), t(4;14), t(14;16) or del(17p) were considered as high-risk aberrations. EMD was defined as any soft-tissue, organ or paramedullary plasma cell infiltration, determined non-contiguous to bone tissue by PET-CT before CAR-T cell infusion. Sample collection Mononuclear cells from peripheral blood (PBMCs) were obtained from LP, on days 21–60 (average 31) and on days 79–169 (average 101) after CAR-T infusion for single cell multi-omics sequencing. PBMC for flow cytometry and blood sera were obtained on the day of LP, and on days 0, 7, 14, 30 and 100. Samples werecryopreserved and stored by trained staff of the Leipzig Medical Biobank according to standard operating procedures. Clinical quantification of CAR-T cell expansion by flow cytometry was performed using fresh whole blood samples as described below. Sample collection and preparation were blinded to the conditions of the patients. Immunophenotyping by flow cytometry PB samples were collected at LP, on the day of CAR-T cell infusion (day 0), and on days 7, 14, 30 and 100 thereafter. Flow cytometry analysis was performed immediately on fresh samples using two separate antibody panels to quantify CAR+ T cells and T cell subsets, as described previously 41 . In brief, the samples were stained with fluorochrome-conjugated antibodies for 15 min at room temperature. Then, red blood cell lysis was performed by adding 2 ml of lyse solution (Becton Dickinson Biosciences, BD) for 10 min. The cells were centrifuged at 500g for 5 min at room temperature. In the following, the cell pellets were resuspended and washed in 2ml PBS for 5 min at room temperature. At last, the cells were subjected to flow cytometry using a BD FACSLyric TM and data analysis was performed using the BD FACSuite TM software (see Supplementary Fig. 2-4 for the gating strategy). For Fig. 2i , the proportions of CD4 naïve and central memory T cells estimated by flow cytometry were added together. The rationale for this is that the naïve T cells from the CD4 reference atlas were determined on the basis of scRNA-Seq 42 . Since quantification in single-cell RNA analyses is performed at the gene level, it is not feasible to differentiate between the CD45RA and CD45RO isoforms and thus between naïve and central memory CD4 cells. The CD8 effector cells estimated by flow cytometry were re-annotated as EMRA cells for correlation analysis with scRNA-Seq. Immunoassay for the quantification of serum BCMA concentration Serum samples were used to determine soluble B-cell maturation antigen (sBCMA) by ELISA (Bio-Techne, cat. DY193 and DY008B), and absorbance was recorded using the FLUOstar OMEGA. Survival analyses We performed survival analyses using the R survival package v3.5.7. Kaplan–Meier plots for patients were generated using the ggsurvplot function from the survminer v0.4.9 R package. Primary endpoint was progression free survival (PFS), defined as time from CAR-T infusion to progression or death. Log-rank tests were performed to compare probabilities of PFS between groups using the survdiff function implemented in the survival package. Cox proportional hazards models were computed using the coxph function from the survival package. In Fig. 1h and Extended Data Fig. 3h , we applied multivariable Cox-regression to assess the performance of the covariates with adjustment for the CAR product. For the dichotomized covariate sBCMA ( Fig. 1h ), the cutoff was determined using the median of the cohort at the time of leukapheresis before the administration of bridging therapy to identify patients with high tumor burden at CAR-T infusion 28 . For the analysis of response conversion, the categories SD/PD, VGPR/PR and CR were encoded as numerical factors and the difference between best response within 6 months after CAR-T infusion and baseline response was calculated for each patient. Then, Wilcoxon rank sum test was used to test whether conversion rates were significantly different between patients who received Cilta-cel and Ide-Cel. B-cell recovery was defined as the first time point with any detectable B-cells from routine clinical blood counts. Patients were censored if they were in ongoing B-cell aplasia at last FU. Inverse Kaplan-Meier curves were computed to compare B-cell recovery in patients treated with Cilta-cel and Ide-cel. Single-cell RNA, TCR, BCR and ADT sequencing Four sequencing runs were performed for the multi-omics analysis, with the first run carried out as described previously 7 . For the sequencing runs 2-4, PBMCs were isolated following standardized operating procedures using Leucosep tubes (Greiner Bio One, Frickenhausen, Germany) with a Ficoll gradient (Ficoll-Paque™ PLUS, VWR Avantor, Spring House, Pennsylvania). Cells were frozen using a CoolCell LX container (Corning, USA) and stored in the vapour phase of liquid nitrogen in an Askion HS200S storage system (Askion, Gera, Germany). For single-cell experiments, samples were thawed and stained with acridine orange and propidium iodide (AO/PI) dye, and quality was assessed using LUNA-FX7 TM automated cell counter (Logos Biosystems, United Sates). Samples with initial viability <80% were subjected to dead cell removal using Viahance TM dead-cell removal kit (BioPAL, Inc., United States) following the manufacturer’s instructions. To enable single-cell protein expression profiling, the cells were stained with a custom panel of TotalSeq TM -C antibodies (BioLegend, United States), optimized, and prepared by Singleron Biotechnologies, Singapore. These TotalSeq TM -C antibodies are oligonucleotide-coupled antibodies directed against selected cell surface proteins. The antibodies are conjugated via a linker to oligonucleotides, which consist of a polymerase chain reaction (PCR) handle, an antibody-specific barcode, and a capture sequence. These antibody-bound oligonucleotides can be captured during single-cell library preparation, allowing simultaneous surface protein and gene expression profiling. For sequencing runs 2-4, samples of up to two donors were pooled and processed together. Specific to the second run, multiplexed samples were stained with a unique TotalSeq TM -C anti-human hashtag antibodies (BioLegend, United States) prior to pooling. The custom panel comprised 62 antibodies for sequencing run 1, 63 antibodies for run 2 and 70 for run 3 and 4 ( Supplementary Table 7-9 ). The single-cell gene expression, surface protein expression, B cell receptor (BCR) and T cell receptor (TCR) libraries were constructed using Chromium Next GEM Single Cell 5’ reagent kits (v2 chemistry; 10x Genomics, Inc., United States), as per the manufacturer’s instructions. Libraries were sequenced on an Illumina NovaSeq 6000 using S4 flow cells (sequencing runs 1-3) or NovaSeq X plus platform using 10B or 25B flow cells (sequencing run 4) and PE150 chemistry for all sequencing runs. Pre-processing and d emultiplexing of single-cell data The CAR sequences for Ide-cel 43 and Cilta-cel 44 were obtained from the respective patents. Two separate custom references for Cilta and Ide-cel have been created by adding the corresponding CAR sequences to the reference genome (GRCh38, Ensembl release 98) and GENCODE v32 annotation using the mkref function from Cellranger v7.1.0. The Cell Ranger multi pipeline was used to process the single-cell gene expression, ADT (antibody-derived tag) expression, and V(D)J libraries. Since some of the libraries in sequencing run 3 consist of Ide and Cilta patients, Cell Ranger was performed once with Ide and once with the custom Cilta reference. The feature reference CSV file, declaring antibody capture constructs and associated barcodes is provided in Supplementary Table 7-9 . The reference dataset required for the V(D)J contigs is provided by 10X Genomics (GRCh38 Human V(D)J Reference - 7.1.0 (December 7, 2022)). Supplementary Table 4 summarizes the quality control data for each 10x library. For sequencing runs 2-4, we used the R package souporcell v.2.0 45 with parameter k = 2 to demultiplex the samples in each 10X scRNA-seq library. This resulted in two cell clusters based on common variants for each sample. For sequencing run 2, where anti-human hashtag antibodies were used prior to pooling, clusters were associated with patient IDs based on the mean hashtag expression values per cluster. Since one male and one female donor per multplexed 10X scRNA-seq library were used for runs 3 and 4, the mean expression of the sex-specific genes XIST and RPS4Y1 was used to associate the clusters with the respective sample ID. Supplementary Table 3 provides an overview of whether and which samples were multiplexed. For all subsequent analyses, R v4.3.2 was used. As suggested by Neavin et al. 46 , we used the R packages scDblFinder v1.16.0 47 and the cxds_bcds_hybrid models from the package scds v1.18.0 48 in combination to remove barcodes (cells) that contained potential doublets. In addition, cell barcodes that met any of the following criteria were excluded: (1) fewer than 250 genes; (2) more than 8,000 genes; (3) fewer than 1,000 UMIs; (4) more than 10,000 UMIs; or (5) mitochondrial transcripts fraction more than 15%; and (6) cell complexity less than 0.8. Complexity was defined as follows: log10(nbr. of present genes)/log10(nbr. of UMIs). The numbers of cells for each sample before and after the filtering steps are listed in Supplementary Table 3 . Raw gene expression data was normalized using the LogNormalize method of the NormalizeData function in Seurat v5.1.0 49 . ADT data was normalized by the centered log-ratio transformation (CLR) with parameters normalization.method = "CLR" and margin = 2 using the NormalizeData function. Quality control showed a correlation between the number of genes/UMIs detected and the sequencing runs (In Extended Data Fig. 2b and c ). Further evaluation of the potential batch effect between sequencing runs at cell type level showed no significant correlation except for plasma and B cells ( Extended Data Fig. 2d-j ), caused by the fact that the plasma cells in a batch essentially originate from the same patient ( Fig. 6c and Extended Data Fig. 9b ). Annotation of cell identities Transfer learning was performed to annotate the PBMC cells. To this end, we used the PBMC reference dataset form Butler et al. 50 with the RunAzimuth function from the R package Azimuth v0.5.0 49 . Only cells with a confidence score greater than 0.5 were retained for downstream analysis. As a secondary cell type annotation strategy, we used the R package scGate v1.6.0 51 , which requires (i) a normalized gene expression matrix and (ii) a 'gating model' (GM) consisting of gene sets representing the cell populations of interest. The GM for T cells, gamma delta T cells, B cells, plasma cells, NK cells and myeloid cells were obtained using the get_scGateDB function. Cells that did not have matching cell type annotations using both the Azimuth and scGate methods were removed. Both annotation methods were applied per sample. Cells were also excluded if a VDJ sequence was found by TCR or BCR sequencing but were not classified as T or B cells. In addition, cells annotated as erythrocytes, erythroid progenitor cells, and platelets were excluded from subsequent analysis. The PBMC reference dataset used includes a coarse annotation of T cell subtypes. To cover a broader range of the T cell compartment, we re-annotated the T cells using the R package ProjecTILs v3.3.1 13 . To classify CD4 and CD8 T cell subtypes, we projected the two T cell lineages separately onto two reference maps. Here, we used the single-cell atlases for CD4 and CD8 cells from the SPICA database (https://spica.unil.ch/), which are part of the ProjecTILs framework. Before applying transfer learning strategy by ProjecTILs, we first annotated CD4 and CD8 T cell lineages based on the gene expression data using the following strategy: Given the raw counts, one cell was considered as CD8 positive or negative if the count value of CD8A or CD8B was >0 or ≤0, respectively. One cell was considered as CD4 positive or negative if the count value of CD4 was >0 or ≤0, respectively. This resulted in a classification of CD8-CD4+, CD8+CD4-, CD8+CD4+, and CD8-CD4- T cells. Double positive cells (CD8+CD4+) were removed from downstream analysis. In addition, based on the TCR-Seq data, clones consisting of CD4 and CD8 cells were removed (453 cells in total). We also excluded 36 clones (188 cells in total) from downstream analysis that were not donor specific. About 23% of all CD8-CD4- cells (total = 35,784) could be assigned to either CD4 or CD8 clones. For all remaining CD4-CD8- T cells, the following strategy was applied: For all T cells, we computed a nearest-neighbor graph using k=10 and then calculated the smoothed expression of CD4 and CD8B by averaging across the 10 nearest neighbors of each cell using SmoothKNN from the R package UCell v2.6.2 52 . The remaining CD4-CD8- T cells were then classified by gating on these expression values. The thresholds of the smoothed expression values used for classification were 0.2 for CD4 and 0.3 for CD8A . CD-CD8- cells >0.2 & 0.3 & <0.2 were annotated as CD4 or CD8 cells, respectively. Using this strategy, 70% of all CD8-CD4- cells could be assigned to the CD4 or CD8 lineages. Lastly, we aimed to annotate the remaining 8350 CD4-CD8- cells using ADT-Seq. The threshold of the normalized ADT expression values used for classification was 1 for CD4 and CD8A. CD4-CD8- cells >1 & 1 & <1 were annotated as CD4 or CD8 cells. A total of 2320 of the 8350 CD4-CD8- cells could be assigned to the CD4 or CD8 lineages. Gamma Delta T cells were annotated using the scGate method with the gating model “gdT” from the function get_scGateDB. All T cells annotated as CD4 or CD8 were used as input for the ProjecTILs.classifier function with the corresponding CD4 and CD8 atlas as reference. To identify proliferating cells, we applied the run_gsea function implemented in the clustifyr R package v1.14 53 using previously published G2/M and S-phase gene sets from the Seurat package as queries. As suggested by the authors, an over-clustering for each sample was performed. For this purpose, the cluster resolution was set to 1. For gene set enrichment 1,000 permutations were performed. One cell cluster was significantly (p-value <0.05) enriched with cell cycle genes from the G2/M and S phases and was annotated as cyclic accordingly. Marker genes for all estimated T-cell identities are shown in Extended Data Fig. 3 . Integration and dimension reduction Integration and clustering analysis were performed using Seurat v5.1.0. For each sample, the 2,000 most variable genes were estimated using the "vst" method of the FindVariableFeatures function. Mitochondrial, Immunoglobulin variable chain, T cell receptor and sex-specific genes were not included in the estimation of variable genes. We then used the SelectIntegrationFeatures function to select 2,000 genes that were repeatedly shown to be highly variable in multiple samples. Thereafter, the data were standardized with ScaleData, and principal component analysis (PCA) was performed on the standardized expression data of the variable genes using the function RunPCA. We used 30 principal components, which explained approx. 90% of the variance, to integrate all samples using the Harmony 54 method (v1.2) with the Seurat wrapper function RunHarmony with additional parameters (dim.use = 1:30, group.by.vars = c("PATIENT_ID")). Using the Harmony-corrected cell embeddings, we performed clustering by computing a shared nearest neighbors (SNN) graph, as implemented in FindNeighbors(reduction = "harmony", dims = 1:30). Cluster identification using the SNN graph was computed by the FindCluster function (Louvain algorithm). The same strategy was also used for integration of the B and T cells. Here, 1,000 variable genes and 15 principal components were used. Differential gene expression and enrichment analysis Differentially expressed genes between groups were identified by the FindMarkers function implemented in Seurat, using the MAST algorithm 55 . When fitting the hurdle model, in addition to the covariates of interest both sequencing run as well as the number of detected genes per cell were used to reduce false positives due to different sequencing time points or mRNA capture rates (see Extended Data Fig. 2a-j) . In Fig. 3i, 5d and g , clinical outcome was also used as further covariate. To retain only genes that were robustly expressed, differential gene expression was calculated only for genes that were expressed in at least 20% of cells in one analyzed group (parameter: min.pct = 0.20). Only cell types with at least 200 cells in both groups were analyzed. To minimize sample bias due to high cell counts, the cells of each sample and each cell type were downsampled to 250 (100 cells for Fig. 3i, 5d and g ). Genes with an adjusted p-value <0.05 (Bonferroni correction) and an absolute fold change of 1.25 were considered as significantly differentially expressed (DE) genes. In Fig. 3i and 5e , we have used custom gene sets that functionally characterize CAR-T cells as previously defined by Anderson and colleagues 56 . DE genes were further used for overrepresentation analysis based on hypergeometric test analysis with the fgsea package v1.28.0 57 . GO terms for biological processes from the MSigDB 58 and curated gene signatures for T cells from Chu et al. 59 were used as references. For GO terms, the mgoSim function of the R package GOSemSim v2.28.0 60 was used to remove redundant terms. Terms and T cell signatures with adjusted p-value <0.05 (Benjamini Hochberg correction) were considered significant. To depict T cell-relevant GO terms in Fig. 3b and 5h , we filtered for metabolism-, T cell- and immune-relevant terms. Significant terms were ranked by rich factor, which is the number of DE genes in the term divided by the number of background genes in that term. In addition, a pathway direction score was calculated as the number of DE genes with log2FC >0 minus the number of DE genes with log2FC <0 divided by the square root of the number of genes associated with the pathway. Differences in cell type proportions We used the propeller method 61 , as implemented in the speckle R package v1.20, to test for differences in cell type proportions between groups. To filter out samples with low cell type complexity, samples consisting of less than 10 cell types were removed from this analysis. In addition, only cell types with >100 cells in one group were used for downstream analysis. Cell type proportions were transformed with arcsin square root transformation function implemented in speckle. A linear model was fitted for each cell type with group, encoding CRS ( Fig 4d ) or product ( Fig 5b ) as the explanatory variable. For Fig 5c , the linear model was fitted using the formula: ~block + contrast, where block encoded the patient (to account for patient‐specific differences in cell type proportion) and contrast containing the group information. Statistical significance was estimated using an empirical Bayes moderated t-test. In case a group for a cell type contained 0 cells and therefore the log2 fold change could not be calculated, the minimum or maximum log2 fold change of all other comparisons was used as a replacement. Detection of polyfunctional cells The RNA expression data was used to determine whether a cell could be associated with multiple functional classifications of secreted cytokines (regulatory, effector, stimulatory, chemoattractive and inflammatory). Each functional group is determined by a list of genes associated as outlined in the study by Rossi et al. 62 . A cell was linked to a functional group whenever the expression of at least one gene in the corresponding list was higher than the median of non-zero counts for that given gene across the dataset. Whenever a cell was assigned to more than one functional group, the cell was classified as polyfunctional, and the fraction of polyfunctional cells was calculated for each patient and time point. Assessing the TCR/BCR repertoire using V(D)J analysis V(D)J data were analyzed using scRepertoire v.2.0.0 63 . Clonotypes were identified using the combineTCR and combineBCR functions. A TCR clonotype is defined as the combination of genes comprising the TCR and the nucleotide sequence of the CDR3 region (gene + nucleotide) for paired TCR alpha and beta chains. BCR clonotypes based on the V gene and >85% normalized Levenshtein distance of the nucleotide sequence. Clonality in Fig. 3g was defined as the complement of evenness (normalized Shannon entropy) as previously described by Zhang and colleagues 64 . The evenness value lies between 0 and 1, with a high value indicating a more equal distribution of TCRs and a low value indicating TCR skewing due to clonal expansion. Clonality, which reflects the dominance of particular clones across the TCR repertoire, was calculated per sample and time point. For CD8 and CD4 cells, the R package STARTRAC v0.10.0 64 was applied to calculate clonality (referred to by STARTRAC as expansion index). In Fig. 4e , the average enrichment scores for cytotoxicity-associated genes ( PRF1 , GNLY , GZMB , GZMH , GZMK and NKG7) were calculated for each clonotype using the R package TCRanker (https://github.com/carmonalab/TCRanker). Trajectory inference analysis Analysis was performed separately for CD8+ and CD4+ cells, excluding gamma delta and proliferating T cells. PCA-based dimensionality reduction was performed with variable genes, followed by integration (see section Integration and dimension reduction ). To model the state transition between cells, the DiffusionMap function from the destiny R-package v3.16.0 65 were used with the harmony-corrected cell embeddings as input. The function slingshot from the R package slingshot v.2.10.0 66 was applied with the first two diffusion components to calculate the trajectories (referred to by slingshot as lineages). Naive T cells cluster are considered as a root for estimation of the trajectories. Cells were grouped into cell clusters according to the nearest node along the trajectories. Cell-Cell communication For RNA expression data, ligand-receptor signaling was inferred with iTalk 51 using default settings. LIANA v0.1.12 52 was used as a reference database consisting of manually curated resources in the context of cell-cell communication. All DE genes (non-CR vs CR) in pre-infusional PBMC were used as input. Supervised pathway deconvolution Supervised non-matrix factorization was performed for all T cells using the Spectra v0.1.0 python package 25 . For each sample, the 5,000 most variable genes were estimated using the "vst" method of the FindVariableFeatures function implement in Seurat. Mitochondrial, Immunoglobulin variable chain and T cell receptor genes were not included in the estimation of variable genes. We then used the SelectIntegrationFeatures function to select 3,000 genes that were repeatedly shown to be highly variable in multiple samples. As suggested by the authors from Spectra, the RNA expression values were normalized using the logNormCounts function from the Scran v1.30 package 67 , which uses the outputs of the quickCluster function followed by computeSumFactors. Global and T cell type specific gene programs were obtained from Spectra