Multiomics profiling of mouse polycystic kidney disease progression at a single-cell resolution

Significance

We generated a single-nucleus multiomic atlas to characterize cell-type-specific molecular changes in a mouse polycystic kidney disease (PKD) model during disease progression. By comparatively analyzing with a human PKD dataset we previously generated, we identified disease-specific cell states that may contribute to cystogenesis and cyst growth. Our analysis sheds light on the role of failed-repair epithelia in cyst growth as well as a shared cyst lining cell marker GPRC5A in human and mouse PKD. Our dataset and analytic strategy contribute to better understanding of PKD mechanism and ultimately toward development of better therapeutic approaches.

Abstract

Autosomal dominant polycystic kidney disease (ADPKD) is the most common hereditary kidney disease and causes significant morbidity, ultimately leading to kidney failure. PKD pathogenesis is characterized by complex and dynamic alterations in multiple cell types during disease progression, hampering a deeper understanding of disease mechanism and the development of therapeutic approaches. Here, we generate a single-nucleus multimodal atlas of an orthologous mouse PKD model at early, mid, and late timepoints, consisting of 125,434 single-nucleus transcriptomic and epigenetic multiomes. We catalog differentially expressed genes and activated epigenetic regions in each cell type during PKD progression, characterizing cell-type-specific responses to Pkd1 deletion. We describe heterogeneous, atypical collecting duct cells as well as proximal tubular cells that constitute cyst epithelia in PKD. The transcriptional regulation of the cyst lining cell marker GPRC5A is conserved between mouse and human PKD cystic epithelia, suggesting shared gene regulatory pathways. Our single-nucleus multiomic analysis of mouse PKD provides a foundation to understand the earliest changes molecular deregulation in a mouse model of PKD at a single-cell resolution.

Autosomal dominant polycystic kidney disease (ADPKD) is a monogenic kidney disease affecting approximately 0.1 to 0.2% of the population worldwide (1). The cysts in ADPKD kidneys are induced by a mutation of the PKD1 or PKD2 gene, and following deregulation of numerous signaling pathways associated with cAMP response, mammalian target of rapamycin complex (mTORC), WNT, and Hippo pathway, among others (1, 2). In addition to disturbed signaling pathways in cystic epithelia, there is also metabolic deregulation of glycolysis, oxidative phosphorylation, and lipid metabolism (3). Subsequently, growing cysts compress and injure the adjacent kidney parenchyma and tubular structures, leading to chronic kidney disease (CKD) and ultimately to kidney failure. The vasopressin receptor antagonist tolvaptan slows cyst growth by decreasing cAMP signaling, although this therapy is associated with polyuria, limiting its wide use (4). Therefore, better understanding of disease mechanisms and development of new therapeutic approaches to ADPKD is of paramount importance. However, the dynamic cellular and molecular complexities in ADPKD progression as well as extremely limited access to early stage human ADPKD samples have hampered a deeper understanding of cellular and molecular mechanisms.

Recent evolution in single-cell analysis has advanced our knowledge about cellular heterogeneity in healthy and diseased kidneys (5–9). We and others have generated multimodal single-cell atlases of human and mouse kidneys, describing previously unrecognized cellular heterogeneity that revealed a proinflammatory, profibrotic proximal tubular cell type—failed-repair proximal tubular cells (FR-PTC) (8–12). We have applied our single-nucleus multiomics approach to the human ADPKD kidneys, and characterized ADPKD-specific cell states and their molecular signatures (13). We have identified GPRC5A as a cyst lining cell marker and its transcriptional deregulation driven by ADPKD-specific epigenetic remodeling. We also revealed cellular heterogeneity of failed-repair proximal tubular cells expanded in human ADPKD kidneys, which may promote PKD progression by modifying the microenvironment (13). Nevertheless, our previous study was limited by the end-stage nature of the ADPKD samples. The cell-type-specific molecular signatures during early stages of PKD progression have remained elusive.

To fill this knowledge gap, we have performed simultaneous snRNA-seq and snATAC-seq on a mouse polycystic kidney disease (PKD) model over time following Pkd1 ablation, to comprehensively characterize the cell states and their dynamics during PKD progression. This approach has allowed us to describe the cell-type-specific molecular signatures at an early and middle stage of PKD progression. The atypical collecting duct principal cells as well as failed-repair proximal tubular cells were identified in the PKD mouse kidneys. Comparative analysis of human ADPKD and mouse PKD model atlas sheds light on GPRC5A as a shared cyst lining cell marker, with its transcriptional regulation has been conserved between human and mouse PKD. Our analysis also underscores cell-type-specificity of a molecular response to Pkd1 deletion during PKD progression, rather than a shared response to Pkd1 deficiency across tubular compartments. To maximize the usefulness of our dataset in the research community, an interactive data visualization tool for this multiomics dataset is available (http://humphreyslab.com/SingleCell/). This single-nucleus multiomics atlas as well as our findings through the analysis will be the foundation to better understand the mechanism of PKD and develop therapeutic approaches.

Results

Single-Nucleus Multimodal Profiling of Slow-Onset PKD Model Mouse Kidneys.

To comprehensively understand the cell-type-specific molecular mechanism of PKD progression, we applied single-nucleus multiomics to an orthologous mouse model of ADPKD—the Pkd1^fl/fl; Pax8-rtTA; Tet-O-Cre mouse that utilizes the Pax8 promoter to drive expression of a reverse tetracycline-dependent transactivator (rtTA) in all renal tubular compartments including proximal and distal tubules as well as collecting ducts (14, 15). We induced Pkd1 deletion in the tubular cells of these male mice from postnatal day 27 (P27) to 56 (P56) with doxycycline injection. The littermate male controls lacked either the Pax8-rtTA (P66) or the Tet-O-Cre (P100, P130, SI Appendix, Table S1) but were treated with doxycycline in the same manner. The kidneys were harvested and snap-frozen at postnatal day 66 (P66), 100 (P100), or 130 (P130) (Fig. 1A and SI Appendix, Table S1). The PKD mouse kidneys demonstrated a gradual increase in the ratio of kidney weight to body weight along the progression of PKD (SI Appendix, Fig. S1A). Immunostaining of tubular lineage markers indicated that the cysts in this PKD model kidneys derived from multiple tubular compartments including lotus tetragonolobus lectin (LTL)+ proximal tubules, uromodulin (UMOD)+ thick ascending limbs of Henle loop and lectin Dolichos biflorus agglutinin (DBA)+ collecting ducts (Fig. 1B). Of note, DBA+ collecting ducts already mildly dilated as early as at P66 in corticomedullary junction (Fig. 1B), suggesting ongoing cystic transformation of these cells. In contrast, tubules in the cortex did not show apparent morphologic changes at P66 (Fig. 1B) except occasional small cysts (SI Appendix, Fig. S1B). Both cortical and corticomedullary regions demonstrated numerous cysts at P100, and the cystic lesions were most prominent at P130 (SI Appendix, Fig. S1B). These findings suggest that our multiomics atlas includes the major PKD stages from a precystic to late cystic stages. We generated and sequenced multimodal single-nucleus libraries on a total of 16 mouse kidney samples (8 PKD and 8 control mouse kidneys) along a time course following Pkd1 deletion (Fig. 1A).

