Single-cell transcriptomics reveals a mechanosensitive injury signaling pathway in early diabetic nephropathy

Animals

All animal experiments were conducted according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals as well as the German law for the welfare of animals. Mice were housed in a specific pathogen-free facility with free access to chow and water and a 12-h/12-h day/night cycle. Breeding and genotyping were performed according to standard procedures.

BTBR ob/ob podocyte-reporter mice

Animal experiments were approved by the veterinary administration of the City of Hamburg under the license Ü003-2018. BTBR ob/ + (BTBR.Cg-Lepob/wt WiscJ) heterozygous animals (Jax No. 004824) and corresponding wild-type mice BTBR (Jax No 002282) were purchased from JAX (Bar Harbor, ME, USA). Podocyte-reporter mice (Gt(ROSA)26Sortm4(ACTB−tdTomato,−EGFP)Luo;Tg(NPHS2-cre)295Lbh) were also purchased from JAX (Jax No. 007576) and crossed for at least 7 generations with BTBR wild-type animals before BTBR.Cg-Lepob/wt WiscJ;Gt(ROSA)26Sortm4(ACTB−tdTomato,−EGFP)Luo;Tg(NPHS2-cre)295Lbh animals were generated. The resultant offspring was intercrossed to yield BTBR.Cg-Lepob/ob WiscJ;Gt(ROSA)26Sortm4(ACTB−tdTomato,−EGFP)Luo;Tg(NPHS2-cre)295Lbh and BTBR.Cg-Lepwt/wt WiscJ;Gt(ROSA)26Sortm4(ACTB−tdTomato,−EGFP)Luo;Tg(NPHS2-cre)295Lbh animals.

Mouse kidney normothermic machine perfusion

All procedures involving mouse kidney perfusion described in this manuscript were conducted according to German and Hamburg law and approved by the veterinary administration of the City of Hamburg under the license N002/2020. As previously described [12], male C57BL/6 kidneys were collected under anesthesia (i.p. injection of 100 μL/10 g body weight of a solution containing 10 mg/ml Ketamine and 1.6 mg/ml Xylazine). Briefly, a laparotomy was performed, the aorta was cannulated, ligatures placed and closed around all main vessels, and perfusion initiated only in the right kidney. The left unperfused kidney was then removed and used as control. After this, the right kidney was removed from the mouse during ongoing perfusion. Perfusion took place ex vivo at 100 mmHg at 37 °C with 100% GBSS solution without the addition of albumin or red blood cells (Sigma G9779) continuously gassed in a dialyzer with 100% carbogen (95% O2, 5% CO2). Perfusion was stopped after 60 min. Isolation of glomeruli was performed as described in Additional file 1: Supplementary methods (FACS-sorted glomerular cells).

Pig kidney normothermic machine perfusion

Pig kidneys were collected from a German slaughterhouse. Kidneys were received directly after slaughter and immediately flushed with 400–500 ml 1 × PBS with 2 ml Heparin 10.000 IE/ml. Kidney were stored at + 4 °C for around 2 h, after which normothermic perfusion was initiated. Perfusion took place at 100 mmHg at 37 °C with increased glomerular pressure and hyperfiltration as described above. Renal cortex tissue was collected either before perfusion (control) or 120 min after start of perfusion (perfused).