using the function default_gene_sets.load. Factor (gene program) modeling was performed using the est_spectra function with hyperparameter lam = 0.01 (other parameters are default settings). Single-cell factor scores were then extracted from the model results. Differentially expressed factors between two groups were estimated using the FindMarkers function implemented in Seurat (two-sided Wilcoxon rank-sum test). To minimize sample bias due to high cell counts, the cells of each sample and each cell type were downsampled to 250. To analyze only factors that were robustly enriched, factors that had a loading value of >0.001 in at least 25% of the cells in one analyzed group were retained. Factors with an adjusted p-value <0.05 (Bonferroni correction) and an absolute fold change of 1.25 were considered as significantly differentially enriched factors. Statistics and reproducibility Statistical analysis was performed in R v.4.3.2. Since the scRNA-seq data conforms to a negative binomial distribution, comparisons between groups on gene expression level were carried out with the MAST method which is based on a zero inflated negative binomial model. For testing differences in the cell type proportions, a normal distribution of the data was not assumed. Using the speckle R package, an arcsin square root transformation was performed. Significant differences were estimated using the empirical Bayes moderated t-statistics (two-sided) implemented in the speckle package. Gene ontology term enrichment was calculated using a hypergeometric test. Estimated probabilities of PFS were calculated using the Kaplan–Meier method and the log-rank test to evaluate differences between survival distributions. Each sample was considered a biological replicate, and no technical replicates were performed in this study. All statistical tests are listed in the corresponding method sections and figure legends for clarity. Declarations Acknowledgement This work was supported by the CERTAINTY project funded by the European Union (Grant Agreement 101136379). Views and opinions expressed are those of the author(s) only and do not necessarily reflect those of the European Union or the Health and Digital Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. M. Merz is further supported by a Translational Research Award from the International Myeloma Society and the Deutsche Jose Carreras Leukämie Stiftung. Additionally, M. Merz has received research funding from Janssen, SpringWorks, and Roche/Genentech. M. Merz, H. Weidner and L. C. Hofbauer were supported by the German Research Foundation (µbone, SPP2084). K. Reiche and U. Köhl are further supported by imSAVAR (Innovative Medicine Initiative 2 Joint Undertaking No 853988), T2Evolve (Innovative Medicine Initiative 2 Joint Undertaking No 945393), SaxoCell (BMBF Clusters4Future), DAAD project 57616814 (SECAI, School of Embedded Composite AI) and the German José-Carreras Leukemia Foundation (DJCLS 08 R/2023). V. Vucinic is supported by SaxoCell (ECP-CAR, BMBF Clusters4Future). We would like to extend our gratitude to S. Scharf from the Leipzig Medical Biobank for handling cryopreserved samples and D. Bretschneider, C. Mueller, K. Wildenberger and C. Wilhelm for performing fluorescence in situ hybridization analyses on clinical samples as well as C. Hertel for handling sample shipment. In addition, we thank Andreas Schmidt and Regina Ohmer from Singleron Biotechnologies for data management as well as D. Dudziak for her input regarding pDCs. Most importantly, we express our deepest thanks to the patients for their invaluable contribution. Authors' Contributions M.M., M.R, D.F., M.K. K.R. designed the experiments, supervised the work, contributed equally to writing, and reviewed and edited the final paper. L.F., S.S., T.W., P.B., H.W., L.C.H., R.B., S.Y.W, E.B., S.H, K.M., J.S., M.H., C.H., M.J., G-N.F, A.B., M.F., U.K., U.P., V.V. and M.M. were responsible for acquisition of data (acquired and managed patients, provided facilities, FISH, Flow cytometry, biobanking, in vitro studies etc.) M.R., D.F., M.K., F.K., P.B., Z.S., J.K. and L.F performed analyses and interpretation of data (e.g., statistical analysis, computational analysis). M.R., D.F. and M.K. wrote the original draft which was reviewed and revised by K.R. and M.M. All authors have substantively revised the paper and approved the submitted version. Competing interests MM gave advisory boards and received honoraria and research support from Amgen, BMS, Celgene, Gilead, Janssen, Stemline, Springworks, Sanofi, and Takeda MH gave advisory boards and received honoraria from Abbvie, Beigene, Jazz, Janssen, Stemline Menarini and Takeda, and received research support from EDO-Mundipharma, Janpix, Novartis, and Roche. S.F.: consultant and/or speaker fees from Novartis Pharma, Janssen-Cilag, Vertex Pharmaceuticals (Germany), Kite/Gilead Sciences, MSGO and Bristol-Myers Squibb. U.K.: consultant and/or speaker fees from AstraZeneca, Affimed, Glycostem, GammaDelta, Zelluna, Miltenyi Biotec and Novartis Pharma and Bristol-Myers Squibb. K.H.M.: BMS (consultancy and honoraria), AbbVie (honoraria, research funding), Pfizer (honoraria), Otsuka (honoraria), Janssen (honoraria) and Novartis (consultancy). U.P.: Syros (consultancy, honoraria, research funding), MDS Foundation (membership on an entity’s Board of Directors or advisory committees), Silence Therapeutics (consultancy, honoraria, research funding), Celgene (honoraria), Takeda (consultancy, honoraria, research funding), Fibrogen (research funding), Servier (consultancy, honoraria, research funding), Roche (research funding), Merck (research funding), Amgen (consultancy, research funding), Novartis (consultancy, honoraria, research funding), AbbVie (consultancy), Curis (consultancy, research funding), Janssen Biotech (consultancy, research funding), Jazz (consultancy, honoraria, research funding), BeiGene (research funding), Geron (consultancy, research funding) and Bristol-Myers Squibb (consultancy, honoraria, membership on an entity’s Board of Directors or advisory committees, other, travel support, medical writing support, research funding). M.J.: Novartis (honoraria), Amgen (honoraria), Pfizer (honoraria), Blueprint Medicine (honoraria), BMS (honoraria) and Jazz (honoraria). L. C. H has received funding for clinical trials from Alexion, Amolyte, and Ascendis to his institution, and honoraria from Ascendis, Amgen, and UCB for consultancies and adboards to himself. VV gave advisory boards for Janssen Cilag, BMS Celgene, MSD, Novartis, Sobi, Caribou and received honoraria from Novartis, Gilead Kite, BMS Celgene, Janssen Cilag, Sobi, Amgen, Abbvie, Takeda. All other authors declare no competing interests. Data availability The single-cell sequencing data (RNA, BCR, TCR and ADT) supporting the study's findings have been deposited in the Gene Expression Omnibus under accession code GSE234261 7 for sequencing run 1. Raw data and Cell ranger outputs for sequencing run 2-4 and pre-processed Seurat objects are available in the European Genome-Phenome Archive (accession code EGAXXXXXXXX) under restricted access. Code availability Processing and analysis code related to this study is deposited in a GitHub repository at https://github.com/fraunhofer-izi/Rade_et_al_CAR_2025 References Swan, D., Madduri, D. & Hocking, J. CAR-T cell therapy in Multiple Myeloma: current status and future challenges. Blood cancer journal 14, 206; 10.1038/s41408-024-01191-8 (2024). Munshi, N. C. et al. Idecabtagene Vicleucel in Relapsed and Refractory Multiple Myeloma. The New England journal of medicine 384, 705–716; 10.1056/NEJMoa2024850 (2021). Berdeja, J. G. et al. Ciltacabtagene autoleucel, a B-cell maturation antigen-directed chimeric antigen receptor T-cell therapy in patients with relapsed or refractory multiple myeloma (CARTITUDE-1): a phase 1b/2 open-label study. Lancet (London, England) 398, 314–324; 10.1016/S0140-6736(21)00933-8 (2021). Cordas Dos Santos, D. M. et al. A systematic review and meta-analysis of nonrelapse mortality after CAR T cell therapy. Nature medicine 30, 2667–2678; 10.1038/s41591-024-03084-6 (2024). Merz, M. et al. Idecabtagene vicleucel or ciltacabtagene autoleucel for relapsed or refractory multiple myeloma: An international multicenter study. HemaSphere 9, e70070; 10.1002/hem3.70070 (2025). Sun, Y. et al. Antigen-induced chimeric antigen receptor multimerization amplifies on-tumor cytotoxicity. Signal transduction and targeted therapy 8, 445; 10.1038/s41392-023-01686-z (2023). Rade, M. et al. Single-cell multiomic dissection of response and resistance to chimeric antigen receptor T cells against BCMA in relapsed multiple myeloma. Nature cancer 5, 1318–1333; 10.1038/s43018-024-00763-8 (2024). Haradhvala, N. J. et al. Distinct cellular dynamics associated with response to CAR-T therapy for refractory B cell lymphoma. Nature medicine 28, 1848–1859; 10.1038/s41591-022-01959-0 (2022). Qin, C. et al. Single-cell analysis of anti-BCMA CAR T cell therapy in patients with central nervous system autoimmunity. Science immunology 9, eadj9730; 10.1126/sciimmunol.adj9730 (2024). Tian, D.-S. et al. B cell lineage reconstitution underlies CAR-T cell therapeutic efficacy in patients with refractory myasthenia gravis. EMBO molecular medicine 16, 966–987; 10.1038/s44321-024-00043-z (2024). Hansen, D. K. et al. Comparison of Standard-of-Care Idecabtagene Vicleucel and Ciltacabtagene Autoleucel in Relapsed/Refractory Multiple Myeloma. Journal of clinical oncology : official journal of the American Society of Clinical Oncology, JCO2401730; 10.1200/JCO-24-01730 (2025). Fandrei, D. et al. Bispecific antibodies as bridging to BCMA CAR-T cell therapy for relapsed/refractory multiple myeloma. Blood cancer discovery ; 10.1158/2643-3230.BCD-24-0118 (2024). Andreatta, M. et al. Interpretation of T cell states from single-cell transcriptomics data using reference atlases. Nature communications 12, 2965; 10.1038/s41467-021-23324-4 (2021). Kharchenko, P. V., Silberstein, L. & Scadden, D. T. Bayesian approach to single-cell differential expression analysis. Nature methods 11, 740–742; 10.1038/nmeth.2967 (2014). Mensurado, S., Blanco-Domínguez, R. & Silva-Santos, B. The emerging roles of γδ T cells in cancer immunotherapy. Nature reviews. Clinical oncology 20, 178–191; 10.1038/s41571-022-00722-1 (2023). Kang, J. et al. Systematic dissection of tumor-normal single-cell ecosystems across a thousand tumors of 30 cancer types. Nature communications 15, 4067; 10.1038/s41467-024-48310-4 (2024). Wu, H. et al. Mitochondrial dysfunction promotes the transition of precursor to terminally exhausted T cells through HIF-1α-mediated glycolytic reprogramming. Nature communications 14, 6858; 10.1038/s41467-023-42634-3 (2023). Yu, Y.-R. et al. Disturbed mitochondrial dynamics in CD8+ TILs reinforce T cell exhaustion. Nature immunology 21, 1540–1551; 10.1038/s41590-020-0793-3 (2020). Huang, Y. et al. Inhibition of CD38 enzymatic activity enhances CAR-T cell immune-therapeutic efficacy by repressing glycolytic metabolism. Cell reports. Medicine 5, 101400; 10.1016/j.xcrm.2024.101400 (2024). Wagner, T. R. et al. Tumor Reactive T Cells Drive Clinical Responses to Multiple Myeloma. Blood 144, 1890; 10.1182/blood-2024-210272 (2024). Liu, T., Zhang, L., Joo, D. & Sun, S.-C. NF-κB signaling in inflammation. Signal transduction and targeted therapy 2, 17023-; 10.1038/sigtrans.2017.23 (2017). Hendriks, J. et al. CD27 is required for generation and long-term maintenance of T cell immunity. Nature immunology 1, 433–440; 10.1038/80877 (2000). Jaeger-Ruckstuhl, C. A. et al. Signaling via a CD27-TRAF2-SHP-1 axis during naive T cell activation promotes memory-associated gene regulatory networks. Immunity 57, 287-302.e12; 10.1016/j.immuni.2024.01.011 (2024). Cibrián, D. & Sánchez-Madrid, F. CD69: from activation marker to metabolic gatekeeper. European journal of immunology 47, 946–953; 10.1002/eji.201646837 (2017). Kunes, R. Z., Walle, T., Land, M., Nawy, T. & Pe'er, D. Supervised discovery of interpretable gene programs from single-cell data. Nature biotechnology 42, 1084–1095; 10.1038/s41587-023-01940-3 (2024). Neitzke-Montinelli, V. et al. Differentiation of Memory CD8 T Cells Unravel Gene Expression Pattern Common to Effector and Memory Precursors. Frontiers in immunology 13, 840203; 10.3389/fimmu.2022.840203 (2022). Lee, H. et al. Impact of soluble BCMA and non-T-cell factors on refractoriness to BCMA-targeting T-cell engagers in multiple myeloma. Blood 144, 2637–2651; 10.1182/blood.2024026212 (2024). Freeman, C. L. et al. Tumor burden quantified by Soluble B-Cell Maturation Antigen and Metabolic Tumor Volume determine myeloma CAR-T outcomes. Blood ; 10.1182/blood.2024024965 (2024). Hansen, D. K. et al. Idecabtagene Vicleucel for Relapsed/Refractory Multiple Myeloma: Real-World Experience From the Myeloma CAR T Consortium. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 41, 2087–2097; 10.1200/JCO.22.01365 (2023). Sidana, S. et al. Safety and efficacy of standard-of-care ciltacabtagene autoleucel for relapsed/refractory multiple myeloma. Blood 145, 85–97; 10.1182/blood.2024025945 (2025). Gagelmann, N. et al. Development and Validation of a Prediction Model of Outcome After B-Cell Maturation Antigen-Directed Chimeric Antigen Receptor T-Cell Therapy in Relapsed/Refractory Multiple Myeloma. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 42, 1665–1675; 10.1200/JCO.23.02232 (2024). Furqan, F. et al. Outpatient administration of CAR T-cell therapies using a strategy of no remote monitoring and early CRS intervention. Blood advances 8, 4320–4329; 10.1182/bloodadvances.2024013239 (2024). Kohler, M. E. & Fry, T. J. CD4+ CAR T cells - more than helpers. Nature cancer 4, 928–929; 10.1038/s43018-023-00567-2 (2023). Melenhorst, J. J. et al. Decade-long leukaemia remissions with persistence of CD4+ CAR T cells. Nature 602, 503–509; 10.1038/s41586-021-04390-6 (2022). Lee, H. et al. Mechanisms of antigen escape from BCMA- or GPRC5D-targeted immunotherapies in multiple myeloma. Nature medicine 29, 2295–2306; 10.1038/s41591-023-02491-5 (2023). Majzner, R. G. et al. Tuning the Antigen Density Requirement for CAR T-cell Activity. Cancer discovery 10, 702–723; 10.1158/2159-8290.CD-19-0945 (2020). Karnell, J. L. et al. Depleting plasmacytoid dendritic cells reduces local type I interferon responses and disease activity in patients with cutaneous lupus. Science translational medicine 13 ; 10.1126/scitranslmed.abf8442 (2021). Griffin, G. K. et al. Ultraviolet radiation shapes dendritic cell leukaemia transformation in the skin. Nature 618, 834–841; 10.1038/s41586-023-06156-8 (2023). Kumar, S. et al. International Myeloma Working Group consensus criteria for response and minimal residual disease assessment in multiple myeloma. The Lancet. Oncology 17, e328-e346; 10.1016/S1470-2045(16)30206-6 (2016). Lee, D. W. et al. ASTCT Consensus Grading for Cytokine Release Syndrome and Neurologic Toxicity Associated with Immune Effector Cells. Biology of blood and marrow transplantation : journal of the American Society for Blood and Marrow Transplantation 25, 625–638; 10.1016/j.bbmt.2018.12.758 (2019). Blache, U. et al. Advanced Flow Cytometry Assays for Immune Monitoring of CAR-T Cell Applications. Frontiers in immunology 12, 658314; 10.3389/fimmu.2021.658314 (2021). Zheng, L. et al. Pan-cancer single-cell landscape of tumor-infiltrating T cells. Science (New York, N.Y.) 374, abe6474; 10.1126/science.abe6474 (2021). Campbell, T. et al. Uses of anti-BCMA chimeric antigen receptors. Worldwide patent WO2021091978A1 (2021). , Schecter, J. M. & Fan, X. BCMA-targeted CAR-T cell therapy for multiple myeloma. Worldwide patent WO2022116086A1 (2022). , Heaton, H. et al. Souporcell: robust clustering of single-cell RNA-seq data by genotype without reference genotypes. Nature methods 17, 615–620; 10.1038/s41592-020-0820-1 (2020). Neavin, D. et al. Demuxafy: improvement in droplet assignment by integrating multiple single-cell demultiplexing and doublet detection methods. Genome biology 25, 94; 10.1186/s13059-024-03224-8 (2024). Germain, P.-L., Lun, A., Garcia Meixide, C., Macnair, W. & Robinson, M. D. Doublet identification in single-cell sequencing data using scDblFinder. F1000Research 10, 979; 10.12688/f1000research.73600.2 (2021). Bais, A. S. & Kostka, D. scds: computational annotation of doublets in single-cell RNA sequencing data. Bioinformatics (Oxford, England) 36, 1150–1158; 10.1093/bioinformatics/btz698 (2020). Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573-3587.e29; 10.1016/j.cell.2021.04.048 (2021). Stuart, T. et al. Comprehensive Integration of Single-Cell Data. Cell 177, 1888-1902.e21; 10.1016/j.cell.2019.05.031 (2019). Andreatta, M., Berenstein, A. J. & Carmona, S. J. scGate: marker-based purification of cell types from heterogeneous single-cell RNA-seq datasets. Bioinformatics (Oxford, England) 38, 2642–2644; 10.1093/bioinformatics/btac141 (2022). Andreatta, M. & Carmona, S. J. UCell: Robust and scalable single-cell gene signature scoring. Computational and structural biotechnology journal 19, 3796–3798; 10.1016/j.csbj.2021.06.043 (2021). Fu, R. et al. clustifyr: an R package for automated single-cell RNA sequencing cluster classification. F1000Research 9, 223; 10.12688/f1000research.22969.2 (2020). Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nature methods 16, 1289–1296; 10.1038/s41592-019-0619-0 (2019). Finak, G. et al. MAST: a flexible statistical framework for assessing transcriptional changes and characterizing heterogeneity in single-cell RNA sequencing data. Genome biology 16, 278; 10.1186/s13059-015-0844-5 (2015). Anderson, N. D. et al. Transcriptional signatures associated with persisting CD19 CAR-T cells in children with leukemia. Nature medicine 29, 1700–1709; 10.1038/s41591-023-02415-3 (2023). Korotkevich, G. et al. Fast gene set enrichment analysis (2016). Liberzon, A. et al. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell systems 1, 417–425; 10.1016/j.cels.2015.12.004 (2015). Chu, Y. et al. Pan-cancer T cell atlas links a cellular stress response state to immunotherapy resistance. Nature medicine 29, 1550–1562; 10.1038/s41591-023-02371-y (2023). Yu, G. Gene Ontology Semantic Similarity Analysis Using GOSemSim. Methods in molecular biology (Clifton, N.J.) 2117, 207–215; 10.1007/978-1-0716-0301-7_11 (2020). Phipson, B. et al. propeller: testing for differences in cell type proportions in single cell data. Bioinformatics (Oxford, England) 38, 4720–4726; 10.1093/bioinformatics/btac582 (2022). Rossi, J. et al. Preinfusion polyfunctional anti-CD19 chimeric antigen receptor T cells are associated with clinical outcomes in NHL. Blood 132, 804–814; 10.1182/blood-2018-01-828343 (2018). Borcherding, N., Bormann, N. L. & Kraus, G. scRepertoire: An R-based toolkit for single-cell immune receptor analysis. F1000Research 9, 47; 10.12688/f1000research.22139.2. (2020). Zhang, L. et al. Lineage tracking reveals dynamic relationships of T cells in colorectal cancer. Nature 564, 268–272; 10.1038/s41586-018-0694-x (2018). Angerer, P. et al. destiny: diffusion maps for large-scale single-cell data in R. Bioinformatics (Oxford, England) 32, 1241–1243; 10.1093/bioinformatics/btv715 (2016). Street, K. et al. Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics. BMC genomics 19, 477; 10.1186/s12864-018-4772-0 (2018). Lun, A. T. L., McCarthy, D. J. & Marioni, J. C. A step-by-step workflow for low-level analysis of single-cell RNA-seq data with Bioconductor. F1000Research 5, 2122; 10.12688/f1000research.9501.2 (2016). Additional Declarations Yes there is potential Competing Interest. MM gave advisory boards and received honoraria and research support from Amgen, BMS, Celgene, Gilead, Janssen, Stemline, Springworks, Sanofi, and Takeda MH gave advisory boards and received honoraria from Abbvie, Beigene, Jazz, Janssen, Stemline Menarini and Takeda, and received research support from EDO-Mundipharma, Janpix, Novartis, and Roche. S.F.: consultant and/or speaker fees from Novartis Pharma, Janssen-Cilag, Vertex Pharmaceuticals (Germany), Kite/Gilead Sciences, MSGO and Bristol-Myers Squibb. U.K.: consultant and/or speaker fees from AstraZeneca, Affimed, Glycostem, GammaDelta, Zelluna, Miltenyi Biotec and Novartis Pharma and Bristol-Myers Squibb. K.H.M.: BMS (consultancy and honoraria), AbbVie (honoraria, research funding), Pfizer (honoraria), Otsuka (honoraria), Janssen (honoraria) and Novartis (consultancy). U.P.: Syros (consultancy, honoraria, research funding), MDS Foundation (membership on an entity’s Board of Directors or advisory committees), Silence Therapeutics (consultancy, honoraria, research funding), Celgene (honoraria), Takeda (consultancy, honoraria, research funding), Fibrogen (research funding), Servier (consultancy, honoraria, research funding), Roche (research funding), Merck (research funding), Amgen (consultancy, research funding), Novartis (consultancy, honoraria, research funding), AbbVie (consultancy), Curis (consultancy, research funding), Janssen Biotech (consultancy, research funding), Jazz (consultancy, honoraria, research funding), BeiGene (research funding), Geron (consultancy, research funding) and Bristol-Myers Squibb (consultancy, honoraria, membership on an entity’s Board of Directors or advisory committees, other, travel support, medical writing support, research funding). M.J.: Novartis (honoraria), Amgen (honoraria), Pfizer (honoraria), Blueprint Medicine (honoraria), BMS (honoraria) and Jazz (honoraria). L. C. H has received funding for clinical trials from Alexion, Amolyte, and Ascendis to his institution, and honoraria from Ascendis, Amgen, and UCB for consultancies and adboards to himself. VV gave advisory boards for Janssen Cilag, BMS Celgene, MSD, Novartis, Sobi, Caribou and received honoraria from Novartis, Gilead Kite, BMS Celgene, Janssen Cilag, Sobi, Amgen, Abbvie, Takeda. All other authors declare no competing interests. Supplementary Files SupplementaryInformation.docx Supplementary Information SupplementaryTables.xlsx Supplementary Tables 1-9 ExtendedDataFigures.docx Extended Data Figures NMEDA140070epc.pdf Editorial Policy Checklist NMEDA140070rs.pdf Reporting Summary Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6165798","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":429887184,"identity":"b91211b3-83bd-432a-afcd-8ab920c4d426","order_by":0,"name":"Michael Rade","email":"","orcid":"https://orcid.org/0000-0001-6247-9289","institution":"Fraunhofer Institute for Cell Therapy and Immunology IZI","correspondingAuthor":false,"prefix":"","firstName":"Michael","middleName":"","lastName":"Rade","suffix":""},{"id":429887185,"identity":"6afa6d28-6755-43f4-9bde-819202c6c426","order_by":1,"name":"David Fandrei","email":"","orcid":"https://orcid.org/0009-0006-4014-6030","institution":"Fraunhofer Institute for Cell Therapy and Immunology IZI","correspondingAuthor":false,"prefix":"","firstName":"David","middleName":"","lastName":"Fandrei","suffix":""},{"id":429887186,"identity":"59072536-b81e-478d-af86-2a5b5860f328","order_by":2,"name":"Markus Kreuz","email":"","orcid":"","institution":"Fraunhofer Institute for Cell Therapy and Immunology IZI","correspondingAuthor":false,"prefix":"","firstName":"Markus","middleName":"","lastName":"Kreuz","suffix":""},{"id":429887187,"identity":"600c6290-c882-49c1-bd18-3a8e82a9a150","order_by":3,"name":"Sabine