Delineating mouse β-cell identity during lifetime and in diabetes with a single cell atlas

Abstract

Multiple single-cell RNA sequencing (scRNA-seq) datasets have been generated to study pancreatic islet cells during development, homeostasis, and diabetes progression. However, despite the time and resources invested into the past scRNA-seq studies, there is still no consensus on islet cell states and associated pathways in health and dysfunction as well as the value of frequently used preclinical mouse diabetes models. Since these challenges can be only resolved with a joint analysis of multiple datasets, we present a scRNA-seq cross-condition mouse islet atlas (MIA). We integrated over 300,000 cells from nine datasets with 56 samples, varying in age, sex, and diabetes models, including autoimmune type 1 diabetes (T1D) model (NOD), gluco-/lipotoxicity T2D model (db/db), and chemical streptozotocin (STZ) β-cell ablation model. MIA is a curated resource that enables interactive exploration of gene expression and transfer of cell types and states. We use MIA to obtain new insights into islet cells in health and disease that cannot be reached from individual datasets. Based on the MIA β-cell landscape we report cross-publication differences between previously suggested marker genes of individual phenotypes. We further show that in the STZ model β-cells transcriptionally correlate to human T2D and mouse db/db model β-cells, but are less similar to human T1D and mouse NOD model β-cells. We define new cell states involved in disease progression across diabetes models. We also observe different pathways shared between immature, aged, and diabetes model β-cell states. In conclusion, our work presents the first comprehensive analysis of β-cell responses to different stressors, providing a roadmap for the understanding of β-cell plasticity, compensation, and demise.

Integrating single-cell RNA-seq datasets with substantial batch effects

“…Data for each of the three use cases was prepared separately as described below. For the mouse-human data, we used pancreatic islet datasets of mouse 1 (without embryonic and low-quality cells) and human 58 , for the organoid-tissue scenario we used a retinal dataset 7 , and for the cell-nuclei scenario we used adipose dataset 59 (using the SAT fat type). We obtained published count data and cell annotation for all datasets (see Data availability section) and removed unannotated cells.…”

“…The joint analysis of multiple single-cell RNA sequencing (scRNA-seq) datasets has recently provided new insights that could not have been obtained from individual datasets. For example, pooling of datasets generated in different studies enabled cross-condition comparisons 1,2 , population-level analysis 3,4 , and revealed evolutionary relationships between individual cell types 5 . The selection of pre-clinical models, such as organoids and animals, and the characterization of their limitations likewise rely on comparison with human tissues [6][7][8][9][10][11][12][13][14] .…”

Pancreatic β-cell failure, clinical implications, and therapeutic strategies in type 2 diabetes

“…Recent accumulating evidence in rodents has shown that aged animals and senescent islets tend to express less β-cell transcription factors ( Pdx-1 , Nkx6.1 , FoxO1 , and MafA ) and functionally important genes ( Ins1 , Ins2 , Slc2a2 , and Slc30a8 ), along with more expression of β-cell dedifferentiation markers ( Sox9 and Aldh1a3 ). [38] Similarly, in the human pancreas, PDX1 expression is markedly downregulated with aging. [39] Remarkably, a single-cell RNA-seq analysis revealed that the number of bihormonal cells that simultaneously expressed both insulin and glucagon mRNA increased with age, and the presence of such transcriptionally noisy cells suggests a fate drift between α- and β-cells in the aging pancreas, resulting in a loss of β-cell identity and function.…”

Interpreting single-cell and spatial omics data using deep networks training dynamics

“…STZ is a naturally occurring diabetogenic compound that targets and destroys insulin-secreting pancreatic β cells, leading to reduced insulin levels and elevated blood glucose levels [52][53][54]. In the mSTZinduced diabetic models, STZ is applied in multiple low doses, partially degrading β-cell activity [55], where the level of degradation may differ between individual cells [56]. To quantify the damage to individual cells extracted from mSTZ-induced diabetic models, we evaluate their trainability for a disease label against the backdrop of control cells annotated as healthy.…”