Smad4 controls proliferation of interstitial cells in the neonatal kidney

Interstitial cells are any cells located between the functional cells, or parenchyma, of a tissue. In the kidney, cells of the nephron and blood vessel are considered the functional components and the interstitial cell population is made up largely of PDGFRβ-expressing fibroblasts. These cells derive from a Foxd1-expressing progenitor, which also gives rise to mesangial cells, the specialized pericytes of the glomerulus. Expansion of interstitial cells in the adult kidney is a hallmark of chronic disease, with unopposed interstitial cell proliferation leading to progressive scarring (fibrosis) and concomitant loss of parenchyma (Humphreys et al., 2010). In contrast, proliferation of interstitial cells plays an essential role in fetal development and somatic growth of the kidney (Boivin and Bridgewater, 2018; Das et al., 2013; Fetting et al., 2014; Hatini et al., 1996). An intriguing difference between adult and neonatal kidneys is that the neonatal kidney has the capacity to control interstitial cell proliferation when the target number has been reached, and our study aims to characterize the basic mechanisms of this proliferation control.

Gene inactivation and experimental therapeutics targeting the TGFβ signaling pathway have been used to reduce fibrosis in kidney injury models (Inazaki et al., 2004; Morishita et al., 2014), indicating that the pathway is a significant driver of interstitial expansion. We therefore hypothesized that the TGFβ pathway may control interstitial cell proliferation in the neonatal kidney. Although TGFβ signaling has been a focus of research in homeostasis of adult interstitial cells, little is known about its function in differentiation of the renal interstitium. TGFβ signaling is essential for development of the kidney (Dudley et al., 1995; Ikeya et al., 2010; Oxburgh et al., 2004), and controls formation of both the collecting ducts and nephrons (Brown et al., 2013; Hartwig et al., 2008). Recessive mutations in genes within this signaling pathway have been identified as monogenic causes of congenital anomalies of the kidney and urinary tract (CAKUT), indicating that TGFβ pathway dysregulation may be an important factor in neonatal kidney disease. An important example is the identification of mutations in bone morphogenetic protein 4 (BMP4) in patients with renal hypodysplasia (Kohl et al., 2014; Weber et al., 2008).

TGFβ superfamily ligands signal through two distinct intracellular pathways: the Smad pathway, which is initiated by phosphorylation of receptor-associated Smad (R-Smad) transcription factors, and the mitogen-associated protein kinase (MAPK) pathway, which is initiated by the TGFβ-associated kinase MAP3K7 (also known as TAK1). Kidneys of mice with conditional inactivation of Map3k7 in the Foxd1 lineage develop spontaneous neonatal mesangiosclerosis (Karolak et al., 2018), suggesting that MAPK is selectively required for mesangial differentiation.

The current study addresses the role of Smad pathway signaling in the Foxd1 lineage. Phosphorylated R-Smads associate with the common mediator Smad (Smad4) to accumulate in the nucleus. R-Smads and Smad4 bind DNA and interact with a variety of other transcription factors, facilitating highly context-dependent responses. Because Smad4 is unique and essential for Smad-mediated responses, we selected it as a tractable node in the pathway for conditional gene inactivation. We report that Smad4 controls proliferation within the interstitium of the mouse neonatal kidney by attenuating Wnt signaling.

To understand the role of TGFβ/BMP in the developing renal interstitium, Smad4 was inactivated in interstitial cell progenitors using Foxd1^Cre. To sensitize the strain for Cre-mediated recombination, Foxd1^Cre was combined with one null Smad4 allele and one loxP-flanked allele (Chu et al., 2004). This cross generates offspring of four genotypes: Foxd1^+/+;Smad4^−/loxp, Foxd1^+/+;Smad4^+/loxp, Foxd1^+/Cre;Smad4^+/loxp and Foxd1^+/Cre;Smad4^−/loxp. Foxd1 is disrupted in the Foxd1^cre strain and Foxd1^+/Cre;Smad4^+/loxp was selected as the control group (referred to as Smad4^con) to control for any subtle effects of compound heterozygous inactivation of Foxd1 and Smad4. Foxd1^+/Cre;Smad4^−/loxp is referred to as Smad4^IC. In this comparison, the Smad4^IC group is globally heterozygous for Smad4. For this study, we used the Smad4^tm1Rob/+ strain, in which the first coding exon of Smad4 is removed (Chu et al., 2004). Smad4^con mice did not develop any overt phenotypes. Smad4^IC mice were distinguishable from Smad4^con littermates at birth because they had kinked tails, and they developed hindlimb paralysis during the first postnatal week. These phenotypes are most likely explained by expression of Foxd1 in the somite and in the central nervous system (Hatini et al., 1994; Robinton et al., 2019). To avoid confounding effects of hindlimb paralysis on postnatal kidney development, analyses were carried out at embryonic time points up to postnatal day (P) 1.

