Single-cell RNAseq of human pancreatic islets from control and type 2 diabetes donors

Perturbed secretion of insulin and other pancreatic hormones is the main cause of type 2 diabetes (T2D). The pancreatic islets harbor five cell types that are potentially altered differently by T2D. Whole-islet transcriptomics and single-cell RNA-sequencing (scRNAseq) studies have revealed differentially expressed genes but unfortunately without reaching consensus. We propose that, rather than analyzing expression of individual genes or structural changes in networks of genes, building and analyzing networks of differentially synchronized genes on single-cell level gives unprecedented insights to disease. To this end, we developed a differential gene synchronization network analysis (dGSNA) algorithm and used it to analyze a high-quality pancreatic islet scRNAseq dataset from T2D and non-T2D individuals. dGSNA on beta cells revealed T2D-induced changes in twelve networks of genes, representing both canonical and less studied biological processes in the context of T2D. Analysis of these networks suggested that in T2D, beta cell energy metabolism, cell growth, differentiation and proteolysis programs are perturbed, whereas exocytosis, insulin translation and the unfolded protein response programs instead are enhanced. To validate our method, we showed that ten out of eleven selected node genes regulated insulin mRNA levels and/or secretion in INS-1 832/13 cells. Further, knockout mouse models for two of these genes (Tmem176a/b and Cebpg) exhibited reduced beta cell mass and perturbed insulin secretion. The dGSNA algorithm thus predicts central disease processes in T2D, which was replicated in an independent data set, revealing targetable cellular functions. We conclude that analysis of networks of differentially synchronized genes provides insight into the pathophysiology of T2D. This approach is likely generally applicable to other diseases where scRNAseq data can be obtained. The single-cell transcriptome data (6 control and 5 type 2 diabetes human donors) was generated at the Eukaryotic Single-cell Genomics facility at Science for Life Laboratory in Stockholm, Sweden.

胰岛素及其他胰腺激素的分泌异常是2型糖尿病（type 2 diabetes, T2D）的核心致病原因。胰腺胰岛包含五种细胞类型，它们在T2D的影响下可能发生不同程度的改变。全胰岛转录组学与单细胞RNA测序（single-cell RNA-sequencing, scRNAseq）相关研究已鉴定出差异表达基因，但遗憾的是尚未达成统一共识。我们认为，相较于单独分析单个基因的表达水平或基因网络的结构变化，在单细胞层面构建并解析差异同步化基因网络，可为疾病研究提供前所未有的视角。为此，我们开发了差异基因同步化网络分析（differential gene synchronization network analysis, dGSNA）算法，并利用该算法分析了一组高质量的胰腺胰岛scRNAseq数据集，该数据集来自T2D患者与非T2D个体。对β细胞开展dGSNA分析后，我们发现T2D诱导了12个基因网络的改变，这些网络涵盖了T2D研究中经典的及较少被关注的生物学过程。对这些基因网络的分析显示，在T2D状态下，β细胞的能量代谢、细胞增殖、分化及蛋白水解程序均发生紊乱，而胞吐作用、胰岛素翻译及未折叠蛋白反应（unfolded protein response）程序则被激活增强。为验证该算法的有效性，我们选取了11个核心节点基因，其中10个可在INS-1 832/13细胞中调控胰岛素mRNA水平及/或胰岛素分泌。此外，针对其中两个基因（Tmem176a/b与Cebpg）的敲除小鼠模型表现出β细胞质量降低及胰岛素分泌紊乱的表型。综上，dGSNA算法可预测T2D的核心疾病进程，该结果在独立数据集上得到验证，揭示了可靶向调控的细胞功能。我们的研究结论表明，差异同步化基因网络分析可为T2D的病理生理学研究提供新见解，该方法或可推广至所有可获取scRNAseq数据的其他疾病研究中。本研究中的单细胞转录组数据（包含6名健康对照个体与5名T2D患者的样本）由瑞典斯德哥尔摩生命科学实验室（Science for Life Laboratory）下属的真核单细胞基因组学平台完成测序。