A Comprehensive Immune Cell Atlas of Cystic Kidney Disease Reveals the Involvement of Adaptive Immune Cells in Injury-Mediated Cyst Progression in Mice

Significance Statement

It is widely accepted that injuries to cilia mutant mice accelerate the rate of cystic kidney disease. However, cellular factors that accelerate cystic disease are unknown. By performing single-cell RNA sequencing of all CD45⁺ immune cells, we found that the subtypes and gene expression profiles of adaptive immune cells are significantly altered among non-injured, aged cystic mice; injury-accelerated cystic mice; and noncystic controls. Surprisingly, deletion of all adaptive immune cells reduced cystic disease in the injury-accelerated model but had no effect on cystic disease in the non-injured model. This differential rescue may be due to unique adaptive immune cell subtypes and ligands that are only present in the injury-accelerated model of cystic disease.

Visual Abstract

Abstract

Background

Inducible disruption of cilia-related genes in adult mice results in slowly progressive cystic disease, which can be greatly accelerated by renal injury.

Methods

To identify in an unbiased manner modifier cells that may be influencing the differential rate of cyst growth in injured versus non-injured cilia mutant kidneys at a time of similar cyst severity, we generated a single-cell atlas of cystic kidney disease. We conducted RNA-seq on 79,355 cells from control mice and adult-induced conditional Ift88 mice (hereafter referred to as cilia mutant mice) that were harvested approximately 7 months post-induction or 8 weeks post 30-minute unilateral ischemia reperfusion injury.

Results

Analyses of single-cell RNA-seq data of CD45⁺ immune cells revealed that adaptive immune cells differed more in cluster composition, cell proportion, and gene expression than cells of myeloid origin when comparing cystic models with one another and with non-cystic controls. Surprisingly, genetic deletion of adaptive immune cells significantly reduced injury-accelerated cystic disease but had no effect on cyst growth in non-injured cilia mutant mice, independent of the rate of cyst growth or underlying genetic mutation. Using NicheNet, we identified a list of candidate cell types and ligands that were enriched in injured cilia mutant mice compared with aged cilia mutant mice and non-cystic controls that may be responsible for the observed dependence on adaptive immune cells during injury-accelerated cystic disease.

Conclusions

Collectively, these data highlight the diversity of immune cell involvement in cystic kidney disease.

Cystic kidney disease is a common phenotype that is observed in ciliopathy patients and in ciliopathy mouse models.¹ Despite the fact that each cell of the nephron carries the same germline mutation,² cyst growth does not occur uniformly across the kidney. The variability in the rate of cyst growth in patients has been attributed to a combination of differences in secondary somatic mutations, genetic modifiers, and alterations in the renal microenvironment.² The differential rate of cyst growth observed in humans can be recapitulated in mouse models. For example, disruption of ciliary genes in adult mice leads to slow, focal cyst formation despite the fact that nearly every tubule in the kidney lacks functional cilia.^3,4 Because these mice have a uniform genetic mutation across all tubules, it is likely that the focal nature of cysts in these mice is due to environmental and cellular factors that are only found around certain tubules. One environmental factor is likely to be renal injury as introduction of ischemia-reperfusion injury (IRI) greatly increases the number of tubules that become cystic and their rate of expansion.^5–7

Recent data have identified a critical role for the innate immune system and macrophages in promoting cyst progression.^8–18 However, the contribution of other immune cells to renal cystic disease is poorly understood. Zeier and colleagues made the seminal observation that patients with polycystic kidney disease (PKD) had increased histologic staining for T-cell markers, mainly CD4⁺ T cells, in interstitial regions of the kidney.¹⁹ The presence of lymphocytes has also been observed in other rodent models of cystic disease.^20–22 Additional data indicate that the number of CD4⁺ and CD8⁺ T cells is increased in the Pkd1^RC/RC mouse model of autosomal dominant PKD (ADPKD), and that depletion of CD8⁺ T cells worsens cystic disease.²³ Further, the number of CD4⁺ and CD8⁺ T cells is increased in patients with ADPKD and the number of urinary CD4⁺ T cells correlates with the rate of decline in renal function over a 5-year period.²⁴ Collectively, these data suggest that CD8⁺ T cells are protective in cystic disease whereas CD4⁺ T cells may be promoting disease progression.

By performing single-cell RNA sequencing (scRNAseq) of kidney cells isolated from two models of cystic disease (rapid, injury-accelerated and slow, non-injured, aged) that have the same underlying genetic mutation (Ift88), we find that the number and gene expression of adaptive immune cells is significantly altered between models. Strikingly, loss of adaptive immune cells, using the genetic Rag1 knock-out model, significantly reduced cystic disease in the injury-accelerated model of cystogenesis but did not affect cyst progression in the absence of injury, independent of the rate of cyst progression or genetic mutation. Using NicheNet, we identify candidate adaptive immune cell subtypes and ligands that may be driving injury-accelerated cystic disease. Collectively, our data indicate that renal injury induces changes in adaptive immune cells and that these changes are required for rapid cyst growth after injury to cilia mutant (CM) mice.

Methods

Mice

Eight-week-old CAGG-Cre/Esr1/5Amc/J (CAGG-Cre^ERT2) IFT88^f/f, Cagg-Cre^ERT2 IFT88^f/f Rag1^−/−, and C57BL/6 and BALB/c Pkd1^RC/RC Rag1^−/− mice were bred in-house. Ift88 animals were maintained in Association for Assessment and Accreditation of Laboratory Animal Care (AALAC) International–accredited facilities in accordance with Institutional Animal Care and Use Committee regulations at the University of Alabama at Birmingham (UAB) under approval numbers 10130 and 21072 and at the University of Oklahoma Health Sciences Center under approval number 20–010-SCH. Pkd1^RC/RC Rag1^−/− animals were maintained in AALAC-accredited facilities in accordance with Institutional Animal Care and Use Committee regulations at the University of Colorado under approval numbers 33 and 685.

Cre Induction

To induce Cre recombinase activity in 7–9-week-old CAGG-Cre^ERT2 IFT88^f/f animals, a single intraperitoneal injection of tamoxifen at 6 mg/40 g body wt was administered once daily for 3 consecutive days.

Renal IRI

Three weeks after tamoxifen induction, mice underwent unilateral IRI for 30 minutes or sham surgery according to our previous description.¹⁸ Mice were weighed and 15 mg/kg body wt avertin was administered via intraperitoneal injection before surgery. Mice were maintained under 1% isoflurane during the surgical procedure. Surgery was carried out through a small incision made on the back with animals on a heating pad to maintain body temperature at 37°C. After surgery, mice were injected with pain medication (buprenorphine [0.05 mg/kg]) and allowed to recover on a 37°C heating pad. After the desired number of days, animals were anesthetized with avertin followed by whole-body perfusion with 1% PBS until the kidneys had a pale appearance. Harvested kidneys were minced for flow cytometry analysis or fixed in 4% (wt/vol) paraformaldehyde (PFA) for histology.

Fixation and Tissue Processing

After harvesting, mouse kidneys were cut in half and specimens were immediately immersed in 4% (wt/vol) PFA overnight at 4°C. After PFA fixation, kidneys were switched to 70% ethanol overnight, embedded in paraffin, sectioned at 5 µm, and stained using hematoxylin and eosin (H&E).