To evaluate the efficiency of Cre recombination, we first compared Smad4 transcript levels, and found that they are reduced by approximately 60% in Smad4^IC kidneys versus Smad4^con (Fig. 1A). To measure recombination efficiency specifically in the Foxd1 lineage, we selected cells expressing the interstitial cell surface marker PDGFRα (Fig. S1A,B), and single-cell genotyped them (Fig. S1C,D). Fig. 1B summarizes the frequency of recombination of interstitial cells in the Smad4^IC genetic model; 87.1% of cells were recombined at the Smad4 locus and thus null for Smad4.

To determine whether loss of Smad4 in the Foxd1-expressing interstitial cell progenitor affects lineage commitment, we introduced the Rosa26R reporter gene. The localization of Foxd1 lineage cells in the Smad4^IC;Rosa26R strain was indistinguishable from Smad4^con;Rosa26R, and thus we conclude that lineage commitment is unperturbed (Fig. S2). Transverse sections were cut through the center of the kidney to reveal the cortex and medulla. The cortex contains glomeruli, whereas the medulla does not. The medulla is divided into an outer medulla directly adjacent to the cortex and an inner medulla or papilla that extends into the pelvis of the kidney. For clarity, we refer to three distinct regions: cortex, outer medulla, and papilla (equivalent to inner medulla). Compared with Smad4^con, Smad4^IC showed abundant stroma in the outer medulla and papilla, and a paucity of epithelial structures in both of these zones (Fig. 1C,D). The R26R reporter serves as a helpful marker that can be used in parallel with histological analysis to understand the abundance of Foxd1 lineage cells. Comparison of Smad4^con (Fig. 1E-J) and Smad4^IC (Fig. 1K-P) revealed a marked expansion of Foxd1-lineage interstitium, which was regionalized in the tissue; a modest difference was noted in cortical interstitium (Fig. 1H,N), whereas pronounced pockets of stroma were seen in the outer medulla and papilla (Fig. 1I,J,O,P). We conclude that Smad4 is required in the Foxd1 lineage for appropriate formation of the renal interstitium.

During renal development, interstitial cells transition from a FOXD1/PDGFRα double-positive state in the cortex to an α-SMA/PDGFRβ double-positive state in the outer medulla. To determine whether the stromal expansion observed in Smad4^IC mice is associated with impaired interstitial cell differentiation, expression of the interstitial markers α-SMA and PDGFRβ were examined in kidneys from embryonic and postnatal mice. Though comparable to control at E17.5 (Fig. S3), differences in α-SMA and PDGFRβ expression are observed in Smad4^IC mouse kidneys by P0. Smad4^IC mice displayed reduced α-SMA expression in the outer medulla compared with control (Fig. 2A,B). In contrast, PDGFRβ expression was increased in both the outer medulla and papilla of Smad4^IC mice compared with control (Fig. 2C,D). Confirming these findings, immunoblots of lysates from whole kidneys (Fig. 2E) showed a significant increase of PDGFRβ (Fig. 2F) and decrease of α-SMA (Fig. 2G) in Smad4^IC mice. The lack of expression of α-SMA in the abundant Foxd1-derived interstitial cells of the outer medulla suggests that Smad4 is required for appropriate expression of their differentiation program.