Seiffert","email":"","orcid":"","institution":"Institute for Clinical Immunology, University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Sabine","middleName":"","lastName":"Seiffert","suffix":""},{"id":429887188,"identity":"45868b34-ed38-40b7-a917-63b27a861950","order_by":4,"name":"Thomas Wiemers","email":"","orcid":"https://orcid.org/0009-0000-2053-3857","institution":"University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Thomas","middleName":"","lastName":"Wiemers","suffix":""},{"id":429887189,"identity":"bf807118-0f91-4324-9cea-68659ead27af","order_by":5,"name":"Patrick Born","email":"","orcid":"","institution":"Department of Hematology, Cell Therapy and Hemostaseology, University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Patrick","middleName":"","lastName":"Born","suffix":""},{"id":429887190,"identity":"84d4f6a1-406d-43e8-9a3d-24c062a67a99","order_by":6,"name":"Luise Fischer","email":"","orcid":"https://orcid.org/0009-0006-3619-2958","institution":"Department of Hematology, Cell Therapy and Hemostaseology, University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Luise","middleName":"","lastName":"Fischer","suffix":""},{"id":429887191,"identity":"f079f44e-f5c9-4310-ab9b-09cf07c711ec","order_by":7,"name":"Heike Weidner","email":"","orcid":"","institution":"Division of Endocrinology, Diabetes, and Bone Diseases, Department of Medicine III \u0026 Center for Healthy Aging, Medical","correspondingAuthor":false,"prefix":"","firstName":"Heike","middleName":"","lastName":"Weidner","suffix":""},{"id":429887192,"identity":"f322956e-8c3b-4ddb-8650-3943dcb6eb94","order_by":8,"name":"Lorenz Hofbauer","email":"","orcid":"","institution":"Technische Universität Dresden","correspondingAuthor":false,"prefix":"","firstName":"Lorenz","middleName":"","lastName":"Hofbauer","suffix":""},{"id":429887193,"identity":"b0011878-fec8-404f-940a-9a82b98331ab","order_by":9,"name":"Ronny Baber","email":"","orcid":"","institution":"Leipzig University","correspondingAuthor":false,"prefix":"","firstName":"Ronny","middleName":"","lastName":"Baber","suffix":""},{"id":429887194,"identity":"43e4199a-5e16-47a9-96da-7a7a471e490e","order_by":10,"name":"Song-Yau Wang","email":"","orcid":"","institution":"Department of Hematology, Cell Therapy and Hemostaseology, University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Song-Yau","middleName":"","lastName":"Wang","suffix":""},{"id":429887195,"identity":"bba4b840-6690-4118-bfdb-433479cfacb7","order_by":11,"name":"Enrica Bach","email":"","orcid":"","institution":"Department of Hematology, Cell Therapy and Hemostaseology, University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Enrica","middleName":"","lastName":"Bach","suffix":""},{"id":429887196,"identity":"516ad5a5-83f5-49c0-b8bb-b783c7864908","order_by":12,"name":"Sandra Hoffmann","email":"","orcid":"","institution":"Department of Hematology, Cell Therapy and Hemostaseology, University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Sandra","middleName":"","lastName":"Hoffmann","suffix":""},{"id":429887197,"identity":"14a59946-5854-472f-b3a7-1f9a38844f94","order_by":13,"name":"Jonathan Scolnick","email":"","orcid":"","institution":"Singleron Biotechnologies","correspondingAuthor":false,"prefix":"","firstName":"Jonathan","middleName":"","lastName":"Scolnick","suffix":""},{"id":429887198,"identity":"53377bbc-e130-4f0c-a124-d96a2fc81f53","order_by":14,"name":"Mirco Friedrich","email":"","orcid":"","institution":"Broad Institute of MIT and Harvard","correspondingAuthor":false,"prefix":"","firstName":"Mirco","middleName":"","lastName":"Friedrich","suffix":""},{"id":429887199,"identity":"ed22f854-786e-4bae-939b-ce9476aefa7b","order_by":15,"name":"Farid Keramati","email":"","orcid":"","institution":"Princess Máxima Center for Pediatric Oncology","correspondingAuthor":false,"prefix":"","firstName":"Farid","middleName":"","lastName":"Keramati","suffix":""},{"id":429887200,"identity":"a53244c0-7a68-4800-b14e-bfba77333893","order_by":16,"name":"Peter Brazda","email":"","orcid":"","institution":"Princess Máxima Center for Pediatric Oncology","correspondingAuthor":false,"prefix":"","firstName":"Peter","middleName":"","lastName":"Brazda","suffix":""},{"id":429887201,"identity":"989c73f3-5de6-4d5f-824e-08162ccbbbb6","order_by":17,"name":"Zsolt Sebestyen","email":"","orcid":"https://orcid.org/0000-0001-6184-7676","institution":"University Medical Center Utrecht","correspondingAuthor":false,"prefix":"","firstName":"Zsolt","middleName":"","lastName":"Sebestyen","suffix":""},{"id":429887202,"identity":"61fac05d-5499-459b-a3cd-3a0d7476bca1","order_by":18,"name":"Jurgen Kuball","email":"","orcid":"https://orcid.org/0000-0002-3914-7806","institution":"University Medical Center Utrecht","correspondingAuthor":false,"prefix":"","firstName":"Jurgen","middleName":"","lastName":"Kuball","suffix":""},{"id":429887203,"identity":"90b6311a-f563-4522-9d7e-48d400785a4e","order_by":19,"name":"Klaus Metzeler","email":"","orcid":"https://orcid.org/0000-0003-3920-7490","institution":"University Hospital Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Klaus","middleName":"","lastName":"Metzeler","suffix":""},{"id":429887204,"identity":"866fc6f9-6c35-4884-9791-2fbcd81038d9","order_by":20,"name":"Marco Herling","email":"","orcid":"","institution":"Department of Hematology, Cellular Therapy, Hemostaseology, Infectious Diseases University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Marco","middleName":"","lastName":"Herling","suffix":""},{"id":429887205,"identity":"b7a2d915-1844-41a0-a868-c88b66bd7914","order_by":21,"name":"Carmen Herling","email":"","orcid":"","institution":"Department of Hematology, Cell Therapy and Hemostaseology, University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Carmen","middleName":"","lastName":"Herling","suffix":""},{"id":429887206,"identity":"1449ff1c-81b9-4f80-ba85-5e2a9c08a007","order_by":22,"name":"Madlen Jentzsch","email":"","orcid":"","institution":"Department of Hematology, Cell Therapy and Hemostaseology, University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Madlen","middleName":"","lastName":"Jentzsch","suffix":""},{"id":429887207,"identity":"41c03001-9ea5-49fd-b539-e3d3c2576303","order_by":23,"name":"Georg-Nikolaus Franke","email":"","orcid":"","institution":"Department of Hematology, Cell Therapy and Hemostaseology, University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Georg-Nikolaus","middleName":"","lastName":"Franke","suffix":""},{"id":429887208,"identity":"4bac260a-9506-47b3-8e5b-97309889593e","order_by":24,"name":"Andreas Boldt","email":"","orcid":"","institution":"University of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Andreas","middleName":"","lastName":"Boldt","suffix":""},{"id":429887209,"identity":"c908d099-8dd2-4e25-b909-fa01093ccf05","order_by":25,"name":"Anja Grahnert","email":"","orcid":"","institution":"Institute for Clinical Immunology, University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Anja","middleName":"","lastName":"Grahnert","suffix":""},{"id":429887210,"identity":"1ac5c2c6-0fd9-46d8-8c1b-a36f8bd3dc1e","order_by":26,"name":"Maik Friedrich","email":"","orcid":"","institution":"Institute for Clinical Immunology, University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Maik","middleName":"","lastName":"Friedrich","suffix":""},{"id":429887211,"identity":"d19eca0e-96dc-4e42-8250-c4f626b1e158","order_by":27,"name":"Ulrike Köhl","email":"","orcid":"","institution":"Institute for Clinical Immunology, University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Ulrike","middleName":"","lastName":"Köhl","suffix":""},{"id":429887212,"identity":"b1ba267f-facf-4e52-bf80-a7beec778d64","order_by":28,"name":"Uwe Platzbecker","email":"","orcid":"https://orcid.org/0000-0003-1863-3239","institution":"University of Dresden, Germany","correspondingAuthor":false,"prefix":"","firstName":"Uwe","middleName":"","lastName":"Platzbecker","suffix":""},{"id":429887213,"identity":"94412636-eb43-4bca-9ffe-fa879527bb4d","order_by":29,"name":"Vladan Vucinic","email":"","orcid":"","institution":"Department of Hematology, Cell Therapy and Hemostaseology, University Hospital of Leipzig","correspondingAuthor":false,"prefix":"","firstName":"Vladan","middleName":"","lastName":"Vucinic","suffix":""},{"id":429887214,"identity":"5552dc64-7f90-4fcd-9a60-10d135267701","order_by":30,"name":"Kristin Reiche","email":"","orcid":"https://orcid.org/0000-0002-4452-4872","institution":"Fraunhofer Institute for Cell Therapy and Immunology","correspondingAuthor":false,"prefix":"","firstName":"Kristin","middleName":"","lastName":"Reiche","suffix":""},{"id":429887183,"identity":"98beb49a-a936-4c71-95b6-ce938e6ef822","order_by":31,"name":"Maximilian Merz","email":"data:image/png;base64,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","orcid":"https://orcid.org/0000-0002-2805-5973","institution":"University of Leipzig","correspondingAuthor":true,"prefix":"","firstName":"Maximilian","middleName":"","lastName":"Merz","suffix":""}],"badges":[],"createdAt":"2025-03-05 23:10:34","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6165798/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6165798/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":78735651,"identity":"3dbe5154-a7d9-487c-b58c-883afb44a428","added_by":"auto","created_at":"2025-03-18 08:09:43","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":919129,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStudy overview, summary of treatment and clinical outcomes\u003c/strong\u003e. \u003cstrong\u003ea\u003c/strong\u003e, Schematic of time points and availability of PBMC samples used for single-cell multi-omics and flow cytometry, and serum samples used for BCMA detection by immunoassay. \u003cstrong\u003eb\u003c/strong\u003e, Swim plots of the 34 patients who received idecabtagene vicleucel (Ide-cel) and \u003cstrong\u003ec,\u003c/strong\u003e of the 27 patients who received ciltacabtagene autoleucel (Cilta-cel). The colors of the bars represent evolution of response from leukapheresis to last follow-up. \u003cstrong\u003ed-f, \u003c/strong\u003eKaplan-Meier curves showing the impact of CAR-T product, best response within the first 6 months after infusion and refractoriness to prior lines of therapy on progression-free survival (PFS). The curves were truncated if the number of patients at risk dropped below 2 in one group. Log-rank tests were performed to evaluate probabilities of PFS between two patient groups. The tables show the estimated log hazard ratios (logHR) using a univariate Cox-regression analysis. Statistical significance was estimated using the Wald test. CI = confidence interval. \u003cstrong\u003eg, \u003c/strong\u003eConversion of response from day 0 to best response within 6 months after infusion of Cilta-cel or Ide-cel. \u003cstrong\u003eh,\u003c/strong\u003e Forest plot showing log hazard ratios for PFS and 90% confidence intervals of baseline clinical characteristics adjusted for CAR-T cell product. \u003cstrong\u003ei\u003c/strong\u003e and \u003cstrong\u003ej,\u003c/strong\u003eComparison of CAR-T cell expansion between CR and non-CR patients over time using flow cytometry data from day 0 to 100 (areas marked in gray are extrapolated). The y-axes show the percentage of CAR T cells in the CD3+ T cell compartment and CAR+ CD4/CD8 ratio. Regression lines and 95% confidence intervals were computed with the locally estimated scatterplot smoothing (LOESS) method. Two-sided Wilcoxon rank-sum tests were performed to calculate p-values between the groups at each time point, respectively. P-values \u0026lt;0.1 are shown.\u003c/p\u003e\n\u003cp\u003eAbbreviations: IMWG: International working group of Multiple Myeloma, CR: complete response, PFS: progression-free survival, TCexposed: triple class exposed, TCRRMM: triple class refractory multiple myeloma, PentaRRMM: penta class refractory multiple myeloma, HR cytogenetics: high-risk cytogenetics, EMD: extra-medullary disease\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6165798/v1/69604b03e9418b0a4ce7a04e.jpeg"},{"id":78736176,"identity":"ad287db8-b9a4-45d4-93e3-ac2b3f116be9","added_by":"auto","created_at":"2025-03-18 08:17:43","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":2614160,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLongitudinal profiling of BCMA-targeting CAR T cell therapy in multiple myeloma at single-cell resolution\u003c/strong\u003e. \u003cstrong\u003ea\u003c/strong\u003e, Overview of sample availability for single-cell analysis. \u003cstrong\u003eb\u003c/strong\u003e, Around 450,000 cells of 135 samples were embedded into a two-dimensional space by the Uniform Manifold Approximation and Projection (UMAP) method. Cells are colored according to coarse annotation of the cell identities. UMAP in \u003cstrong\u003ec-f\u003c/strong\u003e color-coded by cell type marker, enrichment score of proliferation markers (for G2/M and S phase), availability of VDJ sequences from TCR/BCR-Seq and mitochondrial transcript fraction. \u003cstrong\u003eg\u003c/strong\u003e, Bar graphs show summary statistics of the composition of cell types for each sample time point. Cell type color codes are consistent with \u003cstrong\u003eb\u003c/strong\u003e. \u003cstrong\u003eh,\u003c/strong\u003e Approx. 60,000 T cells were embedded using the UMAP method. The cells are colored according to T cell subtypes. \u003cstrong\u003ei,\u003c/strong\u003e Spearman correlation and corresponding p-values between the estimated percentages of cell types from flow cytometry and scRNA-Seq for each patient. The base value for regulatory cells are all T cells, for CD8 cells all CD8 T cells (correspondingly for CD4 all CD4 cells). \u003cstrong\u003ej\u003c/strong\u003e, Same analysis as in \u003cstrong\u003ei\u003c/strong\u003e, but for the estimated percentages of CAR+ cells. \u003cstrong\u003ek\u003c/strong\u003e, Kaplan-Meier curves for patients from late time points with scaled cell type proportion \u0026gt; median compared to patients with proportion ≤ median are shown. Log-rank tests were performed to evaluate probabilities of PFS between the two groups. The curves were truncated if the number of patients at risk dropped below 2 in both groups. The colored numbers above the x-axis indicate the number of patients at risk. \u003cstrong\u003el\u003c/strong\u003e, Forest plot of the overall log hazard ratios and corresponding 90% confidence intervals (90% CI) estimated by Cox-regression with standardized cell type fraction on a continuous scale. log hazard ratios with a p-value \u0026lt; 0.1 are highlighted with an asterisk.​ Abbreviations: \u003cstrong\u003eb\u003c/strong\u003e, NK = natural killer; gd T-Cell = gamma delta T cell; Mono = monocyte; cDC = classical dendritic cell; pDC = plasmacytoid dendritic cell. \u003cstrong\u003eh\u003c/strong\u003e, CTL EOMES, eomesodermin-expressing cytotoxic lymphocytes; CTL GNLY, granulysin-expressing cytotoxic lymphocytes; Tfh, follicular helper T cells; Treg, regulatory T cells, CM, central memory T cells; EM, effector memory T cells; EMRA = effector memory T cells re-expressing CD45RA; TEX, exhausted T cells; MAIT, mucosal-associated invariant T cells\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6165798/v1/902873d4704b27a47fc26365.png"},{"id":78735653,"identity":"f14b965d-5d62-4593-a39c-452bce6813f9","added_by":"auto","created_at":"2025-03-18 08:09:43","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":626793,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAnalysis of T cell landscape following CAR T cell infusion.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e, Differential gene expression analysis for T cell subtypes comparing non-CR with CR at late time points. Only subtypes with significant (adjusted p-value \u0026lt; 0.05) changes in gene expression are shown. Shown are the highest ranked (sorted by log2FC) protein-coding DE genes (plus the CAR construct, if DE). \u003cstrong\u003eb\u003c/strong\u003e, Gene ontology term (for biological processes) enrichment analysis for DE genes from \u003cstrong\u003ea\u003c/strong\u003e. Terms are ranked by rich factor, which is the number of DE genes in the term divided by the number of background genes in that term. The dot plot depicts the most enriched GO terms (adjusted p-value \u0026lt; 0.05). The color indicates the term direction, which is the number of DE genes with a log fold change of \u0026gt;0 minus the number of DE genes with a log fold change of \u0026lt;0 divided by the square root of the number of term-associated genes. \u003cstrong\u003ec\u003c/strong\u003e, Diffusion map for CD4+ and CD8+ cells (left). T cells were subjected to trajectory inference modelling using slingshot and subsequently split into nodes across the trajectory. Cells were grouped into cell clusters according to the nearest node along the trajectories (right). The root node is depicted by a dot. For \u003cstrong\u003ec\u003c/strong\u003e and \u003cstrong\u003ed\u003c/strong\u003e, the composition of cell types and clonotype groups is shown as pie charts. For clonotype groups, the cells were divided into groups based on their cell number. \u003cstrong\u003ee\u003c/strong\u003e, Trajectory of the T cells, coupled with the mean cell composition fold change differences (log2 scale) between non-CR and CR patients, depicted in a blue-to-red color scale. Significant differences (p-value \u0026lt;0.05) between groups were estimated using a permutation test in cell composition and are indicated with an asterisk. The numbers indicate the percentage of CAR+ cells in each cluster (the sum is 100). Percentage of clones consisting of one cell (\u003cstrong\u003ef\u003c/strong\u003e) and T-cell clonality (\u003cstrong\u003eg\u003c/strong\u003e) per patient and time point. The base value in \u003cstrong\u003ef\u003c/strong\u003e is the number of CD4 or CD8 cells in each case \u003cstrong\u003eh\u003c/strong\u003e, Subtype composition for singletons and expanded (consisting of \u0026gt;1 cell) clones grouped by time point as well as CAR positive and negative cells. Only cell subtypes with more than 10% are labelled. \u003cstrong\u003ei,\u003c/strong\u003e DGEA comparing CAR+ cells with CAR- cells (both from late time point) or with time of leukapheresis. Bar plots show custom gene modules with DE genes (adjusted p-value \u0026lt; 0.05) that functionally characterize T cells. \u003cstrong\u003ej\u003c/strong\u003e, Violin plots show the enrichment of potentially tumor-reactive T (pTRT) cells in cell types estimated with the UCell tool. \u003cstrong\u003ek\u003c/strong\u003e, Forest plot of the overall log hazard ratios and corresponding 90% confidence intervals (90% CI) estimated by Cox-regression with the fraction of pTRTs per patient and time point on a continuous scale. A cell was defined as pTRT if the standardized enrichment of all estimated enrichment scores per cell was greater than the median. Log hazard ratios with a p-value \u0026lt; 0.1 are highlighted with an asterisk.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6165798/v1/0655cd73efbc45faf2a967bf.png"},{"id":78736950,"identity":"0b65d0ef-6aec-4ebf-ac43-81008c4bc17f","added_by":"auto","created_at":"2025-03-18 08:25:43","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":818057,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDifferences of cellular composition and transcriptome between patients with CRS and non-CRS. a\u003c/strong\u003e, Comparison of the maximum CRP level in serum of patients grouped by CRS grade. \u003cstrong\u003eb\u003c/strong\u003e, Spearman's correlation between max CRP and proportion of CAR T cells within CD3+ T cell compartment, 7 days after infusion. \u003cstrong\u003ec\u003c/strong\u003e, For late time points, the percentage of CAR T cells is compared with the CRS grade. For \u003cstrong\u003ea\u003c/strong\u003e and \u003cstrong\u003ec\u003c/strong\u003e, the statistical significance between the groups was estimated using a two-sided Jonckheere-Terpstra test. \u003cstrong\u003ed\u003c/strong\u003e, Differences in cell type composition between patients with CRS 1 or 2 (LP n = 34; Late n = 31; Very Late n = 23) and without CRS (LP n = 18; Late n = 15; Very Late n = 7). For each time point and cell type, the log2 fold change (log2FC) in mean cell fraction between CRS and non-CRS samples was calculated and color-coded. Significant differences (unadjusted p-values) were estimated using empirical Bayes moderated t-statistics (two sided) implemented in the speckle package. \u003cstrong\u003ee\u003c/strong\u003e, Average enrichment scores for cytotoxicity-associated genes were calculated for each clonotype. For CD4/CD8 cells and each time point, the 50 proportionally largest clonotypes are shown. \u003cstrong\u003ef\u003c/strong\u003e, Polyfunctionality of T cells were measured using a prespecified panel of key immunologically relevant gene markers across following categories: regulatory, effector, stimulatory, chemoattractive and inflammatory. Whenever a cell was assigned to more than one category, the cell was classified as polyfunctional, and the fraction of polyfunctional cells was calculated for each patient. For \u003cstrong\u003ee\u003c/strong\u003e and \u003cstrong\u003ef\u003c/strong\u003e, statistical significance between the two groups was estimated using a two-sided Wilcoxon rank-sum test. \u003cstrong\u003eg\u003c/strong\u003e, Differential gene expression analysis for T cells and late time points comparing CRS with non-CRS patients. Shown are the highest ranked (sorted by log2FC) protein-coding DE genes (plus the CAR construct, if DE). A positive log2 fold change indicates upregulation in patients with CRS. \u003cstrong\u003eh\u003c/strong\u003e, Enrichment analysis for DE genes from all time points for T cells. \u003cstrong\u003ei\u003c/strong\u003e, Dissection of cell–cell interactions based on DE genes from all cell types for LP and late time points. The plot depicts potential ligand–receptor interactions. The triangles show the direction of the log2FC of the DE genes. An upward-pointing triangle means that the gene is upregulated in patients with CRS. \u003cstrong\u003ej\u003c/strong\u003e, For the late time points, supervised pathway deconvolution was performed with the RNA expression data of T cells. Differences in enrichment of factors/gene programs (sets of co-expressed genes) between patients with CRS and without CRS were estimated using a two-sided Wilcoxon rank-sum test (* p = 0.05, ** p = 0.01, *** p = 0.001, p \u0026lt; 0.0001 = ****). Colors are mapped to the 1st and 99th quantile. \u003cstrong\u003ek\u003c/strong\u003e, Gene loading plots are shown for four gene programs. Gene names are sorted in decreasing order of magnitude of their factor (gene program) contribution. UMAP plots show the corresponding cell factor loading values.