Quantification of Cystic Severity

First, 5-μm sections were cut from paraffin-embedded kidneys and stained with H&E. To identify cysts, we began by quantifying the average diameter and standard deviation of a normal tubule using control, Cre-negative mice. Cysts were identified as any opening (white space) that was >3 SD above the average diameter of a normal tubule. To ensure that we did not include blood vessels and other artifacts that were the result of tissue processing in our cystic calculations, we went back and individually analyzed all particles that were identified as cysts using ImageJ. After removal of noncystic structures, we calculated cystic index as the total cystic area divided by the total area of the sectioned kidney. This quantification was done using at least half of the kidney for all animals. A technician blinded to the treatment modality performed the quantification and analysis. Cystic severity as measured by %KW/BW, cystic index, cyst size, and cyst number for the Pkd1^RC/RC;Rag1 studies was performed as described by Kleczko et al.²³

Tissue Processing for Flow Cytometry

After perfusion of the mouse with PBS, the left kidney (sham, injured, or aged) was removed and put into RPMI 1640 on ice. For the initial comparison of digestion methods, kidneys were minced and digested in either (1) 1 ml of RPMI 1640 containing 1 mg/ml collagenase type I (Sigma-Aldrich) and 100 U/ml DNase I (Sigma-Aldrich) for 30 minutes at 37°C with mixing, or (2) 1 ml of 2.5 mg/ml type II collagenase (Worthington, catalog no. LS004176), 7.5 mg/ml B. licheniformis cold-activated protease (Creative Enzymes, catalog no. NATE-0633), and 125 U/ml DNase in D-PBS for approximately 45 minutes at 12°C. Before digestion, a piece of minced kidney tissue was directly put in TRIzol followed by isolation of mRNA. After digestion, kidney tissue from both groups was individually passed through 70-µm cell-strainers (Falcon; BD Biosciences), yielding single-cell suspensions. Cells were centrifuged at 1300 rpm for 5 minutes at 4°C, resuspended in ACK red blood cell lysis buffer, and incubated at 37°C for 5 minutes. After 5 minutes, 10 ml of RPMI was added to each tube, cells were spun at 1300 rpm for 5 minutes at 4°C, and cells were resuspended in 1% BSA with FC blocking solution (1:200; BioXcell, catalog no. BE0307) for 30 minutes on ice. For all subsequent scRNAseq experiments, we used the 12°C digestion protocol on isolated kidney tissue.

Flow Cytometry for Single-Cell Experiments

After blocking, cells were spun and incubated with the following antibodies in 1% BSA for 30 minutes: FITC-conjugated Lotus tetragonolobus agglutinin (LTA) (catalog no. FL-1321; Vector Laboratories), PE rat anti-mouse CD45 (catalog nos. 12–0451, 30-F11; eBioscience), and Fixable Aqua Dead Cell Stain (catalog no. L34957; Invitrogen). After staining for 30 minutes, cells were spun, washed with 1% BSA, and sorted using a Becton-Dickenson FACSAriaII. For the initial experiments testing the effect of digestion on stress response genes, cells were sorted directly into 1 ml of TRIzol. For subsequent experiments involving scRNAseq, we sorted approximately 25,000 live, CD45⁺, LTA⁺, and CD45⁻/LTA⁻ cells into individual BSA-coated tubes. Cells were counted and approximately 3300 cells from each of the three groups (CD45⁺, LTA⁺, CD45⁻/LTA⁻) were combined into a single tube followed by performing bead emulsion and 10x Genomics. This was done for n=2 animals from each experimental group; all animals were female.

Flow Cytometry Validation of Key Immune Cell Subsets

To validate key immune cell subsets, we performed flow cytometry on individual mice from all five experimental groups. After perfusion of the mouse with PBS, the left kidney (sham, injured, or aged) was removed and put into RPMI 1640 on ice. Kidneys were minced and digested in 1 ml of RPMI 1640 containing 1 mg/ml collagenase type I (Sigma-Aldrich) and 100 U/ml DNase I (Sigma-Aldrich) for 30 minutes at 37°C. Kidney fragments were passed through a 70-µm mesh (Falcon; BD Biosciences), yielding single-cell suspensions. Red blood cells were lysed and cells were resuspended in 1 ml of PBS containing 1% BSA with Fc blocking solution for 30 minutes on ice. Then, 2 × 10⁶ cells were stained for 30 minutes at room temperature with primary antibodies listed in Supplemental Data. Cells were washed, fixed in 2% PFA for 30 minutes, and resuspended in FACS staining buffer.

For intracellular staining of T-cell cytokines, cell suspensions were incubated with Foxp3 Fixation/Permeabilization solution (ThermoFisher) overnight at 4°C after addition of the primary antibodies. The next morning, cells were washed twice with permeabilization buffer and stained with intracellular antibodies (Supplemental Data) made in 1% BSA for 30 minutes at room temperature. Cells were washed with perm buffer, spun, and resuspended in PBS. Samples were run on a BD LSRII and analyzed using FlowJo software.

10x Genomics

10x Chromium single-cell libraries were prepared according to the standard protocol outlined in the manual. Briefly, sorted single-cell suspensions, 10x barcoded gel beads, and oil were loaded into Chromium Single Cell Chip B to capture single cells in nanoliter-scale oil droplets by Chromium Controller and to generate Gel Bead-In-EMulsions (GEMs). We used “Chromium Single Cell 3' GEM, Library & Gel Bead Kit v3” and Chip B for these experiments. Full-length cDNA libraries were prepared by incubation of GEMs in a thermocycler machine. GEMs containing cDNAs were broken and all single-cell cDNA libraries were pooled together, cleaned using DynaBeads MyOne Silane beads (Fisher, PN 37002D), and preamplified by PCR to generate sufficient mass for sequencing library construction. Sequencing libraries were constructed by the following steps: cDNA fragmentation, end repair and A-tailing, size selection by SPRIselect beads (Beckman Coulter, PN B23318), adaptor ligation, sample index PCR amplification, and a repeat of SPRIselect beads size selection. The final constructed single-cell libraries were sequenced by an Illumina Nextseq machine with total reads per cell targeted for a minimum of 25,000.

Single-Cell Sequencing Data Processing

The 10x Genomics Cell Ranger software (version 2.1.1), “mkfastq”, was used to create the fastq files from the sequencer. After fastq file generation, Cell Ranger “count” was used to align the raw sequence reads to the reference genome using STAR. The “count” software created three data files (barcodes.tsv, genes.tsv, matrix.mtx) from the “filtered_gene_bc_matrices” folders that were loaded into the R package Seurat version 3.1.5,²⁵ which allows for selection and filtration of cells on the basis of QC metrics, data normalization and scaling, and detection of highly variable genes. We followed the Seurat vignette (https://satijalab.org/seurat/pbmc3k_tutorial.html) to create the Seurat data matrix object. We then combined Seurat objects from each individual experiment using the “merge” function. To ensure that the data were reproducible when harvested on individual days, we integrated Seurat objects from experiments that were done on different days using the “IntegrateData” function and dims=1:30. We confirmed that all datasets had significant integration across experimental replicates before combining. After merging, we processed data to remove low-quality cells by keeping all genes expressed in >3 cells and cells with at least 200 detected copies. Cells with mitochondrial gene percentages >50% and unique gene counts >3000 or <200 were discarded. For these analyses, we initially set the percentage of mitochondrial reads that were permitted at 50% due to the high mitochondrial activity in the proximal tubule epithelium, as previously reported in single-cell datasets from the kidney.^26,27 For subsequent analysis of immune cell populations, we reduced the permitted mitochondrial reads to 10%, similar to previous reports.²⁸ The data were normalized using Seurat’s “NormalizeData” function, which uses a global-scaling normalization method, LogNormalize, to normalize the gene expression measurements for each cell to the total gene expression. The result is multiplied by a scale factor of 1 × 10⁴ and the result is log-transformed. Highly variable genes were then identified using the function “FindVariableGenes” in Seurat. Genes were placed into 20 bins on the basis of their average expression and removed using 0.0125 low cutoff, 3 high cutoff, and a z-score cutoff of 0.5. We also regressed out the variation arising from library size and percentage of mitochondrial genes using the function “ScaleData” in Seurat. We performed principal component (PC) analysis of the variable genes as input and determined significant PCs on the basis of the “JackStraw” function in Seurat. The first ten PCs were selected as input for Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction using the functions “FindClusters” and “DimPlot” in Seurat. To identify differentially expressed genes (DEGs) in each cell cluster, we used the function “FindAllMarkers” in Seurat on the normalized gene expression data.

Integration of AKI and Unilateral Ureteral Obstruction Single-Cell Datasets

To compare immune cells in cystic disease with immune cells from other models of kidney injury/disease, we downloaded raw scRNAseq data from mouse models of AKI (E-MTAB-8002)²⁹ and unilateral ureteral obstruction (UUO) (GSE140023).³⁰ Cell Ranger “count” was used to align the raw sequence reads to the reference genome using STAR. The “count” software created three data files (barcodes.tsv, features.tsv, matrix.mtx) from the “filtered_feature_bc_matrix” folder which were loaded into the R package Seurat version 4.0.4. Data integration and merging were performed using the “Introduction to scRNA-seq integration” vignette for Seurat (https://satijalab.org/seurat/articles/integration_introduction.html). After integration, we processed data to remove low-quality cells by keeping all genes expressed in >3 cells and cells with at least 200 detected gene copies. Cells with mitochondrial gene percentages >10% and unique gene counts >3000 or <200 were discarded. After integration and preprocessing, data were normalized and clustered as described above. An analysis of cells that passed quality control and genes enriched in each cluster are shown in the Supplemental Material.