Single-cell/nucleus suspension and library preparationSingle-cell suspension

Glomeruli were enriched by depleting the PT segment from the kidneys using Percoll density gradient centrifugation described previously [13]. Briefly, the mouse was euthanized by cervical dislocation under anesthesia (4% isoflurane for 2 min). Both kidneys were harvested and the renal papilla of each kidney was removed. The renal tissue was minced into small pieces and transferred to a C tube (130–093-237, Miltenyi Biotec) filled with 5 ml of collagenase solution (1 mg/ml collagenase from Clostridium histolyticum [C9263, Sigma-Aldrich] and 0.25% bovine serum albumin [BSA] in DMEM/F-12 medium [11320074, Gibco]). The renal tissue was incubated at 37 °C for 5 min and homogenized using a gentleMACS™ Dissociator (Miltenyi Biotec) with the program “m. spleen 01_01”; this process was repeated once. Subsequently, the renal tissue was passed through a 250-μm sieve and separated by Percoll density gradient centrifugation using a solution that contained 45% Percoll® (GE17-0891–02, GE Healthcare) and 55% 2 × PBS-glucose at 17,500 rpm for 15 min at 4 °C. Red blood cells were removed with a Percoll density gradient. PTs that accumulated in the interphase of the gradient were discarded. The rest of the kidney tissue that accumulated in the upper phase of the gradient was taken, washed twice in ice-cold HBSS, and centrifuged at 400 × g for 4 min at 4 °C. A single-cell suspension was prepared using cold-active protease (CAP) as described in previous studies [14]. The pellet was suspended in 2 ml of CAP solution (10 mg/ml Bacillus licheniformis protease [P5380, Sigma-Aldrich], 5 mM CaCl2, 20 U/ml DNase [4716728001, Roche] in PBS) for 20 min on ice with repeated trituration steps for 20 s every 5 min. The digestion was neutralized by 15 ml PBS supplemented with 10% FBS and passed through a 40-μm Corning® cell strainer. Cells were centrifuged at 400 × g for 4 min at 4 °C and washed twice in 20 ml PBS supplemented with 0.5% BSA. Dead cells were removed with a dead cell removal kit (130–090-101, Miltenyi Biotec) according to the manufacturer’s protocol. The living cells were passed through a 30-μm cell strainer (04–004-2326, Sysmex), and the cell concentration was determined using a TC20™ automated cell counter (Bio-Rad). The cells were loaded onto a 10 × Genomics Chromium single-cell instrument. All steps were performed according to the standard protocol of the Chromium single-cell 3’ v2/v3 kits to generate high-quality cDNA libraries.

Single-nucleus suspension

A 1 × 3 mm piece of snap frozen renal cortex tissue was thawed and chopped with a razor blade in a petri dish on ice and homogenized using a Dounce homogenizer (D8938-1 SET, Sigma-Aldrich) in 200-μl ice-cold lysis solution and incubated on ice for 20 min with additional 3.8 ml of ice-cold lysis solution. Lysis solution was prepared with Nuclei PURE lysis buffer (NUC-201, Sigma-Aldrich), 1 mM dithiotreitanol (D9779, Sigma-Aldrich), and 0.1% Triton X-100 (NUC-201, Sigma-Aldrich) according to manufacturer’s protocol and a RNAse inhibitor mix (0.04 U/μl SUPERaseIN RNAse Inhibitor [AM 2696, Thermo Fisher]; 0.04 U/μl RNAsin Plus RNAse Inhibitor [N2615, Promega]) was added. The single nuclei suspension was filtered through a 30-μm filter (04–004-2326, Sysmex) and centrifuged at 500 g for 5 min at 4 °C. The pellet was resuspended and incubated for 2 min in 1 ml red blood cell lysing buffer hybri-max™ (R7757-100 ml, Sigma-Aldrich), filtered through a 5-μm filter (04–004-2323, Sysmex), and washed with 4 ml of ice-cold 0.01% BSA (AM2616, Thermo Fisher) in DPBS (59331C; Sigma) with 0.04 U/μl SUPERaseIN RNAse Inhibitor and 0.04 U/μl RNAsin Plus RNAse Inhibitor at 500 g for 5 min at 4 °C. The pellet was resuspended in 1% BSA in DPBS with 0.04 U/μl SUPERaseIN RNAse Inhibitor and 0.04 U/μl RNAsin Plus RNAse Inhibitor, and nuclei number was counted. The libraries were prepared with the Chromium NEXT GEM Single Cell 3’ Reagent kits v3.1 according to manufacturer’s protocol. The libraries were sequenced on an Illumina Novaseq6000 platform as symmetrically paired end runs (150 bases) with 200 million raw sequencing reads per sample.

Single-cell data analysisPreprocessing and quality control (QC) of scRNA-seq data

10 × Genomics raw sequencing data were processed using CellRanger software (version 3.0.2, 10 × Genomics, Pleasanton, CA), and the 10 × Genomics mouse genome mm10 3.0.0 release was used as the reference genome (function cellranger count). The matrices of cells and the unique molecular identifier (UMI) count were obtained and further processed with the R package Seurat (version 3.1.1) [15]. For QC, we first filtered out genes detected in fewer than 3 cells and data for cells in which fewer than 200 genes had nonzero counts. To remove potential doublets, cells with more than 7000 expressed genes (nFeature) were excluded. We removed low-quality cells with more than 50% mitochondrial genes among all detected genes, as is recommended for kidney tubular cells [16]. After clustering and cell type identification, we performed curated doublet and high-mitochondrial gene cell removal (see below) based on known lineage-restricted markers.