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-6165798/v1/1326b9a2ef0df6daf1e86085.png"},{"id":78735665,"identity":"d12ddfac-0145-45ec-8d6f-8e63516ad4c7","added_by":"auto","created_at":"2025-03-18 08:09:44","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":543459,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eComparison between patients treated with Cilta-cel and Ide-cel. a,\u003c/strong\u003e Comparison of CAR-T cell expansion between Cilta-cel and Ide-cel over time using flow cytometry data from day 0 to 100. The y-axes show the percentage of CAR T cells in the CD3+ T cell compartment and CAR+ CD4/CD8 ratio. Regression lines and 95% confidence intervals were computed with the LOESS method. Two-sided Wilcoxon rank-sum tests were performed to calculate p-values between the groups at each time point, respectively. P-values \u0026lt;0.1 are shown. \u003cstrong\u003eb,\u003c/strong\u003e Differences in cell type composition between patients treated with Cilta-cel (LP \u003cem\u003en\u003c/em\u003e = 21; Late \u003cem\u003en\u003c/em\u003e = 21; Very Late \u003cem\u003en\u003c/em\u003e = 10) and Ide-cel (LP \u003cem\u003en\u003c/em\u003e= 31; Late \u003cem\u003en\u003c/em\u003e = 25; Very Late \u003cem\u003en\u003c/em\u003e = 20). For each time point and cell type, the log2 fold change (log2FC) in mean cell fraction between Cilta and Ide samples was calculated and color-coded. Significant differences (unadjusted p-values) were estimated using empirical Bayes moderated t-statistics (two sided) implemented in the speckle package. \u003cstrong\u003ec,\u003c/strong\u003eDifferences in the cell type proportions between time points for patients treated with Cilta-cel or Ide-cel are shown (y-axis). P-values were estimated using the speckle package. Here, the design matrix takes the pairing information (Patient ID) into account. Shown are p-values \u0026lt;0.1 \u003cstrong\u003ed,\u003c/strong\u003e Differential gene expression analysis for T cell subtypes from Cilta-cel and Ide-cel samples comparing Late with LP time points. Bar plots show custom gene modules with DE genes (adjusted p-value \u0026lt;0.05) that functionally characterize T cells. \u003cstrong\u003ee,\u003c/strong\u003e DE genes with opposite direction of log2 fold change between Cilta-cel and Ide-cel are shown. \u003cstrong\u003ef,\u003c/strong\u003e Each row in the plot depicts a T cell clone with more than one CAR+ cell over time in at least two time points. The y-axis text indicates from which patient the clones originate. The pie charts show the cell type composition of the clones. The 10 most abundant clonotypes per product and T cell lineages are shown (sorted by relative frequency). \u003cstrong\u003eg,\u003c/strong\u003e DGEA for CD8+CAR+ cells comparing Cilta-cel with Ide-cel. Shown are the highest ranked (sorted by log2FC) protein-coding DE genes with a log2FC \u0026gt; 0 and \u0026lt; 0. A positive log fold change indicates upregulation in patients treated with Cilta-cel. \u003cstrong\u003eh,\u003c/strong\u003e Gene ontology term enrichment analysis of DE genes from \u003cstrong\u003eg\u003c/strong\u003e.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-6165798/v1/d5fba52cd2e91f97a4ca7f14.png"},{"id":78735661,"identity":"d055975f-ed6d-449e-988c-acd91f74d907","added_by":"auto","created_at":"2025-03-18 08:09:43","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":2060291,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLongitudinal characterization of BCMA expressing cells\u003c/strong\u003e. \u003cstrong\u003ea\u003c/strong\u003e, Two-dimensional embedding of B- and plasma cells by UMAP. In the upper panel the cells are colored by cell type and in the lower by the size of the respective B-cell receptor (BCR) clone. \u003cstrong\u003eb\u003c/strong\u003e, Clonality of the BCR repertoire per patient and time point comparing B- and plasma cells using the two-sided Wilcoxon rank-sum test. \u003cstrong\u003ec\u003c/strong\u003e, Distribution of ~13,600 B- and plasma cells among the patients. The ~6,000 circulating plasma cells mainly originate from 5 patients (patients with ≥50 plasma cells across all time points are marked by *). We restricted the plot to patients with ≥100 B-/plasma cells. In \u003cstrong\u003ed\u003c/strong\u003e, cells are color-coded by the expression of therapeutic target genes in MM. For B-cells (\u003cstrong\u003ee\u003c/strong\u003e), pDCs (\u003cstrong\u003ef\u003c/strong\u003e) and plasma cells (\u003cstrong\u003eg\u003c/strong\u003e) comparison of the cell proportions derived from scRNA-seq between CR and non-CR patients separately for each time point. Response groups were compared by Wilcoxon rank-sum test. For plasma cells, cell fraction is shown on the square root scale. \u003cstrong\u003eh\u003c/strong\u003e, the number of B-cells per µl from routine blood counts is compared between CR and non-CR patients. Thin lines represent individual patients. B-cell levels were linearly interpolated between -10 days up to 200 days after CAR-T infusion with a step size of 10 highlighted by circles and compared by Wilcoxon rank-sum test (ns: p\u0026gt;0.05; *: 0.01\u0026lt;p≤0.05; **: p≤0.01). Bold lines represent the mean number of B-cells per response group. \u003cstrong\u003ei\u003c/strong\u003e, comparison of B-cell recovery for patients treated with Cilta-cel and Ide-cel by inverse Kaplan-Meier method. \u003cstrong\u003ej\u003c/strong\u003e, sBCMA concentration (ng/ml) on days -5, 0, 7, 30, and 100 after CAR-T infusion of CR vs non-CR patients. Wilcoxon rank-sum tests were performed for each timepoint (ns: p\u0026gt;0.05; **: 0.001\u0026lt;p≤0.01; ****: p≤0.0001). Regression lines and 95% confidence intervals were computed with the locally estimated scatterplot smoothing (LOESS) method.\u003cstrong\u003e k\u003c/strong\u003e, Spearman correlation of the relative proportions of B-, pDC and plasma cells and sBCMA concentration in ng/ml separately for the time points leukapheresis (LP), late and very late. Non-significant correlations (p\u0026gt;0.05) are marked by gray crosses. Spearman correlation of log2 fold changes of sBCMA reduction (day30 - day0) and maximal CAR T expansion (highest proportion of observed CAR+ cells among all CD3+ cells in percent by flow cytometry) (\u003cstrong\u003el\u003c/strong\u003e) and maximal CRP levels (\u003cstrong\u003em\u003c/strong\u003e). Significance was determined by correlation test and the grade of observed CRS is color coded for each patient.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-6165798/v1/3b7d05c4b6ffbcbad0892798.png"},{"id":80324100,"identity":"32cdbbc6-b76c-4ffb-b5cf-670eef987867","added_by":"auto","created_at":"2025-04-10 14:05:36","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":8993097,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6165798/v1/7c9b61aa-c637-4fcc-869d-319c41521d7c.pdf"},{"id":78735654,"identity":"f288a96b-fea5-4eba-b347-94e00cd3b5b4","added_by":"auto","created_at":"2025-03-18 08:09:43","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":979205,"visible":true,"origin":"","legend":"Supplementary Information","description":"","filename":"SupplementaryInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-6165798/v1/2cb4cab988bd1148c7cc90fa.docx"},{"id":78736173,"identity":"6433b62f-2b03-4d0f-b28d-f1925acb8c6a","added_by":"auto","created_at":"2025-03-18 08:17:43","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":428836,"visible":true,"origin":"","legend":"Supplementary Tables 1-9","description":"","filename":"SupplementaryTables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6165798/v1/05160f9890ea83185a4773da.xlsx"},{"id":78735666,"identity":"aba2b093-06b2-454e-a013-64fd29967bdc","added_by":"auto","created_at":"2025-03-18 08:09:44","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":4918667,"visible":true,"origin":"","legend":"Extended Data Figures","description":"","filename":"ExtendedDataFigures.docx","url":"https://assets-eu.researchsquare.com/files/rs-6165798/v1/ca920ecd615bedcfc4002b5c.docx"},{"id":78735657,"identity":"3e09d120-e1f8-4ae7-ab5a-80d768bdb958","added_by":"auto","created_at":"2025-03-18 08:09:43","extension":"pdf","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":2204302,"visible":true,"origin":"","legend":"Editorial Policy Checklist","description":"","filename":"NMEDA140070epc.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6165798/v1/9d075cd49b14d3aa5637850f.pdf"},{"id":78736951,"identity":"c9305971-9b97-47fd-9a71-1fa8f75dbe82","added_by":"auto","created_at":"2025-03-18 08:25:44","extension":"pdf","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":2255852,"visible":true,"origin":"","legend":"Reporting Summary","description":"","filename":"NMEDA140070rs.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6165798/v1/36a812bae7b02a42641a9118.pdf"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential Competing Interest.\nMM gave advisory boards and received honoraria and research support from Amgen, BMS, Celgene, Gilead, Janssen, Stemline, Springworks, Sanofi, and Takeda MH gave advisory boards and received honoraria from Abbvie, Beigene, Jazz, Janssen, Stemline Menarini and Takeda, and received research support from EDO-Mundipharma, Janpix, Novartis, and Roche. S.F.: consultant and/or speaker fees from Novartis Pharma, Janssen-Cilag, Vertex Pharmaceuticals (Germany), Kite/Gilead Sciences, MSGO and Bristol-Myers Squibb. U.K.: consultant and/or speaker fees from AstraZeneca, Affimed, Glycostem, GammaDelta, Zelluna, Miltenyi Biotec and Novartis Pharma and Bristol-Myers Squibb. K.H.M.: BMS (consultancy and honoraria), AbbVie (honoraria, research funding), Pfizer (honoraria), Otsuka (honoraria), Janssen (honoraria) and Novartis (consultancy). U.P.: Syros (consultancy, honoraria, research funding), MDS Foundation (membership on an entity’s Board of Directors or advisory committees), Silence Therapeutics (consultancy, honoraria, research funding), Celgene (honoraria), Takeda (consultancy, honoraria, research funding), Fibrogen (research funding), Servier (consultancy, honoraria, research funding), Roche (research funding), Merck (research funding), Amgen (consultancy, research funding), Novartis (consultancy, honoraria, research funding), AbbVie (consultancy), Curis (consultancy, research funding), Janssen Biotech (consultancy, research funding), Jazz (consultancy, honoraria, research funding), BeiGene (research funding), Geron (consultancy, research funding) and Bristol-Myers Squibb (consultancy, honoraria, membership on an entity’s Board of Directors or advisory committees, other, travel support, medical writing support, research funding). M.J.: Novartis (honoraria), Amgen (honoraria), Pfizer (honoraria), Blueprint Medicine (honoraria), BMS (honoraria) and Jazz (honoraria). L. C. H has received funding for clinical trials from Alexion, Amolyte, and Ascendis to his institution, and honoraria from Ascendis, Amgen, and UCB for consultancies and adboards to himself. VV gave advisory boards for Janssen Cilag, BMS Celgene, MSD, Novartis, Sobi, Caribou and received honoraria from Novartis, Gilead Kite, BMS Celgene, Janssen Cilag, Sobi, Amgen, Abbvie, Takeda. All other authors declare no competing interests.","formattedTitle":"\u003cp\u003eA longitudinal single-cell atlas to predict outcome and toxicity after BCMA-directed CAR T cell therapy in multiple myeloma\u003c/p\u003e","fulltext":[{"header":"Main","content":"\u003cp\u003eChimeric Antigen Receptor (CAR) T-cell therapies targeting the B-cell maturation antigen (BCMA) on malignant plasma cells have revolutionized the treatment landscape for relapsed and refractory multiple myeloma (RRMM)\u003csup\u003e1\u003c/sup\u003e. For patients who have become refractory to the three major classes of anti-myeloma drugs, i.e. immunomodulatory drugs (IMiDs), proteasome inhibitors (PIs) as well as anti-CD38 antibodies, idecabtagene vicleucel (Ide-cel) and ciltacabtagene autoleucel (Cilta-cel) were approved based on the pivotal KarMMa-1\u003csup\u003e2\u003c/sup\u003e and CARTITUDE-1\u003csup\u003e3\u003c/sup\u003e trials. Notable differences in responses and side effects between Ide-cel and Cilta-cel have been observed: Patients treated with Cilta-cel exhibited more durable responses compared to those receiving Ide-cel in real-world, retrospective analyses. However, there is conflicting evidence on a possible association of Cilta-cel with an increased risk of severe adverse events such as cytokine release syndrome (CRS), neurotoxicity and infectious complications\u003csup\u003e4,5\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWhile both therapies target BCMA, Cilta-cel is unique in having two BCMA-binding domains\u003csup\u003e6\u003c/sup\u003e, which is widely believed to be the primary factor behind its higher efficacy. Despite this key structural difference, the reasons behind the improved efficacy of Cilta-cel and the distinct toxicity profiles between these two therapies remain poorly understood. Elucidating the biological mechanisms underlying efficacy as well as the side effects of BCMA-directed CAR T-cell therapies would therefore be crucial for optimizing their clinical use and improving patient outcomes. Translational single cell transcriptomic and multi-omic studies generated new insights into the factors correlating with both success and failure of autologous CAR T-cell therapies in MM\u003csup\u003e7\u003c/sup\u003e, B cell neoplasia\u003csup\u003e8\u003c/sup\u003e and auto-immune disorders\u003csup\u003e9,10\u003c/sup\u003e. Nevertheless, a comprehensive longitudinal characterization of the immune cell alterations in MM patients receiving BCMA-directed CAR-T cell therapies has not been performed so far.\u003c/p\u003e\n\u003cp\u003eIn this study, we provide a comprehensive multi-omics single-cell dataset of longitudinal immune dynamics following autologous CAR-T cell therapy that will serve as a resource to identify clinically actionable biomarkers for optimizing patient selection and therapy management.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eStudy overview and clinical characteristics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe recruited a real-world cohort of 61 RRMM patients treated with BCMA-directed CAR-T cells at our center. Samples of 57 patients were available for single-cell multi-omics profiling: A total of 135 peripheral blood mononuclear cell (PBMC) samples at 3 different time points were collected for single cell RNA, B-cell receptor (BCR), T-cell receptor (TCR) and surface protein profiling. These included samples on the day of leukapheresis (LP), at the late (day 21\u0026ndash;60, average of 31) and very late (day 79\u0026ndash;169, average of 101) time points after infusion. In addition, flow cytometry (FC) was performed at 6 different time points using 2 different antibody panels. Soluble BCMA (sBCMA) levels were measured by ELISA at 4 time points (\u003cstrong\u003eFig. 1a\u0026nbsp;\u003c/strong\u003esummarizes the study design and sample collection workflow). In total, 27 (44%) patients received Cilta-cel (\u003cstrong\u003eFig. 1b\u003c/strong\u003e) and 34 (56%) Ide-cel (\u003cstrong\u003eFig. 1c\u003c/strong\u003e). For further analyses, patients were categorized based on best response within 6 months after infusion into complete responders (CR) and non-CR. In line with recent observations\u003csup\u003e5,11\u003c/sup\u003e, we observed a higher overall response rate (93% vs 67%) and CR rate (78% vs 38%) with Cilta-cel compared to Ide-cel. There were no significant differences in age, sex, R-ISS stage, refractoriness, type of bridging therapy, remission status at CAR-T cell infusion and percentages of high-risk cytogenetic aberrations at baseline. However, patients who received Cilta-cel had fewer prior lines of treatment (median 6 vs 8, p=0.04) (\u003cstrong\u003eExtended Data Fig 1a\u0026nbsp;\u003c/strong\u003eand\u003cstrong\u003e\u0026nbsp;Supplementary Table 1\u003c/strong\u003e). After a median follow-up of 13 months (range, 2-34), median progression-free survival (PFS) was not reached (n.r.) in the Cilta-cel group vs 6 months (95% CI, 3-n.r.) in the Ide-cel group (p=0.001) (\u003cstrong\u003eFig. 1d\u003c/strong\u003e). Median PFS was not reached in CR patients vs 3 months (range 2-9) in non-CR patients (p\u0026lt;0.001) (\u003cstrong\u003eFig. 1e\u003c/strong\u003e). Triple-class/penta-drug refractory patients had a significantly shorter median PFS with 10 months (4\u0026ndash;n.r.) vs not reached in exposed, but non-refractory patients (p=0.007) (\u003cstrong\u003eFig. 1f\u003c/strong\u003e). Age greater than 65 years at infusion, type of lymphodepletion or BCMA-directed therapy in an earlier treatment line (excluding bridging) did not influence PFS (\u003cstrong\u003eExtended Data Fig. 1b-d\u003c/strong\u003e). Analysis of conversion from baseline to post-infusion response confirmed that deeper remission before CAR-T infusion was associated with superior responses with both products, emphasizing our recent findings on the importance of bridging therapy\u003csup\u003e12\u003c/sup\u003e. Cilta-cel demonstrated superior conversion rates (p=0.02, Wilcoxon rank sum test, see methods), with 70% of patients upstaged from stable or progressive disease (SD/PD) to complete response (CR) and 80% from very good partial response or partial response (VGPR/PR) to CR. In contrast, Ide-cel led to an 18% conversion from SD/PD to CR and a 46% conversion from VGPR/PR to CR, respectively (\u003cstrong\u003eFig. 1g\u003c/strong\u003e). Concurrently, cox proportional hazard regression adjusted for CAR-T product showed that deeper baseline remission status was associated with a trend towards improved PFS (p=0.07) (\u003cstrong\u003eFig. 1h\u003c/strong\u003e). The presence of extramedullary disease (p=0.004) and elevated sBCMA (\u0026gt;107 ng/mL, p=0.017) were independent negative baseline predictors, whereas high-risk cytogenetics did not show a strong association with PFS (p=0.29).\u003c/p\u003e\n\u003cp\u003eAnalyzing CAR-T cell expansion by flow cytometry on days 0, 7, 14, 30 and 100 after infusion revealed\u0026nbsp;an increased peak expansion (c\u003csub\u003emax\u003c/sub\u003e) in CR patients on day 14 compared to non-CR patients, however not statistically significant (p=0.135).\u0026nbsp;The proportion of CAR-T cells was significantly higher in CR patients at days 30 (p=0.007) and 100 (p=0.027), indicating slower contraction compared to non-CR patients. (\u003cstrong\u003eFig. 1i\u0026nbsp;\u003c/strong\u003eand \u003cstrong\u003eSupplementary Table 2\u003c/strong\u003e). Furthermore, CD4+CAR+/CD8+CAR+ ratio was significantly higher in CR vs non-CR patients on days 7 (p\u0026lt;0.001), 14 (p=0.022) and 30 (p=0.004), emphasizing the critical role of CD4+ CAR-T cells in tumor eradication (\u003cstrong\u003eFig. 1j\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLongitudinal single cell landscape of patients treated with anti-BCMA CAR T cells\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFollowing the identification of key clinical factors associated with patient outcomes, we explored the longitudinal peripheral blood immune dynamics encompassing CAR-T cell infusion. \u003cstrong\u003eFig. 2a\u003c/strong\u003e provides an overview of sample availability for single cell multi-omics analyses. After rigorous quality filtering and cell type annotation, approx. 450,000 cells were retained for downstream analyses (median of 3,053 cells per sample, range 675\u0026ndash; 8,478, \u003cstrong\u003eFig. 2a\u003c/strong\u003e). Coarse level cell type annotation, expression of the corresponding canonical marker genes, quality control metrics, and cell type composition for each sample are shown in \u003cstrong\u003eFig. 2b-g\u003c/strong\u003e and \u003cstrong\u003eExtended Data Fig. 2a-k\u003c/strong\u003e (see \u003cstrong\u003eSupplementary Tables 3, 4\u0026nbsp;\u003c/strong\u003efor further quality control metrics). In addition, we used CD4 and CD8 reference atlases\u003csup\u003e13\u003c/sup\u003e to refine the annotation of T cell identities (\u003cstrong\u003eFig. 2h\u0026nbsp;\u003c/strong\u003eand\u003cstrong\u003e\u0026nbsp;Extended Data Fig. 3d,e\u0026nbsp;\u003c/strong\u003efor marker genes). Two distinct clusters of differentiated effector memory T cells re-expressing \u003cem\u003eCD45RA\u003c/em\u003e (EMRA) were identified. The EMRA 1 subset exhibited high expression of\u003cem\u003e\u0026nbsp;THEMIS\u003c/em\u003e and \u003cem\u003eCD5\u003c/em\u003e, whereas the EMRA 2 subset was characterized by elevated levels of \u003cem\u003eIKZF2\u003c/em\u003e and \u003cem\u003eKLRC2\u003c/em\u003e. (\u003cstrong\u003eExtended Data Fig. 3 a-c\u003c/strong\u003e). We observed a significant (p \u0026lt;0.05) Spearman correlation between the estimated proportions of T cell subtypes by flow cytometry and scRNA-seq (\u003cstrong\u003eFig. 2i\u0026nbsp;\u003c/strong\u003eand \u003cstrong\u003eSupplementary Table 5\u003c/strong\u003e), which was also confirmed for CAR\u003csup\u003e+\u003c/sup\u003e T cells (\u003cstrong\u003eFig. 2j\u003c/strong\u003e). However, we noted an underestimation of the expression of the Cilta-cel CAR construct by scRNA-seq, which may be due to drop-out events\u003csup\u003e14\u003c/sup\u003e caused by lower expression of the Cilta-cel compared to the Ide-cel construct (\u003cstrong\u003eExtended Data Fig. 3f\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe analyzed the influence of cell type composition at each time point on survival outcomes: Using Kaplan-Meier analysis, elevated levels of circulating plasma cells (cPC), CD56-bright NK cells, CD4+ T follicular helper cells (Tfh), and plasmacytoid dendritic cells (pDC) were significantly associated with reduced PFS, while increased proportions of both CD8+ EMRA subsets, central memory CD8+ (CM), and granulysin-expressing cytotoxic (CTL) CD4+ T cells (CD4 GNLY) at late time points were associated with improved PFS. (\u003cstrong\u003eFig. 2k\u003c/strong\u003e, see \u003cstrong\u003eExtended Data Fig. 3g\u0026nbsp;\u003c/strong\u003efor all time points). Univariate cox proportional hazard regression using the scaled cell type fractions confirmed these results (estimated log hazard ratios (logHRs) on a continuous scale for each time point are shown in \u003cstrong\u003eFig. 2l).