NicheNet Analysis

NicheNet analysis was done on cells from injured cilia mutant and aged cilia mutant groups according to the code deposited in GitHub (https://github.com/saeyslab/nichenetr). For these experiments, we used injured control and aged control mice as reference datasets. The list of prioritized ligands for each “sender” cell type was identified on the basis of the top DEGs (versus respective control) that were found in the “receiver” cell type.

RNA Isolation and qRT-PCR

RNA was isolated and transcribed into cDNA, and qRT-PCR performed using TaqMan real-time PCR. The following probes were used: Egr1 (Mm00656724), Fos (Mm00487425), Jun (Mm07296811_s1), Hprt (Mm00446968_m1).

Code

A list of code used to generate the single-cell data in this paper can be found in the Supplemental Material.

Pathway Analysis

The Seurat function “FindAllMarkers” was run to identify DEGs that distinguished one cluster from all others. Pathway analysis was performed on genes enriched in each cluster of cells (adjusted P>0.05) using ingenuity pathway analysis software. The top three (if available) pathways that were enriched in each cluster of cells are shown. Likewise, the top three functions associated with the enriched genes in each cluster of cells are shown.

Immunofluorescence Microscopy

After overnight fixation, kidneys were switched to 30% sucrose for cryopreservation and OCT-embedded tissues were sectioned at 8 μm. For immunofluorescence studies, kidney sections were fixed with 4% PFA for 10 minutes, permeabilized with 1% Triton X-100 in PBS for 8 minutes, and incubated in 1 ml of blocking solution (PBS with 1% BSA, 0.3% Triton X-100, 2% [vol/vol] donkey serum, and 0.02% sodium azide) for 30 minutes at room temperature. Sections were incubated in 200 μl of primary antibody overnight at 4°C, washed with PBS, and incubated with the appropriate secondary antibodies in blocking solution for 1 hour at room temperature. The primary antibodies were: rat anti-mouse F4/80 (ThermoFisher, catalog no. 14–4801–82, Clone: BM8, diluted 1:200 in blocking solution) and rabbit anti-mouse CD206 (Abcam, catalog no. ab64693, Clone: EPR6868 [B], 1:200 dilution in blocking solution). The secondary antibodies were: Alexa Fluor 488–conjugated donkey anti-rabbit (ThermoFisher, catalog no. A-21206, diluted 1:1000 in blocking solution) and Alexa Fluor 647–conjugated goat anti-rat (ThermoFisher, catalog no. A-21247, diluted 1:1000 in blocking solution). After addition of secondary antibodies, nuclei were stained by Hoechst nuclear stain (Sigma-Aldrich, 1:1000 dilution in PBS) and samples mounted using IMMU-MOUNT (Thermo). All fluorescence images were captured on a Fluoview 1000/IX81 Laser Scanning Confocal microscope (Olympus) with an inverted configuration.

Blinding and Randomization

All mice used for scRNAseq were randomized before treatments. On the day of harvest, the flow cytometry and scRNAseq were performed by a scientist who was blinded to experimental groups. Bead capture, cDNA library preparation, sequencing, file conversion, and FASTq file generation were performed by the flow cytometry core at UAB in a blinded manner. Data were processed in R in a blinded manner until the cluster composition and cell numbers were revealed.

Statistical Analyses

Data were presented as mean±SEM. ANOVA and t tests were used for statistical analysis, and differences were considered significant for P<0.05.

Results

A Comprehensive Single-Cell Atlas of Slowly and Rapidly Progressive Models of Cystogenesis

The fact that cysts progress at different rates in injured and non-injured CM mice despite having the same genetic mutation suggests that renal injury influences the mechanism and/or cell types involved in cyst formation and progression. To identify candidate immune cells that modulate the rate of cyst growth, we performed scRNAseq of kidney cells isolated from adult-induced CM mice (CAGGCre^ERT2+ Ift88^f/f) that were harvested approximately 7 months after tamoxifen injection (CM aged) or 2 months post 30 minutes of unilateral IRI (CM IR). These time points were chosen as they represent periods of moderate cyst formation in both models, with cystic indices ranging from 10% to 15% (Supplemental Figure 1, A–D), thereby reducing differences that would be the result of severe cystic disease. We also sequenced cells isolated from the appropriate controls: CM mice that received a 30-minute sham surgery (CM sham) and were harvested 2 months post surgery, CAGGCre^ERT2- Ift88^f/f mice (intact cilia) that received a 30-minute IR injury (cont IR) and were harvested 2 months post surgery, and CAGGCre^ERT2- Ift88^f/f mice (intact cilia) that were aged (cont aged) for approximately 7 months after tamoxifen injection.

For scRNAseq, we isolated kidney cells from biologic replicates (n=2 per group, all female; 10 total mice) using a modified cold-activated protease digestion (hereafter referred to as CAP). This approach was selected on the basis of our data showing improved renal cell viability and reduced stress response signatures in CAP-digested kidney samples compared with 37°C-digested kidney samples in sham or IR injured mice (Supplemental Figure 2, A–C), similar to recent reports.^31,32 Because we hypothesized that the injury-mediated differences in the rate of cyst growth are due to abnormal epithelial-immune cell signaling networks, we sorted an equal number of CD45⁺ immune cells, LTA⁺ proximal tubule epithelium, and CD45⁻, LTA⁻ cells via FACS and performed scRNAseq using 10x Genomics (Figure 1A). We chose this approach to maximize our ability to analyze crosstalk between epithelial and immune cell populations, which would otherwise be difficult due to the high abundance of tubular epithelium in the kidney relative to the other cell types. After preprocessing and removal of low-quality cells, we merged data from biologic replicates and generated a combined UMAP that included all five experimental groups (CM aged, cont aged, CM IR, CM sham, cont IR). A summary of the number of cells sequenced for each replicate, mean reads per cell, median genes per cell, and number of cells that passed quality control for each replicate can be found in the Supplemental Material. Using the top DEGs from each cluster and well-established cell-specific markers, we identified and annotated ten clusters of cells: cluster 0: mononuclear phagocytes (MNPs; Lyz2, C1qa); cluster 1: vascular endothelial cells (Emcn, Meis2); cluster 2: S1/S2 proximal tubule (Miox, Aldob); cluster 3: T cells (Ccl5, Cd3g); cluster 4: collecting duct/loop of Henle (CD/LOH; Umod, Egf); cluster 5: B cells (Igkc, Cd79a); cluster 6: neutrophils (S100a9, S100a8); cluster 7: mesangial/interstitial cells (Acta2, Ctgf); cluster 8: S3 proximal tubule (Kap, Napsa); and cluster 9: podocytes (Nphs2, Podxl; Figure 1B, Supplemental Figure 3).^26,28,32,33

Figure 1.

Single-cell atlas of cystic kidney disease. (A) Schematic of the experimental design. The number of total cells from combined duplicates that passed quality control metrics in each group are listed. (B) UMAP of all sequenced cells that passed quality control in each experimental group. Each experimental group contains n=2 mice (both female) that were harvested, processed, and sequenced separately, followed by merging during the data processing and analysis steps.

The Immune Cell Atlas of Cystic Kidney Disease

Immune cells are known modifiers of disease progression in mouse models of cystic disease.^{8,12,17,18,23,24,34,35} To identify differences in immune cell composition between CM aged, CM IR, and non-cystic control mice that may be causing the differential rate of cyst progression, we reclustered immune cells from Figure 1 (clusters 0, 3, 5, and 6) to generate the immune cell atlas of cystic disease. Using the Immgen database (https://www.immgen.org/)³⁶ and commonly accepted cell-type–specific markers, we identified five broad clusters of immune cells: cluster 0: MNPs (C1qa, Lyz2); cluster 1: T cells (Ccl5, Cd3g); cluster 2: B cells (Cd79a, Cd79b); cluster 3: neutrophils (S100a9, S100a8); and cluster 4: IgA⁺ secretory B cells (Jchain, Igkc; Figure 2, A and B, Supplemental Figure 4). Because all immune cells express the common leukocyte antigen, CD45,³⁷ which we used to presort immune cells before scRNAseq, we can directly compare the number of immune cells across experimental conditions (equates to analyzing flow cytometry data as a percentage of live CD45⁺ cells) to gain insight into how the proportion of immune cells changes between models of cystic disease. The quantification of immune cell cluster composition as a percentage of all CD45⁺ immune cells in each biologic replicate is shown in Figure 2C. We did not observe a significant difference in immune cell cluster composition between groups, likely due to the fact that this was a broad analysis of immune cell populations (Figure 2C). A full list of genes enriched in each immune cell cluster can be found in the Supplemental Material.