One possible explanation for the pockets of interstitial cells seen in the Smad4^IC medulla is that collecting duct (CD) organization is impaired, causing aberrant clustering of interstitial cells; to determine whether this was the case, we compared Smad4^IC kidney tissue with controls using two strategies. Three-dimensional modeling of whole kidneys stained with the CD marker TROMA1 (cytokeratin 8) showed that branching at embryonic day (E) 14.5 is indistinguishable between Smad4^con and Smad4^IC mice (Fig. 3A-C, Fig. S4A,B). Studies of kidney development using the Smad4^tm1Mak strain, which has a deletion in the carboxy-terminal domain of SMAD4 (Sirard et al., 1998), have determined a 12% increase in the number of collecting duct branch-points in heterozygotes at E13.5 (Hartwig et al., 2005). The fact that we did not observe any difference between Smad4^con and Smad4^IC indicates that this phenotype of global Smad4 modification is not reproduced with the Smad4^tm1Rob/+ strain used in this study. To analyze collecting ducts at P0, when the kidney is too large for modeling based on whole-mount immunostaining, we sectioned through the longitudinal plane of the kidney perpendicular to the papilla to obtain transverse sections of the papilla where collecting ducts are closely bundled (Fig. 3D), and immunostained these sections. We found a reduced number of patent TROMA1-positive structures in the papilla of Smad4^IC mice compared with Smad4^con (Fig. 3E,F). High-magnification images of the papilla revealed TROMA1-positive tubules with an atypical, compressed morphology in Smad4^IC kidneys (Fig. 3G,H). When quantified, the compressed TROMA1-expressing CDs account for the reduced number of CDs in Smad4^IC mice (Fig. 3I). We hypothesize that these structures are remnants of functional CDs that are constricted by stromal expansion. CD compression is predicted to cause restricted urine outflow, and, consistent with this, Smad4^IC mutant kidneys displayed nephron tubule expansion (Fig. 3J-L) and distended Bowman's spaces (Fig. 3M-O). Increased mesangial α-SMA expression (Fig. 3P-Q′, Fig. S4C,D) and mesangial area (Fig. 3R) indicate early glomerulosclerosis in Smad4^IC mice. In vivo 5-ethynyl-2′-deoxyuridine (EdU) incorporation revealed an increased proliferative index in mesangial cells of the mutant at P0 (Fig. S4E-G), indicating that this defect is mesangioproliferative. Numbers of glomeruli were comparable between Smad4^con and Smad4^IC (Fig. S4H), suggesting that the features of increased physiological load that we observed are not due to differences in nephron number. In summary, the histological features that we observed are consistent with CD compression causing increased luminal pressure in the nephron with associated sclerosis of the glomerulus.

To understand whether loss of Smad4 affects nuclear accumulation of R-Smads in regions of medullary interstitial cell expansion, protein localization of Smads was compared in Smad4^con and Smad4^IC mouse kidneys. To localize expression in interstitial cells, we co-stained for α-SMA. Signal amplification was used to detect α-SMA reliably in the Smad4^IC kidney. After testing multiple different antibodies for receptor Smads, we concluded that nuclear localization of Smad3 is the most sensitive proxy for TGFβ signaling and pSmad1/5/8 is the most sensitive for BMP. These assays are not directly comparable with each other because one measures localization of total protein and the other measure localization of a phosphorylated form. However, signal intensity for each is well documented to correlate with cell signaling, leading us to conclude that they are valid reporters for TGFβ and BMP signaling, respectively. As anticipated, Smad4 was lost from the interstitium of Smad4^IC mice (Fig. 4A-B′). In addition, Smad3 was strongly reduced in the interstitium of Smad4^IC kidneys compared with control (Fig. 4C-D′), whereas pSmad1/5/8 levels were only slightly reduced (Fig. 4E-F′). These results reveal that TGFβ signaling is primarily affected by Smad4 inactivation.