\u0026nbsp;\u003c/strong\u003eImportantly, logHRs remained consistent after adjustment for the CAR-T product (\u003cstrong\u003eExtended Data Fig. 3h\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eAlthough \u0026gamma;\u0026delta; T cells did not show a significant association with PFS, we noted that a median of 82% of \u0026gamma;\u0026delta; T cells expressed the variable \u0026delta; variant V\u0026delta;1 and only 3% expressed V\u0026delta;2 or V\u0026delta;3. This observation is consistent with previous findings that V\u0026delta;1 is more abundant in mucosal tissues and cancer specimens than \u0026gamma;\u0026delta; T cells from healthy individuals\u003csup\u003e15\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eNext, we hypothesized that cell types linked to unfavorable outcomes might express BCMA, making their reduction a surrogate marker of effective tumor cell killing or, alternatively, a factor that disrupts CAR-T cell interactions with malignant PCs. Querying a scRNA-seq atlas\u003csup\u003e16\u003c/sup\u003e spanning tumor and normal cells across 26 tissues revealed that pDC, which had the strongest negative impact on PFS after cPC, exhibited the highest \u003cem\u003eTNFRSF17\u0026nbsp;\u003c/em\u003e(BCMA) specificity and expression among non-plasma and B cell populations (\u003cstrong\u003eSupplementary Fig. 1\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTemporal evolution of the T-cell repertoire\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe analyzed significantly differentially expressed (DE) genes between non-CR and CR in T cell subtypes (\u003cstrong\u003eFig. 3a\u003c/strong\u003e). DE genes (adjusted p-value \u0026lt; 0.05) at the late time point included the CAR construct, which was significantly lower expressed in eomesodermin (\u003cem\u003eEOMES\u003c/em\u003e) expressing CD4 CTL. Gene ontology (GO) enrichment analysis of DE genes revealed upregulation of GO terms associated with regulation of T cell mediated immunity and activation in CD4 CTL subsets, which were enriched with DE genes such as \u003cem\u003eTCF7\u003c/em\u003e and \u003cem\u003eIL7R\u003c/em\u003e. The metabolic pathways OXPHOS and aerobic respiration enriched with mitochondrial genes were decreased in CD8 effector subsets. Recently, mitochondrial dysfunction has been described as a hallmark of T cell dysfunction by driving metabolic reprogramming triggering terminal T cell exhaustion\u003csup\u003e17,18\u003c/sup\u003e. \u003cem\u003eTIGIT\u003c/em\u003e was the only significantly upregulated checkpoint marker in CD8 effector populations (\u003cstrong\u003eFig. 3b,\u0026nbsp;\u003c/strong\u003eDE genes and enrichment analysis for all time points are shown in \u003cstrong\u003eExtended Data Fig. 4\u003c/strong\u003e), and integration with surface protein data revealed higher expression of CD38\u003cem\u003e\u0026nbsp;\u003c/em\u003eon CD8 effector subsets\u003cem\u003e,\u0026nbsp;\u003c/em\u003ewhich was recently identified\u003cem\u003e\u0026nbsp;\u003c/em\u003eto drive terminal CAR-T cell exhaustion\u003csup\u003e19\u003c/sup\u003e (see \u003cstrong\u003eSupplementary Table 6\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eT cells were subjected to trajectory inference modeling and subsequently partitioned into nodes across the trajectory. The root node is depicted by a dot in \u003cstrong\u003eFig 3c\u003c/strong\u003e. Cells were grouped into clusters according to the nearest node along the trajectories (\u003cstrong\u003eFig. 3c-e\u003c/strong\u003e). The T cell compartment captured the differentiation trajectory spanning from na\u0026iuml;ve like cells to memory and effector cells, providing a framework for analyzing dynamic changes in cellular proportions. Na\u0026iuml;ve like CD4 cells underwent clonal expansion along trajectory 2 into T cells exhibiting increased activation, and cytotoxic effector function (\u003cstrong\u003eFig. 3c-d and Extended Data Fig. 5\u003c/strong\u003e).\u0026nbsp;When comparing the cell abundances of non-CR\u0026nbsp;with CR patients for CD4+ T cells along the trajectory 2, we observed a significantly (p \u0026lt;0.05) lower abundance of cytotoxic cells in non-responders (\u003cstrong\u003eFig. 3e\u003c/strong\u003e). Overall, non-CR showed an increased proportion of na\u0026iuml;ve like CD4 and CD8 cells along the trajectories, with the highest percentage of CAR+ cells in nodes with cytotoxic/effector functions (depicted as numbers in \u003cstrong\u003eFig 3e\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eAnalysis of the TCR repertoire showed that non-CR vs CR patients exhibited a higher proportion of CD4 clones consisting of only one cell (singletons) at all time points (\u003cstrong\u003eFig 3f\u003c/strong\u003e). Conversely, clonotype richness was higher in CR patients (\u003cstrong\u003eFig 3g\u003c/strong\u003e). Significant differences in clonality for CD8 subsets were only observed at leukapheresis. Composition of the T cell compartment, grouped by CAR- and CAR+ cells, showed that CD4+CAR+ singletons are predominantly of the cytotoxic phenotype expressing \u003cem\u003eEOMES\u003c/em\u003e (\u003cstrong\u003eFig 3h)\u003c/strong\u003e. This phenotype was also more pronounced in the expanded CAR+ clones compared to the CAR- clones.\u003c/p\u003e\n\u003cp\u003eThe comparison of gene expression of CAR+ with CAR- (61 DE genes) or pre-infusional (314 DE genes) T cells revealed that CAR+ cells expressed significantly higher levels of genes associated with cytolysis and exhaustion, where this effect was less pronounced when compared to CAR-. Na\u0026iuml;ve-like and memory gene sets, on the other hand, were downregulated in CAR+ cells (\u003cstrong\u003eFig 3i\u0026nbsp;\u003c/strong\u003eand\u003cstrong\u003e\u0026nbsp;Supplementary Table 6)\u003c/strong\u003e. To characterize potentially tumor-reactive T cells (pTRT), we used an \u003cem\u003ein vitro\u003c/em\u003e validated transcriptional T cell signature for anti-tumor reactivity in MM, as described recently\u003csup\u003e20\u003c/sup\u003e. The highest median enrichment score of pTRTs was observed for CD4 GNLY and CD8 subtypes, especially for EMRA identities (\u003cstrong\u003eFig 3j)\u003c/strong\u003e. Cox-regression analysis using the fractions of pTRTs for each subtype on a continuous scale showed a significant association with PFS for late time points (\u003cstrong\u003eFig 3k)\u003c/strong\u003e. Here, a higher proportion of pTRTs in EMRA subsets showed the strongest association with improved PFS.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle cell profiling of cytokine release syndrome during CAR therapy\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFollowing CAR-T infusion, 67% of the patients experienced grade I or II CRS, 37% of whom required tocilizumab application (no CRS: n\u0026thinsp;=\u0026thinsp;20 CRS/without tocilizumab: n\u0026thinsp;=\u0026thinsp;26, CRS/with tocilizumab: n\u0026thinsp;=\u0026thinsp;15). By product, 79% of patients treated with Ide-cel and 52% of patients treated with Cilta-cel developed any grade CRS (p=0.03, Fisher\u0026rsquo;s exact test). Only 3 patients who received Cilta-cel (11%) and one patient (3%) treated with Ide-cel experienced grade I ICANS (p=0.4). As expected, patients with grade I and II CRS had significantly increased peak levels of serum C-reactive protein (CRP) (p\u0026thinsp;=\u0026thinsp;0.002) (\u003cstrong\u003eFig. 4a\u003c/strong\u003e). There were no significant differences in maximum CRP values between response groups (p = 0.28). In addition, we observed a significant Spearman correlation (p = 0.011) between the estimated proportions of CAR+ cells seven days after infusion and the maximum CRP values (\u003cstrong\u003eFig. 4b\u003c/strong\u003e), with proportions of CAR+ cells increasing significantly with CRS grade (p = 0.002) (\u003cstrong\u003eFig. 4c\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAs only 6 patients with available single-cell data developed grade II CRS, we divided the cohort into CRS (grade I and II) and non-CRS groups for subsequent analyses. We observed significant changes in cell type composition of patients who developed any grade CRS: B cell subsets, na\u0026iuml;ve-like CD4 and CD8 T cells were significantly more abundant at the time of leukapheresis, whereas regulatory T cells were enriched at the late time point (\u003cstrong\u003eFig. 4d\u003c/strong\u003e). Comparison of CD4 and CD8 T cell clones between CRS and non-CRS patients revealed a stronger increase in the cytotoxicity enrichment score of CD4 clones compared to CD8 clones at the late time point (\u003cstrong\u003eFig. 4e\u003c/strong\u003e). In previous studies, transcriptional changes involving genes encoding cytokines and the number of polyfunctional CAR+ cells correlated with CRS\u003csup\u003e15\u003c/sup\u003e. Here, we found a significant increase in the proportion of polyfunctional CD8 T cells at the late time point in patients with CRS (\u003cstrong\u003eFig. 4f\u003c/strong\u003e), but no significant differences in fractions of polyfunctional CD4 T cells at any time point.\u003c/p\u003e\n\u003cp\u003eDifferential gene expression analysis (DGEA)\u0026nbsp;comparing CRS with non-CRS patients showed upregulation of\u0026nbsp;genes involved in TNF-\u0026alpha; signaling via NF-\u0026kappa;B (e.g. \u003cem\u003eNFKBIA\u003c/em\u003e, \u003cem\u003eRELB\u003c/em\u003e, \u003cem\u003eJUN, PIM3\u003c/em\u003e) in most lymphoid and myeloid populations at the late time point, reflecting systemic inflammation\n \u003c!--[if supportFields]\u003e\u003cspan lang=EN-US style='mso-ansi-language:EN-US'\u003e\u003cspan style='mso-element:field-begin'\u003e\u003c/span\u003eADDIN CitaviPlaceholder{{"$id":"1","$type":"SwissAcademic.Citavi.Citations.WordPlaceholder, SwissAcademic.Citavi","Entries":[{"$id":"2","$type":"SwissAcademic.Citavi.Citations.WordPlaceholderEntry, SwissAcademic.Citavi","Id":"80fab19a-ebb7-4482-a86e-370473ce2769","RangeLength":2,"ReferenceId":"a2057701-522b-4bdf-9545-789c9c372b35","PageRange":{"$id":"3","$type":"SwissAcademic.PageRange, SwissAcademic","EndPage":{"$id":"4","$type":"SwissAcademic.PageNumber, SwissAcademic","IsFullyNumeric":false,"NumberingType":0,"NumeralSystem":0},"NumberingType":0,"NumeralSystem":0,"StartPage":{"$id":"5","$type":"SwissAcademic.PageNumber, SwissAcademic","IsFullyNumeric":false,"NumberingType":0,"NumeralSystem":0}},"Reference":{"$id":"6","$type":"SwissAcademic.Citavi.Reference, SwissAcademic.Citavi","AbstractComplexity":0,"AbstractSourceTextFormat":0,"Authors":[{"$id":"7","$type":"SwissAcademic.Citavi.Person, SwissAcademic.Citavi","FirstName":"Ting","LastName":"Liu","Protected":false,"Sex":0,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"0c527c05-0d9c-4e1f-ad08-29f85fc3bd6d","ModifiedOn":"2025-01-26T22:25:52","Project":{"$id":"8","$type":"SwissAcademic.Citavi.Project, SwissAcademic.Citavi"}},{"$id":"9","$type":"SwissAcademic.Citavi.Person, SwissAcademic.Citavi","FirstName":"Lingyun","LastName":"Zhang","Protected":false,"Sex":0,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"6f74842b-b038-42c5-85c7-970a6eade8ad","ModifiedOn":"2025-01-26T22:25:52","Project":{"$ref":"8"}},{"$id":"10","$type":"SwissAcademic.Citavi.Person, SwissAcademic.Citavi","FirstName":"Donghyun","LastName":"Joo","Protected":false,"Sex":0,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"24615fa6-e415-4a37-8555-2f2549a42879","ModifiedOn":"2025-01-26T22:25:52","Project":{"$ref":"8"}},{"$id":"11","$type":"SwissAcademic.Citavi.Person, SwissAcademic.Citavi","FirstName":"Shao-Cong","LastName":"Sun","Protected":false,"Sex":0,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"ca56fd92-2aeb-4c23-9c43-a7dfaaaaac55","ModifiedOn":"2025-01-26T22:25:52","Project":{"$ref":"8"}}],"CitationKeyUpdateType":0,"Collaborators":[],"Date2":"14.07.2017","Doi":"10.1038/sigtrans.2017.23","Editors":[],"EvaluationComplexity":0,"EvaluationSourceTextFormat":0,"Groups":[],"HasLabel1":false,"HasLabel2":false,"Keywords":[],"Language":"eng","LanguageCode":"en","Locations":[{"$id":"12","$type":"SwissAcademic.Citavi.Location, SwissAcademic.Citavi","Address":{"$id":"13","$type":"SwissAcademic.Citavi.LinkedResource, SwissAcademic.Citavi","LinkedResourceType":5,"OriginalString":"PMC5661633","UriString":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5661633","LinkedResourceStatus":8,"Properties":{"$id":"14","$type":"SwissAcademic.Citavi.LinkedResourceProperties, SwissAcademic.Citavi"},"SyncFolderType":0,"IsLocalCloudProjectFileLink":false,"IsCloudRestore":false,"IsCloudCopy":false,"AttachmentFolderWasInFallbackMode":false},"Annotations":[],"LocationType":0,"MirrorsReferencePropertyId":208,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"bf0dc913-de16-4f10-be8c-06403ba9a05b","ModifiedOn":"2025-01-26T22:25:52","Project":{"$ref":"8"}},{"$id":"15","$type":"SwissAcademic.Citavi.Location, SwissAcademic.Citavi","Address":{"$id":"16","$type":"SwissAcademic.Citavi.LinkedResource, SwissAcademic.Citavi","LinkedResourceType":5,"OriginalString":"10.1038/sigtrans.2017.23","UriString":"https://doi.org/10.1038/sigtrans.2017.23","LinkedResourceStatus":8,"Properties":{"$id":"17","$type":"SwissAcademic.Citavi.LinkedResourceProperties, SwissAcademic.Citavi"},"SyncFolderType":0,"IsLocalCloudProjectFileLink":false,"IsCloudRestore":false,"IsCloudCopy":false,"AttachmentFolderWasInFallbackMode":false},"Annotations":[],"LocationType":0,"MirrorsReferencePropertyId":128,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"eb890f24-6b76-421e-8c55-10f413fcef14","ModifiedOn":"2025-01-26T22:25:52","Project":{"$ref":"8"}},{"$id":"18","$type":"SwissAcademic.Citavi.Location, SwissAcademic.Citavi","Address":{"$id":"19","$type":"SwissAcademic.Citavi.LinkedResource, SwissAcademic.Citavi","LinkedResourceType":5,"OriginalString":"29158945","UriString":"http://www.ncbi.nlm.nih.gov/pubmed/29158945","LinkedResourceStatus":8,"Properties":{"$id":"20","$type":"SwissAcademic.Citavi.LinkedResourceProperties, SwissAcademic.Citavi"},"SyncFolderType":0,"IsLocalCloudProjectFileLink":false,"IsCloudRestore":false,"IsCloudCopy":false,"AttachmentFolderWasInFallbackMode":false},"Annotations":[],"LocationType":0,"MirrorsReferencePropertyId":164,"CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"75c43b9f-15a0-4faf-b8d6-226d24a7dc3a","ModifiedOn":"2025-01-26T22:25:52","Project":{"$ref":"8"}}],"Organizations":[],"OthersInvolved":[],"PageRange":"<sp>\r\n  <n>17023</n>\r\n  <ns>Omit</ns>\r\n  <os>17023-</os>\r\n  <ps>17023-</ps>\r\n</sp>\r\n<os>17023-</os>","Periodical":{"$id":"21","$type":"SwissAcademic.Citavi.Periodical, SwissAcademic.Citavi","Eissn":"2059-3635","Name":"Signal transduction and targeted therapy","Pagination":0,"Protected":false,"UserAbbreviation1":"Signal Transduct Target Ther","CreatedBy":"_Michael Rade","CreatedOn":"2025-01-15T22:00:15","ModifiedBy":"_Michael Rade","Id":"f30ef87b-d02e-47dd-9bfa-63a163580d1c","ModifiedOn":"2025-01-15T22:00:15","Project":{"$ref":"8"}},"PmcId":"PMC5661633","Publishers":[],"PubMedId":"29158945","Quotations":[],"Rating":0,"ReferenceType":"JournalArticle","ShortTitle":"Liu, Zhang et al. 2017 – NF-κB signaling in inflammation","ShortTitleUpdateType":0,"SourceOfBibliographicInformation":"PubMed","StaticIds":["3d612eab-59f6-45a9-8c44-30426e1d57de"],"TableOfContentsComplexity":0,"TableOfContentsSourceTextFormat":0,"Tasks":[],"Title":"NF-κB signaling in inflammation","Translators":[],"Volume":"2","Year":"2017","YearResolved":"2017","CreatedBy":"_Michael Rade","CreatedOn":"2025-01-26T22:25:52","ModifiedBy":"_Michael Rade","Id":"a2057701-522b-4bdf-9545-789c9c372b35","ModifiedOn":"2025-03-05T18:34:08","Project":{"$ref":"8"}},"UseNumberingTypeOfParentDocument":false}],"FormattedText":{"$id":"22","Count":1,"TextUnits":[{"$id":"23","FontStyle":{"$id":"24","Superscript":true},"ReadingOrder":1,"Text":"21"}]},"Tag":"CitaviPlaceholder#8285fb0c-58a8-41d0-8a25-0c634287b511","Text":"21","WAIVersion":"6.15.2.0"}}\u003c/span\u003e\u003cspan style='mso-element:field-separator'\u003e\u003c/span\u003e\u003c![endif]--\u003e\u003csup\u003e\u003cspan lang=\"EN-US\"\u003e21\u003c/span\u003e\u003c/sup\u003e\n \u003c!--[if supportFields]\u003e\u003cspan style='mso-element:field-end'\u003e\u003c/span\u003e\u003c![endif]--\u003e (\u003cstrong\u003eFig. 4g\u003c/strong\u003e, see \u003cstrong\u003eExtended Data Fig. 6\u0026nbsp;\u003c/strong\u003eand\u003cstrong\u003e\u0026nbsp;Supplementary Table 6\u0026nbsp;\u003c/strong\u003efor DE genes of all cell types and time points). In addition, we observed a\u0026nbsp;higher expression of the CAR construct in\u0026nbsp;CD4 EOMES, along with a higher expression of cytotoxic-related genes \u003cem\u003eGNLY\u003c/em\u003e and \u003cem\u003eGZMB\u003c/em\u003e in CD4 CTL subsets. Furthermore, the costimulatory receptor \u003cem\u003eCD27,\u0026nbsp;\u003c/em\u003ewhich governs an essential signaling cascade stimulating differentiation and long-term survival in activated T cells\u003csup\u003e22,23\u003c/sup\u003e, was upregulated in various CD4 and CD8 populations. In line with these findings, increased TCR signaling was confirmed in both CD8 and CD4 effector subsets at late time points (\u003cstrong\u003eFig. 4h\u003c/strong\u003e). Gene set enrichment analysis further revealed upregulation of the hypoxia signature in post-infusional CD4 and CD8 effector T cells, cDCs and CD14\u003csup\u003e+\u003c/sup\u003e monocytes (\u003cstrong\u003eExtended Data, Fig. 7\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAt the time of LP, naive B cells showed increased MHC-I presentation to CD8 effector subsets in patients with CRS (\u003cstrong\u003eFig. 4i)\u003c/strong\u003e. Potential receptor-ligand interactions in post-infusional T cells were mostly incoming signals to the CD69 receptor, which was upregulated on CD8 effector and \u0026gamma;\u0026delta; T populations and downregulated on CD8 naive cells. The T cell activation marker \u003cem\u003eCD69\u003c/em\u003e has been identified to play a key role in tissue retention and might therefore play a crucial role for CRS pathophysiology\u003csup\u003e24\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eTo detect gene programs within our single-cell dataset independent of cell clustering and annotation, we performed supervised pathway deconvolution analysis using the Spectra method\u003csup\u003e25\u003c/sup\u003e. Spectra decomposes the scRNA-seq into a set of interpretable factors or gene programs (sets of co-expressed genes) using a curated list of global and T cell specific gene sets as prior information for modeling. \u003cstrong\u003eFig. 4j\u0026nbsp;\u003c/strong\u003edepicts the differences in enrichment of estimated gene programs between patients with CRS and without CRS. Gene programs associated with tumor reactivity, exhaustion, and TNF were significantly (adjusted p-value\u0026nbsp;\u0026lt; 0.05) upregulated in patients with CRS while TNK cytotoxicity was upregulated CD4 CTL GNLY and CD8 EMRA subtypes. Genes associated with the programs and the cell factor loadings are shown in F\u003cstrong\u003eig. 4k\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eComparison between patients treated with Cilta-cel and Ide-cel\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe observed distinct expansion profiles between the two CAR-T products: Patients who received Cilta-cel showed increased c\u003csub\u003emax\u003c/sub\u003e and slower expansion kinetics with c\u003csub\u003emax\u003c/sub\u003e reached on day 14 vs day 7. The CD4+CAR+/CD8+CAR+ ratio was significantly elevated with Cilta-cel compared to Ide-cel on days 7, 14 and 30 (\u003cstrong\u003eFig. 5a\u003c/strong\u003e). Single-cell analysis did not show differentially abundant cell populations at LP, but significant changes in post-infusional cell type composition between patients treated with Cilta-cel and Ide-cel. CD4 CTL and memory/effector/na\u0026iuml;ve-like CD8 cells were more abundant in Cilta-cel-treated patients (\u003cstrong\u003eFig. 5b\u003c/strong\u003e). In addition, we performed paired analyses of cell type fractions between the time points (\u003cstrong\u003eFig. 5c)\u003c/strong\u003e. In Cilta-cel-treated patients, CD4 EOMES and CD8 EM cells increased significantly (p \u0026lt;0.05) comparing the late time point to LP and decreased significantly comparing the very late to the late time point (see \u003cstrong\u003eExtended Data Fig. 8\u0026nbsp;\u003c/strong\u003efor all cell types).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eNext, we performed DGEA for T cell subtypes from Cilta-cel and Ide-cel patients comparing Late with LP time points. In Cilta-cel patients, we observed significant upregulation of \u003cem\u003eCD27\u003c/em\u003e and downregulation of \u003cem\u003eCXCR4\u003c/em\u003e, \u003cem\u003eGZMB\u003c/em\u003e and \u003cem\u003eGZMK\u003c/em\u003e in CD4 CTL subsets. Both products showed similar changes in CD8 populations, with upregulation of T-cell activation and exhaustion genes and downregulation of naive and stem-like genes (\u003cstrong\u003eFig. 5d\u003c/strong\u003e). \u003cstrong\u003eFig. 5e\u003c/strong\u003e depicts DE genes (adjusted p-value\u0026nbsp;\u0026lt;0.05) with opposite direction of log2 fold change between Cilta-cel and Ide-cel. Unexpectedly, cytotoxic markers \u003cem\u003eGNLY\u003c/em\u003e and \u003cem\u003eGZMB\u003c/em\u003e were upregulated in T cells of Ide-cel patients and downregulated in Cilta-cel patients.\u003c/p\u003e\n\u003cp\u003eAnalysis of longitudinal dynamics for the most abundant clones with CAR+ cells revealed a distinct CD4 EOMES phenotype in patients treated with Cilta-cel (\u003cstrong\u003eFig. 5f\u003c/strong\u003e). DGEA for the late time point, comparing CD8+CAR+ Cilta-cel with CD8+CAR+ Ide-cel cells showed upregulation of memory-associated (\u003cem\u003eSELL\u003c/em\u003e, \u003cem\u003eIL7R\u003c/em\u003e)\u003csup\u003e26\u003c/sup\u003e and downregulation of effector-associated genes (\u003cem\u003eKLRD1\u003c/em\u003e, \u003cem\u003eGZMB\u003c/em\u003e, \u003cem\u003eGNLY\u003c/em\u003e). Accordingly, enrichment analysis\u0026nbsp;revealed a higher cytotoxicity and effector function in Ide-cel CARs\u0026nbsp;(\u003cstrong\u003eFig. 5g\u0026nbsp;\u003c/strong\u003eand\u003cstrong\u003e\u0026nbsp;h\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003esBCMA dynamics and immune reconstitution\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe examined changes in B-cell lineage subtypes and the B-cell receptor (BCR) landscape (\u003cstrong\u003eFig. 6a\u0026nbsp;\u003c/strong\u003eand \u003cstrong\u003eExtended Data Fig. 9a)\u003c/strong\u003e. This confirmed monoclonal cPCs and suppressed polyclonal B-cells at each analyzed timepoint \u003cstrong\u003e(Fig. 6b).\u0026nbsp;\u003c/strong\u003eFive patients had relevant proportions of cPCs (\u003cstrong\u003eFig. 6c\u003c/strong\u003e and \u003cstrong\u003eExtended Data Fig. 9b\u003c/strong\u003e), which had retained expression of the immuno-therapy targets \u003cem\u003eTNFRSF17\u003c/em\u003e and \u003cem\u003eCD38\u003c/em\u003e and to a lesser extent of putative novel targets \u003cem\u003eFCRL5\u003c/em\u003e and \u003cem\u003eLAMP5\u003c/em\u003e (\u003cstrong\u003eFig. 6d\u003c/strong\u003e). Interestingly, \u003cem\u003eTNFRSF17\u003c/em\u003e expression remained high even at the very late timepoint, suggesting there was no BCMA loss in refractory malignant PCs of these patients (\u003cstrong\u003eExtended Data Fig. 9c\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eWe observed depletion of B-cells, pDCs and cPCs at the late time point and their recovery at the very late time point in scRNA-seq data. As expected, there were significantly lower proportions of all BCMA-expressing cell types in CR compared to non-CR patients at the time points\u0026nbsp;following\u0026nbsp;CAR-T cell infusion (\u003cstrong\u003eFig 6e-g\u003c/strong\u003e).\u0026nbsp;B-cells showed a shift towards early stages of B-cell differentiation for the late and recovery of memory B-cells at the very late time point (\u003cstrong\u003eExtended Data\u0026nbsp;Fig. 9d\u003c/strong\u003e).\u0026nbsp;There was no significant association between the incidence of infections and B-cell proportions at any of the analyzed time points (\u003cstrong\u003eExtended Data Fig. 9e\u003c/strong\u003e).