Fig. 1.

Single-nucleus multiomics profiling for mouse PKD kidneys. (A) Overview of experimental strategy. Single-nucleus multiomics atlas was generated from PKD model mouse kidneys and littermate control kidneys along a time course (three time points; postnatal day 66 (P66), 100 (P100), or 130 (P130) after Pkd1 deletion (n = 2 to 3 pairs for each time point). (B) Representative immunofluorescence images of lotus tetragonolobus lectin (LTL, green), uromodulin (UMOD, red), and lectin Dolichos biflorus agglutinin (DBA, white) in the cortex (Left) or corticomedullary junction (Right) in the PKD kidneys at P66 (n = 2), P100 (n = 3), or P130 (n = 3), or control kidneys (n = 3). Scale bar indicates 100 µm. (C) UMAP plot of the integrated single-nucleus multiomics dataset with weighted nearest neighbor (wnn) clustering. Clusters were annotated by lineage marker gene expression. PTS1/S2/S3, proximal tubule S1/S2/S3 segments; FRPTC_PEC, failed-repair proximal tubular cells and parietal epithelial cells; DTL1/DTL2/ATL, descending thin limb 1/2 and ascending thin limb of Henle’s loop; TAL, thick ascending limb of Henle’s loop; DCT, distal convoluted tubule; CNT, connecting tubule; PC1/2, principle cells 1/2; URO, uroepithelial cells; ICA, Type A intercalated cells; ICB, Type B intercalated cells; PODO, podocytes; ENDO, endothelial cells; FIB, fibroblasts; Myel, myeoloid cells; FAT, adipocyte.

After sequencing (SI Appendix, Table S2), batch quality control (QC) filtering and preprocessing of sequencing data (SI Appendix, Fig. S2), the individual multiomics datasets were integrated with Seurat (16). We obtained a total of 125,434 nuclei (59,089 nuclei from PKD and 66,345 nuclei from control). Both control and PKD data show similar numbers of unique genes and transcripts per nucleus (SI Appendix, Fig. S3), indicating successful generation of high-quality dataset from the samples including those with advanced PKD (SI Appendix, Fig. S1). Following dimensional reduction on both transcriptomic and epigenomic data with batch effect correction with Harmony (17), a Weighted Nearest Neighbor (WNN) graph was calculated and visualized on UMAP space (Fig. 1C). The cell types were annotated by marker gene expression (SI Appendix, Fig. S4A and Dataset S1). The lineage marker expression was largely maintained during progression of PKD (SI Appendix, Fig. S4 B–E), consistent with our previous finding that the cell type marker gene expression was preserved in human advanced ADPKD kidneys (13). All major cell types were identified in both control and PKD mouse kidneys (SI Appendix, Fig. S5 and Table S3) but a cluster representing failed-repair proximal tubular cells (FR-PTC) gradually expanded in PKD along the time course (SI Appendix, Fig. S5 and Table S3). Although immune cell infiltration as well as interstitial fibrosis have been observed in advanced PKD, the increase in the numbers of those cell types was not apparent in our dataset, presumably due to dissociation bias (18).

Next, we cataloged differentially expressed genes in PKD for each cell type at each time point (Datasets S2–S4). Surprisingly, there were only a few differentially regulated genes shared among tubular cell types (PT, TAL, PC1/2) in PKD (P66, Kap, and Fkbp5; P100, Nox4, and Gm42418; P130, no shared genes) (SI Appendix, Fig. S6A). This finding suggests that the molecular response to Pkd1 deletion is cell-type-specific, although the consequences—cystogenesis and cyst growth—are shared across tubular compartments. This observation was in line with the possibility that the mechanism of cystogenesis and cyst growth may be cell-type-specific, underscoring the importance of single-cell analysis to dissect PKD mechanisms.

Cell-Type-Specific Chromatin Accessibility and Motif Availability in PKD.

The cell types annotated through marker gene expressions (SI Appendix, Fig. S4A) demonstrated consistent cell-type-specific chromatin accessibility to marker genes (SI Appendix, Fig. S7A). Dimensional reduction and projection of nuclei on UMAP with snATAC-seq peaks also indicates that clustering based on chromatin accessibility is largely in line with that with transcriptomic data (SI Appendix, Fig. S7) or integrated approach (Fig. 1C). In agreement with the transcriptional data, there was only one shared differentially accessible region (DAR) among PT, TAL, and PC in PKD at P66 (chr2:24478513-24480280, ~3 kbp 5’-distal to Pax8 gene) and no shared DAR at P100 or P130 (Datasets S5–S7 and SI Appendix, Fig. S6B). The shared DAR in the Pax8 locus at P66 PKD simply reflects the Pax8-rtTA transgene which was inserted to chromosome 8. The Pax8-rtTA transgene contains 4.3 kbp of 5’ upstream regulatory sequence of Pax8 gene (14), so the shared DAR is included on the transgene. Since the aligner does not distinguish between the original and cloned region, the transgene theoretically increases the number of the fragments derived from that region. The actually measured accessibility also reflects the interaction of the transgene with the cis-regulatory elements on the inserted genomic region. These findings emphasize the cell-type-specificity of epigenetic changes induced by Pkd1 deletion. Next, we performed transcription factor binding motif enrichment analysis on accessible chromatin regions at a single-nucleus level (SI Appendix, Fig. S8A) with ChromVar (19). The transcription factors enriched in certain cell types were often associated with those lineage specifications (SI Appendix, Fig. S8B) (8).

Next, we compared the enriched transcription factor binding motifs between PKD and control kidneys (Datasets S8–S10). One of the most enriched transcription factor motifs in late-stage PKD (P130, Dataset S10) in multiple tubular cell types (PT, TAL, PC1, PC2) was Transcription factor EB (TFEB), a master regulator of lysosomal function and autophagy (SI Appendix, Fig. S8C). Nuclear localization and noncanonical activation of TFEB in cystic epithelia were previously shown in the mouse and human polycystic kidneys (20). TFEB was also found to drive mTOR hyperactivation and renal cell carcinoma in tuberous sclerosis complex (21). Constitutive activation of TFEB and its family transcription factor TFE3 promote cell proliferation in several types of cancers in addition to renal cell carcinoma (22). These lines of evidence may implicate TFEB in aberrant growth of cyst epithelia in PKD. AP-1 transcription factor binding motifs were also enriched in tubular cell types in PKD (SI Appendix, Fig. S8D), presumably reflecting a cohort of proinflammatory signaling pathways activated in advanced PKD (23). These findings collectively suggest that alteration in the chromatin accessibility landscape was associated with differential transcription factor availability during PKD progression, potentially promoting activation of cystic epithelia and inflammation in PKD.