Dimensionality reduction and clustering

The Seurat R package (version 4.0.2) was used to perform unsupervised clustering analysis on scRNA-seq data. In brief, gene counts for cells that passed QC were normalized by library size and log-transformed (function NormalizeData, normalization.method = “LogNormalize”, scale.factor = 10,000). Then, highly variable genes were detected (function FindVariableFeatures, selection.method = “vst”, nfeatures = 2000). To reduce batch effects, we applied the integration method implemented in Seurat version 3 (functions FindIntegrationAnchors and IntegrateData, dims = 1:30). The integrated matrix was then scaled with the ScaleData function (default parameters). PCA was performed on the scaled data (function RunPCA, npcs = 30) to reduce dimensionality.

The number of principal components used for each clustering round was dataset dependent and was determined on the basis of the elbow of a PCA scree plot. The selected principal components were then used to compute the KNN graph based on the Euclidean distance (function FindNeighbors). Cell clusters were subsequently generated using the function FindClusters. The resolution of the FindClusters function for each dataset was also determined by the exploration of the top marker genes of each cluster. Uniform Manifold Approximation and Projection (UMAP) was used to visualize the clustering results. The top DEGs in each cluster were found using the FindAllMarkers function (min.pct = 0.25, logfc.threshold = 0.25) with Wilcoxon rank-sum tests. The most highly expressed genes were then used to determine the cell type of each cluster.

Curated doublet and high-mitochondrial gene cell removal

After the cell type was determined for clusters, we performed additional curated doublet and high-mitochondrial-cell removal. Based on the literature and the exploration of our datasets, we created a lineage-restricted marker gene list for tubular (Cubn, Epcam) and nontubular (Pecam1, Pdgfrb, Nphs2, Ptprc) cell types. We removed the cells that expressed markers of the opposite lineage. For nontubular cells, we discarded cells with more than 20% mitochondrial genes among all detected genes. There were 86,508 cells before curated doublet removal, 71,831 cells after curated doublet removal, and 770,944 cells after high-mitochondrial gene cell removal.

Differential gene expression analysis

The Seurat FindMarker function was used to perform differential gene expression analysis for each cell type between the control and DN groups. The test method used in the FindMarker function was MAST (v1.12.0) [17].

Single-nucleus data analysisPreprocessing and QC

10 × Genomics raw sequencing data were processed using CellRanger software (version 5.0.1, 10 × Genomics, Pleasanton, CA). Pig genome was built by function “cellranger mkref” using the Sscrofa v103 FASTA file and GTF file (https://www.ensembl.org) and then mapping was done and the count matrices were generated by function “cellranger count” with parameter “include-introns.” For QC, firstly, we applied soupX [18] with default parameters to remove ambient RNA contamination. Then, the nuclei were filtered out if the number of genes detected was less than 500 or greater than 6000 or the percentage of mitochondrial genes detected exceeded 7%. To further remove potential doublets, Scrublet [19] was applied to the data and cells were excluded if they were identified as doublets (default parameters).

Dimensionality reduction and clustering

An updated version of Seurat R package (version 4.0.2) was used to perform unsupervised clustering analysis on snRNA-seq data since the experiments were performed later than the mouse scRNA-seq analysis. Count data normalization, scaling, highly variable gene selection, and sample integration were the same as the mouse single-cell data process described above. PCA was performed on the scaled data (function RunPCA, npcs = 30) and the first 15 PCs were used for clustering. The KNN graph was calculated (function FindNeighbors) and then the clustering result is obtained (function FindClusters, resolution = 0.5). To visualize the clustering result, UMAP coordinates were calculated (function RunUMAP, dims = 1:15). Wilcoxon method was used to perform the differential gene analysis. For each cluster, the marker gene list was determined by log2FoldChange > 0.25 and adj. p value < 0.05 (FindAllMarkers). The most highly expressed genes were then used to determine the cell type of each cluster.

