{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,26]],"date-time":"2026-06-26T12:49:16Z","timestamp":1782478156760,"version":"3.54.5"},"reference-count":33,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,6,12]],"date-time":"2026-06-12T00:00:00Z","timestamp":1781222400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Immunol."],"abstract":"\n Background<\/jats:title>\n \n Cytomegalovirus (CMV) reactivation occurs in\n \n \n \n 30<\/mml:mn>\n \u2212<\/mml:mo>\n 40<\/mml:mn>\n %<\/mml:mo>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n of seropositive patients receiving chimeric antigen receptor T-cell (CAR-T) therapy for B-cell lymphoma and is strongly associated with treatment failure. However, the causal immunological mechanism driving this failure whether through direct viral cytopathic effects, T-cell exhaustion, or resource competition remains undefined. Furthermore, existing predictive models lack the mechanistic insight needed to guide intervention. We developed a privacy-preserving, mechanistic digital twin to test the \u201ccytokine sink\u201d hypothesis, wherein CMV-specific CD8+ T cells compete with CAR-T cells for the limiting homeostatic cytokine\n \n \n \n I<\/mml:mi>\n L<\/mml:mi>\n \u2212<\/mml:mo>\n 15<\/mml:mn>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n .\n <\/jats:p>\n <\/jats:sec>\n \n Methods<\/jats:title>\n \n We constructed a system of ordinary differential equations formalizing this competition for\n \n \n \n I<\/mml:mi>\n L<\/mml:mi>\n \u2212<\/mml:mo>\n 15<\/mml:mn>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n . The model was trained using a Hierarchical Bayesian Federated Averaging algorithm on multi-institutional data from\n \n \n \n 414<\/mml:mn>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n patients across five academic cancer centres (Sites A-E). without centralizing raw patient data. An independent sixth centre (Site F,\n \n \n \n n<\/mml:mi>\n =<\/mml:mo>\n 89<\/mml:mn>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n ) served exclusively as the prospective validation site and contributed no data to the training process. In silico counterfactual analyses and simulations of antiviral prophylaxis were performed.\n <\/jats:p>\n <\/jats:sec>\n \n Results<\/jats:title>\n \n The digital twin predicted clinically significant CMV reactivation defined as CMV viral load\n \n \n \n \u2265<\/mml:mo>\n 10<\/mml:mn>\n ,<\/mml:mo>\n 000<\/mml:mn>\n \u00a0<\/mml:mo>\n I<\/mml:mi>\n U<\/mml:mi>\n \/<\/mml:mo>\n m<\/mml:mi>\n L<\/mml:mi>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n by\n \n \n \n D<\/mml:mi>\n a<\/mml:mi>\n y<\/mml:mi>\n \u00a0<\/mml:mo>\n 28<\/mml:mn>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n post-infusion (primary outcome; threshold selected based on\n \n \n \n E<\/mml:mi>\n C<\/mml:mi>\n I<\/mml:mi>\n L<\/mml:mi>\n \u2212<\/mml:mo>\n 7<\/mml:mn>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n guidance for threshold-triggered pre-emptive antiviral therapy in T-cell-depleted immunotherapy) with an AUROC of\n \n \n \n 0.91<\/mml:mn>\n \u00a0<\/mml:mo>\n (<\/mml:mo>\n 95<\/mml:mn>\n %<\/mml:mo>\n \u00a0<\/mml:mo>\n C<\/mml:mi>\n I<\/mml:mi>\n :<\/mml:mo>\n 0.88<\/mml:mn>\n \u2212<\/mml:mo>\n 0.94<\/mml:mn>\n )<\/mml:mo>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n , significantly outperforming existing clinical risk scores. Two pre-specified secondary thresholds were also analysed: (i) any detectable reactivation\n \n \n \n (<\/mml:mo>\n \u2265<\/mml:mo>\n 1<\/mml:mn>\n ,<\/mml:mo>\n 000<\/mml:mn>\n \u00a0<\/mml:mo>\n I<\/mml:mi>\n U<\/mml:mi>\n \/<\/mml:mo>\n m<\/mml:mi>\n L<\/mml:mi>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n ) and (ii) severe reactivation\n \n \n \n (<\/mml:mo>\n \u2265<\/mml:mo>\n 50<\/mml:mn>\n ,<\/mml:mo>\n 000<\/mml:mn>\n \u00a0<\/mml:mo>\n I<\/mml:mi>\n U<\/mml:mi>\n \/<\/mml:mo>\n m<\/mml:mi>\n L<\/mml:mi>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n ). Under the structural assumptions of the fitted mechanistic model, the data are quantitatively consistent with\n \n \n \n I<\/mml:mi>\n