Single-Cell Transcriptome Profiling of the Kidney Glomerulus Identifies Key Cell Types and Reactions to Injury

Significance Statement

Single-cell transcriptomics techniques have revolutionized the ability to characterize cells from heterogeneous organs like the kidney. Although glomerular disorders are an important cause of CKD, a thorough characterization of the cells in the glomerulus has remained challenging due to the technical difficulties of isolating undamaged cells, especially from glomeruli of diseased animals. This study provides a comprehensive single-cell atlas, based on approximately 75,000 cells, from glomeruli of healthy mice and mice injured in four ways, including all cell types present. The data set will be a valuable resource for generating precise tools to interrogate specific glomerular cell types and in identifying genes involved in the pathogenesis of glomerular diseases.

Visual Abstract

Abstract

Background

The glomerulus is a specialized capillary bed that is involved in urine production and BP control. Glomerular injury is a major cause of CKD, which is epidemic and without therapeutic options. Single-cell transcriptomics has radically improved our ability to characterize complex organs, such as the kidney. Cells of the glomerulus, however, have been largely underrepresented in previous single-cell kidney studies due to their paucity and intractability.

Methods

Single-cell RNA sequencing comprehensively characterized the types of cells in the glomerulus from healthy mice and from four different disease models (nephrotoxic serum nephritis, diabetes, doxorubicin toxicity, and CD2AP deficiency).

Results

All cell types in the glomerulus were identified using unsupervised clustering analysis. Novel marker genes and gene signatures of mesangial cells, vascular smooth muscle cells of the afferent and efferent arterioles, parietal epithelial cells, and three types of endothelial cells were identified. Analysis of the disease models revealed cell type–specific and injury type–specific responses in the glomerulus, including acute activation of the Hippo pathway in podocytes after nephrotoxic immune injury. Conditional deletion of YAP or TAZ resulted in more severe and prolonged proteinuria in response to injury, as well as worse glomerulosclerosis.

Conclusions

Generation of comprehensive high-resolution, single-cell transcriptomic profiles of the glomerulus from healthy and injured mice provides resources to identify novel disease-related genes and pathways.

The glomerulus, the site of filtration in the kidney, is a capillary bed composed of endothelial cells, podocytes, and mesangial cells, as well as less abundant cell types, such as the parietal epithelial cells (PECs) and vascular smooth muscle cells (SMCs) (Supplemental Figure 1). Loss of glomerular function is the most common cause of CKD, a major health care problem affecting approximately 15% of the population.¹ Glomerular injury is caused by factors such as diabetes and hypertension, as well as by immune injury. Glomeruli are particularly susceptible to injury because podocytes are largely unable to regenerate and, therefore, tissue damage is considered irreversible. It is not known what reparative mechanisms are used to restore function after injury or how acute glomerular injury progresses to chronic fibrosis. Despite their critical roles in kidney function and disease, cells of the glomerulus have largely been underrepresented in previous kidney single-cell RNA sequencing (scRNA-seq) studies due to their paucity and difficulty of isolation.

Here, we performed scRNA-seq using purified glomeruli to characterize all of the cell types and we analyzed their reaction to four common types of kidney injury: immune, metabolic, toxic, and genetic injury. Our work, which includes sequencing of approximately 75,000 glomerular cells, provides a comprehensive transcriptional signature of all cell types in the glomerulus, including those that have not been well characterized previously, such as mesangial cells, PECs, juxtaglomerular (JG) cells, and arteriolar SMCs. Results from four disease models provide new insights into the glomerular response to acute injury and its progression to CKD.

Methods

Reagents

The reagents used were as follows: Dynabeads M-450 Tosylactivated (Thermo Fisher Scientific), Liberase TM (Sigma-Aldrich), DNase I (Sigma-Aldrich), trypsin (Thermo Fisher Scientific), Dispase II (Roche Applied Science), Collagenase D (Roche Applied Science), paraformaldehlyde (Electron Microscopy Sciences), and OCT (Sakura Finetek).

Antibodies

Anti-FHL2, anti-SERPINE2, anti-RGS2, anti-ADAMTS5, anti–calponin 1, and anti–α smooth muscle actin antibodies were purchased from Abcam. Anti-PKCα antibody was purchased from Thermo Fisher Scientific. Anti-PDGFRβ (APB5) antibody was purchased from eBioscience. Anti-YAP/TAZ (D24E4) antibody was purchased from Cell Signaling. Anti–β-actin (AC-15) and anti-vinculin (hVIN-1) antibodies were purchased from Sigma-Aldrich. Phycoerythrin-conjugated anti-CCL2 (2H5) and phycoerythrin/Cy7-conjugated anti–TNF-α (MP6_XT22) antibodies were purchased from Biolegend. Alexa Fluor–conjugated secondary antibodies were purchased from Thermo Fisher Scientific.

Mice

C57BL/6J and BTBR ob/ob (BTBR.Cg-Lep^ob/WiscJ) mice were purchased from Jackson Laboratory. CD2AP-deficient mice have been described previously.² Generation of Wwtr1 (TAZ) and Yap1 (YAP) conditional knockout (CKO) strains and CCL2-YPet reporter strain is described below. All animals were bred and housed at Genentech under specific pathogen-free conditions with free access to chow and water and a 12-hour day/night cycle. Only male mice were used. For the nephrotoxic serum nephritis model, age-matched C57BL/6J mice were injected intravenously with 100 μl (for scRNA-seq analysis) or 2.3 ml/kg body wt (approximately 60 μl, for YAP/TAZ experiments) of sheep anti-rat glomeruli serum (Probetex). For the doxorubicin nephropathy model, age- and weight-matched C57BL/6J mice were given a single intraperitoneal injection of 20 mg/kg doxorubicin hydrochloride (Pfizer). All animal procedures were conducted under protocols approved by the Institutional Animal Care and Use Committee at Genentech, and were performed in accordance with the Guide for the Care and Use of Laboratory Animals.

Generation of Wwtr1 and Yap1 CKO Mice

Homologous recombination and mouse embryonic stem (ES) cell technology^3–5 was used to generate genetically modified mouse strains with a Wwtr1 or Yap1 CKO. For Wwtr1, a gene targeting vector was constructed with a 1415-bp arm of 5′ homology corresponding to GRCm38/mm10 chromosome 3 (chr3): 57,577,558–57,576,144 and a 2066-bp arm of 3′ homology corresponding to chr3: 57,574,633–57,572,568. The 1510-bp region flanked by loxP sites (exon 1+2) corresponds to chr3: 57,576,143–57,574,634. For Yap1, a gene-targeting vector was constructed with a 1990-bp arm of 5′ homology corresponding to GRCm38/mm10 chr9: 8,003,842–8,001,853 and a 2045-bp arm of 3′ homology corresponding to chr9: 8,001,304–7,999,260. The 548-bp region flanked by loxP sites (exon 2) corresponding to chr9: 8,001,852–8,001,305.