Figure 2.

Immune cell atlas of cystic kidney disease. (A) UMAP showing all immune cells from each experimental group. (B) Feature plot showing expression of key immune cell markers that were used to define each cluster of cells. (C) Quantification of the proportion of immune cells in each cluster of biologic duplicates. Data are from n=2 mice; both female.

Only One Cluster of MNPs Is Different between Slowly and Rapidly Progressive Cystic Models

To identify differences in immune cell subsets between cystic models and non-cystic controls, we individually isolated and reclustered MNPs, T cells, and B cells using higher resolution clustering. Analysis of reclustered MNPs revealed ten clusters of cells including: cluster 0: kidney-resident macrophages (KRMs, C1qa, C1qb); cluster 1: Ly6c^hi infiltrating macrophages (Ly6c^hi Φ, Plac8, Chil3); cluster 2: Ly6c^lo infiltrating macrophages (Ly6c^lo Φ, Ear2, Pglyrp1); cluster 3: classic dendritic cells type 2 (cDC2, Cd209a, Clec10a); cluster 4: Mrc1⁺ KRMs (Ccl8, Mrc1); cluster 5: Spp1⁺ KRMs (Gpx3, Spp1); cluster 6: monocyte/dendritic cell mixture (monocyte/DC mixture; Vim, Anxa2; shared expression of Ly6c2 and Cd209a); cluster 7: vascular endothelial cells (Meis2, Emcn); cluster 8: B cells (Cd79a, Ly6d); and cluster 9: plasmacytoid dendritic cells (pDCs, Siglech; Figure 3, A and B, Supplemental Figure 5). A full list of genes enriched in each cluster of cells can be found in the Supplemental Material.

Figure 3.

There are limited differences in MNP clusters between CM aged and CM IR mice. (A) UMAP of MNPs from each experimental group. (B) Feature plot showing expression of key genes that were used to identify each cluster of cells. (C) Quantification of the proportion of MNPs in biologic duplicates shown as a percentage of total MNPs. (D and E) Heatmap showing the frequency of MNP subtypes as a proportion of (D) all MNPs or (E) CD45⁺ immune cells from each group. (F and G) Heatmap showing the frequency of MNP subtypes shown as a proportion of (F) MNPs or (G) CD45⁺ immune cells as determined by flow cytometry. CM aged (n=2), cont aged (n=3), CM IR (n=6), CM sham (n=3), cont IR (n=6). To determine statistical significance in (D)–(G), we performed a two-way ANOVA comparing (1) CM aged versus cont aged/CM sham, (2) CM IR versus CM Sham/cont IR, or (3) CM aged versus CM IR (to identify differences between cystic models). A value was considered significant if the experimental group was different compared with all relevant control groups. *Cluster was significantly different when comparing CM aged with cont aged/CM sham groups; ^$cluster was significantly different when comparing CM IR with CM sham/cont IR groups; ^#cluster was significantly different when comparing cystic models. */^#/^$P<0.05; **/^##/^$$P<0.01; ***/^###/^$$$P<0.001.

To identify MNP subtypes that may be influencing the differential rate of cyst growth observed between models, we quantified the number of cells in each MNP subset as a proportion of total MNPs (Figure 3C, shows each biologic replicate; Figure 3D, shows average of both replicates which was used for calculating statistics) and as a proportion of CD45⁺ immune cells (Figure 3E). In addition to directly comparing MNPs between cystic models, we also compared MNPs from CM aged and CM IR mice with their most appropriate controls (CM aged versus cont aged/CM sham; CM IR versus CM sham/cont IR). Our data indicate that Mrc1⁺ KRMs (cluster 4) were enriched in CM aged mice compared with both CM sham and cont aged groups, both as a proportion of all MNPs (Figure 3D) and as a proportion of CD45⁺ immune cells (Figure 3E). Likewise, cluster 6 (monocyte/DC mixture) was significantly enriched in CM IR mice compared with CM aged mice and controls (Figure 3, D and E).

Next, we designed a flow cytometry panel to validate our scRNAseq data (see Supplemental Figure 6A for gating strategy and markers used to identify each cluster). We focused our validation studies on clusters 0–4 because these clusters represented the overwhelming majority of cells in each experimental group. Analysis of flow cytometry data confirms the significant enrichment of Mrc1⁺ KRMs (identified via CD206 protein expression) in CM aged mice compared with CM IR and control mice as a percentage of MNPs (Figure 3F), CD45⁺ cells (Figure 3G), and live single cells (Supplemental Figure 6B). Further, Mrc1⁺ KRMs were localized to cystic regions of CM aged kidneys but were only sparsely found in cystic regions of CM IR mice (Supplemental Figure 6C). Unfortunately, we were unable to design a flow cytometry panel to validate cluster 6 because this cluster shared markers with several other cell lineages.

To gain insight into the possible function of Mrc1⁺ KRMs, which were enriched in CM aged mice, we performed an Ingenuity Pathway Analysis on genes that were significantly enriched in this cluster. Pathway analyses of the genes enriched in Mrc1⁺ KRMs indicate an association with the complement system (C1qa, C3ar1, C5ar1), phagosome formation (Fcrls, Mrc1, C5ar1), and IL-10 signaling (Fcgr2a, Ccr5, Il10rb; Supplemental Figure 7A). These genes are associated with functions including hematologic system development and function, cellular movement, and immune cell trafficking (Supplemental Figure 7B). Of note, excess complement activation^11,38 and IL-10 signaling¹⁰ are associated with worsened cystic disease, suggesting that Mrc1⁺ KRMs may be enhancing cyst growth in CM aged mice through these signaling networks.

Although analysis of cluster abundance suggests that Mrc1⁺ KRMs influence cystogenesis in CM aged mice, this analysis does not consider the possibility that the number of cells within a cluster could be the same between models although the gene expression and function are different. To address this possibility, we identified genes in clusters 0–5 that were significantly different (adjusted P <0.05) between CM aged and CM IR mice and performed a pathway analysis. When we did this, we noticed that MNPs from CM aged mice had an enrichment of genes associated with inflammatory response (clusters 0 and 1), leukocyte chemotaxis and differentiation (clusters 0 and 1), and phagocytosis/endocytosis (clusters 0, 3, and 5; Supplemental Figure 7C). In contrast, MNPs from CM IR mice had an enrichment of genes associated with leukocyte adhesion and migration (clusters 0 and 1), epithelial cell differentiation and ion transport (clusters 0 and 2), and antigen processing and presentation (clusters 4 and 5; Supplemental Figure 7C). Overall, these data suggest that MNPs from both models are being recruited to the kidney to influence cystic disease; MNPs from CM aged mice possibly phagocytose biologic material and produce inflammatory cytokines whereas MNPs from CM IR mice activate the adaptive immune system and produce factors that may influence epithelial function.

T-Cell Cluster Composition Is Significantly Different in Injured, Rapidly Progressive versus Non-injured, Slowly Progressive Renal Cystic Disease

Analysis of reclustered T cells (cluster 1, Figure 2) reveals 13 clusters, including: cluster 0: CD8⁺ T cells (Cd8a, Gzmk); cluster 1: CD4⁺ T regulatory cells (Tregs; Ikzf2, Foxp3); cluster 2: effector CD4⁺ T cells (Cxcr3, Cd40lg); cluster 3: naïve/central memory T cells (Sell, Ccr7, Lef1, Cd69); cluster 4: CD4⁺ Th17 cells (Rorc, Tmem176a, Lgals3); cluster 5: type I natural killer T cells (NKT1 cells; Klrk1, Ly6c2, Anxa1); cluster 6: Gzma⁺ natural killer cells (Gzma⁺ NK cells; Gzma, Tyrobp); cluster 7: Gzma⁺ CD8⁺ T cells (Gzma, Cx3cr1, Zeb2); cluster 8: Gzma^lo NK cells (Cd7, Xcl1); cluster 9: IFN-γ–responsive CD4⁺ T cells (CD4⁺ T cells–IFN-γ responsive; Ifit1, Ifit3); cluster 10: innate lymphoid cells (ILCs; Il1rl1, Arg1); cluster 11: epithelium (Gpx3, Aldob); and cluster 12: proliferating cells (Birc5, Stmn1; Figure 4, A and B, Supplemental Figures 8 and 9). A full list of genes enriched in each cluster can be found in the Supplemental Material.