Deregulation of Molecular Pathways in PKD Collecting Duct Cells.

We have previously characterized cellular heterogeneity of collecting duct cells in human ADPKD, revealing activation of inflammatory pathways as well as metabolic deregulation in cystic cell types (13). The mouse PKD kidneys also have numerous DBA+ distal nephron cysts (Fig. 1B). To characterize the mouse PKD distal nephron cells, we performed subclustering analysis for the distal nephron cell types; connecting tubules (CNT), principal cells (PC1 and PC2), and uroepithelial cells (URO) in the whole dataset (Fig. 1C), identifying 8 subpopulations (Fig. 2A). These subtypes express unique marker genes (Fig. 2B and Dataset S11). CNT subtypes express Ptprd, Slc8a1, and Calb1 as shared marker genes. The cortical principal cells (cPC1/2) and medullary principal cells (mPC1/2) express Apq2, which encodes aquaporin2 (AQP2). All these subtypes were detected in both PKD and control at each time point (SI Appendix, Fig. S9). Among the subtypes of the distal nephron cells, CNT3 shows highest frequency of PKD-derived cells (SI Appendix, Fig. S10A). Further subclustering of CNT3 identified almost PKD-specific (~96%) subpopulation (cluster1; CNT3_Pkd), suggesting the distinctive CNT cell state driven by the Pkd1 loss (SI Appendix, Fig. S10 B and C). CNT3_Pkd3 expressed unique genes including Dcdc2a and Creb5 (SI Appendix, Fig. S10D), which have been shown to be up-regulated in the failed-repair PTC (FR-PTC) emerging after epithelial injury(10, 24). CNT3_Pkd3 lacks expression of other FR-PTC signature genes like Havcr1 or Vcam1, although this subpopulation also highly up-regulated Lcn2 expression, which was shown to be another epithelial injury marker (25) (SI Appendix, Fig. S10D). These observations indicate that CNT_Pkd3 may be the dedifferentiated CNT cell state induced by the loss of Pkd1, resembling FR-PTC.

Fig. 2.

Heterogeneity of collecting duct principal cells in mouse PKD. (A) Subclustering of distal nephron clusters (CNT, PC1, PC2, and URO) on the UMAP plot, colored by unsupervised clustering (Left) or genotype (Right). cPC, cortical principal cells; mPC, medullary principal cells. (B) Dot plot showing expressions of the genes enriched in each of the subtypes. The diameter of the dot corresponds to the proportion of cells expressing the indicated gene and the density of the dot corresponds to average expression. (C) Representative immunofluorescence images of CALB1 (green) and AQP2 (red) in the PKD kidneys at P130 (n = 3). Arrowheads indicate AQP2+ cyst lining. Arrows mark AQP2+/CALB1+ cyst lining. Scale bar indicates 50 µm. (D) Distal nephron subtypes in the human advanced ADPKD data are label-transferred from those in the mouse PKD data, and the frequencies of predicted mouse subtypes in each human subtype are shown on the heatmap. (E) Heatmap showing enrichment of gene expressions of the hallmark gene sets among cPC2 and mPC1 clusters at each time point. The pathways associated with DNA damage response are in blue characters and those associated with metabolic regulation are in red characters. (F) Volcano plot showing differentially expressed genes in mPC1 of PKD mice compared to that of control at P66 (Upper) or P130 (Middle). The x-axis represents the log fold change, and the y-axis represents the P values. Dot plot (Lower) showing representative differentially expressed genes among control and PKD mPC1 at P66 and P130. The diameter of the dot corresponds to the proportion of cells expressing the indicated gene and the density of the dot corresponds to average expression. (G) Representative immunofluorescence images of DBA (green) and TMSB4X (red) in the PKD kidneys at P130 (n = 3). An arrow indicates a glomerulus. Scale bar indicates 50 µm.

Immunostaining of CALB1 and AQP2 proteins in the late-stage PKD kidneys revealed that most of cystic CALB1+ cells were also AQP2+/DBA+, and that CALB1+ tubules lacking AQP2 expression rarely contributed to cyst formation (Fig. 2C and SI Appendix, Fig. S11), suggesting that the majority of DBA+ cysts observed in PKD kidneys were originated from collecting ducts. To compare the mouse PKD principal cells with human ADPKD principal cells (13), we predict the corresponding subtypes of mouse PKD distal nephron in the human ADPKD distal nephron cells. Label transfer of subtype annotations from mouse PKD to human ADPKD indicates that the majority of human ADPKD cystic collecting duct cells (PKD-CDC)1 were related to the mPC2 population, and human PKD-CDC2 cells were associated with cPC1 cluster (Fig. 2D) (13). The gene expression signature of PKD-CDC1/2 in human ADPKD was associated with inflammation, hypoxia, and cellular senescence, with upregulation of cyst lining cell marker GPRC5A (13). The cell-type prediction suggests that the cellular heterogeneity of collecting duct cells is conserved between humans and mice.

To better understand the molecular signature of mPC2 and cPC1 subtypes in mouse PKD, we applied gene set enrichment analysis on cPC2 and mPC1 subtypes (Fig. 2E). Interestingly, the differentially regulated signaling pathways in these subpopulations were dynamically changed during PKD progression. We observed activation of a cohort of genes associated DNA damage response (Fig. 2E, pathways in blue characters) and metabolic alterations (Fig. 2E, pathways in red characters) at P66 in mPC1, implicating activation of these pathways in early morphological changes in the collecting duct (Fig. 1B). In contrast, gene expression associated with inflammatory pathways are up-regulated in mPC2 at the late-stage PKD samples (P130), presumably reflecting inflammation secondary to tissue destruction by large cysts as well as recruitment of immune cells (Fig. 2E). This observation in late-stage PKD was also in line with activation of inflammatory pathways in PKD-CDC1/2 in human advanced ADPKD (13). Consistent with dynamic alterations of gene set enrichment in mPC1, the differentially expressed genes in mPC1 (Fig. 2F and Dataset S12) as well as cPC2 (SI Appendix, Fig. S12) are disease-stage-specific (Fig. 2F). Among the differentially up-regulated genes in the late-stage mPC2, the most up-regulated gene was Tmsb4x that encodes thymosin beta 4 (TMSB4X) protein, which is an actin sequestering protein playing a role in regulation of cytoskeleton (26, 27). TMSB4X has been shown to be beneficial in diverse pathological conditions including myocardial infarction, neuronal damage, and diabetic nephropathy (26, 27). In contrast, TMSB4X has been found to promote tumor progression and to be associated with poor prognosis in various cancers (26). Immunostaining analysis identified expression of TMSB4X in DBA+ cysts (Fig. 2G), implicating TMSB4X in cyst growth in advanced PKD. Tmsb4x is broadly up-regulated in multiple cell types including parenchymal cell types including endothelial cells and immune cells (SI Appendix, Fig. S13), in line with interstitial staining of TMSB4X in addition to cyst lining cells (Fig. 2G). The relatively high level of Tmsb4x expression in podocytes was also consistent with previous study (Fig. 2G and SI Appendix, Fig. S13) (27). These findings collectively suggest that heterogeneous principal cell states show dynamic alterations in their molecular signatures and signaling pathways during PKD progression.