The final vector was confirmed by DNA sequencing, linearized, and used to target C2 (C57BL/6N) ES cells using standard methods (G418-positive and ganciclovir-negative selection).⁶ C57BL/6N C2 ES cells⁷ were electroporated with 20 μg of linearized targeting vector DNA and cultured under drug selection essentially as described. Positive clones were identified using long-range PCR followed by sequence confirmation.

Correctly targeted ES cells were subjected to karyotyping. Euploid gene-targeted ES cell clones were treated with Adeno-FLP to remove PGK neomycin, ES cell clones were tested to identify clones with no copies of the PGK neomycin cassette, and the correct sequence of the targeted allele was verified. The presence of the Y chromosome was verified before microinjection into albino C57BL/6N embryos. Germline transmission was obtained after crossing resulting chimeras with C57BL/6N females. Genomic DNA from born pups was screened by long-range PCR to verify the desired gene-targeted structure before mouse colony expansion. Genotyping primers used to identify germline transmission were the following: Wwtr1.CKO primers were (1) TGG​TCA​CAA​GCG​TTA​AGC, (2) TGG​TTC​AAG​CCT​GTT​AAA​TCA, and (3) CCTACTCACCTGGCTGT; expected amplicon sizes were 255 bp for wild type, 289 bp for floxed, 440 bp for knockout. The Yap1.CKO primers were (1) TTG​AGT​TAT​GTA​GGA​TGA​GCA​TTA, (2) GTA​TGT​CAC​GGC​AAC​CAA, and (3) TGA​CCA​ACC​CTA​AAG​AGA​GA; expected amplicon sizes were 246 bp for wild type, 280 bp for floxed, 320 bp for knockout.

Generation of CCL2-Ypet Reporter Mice

A bacterial artificial chromosome (BAC) clone harboring the mouse CCL2 gene (RP23-328G11) was obtained from the BACPAC Resources Center at Children’s Hospital Oakland Research Institute. The BAC was modified by inserting YPet-bGHpA (bovine growth hormone polyadenylation signal) into exon 1 of the CCL2 locus using the Counter Selection BAC Modification Kit (Gene Bridges, Heidelberg, Germany). The 12.8-kb segment of the modified BAC including the YPet-bGHpA modified CCL2 gene and the 5-kb upstream and 5-kb downstream regions was subcloned using homologous recombination into a linearized pBluescript II SK plasmid (Agilent) that contained homology arms with flanking XhoI sites. The plasmid construct was amplified in Escherichia coli and purified using the EndoFree Plasmid Maxi Kit (Qiagen). The purified plasmid was digested with XhoI and the fragment containing the CCL2 reporter construct was gel purified for injection. C57BL/6 mice were injected by the Mouse Genetics Core at Washington University (St. Louis, MO). Genotyping primers used to identify founders carrying the transgene were the following: (1) GCA​TCG​ACT​TCA​AGG​AGG​AC and (2) GTC​AGG​AAC​TCC​AGC​AGC​AC, with expected amplicon size of 295bp.

Preparation of Single Cells from Glomeruli

Single-cell suspensions of glomerular cells were prepared from biologic replicates for each sample using a previously described method with modifications.⁸ First, glomeruli were isolated using magnetic beads. Kidneys were removed from mice en bloc with the abdominal aorta attached and transferred to petri dishes filled with ice-cold HBSS without calcium and magnesium ions. After cutting open the abdominal aorta along its length, each kidney was perfused directly through the renal arteries with 1 ml ice-cold HBSS containing magnetic beads (200 μl Dynabeads M-450 rinsed and resuspended in 10 ml HBSS). The kidneys were then minced into small pieces (<1 mm³) with a razor blade and incubated in 3 ml prewarmed digestion buffer (1.5 U/ml Liberase TM, 100 U/ml DNase I in HBSS) at 37°C for 20 minutes with constant agitation (800 RPM) in a ThermoMixer (Eppendorf). All subsequent steps were performed on ice or at 4°C unless specified otherwise. The digested kidneys were passed through a 100-μm cell strainer twice. The samples were washed twice with ice-cold HBSS and resuspended in 1 ml HBSS. The samples were then placed in a magnetic separator to collect the glomeruli and washed four to five times with HBSS until samples were >98% pure by visual inspection under a microscope.

For preparation of single-cell suspensions, the purified glomeruli were resuspended in 1.25 ml of digestion buffer (0.5% trypsin, 2.0 U/ml Dispase II, 2 U/ml Collagenase D, 10 U/ml DNase I in prewarmed PBS without calcium and magnesium ions) and incubated at 37°C for 40–60 minutes with constant agitation (800 RPM) in a ThermoMixer. (We tested the use of cold active protease from Bacillus licheniformis, but it was not effective at dissociating glomeruli.) During the incubation, samples were triturated by pipetting every 5 minutes for the first 30 minutes. After 30 minutes, the glomeruli were mechanically sheared by passing through a 27 1/2 gauge needle twice. The samples were then incubated for an additional 10–30 minutes until >98% of the cells had been dissociated into single cells by visual inspection under a microscope. Shearing of the glomeruli with a syringe needle at earlier time points resulted in selective loss of podocytes. Longer incubation increased the number of cells isolated, but decreased cell viability. The digestion was stopped by adding 10 ml ice-cold PBS with 10% FBS. The samples were then placed in a magnetic separator to remove magnetic beads. The supernatant was collected, passed through a 40-μm cell strainer to remove residual cell aggregates, centrifuged at 300 × g for 5 minutes at 4°C, and resuspended in 1 ml of PBS with 0.1% BSA. Cell density and viability of single-cell suspensions were determined using the Vi-CELL XR Cell Counter (Beckman Coulter). The method typically generated 0.5×10⁶–1×10⁶ cells per mouse with >90% viability.

Sample Processing, Library Preparation, and Sequencing

Sample processing for scRNA-seq was done using Chromium Single Cell 3′ Library and Gel Bead Kit version 2 according to the manufacturer’s instructions (10× Genomics). Total cell density was used to calculate the volume of single-cell suspension needed in the reverse transcription master mix, aiming to achieve approximately 6000 cells per sample. cDNAs and libraries were prepared according to the manufacturer’s instructions (10× Genomics). cDNA amplification and preparation of indexed libraries were performed using 12 and 14 cycles of PCR, respectively. Libraries were profiled using the Bioanalyzer High Sensitivity DNA Kit (Agilent) and quantified using the Kapa Library Quantification Kit (Kapa Biosystems). Each library was sequenced in one lane of the HiSeq 4000 (Illumina) to achieve approximately 300 million reads following the manufacturer’s sequencing specifications (10× Genomics).