Smad4^IC kidneys are comparable to Smad4^con at E17.5 but show a phenotype in the newborn, so we were interested to understand whether there were dynamic changes in the TGFβ pathway at late developmental time points. A developmental time course of gene expression in the kidney has been reported (Challen et al., 2005), and we screened developmental time points from E15.5 to P0 for components of the TGFβ signaling pathway. We based the selection of pathway genes on the KEGG pathway hsa04350, which includes 91 genes. Of these, 14 were differentially expressed in the kidney developmental time course. From this transcriptomic analysis, there was no obvious evidence for a coordinated change in the expression of TGFβ pathway components as an equivalent number of genes were up- and downregulated between E17.5 and P0 (Fig. S5).

To determine whether aberrant proliferation is responsible for the interstitial expansion observed in Smad4^IC mice, we performed in vivo EdU incorporation at P0 (Fig. 5). To localize proliferating cells in the interstitium, we co-stained tissue with a cocktail of antibodies for α-SMA and PDGFRβ. These two interstitial cell markers differ strongly in expression level between Smad4^IC and Smad4^con, and by detecting both primary antibodies with secondary antibodies conjugated to the same fluorophore we were able to achieve a comparable signal intensity between the two genotypes. Whereas EdU⁺ cycling interstitial cells were largely limited to the cortex and outer medulla in Smad4^con kidneys, they were abundant in the papilla of Smad4^IC mice (Fig. 5A-B′). Three-dimensional modeling and quantification of comparable volumes confirmed a higher proliferation rate in the papillary interstitium of Smad4^IC mice compared with Smad4^con (Fig. 5C-E). We conclude that there is a failure of interstitial cell growth restriction in the transition zone between the outer medulla and papilla.

Several lines of evidence indicate that Wnt/β-catenin signaling is required for interstitial cell maintenance: loss of Wnt7B from the collecting duct epithelium or inactivation of β-catenin in interstitial cell precursors both result in medullary hypoplasia (Boivin et al., 2016; Yu et al., 2009). Thus, developmental genetic studies support a model in which collecting duct-derived Wnt drives medullary interstitial cell proliferation through a β-catenin-dependent signaling mechanism. We were therefore curious to understand whether Wnt/β-catenin signaling was perturbed in the Smad4^IC kidney. Expression of the feedback inhibitor axin 2 is a sensitive read-out of Wnt/β-catenin signaling, and we compared its expression level in Smad4^con and Smad4^IC kidneys. In situ hybridization revealed axin 2 expression in the interstitium of the outer medulla and papilla of Smad4^con, but undetectable levels in Smad4^IC (Fig. 6A,B, Fig. S6). Reduction in activation of feedback inhibitors can be interpreted as evidence of reduced Wnt/β-catenin signaling, which would be unanticipated considering the proliferative phenotype of Smad4^IC. We therefore evaluated expression of a panel of other Wnt/β-catenin targets. LEF1 was elevated in the outer medulla and papillary interstitium of Smad4^IC kidneys (Fig. 6C-E), along with the Wnt/β-catenin-activated cell cycle regulator CCND1 (Fig. 6F-H), suggesting an increase in Wnt/β-catenin signaling and proliferation relative to Smad4^con. We also found increased expression of the Wnt-responsive cell cycle regulator CDKN1C (p57Kip2) (Fig. 6I-K), which is required for renal medulla formation (Yu et al., 2009; Zhang et al., 1997).

Considering the central role that Smad4 plays in TGFβ superfamily signaling, it is not surprising that the Foxd1^IC strain displays a profound phenotype in the neonate. TGFβ superfamily signaling in kidney fibroblasts has mainly focused on the study of TGFβ ligands, because foundational work showed that they promote proliferation (Roberts et al., 1985), extracellular matrix deposition (Edwards et al., 1987; Ignotz and Massagué, 1986) and myofibroblast transition (Rønnov-Jessen and Petersen, 1993). In previous work, we explored the consequences of inactivating the TGFβ/MAPK pathway in interstitial cells by inactivating TGFβ-associated kinase 1 (Map3k7) using Foxd1^+/Cre (Karolak et al., 2018). In this study, we define the consequences of inactivating the TGFβ/Smad response by inactivating Smad4.