L<\/mml:mi>\n \u2212<\/mml:mo>\n 15<\/mml:mn>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n resource competition as a mechanism linking CMV reactivation to CAR-T impairment. Mechanistically, the model revealed that CMV reactivation was associated with reduced peak CAR-T expansion by\n \n \n \n 41.8<\/mml:mn>\n %<\/mml:mo>\n \u00a0<\/mml:mo>\n (<\/mml:mo>\n p<\/mml:mi>\n <<\/mml:mo>\n 0.001<\/mml:mn>\n )<\/mml:mo>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n in counterfactual analysis. Global sensitivity analysis identified the pre-infusion frequency of CMV-specific T-cell precursors\n \n \n \n (<\/mml:mo>\n \u03be<\/mml:mi>\n )<\/mml:mo>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n and the resource competition coefficient\n \n \n \n (<\/mml:mo>\n \u03b1<\/mml:mi>\n )<\/mml:mo>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n as the primary drivers of this effect, explaining\n \n \n \n 38<\/mml:mn>\n %<\/mml:mo>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n and\n \n \n \n 29<\/mml:mn>\n %<\/mml:mo>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n of output variance, respectively. In silico simulation of a risk-adapted, digital-twin-guided antiviral prophylaxis strategy reduced projected six-month progression by\n \n \n \n 32<\/mml:mn>\n %<\/mml:mo>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n while reducing aggregate drug exposure by\n \n \n \n 36<\/mml:mn>\n %<\/mml:mo>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n . In the prospective validation cohort, the model-predicted kinetic impairment independently predicted progression-free survival (hazard ratio [HR]\n \n \n \n 3.4<\/mml:mn>\n ,<\/mml:mo>\n \u00a0<\/mml:mo>\n 95<\/mml:mn>\n %<\/mml:mo>\n \u00a0<\/mml:mo>\n C<\/mml:mi>\n I<\/mml:mi>\n :<\/mml:mo>\n 1.5<\/mml:mn>\n \u2212<\/mml:mo>\n 7.8<\/mml:mn>\n ,<\/mml:mo>\n \u00a0<\/mml:mo>\n p<\/mml:mi>\n =<\/mml:mo>\n 0.004<\/mml:mn>\n )<\/mml:mo>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n . Formal model competition analysis against three alternative mechanistic hypotheses (Results Section 3.3) further supports the cytokine sink hypothesis over exhaustion-based or mediation-only alternatives.\n <\/jats:p>\n <\/jats:sec>\n \n Conclusion<\/jats:title>\n \n Under the structural assumptions of the fitted mechanistic model, the data are quantitatively consistent with\n \n \n \n I<\/mml:mi>\n L<\/mml:mi>\n \u2212<\/mml:mo>\n 15<\/mml:mn>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n resource competition as a mechanism linking CMV reactivation to CAR-T impairment. Formal model competition analysis against three alternative mechanistic hypotheses further supports the cytokine sink hypothesis; randomised interventional evidence is required for definitive causal proof. We further demonstrate that a privacy-preserving, mechanistic digital twin can serve as a clinically actionable tool for early risk stratification and personalized intervention, while providing a scalable blueprint for collaborative systems immunology research. Definitive validation of clinical utility requires a randomised controlled trial in which treatment decisions are prospectively guided by the digital twin\u2019s predictions.\n <\/jats:p>\n <\/jats:sec>","DOI":"10.3389\/fimmu.2026.1844927","type":"journal-article","created":{"date-parts":[[2026,6,12]],"date-time":"2026-06-12T05:48:36Z","timestamp":1781243316000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["A federated digital twin reveals cytomegalovirus reactivation impairs CAR-T cell therapy via IL-15-mediated cytokine competition in B-Cell lymphoma"],"prefix":"10.3389","volume":"17","author":[{"given":"Padmasri","family":"Sridharan","sequence":"first","affiliation":[{"name":"Department of Mathematics, School of Advanced Sciences, Vellore Institute of Technology","place":["Chennai, India"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mini","family":"Ghosh","sequence":"additional","affiliation":[{"name":"Department of Mathematics, School of Advanced Sciences, Vellore Institute of