AUTHORS
xiaoyu meng, tianying chen2, chenxi Cai3, xiaobin song4, Yanan Yue5, zhixiang yan6, Jianxiang Wang Sr, MD, PhD7, tony ho8, yan zhang9, george church10, Luhan Yang, PhD11
AFFILIATIONS
1 Qihan Biotechnology, Hangzhou, China
2 Qihan Biotechnology, Hangzhou, China
3 Department of Pharmacology and Department of Pathology of Sir Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou, China
4 Qihan Biotechnology, Hangzhou, China
5 Qihan Biotechnology, Hangzhou, China
6 Qihan Biotechnology, Hangzhou, China
7 Institute of Hematology & Blood Diseases Hospital, CAMS/PUMC, tianjing, China
8 Qihan Biotechnology, Hangzhou, China
9 Department of Pharmacology and Department of Pathology of Sir Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou, China
10 Department of Genetics, Harvard Medical School, boston, United States
11 Qihan Biotechnology, Hangzhou, China
ABSTRACT
Background
Allogeneic CD19/BCMA dual-target CAR-T therapy offers an off-the-shelf approach to B cell–mediated autoimmune disease by simultaneously depleting pathogenic B cells and long-lived plasma cells. Our first-generation product (QT-019B) — in which TCR, HLA-A, HLA-B, and CIITA were disrupted while HLA-X was retained — achieved favorable persistence and encouraging clinical activity. However, retaining native HLA-X imposes donor–recipient HLA-X matching, which restrcits donor availability and undermines the off-the-shelf premise.
To overcome this limitation, we engineered the QT-019B backbone by knocking out HLA-X and introducing Edit-K, a set of AI-designed NK-biased ligands optimized to reduce NK-mediated rejection while limiting stimulation of allogeneic T cells. The resulting next-generation CD19/BCMA UCAR-T platform is designed to eliminate the need for HLA matching while preserving immune evasion, persistence, and therapeutic potency.
Methods
AI-guided design.
Structural models of Edit-K were predicted using AlphaFold 3 and evaluated by Rosetta-based interface energy scoring. We integrated structural modeling, interface analysis, and multi-objective optimization to select Edit-K candidates predicted to retain NK-inhibitory activity while reducing alloreactive recognition. Hundreds of candidate designs were computationally assessed, and the top 50 candidates were advanced to experimental testing.
Cell engineering and validation.
QT-219C CD19/BCMA dual-target UCAR-T cells were generated by multiplex editing on a fully HLA class I/II-null backbone together with overexpression of Edit-K and an additional efficacy-enhancing component. Candidates were screened for resistance to primary NK cytotoxicity across multiple healthy donors, and for allogeneic T-cell activation by mixed lymphocyte reaction. CAR-mediated cytotoxicity, in vivo efficacy, and GLP-oriented nonclinical safety were assessed. Clinical pharmacokinetics and safety were evaluated in the first autoimmune patient treated with QT-219C after standard lymphodepletion.
Results
Experimental validation confirmed the immune-evasive efficacy of Edit-K.
Even upon complete HLA class I depletion, Edit-K-expressing cells resisted primary NK-cell-mediated lysis, demonstrating protection against innate surveillance. In parallel, MLR showed allogeneic T-cell activation reduced to near negative-control levels. These findings showed that Edit-K was capable of balancing broad immune compatibility with persistent NK-cell resistance.
Potency and safety preserved.
Neither multiplex editing nor Edit-K expression impaired CAR-mediated cytotoxicity in vitro or antitumor/depletion efficacy in vivo. Nonclinical studies identified no treatment-related toxicological findings.
First-in-human pharmacokinetics.
In the first autoimmune patient treated with QT-219C following standard lymphodepletion, engineered cells showed in vivo expansion, reaching a peak of 1679 cells/ul on Day11 with sustained detectable persistence through Day 21, comparable to the HLA-X–matched first-generation platform despite complete HLA mismatch. No significant treatment-related adverse events occurred during early follow-up.
Conclusions
By treating NK evasion as a structural design problem rather than a matching constraint, we used AI-guided interface modeling to engineer Edit-K — NK-biased ligands that confer broad immune compatibility while maintaining innate immune resistance. Incorporated into a fully HLA class I/II–null CD19/BCMA UCAR-T backbone, Edit-K reproduced the NK protection of HLA-X matching without requiring it, retained therapeutic potency, and supported robust in vivo expansion and persistence in the first treated patient. This establishes computationally designed immune-evasion ligands as a viable route to donor-unrestricted CAR-T therapy for autoimmune disease.
DISCLOSURES
Luhan Yang, PhD: No relevant conflicts of interest to declare.