Sequencing Data Quality Control and Preprocessing

scRNA-seq data were processed with a custom pipeline. Briefly, reads were demultiplexed based on perfect matches to expected cell bar codes. Transcript reads were aligned to the mouse reference genome (mm10) using GSNAP.⁹ Per-gene transcript counts were determined based on the number of unique molecular identifiers for mapped reads overlapping exons in the sense orientation, allowing for one mismatch when collapsing unique molecular identifier sequences. To be considered for downstream analysis, cells were required to exceed a minimum number of detected transcripts, where a sample-specific cutoff was set to 0.1 times the total transcript count for cells at rank 30 (the 99th percentile for 3000 cells). Cells with >5% mitochondrial gene counts and genes detected in <0.1% of cells were excluded from further analysis. Genes in the protocadherin γ family (Pcdhga1-12, Pcdhgb1-8, Pcdhgc1-5) were removed from further analysis due to multimapping. Doublets were removed using a combination of high-count thresholding and gating for simultaneous expression of two or more cell type–specific marker genes. Per-gene transcript counts were normalized by dividing by the total transcript count for a given cell and multiplying by a scale factor of 10,000. Normalized counts were transformed using a log₂(x + 1) transformation.

Dimensionality Reduction, Clustering Analysis, and T-Distributed Neighbor Embedding/Uniform Manifold Approximation and Projection Visualization

Partek Flow software (version 8.0.19.0707) was used for analysis of single-cell data. Principal component (PC) analysis (PCA) was performed on the highly variable genes, and PCs with eigenvalues ≥1 were chosen (maximum 100 PCs) to be used for downstream clustering analysis and t-distributed neighbor embedding or Uniform Manifold Approximation and Projection (UMAP) visualization. To prevent clustering artifacts due to tissue dissociation–induced stress, the 140 dissociation-induced genes reported by van den Brink et al.¹⁰ were excluded from PCA and downstream clustering analyses. Clusters were identified using the Louvain clustering algorithm. For identification of subclusters within cell types, cells belonging to each cell type were separated from the total population and reanalyzed as a separate sample.

Identification of Marker Genes and Differentially Expressed Genes

Differential gene expression analysis was performed using the Partek GSA (gene-specific analysis) function, which identifies and uses the most suitable statistical model for each transcript (details can be found at https://documentation.partek.com/display/FLOWDOC/Gene-specific+Analysis). Marker genes for each cluster were identified by comparing cells in a specific cluster with all remaining cells (mean fold change, >1.5; false-discovery rate [FDR]–adjusted P<0.01). Clusters with less than five marker genes were considered overclustering artifacts. Expression of canonical marker genes (Supplemental Table 1) was used to identify and assign cell types (podocytes, mesangial cells, endothelial cells, vascular SMCs, immune cells, PECs, tubular epithelial cells [TECs]) to clusters. PECs were identified by expression of Cldn1 and Pax8 in addition to podocyte marker genes. TECs were excluded from analysis in Figure 1 and subsequent figures. To identify genes specific to PECs and mesangial cells, we first identified genes highly enriched in each cell type compared with all other cell types (mean fold change, >3; FDR-adjusted P<0.01). The list was then further refined by removing the genes with expression levels greater than or equal to LSMean 2 in any other cell type. Hierarchic clustering was performed using Ward linkage and Euclidean distance and shown as z-scores of normalized expression levels.

Figure 1.

Unbiased clustering of scRNA-seq data reveals the major cell types of the glomerulus. (A) T-distributed neighbor embedding (t-SNE) representation of 5287 glomerular cells from a healthy C57BL/6J mouse. Labels indicate clusters identified by unsupervised clustering analysis. (B) Violin plots of marker gene expression in each cluster shown in (A). (C) Violin plots showing expression of glomerular capillary endothelial cell (Ehd3) or arteriolar endothelial cell (Fbln2, Mgp, Trpv4, Bmx) marker genes in clusters 2 and 4 (D). PCA plot of normal glomerular cells. Cluster labels are identical to (A). Proximity of clusters 3 and 5 indicate high degree of similarity between the two clusters. (E) Violin plots showing expression levels of genes specific to mesangial cells or SMCs/JG cells. The numbers in parentheses indicate cluster number. (F) Immunofluorescence staining of kidney sections shows mesangial-specific expression of the identified marker genes. PDGFRβ, which stains mesangial cells in the glomerulus (dotted circle) and stromal cells outside the glomerulus, was used as reference. Scale bars, 20 μm. (G) Expression levels of the indicated genes are shown in a t-SNE plot of SMCs/JG cells. The two subclusters corresponding to AA SMCs and EA SMCs are outlined. (H) Immunofluorescence staining of kidney sections shows specific staining of calponin 1 in the vascular SMCs of the AA. Scale bar, 10 μm. Normalized expression levels are shown in the violin plots. DAPI, 4′,6-diamidino-2-phenylindole; α-SMA, α-smooth muscle actin.

Mapping of Disease-Associated Genes

For analysis of cell type–specific expression of FSGS disease– and CKD-associated genes, we generated pseudo-bulk data for each cell type by calculating the mean expression levels for each gene. Heatmaps were generated to depict z-scores of normalized expression levels.

Pathway Analysis

Pathway analyses were performed using the biologic interpretation function in Partek Flow. Lists of differentially expressed (DE) genes (gene-specific analysis mean fold change, >1.5; FDR-adjusted P<0.01) were used as input to identify enriched Gene Ontology terms or KEGG pathways.

Immunofluorescence Microscopy

For immunofluorescence, kidneys were perfusion fixed by transcardial perfusion with fixation buffer (4% paraformaldehlyde in PBS), immersion fixed in fixation buffer for an additional hour, and immersed in 30% sucrose in PBS at 4°C overnight. The kidneys were frozen in OCT and cut into 6- to 15-μm sections with a cryostat (Leica Biosystems). The sections were blocked and permeabilized with Image-iT FX (Thermo Fisher Scientific) or 10% normal goat serum or normal donkey serum (Jackson ImmunoResearch) in PBS with 0.3% Triton X-100 (Sigma-Aldrich). The slides were incubated with primary antibodies in PBS with 0.3% Triton X-100 and 1% BSA (Sigma-Aldrich) overnight at 4°C. After three washes with PBS, the slides were incubated with fluorescent dye–conjugated secondary antibodies at room temperature for 1 hour. After three washes with PBS, the slides were mounted with ProLong Gold Antifade Mountant with 4′,6-diamidino-2-phenylindole (Thermo Fisher Scientific) and cured overnight. Images were obtained using a Nikon A1R confocal microscope (Nikon).

Measurement of Albuminuria

Spot urine samples were collected at the indicated time points. Urinary albumin (Bethyl Laboratories) and creatinine (BioAssay Systems) levels were quantified by ELISA according to the manufacturers’ instructions.