Heterogeneous Failed-Repair Proximal Tubular Cells in Cystic Epithelia of Mouse PKD.

We and others have described Vcam1-expressing failed-repair proximal tubular cells (FR-PTC) after kidney injury in mice (10–12, 28) and humans (8, 9, 13, 29). This subpopulation is characterized by proinflammatory, profibrotic gene expression signature, implicated in progression of kidney diseases (13, 29). Of note, the frequency of FR-PTC in advanced ADPKD was markedly increased, replacing normal PTC (13), suggestive of its significant role in ADPKD progression. Furthermore, we also revealed the heterogeneity of FR-PTC in human ADPKD kidneys (13).

We also observed expansion of FR-PTC cluster during mouse PKD progression (Fig. 1C and SI Appendix, Fig. S5 and Table S3). To understand the heterogeneity of FR-PTC in PKD mouse kidneys, we performed subclustering analysis on the FRPTC_PEC cluster (Fig. 1C), identifying 4 epithelial subtypes besides PEC (Fig. 3A and Dataset S13). Cluster 0 express normal PT lineage markers, although they are down-regulated compared to normal epithelia (Fig. 3B), suggesting that they may be a transitional state (Transitional PTC). FR-PTC1 (Cluster 2) highly express previously described FR-PTC marker genes; Havcr1 and Vcam1 (Fig. 3 B and C). In agreement with this observation, Vcam1 gene promoter is more accessible in FR-PTC1 (Fig. 3C) compared to other PT lineages. Hnf4a expression was down-regulated in FR-PTC subtypes (Fig. 3D), while Cdh6 expression was up-regulated in FR-PTC2/3 (Cluster 1/3) (Fig. 3D). Immunostaining analysis of PKD kidneys along the time course revealed a gradual increase in the abundance of VCAM1+ cells during PKD progression (Fig. 3 E and F). The abundance of VCAM1+ cells in cystic region was much more than that in noncystic or mildly cystic region in the same kidney in both P100 and P130 PKD kidneys (Fig. 3F), suggesting a skewed distribution of VCAM1+ cells in cystic lesions. VCAM1 protein expression was mutually exclusive to LTL staining even in the same cysts (Fig. 3F), in line with the previous finding indicating that VCAM1 is a marker for dedifferentiation (8, 10, 13). Furthermore, cysts with VCAM1+ epithelia often acquire epithelial lining lacking both LTL and VCAM1 (Fig. 3F). Consistent with this observation, FR-PTC2/3 subsets demonstrated much lower levels of Vcam1 expression compared to FR-PTC1 (Fig. 3 B and C). FR-PTC2 and FR-PTC3 demonstrated similar gene expression signatures (Fig. 3B). However, the general chromatin accessibility of FR-PTC3 was limited compared to other subtypes (SI Appendix, Fig. S14). The poor chromatin accessibility in FR-PTC3 may be due to extensive chromatin condensation.

Fig. 3.

Heterogeneity of failed-repair proximal tubular cells in mouse PKD. (A) Subclustering of FRPTC_PEC cluster on the UMAP plot. (B) Dot plot showing expressions of the genes enriched in each of the subtypes. The diameter of the dot corresponds to the proportion of cells expressing the indicated gene and the density of the dot corresponds to average expression. (C) UMAP plot showing a gene expression level of Vcam1 (Left) and coverage plot showing accessible regions around Vcam1 promoter in each subtype is also shown (Right). The color scale represents a normalized log-fold-change (LFC, Left). The scale bar indicates 2 kbp (Right). (D) UMAP plot displaying Hnf4a (Left) or Cdh6 (Right) expression. The color scale represents LFC. (E) Representative immunofluorescence images of LTL (green) and VCAM1 (red) in the PKD or control kidneys at P66 (n = 2). Scale bar indicates 50 µm. (F) Representative immunofluorescence images of LTL (green) and VCAM1 (red) in the mildly or severely cystic regions in PKD kidneys at P100 or P130 (n = 3). The arrow indicates VCAM1+ PTC. Arrowheads mark cystic epithelial lining losing both VCAM1 and LTL staining. Scale bar indicates 50 µm. (G) Dot plot showing expressions of the differentially expressed genes among different time points in FR-PTC1 (Left) or FR-PTC2 (Right). The diameter of the dot corresponds to the proportion of cells expressing the indicated gene and the density of the dot corresponds to average expression. Spp1 expression is up-regulated at P130 in both subtypes (red). (H) Representative immunofluorescence images of VCAM1 (green), SPP1 (red), and LTL (white) in the PKD kidneys at P130 (n = 3). Scale bar indicates 50 µm.

We identified presumptive PEC cluster abundantly expressing Ncam1 (Fig. 3B and SI Appendix, Fig. S15A), which has been found to be specifically expressed in PEC at a protein level in the healthy adult mouse kidneys (30). Given high-level expressions of Ncam1 as well as another PEC marker: Cldn1 in this population of the control kidneys (SI Appendix, Fig. S15A), we have annotated this as PEC. However, NCAM1 as well as CLDN1 protein has been shown to be up-regulated in injured tubular epithelia (31, 32), in agreement with the upregulation of these genes in FR-PTC subtypes in our dataset (SI Appendix, Fig. S15B). Therefore, there is a remaining possibility that a FR-PTC subtype may constitute a part of this cluster especially in PKD. Future spatial omics approach to mouse PKD will further our understanding of PEC-specific gene expression signature.