Bulk RNA-seqLibrary preparation and RNA-seq

A small amount of RNA (2 ng) was used as input material, and libraries were prepared with a SMART-Seq Stranded Kit according to the user manual (Takara Bio USA, Mountain View, CA, USA). In brief, samples were fragmented at 85 °C for 6 min prior to first-strand synthesis. Illumina adaptors and indexes were added to single-stranded cDNA via 5 cycles of PCR. After library purification with AMPure beads and depletion of ribosomal cDNA with scZapR, final RNA-seq library amplification (13 cycles) was conducted, and the final RNA-seq library was purified with AMPure beads. The library samples were quantified using Quant-iT PicoGreen dsDNA Reagent (Invitrogen; Thermo Fisher Scientific, Waltham, MA, USA) on a ClarioStar microplate reader according to the manufacturer’s instructions (BMG LABTECH, Ortenberg, Germany). The quality, including fragment size, of the cDNA was assessed on an Agilent Technologies Bioanalyzer 2100 using an Agilent DNA 1000 kit according to the manufacturer’s instructions (Agilent Technologies, Palo Alto, CA, USA).

Pooled samples were quantified with a Qubit 1X dsDNA HS Assay Kit on a Qubit fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). Single-read sequencing was performed on a NovaSeq 6000 device using an S2 Reagent kit (100 cycles) according to the manufacturer’s instructions (Illumina Inc., CA, USA).

Bulk RNA-seq data analysis

The quality of the bulk RNA-seq reads was assessed using FastQC (v0.11.5), and the reads were aligned to the mouse reference genome (mm10) with Bowtie2 (v2.3.3.1) [20] using RSEM (v1.3.0) [21] with the default parameters. The function rsem-calculate-expression was used to align the reads and quantify the gene and isoform abundance. The output of rsem-calculate-expression separately gives the read count and transcripts per million (TPM) value for each gene and isoform. Differential expression analysis was carried out using gene read counts with the DESeq2 package (v1.22.2) [22] to produce log2FC values and corresponding p values and adjusted p values. Principal component analysis (PCA) was performed using regularized log transformation of the count data, and the results were visualized using gplots (v3.0.1.1).

Integrated analysisCorrelation of bulk RNA-seq and scRNA-seq data

For podocytes, ECs, and mesangial cells, the single-cell data were correlated with bulk RNA-seq data. The cell type markers (top DEGs of each cell type identified by the Seurat FindAllMarkers function (min.pct = 0.25, logfc.threshold = 0.25) were used for correlation analysis. The normalized gene expression count matrices of cell type markers were extracted from the single-cell data and bulk RNA-seq data separately, and then Pearson correlation coefficients were calculated between every pair of single cells and replicates of bulk RNA-seq data.

CellChat

We applied Cellchat (v 1.4.0) [23] to infer cell–cell communications across all kidney cell types and the glomerular cell types. We used the mouse database curated in CellChat including the “Secreted Signaling,” “Cell–cell Contact,” and “ECM-Receptor.” The comparison analysis between diseased samples and control samples was performed according to the Cellchat tutorial (https://htmlpreview.github.io/?https://github.com/sqjin/CellChat/blob/master/tutorial/Comparison_analysis_of_multiple_datasets.html). The communication probabilities were compared to identify the unregulated and downregulated signaling ligand-receptor pairs between the two conditions.

SCENIC

The gene regulatory network was inferred using pySCENIC (v 0.11.2, a lightning-fast python implementation of the SCENIC pipeline) [24, 25]. Firstly, the GRNboost2 algorithm was used to infer gene regulatory network and generate co-expression modules (pyscenic grn). Next, the regulon prediction step was performed (pyscenic ctx, using the default parameters and mm10__refseq-r80__10kb_up_and_down_tss.mc9nr and mm10__refseq-r80__500bp_up_and_100bp_down_tss.mc9nr motif collections). Finally, the AUCell matrix was generated with a threshold of 0.01 (pyscenic aucell). Cell type-specific regulators were identified based on Z-score of the AUCell values for the cells of a given type.

Upstream analysis

We performed Qiagen’s ingenuity Pathway Analysis (IPA) tool to identify transcriptional regulators (significance: p value < 0.05). Positive and negative z-score values indicated the activation or inhibition of transcriptional regulators, respectively.

Enrichment analysis

GO biological processes, KEGG, WikiPathways, and Reactome Gene Sets were performed with Metascape [26].