Histopathology Analysis

Kidney tissues were fixed for 24 hours at ambient temperature in 10% neutral buffered formalin (VWR), and then processed and embedded using a Tissue-Tek VIP processor (Sakura). Sections were mounted on Superfrost Plus glass slides (Richard-Allan) and stained with Periodic acid–Schiff (PAS) or Masson trichrome using an automated stainer (DAKO) according to the manufacturer’s instructions. Severity of glomerular injury was visually scored in a blinded fashion on PAS-stained slides using a subjective, semiquantitative, five-point scale as follows: 0, within normal limits; 1, mild, segmental mesangial expansion with or without increased cellularity; 2, moderate, segmental mesangial expansion frequently associated with increased cellularity; 3, global membranoproliferative GN; and 4, glomerulosclerosis. The average score for each mouse was calculated from 20 glomeruli. Severity of fibrosis was visually scored in a blinded fashion on trichrome-stained slides using a subjective, semiquantitative, five-point scale as follows: 0, no appreciable increase in periglomerular fibrosis; 1, one to three glomeruli or tubules with increased surrounding fibrosis and/or sclerosis; 2, four to ten glomeruli or tubules with increased surrounding fibrosis and/or sclerosis; 3, 11–20 glomeruli or tubules with increased surrounding fibrosis and/or sclerosis; and 4, >20 glomeruli or tubules with increased surrounding fibrosis and/or sclerosis. PAS- and trichrome-stained slides were imaged with a Nanozoomer 2.0-HT automated slide scanning platform (Hamamatsu) at 200× final magnification.

Immunoblotting

Kidney or glomeruli isolated from mice were lysed in ice-cold TNET lysis buffer (50 mM Tris-hydrochloride [pH 7.4], 150 mM sodium chloride, 1 mM EDTA, 1% Triton X-100, cOmplete Protease Inhibitor Cocktail; Roche Applied Science) for 15 minutes on ice. The protein content of the lysates was quantified with the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Equivalent amounts of each sample were separated by SDS-PAGE, transferred to nitrocellulose membranes, and immunoblotted for YAP/TAZ, β-actin, or vinculin. The blots were visualized by infrared imaging on a LI-COR Odyssey system (LI-COR).

Real-Time Quantitative PCR Analysis

RNA was isolated using TRIzol Reagent (Invitrogen), and cDNA was generated using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Real-time quantitative PCR (qPCR) was performed on the ABI 7500 (Applied Biosystems) using the TaqMan Universal PCR Master Mix (Applied Biosystems) and the following TaqMan Gene Expression Assays (Thermo Fisher Scientific): Angptl2 (Mm00507897_m1), Cebpb (Mm00843434_s1), Col15a1(Mm00456551_m1), Csrp1 (Mm00456002_g1), Ctgf (Mm01192933_g1), Ddn (Mm02020418_s1), Ednrb (Mm00432989_m1), H2-q7 (Mm00843895_s1), Lgals1 (Mm00839408_g1), Lifr (Mm00442942_m1), Mafb (Mm00627481_s1), Mest (Mm00485003_m1), Nphs1 (Mm01176615_g1), Rcan2 (Mm00472671_m1), Rpl13a (Mm05910660_g1), Sema3g (Mm01219781_m1), S100a6 (Mm00771682_g1), Thbd (Mm00437014_s1), Tpm1 (Mm00445895_g1), and Vegfa (Mm00437306_m1).

Validation of CCL2-YPet Reporter Mice

For in vitro validation of the CCL2-YPet reporter mice, YPet expression was measured in isolated peritoneal macrophages after stimulation with LPS and polyinosinic:polycytidylic acid (poly[I:C]). First, mice were injected with 1 ml of 3% Brewer thioglyocollate (Sigma-Aldrich) medium into the peritoneal cavity. Peritoneal macrophages were isolated from the mice 4 days after injection by flushing the peritoneal cavity with ice-cold PBS with 10% FBS and cultured overnight in DMEM with 10% FBS before stimulation. For FACS analysis of YPet expression, cells were treated with 100 ng/ml LPS (Sigma-Aldrich) and 50 μg/ml poly(I:C) (InvivoGen) for 24 hours. For FACS analysis of intracellular CCL2 and TNF-α, cells were pretreated with 2 μM monensin (BioLegend) and 5 μg/ml brefeldin A (BioLegend) for 30 minutes, then treated with 200 ng/ml LPS and 50 μg/ml poly(I:C) for 3 hours. Cells were harvested, stained for intracellular CCL2 and TNF-α using the Cytofix/Cytoperm kit (BD biosciences), and analyzed by flow cytometry using the LSRII or LSRFortessa (BD biosciences).

For qPCR analysis, peritoneal macrophages treated with 100 ng/ml LPS and 50 μg/ml poly(I:C) for 24 hours were sorted into YPet-positive and YPet-negative cells using the FACSAria II (BD biosciences). RNA was isolated from the sorted cells and used to generate cDNA. qPCR was performed using the SYBR Green PCR Master Mix (Applied Biosystems) and the following primers: YPet forward, ACG​ACG​GCA​ACT​ACA​AGA​CC; YPet reverse, GTC​CTC​CTT​GAA​GTC​GAT​GC; Ccl2 forward, CAA​GAA​GGA​ATG​GGT​CCA​GA; Ccl2 reverse, GCT​GAA​GAC​CTT​AGG​GCA​GA; Ppia forward, AGC​ATA​CAG​GTC​CTG​GCA​TC; and Ppia reverse, CCA​TCC​AGC​CAT​TCA​GTC​TT.

To measure CCL2 secretion from peritoneal macrophages, cells were stimulated with varying doses of LPS (0.1–100 ng/ml) for 24 hours. The amount of CCL2 in the supernatant was measured using a mouse CCL2 ELISA kit according to the manufacturer’s instructions (R&D systems).

For in vivo validation of the CCL2-YPet reporter mice, the animals were either maintained on a normal chow diet (Picolab rodent diet 20; LabDiet) or a high-fat diet (60 kcal% fat, D12492; Research Diets) for 10 weeks. For two-photon imaging of the epididymal fat pads, animals were injected intravenously with QTRACKER Qdot 655 (Thermo Fisher Scientific) 10 minutes before euthanasia to label the vasculature. The fat pads were removed from euthanized mice, rinsed in PBS, and mounted for imaging. Images were collected using a customized Leica SP8 Two-Photon Microscope (Leica Microsystems) equipped with a 25× and 0.95 numerical aperture water immersion objective and a Mai Tai HP DeepSee Laser (Spectra-Physics). Induction of YPet expression in fat cells and infiltrating macrophages in the mice fed a high-fat diet confirmed the CCL2 reporter mouse is a valid tool to detect CCL2 expression in vivo.¹¹

Results

scRNA-Seq Analysis of Healthy Glomeruli

Because glomerular cells constitute only a minute fraction (<1%) of the kidney, we first isolated glomeruli from C57BL/6J mice using a magnetic beads–based method⁸ that was optimized to consistently yield approximately 0.5×10⁶ cells per mouse with >90% viability. We began by sequencing >6000 glomerular cells from a healthy adult male mouse. We analyzed 5488 cells that remained after quality control and removal of doublets.