Despite a lack of Vcam1 expression, FR-PTC2/3 subtypes shared the expression of Cftr, Sema5a, and Spp1 with FR-PTC1 (Fig. 3B). Interestingly, Cftr expression was specifically up-regulated in FR-PTCs in PKD (SI Appendix, Figs. S16 and S17A). The Cftr gene encodes cystic fibrosis transmembrane conductance regulator (CFTR) protein, which has been implicated in fluid accumulation in cysts (33). The gene expression pattern in each FR-PTC subtype also changed dynamically during PKD progression (Fig. 3G), which was also observed in collecting duct subpopulations (Fig. 2). The genes with dynamically changed expression during PKD progression include Spp1 and Cftr in both FR-PTC1 and FR-PTC2 (Fig. 3G). Spp1 encodes a pleiotropic glycoprotein; osteopontin (SPP1), which has been shown to be involved in inflammation, angiogenesis, and apoptosis in pathologic conditions including ADPKD (34). Spp1 expression was widely detected and up-regulated in other tubular and nontubular cell types in PKD kidneys (SI Appendix, Fig. S17B), consistent with a recent line of evidence suggesting that Spp1 expression in the pericystic endothelium in PKD (35). We observed SPP1 protein localized in cystic epithelia with PT lineage marker expression (Fig. 3H) in PKD. In contrast, SPP1 protein was stained in the distal nephron tubules in the cortex as well as more notably in the uroepithelial cells on the papilla (SI Appendix, Fig. S18) in control, largely consistent with our snRNA-seq analysis (SI Appendix, Fig. S17B). SPP1 expression in cyst epithelia (Fig. 3 B and H) may alter surrounding microenvironment, potentially promoting PKD progression.

Next, we asked whether the heterogeneity of FR-PTC in mouse PKD is conserved in human ADPKD. We have previously generated single-nucleus transcriptomic atlas for human advanced ADPKD kidneys (13). We performed subclustering analysis on all ADPKD PT cells with FR-PTC alone from healthy kidneys to characterize PT cells in human ADPKD kidneys, identifying 4 FR-PTC subtypes (PT1-PT4) in our previous study (13). Both FR-PTC in healthy kidneys as well as ADPKD kidneys include PT1/2, while PT3/4 were mostly ADPKD-specific (SI Appendix, Fig. S19A). PT3 was characterized by the enrichment of TGFβ signaling pathway gene expressions (13) as well as lower VCAM1 expression (SI Appendix, Fig. S19B).

Label transfer of these FR-PTC subtype annotations from mouse PKD (Fig. 4A) to human ADPKD FR-PTC predicted that the gene expression signature of mouse FR-PTC2 was closest to that of human ADPKD PT3 subtype (Fig. 4A), which was characterized by a low expression level of VCAM1 (13). This finding indicates conservation of cellular heterogeneity in FR-PTC between human and mouse PKD. Together, these observations implicate heterogeneous FR-PTC subtypes in PKD progression as components of cystic epithelia. Furthermore, the gene expression signature of each FR-PTC subtype is dynamically altered during PKD progression (Fig. 3G).

Fig. 4.

Differentially activated molecular signaling pathways among PT subtypes in PKD. (A) The proximal tubular cell subtypes of human advanced ADPKD data are label-transferred from FR-PTC subtypes in mouse PKD data, and the frequencies of predicted mouse subtypes in each human subtype are shown on the heatmap. (B) Frequency of each PT subtype among whole PTC for each time point in PKD or control data. (C) Heatmap showing relative transcription factor binding motif enrichment among PT subtypes in snATAC-seq. Most enriched transcription binding motifs in each subtype are shown. (D) Violin plots displaying relative motif enrichment (chromVAR score) among PTC for TEAD3 (MA0808.1) (E) Violin plots displaying relative gene set enrichment among PTC for Hippo pathway genes (REACTOME-SIGNALING-BY-HIPPO).

Atypical Failed-Repair Proximal Tubular Cells Emerging after PKD1 Inactivation.

The frequency of FR-PTC1 subtype increased during PKD progression in the dataset (Fig. 4B and SI Appendix, Fig. S21A), indicating the accumulation of VCAM1+ cells in PKD kidneys. The important caveat is that there is often a discrepancy between cell type frequencies measured by a single-cell and histological approach, probably due to the artifact induced by tissue dissociation and nuclei isolation (5, 13, 18). Indeed, we did not observe any VCAM1+ cells at P66 PKD as well as control kidney sections (Fig. 3E), although a small percentage of FR-PTC1 has been already detected in the P66 kidneys in the dataset (Fig. 4B and SI Appendix, Fig. S21A). Furthermore, there is heterogeneity of cyst burden ranging from noncystic or mildly cystic to severely cystic regions even in the same PKD kidney, preventing accurate histological quantification in the cystic kidneys. Nevertheless, we have confirmed the gradual increase in the abundance of VCAM1+ cells during PKD progression by immunostaining (Fig. 3F), although the abundance of VCAM1+ cells in cystic region was much more than that in mildly cystic region in the same kidney in both P100 and P130 PKD kidneys (Fig. 3F), in agreement with the view that VCAM1+ cells were associated with cysts. In contrast, the frequency of FR-PTC2 was relatively steady during PKD progression, suggesting that FR-PTC2 may be a reversible cell state. FR-PTC2 specifically up-regulated the expression of Cdh6 encoding cadherin 6 (CDH6) (Fig. 3 B and D). Cdh6 expression has been shown to be up-regulated among the PT progenitors during development as well as PT following epithelial injury in adult kidneys (11, 28, 36). Interestingly, Cdh6 reexpression was also recently found to be associated with regenerative PT cell states following epithelial injury and deregulation of cell polarity (37, 38) in adult mice. These studies are in line with the notion that FR-PTC2 may be the dedifferentiated cell state induced by Pkd1 deletion. In contrast, CDH6 expression level in PT3, which was closest to mouse FR-PTC2 in the label-transfer prediction, was similar to that in PT1 in human ADPKD kidneys (SI Appendix, Fig. S19B). Furthermore, CDH6 mRNA as well as protein was detected even in the healthy PT cells of human healthy kidneys, although its mRNA expression level was higher in FR-PTC (SI Appendix, Fig. S20). These observations suggest that CDH6 expression may be less specific as a dedifferentiated PT marker in human kidneys compared to mouse kidneys. Pseudotemporal analysis suggested that the trajectory from transitional PTC to FR-PTC1 was independent from that to FR-PTC2 (SI Appendix, Fig. S21B), indicating that FR-PTC1 and FR-PTC2 may be separately dedifferentiated from normal epithelia.

