α-Parvin regulation of cell rearrangement is critical for ureteric bud branching morphogenesis

Abstract

All branched tubular structures, including the kidney collecting system, are formed by branching morphogenesis, a process that includes tip branching and trunk narrowing. Tight control of cell movement and rearrangement is a prerequisite for branching morphogenesis. The role of integrin-associated adhesion proteins in coordinating actin dynamics and cell rearrangement during branching morphogenesis is poorly understood. Here, we used three-dimensional live imaging of mouse ureteric bud branching to show that α-parvin, a component of the integrin binding ILK-PINCH-parvin complex, regulates tip branching and tubule thinning by inhibiting excessive cell adhesion and actin polymerization. Mechanistically, α-parvin promotes actin turnover by inhibiting activation of the small guanosine triphosphatases RhoA and Cdc42, which in turn enhances the severing function of the actin regulatory protein, cofilin. These results underscore the importance of adhesion protein–regulated actin dynamics in the critical process of cell rearrangement, which is required for branching morphogenesis.

INTRODUCTION

Branching morphogenesis underlies the formation of all branched tubular structures like the developing lung, mammary gland, and kidney (1). This process requires tightly coordinated cell movement and rearrangement (2). While extensive information exists on how growth factor–dependent signaling regulates branching morphogenesis (3), much less is known about the role of cell–extracellular matrix (ECM) interactions and actin dynamics during this process, especially in kidney development (4, 5).

The kidney collecting duct (CD) system is formed from the ureteric bud (UB), a single-layered epithelial tube with a lumen that grows out of the nephric duct into the metanephric mesenchyme (MM) (6, 7). The mouse UB tips undergo multiple cycles of branching, and the newly formed tubules undergo extensive thinning and elongation giving rise to the CDs (8–10). The mechanisms and molecules that drive this complex morphogenic process are still poorly understood.

Cell-ECM interactions mediated by integrins are critical for UB branching morphogenesis. Integrins are transmembrane α/β heterodimers that link extracellular cues to intracellular processes like signal transduction and cytoskeletal remodeling, which allow them to mediate cell functions like adhesion, migration, and cytoskeletal organization. Integrin function is mediated by the extracellular domain binding to ECM and the cytoplasmic tails recruiting scaffold proteins (11) including the ILK/Pinch/parvin (IPP) complex (12, 13). ILK (14), the best studied component of this complex, promotes integrin-dependent adhesion and migration (15, 16) and reduces actomyosin contractility (17, 18). Parvins (14, 19) are a three-member protein family consisting of the ubiquitously expressed α-parvin; β-parvin, which is primarily expressed in heart and skeletal muscle; and hematopoietic γ-parvin (20). In addition to binding to ILK, parvins interact with filamentous actin (F-actin) (21) and actin-regulatory proteins including regulators of Rho guanosine triphosphatases (GTPases) (22, 23). They are thought to have similar functions to ILK (24, 25). Germ line deletion of α-parvin in mice led to kidney agenesis (26); however, the role of α-parvin in UB branching morphogenesis is still unknown.

In this study, we used state-of-the-art three-dimensional (3D) live-organ imaging to track individual cell behaviors in whole kidneys, to show that α-parvin promotes coordinated cell rearrangement that drives both UB branching and tubule thinning. Unexpectedly, loss of α-parvin caused increased cell adhesion, aberrant migration, and excessive F-actin formation, which was unlike the phenotype seen when β1 integrins or ILK were deleted in the developing UB. Mechanistically, α-parvin promoted actin turnover by inhibiting RhoA and Cdc42 activation, which in turn enhanced the actin-severing function of cofilin. These findings uncover unexpected divergent roles between α-parvin and ILK during epithelial morphogenesis and provide insights into how various components within adhesion complexes differentially regulate actin dynamics during tissue morphogenesis. It also underscores the general importance of regulated actin remodeling in branching morphogenesis, a process fundamental to the development of multiple organ systems including the kidney, lung, mammary gland, and pancreas.