\u0026nbsp;Next, we analyzed longitudinal B cell frequencies from routine clinical blood counts (n=55), revealing that all patients except for one (98%) had B cell aplasia following CAR-T cell infusion. \u0026nbsp;B-cell recovery defined as any detectable B-cells occurred significantly earlier in non-CR vs CR patients (\u003cstrong\u003eFig. 6h)\u003c/strong\u003e, which was driven by the Ide-cel subgroup (\u003cstrong\u003eExtended Data Fig. 9f\u003c/strong\u003e). However, most patients remained below the lower normal range even 200 days after infusion. Direct comparison of B-cell recovery dynamics between patients treated with Cilta-cel and Ide-cel using the inverse Kaplan-Meier method showed significantly slower B-cell recovery after Cilta-cel infusion (\u003cstrong\u003eFig. 6i\u003c/strong\u003e). This was confirmed using a Cox proportional hazard model adjusted for clinical response (p=0.04).\u003c/p\u003e\n\u003cp\u003eLongitudinal analysis of sBCMA showed significantly lower levels in CR vs non-CR patients on days 0, 30 and 100, consistent with recent data emphasizing its role as a putative surrogate marker for MM tumor burden\u003csup\u003e27,28\u003c/sup\u003e (\u003cstrong\u003eFig. 6j\u003c/strong\u003e). To determine the likely source of sBCMA, we conducted correlation analyses with BCMA-expressing cell types at LP, on day 30 and 100 after infusion (\u003cstrong\u003eFig. 6k\u003c/strong\u003e). No correlation was observed at LP, whereas there was a significant correlation between sBCMA, pDC and cPC fractions at the late time point, confirming similar depletion of all BCMA-expressing cell types. At the very late time point, sBCMA levels and the cPC fraction showed the strongest correlation, suggesting that sBCMA after CAR-T cell infusion is mostly derived from cPCs. Finally, we investigated the relationship between sBCMA reduction between days 0 and 30 and CAR-T cell expansion (\u003cstrong\u003eFig. 6l\u003c/strong\u003e). We observed a significant correlation between c\u003csub\u003emax\u003c/sub\u003e and sBCMA reduction (p=0.00013). Furthermore, maximal CRP levels significantly correlated with sBCMA reduction (p=0.0064; \u003cstrong\u003eFig. 6m\u003c/strong\u003e and \u003cstrong\u003eExtended Data Fig. 9g\u003c/strong\u003e) providing a connection between tumor debulking and the severity of system inflammation following CAR-T cell infusion. Taken together, these findings validate the role of sBCMA as a prognostic biomarker for BCMA-directed CAR-T cells in RRMM and link high initial tumor burden and effective CAR-T cell expansion with CAR-T related toxicities.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eOur study provides a comprehensive clinical and molecular characterization of patients treated with BCMA-directed CAR T-cell therapies. Consistent with previous real-world studies\u003csup\u003e5,29\u0026ndash;31\u003c/sup\u003e, Cilta-cel exhibited superior response rates, with a significantly higher CR rate (78% vs. 38%) and PFS compared to Ide-cel. We observed distinct cellular kinetics between the two products. Cilta-cel exhibited higher and delayed peak expansion aligning with the later onset of CRS reported in recent real-world studies\u003csup\u003e5\u003c/sup\u003e. This delayed expansion has important implications for the ambulatory application of Cilta-cel and individualized patient monitoring\u003csup\u003e32\u003c/sup\u003e. Given that peak CAR-T expansion was associated with higher-grade CRS, preemptive strategies such as dexamethasone administration in patients with high CAR-T cell counts are currently being explored in clinical trials. Furthermore, Cilta-cel-treated patients had a significantly higher CD4+CAR+/CD8+CAR+ ratio on days 7, 14, and 30 compared to Ide-cel patients, suggesting an important role for CD4+ CAR-T cells in mediating durable anti-tumor responses. Recent single cell transcriptomic studies of patients who received anti-CD19 CAR-T cell therapy for B cell neoplasia emphasized the role of CD4+CAR+ T cells for tumor-clearance and long-term remission\u003csup\u003e33,34\u003c/sup\u003e. In line with the expansion data, single cell analyses revealed that CD4 CTL subsets were more abundant in Cilta-cel vs Ide-cel and CR vs non-CR patients. Additionally, the CD4 CTL population of patients treated with Cilta-cel showed a distinct transcriptional profile with upregulation of \u003cem\u003eCD27\u003c/em\u003e and downregulation of \u003cem\u003eCXCR4\u003c/em\u003e, \u003cem\u003eGZMB\u003c/em\u003e and \u003cem\u003eGNLY\u003c/em\u003e. Recent analyses demonstrated superior antitumor efficacy in CD27-costimulated CAR-T cells\u003csup\u003e23\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eDGEA of non-CR compared to CR patients revealed a transcriptional signature of impaired effector function, mitochondrial dysfunction and upregulation of exhaustion-related genes such as \u003cem\u003eTIGIT\u003c/em\u003e and \u003cem\u003eCD38\u0026nbsp;\u003c/em\u003ein CD8 T effector subsets at the late time point. Therefore, impaired T cell effector\u0026nbsp;function is a main driver of suboptimal clinical responses to anti-BCMA CAR-T cell therapy in RRMM. However, recent studies have highlighted that predicting response and toxicity in CAR T-cell therapy is not solely dependent on CAR T-cell properties and host immune fitness, but also on tumor burden, with sBCMA emerging as a key predictive factor influencing both efficacy and side effects\u003csup\u003e27,28,35\u003c/sup\u003e. sBCMA levels correlated with initial tumor burden, with greater reductions linked to stronger CAR-T expansion. Interestingly, patients with greater sBCMA reduction also exhibited higher peak CRP levels, suggesting a connection between tumor debulking and systemic inflammation. This finding aligns with prior observations that high antigen density can drive more robust CAR-T expansion\u003csup\u003e36\u003c/sup\u003e and inflammatory responses, potentially contributing to CRS pathophysiology.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe identified pDCs to potentially contribute to ongoing BCMA ligand availability, which might impair CAR T-cell efficacy by acting as antigen decoys.\u0026nbsp;Given their high BCMA\u003cem\u003e\u0026nbsp;\u003c/em\u003eexpression, pDCs might interfere with CAR-T functionality by occupying BCMA-targeted CAR constructs, reducing effective antigen engagement on malignant plasma cells. The persistence of pDCs in circulation post-infusion and their association with lower response rates suggest that they may serve as a surrogate marker for therapeutic efficacy. Our findings may also have broader implications for dendritic cell-associated diseases, including autoimmune disorders\u003csup\u003e37\u003c/sup\u003e and blastic plasmacytoid dendritic cell neoplasm (BPDCN), a clinically aggressive neoplasm derived from non-activated precursors of pDCs\u003csup\u003e38\u003c/sup\u003e. Taken together, our findings highlight the superior remission conversion potential of Cilta-cel, its distinct CAR-T expansion and transcriptional profile, and the interplay between tumor burden, T cell fitness, and systemic inflammation.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eStudy design\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData and sample collection were approved by the local ethics committee (Leipzig Medical Biobank/LMB No. 311-15-24082015 and LMB-UCCL-2022_06 // 361/22-ek) and performed in accordance with the Declaration of Helsinki after obtaining written informed consent. All 61 RMM patients who received Ide-cel or Cilta-cel outside of clinical trials at University Hospital Leipzig (Leipzig, Germany) between 01/06/2022 and 01/04/2024 were included in this study. The median age at CAR-T infusion was 64 (range 31-75) years and 44% of patients were female. All patients were of Non-Hispanic European origin. The median number of prior lines of therapy before CAR-T was 7 (1-15) across all patients. Manufacturing failure occurred in 10 patients, and 4 patients received products that were out-of-specification of the manufacturer\u0026apos;s quality criteria (\u003cstrong\u003eExtended Data Fig. 1a\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBCMA-directed therapies in earlier treatment lines (excluding bridging) were administered to 8 (13%) patients. All patients received bridging therapy with either chemotherapy (31%), CD38 antibody-based, (31%) SLAMF7 antibody-based (13%) regimens or bispecific T cell engaging antibodies (25%). Lymphodepletion was performed using fludarabine and cyclophosphamide (n=56, 92%) or bendamustine (n=5, 8%) in patients with severe renal insufficiency. Patients with mild/moderate renal insufficiency received a Flu/Cy dose adjusted for creatinine clearance determined by CKD-EPI (n=5, 8%). The patients were followed up in our outpatient clinic and response was assessed according to International Myeloma Working Group (IMWG) criteria\u003csup\u003e39\u003c/sup\u003e at the time point of LP, on the day of CAR-T infusion, and at months 1, 3, 6, and 12 thereafter. Overall response rate was defined as the proportion of patients who were in partial response or better and participants were categorized into CR and non-CR based on the best response within 6 months post CAR-T infusion. CRS and ICANS were assessed according to the American Society for Transplantation and Cellular Therapy (ASTCT) recommendations\u003csup\u003e40\u003c/sup\u003e. Cytogenetic analyses were performed using fluorescence in-situ hybridization. The presence of gain(1q) (3 copies) or amp(1q) (4 or more copies), t(4;14), t(14;16) or del(17p) were considered as high-risk aberrations. EMD was defined as any soft-tissue, organ or paramedullary plasma cell infiltration, determined non-contiguous to bone tissue by PET-CT before CAR-T cell infusion.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSample collection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMononuclear cells from peripheral blood (PBMCs) were obtained from LP, on days 21\u0026ndash;60 (average 31) and on days 79\u0026ndash;169 (average 101) after CAR-T infusion for single cell multi-omics sequencing. PBMC for flow cytometry and blood sera were obtained on the day of LP, and on days 0, 7, 14, 30 and 100. Samples werecryopreserved and stored by trained staff of the Leipzig Medical Biobank according to standard operating procedures. Clinical quantification of CAR-T cell expansion by flow cytometry was performed using fresh whole blood samples as described below. Sample collection and preparation were blinded to the conditions of the patients.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImmunophenotyping by flow cytometry\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePB samples were collected at LP, on the day of CAR-T cell infusion (day 0), and on days 7, 14, 30 and 100 thereafter. Flow cytometry analysis was performed immediately on fresh samples using two separate antibody panels to quantify CAR+ T cells and T cell subsets, as described previously\u003csup\u003e41\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn brief, the samples were stained with fluorochrome-conjugated antibodies for 15 min at room temperature. Then, red blood cell lysis was performed by adding 2 ml of lyse solution (Becton Dickinson Biosciences, BD) for 10 min. The cells were centrifuged at 500g for 5 min at room temperature. In the following, the cell pellets were resuspended and washed in 2ml PBS for 5 min at room temperature. At last, the cells were subjected to flow cytometry using a BD FACSLyric\u003csup\u003eTM\u003c/sup\u003e and data analysis was performed using the BD FACSuite\u003csup\u003eTM\u0026nbsp;\u003c/sup\u003esoftware (see \u003cstrong\u003eSupplementary Fig. 2-4\u003c/strong\u003e for the gating strategy).\u003c/p\u003e\n\u003cp\u003eFor \u003cstrong\u003eFig. 2i\u003c/strong\u003e, the proportions of CD4 na\u0026iuml;ve and central memory T cells estimated by flow cytometry were added together. The rationale for this is that the na\u0026iuml;ve T cells from the CD4 reference atlas were determined on the basis of scRNA-Seq\u003csup\u003e42\u003c/sup\u003e. Since quantification in single-cell RNA analyses is performed at the gene level, it is not feasible to differentiate between the CD45RA and CD45RO isoforms and thus between na\u0026iuml;ve and central memory CD4 cells. The CD8 effector cells estimated by flow cytometry were re-annotated as EMRA cells for correlation analysis with scRNA-Seq.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImmunoassay for the quantification of serum BCMA concentration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSerum samples were used to determine soluble B-cell maturation antigen (sBCMA) by ELISA (Bio-Techne, cat. DY193 and DY008B), and absorbance was recorded using the FLUOstar OMEGA.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSurvival analyses\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe performed survival analyses using the R survival package v3.5.7. Kaplan\u0026ndash;Meier plots for patients were generated using the ggsurvplot function from the survminer v0.4.9 R package. Primary endpoint was progression free survival (PFS), defined as time from CAR-T infusion to progression or death. Log-rank tests were performed to compare probabilities of PFS between groups using the survdiff function implemented in the survival package. Cox proportional hazards models were computed using the coxph function from the survival package. In \u003cstrong\u003eFig. 1h and Extended Data Fig. 3h\u003c/strong\u003e, we applied multivariable Cox-regression to assess the performance of the covariates with adjustment for the CAR product. For the dichotomized covariate sBCMA (\u003cstrong\u003eFig. 1h\u003c/strong\u003e), the cutoff was determined using the median of the cohort at the time of leukapheresis before the administration of bridging therapy to identify patients with high tumor burden at CAR-T infusion\u003csup\u003e28\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eFor the analysis of response conversion, the categories SD/PD, VGPR/PR and CR were encoded as numerical factors and the difference between best response within 6 months after CAR-T infusion and baseline response was calculated for each patient. Then, Wilcoxon rank sum test was used to test whether conversion rates were significantly different between patients who received Cilta-cel and Ide-Cel. B-cell recovery was defined as the first time point with any detectable B-cells from routine clinical blood counts. Patients were censored if they were in ongoing B-cell aplasia at last FU. Inverse Kaplan-Meier curves were computed to compare B-cell recovery in patients treated with Cilta-cel and Ide-cel.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle-cell RNA, TCR, BCR and ADT sequencing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFour sequencing runs were performed for the multi-omics analysis, with the first run carried out as described previously\u003csup\u003e7\u003c/sup\u003e. For the sequencing runs 2-4, PBMCs were isolated following standardized operating procedures using Leucosep tubes (Greiner Bio One, Frickenhausen, Germany) with a Ficoll gradient (Ficoll-Paque\u0026trade; PLUS, VWR Avantor, Spring House, Pennsylvania). Cells were frozen using a CoolCell LX container (Corning, USA) and stored in the vapour phase of liquid nitrogen in an Askion HS200S storage system (Askion, Gera, Germany).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFor single-cell experiments, samples were thawed and stained with acridine orange and propidium iodide (AO/PI) dye, and quality was assessed using LUNA-FX7\u003csup\u003eTM\u003c/sup\u003e automated cell counter (Logos Biosystems, United Sates). Samples with initial viability \u0026lt;80% were subjected to dead cell removal using Viahance\u003csup\u003eTM\u003c/sup\u003e dead-cell removal kit (BioPAL, Inc., United States) following the manufacturer\u0026rsquo;s instructions.\u003c/p\u003e\n\u003cp\u003eTo enable single-cell protein expression profiling, the cells were stained with a custom panel of TotalSeq\u003csup\u003eTM\u003c/sup\u003e-C antibodies (BioLegend, United States), optimized, and prepared by Singleron Biotechnologies, Singapore. These TotalSeq\u003csup\u003eTM\u003c/sup\u003e-C antibodies are oligonucleotide-coupled antibodies directed against selected cell surface proteins. The antibodies are conjugated via a linker to oligonucleotides, which consist of a polymerase chain reaction (PCR) handle, an antibody-specific barcode, and a capture sequence. These antibody-bound oligonucleotides can be captured during single-cell library preparation, allowing simultaneous surface protein and gene expression profiling. For sequencing runs 2-4, samples of up to two donors were pooled and processed together. Specific to the second run, multiplexed samples were stained with a unique TotalSeq\u003csup\u003eTM\u003c/sup\u003e-C anti-human hashtag antibodies (BioLegend, United States) prior to pooling. The custom panel comprised 62 antibodies for sequencing run 1, 63 antibodies for run 2 and 70 for run 3 and 4 (\u003cstrong\u003eSupplementary Table 7-9\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eThe single-cell gene expression, surface protein expression, B cell receptor (BCR) and T cell receptor (TCR) libraries were constructed using Chromium Next GEM Single Cell 5\u0026rsquo; reagent kits (v2 chemistry; 10x Genomics, Inc., United States), as per the manufacturer\u0026rsquo;s instructions. Libraries were sequenced on an Illumina NovaSeq 6000 using S4 flow cells (sequencing runs 1-3) or NovaSeq X plus platform using 10B or 25B flow cells (sequencing run 4) and PE150 chemistry for all sequencing runs.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePre-processing and d\u003c/strong\u003e\u003cstrong\u003eemultiplexing\u0026nbsp;of single-cell data\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe CAR sequences for Ide-cel\u003csup\u003e43\u003c/sup\u003e and Cilta-cel\u003csup\u003e44\u003c/sup\u003e were obtained from the respective patents. Two separate custom references for Cilta and Ide-cel have been created by adding the corresponding CAR sequences to the reference genome (GRCh38, Ensembl release 98) and GENCODE v32 annotation using the mkref function from Cellranger v7.1.0. The Cell Ranger multi pipeline was used to process the single-cell gene expression, ADT (antibody-derived tag) expression, and V(D)J libraries. Since some of the libraries in sequencing run 3 consist of Ide and Cilta patients, Cell Ranger was performed once with Ide and once with the custom Cilta reference. The feature reference CSV file, declaring antibody capture constructs and associated barcodes is provided in \u003cstrong\u003eSupplementary Table 7-9\u003c/strong\u003e. The reference dataset required for the V(D)J contigs is provided by 10X Genomics (GRCh38 Human V(D)J Reference - 7.1.0 (December 7, 2022)). \u003cstrong\u003eSupplementary Table 4\u0026nbsp;\u003c/strong\u003esummarizes the quality control data for each 10x library.\u003c/p\u003e\n\u003cp\u003eFor sequencing runs 2-4, we used the R package souporcell v.2.0\u003csup\u003e45\u003c/sup\u003e with parameter k = 2 to demultiplex the samples in each 10X scRNA-seq library. This resulted in two cell clusters based on common variants for each sample. For sequencing run 2, where anti-human hashtag antibodies were used prior to pooling, clusters were associated with patient IDs based on the mean hashtag expression values per cluster. Since one male and one female donor per multplexed 10X scRNA-seq library were used for runs 3 and 4, the mean expression of the sex-specific genes \u003cem\u003eXIST\u003c/em\u003e and \u003cem\u003eRPS4Y1\u003c/em\u003e was used to associate the clusters with the respective sample ID. \u003cstrong\u003eSupplementary Table 3\u003c/strong\u003e provides an overview of whether and which samples were multiplexed.\u003c/p\u003e\n\u003cp\u003eFor all subsequent analyses, R v4.3.2 was used. As suggested by Neavin et al.\u003csup\u003e46\u003c/sup\u003e, we used the R packages scDblFinder v1.16.0\u003csup\u003e47\u003c/sup\u003e and the cxds_bcds_hybrid models from the package scds v1.18.0\u003csup\u003e48\u003c/sup\u003e in combination to remove barcodes (cells) that contained potential doublets. In addition, cell barcodes that met any of the following criteria were excluded: (1) fewer than 250 genes; (2) more than 8,000 genes; (3) fewer than 1,000 UMIs; (4) more than 10,000 UMIs; or (5) mitochondrial transcripts fraction more than 15%; and (6) cell complexity less than 0.8. Complexity was defined as follows: log10(nbr. of present genes)/log10(nbr. of UMIs). The numbers of cells for each sample before and after the filtering steps are listed in \u003cstrong\u003eSupplementary Table 3\u003c/strong\u003e. Raw gene expression data was normalized using the LogNormalize method of the NormalizeData function in Seurat v5.1.0\u003csup\u003e49\u003c/sup\u003e. ADT data was normalized by the centered log-ratio transformation (CLR) with parameters normalization.method = \u0026quot;CLR\u0026quot; and margin = 2 using the NormalizeData function.\u003c/p\u003e\n\u003cp\u003eQuality control showed a correlation between the number of genes/UMIs detected and the sequencing runs (In \u003cstrong\u003eExtended Data Fig. 2b\u0026nbsp;\u003c/strong\u003eand\u003cstrong\u003e\u0026nbsp;c\u003c/strong\u003e). Further evaluation of the potential batch effect between sequencing runs at cell type level showed no significant correlation except for plasma and B cells (\u003cstrong\u003eExtended Data Fig. 2d-j\u003c/strong\u003e), caused by the fact that the plasma cells in a batch essentially originate from the same patient (\u003cstrong\u003eFig. 6c\u003c/strong\u003e and \u003cstrong\u003eExtended Data Fig. 9b\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAnnotation of cell identities\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTransfer learning was performed to annotate the PBMC cells. To this end, we used the PBMC reference dataset form Butler et al.