Background
Ambient RNAs contamination in single-nuclei RNA sequencing (snRNA-seq) is a challenging problem, but the consequences of ambient RNAs contamination of damaged and/or diseased tissues are poorly understood. Cognitive impairments and white/gray matter injuries are characteristic of deeper cerebral hypoperfusion mouse models induced by bilateral carotid artery stenosis (BCAS), but the molecular mechanisms still need to be further explored. More importantly, the BCAS mice can also offer an excellent model to examine the signatures of ambient RNAs contamination in damaged tissues when performing snRNA-seq.
Conclusions
Taken together, our current study unravels the features of ambient RNAs in snRNA-seq datasets under diseased conditions, and the in silico approaches can effectively eliminate the incorrected cell annotation and following misleading analysis. In the future, snRNA-seq data analysis should be carefully revisited, and ambient RNAs removal needs to be taken into consideration, especially for those diseased tissues. To our best knowledge, our study also offers the first cortex-specific snRNA-seq data of deeper cerebral hypoperfusion, which provides with novel therapeutic targets.
Methods
After the sham and BCAS mice were established, cortex-specific single-nuclei libraries were constructed. Single-nuclei transcriptomes were described informatically by the R package Seurat, and ambient RNA markers of were identified in each library. Then, after removing ambient RNAs in each sample using the in silico approaches, the combination of CellBender and subcluster cleaning, single-nuclei transcriptomes were reconstructed. Next, the comparison of ambient RNA contamination was performed using irGSEA analysis before and after the in silico approaches. Finally, further bioinformatic analyses were performed.
Results
The ambient RNAs are more predominant in the BCAS group than the sham group. The contamination mainly originated from damaged neuronal nuclei, but could be reduced largely using the in silico approaches. The integrative analysis of cortex-specific snRNA-seq data and the published bulk transcriptome revealed that microglia and other immune cells were the primary effectors. In the sequential microglia/immune subgroups analysis, the subgroup of Apoe+ MG/Mac (microglia/macrophages) was identified. Interestingly, this subgroup mainly participated in the pathways of lipid metabolism, associated with the phagocytosis of cell debris. Conclusions: Taken together, our current study unravels the features of ambient RNAs in snRNA-seq datasets under diseased conditions, and the in silico approaches can effectively eliminate the incorrected cell annotation and following misleading analysis. In the future, snRNA-seq data analysis should be carefully revisited, and ambient RNAs removal needs to be taken into consideration, especially for those diseased tissues. To our best knowledge, our study also offers the first cortex-specific snRNA-seq data of deeper cerebral hypoperfusion, which provides with novel therapeutic targets.