Figure 4.

T-cell clusters are significantly different in CM IR and CM aged cystic models compared with each other and with noncystic controls. (A) UMAP of T cells found in each experimental group. (B) Feature plot showing expression of key genes that were used to identify each cluster of cells. (C) Quantification of cluster frequency in biological duplicates shown as a percentage of total T cells. (D and E) Heatmap showing the frequency of T-cell subtypes as a percentage of (D) all T cells or (E) CD45⁺ immune cells. (F and G) Heatmap showing the frequency of T-cell subtypes shown as a proportion of (F) T cells or (G) CD45⁺ immune cells as determined by flow cytometry. CM aged (n=2), cont aged (n=3), CM IR (n=6), CM sham (n=3), cont IR (n=6). Statistical significance was determined by two-way ANOVA, as described in Figure 3. */^#/^$P<0.05; **/^##/^$$P<0.01; ***/^###/^$$$P<0.001.

We once again quantified the number of cells in each cluster as a proportion of total T cells (Figure 4, C and D) and as a proportion of total CD45⁺ immune cells (Figure 4E). These analyses revealed that CM aged mice had a significant increase in the proportion of CD8⁺ T cells (cluster 0), CD4⁺ Tregs (cluster 1), and CD4⁺ Th17 cells (cluster 4) compared with both control groups (cont aged and CM sham) and a significant reduction in the proportion of effector CD4⁺ T cells (cluster 2) and naïve/central memory T cells (cluster 3) compared with both control groups (Figure 4, D and E). In contrast, only the proportion of naïve/central memory T cells (cluster 3) was significantly enriched in CM IR mice compared with both control (CM sham and cont IR) groups (Figure 4, D and E). Direct comparison of the two cystic models confirms that the proportions of CD8⁺ T cells (cluster 0), CD4⁺ Tregs (cluster 1), and CD4⁺ Th17 cells (cluster 4) were significantly enriched in CM aged mice compared with CM IR mice, whereas effector CD4⁺ T cells (cluster 2) and naïve/central memory T cells (cluster 3) were significantly enriched in CM IR mice compared with CM aged mice (Figure 4, D and E). Our analyses also found that NKT1 (cluster 5) and Gzma⁺ NK cells (cluster 6) were enriched in CM IR mice compared with CM aged mice as a proportion of T cells; Gzma⁺ NK cells were not different between groups as a percentage of CD45⁺ immune cells (Figure 4, D and E).

We once again performed flow cytometry to validate differences in key T-cell populations identified via scRNAseq (see Supplemental Figure 10A for markers used and gating strategy for each cluster). Analysis of flow cytometry data reveals an enrichment of CD8⁺ T cells in CM aged mice compared with CM IR mice as a proportion of T cells (Figure 4F) and as a proportion of live single cells (Supplemental Figure 10B). Likewise, we observed a significant enrichment of naïve/central memory T cells in CM IR mice compared with CM aged mice as a proportion of T cells (Figure 4F) and as a proportion of CD45⁺ immune cells (Figure 4G). We also observed a trend toward increased effector CD4⁺ T cells in CM IR mice compared with CM aged mice and controls and increased CD4⁺ Th17 cells in CM aged mice compared with CM IR mice and controls (Figure 4, F and G). Surprisingly, when we attempted to validate the increased proportion (or number) of CD4⁺ Tregs in CM aged mice compared with controls, we were unable to do so (Figure 4, F and G, Supplemental Figure 10B). This is likely because only a fraction of CD4⁺ Tregs (cluster 1) express the transcription factor Foxp3 by scRNAseq (Supplemental Figure 11A) and Foxp3 was used to identify Tregs via flow cytometry. Despite the lack of uniform Foxp3 expression in cluster 1, a majority of cells in this cluster expressed the transcription factor Ikzf2 (Helios; Supplemental Figure 11, B and C), which is a specific marker of Tregs.³⁹ Thus, it is likely that these cells are bona fide CD4⁺ Tregs.

We also performed pathway analyses of the top DEGs within each of the enriched T-cell clusters and did not find pathways/functions beyond what is already known for these T-cell subtypes (e.g., CD8⁺ T cells are predominantly involved in cytotoxicity-related pathways). However, a closer examination of genes associated with T-cell activation and exhaustion⁴⁰ revealed that CD8 T cells (cluster 0) expressed genes associated with T-cell exhaustion (Pdcd1; PD-1) whereas effector CD4⁺ T cells (cluster 2) expressed genes associated with T-cell activation (Cd40lg, Tnfrsf4, Tnfrsf18; Supplemental Figure 12, A and B). CD4⁺ Tregs (cluster 1) expressed markers of both activation and exhaustion (Supplemental Figure 12, A and B). These data suggest that CD8 T cells, which are enriched in CM aged mice, are exhausted whereas effector CD4⁺ T cells, which are enriched in CM IR mice, are activated. We also performed a pathway analysis on genes that were enriched in clusters of lymphoid cells isolated from CM aged and CM IR mice. These data indicate that CD8⁺ T cells and effector CD4⁺ T cells from CM IR mice had an enrichment of genes associated with ion transport, a hallmark of cystic kidney disease (Supplemental Figure 12, C and D).⁴¹ Unfortunately, we were unable to perform pathway analysis on the remaining lymphoid clusters due to the limited number of cells in each cluster.

B-Cell Clusters Are Different between Models of Cystic Disease

Reclustering of B cells (cluster 2, Figure 2) resulted in identification of six subclusters, including: cluster 0: B1 B cells (Ccr7, Cd69); cluster 1: T3/follicular helper B cells (Cr2, Ccr6); cluster 2: T1 B cells/memory B cells (Spib, Iglc1); cluster 3: germinal center/plasma blasts/plasma cells (Zbtb20, Apoe); cluster 4: memory B cells (Ifit3, Isg15); and cluster 5: T cells (Cd3g, Ms4a4b; Figure 5, A and B, Supplemental Figure 13). Comparison of B-cell cluster composition in CM aged and control mice indicates that the number of germinal center/plasma blasts/plasma cells (cluster 3) is significantly enriched in CM aged mice compared with both controls as a percentage of all B cells (Figure 5, C and D). However, when analyzing these cells as a percentage of CD45⁺ immune cells, the differences were not significant, likely due to the low overall number of B cells in the kidney in relation to total immune cells (Figure 5E). Comparison of B-cell cluster composition in CM IR mice and controls did not identify any significant differences (Figure 5, D and E). A direct comparison of B-cell subsets between cystic models reveals that B1 B cells (cluster 0), T3/follicular helper B cells (cluster 1), and germinal center/plasma blasts/plasma cells (cluster 3) were significantly different between cystic models as a percentage of total B cells (Figure 5D), although only B1 B cells were significantly different between groups as a percentage of CD45⁺ cells (Figure 5E). Thus, there are minimal, albeit significant, differences in B cells between cystic models. A full list of genes enriched in each B-cell cluster can be found in the Supplemental Material.

Figure 5.

The composition of B-cell clusters is different in CM aged and CM IR mice compared with one another and with noncystic controls. (A) UMAP of B cells from each experimental group. (B) Feature plot showing expression of key immune cell genes that were used to identify each cluster of B cells. (C) Quantification of the proportion of B cells in biological duplicates shown as a percentage of total B cells. (D and E) Heatmap showing the frequency of B-cell subtypes as a percentage of (D) all B cells or (E) CD45⁺ immune cells. Statistical significance was determined by two-way ANOVA, as described in Figure 3. */^#/^$P<0.05; **/^##/^$$P<0.01; ***/^###/^$$$P<0.001.

