Single-cell analysis of dissociation-induced compositional and transcriptional bias in human breast tissue samples

The use of single cell transcriptomics provides previously inaccessible insights into cellular heterogeneity and lineage dynamics of the mammary gland allowing for a better understanding of normal mammary gland function as well as breast cancer initiation and progression. Especially for human mammary gland research, limited tissue accessibility and restriction to ex vivo techniques reinforce the importance of reliable cross-study comparison of single-cell transcriptomic data. However, it is unclear to what extent differences in breast tissue dissociation influence composition and transcriptomic profiles of isolated cells. Here, we used single-cell RNA sequencing to compare human mammary cell populations isolated from a single mammoplasty patient by varying enzymatic dissociation protocols differing in duration (3 or 16 hours) and agitation speed (10 rpm or 100 rpm). Protocol A (3 hours, 100 rpm), protocol B (16 hours, 100 rpm) and protocol C (16 hours, 10 rpm) were used to extract cell fragments from tissue which were either frozen down directly or further dissociated into single cells prior to cryopreservation. Samples were then prepared together for 10x scRNA-sequencing, where the fragments were defrosted and freshly dissociated and were loaded together with the defrosted single cells. From each of the protocols we sequenced similar numbers of cells isolated from fragments (Protocol A: 3,586 cells, Protocol B: 2,809 cells and Protocol C: 4,796 cells), finding an average of 7,427 unique molecular identifiers and 2,445 genes detected per cell. Overall, we detected a greater abundance and heterogeneity of stromal cell types, such as fibroblasts and endothelial cells at a lower agitation speed. Moreover, an extended duration of tissue dissociation governed an overall cellular oxidative stress response together with a downregulation of breast cancer associated genes and a cell-type specific downregulation of lineage markers. Thus, our systematic analysis of dissociation-induced compositional and transcriptional bias in human breast tissue samples yields useful information to avoid misinterpretation of cellular heterogeneity and lineage composition.

单细胞转录组学 (single cell transcriptomics) 的应用，为解析乳腺的细胞异质性与谱系动态特征提供了此前难以获取的研究视角，助力我们更深入地理解乳腺正常生理功能，以及乳腺癌的发生与发展进程。针对人类乳腺研究而言，组织获取难度有限且仅能采用体外 (ex vivo) 实验技术的现状，进一步凸显了对单细胞转录组学数据开展可靠跨研究比对的重要性。然而，目前尚不明确乳腺组织解离流程的差异，会在多大程度上影响分离所得细胞的组成与转录组特征。
本研究中，我们借助单细胞RNA测序 (single-cell RNA sequencing) 技术，对比了同一名乳腺整形手术患者的乳腺细胞群——通过调整酶解解离方案的参数（解离时长为3小时或16小时，振荡转速为10转每分钟 (rpm) 或100转每分钟 (rpm)）分离得到的细胞群。本研究采用三种解离方案：方案A（3小时，100 rpm）、方案B（16小时，100 rpm）与方案C（16小时，10 rpm），从组织中提取细胞片段；这些片段可直接冻存，或在冻存前进一步解离为单细胞悬液。随后将所有样本统一制备以进行10x scRNA-sequencing测序：先将细胞片段解冻并进行新鲜解离，再将其与解冻后的单细胞悬液一同上样。针对每个解离方案，我们对从细胞片段中分离得到的细胞进行了相近规模的测序：方案A为3586个细胞，方案B为2809个细胞，方案C为4796个细胞；每个细胞平均可检测到7427个独特分子标识符 (unique molecular identifiers) 与2445个表达基因。
总体而言，当振荡转速较低时，我们可检测到丰度更高、异质性更强的间质细胞类群，例如成纤维细胞与内皮细胞。此外，延长组织解离时长会引发整体细胞的氧化应激反应，同时伴随乳腺癌相关基因的表达下调，以及谱系标记物的细胞类型特异性下调。综上，本研究针对人类乳腺组织样本中解离流程诱导的细胞组成与转录偏倚开展了系统性分析，所得结果可为避免对细胞异质性与谱系组成的误判提供有价值的参考依据。