\u003csup\u003e50\u003c/sup\u003e with the RunAzimuth function from the R package Azimuth v0.5.0\u003csup\u003e49\u003c/sup\u003e. Only cells with a confidence score greater than 0.5 were retained for downstream analysis. As a secondary cell type annotation strategy, we used the R package scGate v1.6.0\u003csup\u003e51\u003c/sup\u003e, which requires (i) a normalized gene expression matrix and (ii) a \u0026apos;gating model\u0026apos; (GM) consisting of gene sets representing the cell populations of interest. The GM for T cells, gamma delta T cells, B cells, plasma cells, NK cells and myeloid cells were obtained using the get_scGateDB function. Cells that did not have matching cell type annotations using both the Azimuth and scGate methods were removed. Both annotation methods were applied per sample. Cells were also excluded if a VDJ sequence was found by TCR or BCR sequencing but were not classified as T or B cells. In addition, cells annotated as erythrocytes, erythroid progenitor cells, and platelets were excluded from subsequent analysis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe PBMC reference dataset used includes a coarse annotation of T cell subtypes. To cover a broader range of the T cell compartment, we re-annotated the T cells using the R package ProjecTILs v3.3.1\u003csup\u003e13\u003c/sup\u003e. To classify CD4 and CD8 T cell subtypes, we projected the two T cell lineages separately onto two reference maps. Here, we used the single-cell atlases for CD4 and CD8 cells from the SPICA database (https://spica.unil.ch/), which are part of the ProjecTILs framework. Before applying transfer learning strategy by ProjecTILs, we first annotated CD4 and CD8 T cell lineages based on the gene expression data using the following strategy: Given the raw counts, one cell was considered as CD8 positive or negative if the count value of \u003cem\u003eCD8A\u003c/em\u003e or \u003cem\u003eCD8B\u003c/em\u003e was \u0026gt;0 or \u0026le;0, respectively. One cell was considered as CD4 positive or negative if the count value of \u003cem\u003eCD4\u003c/em\u003e was \u0026gt;0 or \u0026le;0, respectively. This resulted in a classification of CD8-CD4+, CD8+CD4-, CD8+CD4+, and CD8-CD4- T cells. Double positive cells (CD8+CD4+) were removed from downstream analysis. In addition, based on the TCR-Seq data, clones consisting of CD4 and CD8 cells were removed (453 cells in total). We also excluded 36 clones (188 cells in total) from downstream analysis that were not donor specific. About 23% of all CD8-CD4- cells (total = 35,784) could be assigned to either CD4 or CD8 clones. For all remaining CD4-CD8- T cells, the following strategy was applied: For all T cells, we computed a nearest-neighbor graph using k=10 and then calculated the smoothed expression of \u003cem\u003eCD4\u003c/em\u003e and \u003cem\u003eCD8B\u003c/em\u003e by averaging across the 10 nearest neighbors of each cell using SmoothKNN from the R package UCell v2.6.2\u003csup\u003e52\u003c/sup\u003e. The remaining CD4-CD8- T cells were then classified by gating on these expression values. The thresholds of the smoothed expression values used for classification were 0.2 for \u003cem\u003eCD4\u003c/em\u003e and 0.3 for \u003cem\u003eCD8A\u003c/em\u003e. CD-CD8- cells \u0026gt;0.2 \u0026amp; \u0026lt;0.3 or \u0026gt;0.3 \u0026amp; \u0026lt;0.2 were annotated as CD4 or CD8 cells, respectively. Using this strategy, 70% of all CD8-CD4- cells could be assigned to the CD4 or CD8 lineages. Lastly, we aimed to annotate the remaining 8350 CD4-CD8- cells using ADT-Seq. The threshold of the normalized ADT expression values used for classification was 1 for CD4 and CD8A. CD4-CD8- cells \u0026gt;1 \u0026amp; \u0026lt;1 or \u0026gt;1 \u0026amp; \u0026lt;1 were annotated as CD4 or CD8 cells. A total of 2320 of the 8350 CD4-CD8- cells could be assigned to the CD4 or CD8 lineages.\u003c/p\u003e\n\u003cp\u003eGamma Delta T cells were annotated using the scGate method with the gating model \u0026ldquo;gdT\u0026rdquo; from the function get_scGateDB. All T cells annotated as CD4 or CD8 were used as input for the ProjecTILs.classifier function with the corresponding CD4 and CD8 atlas as reference. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo identify proliferating cells, we applied the run_gsea function implemented in the clustifyr R package v1.14\u003csup\u003e53\u003c/sup\u003e using previously published G2/M and S-phase gene sets from the Seurat package as queries. As suggested by the authors, an over-clustering for each sample was performed. For this purpose, the cluster resolution was set to 1. For gene set enrichment 1,000 permutations were performed. One cell cluster was significantly (p-value \u0026lt;0.05) enriched with cell cycle genes from the G2/M and S phases and was annotated as cyclic accordingly. Marker genes for all estimated T-cell identities are shown in \u003cstrong\u003eExtended Data Fig. 3\u003c/strong\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIntegration and dimension reduction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIntegration and clustering analysis were performed using Seurat v5.1.0. For each sample, the 2,000 most variable genes were estimated using the \u0026quot;vst\u0026quot; method of the FindVariableFeatures function. Mitochondrial, Immunoglobulin variable chain, T cell receptor and sex-specific genes were not included in the estimation of variable genes. We then used the SelectIntegrationFeatures function to select 2,000 genes that were repeatedly shown to be highly variable in multiple samples. Thereafter, the data were standardized with ScaleData, and principal component analysis (PCA) was performed on the standardized expression data of the variable genes using the function RunPCA. We used 30 principal components, which explained approx. 90% of the variance, to integrate all samples using the Harmony\u003csup\u003e54\u003c/sup\u003e method (v1.2) with the Seurat wrapper function RunHarmony with additional parameters (dim.use = 1:30, group.by.vars = c(\u0026quot;PATIENT_ID\u0026quot;)). Using the Harmony-corrected cell embeddings, we performed clustering by computing a shared nearest neighbors (SNN) graph, as implemented in FindNeighbors(reduction = \u0026quot;harmony\u0026quot;, dims = 1:30). Cluster identification using the SNN graph was computed by the FindCluster function (Louvain algorithm). The same strategy was also used for integration of the B and T cells. Here, 1,000 variable genes and 15 principal components were used.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDifferential gene expression and enrichment analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDifferentially expressed genes between groups were identified by the FindMarkers function implemented in Seurat, using the MAST algorithm\u003csup\u003e55\u003c/sup\u003e. When fitting the hurdle model, in addition to the covariates of interest both sequencing run as well as the number of detected genes per cell were used to reduce false positives due to different sequencing time points or mRNA capture rates (see \u003cstrong\u003eExtended Data Fig. 2a-j)\u003c/strong\u003e. In \u003cstrong\u003eFig. 3i, 5d\u003c/strong\u003e and \u003cstrong\u003eg\u003c/strong\u003e, clinical outcome was also used as further covariate. To retain only genes that were robustly expressed, differential gene expression was calculated only for genes that were expressed in at least 20% of cells in one analyzed group (parameter: min.pct = 0.20). Only cell types with at least 200 cells in both groups were analyzed. To minimize sample bias due to high cell counts, the cells of each sample and each cell type were downsampled to 250 (100 cells for \u003cstrong\u003eFig. 3i, 5d\u003c/strong\u003e and \u003cstrong\u003eg\u003c/strong\u003e). Genes with an adjusted p-value \u0026lt;0.05 (Bonferroni correction) and an absolute fold change of 1.25 were considered as significantly differentially expressed (DE) genes. In \u003cstrong\u003eFig. 3i\u003c/strong\u003e and \u003cstrong\u003e5e\u003c/strong\u003e, we have used custom gene sets that functionally characterize CAR-T cells as previously defined by Anderson and colleagues\u003csup\u003e56\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eDE genes were further used for overrepresentation analysis based on hypergeometric test analysis with the fgsea package v1.28.0\u003csup\u003e57\u003c/sup\u003e. GO terms for biological processes from the MSigDB\u003csup\u003e58\u003c/sup\u003e and curated gene signatures for T cells from Chu et al.\u003csup\u003e59\u003c/sup\u003e were used as references. For GO terms, the \u003cem\u003emgoSim\u003c/em\u003e function of the R package GOSemSim v2.28.0\u003csup\u003e60\u003c/sup\u003e was used to remove redundant terms. Terms and T cell signatures with adjusted p-value \u0026lt;0.05 (Benjamini Hochberg correction) were considered significant. To depict T cell-relevant GO terms in \u003cstrong\u003eFig. 3b\u003c/strong\u003e and \u003cstrong\u003e5h\u003c/strong\u003e, we filtered for metabolism-, T cell- and immune-relevant terms. Significant terms were ranked by rich factor, which is the number of DE genes in the term divided by the number of background genes in that term. In addition, a pathway direction score was calculated as the number of DE genes with log2FC \u0026gt;0 minus the number of DE genes with log2FC \u0026lt;0 divided by the square root of the number of genes associated with the pathway.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDifferences in cell type proportions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe used the propeller method\u003csup\u003e61\u003c/sup\u003e, as implemented in the speckle R package v1.20, to test for differences in cell type proportions between groups. To filter out samples with low cell type complexity, samples consisting of less than 10 cell types were removed from this analysis. In addition, only cell types with \u0026gt;100 cells in one group were used for downstream analysis. Cell type proportions were transformed with arcsin square root transformation function implemented in speckle. A linear model was fitted for each cell type with group, encoding CRS (\u003cstrong\u003eFig 4d\u003c/strong\u003e) or product (\u003cstrong\u003eFig 5b\u003c/strong\u003e) as the explanatory variable. For \u003cstrong\u003eFig 5c\u003c/strong\u003e, the linear model was fitted using the formula: ~block + contrast, where block encoded the patient (to account for patient‐specific differences in cell type proportion) and contrast containing the group information. Statistical significance was estimated using an empirical Bayes moderated t-test. In case a group for a cell type contained 0 cells and therefore the log2 fold change could not be calculated, the minimum or maximum log2 fold change of all other comparisons was used as a replacement.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDetection of polyfunctional cells\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe RNA expression data was used to determine whether a cell could be associated with multiple functional classifications of secreted cytokines (regulatory, effector, stimulatory, chemoattractive and inflammatory). Each functional group is determined by a list of genes associated as outlined in the study by Rossi et al.\u003csup\u003e62\u003c/sup\u003e. A cell was linked to a functional group whenever the expression of at least one gene in the corresponding list was higher than the median of non-zero counts for that given gene across the dataset. Whenever a cell was assigned to more than one functional group, the cell was classified as polyfunctional, and the fraction of polyfunctional cells was calculated for each patient and time point.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAssessing the TCR/BCR repertoire using V(D)J analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eV(D)J data were analyzed using scRepertoire v.2.0.0\u003csup\u003e63\u003c/sup\u003e. Clonotypes were identified using the combineTCR and combineBCR functions. A TCR clonotype is defined as the combination of genes comprising the TCR and the nucleotide sequence of the CDR3 region (gene\u0026thinsp;+\u0026thinsp;nucleotide) for paired TCR alpha and beta chains. BCR clonotypes based on the V gene and \u0026gt;85% normalized Levenshtein distance of the nucleotide sequence.\u003c/p\u003e\n\u003cp\u003eClonality in \u003cstrong\u003eFig. 3g\u003c/strong\u003e was defined as the complement of evenness (normalized Shannon entropy) as previously described by Zhang and colleagues\u003csup\u003e64\u003c/sup\u003e. The evenness value lies between 0 and 1, with a high value indicating a more equal distribution of TCRs and a low value indicating TCR skewing due to clonal expansion. Clonality, which reflects the dominance of particular clones across the TCR repertoire, was calculated per sample and time point. For CD8 and CD4 cells, the R package STARTRAC v0.10.0\u003csup\u003e64\u003c/sup\u003e was applied to calculate clonality (referred to by STARTRAC as expansion index).\u003c/p\u003e\n\u003cp\u003eIn \u003cstrong\u003eFig. 4e\u003c/strong\u003e, the average enrichment scores for cytotoxicity-associated genes (\u003cem\u003ePRF1\u003c/em\u003e, \u003cem\u003eGNLY\u003c/em\u003e, \u003cem\u003eGZMB\u003c/em\u003e, \u003cem\u003eGZMH\u003c/em\u003e, \u003cem\u003eGZMK\u003c/em\u003e and \u003cem\u003eNKG7)\u003c/em\u003e were calculated for each clonotype using the R package TCRanker (https://github.com/carmonalab/TCRanker).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTrajectory inference analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAnalysis was performed separately for CD8+ and CD4+ cells, excluding gamma delta and proliferating T cells. PCA-based dimensionality reduction was performed with variable genes, followed by integration (see section \u003cstrong\u003eIntegration and dimension reduction\u003c/strong\u003e). To model the state transition between cells, the DiffusionMap function from the destiny R-package v3.16.0\u003csup\u003e65\u003c/sup\u003e were used with the harmony-corrected cell embeddings as input. The function slingshot from the R package slingshot v.2.10.0\u003csup\u003e66\u003c/sup\u003e was applied with the first two diffusion components to calculate the trajectories (referred to by slingshot as lineages). Naive T cells cluster are considered as a root for estimation of the trajectories. Cells were grouped into cell clusters according to the nearest node along the trajectories.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCell-Cell communication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor RNA expression data, ligand-receptor signaling was inferred with iTalk\u003csup\u003e51\u003c/sup\u003e using default settings. LIANA v0.1.12\u003csup\u003e52\u003c/sup\u003e was used as a reference database consisting of manually curated resources in the context of cell-cell communication. All DE genes (non-CR vs CR) in pre-infusional PBMC were used as input.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSupervised pathway deconvolution\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSupervised non-matrix factorization was performed for all T cells using the Spectra v0.1.0 python package\u003csup\u003e25\u003c/sup\u003e.\u0026nbsp;For each sample, the 5,000 most variable genes were estimated using the \u0026quot;vst\u0026quot; method of the FindVariableFeatures function implement in Seurat. Mitochondrial, Immunoglobulin variable chain and T cell receptor genes were not included in the estimation of variable genes. We then used the SelectIntegrationFeatures function to select 3,000 genes that were repeatedly shown to be highly variable in multiple samples. As suggested by the authors from Spectra, the RNA expression values were normalized using the logNormCounts function from the Scran v1.30 package\u003csup\u003e67\u003c/sup\u003e, which uses the outputs of the quickCluster function followed by computeSumFactors.\u003c/p\u003e\n\u003cp\u003eGlobal and T cell type specific gene programs were obtained from Spectra using the function default_gene_sets.load. Factor (gene program) modeling was performed using the est_spectra function with hyperparameter lam\u0026thinsp;=\u0026thinsp;0.01 (other parameters are default settings). Single-cell factor scores were then extracted from the model results. Differentially expressed factors between two groups were estimated using the FindMarkers function implemented in Seurat (two-sided Wilcoxon rank-sum test). To minimize sample bias due to high cell counts, the cells of each sample and each cell type were downsampled to 250. To analyze only factors that were robustly enriched, factors that had a loading value of \u0026gt;0.001 in at least 25% of the cells in one analyzed group were retained. Factors with an adjusted p-value \u0026lt;0.05 (Bonferroni correction) and an absolute fold change of 1.25 were considered as significantly differentially enriched factors.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistics and reproducibility\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStatistical analysis was performed in R v.4.3.2. Since the scRNA-seq data conforms to a negative binomial distribution, comparisons between groups on gene expression level were carried out with the MAST method which is based on a zero inflated negative binomial model. For testing differences in the cell type proportions, a normal distribution of the data was not assumed. Using the speckle R package, an arcsin square root transformation was performed. Significant differences were estimated using the empirical Bayes moderated t-statistics (two-sided) implemented in the speckle package. Gene ontology term enrichment was calculated using a hypergeometric test. Estimated probabilities of PFS were calculated using the Kaplan\u0026ndash;Meier method and the log-rank test to evaluate differences between survival distributions. Each sample was considered a biological replicate, and no technical replicates were performed in this study. All statistical tests are listed in the corresponding method sections and figure legends for clarity.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAcknowledgement\u003c/h2\u003e\n\u003cp\u003eThis work was supported by the CERTAINTY project funded by the European Union (Grant Agreement 101136379). Views and opinions expressed are those of the author(s) only and do not necessarily reflect those of the European Union or the Health and Digital Executive Agency. Neither the European Union nor the granting authority can be held responsible for them.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eM. Merz is further supported by a Translational Research Award from the International Myeloma Society and the Deutsche Jose Carreras Leuk\u0026auml;mie Stiftung. Additionally, M. Merz has received research funding from Janssen, SpringWorks, and Roche/Genentech. M. Merz, H. Weidner and L. C. Hofbauer were supported by the German Research Foundation (\u0026micro;bone, SPP2084). K. Reiche and U. K\u0026ouml;hl are further supported by imSAVAR (Innovative Medicine Initiative 2 Joint Undertaking No 853988), T2Evolve (Innovative Medicine Initiative 2 Joint Undertaking No 945393), SaxoCell (BMBF Clusters4Future), DAAD project 57616814 (SECAI, School of Embedded Composite AI) and the German Jos\u0026eacute;-Carreras Leukemia Foundation (DJCLS 08 R/2023). V. Vucinic is supported by SaxoCell (ECP-CAR, BMBF Clusters4Future).\u003c/p\u003e\n\u003cp\u003eWe would like to extend our gratitude to S. Scharf from the Leipzig Medical Biobank for handling cryopreserved samples and D. Bretschneider, C. Mueller, K. Wildenberger and C. Wilhelm for performing fluorescence in situ hybridization analyses on clinical samples as well as C. Hertel for handling sample shipment. In addition, we thank Andreas Schmidt and Regina Ohmer from Singleron Biotechnologies for data management as well as D. Dudziak for her input regarding pDCs. Most importantly, we express our deepest thanks to the patients for their invaluable contribution.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eAuthors\u0026apos; Contributions\u003c/h2\u003e\n\u003cp\u003eM.M., M.R, D.F., M.K. K.R. designed the experiments, supervised the work, contributed equally to writing, and reviewed and edited the final paper. L.F., S.S., T.W., P.B., H.W., L.C.H., R.B., S.Y.W, E.B., S.H, K.M., J.S., M.H., C.H., M.J., G-N.F, A.B., M.F., U.K., U.P., V.V. and M.M. were responsible for acquisition of data (acquired and managed patients, provided facilities, FISH, Flow cytometry, biobanking, in vitro studies etc.) M.R., D.F., M.K., F.K., P.B., Z.S., J.K. and L.F performed analyses and interpretation of data (e.g., statistical analysis, computational analysis). M.R., D.F. and M.K. wrote the original draft which was reviewed and revised by K.R. and M.M. All authors have substantively revised the paper and approved the submitted version.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eCompeting interests\u003c/h2\u003e\n\u003cp\u003eMM gave advisory boards and received honoraria and research support from Amgen, BMS, Celgene, Gilead, Janssen, Stemline, Springworks, Sanofi, and Takeda MH gave advisory boards and received honoraria from Abbvie, Beigene, Jazz, Janssen, Stemline Menarini and Takeda, and received research support from EDO-Mundipharma, Janpix, Novartis, and Roche. S.F.: consultant and/or speaker fees from Novartis Pharma, Janssen-Cilag, Vertex Pharmaceuticals (Germany), Kite/Gilead Sciences, MSGO and Bristol-Myers Squibb. U.K.: consultant and/or speaker fees from AstraZeneca, Affimed, Glycostem, GammaDelta, Zelluna, Miltenyi Biotec and Novartis Pharma and Bristol-Myers Squibb. K.H.M.: BMS (consultancy and honoraria), AbbVie (honoraria, research funding), Pfizer (honoraria), Otsuka (honoraria), Janssen (honoraria) and Novartis (consultancy). U.P.: Syros (consultancy, honoraria, research funding), MDS Foundation (membership on an entity\u0026rsquo;s Board of Directors or advisory committees), Silence Therapeutics (consultancy, honoraria, research funding), Celgene (honoraria), Takeda (consultancy, honoraria, research funding), Fibrogen (research funding), Servier (consultancy, honoraria, research funding), Roche (research funding), Merck (research funding), Amgen (consultancy, research funding), Novartis (consultancy, honoraria, research funding), AbbVie (consultancy), Curis (consultancy, research funding), Janssen Biotech (consultancy, research funding), Jazz (consultancy, honoraria, research funding), BeiGene (research funding), Geron (consultancy, research funding) and Bristol-Myers Squibb (consultancy, honoraria, membership on an entity\u0026rsquo;s Board of Directors or advisory committees, other, travel support, medical writing support, research funding). M.J.: Novartis (honoraria), Amgen (honoraria), Pfizer (honoraria), Blueprint Medicine (honoraria), BMS (honoraria) and Jazz (honoraria). L. C. H has received funding for clinical trials from Alexion, Amolyte, and Ascendis to his institution, and honoraria from Ascendis, Amgen, and UCB for consultancies and adboards to himself. VV gave advisory boards for Janssen Cilag, BMS Celgene, MSD, Novartis, Sobi, Caribou and received honoraria from Novartis, Gilead Kite, BMS Celgene, Janssen Cilag, Sobi, Amgen, Abbvie, Takeda. All other authors declare no competing interests.\u003c/p\u003e\u003ch2\u003eData availability\u003c/h2\u003e\n\u003cp\u003eThe single-cell sequencing data (RNA, BCR, TCR and ADT) supporting the study\u0026apos;s findings have been deposited in the Gene Expression Omnibus under accession code GSE234261\u003csup\u003e7\u003c/sup\u003e for sequencing run 1. Raw data and Cell ranger outputs for sequencing run 2-4 and pre-processed Seurat objects are available in the European Genome-Phenome Archive (accession code EGAXXXXXXXX) under restricted access.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eCode availability\u003c/h2\u003e\n\u003cp\u003eProcessing and analysis code related to this study is deposited in a GitHub repository at https://github.com/fraunhofer-izi/Rade_et_al_CAR_2025\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSwan, D., Madduri, D. \u0026amp; Hocking, J. CAR-T cell therapy in Multiple Myeloma: current status and future challenges. \u003cem\u003eBlood cancer journal \u003c/em\u003e\u003cstrong\u003e14, \u003c/strong\u003e206; 10.1038/s41408-024-01191-8 (2024).\u003c/li\u003e\n\u003cli\u003eMunshi, N. C.\u003cem\u003e et al. \u003c/em\u003eIdecabtagene Vicleucel in Relapsed and Refractory Multiple Myeloma. \u003cem\u003eThe New England journal of medicine \u003c/em\u003e\u003cstrong\u003e384, \u003c/strong\u003e705\u0026ndash;716; 10.1056/NEJMoa2024850 (2021).