The Immune Cell Environment in Cystic Kidney Disease Shares Features with the Immune Cell Environment in AKI and UUO

To compare the immune cell populations in cystic kidney disease with other kidney diseases, we merged and integrated our immune-focused scRNAseq data with two other immune-focused scRNAseq studies from AKI²⁹ and UUO.³⁰ Subsequent analysis of the 48,934 cells from all three combined datasets (cystic, AKI, UUO) that passed quality control revealed the presence of 21 clusters of cells, including KRMs, T cells, B cells, and neutrophils (Supplemental Figure 14, A, B, and D). Quantification of cluster abundance reveals that each cell type was present across models, although the frequency of clusters differed between diseases (Supplemental Figure 14 C). For example, KRMs (C1qa, C1qc) were the most abundant cluster of immune cells in each model. Other interesting findings include: a significant increase in the frequency of CD4⁺ T cells and B cells in cystic disease compared with AKI and UUO; an increased proportion of Ccl2⁺ KRMs in AKI compared with cystic disease and UUO; and an increased frequency of Ly6c^hi Φ in UUO compared with cystic disease and AKI (Supplemental Figure 14C). These data suggest that cystic disease may be more heavily influenced by lymphoid cells whereas AKI and UUO phenotypes may be driven by myeloid cells.

Loss of Adaptive Immune Cells Reduces Cystic Disease in CM IR, but Not CM Aged, Mice

On the basis of our data, we hypothesized that adaptive immune cells may be involved in cystic disease. To test this hypothesis, we crossed CM mice to Rag1-deficient mice that lack all adaptive immune cells (T and B cells).⁴² Strikingly, our data indicate that deletion of all adaptive immune cells significantly attenuated cystic disease in CM IR mice, but had no effect on cyst number or severity in the CM aged model (Figure 6, A–F). These data suggest the involvement of adaptive immune cells and the overall mechanism of cyst formation may be different between rapid, injury-induced and slow, non-injured models of cystic disease. This also suggests that the T or B cells that are enriched in CM IR mice may affect cyst progression.

Figure 6.

Adaptive immune cells promote cyst formation in CM IR, but not CM aged, mice. (A) H&E image of CM IR Rag1 control and CM IR Rag1^−/− mice 56 days post IR injury. (B and C) Quantification of (B) cystic index or (C) cystic number in CM IR Rag1 control or CM IR Rag1^−/− mice 56 days post injury. (D) H&E image of CM aged Rag1 control or CM aged Rag1^−/− mice 9 months post tamoxifen induction. (E, F) Quantification of (E) cystic index or (F) cystic number in CM aged Rag1 control or CM aged Rag1^−/− mice 9 months post induction. Significance was determined using a t test. *P<0.05; **P<0.01.

As previous data indicate that the ability of immune cells to promote cystogenesis may be different between mice with mutations in the polycystin genes (Pkd1/Pkd2) versus mice with mutations in genes encoding other cilia-related proteins (Pkhd1, Cys1, Ift88),^{8,12,14,43,44} we also analyzed the severity of renal cysts in the Pkd1^RC/RC Rag1^−/− model of cystogenesis. Pkd1^RC/RC mice have a knock-in missense mutation in the Pkd1 gene resulting in reduced levels of functional PC1 protein and slowly progressing cystic disease.^23,45 In the C57Bl/6 background, Pkd1^RC/RC mice have slowly progressing disease (approximately 20% cystic index after 6 months) whereas in the BALB/c background the rate of cyst progression is greatly increased (approximately 40% cystic index after 3 months), thereby providing us with both slow and rapid models of cystogenesis in the absence of renal injury.²³ Strikingly, our data indicate that loss of adaptive immune cells did not affect kidney weight/body weight, cystic index, cystic size, or cystic number in 3- or 6-month-old C57BL/6 or BALB/c Pkd1^RC/RC Rag1^−/− mice compared with control mice (Pkd1^RC/RC Rag1^+/+, Supplemental Figure 15, A–E). The lack of rescue in Pkd1^RC/RC mice occurred despite the fact that these mice have increased numbers of both CD4⁺ and CD8⁺ T cells at these time points,²³ suggesting that the adaptive immune cells that specifically accumulate in response to injury, but not aging, may be modifying the rate of cyst growth.

NicheNet Identifies Potential Mechanisms through which T Cells Promote Cyst Progression in CM IR Mice

To understand how T and B cells from CM IR mice, but not CM aged mice, may be accelerating cyst growth, we utilized NicheNet, a software that identifies changes in ligand-receptor-gene regulatory networks in scRNAseq datasets.⁴⁶ This software provides a list of prioritized ligands produced by a “sender” cell and calculates their potential (regulatory potential) to drive the differential gene expression observed in a “receiver” cell, thus allowing us to identify ligands coming from CM IR T and B cells that may be driving the DEGs in CM IR epithelium, eventually resulting in accelerated cyst growth. For these analyses, we independently assigned T and B cells (clusters 3 and 5; Figure 1) as “senders” and the cilia dysfunctional epithelial clusters (S1/S2 proximal tubule epithelium, cluster 2; Figure 1; CD/LOH, cluster 4; Figure 1) as receivers in both CM IR and CM aged mice. Analyses of the data indicate that ligands expressed by CM IR T cells could possibly explain four of the 33 DEGs (12.1%) that were found in the CM IR epithelium (versus cont IR epithelium; Figure 7A) whereas CM IR B cells could possibly explain five out of the 33 DEGs (15.2%) observed in CM IR epithelium (Figure 7B). Our analyses indicate that CM aged T cells express ligands that could possibly explain 41 of the 212 DEGs (19.3%) observed in CM aged epithelium whereas CM aged B cells expressed ligands that could explain 42 of the 212 DEGs (19.8%) observed in the CM aged epithelium (Figure 7, C and D). Thus, T and B cells from both models would be expected to have some influence on cyst progression. However, on the basis of data showing that T and B cells only influence cystic disease in the CM IR model, we hypothesized that there may be some unique ligands coming from CM IR T and B cells that could explain why this model was affected by loss of adaptive immune cells whereas the CM aged model was not affected. Therefore, we analyzed ligand composition in both cell types and models and identified five ligands that were only expressed by CM IR T cells (Pkd1, Egf, Psen1, Gzmb, and Icam2) and two ligands that were only expressed by CM IR B cells (Adam17 and Egf) when compared with their CM aged counterparts (Figure 7, E and F). Thus, our analyses suggest that the interaction between these T- and B-cell ligands and the CM IR epithelium may explain why loss of adaptive immune cells rescues cystic disease in the CM IR model, but not the CM aged model of cystic disease.

Figure 7.

NicheNet analysis of T- and B-cell communication with the CM epithelium in CM IR and CM aged mice. (A and B) Top prioritized ligands expressed in (A) T cells or (B) B cells (senders) and a list of target genes that are differentially expressed in the CM epithelium (receivers) in CM IR mice in relation to cont IR mice. (C and D) Top prioritized ligands expressed by (C) T cells or (D) B cells (senders) and a list of target genes that are differentially expressed in the CM epithelium (receivers) in CM aged mice in relation to cont aged mice. (E) Venn diagram comparing the list of ligands that are expressed in T cells from CM IR mice with ligands expressed in T cells from CM aged mice. (F) Venn diagram comparing the list of ligands that are expressed in B cells from CM IR mice with ligands expressed in B cells from CM aged mice.

Because adaptive immune cells often function by altering MNP (macrophages, dendritic cells) activation, we also analyzed if T- and B-cell–derived ligands from both models could explain the DEGs found in MNPs from CM IR and CM aged mice (cluster 0, Figure 1). This analysis revealed that ligands expressed in CM IR T cells may explain 43 out of a possible 153 DEGs found in CM IR MNPs (28.1%; Figure 8A) whereas B-cell–expressed ligands may explain 44 out of 153 DEGs in CM IR MNPs (28.8%; Figure 8B). The data also show that ligands coming from CM aged T cells may explain 40 out of 136 (29.4%) of the DEGs observed in CM aged MNPs (Figure 8C). Likewise, ligands coming from CM aged B cells may explain 39 out of 136 (28.6%) of the DEGs observed in CM aged MNPs (Figure 8D). Once again, we analyzed what ligands were specifically expressed in CM IR T and B cells to identify a possible reason why this model has reduced cystic disease when adaptive immune cells are lost. This comparison revealed that CM IR T cells express 11 ligands (Ifng, Itga4, Plat, Pkd1, Sell, Egf, Rps19, Psen1, Aimp1, Manf, Gzmb) whereas CM IR B cells express nine ligands (Adam17, Itga4, Plat, Egf, Rps19, Hmgb2, Mif, Itgal, Fbrs) that are not found in T and B cells from CM aged mice (Figure 8, E and F). When we analyzed expression of these ligands in our single-cell immune atlas, we found that the majority of T- and B-cell–derived ligands identified by NicheNet were not uniquely expressed by T and B cells, with the exception of Gzmb and Ifng, which were only expressed by T cells (Figure 9, A–C). These data suggest that T cells are likely the cell type responsible for driving injury-accelerated cyst progression in CM IR mice, in agreement with our data showing that six of 13 T-cell clusters were different between cystic models whereas only one of six B-cell clusters were different between cystic models. This idea is further supported by our data showing that T cells enriched in CM IR mice (effector CD4⁺ T cells) have an activated phenotype whereas T cells enriched in CM aged mice (CD8⁺ T cells, CD4⁺ Tregs) have an exhausted phenotype. Likewise, the idea that activated T-cell–specific ligands (Ifng, Gzmb) influence cyst progression in CM IR mice in a MNP-dependent manner is supported by our pathway analysis data showing that MNPs from CM IR (KRMs, Ly6c^lo Φ) mice have an enrichment of genes associated with ion transport and epithelial cell differentiation (Supplemental Figure 7).