Next, we performed transcription factor binding motif enrichment analysis to understand the epigenetic mechanism driving the heterogeneity of FRPTCs (Fig. 4C). FR-PTC1 enriched AP-1 transcription factor (Fig. 4C and SI Appendix, Fig. S22 and Dataset S14) as well as NF-kB family transcription factor binding motifs (SI Appendix, Fig. S22 and Dataset S14), consistent with previous findings that Vcam1+ FR-PTC demonstrates an increase in AP-1 and RELA binding motif availability(8, 12, 24). In contrast, the FR-PTC2 subtype showed lowest motif enrichment for HNF4A among PTC (SI Appendix, Fig. S22) as well as the motif enrichment of TEAD family transcription factors (Fig. 4 C, D), through which the Hippo signaling pathway regulates expression of the downstream target genes. The Hippo pathway has been implicated in regulation of proliferation during development and pathologic condition including various cancers (39) and ADPKD (40). The most differentially up-regulated gene in FR-PTC2 was Rbms3 encoding an RNA-binding protein (Fig. 3B). Rbms3 was found to be one of the common Hippo pathway target genes (39), in agreement with Hippo pathway deregulation in this subtype. Furthermore, gene set enrichment analysis indicated upregulation of genes related to Hippo pathway in FR-PTC, especially FR-PTC2 (Fig. 4E). Gene set enrichment analysis for hallmark gene sets suggested enrichment of genes related to cell polarity (apical surface) as well as notch signaling in FR-PTC2 (pathways in blue characters in SI Appendix, Fig. S23). Both of these pathways have been implicated in deregulation of the Hippo pathway(41, 42). In contrast, FR-PTC1 enriched inflammatory pathway genes (pathways in red characters in SI Appendix, Fig. S23).

Interestingly, gene set enrichment analysis has also associated human ADPKD PT3 with Hippo pathway (SI Appendix, Fig. S19 C and D). Furthermore, the representative TEAD target gene RBMS3 expression was up-regulated in human ADPKD PT3 (SI Appendix, Fig. S19B). These observations suggest that Hippo pathway deregulation may be a shared feature of atypical FR-PTC between mouse (FR-PTC2) and human kidneys (PT3).

Next, we evaluated the correlation of FRPTC1-3 with the FR-PTC state after ischemia–reperfusion injury (IRI) (10–12, 24). We have performed label transfer from mouse PKD FR-PTC subtypes to IRI-associated FR-PTC in our snRNA-seq dataset (10) to predict the frequency of FR-PTC1-3 emerging after IRI. After label transfer, the predicted FR-PTC2 highly expressed Cdh6, confirming successful prediction (SI Appendix, Fig. S24A). The majority of FR-PTC in IRI was predicted as FR-PTC1 (SI Appendix, Fig. S24B). FR-PTC2 and FR-PTC3 were also predicted to exist in each time point (2, 14, or 42 d following IRI), although their frequencies (~10-20% of total FR-PTC) were less than that of FR-PTC1 (SI Appendix, Fig. S24B). This finding is in line with the previous studies describing that CDH6+ PTC emerges after epithelial injury (11, 28, 37, 38). The caveat is that there may be a possible discrepancy of individual gene expressions between datasets with difference in modality or sequencing depth (SI Appendix, Fig. S25). Nevertheless, our analysis indicates that atypical FR-PTC may have distinct roles both in PKD and other kidney diseases including acute kidney injury.

Collectively, our analysis suggests each of FR-PTC subtypes separately dedifferentiated from normal epithelia by Pkd1 deletion, with unique activation of transcription factors and downstream molecular signaling. These FR-PTC subtypes may have unique roles in PKD progression.

GPRC5A as a Cyst Lining Cell Marker for both Human and Mouse PKD.

We have previously identified GPRC5A as a marker for cyst-lining cells in human advanced ADPKD samples(13). GPRC5A is a G-protein-coupled receptor which is deregulated in many forms of cancer(43). Recently, aromatic monoamines were found to bind GPRC5A and stimulate β-arrestin recruitment (44), although the physiological ligands in vivo as well as its biological function have been elusive. Gprc5a expression was detected in many cell types although its expression was most abundant in FRPTC/PEC, DTL as well as some distal nephron cell types in our dataset (Fig. 5A). In control mouse kidneys, GPRC5A protein was mainly detected in the tubules lacking both LTL and DBA in the medullary region (SI Appendix, Fig. S26). These tubules were often continuous to the LTL+ cells in the cortex (SI Appendix, Fig. S26), suggesting that these structures are DTL. The GPRC5A expression in DTL was also maintained through the progression of PKD (SI Appendix, Fig. S26). In the distal nephron subclustering, Gprc5a expression was mainly detected in cPC2, mPC1/2 as well as URO (Fig. 5B). GPRC5A protein expression was often observed in DBA+ collecting duct cyst lining cells at P100 and P130 (Fig. 5C and SI Appendix, Fig. S27A). Gprc5a expression was also widely detected in FR-PTC subtypes, while it was hardly detected in transitional PTC (Fig. 5D), consistent with the observation that FR-PTC composes cystic epithelia in the PKD kidneys (Fig. 3). Immunostaining analysis revealed GPRC5A protein was often detected in proximal tubular cysts at P100 and P130 (Fig. 5E and SI Appendix, Fig. S27B). Similar to VCAM1, GPRC5A+ epithelia lost LTL staining, indicating GPRC5A is a marker for dedifferentiation in PT-derived cyst epithelia. In agreement with this notion, GPRC5A expression was often colocalized with VCAM1 expression (Fig. 5F). GPRC5A expression was not observed in VCAM1+ noncystic, atrophic tubules, suggesting that GPRC5A is a cyst cell marker but not FR-PTC marker (Fig. 5F). GPRC5A expression was not detected in the P66 cortex (SI Appendix, Fig. S27A), indicating that cystic GPRC5A upregulation starts between P66 and P100 during development of cystic lesions. The median size of GPRC5A+ cysts was larger than that of adjacent cysts lacking GPRC5A expression (Fig. 5G), suggesting that GPRC5A may be the marker of growing cysts rather than early cystogenesis. These findings collectively indicated that GPRC5A is a conserved cyst lining cell marker between human and mouse PKD, and that it marks the cyst lining cells regardless of their original lineages.

Fig. 5.

