A single-cell atlas of CD19 chimeric antigen receptor T cells.
A single-cell atlas of CD19 chimeric antigen receptor T cells.
复制标题
CD19 嵌合抗原受体 T 细胞的单细胞图谱。
DOI：
10.1016/j.ccell.2023.08.015
复制
发表时间：
2023
期刊：
影响因子：
50.3
通讯作者：
Green,MichaelR
中科院分区：
文献类型：
--
作者：
Li,Xubin;Henderson,Jared;Gordon,MaxJ;Sheikh,Irtiza;Nastoupil,LorettaJ;Westin,Jason;Flowers,Christopher;Ahmed,Sairah;Wang,Linghua;Neelapu,SattvaS;Strati,Paolo;Deng,Qing;Green,MichaelR
针对CD19的自体嵌合抗原受体T细胞（CART 19）显著改善了复发性或难治性大B细胞淋巴瘤（rrLBCL）的结果。然而，大多数患者没有持久的反应;因此，迫切需要改善结局的策略。CART 19治疗失败的机制是复杂的，可能是相互关联的。它们可以包括在单采期间收获的T细胞的初始功能状态和组成、CART 19输注产品的后续功能特征和肿瘤固有特性，例如肿瘤对T细胞浸润和/或细胞毒性的抗性。使用基于组学的策略（如单细胞RNA测序（scRNA-seq））仔细检查这些T细胞表型，揭示了与CAR T细胞耐药性相关的特征（在Yang et al. 2中综述）。CAR T细胞输注产品的scRNA-seq分析突出了与记忆、疲惫和调节性T细胞状态的重要关联。3，4然而，这些研究中的样本量有限，无论是患者数量还是细胞数量。在这里，我们呈现了来自用标准护理的阿昔卡他烯西洛糖胞苷（Axi-cel; 417，167个高质量单细胞）治疗的59名rrLBCL患者的输注产品的scRNA-seq数据的资源（表S1A），我们已经公开了这些scRNA-seq数据，以促进将改善患者结果的正在进行的和未来的发现努力。作为该数据集实用性的原理证明，我们通过PET/CT确定了3个月随访时缓解者（完全缓解[CR]）和非缓解者（疾病稳定、部分缓解和疾病进展）产品之间的显著差异特征，这是长期结局的重要标志。来自所有59名患者的细胞被包括用于无监督聚类以定义主要细胞类型。然而，只有35例患者被纳入结局比较，因为5例患者不可评价，19例患者在输注后接受了可能改变其结局的临床试验中的额外治疗。5所有细胞的聚类，这些细胞鉴定为先前报道的ICANS相关细胞（IAC）和非ICANS相关骨髓样细胞3以及CD 4+、CD 8+和双阴性聚类（图S1 A;表S1 B）。为了对T细胞进行分类，我们采用K-最近邻方法4来计算CD3（CD3D、CD3E、CD3G和CD247的平均值）、CD8（CD8A和CD8B的平均值）和CD4的平滑表达。我们使用平滑的基因表达严格定义了CD 4+和CD 8+细胞（图S1 B），并重新聚类这些子集以定义转录上不同的亚群。CD4 T细胞簇（图S1C;表S1C）包括T调节性（F0XP3阳性）和细胞毒性（细胞毒性特征高）亚群，其先前分别涉及CAR T细胞衰竭4和持久应答6。然而，使用scCODA 7，这些和其他CD4 T细胞簇的相对比例在响应者（R，n = 15）和非响应者（NR，n = 20）之间没有显著差异（图S1D）。类似地，CD8 T细胞的亚聚类鉴定了先前在CAR T细胞应答3中涉及的记忆、效应子和功能障碍簇（图S1E;表S1D），但这些簇的频率在应答者和非应答者之间没有显著差异（图S1F）。除了先前描述的与IACs簇的相关性之外，这种方法也没有发现CD4或...
Autologous chimeric antigen receptor T cells directed toward CD19 (CART19) have significantly improved the outcomes of relapsed or refractory large B cell lymphoma (rrLBCL). 1 However, the majority of patients do not have durable responses; thus, strategies to improve outcomes are critically needed. The mechanisms underlying CART19 treatment failure are complex and likely interrelated. They may include the initial functional states and composition of T cells harvested during apheresis, the subsequent functional characteristics of the CART19 infusion product, and tumor-intrinsic properties such as the tumor’s resistance to T cell infiltration and/or cytotoxicity. Careful examination of these T cell phenotypes using omics-based strategies such as single-cell RNA sequencing (scRNA-seq) has revealed characteristics associated with CAR T cell resistance (reviewed in Yang et al. 2). scRNA-seq analysis of CAR T cell infusion products has highlighted important associations with memory, exhausted, and regulatory T cell states. 3, 4 However, the sample sizes in these studies have been limited, both in terms of the number of patients and the number of cells. Here we present a resource of scRNA-seq data from the infusion products of 59 rrLBCL patients (Table S1 A) treated with standard-of-care axicabtagene ciloleucel (Axi-cel; 417,167 high-quality single cells) that we have made publicly available to facilitate ongoing and future discovery efforts that will improve patient outcomes. As a proof-of-principle for the utility of this dataset, we identified features that were significantly different between products from responders (complete response [CR]) and non-responders (stable disease, partial response, and progressive disease) at 3-month followup by PET/CT, an important landmark for long-term outcomes.A detailed description of the methods is provided in the supplemental information. Cells from all 59 patients were included for unsupervised clustering to define major cell types. However, only 35 patients were included for outcome comparisons because 5 patients were not evaluable and 19 patients received additional therapy on a clinical trial following their infusion that likely modified their outcomes. 5 Clustering of all cells identified previously reported ICANS-associated cells (IACs) and non-ICANS-associated myeloid-like cells, 3 as well as CD4+, CD8+, and double-negative clusters (Figure S1 A; Table S1 B). To classify T cells, we employed a K-nearest neighbors approach 4 to calculate smoothed expressions of CD3 (average of CD3D, CD3E, CD3G, and CD247), CD8 (average of CD8A and CD8B), and CD4. We stringently defined CD4+ and CD8+ cells using smoothed gene expressions (Figure S1 B) and re-clustered these subsets to define transcriptionally distinct subclusters. The CD4 T cell clusters (Figure S1 C; Table S1 C) included T-regulatory (FOXP3-positive) and cytotoxic (cytotoxicity signature high) subsets that have previously been implicated in CAR T cell failure 4 and durable response, 6 respectively. However, the relative proportions of these and other CD4 T cell clusters were not significantly different between responders (R, n= 15) and non-responders (NR, n= 20) using scCODA 7 (Figure S1 D). Similarly, sub-clustering of CD8 T cells identified memory, effector, and dysfunctional clusters that have previously been implicated in CAR T cell response 3 (Figure S1 E; Table S1 D), but the frequencies of these clusters were not significantly different between responders and non-responders (Figure S1 F). Beyond the previously described association with the IACs cluster, 3 this approach also did not identify significant associations between CD4 or …