TGFβ superfamily ligands promote phosphorylation of the R-Smads 1, 2, 3 and 5 (Ramachandran et al., 2018; Zhang et al., 1996), and therefore one tractable strategy for studying Smad responses is to inactivate Smad4, which is required for nuclear retention and transcriptional activity of activated R-Smads (Lagna et al., 1996; Schmierer and Hill, 2005). Unexpectedly, we find that Smad4 inactivation eliminates nuclear accumulation of the Smad3 transcription factor, but only slightly reduces nuclear accumulation of the Smad1/5 transcription factors. This is in line with the Smad transcriptional response to TGFβ1, which is estimated to comprise approximately 75% Smad2/3 transcriptional response and 25% Smad1/5 transcriptional response (Ramachandran et al., 2018). The finding that Smad1/5 nuclear accumulation is not lost indicates that the BMP R-Smad response is only marginally affected by Smad4 inactivation. Thus, from the perspective of Smad activation, our genetic model reflects the effects of eliminating TGFβ rather than BMP signaling.

The number of interstitial cells in the kidney must be carefully balanced both during development and in the adult. In development, interstitial cells provide essential signals that guide differentiation of the surrounding epithelia (Das et al., 2013; Fetting et al., 2014). In the adult, interstitial cells provide essential endocrine functions, such as erythropoietin production (Kobayashi et al., 2016), but their uncontrolled expansion is the basis for organ fibrosis whereby functional tissue is marginalized (Humphreys et al., 2010). Our observation that reducing TGFβ/Smad signaling in the interstitial cell lineage leads to increased proliferation in the neonate agrees with findings from studies of multiple cell lines showing that growth is inhibited by TGFβ (Roberts et al., 1985). An interesting feature of our study is that the effect is limited to a particular zone of interstitial cells at the border between the outer medulla and the papilla. Genetic studies have suggested that formation of the medullary interstitium requires Wnt/β-catenin signaling and our work indicates that TGFβ/Smad signaling inhibits this proliferative stimulus in a regional manner to ensure appropriate structural differentiation of the medulla. Interestingly, the effect of knocking out the Wnt/β-catenin pathway in the interstitial lineage is loss of the kidney medulla, but the cortex of these kidneys is preserved (Boivin and Bridgewater, 2018; Yu et al., 2009), suggesting specificity of Wnt/β-catenin for medullary interstitium. Similarly, our inactivation study of Smad4 reveals overproliferation of the medullary interstitium, indicating that this Wnt/β-catenin–TGFβ/Smad circuit for growth control primarily affects this region.

Loss of Smad4 causes differential effects on Wnt/β-catenin targets, with a loss of expression of the Wnt feedback inhibitor axin 2 and increased LEF1 and p57Kip2. This supports a model in which loss of feedback inhibition causes increased Wnt/β-catenin signaling resulting in increased proliferation. We hypothesize that the loss of feedback inhibition could be a primary event in the deregulation of interstitial cell proliferation. Multiple Wnts, including Wnt4 and Wnt7b, are expressed in the medullary interstitium (Yu et al., 2009), facilitating combinatorial effects. TGFβ sources have been less well studied, but review of the GenePaint in situ hybridization database shows Tgfb1 expression in the medulla (https://gp3.mpg.de/; Visel et al., 2004). Recent single-cell expression analysis of E18.5 kidneys has shown a spatial intersection of Lef1 expression with expression of the TGFβ-induced transcript Tgfb1i1, supporting a model in which Wnt/β-catenin mediated proliferation may be controlled by TGFβ (England et al., 2020). Relative amplitudes of these different signaling pathways could be regulated at the level of the ligands or perhaps more likely by other intracellular signaling components or distinct signaling pathways.

In summary, our findings suggest that Smads and Wnt/β-catenin antagonistically control cell proliferation in the medullary interstitium and that the balance of these signaling pathways determines interstitial cell abundance in the postnatal kidney. Understanding this molecular crosstalk will not only contribute important concepts to the pathogenesis of CAKUT, but may also identify therapeutically tractable mechanisms that control kidney fibrosis in the adult.