Figure 8.

NicheNet analysis of T- and B-cell communication with MNPs from CM IR and CM aged mice. (A and B) Top prioritized ligands expressed by (A) T cells or (B) B cells (senders) and a list of target genes that are differentially expressed in the MNPs (receivers) in CM IR mice in relation to cont IR mice. (C and D) Top prioritized ligands expressed by (C) T cells or (D) B cells (senders) and a list of target genes that are differentially expressed in the MNPs (receivers) in CM aged mice in relation to cont aged mice. (E) Venn diagram comparing the list of ligands that are expressed in T cells from CM IR mice with ligands expressed in T cells from CM aged mice. (F) Venn diagram comparing the list of ligands that are expressed in B cells from CM IR mice with ligands expressed in B cells from CM aged mice.

Figure 9.

Analysis of priority ligand expression in immune cells. (A–C) Violin plots showing expression of ligands identified by NicheNet in clusters of cells from the immune cell atlas (Figure 1). Data are broken down by (A) NicheNet ligands that were expressed by both T and B cells in CM IR mice, (B) NicheNet ligands expressed by just T cells in CM IR mice, and (C) NicheNet ligands expressed by just B cells in CM IR mice. (D) Quantification of Ifng mRNA counts (single-cell data) or Ifng protein expression (flow cytometry) in broad T-cell subtypes. (E) Violin plots showing expression of Ifng in clusters of T cells from Figure 4. Also shown is quantification of Ifng protein expression (flow cytometry) in clusters of T cells identified in Figure 4.

To identify the T cell that produced these ligands, we began by quantifying the total Ifng mRNA counts in broad groups of T cells (i.e., CD4⁺ T cells, CD8⁺ T cells, NKT cells, NK cells, ILCs, and others) on the basis of the T-cell clustering in Figure 4. We chose to focus this analysis on Ifng due to its high ligand priority score (Figure 8A) and its known role in influencing MNP function.^47,48 This analysis revealed that CD4⁺ and CD8⁺ T cells expressed the majority of Ifng (Figure 9D), which was confirmed at the protein level by flow cytometry (see Supplemental Figure 16A for gating strategy). To identify which specific T-cell subsets expressed these cytokines, we analyzed Ifng expression in T-cell clusters from Figure 4. Analyses of violin plots indicate that CD8⁺ T cells (cluster 0) and effector CD4⁺ T cells (cluster 2) expressed the greatest amount of Ifng (Figure 9E). Once again, we validated these data at the protein level via flow cytometry (Figure 9E, Supplemental Figure 16B). Of interest, renal injury increased Ifng mean fluorescence intensity in both CM and cont mice (Supplemental Figure 16C). These data, combined with data in Figure 4 showing that CM IR mice have an increased proportion of effector CD4⁺ T cells compared with CM aged mice, suggest that Ifng-expressing effector CD4⁺ T cells promote injury-induced cyst progression in CM IR mice by driving MNP activation.

Discussion

Using scRNAseq, our data indicate that the composition of T- and B-cell subsets is significantly different between CM IR and CM aged models of renal cystic disease. Despite the fact that both the CM IR and CM aged models of cystic disease had alterations in T- and B-cell subsets compared with each other and noncystic controls, genetic deletion of all adaptive immune cells (Rag1^−/− mice) only affected cystic disease in the CM IR model. Using NicheNet, we identify a candidate ligand produced by T cells from CM IR mice (Ifng) that may be responsible for the accelerated rate of cystic disease after injury. These data highlight the diversity of the immune environment in different models of cystic disease and suggest that the involvement of adaptive immune cells may be different depending on the model used, even when cystic disease is caused by a common underlying genetic mutation. Because this is the first comprehensive immune cell atlas of cystic kidney disease, we have created a user-friendly interface for researchers to query their gene of interest in our single-cell immune dataset (https://zimmerman-lab.shinyapps.io/cystic-immune-cell-dataset/).

We hypothesize that the reduced cyst severity observed in CM IR Rag1^−/− mice compared with CM IR Rag1 control mice is due to the production of injury-specific ligands (i.e., cytokines) that drive accelerated cystic disease. On the basis of the NicheNet data, we have identified a candidate ligand (Ifng) that is produced by CM IR T cells that can explain several of the DEGs found in CM IR MNPs. We also show that this candidate ligand is mainly produced by effector CD4⁺ T cells, at both the RNA and the protein level, suggesting that targeting this cell type, ligand, or downstream signaling pathway may restrict cystic disease in CM IR mice. Whether Pkd1^RC/RC mice on the C57BL/6 or BALB/c background also experience accelerated cystic disease and an enrichment of this T-cell subtype and ligand after injury is unknown. Overall, we propose that effector CD4⁺ T cells in CM IR mice produce Ifng to drive MNP activation and accelerated cystic disease, a hypothesis that is in line with our previous studies showing that resident macrophages, a type of MNP, promote cystic disease in CM IR mice.¹⁸

This study has several possible limitations. One possible limitation is the fact that the CM aged mice used for scRNAseq had a greater cystic size when compared with CM IR mice and an overall difference in histologic appearance. This occurred even though the average cystic index and cyst number between CM aged and CM IR mice was similar at their respective time points (Supplemental Figure 1). Thus, it appears that the animals randomly chosen for scRNAseq were on opposite ends of the cystic spectrum within each group. This may be due to the fact that all scRNAseq studies used female mice, which are known to have reduced tubular damage after IR injury.⁴⁹ Despite the differences in cystic appearance in the animals undergoing scRNAseq, both cystic models had large regions of the kidney with cystic and normal-appearing tissue, suggesting that differences in immune profiles may not be heavily influenced by the differences in cyst severity. Further, we were able to verify via flow cytometry the differences in cluster composition between groups in male and female mice. A final limitation is that our scRNAseq data are presented as a proportion of cells rather than as an absolute number of cells due to presorting using the pan-immune cell marker CD45. Overall, our immune cell atlas of cystic kidney disease identifies immune cell subsets that are associated with both the slow, aged model of cystic disease (CM aged) and the rapid, injury-accelerated model (CM IR) of cystogenesis.