Unsupervised clustering analysis after removing dissociation-induced genes¹⁰ identified seven distinct clusters of cells (Figure 1A, Supplemental Figure 2, A and B, Supplemental Table 2). The cell types of each cluster were determined using expression of known marker genes (Figure 1B, Supplemental Table 1). Podocytes, endothelial cells, and mesangial cells were present in similar proportions and together comprised >90% of the cells (Supplemental Table 3), demonstrating the purity of the isolated glomeruli and the efficiency of our tissue-disassociation method.¹² The remaining cells corresponded to vascular SMCs/JG cells (cluster 5), immune cells (cluster 6; Supplemental Figure 2C), and PECs (cluster 7) (Figure 1B). Small numbers of contaminating TECs were excluded from further analysis. The top 50 genes enriched in each cell type are listed in Supplemental Table 4.

Podocytes and mesangial cells were each grouped into a single cluster, indicating they are relatively homogenous in homeostatic conditions. Despite their close similarity to podocytes, we were able to identify genes highly specific for PECs, which have been implicated as potential glomerular stem cells (Supplemental Figure 2D).¹³ Analysis of endothelial cell subclusters showed high levels of Ehd3 in cluster 2, whereas Fbln2, Mgp, Trpv4, and Bmx were preferentially expressed in cluster 4 (Figure 1C). Based on previous studies, we concluded that cluster 2 represents capillary endothelial cells and that cluster 4 represents arteriolar endothelial cells from the afferent arteriole (AA) and efferent arteriole (EA). We also identified a novel subpopulation of endothelial cells that highly express the Notch ligand Dlk1 and the endothelin receptor (Ednrb) (Supplemental Figure 2E). The identity of these cells is unclear, but DLK1 is known to inhibit angiogenesis.¹⁷

Identification of Mesangial Cell and SMC Markers

Mesangial cells function as the central structural support of the glomerulus, interacting with both endothelial cells and podocytes.¹⁸ They are important producers of extracellular matrix, regulate capillary blood flow, phagocytose extracellular debris, and contribute to the homeostasis of the glomerulus by secreting growth factors. Marker genes commonly used to identify mesangial cells, such as Pdgfrb and Gata3, are not specific to mesangial cells because these genes are also expressed by vascular SMCs, as well as stromal cells outside of the glomerulus. α-Smooth muscle actin (Acta2) is also often used as a mesangial marker, but it is not highly expressed in uninjured adult mesangial cells and is more specific to SMCs.²¹ Such lack of a precise genetic signature for mesangial cells has been an obstacle to understanding the specific function of mesangial cells in the kidney.

Clusters 3 and 5 both expressed Gata3 and Pdgfrb (Figure 1, A and B) and were closely related transcriptionally, as indicated by their proximity in the PCA plot (Figure 1D). Expression of Acta2, Tagln, Myh11, and JG cell markers Ren1 (renin), Akr1b7, and Rgs5 revealed that cluster 5 includes the vascular SMCs and JG cells²² (Figure 1E, Supplemental Table 5). Cluster 3, therefore, represents bona fide mesangial cells.

To define a precise signature for mesangial cells, we first searched for genes that were highly expressed in cluster 3 compared with all other cells (Supplemental Table 6). For a subset of these genes, we confirmed mesangial expression at the protein level using immunofluorescence staining (Figure 1F, Supplemental Figure 2F). Some of the identified genes—such as Serpine2, Fhl2, Des (desmin), and Pdgfrb—were also expressed at low levels in SMCs/JG cells (Supplemental Figure 2G), so we identified genes that were exclusively expressed in each of the two cell types. We found specific expression of Plvap, Prkca, Art3, and Nt5e in mesangial cells; whereas SMCs/JG cells specifically expressed Acta2, Myh11, Rergl, Map3k7cl, Ren1, and Akr1b7 (Figure 1E). Our results suggest SMCs have been misidentified as mesangial cells in previously published scRNA-seq studies.^23,24

Analysis of SMCs revealed two subpopulations. One cluster expressed Adra1a (α-adrenergic receptor 1A) and was enriched for renin, suggesting these cells represent SMCs from the AAs (Figure 1G).²⁵ The other cluster, which expressed Adra1b (α-adrenergic receptor 1B), had higher expression of angiotensin receptors (Agtr1a, Agtr1b), suggesting these cells are EA SMCs. We identified Cnn1 (calponin 1) and Cygb (cytoglobin) as novel markers for AA and EA SMCs, respectively. Immunofluorescence staining confirmed specific expression of calponin 1 in the AA SMCs (Figure 1H). The identification of markers to distinguish between AA and EA SMCs is significant because differential targeting of these cells to control glomerular blood flow is the basis of kidney protective antihypertensive drugs.²⁶

Mapping of Disease Susceptibility Genes

Studies over the last couple of decades have led to the identification of susceptibility genes for FSGS, a glomerular disease that is considered to be a podocyte-specific disease.²⁷ Because our data set contains all of the glomerular cell types at high resolution, we were interested in mapping expression patterns of FSGS susceptibility genes on our data. Surprisingly, we found that only 20 out of 50 susceptibility genes were exclusively expressed in podocytes within the glomerulus (Supplemental Figure 3A). Of the remaining genes, some were more highly expressed in other cell types, such as mesangial cells (Supplemental Figure 3B), and some were not detected in any glomerular cell type. These findings suggest these genes may contribute to FSGS susceptibility through novel podocyte-independent pathways.

Genome-wide association studies have identified a large number of potential CKD susceptibility genes which are mainly thought to be expressed in TECs.²⁸ We confirmed that many of the genes are indeed highly expressed in TECs, but we also noted that some of these genes were expressed in glomerular cells (Supplemental Figure 3C). For example, Dach1 and Vegfa were relatively specific to podocytes, whereas Adamts5 and Pik3r1 had specific expression in mesangial cells (Supplemental Figure 3D). We noted specific expression of some CKD genes in SMCs (Jph2, Prkag2) (Supplemental Figure 3D). The implication that SMCs may play a role in CKD is intriguing.

Glomerular Response to Immune-Mediated Injury

Glomerular deposition of circulating immune complexes or local formation of antigen-antibody complexes is the cause of various forms of GN.²⁹ We induced GN in mice using nephrotoxic serum, a surrogate mouse model for the type of glomerular injury that occurs in lupus nephritis or Goodpasture disease (Supplemental Figure 4A).³⁰ Injection of the nephrotoxic serum caused an acute inflammatory response and a transient proteinuria that peaked 1 day after injection and gradually decreased to near baseline levels after 7 days (Supplemental Figure 4B). After 4–6 weeks, significant scarring of glomeruli (glomerulosclerosis) and fibrosis of periglomerular regions was evident, indicating a continued injury response (Supplemental Figure 4C).