Animal care was in accordance with the National Research Council Guide for the Care and Use of Laboratory Animals and protocols were approved by the Institutional Animal Care and Use Committee of Maine Medical Center. The second exon that includes the transcriptional start site is targeted in Smad4^tm1Rob/+ (referred to as Smad4^+/−) and Smad4^{tm1.1Rob/tm1.1Rob} (referred to as Smad4^loxP/loxP) mice (Chu et al., 2004). Foxd1^+/Cre;Smad4^−/+ mice were crossed to Smad4^loxp/loxp mice to produce Foxd1^+/Cre;Smad4^+/loxp (Smad4^con) and Foxd1^+/Cre;Smad4^−/loxp (Smad4^IC) mice. Foxd1^+/Cre;Smad4^−/+ and Smad4^loxp/loxp mice were maintained on an ICR background. R26RlacZ mice were maintained on an FVB/NJ background.

A complete list of antibodies including source, clone name, product number and application are listed in Table S1.

Total protein was extracted from whole kidneys as previously described (Blank et al., 2009). Immunoblotting was performed using standard procedures. Antibodies used were anti-α-SMA (1:1000), anti-PDGFRβ (1:1000), anti-Smad4 (1:1000) and anti-β-tubulin (1:5000) (see Table S1). Protein levels were quantified using Fiji/ImageJ software by measuring the integrated density of the indicated proteins normalized to the β-tubulin loading control.

E14.5 and P0 kidneys were fixed in 4% paraformaldehyde (PFA) at room temperature for 10 or 30 min, respectively, then transferred to 70% ethanol at −20°C for storage. P0 kidneys were longitudinally or transversely vibratome-sectioned at 100 μm directly into 70% ethanol. Whole kidneys and kidney sections were rehydrated in PBS then permeabilized with 1% Triton X-100 in PBS for 10 min at 4°C. Tissue was washed with PBS to remove residual detergent then incubated in blocking solution (0.01% Tween in PBS plus 5% serum of secondary antibody species) for 1 h. Lectins Dolichos Biflorus Agglutinin (DBA; Vector Laboratories) and Lotus Tetragonolobus Lectin (LTL; Vector Laboratories) were diluted 1:200 and primary antibodies anti-α-SMA-Cy3, anti-PDGFRβ, anti-cytokeratin8/TROMA1, anti-LEF1, anti-BRN1, anti-AnnexinA2, p57Kip2, anti-cyclin D1 and anti-Ki67 (see Table S1) were diluted 1:50 in blocking solution. A Tyramide Signal Amplification kit (TSA; Perkin Elmer) was used according to the manufacturer's protocol for whole-mount immunostaining with anti-Smad4 (1:500), anti-Smad3 (1:3000) and anti-pSmad1/5/8 (1:1000) antibodies (see Table S1). Blocked tissue was incubated in diluted lectins/primary antibodies for 24 h at 4°C followed by three washes with blocking solution, the third wash for 24 h at 4°C. Alexa Fluor 488/568/647 secondary antibodies (Molecular Probes) were used at 1:200 and incubated for 24 h followed by three washes with 0.01% Tween in PBS, the third wash for 24 h at 4°C. Tissue was counterstained with DAPI (1:5000), dehydrated in ethanol, and cleared with BABB (1:1 benzyl alcohol:benzyl benzoate) before imaging with a laser-scanning confocal microscope (Leica Microsystems SP8).

P0 kidneys were fixed in 4% PFA overnight at 4°C, washed with PBS for 4 h, then equilibrated in 30% sucrose in PBS overnight at 4°C. After flash-freezing in OCT, kidneys were cryosectioned at 20 μm and sections were fixed in 4% PFA for 10 min, rinsed three times with PBS, incubated in 20 μg/ml proteinase K in PBS for 10 min, followed by three rinses with PBS. Sections were incubated in 1.3% triethalamine and 0.375% acetic anhydride for 10 min, rinsed three times with PBS then incubated in hybridization buffer for 2 h. Digoxigenin (DIG)-labeled axin 2 sense versus antisense riboprobes (Jho et al., 2002) were diluted to 500 ng/ml in hybridization buffer (50% formamide, 5× SSC pH 4.5, 1% SDS, 50 μg/ml yeast tRNA, 50 μg/ml heparin) and sections were hybridized in a humidified chamber at 68°C overnight. After two washes with 0.2× SSC for 30 min at 72°C, sections were rinsed with NTT (0.15 M NaCl, 0.1% Tween-20, 0.1 M Tris-HCl pH 7.5), incubated in blocking buffer (5% heat-inactivated sheep serum, 2% blocking reagent in NTT) for 2 h, followed by incubation in anti-DIG AP-conjugated antibody (diluted 1:4000 in blocking buffer) overnight at 4°C. After three washes with NTT for 30 min each, sections were rinsed with NTTML (0.15 M NaCl, 0.1% Tween-20, 0.1 M Tris-HCl pH 9.5, 50 mM MgCl₂, 2 mM levamisole) then incubated in BM Purple until desired staining was reached. Stained sections were rinsed three times with PBS then mounted in glycerol for imaging.