Disclosures

A. Agarwal reports consultancy: Akebia Therapeutics—Expert Panel to review new therapeutics on the basis of the HIF pathway for AKI, medical advisory board of Creegh Pharmaceuticals, Dynamed—reviews content related to AKI for Dynamed and reviews updated materials prepared by the Dynamed editorial team for AKI topics; ownership interest: Creegh Pharmaceuticals, Goldilocks Therapeutics, Inc.; research funding: Genzyme/Sanofi Fabry Fellowship Award; honoraria: University of Toledo, University of Maryland; advisory or leadership role: Editorial Board of AJP Renal, Kidney International, and Laboratory Investigation; Advisory Board of Goldilocks Therapeutics, an NY-based company investigating delivery of drugs in the kidney using nanotechnology for acute and chronic kidney disease; Advisory Boards of Angion, Alpha Young, LLC, and Creegh Pharmaceuticals; and other interests or relationships: spouse is President for Women in Nephrology (2020–2022). U.K.B. Ahmed reports research funding: University of Oklahoma Health Sciences Center. N.M. Gonzalez reports ownership interest: Amazon; and other interests and relationships: Houston Methodist Research Institute, and Lupus Research Alliance. L.E. Harrington reports consultancy: Tentarix Biotherapeutics; and honoraria: Tentarix Biotherapeutics. K. Hopp reports consultancy: AceLink; research funding: AceLink; honoraria: Otsuka; patents or royalties: Mayo Clinic; and advisory or leadership role: Kidney360. M.L. Lang reports patents or royalties: M.L. Lang holds a patent unrelated to the manuscript under consideration: Clostridium difficile immunogenic compositions and methods of use. Inventors: Mark L. Lang and Jimmy D. Ballard (#50183P011US); M.L. Lang has received royalties from the University of Oklahoma as a contributor to an unrelated invention patented by other colleagues. M. Mrug reports consultancy: Caraway Therapeutics, Chinook, Goldilocks Therapeutics, Natera, Otsuka Corp., Reata, Sanofi; research funding: Chinook, Goldilocks Therapeutics, Otsuka Corp., Sanofi; honoraria: Chinook, Natera, Otsuka Corp., Reata, Sanofi; and advisory or leadership role: PKD Foundation, STAGED-PKD steering committee (Sanofi), Advisory Board (Carraway Therapeutics, Goldilocks Therapeutics, Santa Barbara Nutrients). C.J. Song reports employer: Amgen, and University of Southern California. All remaining authors have nothing to disclose.

Funding

These studies were supported in part by the following research grants: School of Medicine, University of Alabama at Birmingham (UAB) grant AMC21 (B.K. Yoder, M. Mrug, J.F. George); PKD Foundation grants 214g16a (B.K. Yoder), 826369 (K.A. Zimmmerman), and 216g18a (K. Hopp); National Institutes of Health (NIH) grants R01 DK115752 (B.K. Yoder), R01 DK097423 (M. Mrug), R01 NS57563 (E.N. Benveniste), K01DK114164 (K. Hopp), K01DK119375 (K.A. Zimmmerman), and 1-I01-BX002298 and 1I01BX004232-01A2 from the Office of Research and Development, Medical Research Service, US Department of Veterans Affairs (M. Mrug); by the Zell Family Foundation (K. Hopp); by the Detraz Endowed Research Fund in Polycystic Kidney Disease (M. Mrug); by the NIH T32 training grant in Basic Immunology and Immunologic disease 2T32AI007051-38 (K.A. Zimmmerman); by two Pilot and Feasibility Grants from the Baltimore PKD Center, 2P30DK090868 (K. Hopp and K.A. Zimmmerman); by a pilot grant from the UAB Hepato/Renal Fibrocystic Disease Core Center, 5P30DK074038 (K.A. Zimmmerman); by a seed grant from the Presbyterian Health Foundation (K.A. Zimmmerman); by a pilot grant from the Oklahoma Center for Microbial Pathogenesis and Immunity COBRE (1P20GM134973) to K.A. Zimmmerman; by a research grant from the Oklahoma Center for Adult Stem Cell Research (K.A. Zimmmerman); and by a Team Science Grant from the Presbyterian Health Foundation (K.A. Zimmmerman). The following NIH-funded cores provided services for this project: OUHSC COBRE grant P30GM122744, UAB Hepato/Renal Fibrocystic Disease Core Center grant P30-DK074038, UAB-UCSD O’Brien Center for Acute Kidney Injury Research grant P30-DK079337, and the UAB Comprehensive Flow Cytometry Core grants P30-AR048311 and P30-AI27667.

Author Contributions

C.J. Song was responsible for conceptualization, data curation, formal analysis, investigation, methodology, and validation, and reviewed and edited the manuscript; Z. Li was responsible for conceptualization, data curation, formal analysis, investigation, methodology, and validation, and reviewed and edited the manuscript; U.K.B. Ahmed was responsible for formal analysis, investigation, and methodology; wrote the original draft; and reviewed and edited the manuscript; S.J. Bland carried out conceptualization, data curation, formal analysis, project administration, validation, and visualization, and reviewed and edited the manuscript; A. Yashchenko was responsible for data curation, investigation, and methodology, and reviewed and edited the manuscript; S. Liu was responsible for data curation, investigation, and methodology, and reviewed and edited the manuscript; E.J. Aloria performed data curation, formal analysis, and investigation, and reviewed and edited the manuscript; J.M. Lever was responsible for conceptualization, data curation, formal analysis, investigation, and methodology, and reviewed and edited the manuscript; N.M. Gonzalez was responsible for conceptualization, data curation, formal analysis, and visualization, and reviewed and edited the manuscript; M.A. Bickel was responsible for data curation, investigation, and methodology, and reviewed and edited the manuscript; C.B. Giles was responsible for data curation, investigation, methodology, and software, and reviewed and edited the manuscript; C. Georgescu was responsible for data curation, investigation, resources, and software, and reviewed and edited the manuscript; J.D. Wren was responsible for data curation, methodology, resources, and software, and reviewed and edited the manuscript; M.L. Lang was responsible for methodology, validation, and visualization, and reviewed and edited the manuscript; E.N. Benveniste was responsible for methodology, project administration, resources, supervision, validation, and visualization, and reviewed and edited the manuscript; L.E. Harrington was responsible for resources, supervision, validation, and visualization, and reviewed and edited the manuscript; L. Tsiokas was responsible for resources, supervision, validation, and visualization, and reviewed and edited the manuscript; J.F. George was responsible for conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, resources, and supervision, and reviewed and edited the manuscript; K.L. Jones was responsible for data curation, resources, software, and supervision, and reviewed and edited the manuscript; D.K. Crossman was responsible for data curation, formal analysis, investigation, and methodology, and reviewed and edited the manuscript; A. Agarwal was responsible for funding acquisition, investigation, methodology, resources, and supervision, and reviewed and edited the manuscript; M. Mrug was responsible for conceptualization, formal analysis, funding acquisition, investigation, resources, supervision, validation, and visualization, and reviewed and edited the manuscript; B.K. Yoder was responsible for conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, and visualization; wrote the original draft; and reviewed and edited the manuscript; K. Hopp was responsible for conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, resources, supervision, validation, and visualization, wrote the original draft, and reviewed and edited the manuscript; and K.A. Zimmerman was responsible for conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, and visualization, wrote the original draft, and reviewed and edited the manuscript.

Data Sharing Statement

The sequencing data for these experiments have been deposited in the GEO database under accession number: GSE193528.

Supplemental Material

This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2021030278/-/DCSupplemental.

Supplemental Figure 1. CM aged and CM IR mice have similar cystic indices and cyst number at the time points chosen for scRNAseq.

Supplemental Figure 2. CAP digestion increases cell survival while reducing expression of stress response genes compared with standard 37°C digest.

Supplemental Figure 3. Heatmap showing the top five DEGs in each cluster of cells from the whole-kidney single-cell atlas.

Supplemental Figure 4. Heatmap showing the top five DEGs in each cluster of cells from the immune single-cell atlas.

Supplemental Figure 5. Heatmap showing the top five DEGs in each cluster of MNPs.

Supplemental Figure 6. CM aged mice have increased numbers of MNPs localized adjacent to cysts compared with CM IR mice.

Supplemental Figure 7. Pathway analysis of MNP clusters.

Supplemental Figure 8. Heatmap showing the top five DEGs in each cluster of T cells.

Supplemental Figure 9. Naïve/central memory T cells express transcripts associated with naïve and antigen-experienced T cells.

Supplemental Figure 10. CM aged mice have increased numbers of T cells compared with all other groups.

Supplemental Figure 11. Only a small fraction of CD4⁺ Tregs express Foxp3 by single-cell RNA sequencing.

Supplemental Figure 12. Analysis of genes associated with T-cell activation or exhaustion.

Supplemental Figure 13. Heatmap showing the top five DEGs in each cluster of B cells.

Supplemental Figure 14. Integrated single-cell data comparing immune cell clusters in cystic disease, AKI, and UUO.

Supplemental Figure 15. Loss of adaptive immune cells does not affect cyst formation in C57BL/6 (slow) or BALB/c (rapid) Pkd1^RC/RC mice.

Supplemental Figure 16. Gating strategy used to identify Ifng-expressing cells, representative FACS plots showing expression of Ifng in T-cell clusters 0–7, and quantification of Ifng MFI in each experimental condition.

Supplemental Material. Dataset 1.

Supplemental Material. Dataset 2.