GPRC5A as a shared cyst lining cell marker for mouse and human PKD. (A and B) UMAP plot showing a gene expression level of Gprc5a in the whole dataset (A) or distal nephron subclustering (B). The color scale represents a normalized log-fold-change (LFC). (C) Representative immunofluorescence images of LTL (green), GPRC5A (red), and DBA (white) in the PKD kidneys at P130 (n = 3). Arrowheads indicate GPRC5A+DBA+ cyst lining. Scale bar indicates 50 µm. (D) UMAP plot showing a gene expression level of Gprc5a in the FR-PTC subclustering. The color scale represents a normalized LFC. (E and F) Representative immunofluorescence images of VCAM1 (green), GPRC5A (red), and LTL (white) in the PKD kidneys at P130 (n = 3). The arrows indicate VCAM1+ atrophic tubules. Arrowheads indicate cyst lining with GPRC5A mutually exclusive with LTL (E) or colocalized with VCAM1 (F). Scale bar indicates 50 µm. (G) Areas of GPRC5A+ cysts and the adjacent cysts lacking GPRC5A signals in the PKD mouse kidneys at P130. The quantification was performed in five 200× images including GPRC5A+ cysts, taken from each of n = 3 PKD kidneys. The box-and-whisker plots depict the median, quartiles, and range. Wilcoxon rank sum test. (H and I) Representative immunofluorescence images of LTL (green), GPRC5A (red), and DBA (white) in the Pkd1^RC/RC mouse cystic kidneys at 11 mo of age (n = 3) (G) or Six2-Cre; Pkd1^F/F cystic kidneys at P7 (n = 2) (H). Scale bar indicates 50 µm (H), 100 µm (I, Left), or 10 µm (I, Right). (J) Cis-coaccessibility network (CCAN, gray arcs) of a conserved cis-regulatory region (CRE) 5’ distal to Gprc5a promoter in the mouse PKD FR-PTC (lower) or human ADPKD (upper) among accessible regions (red boxes) is shown. (K) Coverage plot showing accessibility of conserved CRE 5’ distal to Gprc5a gene among PT subtypes. The conserved CRE has several TEAD family binding motifs both in humans and mice. (L) Quantitative PCR for GPRC5A expression in primary human PTC with siRNA knockdown of LATS1 and LATS2 (n = 3). Bar graphs represent the mean and error bars are the s.d. Student’s t test (Upper). LATS1/2 knockdown inhibits Hippo pathway, activating TEAD and subsequently up-regulating GPRC5A expression (Lower). Schematic was created with BioRender.

Pkd1^RC/RC is another orthogonal ADPKD model mouse with Arg3277Cys (RC) mutation, which is associated with hypomorphic PKD1 (45). The cystic kidneys with this mouse model demonstrated GPRC5A expression in the cyst lining cells originated from proximal tubular cells as well as collecting duct cells (Fig. 5H). Furthermore, the cystic lesions in a rapidly progressive PKD model mouse (Six2-Cre; Pkd1^F/F) kidneys (46) also demonstrated localization of GPRC5A in the cyst lining cells (Fig. 5I). These findings confirmed GPRC5A is a common cystic cell marker in disparate mouse models of PKD.

Conservation of GPRC5A Distal Enhancer and Regulation of GPRC5A through the Hippo Pathway.

We identified a cis-regulatory region (CRE) 5’ distal to the GPRC5A promoter in human ADPKD kidneys and validated it by CRISPR interference(13). The sequence of the CRE was conserved on the 5’ distal region of the Gprc5a gene also in the mouse genome (remapping ratio > 0.5), and this region was more accessible in FR-PTC compared to normal PT epithelia in the PKD mouse model (Fig. 5 J and K). Cis-coaccessibility network (CCAN) modeling in mouse FR-PTC predicted the interaction between this conserved CRE with the Gprc5a promoter (Fig. 5J), suggesting that the epigenetic mechanism regulating Gprc5a expression is conserved in mouse PKD. We have also shown that the CRE in human ADPKD includes cAMP-responsive element binding protein 1 (CREB1) motif as well as Hippo pathway effector TEAD family transcription factor binding motifs (13). The mouse Gprc5a CRE also has TEAD binding motifs (SI Appendix, Fig. S28). ChIP-seq of TEAD1 in human cancer cell lines has shown that TEAD1 coupled with YAP binds to the GPRC5A CRE (SI Appendix, Fig. S29) (47), suggesting that the CRE may regulate GPRC5A expression through the Hippo pathway, in addition to cAMP and retinoic acid signaling (13). LATS1/2 are the kinases essential for Hippo pathway regulation, phosphorylating YAP for subsequent degradation. Deregulation of the Hippo pathway by siRNA knockdown of LATS1/2 induced GPRC5A upregulation in the primary proximal tubular cells (Fig. 5L and SI Appendix, Fig. S30), in agreement with the notion that GPRC5A is regulated by TEAD family transcription factors in addition to cAMP and retinoic acid signaling pathways. This finding along with our previous study (13) indicates that GPRC5A is regulated by a cohort of signaling pathways associated with PKD progression, including cAMP, retinoic acid, and Hippo pathways. These findings collectively suggest that the epigenetic mechanism of GPRC5A regulation and the gene regulatory network driving PKD progression are largely conserved between human and mouse kidneys.

Discussion

In this study, we generated a single-nucleus multiomic atlas for a mouse model of PKD along the time course following Pkd1 inactivation to comprehensively describe the dynamic cell states during PKD progression. This research strategy allows us to dissect cellular heterogeneity and characterize molecular response to PKD1 deficiency in each cell type at each disease stage. Given the recent evidence that PKD is partially reversible by Pkd1 reactivation (48), the persistent effect of Pkd1 deletion at each PKD stage may be critical for PKD progression, underscoring the importance of analysis at time-course analysis following Pkd1 deletion. Comparing and contrasting with the mouse PKD atlas to human ADPKD dataset (13) shed light on the conserved cellular heterogeneity (Figs. 2–4) in PKD as well as shared molecular mechanism including transcriptional regulation of a cyst lining cell marker; GPRC5A (Fig. 5).

Our single-cell analysis indicated that the differentially expressed genes as well as accessible regions in Pkd1-deficient tubular epithelia are highly cell-type-specific (Datasets S2–S4 and SI Appendix, Fig. S6), although the resultant phenotype; cyst formation is shared across tubular compartments. This observation suggests that the response to Pkd1 deletion as well as the molecular mechanism of cyst formation may be cell-type-specific. Furthermore, the effect of Pkd1 deletion in one cell type may affect the responses to Pkd1 deletion in other cell types through modifying the microenvironment, potentially further complicating the molecular response to Pkd1 deficiency in vivo. In addition, we found that the cell-type-specific response to Pkd1 is dynamically changed along PKD progression (Figs. 2 E and F and 3G and Datasets S2–S4). The medullary principal cell subpopulation initially showed molecular signature consistent with DNA damage response and alteration of metabolic pathways following Pkd1 deletion, and subsequently inflammatory responses are prominent at a later time point (Fig. 2E). Our findings emphasize the importance of single-cell analysis with a time course for dissection of PKD mechanism.

In this study, we described that heterogeneous FR-PTCs with distinct molecular signatures are components of proximal tubular cysts. Furthermore, FR-PTCs in PKD show unique gene expression signature compared to those observed after ischemia–reperfusion injury (10). Pseudotemporal analysis indicated FR-PTC subtypes separately arise from a transitional cell state, and their chromatin accessibilities as well as transcription factor activities are unique (Fig. 4 and SI Appendix, Fig. S21). FRPTC1 showed RELA and AP-1 family activation and subsequent inflammatory molecular signature, while FRPTC2 is characterized by TEAD family activation probably through Hippo pathway deregulation (Fig. 4). However, the differential roles of these heterogeneous FR-PTCs in PKD progression have remained elusive in this study. Future genetically engineered approaches and lineage-tracing in a mouse PKD model may reveal their contributions to PKD progression.