Technology","place":["Chennai, India"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,6,12]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","DOI":"10.1056\/NEJMoa2116133","article-title":"Axicabtagene ciloleucel as second-line therapy for large B-cell lymphoma","volume":"386","author":"Locke","year":"2022","journal-title":"N Engl J Med"},{"key":"B2","doi-asserted-by":"publisher","DOI":"10.1200\/JCO.22.02042","article-title":"CAR T-cell therapy in relapsed\/refractory B-cell Malignancies: 5-year follow-up","volume":"41","author":"Neelapu","year":"2023","journal-title":"J Clin Oncol"},{"key":"B3","doi-asserted-by":"publisher","DOI":"10.1182\/bloodadvances.2021006548","article-title":"Infectious complications of CAR T-cell therapy: a multicenter analysis","volume":"6","author":"Hill","year":"2022","journal-title":"Blood Adv"},{"key":"B4","doi-asserted-by":"publisher","DOI":"10.1038\/s41408-022-00606-y","article-title":"Infection during the first year in patients treated with CD19 CAR T cells for diffuse large B cell lymphoma","volume":"12","author":"Wudhikarn","year":"2022","journal-title":"Blood Cancer J"},{"key":"B5","doi-asserted-by":"publisher","DOI":"10.1038\/s41577-022-00819-7","article-title":"The role of cytokines in CAR T-cell expansion and persistence","volume":"23","author":"Mackall","year":"2023","journal-title":"Nat Rev Immunol"},{"key":"B6","doi-asserted-by":"publisher","DOI":"10.1016\/j.jtct.2023.02.015","article-title":"CMV reactivation after CD19 CAR T-cell therapy: risk factors and impact on survival","volume":"29","author":"Wudhikarn","year":"2023","journal-title":"Transplant Cell Ther"},{"key":"B7","doi-asserted-by":"publisher","DOI":"10.1126\/sciimmunol.add2203","article-title":"Engineering CAR T cells for enhanced cytokine signaling","volume":"8","author":"Guedan","year":"2023","journal-title":"Sci Immunol"},{"key":"B8","doi-asserted-by":"publisher","DOI":"10.1038\/s41590-023-01714-4","article-title":"T-cell stemness and persistence in adoptive immunotherapy","volume":"25","author":"Gattinoni","year":"2024","journal-title":"Nat Immunol"},{"key":"B9","doi-asserted-by":"publisher","DOI":"10.1016\/j.ccell.2023.08.008","article-title":"Mechanisms of CAR T-cell exhaustion and resistance","volume":"41","author":"Singh","year":"2023","journal-title":"Cancer Cell"},{"key":"B10","doi-asserted-by":"publisher","DOI":"10.1038\/s41571-023-00847-6","article-title":"Digital twins in oncology: from theory to clinical application","volume":"21","author":"Bjornson","year":"2024","journal-title":"Nat Rev Clin Oncol"},{"key":"B11","doi-asserted-by":"publisher","DOI":"10.1016\/j.cels.2023.05.003","article-title":"Whole-cell models for personalized cancer therapy","volume":"14","author":"Karr","year":"2023","journal-title":"Cell Syst"},{"key":"B12","doi-asserted-by":"publisher","DOI":"10.1163\/15718093-12665678","article-title":"GDPR compliance in multi-institutional health research","volume":"30","author":"Voigt","year":"2023","journal-title":"Eur J Health Law"},{"key":"B13","doi-asserted-by":"publisher","first-page":"25","DOI":"10.1038\/s42256-023-00776-9","article-title":"Federated learning for medical imaging: challenges and opportunities","volume":"6","author":"Rieke","year":"2024","journal-title":"Nat Mach Intell"},{"key":"B14","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1561\/9781638280334","article-title":"Privacy and robustness in federated learning","volume":"18","author":"Kairouz","year":"2025","journal-title":"Found Trends Mach Learn"},{"key":"B15","doi-asserted-by":"publisher","DOI":"10.1136\/bmj-2023-075333","article-title":"TRIPOD+AI: a guideline for reporting AI-based prediction models","volume":"384","author":"Collins","year":"2024","journal-title":"BMJ"},{"key":"B16","volume-title":"Flexible Imputation of Missing Data","author":"van Buuren","year":"2023"},{"key":"B17","doi-asserted-by":"publisher","DOI":"10.1016\/j.jesp.2013.03.013","article-title":"Detecting outliers: Do not use standard deviation around the mean, use absolute deviation around the median","volume":"49","author":"Leys","year":"2013","journal-title":"J Exp Soc