One EdU pulse (20 mg/kg) was administered to P0 pups by intraperitoneal injection. After 2 h, kidneys were dissected on ice and fixed in 4% PFA for 30 min. Kidneys were vibratome sectioned and whole-mount immunofluorescence staining was performed as described above. Click-iT chemistry was subsequently performed on sections according to the manufacturer's protocol (Thermo Fisher) followed by imaging with a laser-scanning confocal microscope (Leica Microsystems SP8). EdU quantification was performed with Imaris image analysis software (Bitplane). Briefly, equivalent 250×250×50 μm³ image areas were selected in the outer medulla and papilla and the spot function was used to count EdU⁺ nuclei. A filter was applied to the spot function to select for PDGFRβ/α-SMA-labeled interstitial cells. Equivalent image analysis, including identical thresholding, was performed on six biological replicates (n=6).

Paraffin-embedding and Hematoxylin and Eosin (H&E) staining of tissue was performed by the Maine Medical Center Research Institute Histomorphometry Core. For X-gal staining, whole P0 kidneys were vibratome sectioned (300 μM), rinsed in X-gal buffer (5 mM EGTA; 2 mM MgCl₂; 0.02% NP40; 250 µM sodium deoxycholate in PBS) then fixed for 30 min (1% formaldehyde; 0.2% glutaraldehyde in X-gal buffer). Sections were subjected to two 10 min washes with X-gal buffer then stained (5 mM K₃Fe; 5 mM K₄Fe; 0.5 mg/ml X-gal in X-gal buffer) overnight at 37°C. Stained sections were washed twice (10 min each) with X-gal buffer then dehydrated through a series of ethanol/xylene. Dehydrated sections were embedded in paraffin then re-sectioned at 10 µm and counter stained with Nuclear Fast Red.

Nephrogenic zone cells were isolated from E17.5 Smad4^IC mice as previously described (Brown et al., 2015) and labeled with phycoerythrin (PE)-conjugated anti-mouse CD140a/PDGFRα (1:10; 103-102-502, Miltenyi) followed by incubation with anti-PE MicroBeads (1:20; 130-048-801, Miltenyi). Labeled cells were purified by three rounds of magnetic-activated cell sorting and single cells were manually picked from the PE-positive fraction with the aid of an EVOS FL digital microscope. Total DNA was amplified from single cells with the REPLI-g Single Cell Kit (QIAGEN) and genotyping was performed with NovaTaq Hot Start Master Mix (Millipore) according to the manufacturer's cycling parameters (Tm=55°C) and the primers listed in Table S2.

Chi-square testing was performed to verify that all data sets were normally distributed. Two-tailed t-tests were performed comparing normally distributed groups with P>0.05 considered significant. Proliferation analysis of FH535-treated mice and FH535-treated monolayer cells were subject to one-way ANOVA. Asterisks indicate statistical significance as follows: not significant (ns), P>0.05; *P<0.05; **P<0.01; ***P<0.001.

The authors gratefully acknowledge Dr Volkhard Lindner for providing the pSmad1/5/8 antibody and the MMCRI Histomorphometry Core for tissue processing and H&E and CD31 staining. The authors also thank Drs Lucy Liaw, Calvin Vary, Pradeep Sathyanaryana and Ron Korstanje for their critical review of this manuscript.