Cell ranger software pipeline

The Cell Ranger software pipeline (version 2.1.0) developed by 10X Genomics was used to demultiplex cell barcodes, map reads to the transcriptome (GRCh38) using STAR aligner and down-sample reads as required to generate normalized aggregate data across samples. The number of reads per cell barcode was calculated using the BamTagHistogram function in the Drop-seq Alignment Cookbook^{49 (link)}. Subsequently, the number of cells per sample was determined by calculating the cumulative fraction of reads corresponding to each individual cell barcode in a library. Cell barcodes were sorted in decreasing order and the inflection point was identified using the R package Dropbead^{50 (link)} (version 0.3.1) to distinguish between empty droplets with only ambient RNA and true droplets containing a cell. The raw matrix of gene counts versus cells from Cell Ranger output was filtered by cell barcodes identified from Dropbead. We processed the resultant unique molecular identifier (UMI) count matrix using the R package Seurat^{51 (link)} (version 2.3.4).

Considering the abundance of tubular epithelial cells and the small amount of glomerular intrinsic cells in a normal kidney cortex, we enriched glomeruli by sequential filtration through 80 and 140 mesh sieve, prepared single-cell suspensions by digesting glomeruli with collagenase I and subsequently performed scRNA-seq and V(D)J-seq using the Single-Cell Immune Profiling Solution. The concentration of the single-cell suspension was counted and adjusted to 1000 cells/μL for a capture of 7000 cells. All remaining procedures, including library construction, were performed according to the manufacturer’s standard protocol described in Shi’ work [19 (link)]. We used the Cell Ranger software pipeline (version 3.0.0, 10xGenomics, USA) to demultiplex cellular barcodes and map reads to the genome. Loupe Browser (version 4.0.0, 10xGenomics, USA) was used for clustering. The barcodes of podocytes were obtained by the clustering of the 10× Genomics transcriptome, and the fastX-Toolkit (Version 0.0.13, Cold Spring Harbor Laboratory, USA) was used to split the data of immune repertoire and obtained Ig sequences expressed in podocytes. Then MiXCR (Version 3.0.7, MiLaboratory, Russia) was used for mapping analyses of Ig genes with accurate alignment of gene segments.

The Cell Ranger software pipeline (v5.0.1, 10x Genomics) was used for demultiplexing and read count annotation to the human reference genome (refdata-cellranger-GRCh38-3.0.0, 2020-A (July 7, 2020)). The resulting filtered feature-barcode matrix, which contains all gene expression counts per cellular barcode, was further processed in R using the Seurat (v 3.2.0) package(Stuart et al., 2019 (link)). To remove low quality cells and doublets, all cells in which less than 200 or more than 2,500 genes were detected and/or cells where mitochondrial genes accounted for 10 or more percent of all detected genes were filtered out. The remaining cells were log-normalized and globally scaled to factor 10,000. We calculated features that exhibited high variation between cells with the function FindVariableFeatures and method “vst”. Prior to dimension reduction, we used ScaleData as linear transformation. With a principal components analysis (PCA) and elbow plot over the variable features we estimated the dimensionality of the dataset. The majority of robust signals was captured in the first 10 principal components. To cluster the cells we used function FindNeighbors to construct a KNN graph based on Euclidean distance using the dimensions previously defined. Using FindClusters cells were grouped together to graph based clusters. For visualization we used UMAP as dimension reduction plot.

The Cell Ranger software pipeline (version 2.1.1) provided by 10x Genomics was used to demultiplex cellular barcodes and map reads to a genome (GRch38) and transcriptome (STAR aligner), producing a matrix of gene counts versus cells. We processed the unique molecular identifier (UMI) count matrix using the R package Seurat (version 2.3.4). As a quality-control (QC) step, we filtered out genes annotated as ribosomal genes or those found in less than three cells and removed cells with fewer than 100 nonzero count genes or with total UMI counts fewer than 1500. To remove likely multiplet captures, which is a major concern in microdroplet-based experiments, we calculated and excluded cells with a transcript count greater than 3 standard deviations away from the mean. We further discarded low-quality cells, in which more than 10% of the counts belonged to mitochondrial genes. After applying these QC criteria, 85,265 single cells and 18,474 genes were included in downstream analyses.

We used the Cell Ranger software pipeline (version 2.2.0, 10xGenomics) to process raw sequencing data and Seurat (version 2.3.4) [21 (link)] R package for downstream analysis as previously described [20 (link)]. Briefly, principle component analysis (PCA) was performed for dimensional reduction. Clusters were identified using the Seurat “FindClusters” algorithm. Graph-based clustering results on 20 principle components were visualized in 2-dimension using t-SNE. Cell clusters were annotated to known biological cell types using canonical marker genes. A cluster-specific biomarker was found by “FindAllMarkers” function identified when it was expressed in a minimum of 25% of cells and at a minimum log fold change threshold of 0.25.