Human dataERCB-Kröner-Fresenius Biopsy Bank

Human renal indication biopsy specimens were collected in an international multicenter study and deposited in the ERCB-Kröner-Fresenius Biopsy Bank (ERCB-KFB [27, 28]; for participating centers, see Shved et al. [29]). Glomerular samples from patient with different renal diseases were analyzed for mRNA expression levels (GSE32591, GSE35489, GSE37463, GSE47185, GSE99340). The analysis included gene expression profiles from patients with DN (n = 14), hypertensive nephropathy (HT; n = 15), minimal change disease (MCD; n = 14), IgA nephropathy (IgA; n = 27), focal segmental glomerulosclerosis (FSGS; n = 23), membranous nephropathy (MGN; n = 21), lupus nephritis (SLE; n = 32), ANCA-associated glomerulonephritis (RPGN; n = 23), and controls (LDs; n = 41).

Early diabetic nephropathy in Pima Indians

Protocol kidney biopsy specimens were collected from Pima Indians with type 2 diabetes. The study subjects participated in an intervention trial (Renoprotection in Early Diabetic Nephropathy in Pima Indians, clinicaltrials.gov, NCT00340678) [30, 31]. Biopsies were obtained from patients after informed consent was obtained and with the approval of the local ethics committees. The analysis included gene expression profiles from DN patients (n = 68) and LD controls (n = 18). To account for ethical considerations, to ensure privacy protection, and to avoid identifying individual study participants in this vulnerable population of Pima people, the Institutional Review Board of the National Institute of Diabetes and Digestive and Kidney Diseases has stipulated that individual-level gene expression and genotype data from this study cannot be made publicly available.

Microarray analysis

Tissue processing and microdissection protocols were performed similarly on both ERCB and Pima biopsy material. Following renal biopsy, the tissue was transferred to an RNase inhibitor and microdissected into glomeruli and tubulointerstitial tissue. Total RNA was isolated, reverse-transcribed, and amplified. Fragmentation, hybridization, staining, and imaging were performed according to the Affymetrix Expression Analysis Technical Manual (Affymetrix, Santa Clara, CA, USA). CEL file normalization was performed with the robust multichip average method using RMAExpress (Version 1.0.5) and the human Entrez-Gene custom CDF annotation from Brain Array version 18 (PIMA) and 25 (ERCB) (http://brainarray.mbni.med.umich.edu/Brainarray/default.asp). The log-transformed ERCB dataset was corrected for batch effect using ComBat from the GenePattern pipeline (http://www.broadinstitute.org/cancer/software/genepattern/). To identify DEGs, the Significance Analysis of Microarrays (SAM) method [32] was applied using the SAM function in Multiple Experiment Viewer (TiGR MeV, Version 4.9). A q-value below 5% was considered to indicate statistical significance.

Immunofluorescent staining

Paraffin sections from human biopsies or from experimental mice (3 μm thick) were deparaffinized and rehydrated in water. Following antigen retrieval with DAKO pH6 for 30 min at 98 °C, unspecific binding was blocked for 30 min at RT in blocking buffer (0.05% TritonX-100, 5% normal horse serum (Vector) in PBS). Primary antibodies for human biopsies (MRTFA [Sigma HPA030782]; MRTFB [Invitrogen PA5-113519]; CD34 [Leica Biosystems NCL-L-END]; NPHS1 [Progene GP-N2]) and for murine sections (MRTFA [Sigma HPA030782]; MRTFB [Invitrogen PA5-113519]; Endomucin-AF546 [Santa Cruz sc-65495]; NPHS1 [Progene GP-N2]) were diluted in blocking buffer and incubated overnight at 4 °C. Stainings were visualized following incubation with appropriate fluorochrome coupled secondary donkey antibodies (Jackson Immunoresearch Laboratories) and Hoechst (Molecular Probes) and mounted with fluoromount (Thermo Fisher). Analyses were performed on a LSM800 airyscan 1 using the ZENBlue software (all ZEISS).

A piece of pig cortex (1 cm deep) was snap frozen in OCT. Seven-micrometer cryo-sections were fixed immediately in 4% PFA for 10 min at RT and washed out with PBS. The sections were incubated for 30 min with 5% BSA in PBS + 0.1% Triton for blocking and permeabilization and then incubated overnight at 4 °C with primary antibody MEIS1 (Invitrogen MA5-27,191). After washing out the primary antibody, sections were incubated for 1 h at RT with secondary antibody (Life Technologies) and DAPI (4′,6-diamidino-2-phenylindole) (Invitrogen). Slides were mounted and analyzed using the Leica TCS SP5 microscope.

留言 (0)

沒有登入
gif