Our previous single-cell analysis on human ADPKD identified GPRC5A as a cyst lining cell marker with epigenetic deregulation in ADPKD (13). Interestingly, GPRC5A is a cyst marker also in mouse kidneys regardless of their proximal or distal origin (Fig. 5). In the normal kidneys, GPRC5A protein was mainly detected in DTL (SI Appendix, Fig. S26), suggesting that dedifferentiated epithelia induced by the loss of Pkd1 may acquire some features of DTL. Indeed, Rbms3 expression, which was up-regulated FR-PTC2, was the most differentially expressed gene in DTL1 (Fold change: 2.31, Dataset S1). Additionally, hypoxic milieu in advanced PKD may contribute to GPRC5A expression in cystic epithelia.

We also found that the 5’ distal enhancer activating GPRC5A expression in human ADPKD was conserved in mice (Fig. 5). The 5’ distal enhancer is characterized by multiple TEAD binding motifs both in mouse (SI Appendix, Fig. S28) and human kidneys (13). Indeed, Hippo pathway deregulation by LATS inhibition up-regulated GPRC5A expression in human primary PTC (Fig. 5). This finding suggests that upregulation of GPRC5A expression is involved in Hippo pathway deregulation in cystic epithelia. The regulation of GPRC5A expression by the Hippo pathway is in line with the previous study implicating Hippo pathway in PKD progression (41). Our finding suggests that the disturbance of the gene regulatory networks driving GPRC5A expression may be shared with multiple cell types in mouse and human PKD despite the general cell-type-specificity of the responses to Pkd1 deletion (SI Appendix, Fig. S6). However, the function of GPRC5A as well as physiological ligand in vivo remains elusive. Future study with a PKD mouse model with Gprc5a deletion or pharmacologic inhibition may reveal the role of GPRC5A in cyst growth in PKD.

In summary, we performed multiomic single-nucleus analysis for a mouse PKD to characterize temporally dynamic, cell-type-specific deregulation of gene expressions and chromatin accessibilities. By comparing our mouse PKD datasets with human ADPKD datasets, our study sheds light on the conservation of cellular heterogeneity in PKD. Our findings collectively indicated that the intersectional analysis for our mouse and human PKD dataset is a useful approach to describe PKD mechanism and identify potential therapeutic targets. This study is limited by the lack of experimental analysis on the roles of heterogeneous principal cells and FR-PTC subtypes in PKD progression. Furthermore, the mechanistic role of GPRC5A in PKD progression needs to be addressed. Despite these limitations, our single-nucleus multimodal atlas of mouse PKD as well as our analysis provides a foundation to promote future efforts to understand cell-specific mechanism of cystic tubular changes and develop better therapeutic approaches for PKD.

Materials and Methods

PKD Mouse Model.

All animal experimental procedures were conducted following the University of Maryland or Washington University Animal Care and Use Committee (IACUC) guidelines and procedures. The Pkd1 conditional allele (Pkd1^fl/fl) used in these studies has been previously described (15). The Pkd1^fl/fl; Pax8-rtTA; Tet-O-Cre mouse model is an orthologous mouse model of ADPKD that utilizes the murine Pax8 promoter to drive expression of the reverse tetracycline-dependent transactivator (rtTA) to all renal tubular compartments (14). To induce Pkd1 deletion, experimental mice were treated with a single daily intraperitoneal injection of doxycycline hyclate (Sigma-Aldrich, D9891) 50 mg/Kg body weight, diluted in sterile distilled water. Specific time points for induction are indicated P27, P28, P29, P44, and P56. Littermate controls lacked either the Pax8-rtTA or the Tet-O-Cre allele but were treated with doxycycline in the same manner as experimental mice. The samples for sequencing studies were all from males as well as littermates except one control mouse at P130. P66 samples (n = 2) were from 2 litters and each litter has a PKD and control mouse. All the samples at P100 were from one litter. At P130, 3 PKD and 2 controls were from the same litter, but the last control was from a different litter (SI Appendix, Table S1). All mice were maintained on an inbred C57BL/6 background. Kidneys were harvested at P66, P100, or P130. The right kidneys were decapsulated and snap-frozen in liquid nitrogen for single-nucleus multiomics analysis. The kidneys for male Pkd1^fl/fl; Six2-Cre mouse model kidneys were analyzed at P7 (n = 2). The Pkd1p.R3269C knock-in mice [equivalent to human PKD1p.R3277C mutation (Pkd1^RC/RC)] (45, 49) were analyzed at 11 mo of age (2 female and 1 male mouse kidneys).

Single Nucleus Multiomics Library Sequencing and Following BioinFormatics Workflow.

Briefly, single-nucleus multiomics libraries were generated using Chromium Next GEM Single Cell Multiome ATAC+ Gene Expression Reagent Kit (10× Genomics) following nuclear dissociation. Libraries were sequenced on an Illumina Novaseq instrument and counted with CellRanger ARC v2.0.2 (10× Genomics) using refdata-cellranger-arc-mm10-2020-A-2.0.0, which is mm10 reference genome 10× Genomics provided. The output of CellRanger ARC was processed through Seurat v4.0.2 (16). The details of the nuclei preparation, sequencing, and subsequent bioinformatic workflow are described in Supplementary Materials and Method.

Statistical Analysis.

No statistical methods were used to predetermine sample size for single-nucleus analysis. Experiments were not randomized and investigators were not blinded to allocation during library preparation, experiments, or analysis. Bonferroni adjusted P values were used to determine significance for differential accessibility. The quantitative PCR data (Fig. 5L) are presented as mean±s.d. and were compared between groups with two-sided Student’s t-test. The kidney to body weight ratios (SI Appendix, Fig. S1A) were presented as mean±s.d. and one-way ANOVA with the post hoc Tukey test. The Wilcoxon rank sum test was performed for quantification of cyst size (Fig. 5G). A P value of < 0.05 was considered statistically significant.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (CSV)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Dataset S06 (XLSX)

Dataset S07 (XLSX)

Dataset S08 (XLSX)

Dataset S09 (XLSX)

Dataset S10 (XLSX)

Dataset S11 (CSV)

Dataset S12 (XLSX)

Dataset S13 (CSV)

Dataset S14 (CSV)

Data, Materials, and Software Availability

Fastq and processed files (single-cell sequencing) data have been deposited in GEO (NCBI) (GSE268494) (50). All R scripts for preprocessing the single-nucleus multiomics atlas for mouse PKD are available on GitHub at https://github.com/Yoshi-MutoLab/Mouse_PKD_Multiomics (51). Gene expression and ATAC peaks for each cell type are also available via our interactive website Kidney Interactive Transcriptomics (http://humphreyslab.com/SingleCell/) (52).