PO.IM01.06 · 免疫学
用于治疗胰腺癌的靶向GPC1纳米抗体CAR T细胞的结构导向蛋白质工程与人源化
Structure-guided protein engineering and humanization of GPC1-targeted nanobody CAR T cells for treating pancreatic cancer
作者与单位 Authors & Affiliations
摘要 Abstract
中文摘要
磷脂酰肌醇蛋白聚糖-1（GPC1）是一种硫酸乙酰肝素蛋白聚糖，在胰腺导管腺癌（PDAC）中过表达。我们此前使用单峰驼VHH纳米抗体D4和小鼠单克隆IgG抗体HM2开发了GPC1 CAR，它们分别识别膜远端表位和膜近端表位。在此，我们将结构建模与CAR功能数据及动物实验相结合，以定义表位位置和CAR几何结构如何共同决定GPC1 CAR的疗效。对包含CD8或IgG4铰链以及CD8或CD28跨膜（TM）结构域的CAR胞外域进行的建模表明，对于给定表位，只有特定组合才能产生与TCR-pMHC免疫突触相当的膜间距。这些几何预测与功能数据相符，其中D4-IgG4H-CD28TM CAR T细胞在T3M4腹腔内PDAC模型中快速消退了肿瘤的生物发光。总之，这些结果支持这样一个模型，即GPC1上表位的空间位置和CAR铰链/TM结构是靶向GPC1的CAR T细胞的关键设计参数。为促进D4 CAR T细胞的临床开发，VHH D4的人源化非常必要，尽管VHH等纳米抗体的人源化尚未得到充分确立。除了如我们此前在兔和小鼠抗体人源化中所描述的将CDR移植到最接近的种系框架外，我们还使用来自多个平台的AI预测结构模型，将预测的D4框架和CDR区域的空间几何结构和序列相似性与现有的抗体和纳米抗体结构进行比较。具有最高人源性评分、几何结构和相似性评分的人源化构建体被回复突变，以保留纳米抗体框架序列中的关键残基，然后评估其GPC1结合亲和力和细胞表面结合。利用这种AI辅助的结构导向策略，我们旨在为GPC1 CAR T疗法及其他基于纳米抗体的临床应用生成临床适用的人源化VHH纳米抗体支架。
查看英文原文 English abstract
Glypican-1 (GPC1) is a heparan sulfate proteoglycan that is overexpressed in pancreatic ductal adenocarcinoma (PDAC). We previously developed GPC1 CARs using the dromedary VHH nanobody D4 and the mouse monoclonal IgG antibody HM2, which recognize a membrane-distal epitope and a membrane-proximal epitope, respectively. Here, we integrate structural modeling with CAR functional data and animal experiments to define how epitope position and CAR geometry jointly determine GPC1 CAR efficacy. Modeling of CAR ectodomains incorporating CD8 or IgG4 hinges and CD8 or CD28 transmembrane (TM) domains indicated that only specific combinations yield an intermembrane spacing comparable to that of the TCR-pMHC immune synapse for a given epitope. These geometric predictions aligned with functional data in which D4-IgG4H-CD28TM CAR T cells rapidly regressed tumor bioluminescent in a T3M4 intraperitoneal PDAC model. Together, these results support a model in which epitope spatial location on GPC1 and CAR hinge/TM architecture are key design parameters for GPC1-targeting CAR T cells. To facilitate clinical development of D4 CAR T cells, humanization of the VHH D4 is highly desirable, although humanization of nanobodies such as VHHs is not well established. In addition to the CDR grafting to the nearest germline framework as we described previously for humanization of rabbit and mouse antibodies, we used AI-predicted structural models from multiple platforms to compare both the spatial geometry and sequence similarity of predicted D4 framework and CDR regions with available antibody and nanobody structures. Humanized constructs with the highest humanness scores, geometry, and similarity scores were back-mutated to preserve critical residues in the nanobody framework sequences and were then evaluated for GPC1 binding affinity and cell-surface binding. Using this AI-assisted, structure-guided strategy, we aim to generate clinically suitable humanized VHH nanobody scaffolds for GPC1 CAR T therapy and other nanobody-based clinical applications.
利益披露 Disclosure
H. Tsao, None.