Psychol"},{"key":"B18","doi-asserted-by":"publisher","DOI":"10.1016\/j.jbi.2023.104582","article-title":"The OMOP common data model: evolution and applications","volume":"149","author":"Hripcsak","year":"2024","journal-title":"J BioMed Inform"},{"key":"B19","doi-asserted-by":"publisher","first-page":"58","DOI":"10.1007\/s11538-023-01173-4","article-title":"Mathematical models of CAR T-cell--tumor interactions","volume":"85","author":"de Pillis","year":"2023","journal-title":"Bull Math Biol"},{"key":"B20","doi-asserted-by":"publisher","DOI":"10.3389\/fimmu.2024.1402872","article-title":"Systems immunology of CAR T-cell therapy","volume":"15","author":"Kronik","year":"2024","journal-title":"Front Immunol"},{"key":"B21","doi-asserted-by":"publisher","DOI":"10.1145\/1089014.1089020","article-title":"SUNDIALS: Suite of nonlinear and differential\/algebraic equation solvers","volume":"31","author":"Hindmarsh","year":"2005","journal-title":"ACM Trans Math Softw"},{"key":"B22","article-title":"Communication-efficient learning of deep networks from decentralized data","author":"McMahan","year":"2017"},{"key":"B23","doi-asserted-by":"publisher","first-page":"1","DOI":"10.18637\/jss.v105.i01","article-title":"Stan 2.0: a probabilistic programming language for Bayesian inference","volume":"105","author":"Carpenter","year":"2024","journal-title":"J Stat Softw"},{"key":"B24","first-page":"45","article-title":"Practical secure aggregation for federated learning on large-scale data","author":"Bonawitz","year":"2023"},{"key":"B25","doi-asserted-by":"crossref","DOI":"10.1109\/CSF.2017.11","article-title":"R\u00e9nyi differential privacy","volume-title":"2017 IEEE 30th Computer Security Foundations Symposium (CSF)","author":"Mironov","year":"2017"},{"key":"B26","doi-asserted-by":"publisher","DOI":"10.1214\/ss\/1177011136","article-title":"Inference from iterative simulation using multiple sequences","volume":"7","author":"Gelman","year":"1992","journal-title":"Stat Sci"},{"key":"B27","doi-asserted-by":"publisher","DOI":"10.1136\/jitc-2022-006123","article-title":"Harmonization of immune monitoring assays for cellular therapy","volume":"11","author":"Gratama","year":"2023","journal-title":"J Immunother Cancer"},{"key":"B28","doi-asserted-by":"publisher","DOI":"10.1200\/JCO.23.02235","article-title":"Lugano 2023: revised criteria for lymphoma response assessment","volume":"42","author":"Cheson","year":"2024","journal-title":"J Clin Oncol"},{"key":"B29","doi-asserted-by":"publisher","DOI":"10.1016\/S0378-4754(00)00270-6","article-title":"Global sensitivity indices for nonlinear mathematical models and their Monte Carlo estimates","volume":"55","author":"Sobol","year":"2001","journal-title":"Math Comput Simul"},{"key":"B30","doi-asserted-by":"publisher","first-page":"45","DOI":"10.1111\/biom.14025","article-title":"Statistical methods for comparing ROC curves: an update","volume":"80","author":"DeLong","year":"2024","journal-title":"Biometrics"},{"key":"B31","volume-title":"R: A language and environment for statistical computing","year":"2023"},{"key":"B32","doi-asserted-by":"publisher","DOI":"10.1038\/nri3862","article-title":"Molecular and cellular insights into T cell exhaustion","volume":"15","author":"Wherry","year":"2015","journal-title":"Nat Rev Immunol"},{"key":"B33","doi-asserted-by":"publisher","DOI":"10.1038\/s42256-023-00665-3","article-title":"Secure federated learning in medical imaging","volume":"5","author":"Kaissis","year":"2023","journal-title":"Nat Mach Intell"}],"container-title":["Frontiers in Immunology"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fimmu.2026.1844927\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,26]],"date-time":"2026-06-26T12:20:24Z","timestamp":1782476424000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fimmu.2026.1844927\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6,12]]},"references-count":33,"alternative-id":["10.3389\/fimmu.2026.1844927"],"URL":"https:\/\/doi.org\/10.3389\/fimmu.2026.1844927","relation":{},"ISSN":["1664-3224"],"issn-type":[{"value":"1664-3224","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,6,12]]},"article-number":"1844927"}}