\u003c/li\u003e\n\u003cli\u003eBerdeja, J. G.\u003cem\u003e et al. \u003c/em\u003eCiltacabtagene autoleucel, a B-cell maturation antigen-directed chimeric antigen receptor T-cell therapy in patients with relapsed or refractory multiple myeloma (CARTITUDE-1): a phase 1b/2 open-label study. \u003cem\u003eLancet (London, England) \u003c/em\u003e\u003cstrong\u003e398, \u003c/strong\u003e314\u0026ndash;324; 10.1016/S0140-6736(21)00933-8 (2021).\u003c/li\u003e\n\u003cli\u003eCordas Dos Santos, D. M.\u003cem\u003e et al. \u003c/em\u003eA systematic review and meta-analysis of nonrelapse mortality after CAR T cell therapy. \u003cem\u003eNature medicine \u003c/em\u003e\u003cstrong\u003e30, \u003c/strong\u003e2667\u0026ndash;2678; 10.1038/s41591-024-03084-6 (2024).\u003c/li\u003e\n\u003cli\u003eMerz, M.\u003cem\u003e et al. \u003c/em\u003eIdecabtagene vicleucel or ciltacabtagene autoleucel for relapsed or refractory multiple myeloma: An international multicenter study. \u003cem\u003eHemaSphere \u003c/em\u003e\u003cstrong\u003e9, \u003c/strong\u003ee70070; 10.1002/hem3.70070 (2025).\u003c/li\u003e\n\u003cli\u003eSun, Y.\u003cem\u003e et al. \u003c/em\u003eAntigen-induced chimeric antigen receptor multimerization amplifies on-tumor cytotoxicity. \u003cem\u003eSignal transduction and targeted therapy \u003c/em\u003e\u003cstrong\u003e8, \u003c/strong\u003e445; 10.1038/s41392-023-01686-z (2023).\u003c/li\u003e\n\u003cli\u003eRade, M.\u003cem\u003e et al. \u003c/em\u003eSingle-cell multiomic dissection of response and resistance to chimeric antigen receptor T cells against BCMA in relapsed multiple myeloma. \u003cem\u003eNature cancer \u003c/em\u003e\u003cstrong\u003e5, \u003c/strong\u003e1318\u0026ndash;1333; 10.1038/s43018-024-00763-8 (2024).\u003c/li\u003e\n\u003cli\u003eHaradhvala, N. J.\u003cem\u003e et al. \u003c/em\u003eDistinct cellular dynamics associated with response to CAR-T therapy for refractory B cell lymphoma. \u003cem\u003eNature medicine \u003c/em\u003e\u003cstrong\u003e28, \u003c/strong\u003e1848\u0026ndash;1859; 10.1038/s41591-022-01959-0 (2022).\u003c/li\u003e\n\u003cli\u003eQin, C.\u003cem\u003e et al. \u003c/em\u003eSingle-cell analysis of anti-BCMA CAR T cell therapy in patients with central nervous system autoimmunity. \u003cem\u003eScience immunology \u003c/em\u003e\u003cstrong\u003e9, \u003c/strong\u003eeadj9730; 10.1126/sciimmunol.adj9730 (2024).\u003c/li\u003e\n\u003cli\u003eTian, D.-S.\u003cem\u003e et al. \u003c/em\u003eB cell lineage reconstitution underlies CAR-T cell therapeutic efficacy in patients with refractory myasthenia gravis. \u003cem\u003eEMBO molecular medicine \u003c/em\u003e\u003cstrong\u003e16, \u003c/strong\u003e966\u0026ndash;987; 10.1038/s44321-024-00043-z (2024).\u003c/li\u003e\n\u003cli\u003eHansen, D. K.\u003cem\u003e et al. \u003c/em\u003eComparison of Standard-of-Care Idecabtagene Vicleucel and Ciltacabtagene Autoleucel in Relapsed/Refractory Multiple Myeloma. \u003cem\u003eJournal of clinical oncology : official journal of the American Society of Clinical Oncology, \u003c/em\u003eJCO2401730; 10.1200/JCO-24-01730 (2025).\u003c/li\u003e\n\u003cli\u003eFandrei, D.\u003cem\u003e et al. \u003c/em\u003eBispecific antibodies as bridging to BCMA CAR-T cell therapy for relapsed/refractory multiple myeloma. \u003cem\u003eBlood cancer discovery\u003c/em\u003e; 10.1158/2643-3230.BCD-24-0118 (2024).\u003c/li\u003e\n\u003cli\u003eAndreatta, M.\u003cem\u003e et al. \u003c/em\u003eInterpretation of T cell states from single-cell transcriptomics data using reference atlases. \u003cem\u003eNature communications \u003c/em\u003e\u003cstrong\u003e12, \u003c/strong\u003e2965; 10.1038/s41467-021-23324-4 (2021).\u003c/li\u003e\n\u003cli\u003eKharchenko, P. V., Silberstein, L. \u0026amp; Scadden, D. T. Bayesian approach to single-cell differential expression analysis. \u003cem\u003eNature methods \u003c/em\u003e\u003cstrong\u003e11, \u003c/strong\u003e740\u0026ndash;742; 10.1038/nmeth.2967 (2014).\u003c/li\u003e\n\u003cli\u003eMensurado, S., Blanco-Dom\u0026iacute;nguez, R. \u0026amp; Silva-Santos, B. The emerging roles of \u0026gamma;\u0026delta; T cells in cancer immunotherapy. \u003cem\u003eNature reviews. Clinical oncology \u003c/em\u003e\u003cstrong\u003e20, \u003c/strong\u003e178\u0026ndash;191; 10.1038/s41571-022-00722-1 (2023).\u003c/li\u003e\n\u003cli\u003eKang, J.\u003cem\u003e et al. \u003c/em\u003eSystematic dissection of tumor-normal single-cell ecosystems across a thousand tumors of 30 cancer types. \u003cem\u003eNature communications \u003c/em\u003e\u003cstrong\u003e15, \u003c/strong\u003e4067; 10.1038/s41467-024-48310-4 (2024).\u003c/li\u003e\n\u003cli\u003eWu, H.\u003cem\u003e et al. \u003c/em\u003eMitochondrial dysfunction promotes the transition of precursor to terminally exhausted T cells through HIF-1\u0026alpha;-mediated glycolytic reprogramming. \u003cem\u003eNature communications \u003c/em\u003e\u003cstrong\u003e14, \u003c/strong\u003e6858; 10.1038/s41467-023-42634-3 (2023).\u003c/li\u003e\n\u003cli\u003eYu, Y.-R.\u003cem\u003e et al. \u003c/em\u003eDisturbed mitochondrial dynamics in CD8+ TILs reinforce T cell exhaustion. \u003cem\u003eNature immunology \u003c/em\u003e\u003cstrong\u003e21, \u003c/strong\u003e1540\u0026ndash;1551; 10.1038/s41590-020-0793-3 (2020).\u003c/li\u003e\n\u003cli\u003eHuang, Y.\u003cem\u003e et al. \u003c/em\u003eInhibition of CD38 enzymatic activity enhances CAR-T cell immune-therapeutic efficacy by repressing glycolytic metabolism. \u003cem\u003eCell reports. Medicine \u003c/em\u003e\u003cstrong\u003e5, \u003c/strong\u003e101400; 10.1016/j.xcrm.2024.101400 (2024).\u003c/li\u003e\n\u003cli\u003eWagner, T. R.\u003cem\u003e et al. \u003c/em\u003eTumor Reactive T Cells Drive Clinical Responses to Multiple Myeloma. \u003cem\u003eBlood \u003c/em\u003e\u003cstrong\u003e144, \u003c/strong\u003e1890; 10.1182/blood-2024-210272 (2024).\u003c/li\u003e\n\u003cli\u003eLiu, T., Zhang, L., Joo, D. \u0026amp; Sun, S.-C. NF-\u0026kappa;B signaling in inflammation. \u003cem\u003eSignal transduction and targeted therapy \u003c/em\u003e\u003cstrong\u003e2, \u003c/strong\u003e17023-; 10.1038/sigtrans.2017.23 (2017).\u003c/li\u003e\n\u003cli\u003eHendriks, J.\u003cem\u003e et al. \u003c/em\u003eCD27 is required for generation and long-term maintenance of T cell immunity. \u003cem\u003eNature immunology \u003c/em\u003e\u003cstrong\u003e1, \u003c/strong\u003e433\u0026ndash;440; 10.1038/80877 (2000).\u003c/li\u003e\n\u003cli\u003eJaeger-Ruckstuhl, C. A.\u003cem\u003e et al. \u003c/em\u003eSignaling via a CD27-TRAF2-SHP-1 axis during naive T cell activation promotes memory-associated gene regulatory networks. \u003cem\u003eImmunity \u003c/em\u003e\u003cstrong\u003e57, \u003c/strong\u003e287-302.e12; 10.1016/j.immuni.2024.01.011 (2024).\u003c/li\u003e\n\u003cli\u003eCibri\u0026aacute;n, D. \u0026amp; S\u0026aacute;nchez-Madrid, F. CD69: from activation marker to metabolic gatekeeper. \u003cem\u003eEuropean journal of immunology \u003c/em\u003e\u003cstrong\u003e47, \u003c/strong\u003e946\u0026ndash;953; 10.1002/eji.201646837 (2017).\u003c/li\u003e\n\u003cli\u003eKunes, R. Z., Walle, T., Land, M., Nawy, T. \u0026amp; Pe\u0026apos;er, D. Supervised discovery of interpretable gene programs from single-cell data. \u003cem\u003eNature biotechnology \u003c/em\u003e\u003cstrong\u003e42, \u003c/strong\u003e1084\u0026ndash;1095; 10.1038/s41587-023-01940-3 (2024).\u003c/li\u003e\n\u003cli\u003eNeitzke-Montinelli, V.\u003cem\u003e et al. \u003c/em\u003eDifferentiation of Memory CD8 T Cells Unravel Gene Expression Pattern Common to Effector and Memory Precursors. \u003cem\u003eFrontiers in immunology \u003c/em\u003e\u003cstrong\u003e13, \u003c/strong\u003e840203; 10.3389/fimmu.2022.840203 (2022).\u003c/li\u003e\n\u003cli\u003eLee, H.\u003cem\u003e et al. \u003c/em\u003eImpact of soluble BCMA and non-T-cell factors on refractoriness to BCMA-targeting T-cell engagers in multiple myeloma. \u003cem\u003eBlood \u003c/em\u003e\u003cstrong\u003e144, \u003c/strong\u003e2637\u0026ndash;2651; 10.1182/blood.2024026212 (2024).\u003c/li\u003e\n\u003cli\u003eFreeman, C. L.\u003cem\u003e et al. \u003c/em\u003eTumor burden quantified by Soluble B-Cell Maturation Antigen and Metabolic Tumor Volume determine myeloma CAR-T outcomes. \u003cem\u003eBlood\u003c/em\u003e; 10.1182/blood.2024024965 (2024).\u003c/li\u003e\n\u003cli\u003eHansen, D. K.\u003cem\u003e et al. \u003c/em\u003eIdecabtagene Vicleucel for Relapsed/Refractory Multiple Myeloma: Real-World Experience From the Myeloma CAR T Consortium. \u003cem\u003eJournal of clinical oncology : official journal of the American Society of Clinical Oncology \u003c/em\u003e\u003cstrong\u003e41, \u003c/strong\u003e2087\u0026ndash;2097; 10.1200/JCO.22.01365 (2023).\u003c/li\u003e\n\u003cli\u003eSidana, S.\u003cem\u003e et al. \u003c/em\u003eSafety and efficacy of standard-of-care ciltacabtagene autoleucel for relapsed/refractory multiple myeloma. \u003cem\u003eBlood \u003c/em\u003e\u003cstrong\u003e145, \u003c/strong\u003e85\u0026ndash;97; 10.1182/blood.2024025945 (2025).\u003c/li\u003e\n\u003cli\u003eGagelmann, N.\u003cem\u003e et al. \u003c/em\u003eDevelopment and Validation of a Prediction Model of Outcome After B-Cell Maturation Antigen-Directed Chimeric Antigen Receptor T-Cell Therapy in Relapsed/Refractory Multiple Myeloma. \u003cem\u003eJournal of clinical oncology : official journal of the American Society of Clinical Oncology \u003c/em\u003e\u003cstrong\u003e42, \u003c/strong\u003e1665\u0026ndash;1675; 10.1200/JCO.23.02232 (2024).\u003c/li\u003e\n\u003cli\u003eFurqan, F.\u003cem\u003e et al. \u003c/em\u003eOutpatient administration of CAR T-cell therapies using a strategy of no remote monitoring and early CRS intervention. \u003cem\u003eBlood advances \u003c/em\u003e\u003cstrong\u003e8, \u003c/strong\u003e4320\u0026ndash;4329; 10.1182/bloodadvances.2024013239 (2024).\u003c/li\u003e\n\u003cli\u003eKohler, M. E. \u0026amp; Fry, T. J. CD4+ CAR T cells - more than helpers. \u003cem\u003eNature cancer \u003c/em\u003e\u003cstrong\u003e4, \u003c/strong\u003e928\u0026ndash;929; 10.1038/s43018-023-00567-2 (2023).\u003c/li\u003e\n\u003cli\u003eMelenhorst, J. J.\u003cem\u003e et al. \u003c/em\u003eDecade-long leukaemia remissions with persistence of CD4+ CAR T cells. \u003cem\u003eNature \u003c/em\u003e\u003cstrong\u003e602, \u003c/strong\u003e503\u0026ndash;509; 10.1038/s41586-021-04390-6 (2022).\u003c/li\u003e\n\u003cli\u003eLee, H.\u003cem\u003e et al. \u003c/em\u003eMechanisms of antigen escape from BCMA- or GPRC5D-targeted immunotherapies in multiple myeloma. \u003cem\u003eNature medicine \u003c/em\u003e\u003cstrong\u003e29, \u003c/strong\u003e2295\u0026ndash;2306; 10.1038/s41591-023-02491-5 (2023).\u003c/li\u003e\n\u003cli\u003eMajzner, R. G.\u003cem\u003e et al. \u003c/em\u003eTuning the Antigen Density Requirement for CAR T-cell Activity. \u003cem\u003eCancer discovery \u003c/em\u003e\u003cstrong\u003e10, \u003c/strong\u003e702\u0026ndash;723; 10.1158/2159-8290.CD-19-0945 (2020).\u003c/li\u003e\n\u003cli\u003eKarnell, J. L.\u003cem\u003e et al. \u003c/em\u003eDepleting plasmacytoid dendritic cells reduces local type I interferon responses and disease activity in patients with cutaneous lupus. \u003cem\u003eScience translational medicine \u003c/em\u003e\u003cstrong\u003e13\u003c/strong\u003e; 10.1126/scitranslmed.abf8442 (2021).\u003c/li\u003e\n\u003cli\u003eGriffin, G. K.\u003cem\u003e et al. \u003c/em\u003eUltraviolet radiation shapes dendritic cell leukaemia transformation in the skin. \u003cem\u003eNature \u003c/em\u003e\u003cstrong\u003e618, \u003c/strong\u003e834\u0026ndash;841; 10.1038/s41586-023-06156-8 (2023).\u003c/li\u003e\n\u003cli\u003eKumar, S.\u003cem\u003e et al. \u003c/em\u003eInternational Myeloma Working Group consensus criteria for response and minimal residual disease assessment in multiple myeloma. \u003cem\u003eThe Lancet. Oncology \u003c/em\u003e\u003cstrong\u003e17, \u003c/strong\u003ee328-e346; 10.1016/S1470-2045(16)30206-6 (2016).\u003c/li\u003e\n\u003cli\u003eLee, D. W.\u003cem\u003e et al. \u003c/em\u003eASTCT Consensus Grading for Cytokine Release Syndrome and Neurologic Toxicity Associated with Immune Effector Cells. \u003cem\u003eBiology of blood and marrow transplantation : journal of the American Society for Blood and Marrow Transplantation \u003c/em\u003e\u003cstrong\u003e25, \u003c/strong\u003e625\u0026ndash;638; 10.1016/j.bbmt.2018.12.758 (2019).\u003c/li\u003e\n\u003cli\u003eBlache, U.\u003cem\u003e et al. \u003c/em\u003eAdvanced Flow Cytometry Assays for Immune Monitoring of CAR-T Cell Applications. \u003cem\u003eFrontiers in immunology \u003c/em\u003e\u003cstrong\u003e12, \u003c/strong\u003e658314; 10.3389/fimmu.2021.658314 (2021).\u003c/li\u003e\n\u003cli\u003eZheng, L.\u003cem\u003e et al. \u003c/em\u003ePan-cancer single-cell landscape of tumor-infiltrating T cells. \u003cem\u003eScience (New York, N.Y.) \u003c/em\u003e\u003cstrong\u003e374, \u003c/strong\u003eabe6474; 10.1126/science.abe6474 (2021).\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eCampbell, T. et al. Uses of anti-BCMA chimeric antigen receptors. Worldwide patent WO2021091978A1 (2021). ,\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eSchecter, J. M. \u0026amp; Fan, X. BCMA-targeted CAR-T cell therapy for multiple myeloma. Worldwide patent WO2022116086A1 (2022). ,\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003eHeaton, H.\u003cem\u003e et al. \u003c/em\u003eSouporcell: robust clustering of single-cell RNA-seq data by genotype without reference genotypes. \u003cem\u003eNature methods \u003c/em\u003e\u003cstrong\u003e17, \u003c/strong\u003e615\u0026ndash;620; 10.1038/s41592-020-0820-1 (2020).\u003c/li\u003e\n\u003cli\u003eNeavin, D.\u003cem\u003e et al. \u003c/em\u003eDemuxafy: improvement in droplet assignment by integrating multiple single-cell demultiplexing and doublet detection methods. \u003cem\u003eGenome biology \u003c/em\u003e\u003cstrong\u003e25, \u003c/strong\u003e94; 10.1186/s13059-024-03224-8 (2024).\u003c/li\u003e\n\u003cli\u003eGermain, P.-L., Lun, A., Garcia Meixide, C., Macnair, W. \u0026amp; Robinson, M. D. Doublet identification in single-cell sequencing data using scDblFinder. \u003cem\u003eF1000Research \u003c/em\u003e\u003cstrong\u003e10, \u003c/strong\u003e979; 10.12688/f1000research.73600.2 (2021).\u003c/li\u003e\n\u003cli\u003eBais, A. S. \u0026amp; Kostka, D. scds: computational annotation of doublets in single-cell RNA sequencing data. \u003cem\u003eBioinformatics (Oxford, England) \u003c/em\u003e\u003cstrong\u003e36, \u003c/strong\u003e1150\u0026ndash;1158; 10.1093/bioinformatics/btz698 (2020).\u003c/li\u003e\n\u003cli\u003eHao, Y.\u003cem\u003e et al. \u003c/em\u003eIntegrated analysis of multimodal single-cell data. \u003cem\u003eCell \u003c/em\u003e\u003cstrong\u003e184, \u003c/strong\u003e3573-3587.e29; 10.1016/j.cell.2021.04.048 (2021).\u003c/li\u003e\n\u003cli\u003eStuart, T.\u003cem\u003e et al. \u003c/em\u003eComprehensive Integration of Single-Cell Data. \u003cem\u003eCell \u003c/em\u003e\u003cstrong\u003e177, \u003c/strong\u003e1888-1902.e21; 10.1016/j.cell.2019.05.031 (2019).\u003c/li\u003e\n\u003cli\u003eAndreatta, M., Berenstein, A. J. \u0026amp; Carmona, S. J. scGate: marker-based purification of cell types from heterogeneous single-cell RNA-seq datasets. \u003cem\u003eBioinformatics (Oxford, England) \u003c/em\u003e\u003cstrong\u003e38, \u003c/strong\u003e2642\u0026ndash;2644; 10.1093/bioinformatics/btac141 (2022).\u003c/li\u003e\n\u003cli\u003eAndreatta, M. \u0026amp; Carmona, S. J. UCell: Robust and scalable single-cell gene signature scoring. \u003cem\u003eComputational and structural biotechnology journal \u003c/em\u003e\u003cstrong\u003e19, \u003c/strong\u003e3796\u0026ndash;3798; 10.1016/j.csbj.2021.06.043 (2021).\u003c/li\u003e\n\u003cli\u003eFu, R.\u003cem\u003e et al. \u003c/em\u003eclustifyr: an R package for automated single-cell RNA sequencing cluster classification. \u003cem\u003eF1000Research \u003c/em\u003e\u003cstrong\u003e9, \u003c/strong\u003e223; 10.12688/f1000research.22969.2 (2020).\u003c/li\u003e\n\u003cli\u003eKorsunsky, I.\u003cem\u003e et al. \u003c/em\u003eFast, sensitive and accurate integration of single-cell data with Harmony. \u003cem\u003eNature methods \u003c/em\u003e\u003cstrong\u003e16, \u003c/strong\u003e1289\u0026ndash;1296; 10.1038/s41592-019-0619-0 (2019).\u003c/li\u003e\n\u003cli\u003eFinak, G.\u003cem\u003e et al. \u003c/em\u003eMAST: a flexible statistical framework for assessing transcriptional changes and characterizing heterogeneity in single-cell RNA sequencing data. \u003cem\u003eGenome biology \u003c/em\u003e\u003cstrong\u003e16, \u003c/strong\u003e278; 10.1186/s13059-015-0844-5 (2015).\u003c/li\u003e\n\u003cli\u003eAnderson, N. D.\u003cem\u003e et al. \u003c/em\u003eTranscriptional signatures associated with persisting CD19 CAR-T cells in children with leukemia. \u003cem\u003eNature medicine \u003c/em\u003e\u003cstrong\u003e29, \u003c/strong\u003e1700\u0026ndash;1709; 10.1038/s41591-023-02415-3 (2023).\u003c/li\u003e\n\u003cli\u003eKorotkevich, G.\u003cem\u003e et al.\u003c/em\u003e \u003cem\u003eFast gene set enrichment analysis \u003c/em\u003e(2016).\u003c/li\u003e\n\u003cli\u003eLiberzon, A.\u003cem\u003e et al. \u003c/em\u003eThe Molecular Signatures Database (MSigDB) hallmark gene set collection. \u003cem\u003eCell systems \u003c/em\u003e\u003cstrong\u003e1, \u003c/strong\u003e417\u0026ndash;425; 10.1016/j.cels.2015.12.004 (2015).\u003c/li\u003e\n\u003cli\u003eChu, Y.\u003cem\u003e et al. \u003c/em\u003ePan-cancer T cell atlas links a cellular stress response state to immunotherapy resistance. \u003cem\u003eNature medicine \u003c/em\u003e\u003cstrong\u003e29, \u003c/strong\u003e1550\u0026ndash;1562; 10.1038/s41591-023-02371-y (2023).\u003c/li\u003e\n\u003cli\u003eYu, G. Gene Ontology Semantic Similarity Analysis Using GOSemSim. \u003cem\u003eMethods in molecular biology (Clifton, N.J.) \u003c/em\u003e\u003cstrong\u003e2117, \u003c/strong\u003e207\u0026ndash;215; 10.1007/978-1-0716-0301-7_11 (2020).\u003c/li\u003e\n\u003cli\u003ePhipson, B.\u003cem\u003e et al. \u003c/em\u003epropeller: testing for differences in cell type proportions in single cell data. \u003cem\u003eBioinformatics (Oxford, England) \u003c/em\u003e\u003cstrong\u003e38, \u003c/strong\u003e4720\u0026ndash;4726; 10.1093/bioinformatics/btac582 (2022).\u003c/li\u003e\n\u003cli\u003eRossi, J.\u003cem\u003e et al. \u003c/em\u003ePreinfusion polyfunctional anti-CD19 chimeric antigen receptor T cells are associated with clinical outcomes in NHL. \u003cem\u003eBlood \u003c/em\u003e\u003cstrong\u003e132, \u003c/strong\u003e804\u0026ndash;814; 10.1182/blood-2018-01-828343 (2018).\u003c/li\u003e\n\u003cli\u003eBorcherding, N., Bormann, N. L. \u0026amp; Kraus, G. scRepertoire: An R-based toolkit for single-cell immune receptor analysis. \u003cem\u003eF1000Research \u003c/em\u003e\u003cstrong\u003e9, \u003c/strong\u003e47; 10.12688/f1000research.22139.2. (2020).\u003c/li\u003e\n\u003cli\u003eZhang, L.\u003cem\u003e et al. \u003c/em\u003eLineage tracking reveals dynamic relationships of T cells in colorectal cancer. \u003cem\u003eNature \u003c/em\u003e\u003cstrong\u003e564, \u003c/strong\u003e268\u0026ndash;272; 10.1038/s41586-018-0694-x (2018).\u003c/li\u003e\n\u003cli\u003eAngerer, P.\u003cem\u003e et al. \u003c/em\u003edestiny: diffusion maps for large-scale single-cell data in R. \u003cem\u003eBioinformatics (Oxford, England) \u003c/em\u003e\u003cstrong\u003e32, \u003c/strong\u003e1241\u0026ndash;1243; 10.1093/bioinformatics/btv715 (2016).\u003c/li\u003e\n\u003cli\u003eStreet, K.\u003cem\u003e et al. \u003c/em\u003eSlingshot: cell lineage and pseudotime inference for single-cell transcriptomics. \u003cem\u003eBMC genomics \u003c/em\u003e\u003cstrong\u003e19, \u003c/strong\u003e477; 10.1186/s12864-018-4772-0 (2018).\u003c/li\u003e\n\u003cli\u003eLun, A. T. L., McCarthy, D. J. \u0026amp; Marioni, J. C. A step-by-step workflow for low-level analysis of single-cell RNA-seq data with Bioconductor. \u003cem\u003eF1000Research \u003c/em\u003e\u003cstrong\u003e5, \u003c/strong\u003e2122; 10.12688/f1000research.9501.2 (2016).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Multiple myeloma, single-cell sequencing, chimeric antigen receptor T cells, B-cell maturation antigen","lastPublishedDoi":"10.21203/rs.3.rs-6165798/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6165798/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Chimeric Antigen Receptor (CAR) T-cell therapies targeting B-cell maturation antigen (BCMA) have transformed the treatment landscape for relapsed/refractory multiple myeloma (RRMM). In this study, we present a real world cohort of 61 RRMM patients treated with idecabtagene vicleucel (Ide-cel, n=34) and ciltacabtagene autoleucel (Cilta-cel, n=27). Cilta-cel demonstrated superior complete response (CR) rates (CR: 78% vs. 38%, p \u003c 0.001) and longer progression-free survival (PFS), with a distinct CAR-T expansion profile marked by increased CD4+CAR+/CD8+CAR+ ratio. To gain insights into immune dynamics encompassing CAR-T cell infusion with either product, we developed a longitudinal multi-omics single-cell atlas using 135 peripheral blood samples from 57 of the 61 patients. There was a strong association between CD4+ cytotoxic T cells and treatment with Cilta-cel, CR and CRS occurrence. Analysis of T cell receptor repertoires showed higher clonality in CD4 T cells in CR patients at all time points. CD8 T cells of non-CR patients showed transcriptomic changes in line with impaired effector function after CAR-T infusion. The BCMA expressing circulating plasma cells, B-cells and plasmacytoid dendritic cells were depleted after infusion in a response-dependent manner, with Cilta-cel leading to significantly slower B-cell recovery (p=0.03). Increased soluble BCMA reduction between day 0 and 30 was linked to stronger CAR-T expansion and higher CRP levels, suggesting an association of tumor debulking and systemic inflammation (p \u003c 0.01, respectively). Our analyses provide a comprehensive resource for understanding longitudinal cellular kinetics in RRMM patients treated with BCMA-directed CAR-T cells.","manuscriptTitle":"A longitudinal single-cell atlas to predict outcome and toxicity after BCMA-directed CAR T cell therapy in multiple myeloma","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-03-18 08:09:38","doi":"10.21203/rs.3.rs-6165798/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"64780168-e69d-464e-948d-53016d0c067e","owner":[],"postedDate":"March 18th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":45821027,"name":"Health sciences/Medical research/Translational research"},{"id":45821028,"name":"Health sciences/Diseases/Cancer/Haematological cancer/Myeloma"},{"id":45821029,"name":"Biological sciences/Cancer/Cancer therapy"}],"tags":[],"updatedAt":"2025-04-10T13:57:24+00:00","versionOfRecord":[],"versionCreatedAt":"2025-03-18 08:09:38","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6165798","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6165798","identity":"rs-6165798","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}