RESULTS

Deleting α-parvin in the UB causes end stage kidney failure

Deleting α-parvin resulted in renal agenesis, which suggested that it is expressed early in the developing kidney (26). We verified that α-parvin is present in both the stalk and tips of the developing UB at embryonic day 12.5 (E12.5) (fig. S1A), as well as in tubules and glomeruli (derived from the MM) of newborn mice (fig. S1B). To identify the role of α-parvin specifically in the UB, we selectively deleted it by crossing floxed α-parvin mice (α-pv^f/f) with HoxB7^Cre mice, which express cre recombinase in the Wolffian duct and UB starting at E10.5 (27). We confirmed α-parvin deletion in the cytokeratin-positive CDs (which are derived from the UB) in the postnatal day 1 Hoxb7:α-pv^f/f mice (Fig. 1A) and the developing UB at E12.5 (fig. S1A). The Hoxb7:α-pv^f/f mice were born in the predicted Mendelian ratio (fig. S1C); however, most of them died by 6 months of age (Fig. 1B). Kidneys isolated from the Hoxb7:α-pv^f/f mice just before death were hypoplastic and dysmorphic, and there were large areas of inflammation with tissue destruction in both the cortex and medulla, making it difficult to identify mechanisms for the phenotype (Fig. 1C). We therefore examined 2-month-old mice, where the tissue destruction was less severe. At this age, Hoxb7:α-pv^f/f mice already had severe renal dysfunction with serum blood urea nitrogen (BUN) levels significantly higher than in α-pv^f/f mice (fig. S1D). The kidneys were significantly smaller than the α-pv^f/f kidneys and the ureters inserted into the bladder with no evidence of ureteric obstruction (Fig. 1D). The papillae in these mice were hypoplastic (Fig. 1E), and both the cortical and medullary CDs were disordered and dilated (Fig. 1E). Inflammatory cell infiltration was present (Fig. 1E), and fibrosis, quantified using Sirius red staining, was significantly increased (Fig. 1, F and G). There were decreased AQP2-positive tubules in the Hoxb7:α-pv^f/f kidneys (Fig. 1H), suggesting CD development defects. Thus, deleting α-parvin results in an end stage kidney characterized by a hypoplastic collecting system, loss of various kidney structures, and severe fibrosis and inflammation in both the kidney cortex and medulla.

Fig. 1. Deleting α-parvin in the UB results in a branching morphogenesis defect and kidney failure.

(A) α-Parvin immunostaining (magenta) of kidney sections of newborn α-pv^f/f and Hoxb7:α-pv^f/f mice. CDs were labeled with Dolichos Biflorus agglutinin (DBA). Arrows demarcate the basolateral sides of the CD and (G) the glomerulus. n = 3 mice per group. (B) Kaplan-Meier curve analysis. 100% of Hoxb7:α-pv^f/f mice die by 6 months, with no mortality in α-pv^f/f mice (P < 0.0001). (C) Hematoxylin and eosin (H&E) staining of 4-month-old α-pv^f/f and Hoxb7:α-pv^f/f kidney sections. n = 3 mice per group. (D) Dissected kidneys with ureters and bladder demonstrate similar phenotypes in 2-month-old α-pv^f/f and Hoxb7:α-pv^f/f mice (top). Total kidney weight (bottom right) and kidney-to-body weight ratios (bottom left) of 2-month-old mice revealed significantly smaller kidneys in Hoxb7:α-pv^f/f compared to α-pv^f/f mice (α-pv^f/f, n = 9; Hoxb7:α-pv^f/f, n = 8). Data are shown as means ± SEM. ****P < 0.0001; two-tailed t test. (E) H&E staining of kidneys from 2-month-old α-pv^f/f and Hoxb7:α-pv^f/f mice. Low magnification (left) demonstrated fibrosis and hypoplasia/dysplasia of the Hoxb7:α-pv^f/f kidneys. Higher magnification of the cortex including glomeruli (G) (middle) and medulla (right) demonstrated disorganized CDs and excessive inflammatory cells (arrowheads) in Hoxb7:α-pv^f/f kidneys. n = 8 mice per group. (F and G) Picrosirius red staining of 2-month-old kidneys revealed more fibrosis in Hoxb7:α-pv^f/f mice compared to α-pv^f/f mice (F). The red stained area of five high-power fields per sample was quantified. The data are expressed as means ± SEM. n = 8 mice (G), ****P < 0.0001; two-tailed t test. (H) Representative immunostaining of kidneys from 2-month-old α-pv^f/f and Hoxb7:α-pv^f/f mice for AQP2 revealed less CDs in Hoxb7:α-pv^f/f kidneys. n = 3 mice.