We analyzed purified glomeruli on day 1 (peak proteinuria) and day 5 (proteinuria largely resolved) after administration of nephrotoxic serum (Supplemental Figure 5, A–F). Podocytes, mesangial cells, and endothelial cells from nephritic mice at day 1 each formed clusters that were clearly distinct from their control counterparts, indicating a major shift in gene expression for these three major cell types (Figure 2A). In contrast, PECs and SMCs from control and nephritic mice formed overlapping clusters, suggesting the changes that were seen are not a batch artifact. This conclusion was supported by using a method of random downsampling of cells (see Supplemental Figure 5G), as well as by qPCR validation of DE genes (see below).

Figure 2.

scRNA-seq analysis of nephritic mice reveal cell type–specific responses to injury. (A) T-distributed neighbor embedding (t-SNE) plot of control and nephritic mice at day 1. (B) Violin plots show the downregulation of podocyte-specific genes after nephrotoxic injury. Normalized expression levels are shown. (C) Heatmaps showing the top 100 most variable genes in each cell type. Each column represents a cell, each row represents a gene. (D) PCA plot of podocytes, mesangial cells, and endothelial cells from control and nephritic mice. (E) t-SNE plots of immune cells show increased immune infiltration after nephrotoxic injury. Cell types shown in the right panel were determined by expression of marker genes shown in Supplemental Figure 5L.

We identified 263, 85, and 165 upregulated genes and 312, 56, and 189 downregulated genes in nephritic podocytes, mesangial cells, endothelial cells, respectively (Supplemental Table 7). Changes in expression levels were confirmed by qPCR for a subset of these genes (Supplemental Figure 5H). Pathway analysis showed that podocytes induced programs in cytoskeletal regulation, cell adhesion, and inflammatory response (Supplemental Figure 5I). As expected, pathways related to podocyte differentiation were downregulated, because several podocyte-specific genes were decreased after injury (Figure 2B). Mesangial cells also showed changes in cytoskeletal regulation and cell adhesion, in addition to regulation of BP and immune responses (Supplemental Figure 5I).

On day 5, when proteinuria was nearly resolved, heatmaps of DE genes and PCA revealed the changes observed in podocytes at day 1 had largely normalized with a significant reduction in DE genes (Figure 2, C and D, Supplemental Table 7). In contrast, the transcriptional profiles of mesangial and endothelial cells continued to change at day 5 to a state further distant from the control cells with an increase in the number of DE genes (Figure 2, C and D, Supplemental Table 7). We also identified subclusters of dividing mesangial and endothelial cells (Supplemental Figure 5J). A subcluster of endothelial cells emerged on day 5 that had high expression of proangiogenic factors (Apln, Pgf) and the plasminogen activator inhibitor 1 gene (Serpine1) (Supplemental Figure 5K).

There were increased immune cells after injury with increased numbers of macrophages (F4/80⁺) and dendritic cells (DC-SIGN⁺ or XCR1⁺) on day 1 (Figure 2E, Supplemental Figure 5L). The number of myeloid cells was further increased on day 5, demonstrating that inflammation was persisting even in the absence of proteinuria. Ccl2, a chemokine responsible for monocyte recruitment, was detected in mesangial cells and immune cells at day 1 and was expressed by more cells on day 5 (Supplemental Figure 6A). Although CCL2 has widely been implicated in glomerular diseases, the source has largely been assumed to be from immune cells. We confirmed that CCL2 is produced by mesangial cells by generating a CCL2 reporter mouse (Supplemental Figure 6, B–E). After antibody injury, CCL2 reporter expression increased progressively over several days (Supplemental Figure 6, F and G). Flow cytometry showed >80% of the CCL2-expressing cells were mesangial cells (Supplemental Figure 6H). This suggests injured mesangial cells are an important driver of inflammation.

Hippo Pathway Is Involved in the Podocyte Response to Injury

To validate the podocyte injury expression signature, we focused our analysis on identifying pathways in podocytes induced after injury. Pathway analysis indicated that several signaling pathways, including Hippo, TGF-β, NF-κB, and FoxO pathways were upregulated in injured podocytes (Supplemental Figure 5I). Because the Hippo pathway is activated by disruption of cell junctions and because podocyte injury involves significant junctional reorganization (foot process effacement), we focused on the role of the Hippo signaling pathway.^34,35

The Hippo pathway is mediated by the coactivators YAP and TAZ which, when activated, are stabilized and facilitate transcription of target genes. We saw significantly increased protein levels of YAP and TAZ in glomerular lysates from nephritic mice (Figure 3A), and known targets of YAP and TAZ—including Ctgf, Cyr61, and Axl—were among the highest induced genes in podocytes after injury (Figure 3B).

Figure 3.

The Hippo pathway is required for podocyte response to injury. (A) Immunoblots showing increased protein levels of TAZ and YAP in glomerular lysates 2 days after nephrotoxic injury. Blots are representative of two independent experiments. Vinculin was used as loading control. (B) YAP/TAZ target genes are induced in podocytes after nephrotoxic injury. Normalized expression levels are shown as violin plots. (C) Deletion of TAZ (Wwtr1^−/−) or YAP (Yap1^−/−) exacerbates proteinuria. Albumin-creatinine ratios (ACR) were measured from spot urine collected from mice at the indicated time points after nephrotoxic serum injection. (D) Kidney sections were stained with PAS and Masson trichrome stain 5 weeks after injection with nephrotoxic serum or saline and scored blindly for GN and fibrosis. Scoring criteria are described in the methods. Results were analyzed by multiple t tests using the two-stage linear step-up method (*q<0.05, **q<0.01, ****q<0.0001). WT, wild type.

We generated mice that allowed for conditional deletion of either TAZ (Wwtr1^−/−) or YAP (Yap1^−/−) (Supplemental Figure 7A). In contrast to a previous study that deleted YAP during development,⁴⁰ we found that deletion of YAP or TAZ in adult mice did not, by itself, cause any renal dysfunction (Supplemental Figure 7B). After nephrotoxic serum injection, however, Wwtr1^−/− and Yap1^−/− mice displayed significantly higher levels of proteinuria at day 1 compared with wild-type mice and they also exhibited delayed resolution (Figure 3C). Kidney sections examined at 5 weeks after nephrotoxic serum injection showed increased injury for both knockout mice compared with wild-type controls (Figure 3D). These effects were less severe in the Wwtr1^−/− mice compared with Yap1^−/− mice, which may be due to incomplete deletion of TAZ (Supplemental Figure 7A). This suggests Hippo pathway signaling is important for podocyte recovery after immune injury and validates the use of our database in identifying important reparative pathways.

Injury Response in Different Glomerular Injury Models

We performed scRNA-seq on several other models of glomerular injury. We used BTBR ob/ob mice,⁴¹ doxorubicin treatment,⁴² and CD2AP-deficient mice² to model glomerular injuries caused by metabolic disorder, drug toxicity, and podocyte-specific genetic disease, respectively.

