snRNAseq- Neuroectodermal IL-12R signaling in neuroinflammation/EAE
Description
We performed single-nucleus RNA sequencing (Chromium Next GEM Single Cell 3’ Reagent Kits v3.1 protocol, 10X Genomics) of FANS-isolated Hoechst+ nuclei from the cerebellum, brainstem and cervical spinal cord of NestinCre/+Il12rb2fl/fl mice and their Il12rb2fl/fl littermates at the onset of clinical EAE symptoms (10 dpi) . We performed two independent sequencing experiments and successfully sequenced 8 samples. The first experiment included two samples: Il12rb2fl, NestinCre_Il12rb2fl. The second experiment included six samples: three Il12rb2fl mice and three NestinCre_Il12rb2fl mice. After quality control and doublet exclusion, snRNA-seq yielded a total of 133,151 single-nucleus transcriptomic profiles, among which 21,552 distinct genes were detected. Batches were integrated using SCVI. Unsupervised clustering of these data identified clusters that were assigned to diverse neuronal, glial and other cell types on the basis of known lineage marker genes. This allowed us to assess how the transcriptional profile of the neuroectoderm differed in the presence or absence of IL-12 receptor signaling by interrogating differentially expressed genes (DEGs). This revealed pronounced alterations particularly in excitatory neurons, granule cells, mature oligodendrocytes (MOLs) and myelin forming oligodendrocytes (MOLs1) suggesting a specific/concentrated action of IL-12 on these populations in the inflamed CNS. Our data suggest IL-12 to be critically involved in neuroprotection and shaping the trophic factor milieu within the inflamed CNS. Conversely, dysregulation or loss of trophic factor release by neurons in the absence of IL-12 may further propagate the degeneration of neurons, oligodendrocytes and possibly other cell types. Sequencing raw data and processed gene expression data will be deposited into the GEO repository upon publication. Code for analyses will be available upon contacting the corresponding authors: becher@immunology.uzh.ch and mundt@immunology.uzh.ch
Files
Steps to reproduce
RNA sequencing of single nuclei has been performed in two independent batches (batch 1: n=1 NestinCre/+Il12rb2fl/fl and n=1 Il12rb2fl/fl; batch 2: n=3 NestinCre/+Il12rb2fl/f and n=3 Il12rb2fl/fl). Immediately after extraction, 17,500 sorted nuclei per sample were loaded onto a Chromium Single Cell 3′ Chip (10X Genomics) and processed for the single-nucleus cDNA library preparation (Chromium Next GEM Single Cell 3’ Reagent Kits v3.1 protocol). 50.000 reads per nucleus were sequenced using the Illumina Novaseq 6000 #1 platform according to the manufacturer’s instructions without modifications (R1= 28, i7= 10, i5= 10, R2=90). Preparation of cDNA libraries and sequencing were performed at the Functional Genomics Center Zurich. CellRanger software (v6.0.2) was implemented for library demultiplexing, barcode processing, fastq file generation, gene alignment to the mouse genome (GENCODE reference build GRCm39), and unique molecular identifier (UMI) counts. We implemented the “include-introns” option for counting intronic reads, as the snRNA-seq assay captures unspliced pre-mRNA as well as mature mRNA. For each sample, a CellRanger report was obtained with all the available information regarding sequencing and mapping parameters. All samples were merged into a matrix using CellRanger (cellranger -aggr function). Starting from the filtered gene-cell count matrix produced by CellRanger’s built-in cell calling algorithms, we proceeded with the SCANPY workflow in Python. Doublet exclusion was employed individually for each sample using the Scrublet package with variable thresholds. Quality control (QC) processing has been performed individually for each batch. Low-quality nuclei have been filtered out based on the number of unique genes (<500 and <200 in batch 1 and batch 2 respectively), total UMIs (<6000) and fraction of mitochondrial counts (<5%). Genes present in less than three nuclei have been removed. Counts have been log1p transformed normalized to 10,000 counts and scaled for visualization purposes. Individual batches have been integrated into a combined SCVI model consisting of two layers and 30 latent dimensions using the 3000 most variable features. The attached h5ad file contains the raw counts including relevant metadata in the anndata "obs" field to reproduce the analysis.