Deleting α-parvin in the UB causes a branching morphogenesis defect

To define whether the hypoplastic papilla was caused by a UB branching morphogenesis defect, we dissected embryonic kidneys at E10.5 (soon after the initiation of UB development) (8) and visualized the developing UB with E-cadherin and the developing MM with Six2. The UB in the α-pv^f/f mice invaded the MM, and an initial branching step was observed at E10.5. This process was delayed in the Hoxb7:α-pv^f/f mice where, although the UB invaded the MM, it did not branch (Fig. 2A). This branching defect was confirmed in E12.5 kidneys, where the α-pv^f/f mice had approximately twice the number of branch tips compared to the Hoxb7:α-pv^f/f UBs (Fig. 2A) (28). Although Six2⁺ cap mesenchyme was reduced in the Hoxb7:α-pv^f/f kidneys, Six2⁺ nephron progenitors were clearly present at the UB tips, suggesting that the Six2⁺ progenitors were maintained and appropriately associated with the branching UB (Fig. 2A). Histological analysis of Hoxb7:α-pv^f/f embryonic kidneys at E15.5 revealed hypoplastic kidneys with a notable reduction in MM volume and UB branching compared to α-pv^f/f controls (Fig. 2B). A similar but more pronounced phenotype was seen at E18.5, where the mutant kidneys were markedly smaller with a hypoplastic papilla, medulla, and cortex (Fig. 2C). There did not appear to be abnormalities in development of the vasculature as the CD31⁺ endothelial network is present and shows a similar density in developing glomeruli and throughout the parenchyma in both E15.5 Hoxb7:α-pv^f/f and Hoxb7:α-pv^f/f kidneys (fig. S2A). Similarly, stromal organization, as assessed by platelet-derived growth factor receptor β staining in E18.5 Hoxb7:α-pv^f/f and Hoxb7:α-pv^f/f kidneys, was similar (fig. S2B). Last, the mutant kidneys did not have any evidence of ureteric obstruction (fig. S2C). Thus, α-parvin deletion in the UB primarily disrupts UB morphogenesis and nephron progenitor development, but endothelial and stromal compartment formation and patterning are largely preserved.

Fig. 2. Deleting α-parvin in the UB causes a branching morphogenesis and tubule thinning defect.

(A) Whole mounts of E10.5 (left) and E12.5 (right) kidneys of α-pv^f/f and Hoxb7:α-pv^f/f mice were stained with E-cadherin (indicating UB) and Six2 (indicating the MM) and imaged with confocal microscopy. Arrows show UB invasion. Branches number was significantly less in E12.5 Hoxb7:α-pv^f/f kidneys (α-pv^f/f, n = 6; Hoxb7:α-pv^f/f, n = 5). **P < 0.01; two-tailed t test. (B) H&E-stained E15.5 kidneys demonstrated a branching defect and decreased MM development in the Hoxb7:α-pv^f/f mice. Hoxb7:α-pv^f/f kidneys were smaller in total area (top) but not length (bottom) than α-pv^f/f kidneys (α-pv^f/f, n = 10; Hoxb7:α-pv^f/f, n = 5). Data are means ± SEM. *P < 0.05; two-tailed t test. (C) H&E stained E18.5 kidneys demonstrated a branching and MM induction defect in the mutant mice. A disorganized and hypercellular CD was present in Hoxb7:α-pv^f/f mice. Hoxb7:α-pv^f/f were smaller in both total area (top) and length (bottom) than α-pv^f/f kidneys (α-pv^f/f, n = 13; Hoxb7:α-pv^f/f, n = 8). Data are shown as means ± SEM. ****P < 0.0001; two-tailed t test. (D and E) Frozen sections of Hoxb7:α-pv^f/f newborn kidneys labeled with DBA show widened CDs (D). CD tubule diameters are shown as means ± SEM (D) and distribution plots (E). Tubules were pooled from three mice per genotype, ****P < 0.0001; two-tailed t test. (F to H) Thick, frozen kidney sections of neonatal α-pv^f/f and Hoxb7:α-pv^f/f mice were immunolabeled with DBA and 4′,6-diamidino-2-phenylindole (DAPI) and imaged by confocal microscopy in 3D. Tubules only contain a single cell. The height (x-y plane) and number (z plane) of cells were counted in the tubules. Quantifications of cell height (G) and cell number (H) were presented as means ± SEM. Measurements were accumulated over three individual mice per genotype with at least five tubules analyzed, ****P < 0.0001; two-tailed t test.