Leptin-deficient BTBR ob/ob mice, a model of type 2 diabetes, become hyperglycemic around 6–8 weeks of age, exhibited renal injury with progressive proteinuria at around 10 weeks of age, and developed glomerular lesions that closely resemble those observed in human diabetic kidney disease by 20–22 weeks of age.⁴¹ We therefore analyzed glomerular cells at 12 and 21 weeks of age (Supplemental Figure 8A). Comparison of cells from diabetic (ob/ob) and control (ob/+) mice by PCA showed changes in podocytes and mesangial cells, but did not show major changes in endothelial cell gene expression, a surprising finding given that diabetes is considered to be a vascular disease (Figure 4A).⁴³ In addition, the gene expression profiles of podocytes and mesangial cells did not change significantly between 12 weeks and 21 weeks of age, suggesting the changes induced by diabetes are chronic (Figure 4A). Pathway analysis showed changes to glucose and lipid metabolism pathways in both podocytes and mesangial cells (Supplemental Figure 8B). Cell proliferation pathways were induced in mesangial cells, and apoptotic pathways were induced in podocytes. Consistent with this, the number of mesangial cells were increased (41.5% of total cells compared with 11.9% in control mice) whereas podocytes were decreased (4.4% compared with 6.5% in control mice) at 21 weeks. Consistent with the expanded matrix that is a hallmark of diabetic kidney disease, there was increased expression of matrix and matrix-modifying proteins in both podocytes and mesangial cells (Figure 4, B and C).

Figure 4.

Glomerular cells display distinct responses to different types of injury. (A) PCA plots of podocytes, mesangial cells, and endothelial cells from control and diabetic mice. (B and C) Violin plots show increased expression of extracellular matrix and matrix-modifying genes in (B) podocytes and (C) mesangial cells from diabetic mice. (D) T-distributed neighbor embedding (t-SNE) plot of immune cells shows increased numbers of neutrophils and monocytes/macrophages in doxorubicin-treated mice. Immune cell types were determined by expression of marker genes shown in Supplemental Figure 8I. (E) Violin plot shows increased expression of Cxcl1 in mesangial cells from doxorubicin-treated mice. (F) PCA plots of podocytes, mesangial cells, and endothelial cells from wild-type and Cd2ap^−/− mice. (G) Graphs showing the fraction of mesangial cells expressing Ren1 after injury caused by nephrotoxic serum, diabetes, and doxorubicin.

Doxorubicin is a model of glomerular injury due to its toxicity to podocytes and glomerular endothelial cells.^42,44 In mice, doxorubicin induces proteinuria that begins after 7–14 days and progressive kidney deterioration that occurs after several weeks. Analysis of glomeruli 14 days after doxorubicin injection revealed reduced podocyte numbers (Supplemental Figure 8C) and downregulation of podocyte marker genes (Supplemental Figure 8D). p53 signaling was increased in podocytes, demonstrated by increased Cdkn1a and Gadd45g, which was likely secondary to reactive oxygen species–mediated DNA damage (Supplemental Figure 8, E and F).⁴⁵ In support of elevated oxidative stress, the metallothionein genes Mt-1 and Mt-2 were among the highest upregulated genes in both mesangial and endothelial cells (Supplemental Figure 8, G and H). Distinct to this model was a significant increase in the number of neutrophils in the glomerulus after injury (Figure 4D, Supplemental Figure 8I). Expression analysis showed that mesangial production of Cxcl1, a key neutrophil chemokine, is the likely mechanism (Figure 4E).

Lastly, we analyzed mice lacking CD2AP, which is a scaffolding protein expressed in podocytes and associated with sporadic nephrotic syndrome and FSGS in humans.² CD2AP-deficient mice develop proteinuria around 2–3 weeks of age and die around 6–7 weeks of age due to renal failure. We purified glomeruli from 3-week-old animals, soon after the onset of proteinuria. The number of podocytes was significantly reduced in the CD2AP-deficient mice compared with age-matched, wild-type mice (31.5% of total in wild type, 10.7% of total in knockout) (Supplemental Figure 8J). This was associated with upregulation of apoptotic pathways in podocytes (Supplemental Figure 8K). Consistent with expanded mesangial matrix seen in CD2AP-deficient kidneys, podocytes and mesangial cells both showed increased expression of matrix proteins (Supplemental Figure 8, L and M). In contrast, there was very little change detected in endothelial cells or immune cells (Figure 4F).

Discussion

CKD is a major health care problem affecting approximately 15% of the population, but treatment options are limited beyond dialysis and kidney transplantation.¹ With the recognition that many kidney diseases have a hereditary component and with the identification of susceptibility genes for kidney diseases, such as FSGS, IgA nephropathy, and CKD, there has been renewed interest in the molecular basis of kidney diseases. A more precise understanding of glomerular cell injury at the molecular level has the potential to revolutionize diagnosis and lead to new therapies.

For this reason, we used scRNA-seq to characterize the cells of the glomerulus and monitor the changes that occur after several different types of injury. The quality of the data generated by single-cell sequencing is greatly affected by the condition of the prepared cells. Our optimized methods allowed the preparation of >0.5×10⁶ cells from the glomerulus of a single mouse with approximately 95% viability in <3 hours. The low mitochondrial read content; high gene-detection sensitivity; and the balanced sampling of podocytes, mesangial cells, and endothelial cells all attest to the quality of our cell preparations (Supplemental Table 8). Using these methods, we were able to generate a comprehensive and detailed single-cell data set focused on the glomerulus that allowed for in-depth analysis. The distinct clustering of the AA and EA SMCs, which are a minor population of cells, and the subsequent identification of their respective markers, affirms the high resolution of the data set.

We are not the first to use scRNA-seq to examine the glomerulus. Karaiskos et al.²³ sequenced approximately 13,000 glomerular cells, but required 32 mice to generate this number of cells. The cell preparation in this study is dominated by podocytes, suggesting issues with tissue dissociation. Comparison with our data suggest that the population of cells that were identified as mesangial cells are instead mostly SMCs/JG cells as evidenced by the expression of Acta2, Myh11, and Akr1b7. In another study that analyzed glomerular cells from a model of type 1 diabetes (streptozotocin-injected eNOS knockout mice), Fu et al.²⁴ sequenced a total of 644 cells from control and diabetic mice. This study was limited by the small number of cells, the authors were unable to distinguish mesangial cells from SMCs/JG cells and they were unable to obtain sufficient numbers of podocytes from diabetic mice for analysis.

Other studies have identified glomerular cells using single-cell approaches of the whole kidney. In general, these approaches identified relatively few glomerular cells, partly due to their paucity, but also because tissue-dissociation methods optimized for the whole kidney are not ideal for glomeruli. Park et al.²⁸ only identified 78 podocytes from 43,745 total cells, whereas Ransick et al.⁴⁶ found 24 podocytes from 31,265 total cells. It is also difficult to clearly identify mesangial cells in these data sets because of the similarity between mesangial cells and stromal mesenchymal cells. Wu et al.⁴⁷ and Lake et al.⁴⁸, using single-nucleus RNA-seq of whole mouse and human kidneys, identified severalfold more glomerular cells, but the gene detection sensitivity was three- to fivefold lower than in our data set. In addition, the quality control steps, which use mitochondrial read content (indicative of damaged or dead cells) as a metric, are complicated by the high abundance of mitochondrial transcripts in the renal parenchyma when analyzing the whole kidney. This is not a problem when using purified glomerular cells. A table comparing our data with previous studies is shown in Supplemental Table 8.

Our analysis suggests that difficulties in identifying mesangial cell–specific genes is likely due to their close similarity with SMCs. For example, a recent study defined a mesangial gene signature using bulk analysis of cells isolated based on expression of Meis1, a developmental gene of mesangial cells.^49,50 We found that several genes identified in this study are not specific to mesangial cells, and in some cases are specific to SMCs (Myh11, Rergl, Pln, and Olfr558) (Supplemental Figure 9A). Potential explanations for these discrepancies could be the expression of Meis1 in cells other than mesangial cells in adult glomeruli (Supplemental Figure 9B) or impurities in the bulk cell preparations. Our gene signatures suggest that several previous single-cell studies have either misidentified SMCs for mesangial cells or conflated these two cell types together.^23,24 The bona fide mesangial cell markers identified in this study should provide more precise tools to interrogate the function of mesangial cells in vivo.

Based on their location adjacent to endothelial cells, mesangial cells are considered a specialized pericyte.¹⁸ They function in stabilizing the vasculature, synthesizing components of the basement membrane, and participating in controlling capillary vascular tone. However, focus over the last 10 years in glomerular disease pathogenesis has shifted to the podocyte because mesangial cells are considered to play a secondary and bystander role.¹⁸ With the recognition that stromal cells, including fibroblasts and pericytes, play critical roles in promoting inflammation in tumors⁵¹ as well as in liver and lung fibrosis,⁵² by analogy, it is likely that the mesangial cell is playing a similar role.

The four injury models we analyzed displayed a wide spectrum of outcomes with varying rates and patterns of disease progression and cell types affected, with little overlap in the transcriptional responses to each type of injury (Supplemental Figure 10, A and B). However, in all of the injury models, mesangial cells showed persistent induction of genes involved in wound healing (Supplemental Table 9) with distinct expression patterns of matrix (Col4a1, Col4a3, Col6a3, Col8a1) and chemokine genes (Ccl2, Cxcl1, Cxcl13, Cx3cl1) in each injury model. We also noted induction of Ren1 expression in a subset of mesangial cells in all models except CD2AP deficiency (Figure 4G). In diabetic mice, the number of Ren1-expressing mesangial cells increased twofold from week 12 to week 21 and, in doxorubicin-treated mice, >25% of the mesangial cells expressed Ren1. Collectively, these results suggest mesangial cells may shape the character of the inflammation and wound-healing programs in response to distinct types of injuries. It also suggests that a persistent mesangial reaction may drive the chronic decline of kidney function in many diseases.

We noted some novel findings in each of the injury models. The lack of an endothelial reaction in the ob/ob mice was unexpected and suggests that diabetes does not by itself induce transcriptional changes in endothelial cells. Given the prevalence of hypertension in patients with diabetes, hypertension might be important in injuring endothelial cells.⁵³ The fact that BTBR ob/ob mice do not develop hypertension could explain why studies using a different model of diabetic nephropathy, which does become hypertensive, did show transcriptional changes to endothelial cells.²⁴ The Cd2ap^−/− mice differed from the other three models by the lack of Ren1 expression in injured mesangial cells. We speculate this may be due to the young age of the mice. This is an intriguing possibility to explore because children that have nephrotic syndrome often outgrow it by their teen years without fibrotic sequelae. It should be noted that there may be differences in tissue dissociation between different mouse strains and, therefore, cross-strain comparisons in cell numbers between C57BL/6J and BTBR mice are not possible.

Overall, our data illustrate the power of using scRNA-seq to monitor the complex tissue injury process during disease. Focusing on the glomerulus allowed detailed characterization of multiple cell types that exist in small numbers and simultaneous monitoring of their responses to injury. Whereas we were able to identify and validate the Hippo pathway as critical for podocyte repair, other pathways like the TGF-β, Wnt, and Notch pathways are implicated by our data and likely to play important roles. By generating comprehensive snapshots of the altered genetic landscapes in multiple injury models, this work provides data to test the link between injury type, glomerular cell response, inflammation, and histology. Although the exact mechanisms of this process require further investigation, our results provide insight into the underlying pathophysiologic pathways and potential novel therapeutic approaches for glomerular diseases.

Disclosures

Y. Chen, J. Chung, A. Dey, L. Goldstein, J. Lee, Z. Modrusan, M. Roose-Girma, A. Shaw, and J. Webster report employment at Genentech. All remaining authors have nothing to disclose.

Funding

This work was supported by Genentech.

Data Sharing Statement

The scRNA-seq data generated in this study have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus (GSE146912) and can be accessed at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE146912 or the Broad Institute Single Cell Portal at https://singlecell.broadinstitute.org/single_cell/study/SCP975. Mice generated for this study are available from the corresponding author on reasonable request. All other unique reagents used in this study are available from the corresponding author on reasonable request.

Supplemental Material

This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2020020220/-/DCSupplemental.

Supplemental Figure 1. Diagram of the glomerular structure.

Supplemental Figure 2. Unbiased clustering reveals the major cell types of the glomerulus.

Supplemental Figure 3. Mapping of FSGS and CKD susceptibility gene expression to glomerular cell types.

Supplemental Figure 4. Nephrotoxic serum nephritis in C57BL/6J mice.

Supplemental Figure 5. scRNA-Seq analysis of glomerular cells from nephritic mice.

Supplemental Figure 6. Induction of Ccl2 expression in nephrotoxic serum nephritis.

Supplemental Figure 7. Conditional deletion of TAZ and YAP.

Supplemental Figure 8. scRNA-Seq analysis of glomerular cells from diabetic, doxorubicin-treated, and CD2AP-deficient mice.

Supplemental Figure 9. Validation of previously reported mesangial cell-specific genes.

Supplemental Figure 10. Glomerular cells show distinct responses to different types of injury.

Supplemental Table 1. Genes differentially expressed between two podocyte sub-clusters (2 and 4) in Supplemental Figure 2A.

Supplemental Table 2. List of canonical cell type markers used to determine the identity of cell clusters.

Supplemental Table 3. Composition of glomerular cells from a C57BL/6J mouse analyzed by scRNA-Seq.

Supplemental Table 4. Top 50 cell type marker genes.

Supplemental Table 5. Genes differentially expressed between mesangial cells (cluster 3) and SMCs/JG cells (cluster 5).

Supplemental Table 6. Genes enriched in mesangial cells.

Supplemental Table 7. Number of differentially expressed genes in nephritic mice.

Supplemental Table 8. Comparison of kidney single cell datasets.

Supplemental Table 9. Genes commonly induced in mesangial cells after different injury types.