One of the major deficiencies noted in the Hoxb7:α-pv^f/f kidneys was that the typical monolayered epithelial cell CD structure was disrupted, and cells appeared to pile up on each other within the CD lumens, resulting in enlarged tubules in the Hoxb7:α-pv^f/f mice when compared to the α-pv^f/f mice (Fig. 2C). When we quantified the size of the CDs by measuring their diameters in kidney cross sections, the average diameter in Hoxb7:α-pv^f/f kidneys was 30.55 ± 9.3 μm compared to 20.91 ± 10.51 μm in the α-pv^f/f kidneys, and there was a right shift of the distribution of tubular size with an increased diameter in the mutant mice (Fig. 2, D and E). To determine whether this expanded tubular size was due to increased cellular proliferation, we labeled the CD with Ki67. We found a significant decrease in cellular proliferation within the developing Hoxb7:α-pv^f/f CDs that were identified by cytokeratin staining (fig. S2D), which suggests that CD hyperproliferation was not the cause of the luminal hypercellularity. To investigate whether the widened CDs in the Hoxb7:α-pv^f/f kidneys were due to abnormal multilayering of the epithelium, we performed 3D confocal imaging of these tubules in thick frozen sections. The orthogonal views (z axis) (Fig. 2F) revealed that the tubules were a monolayer; however, the cells were significantly taller (Fig. 2, F and G), and there was a significant increase in the number of cells per cross section (Fig. 2H) in the Hoxb7:α-pv^f/f kidneys when compared to the α-pv^f/f kidneys. These data suggest that the widening of the tubules was due to a morphological defect of the CD cells. Thus, deleting α-parvin in the UB appears to cause abnormal epithelial cell organization resulting in the inability to branch and a failure in establishing normal tubular geometry.

α-Parvin regulates CD cell neighbor exchange during UB branching

To determine the cellular mechanisms underlying the defective branching morphogenesis and tubule thinning of the UB the Hoxb7:α-pv^f/f mice, we used mTmG mice in which cells initially express the membrane-targeted red fluorescent protein “mT” but, upon Cre-mediated recombination, switches to express the membrane-targeted green fluorescent protein (GFP) “mG” (29). We crossed mTmG mice with Hoxb7:α-pv^f/f transgenic mice and performed volumetric confocal imaging at regular intervals. This allowed us to identify and track individual UB cells over an extended period within ex vivo–cultured kidneys using a methodology adopted from salivary gland (30) and embryonic kidney cultures (31). We isolated kidneys from these mice at E12.5, grew them on transwells (30, 31) (Fig. 3A and movies S1 and S2), and used 3D confocal microscopy to capture branching morphogenesis at 10-min intervals for a total of 22 hours. As expected, the UB tip in the α-pv^f/f kidney first enlarged to form an ampulla and then underwent remodeling to form two tips (Fig. 3B, top). In contrast, the UB from Hoxb7:α-pv^f/f mice showed delayed ampulla enlargement and decreased trunk remodeling resulting in wide unbranched tubules (Fig. 3B, bottom). To define the dynamics of tubule remodeling in greater detail, we examined trunk narrowing using orthogonal views of the time-lapse z-stacks (Fig. 3C, orange line, and movies S3 and S4). The tubules were aligned with the x axis, and the cross-sectional area was measured over time. The elongating tubules in the α-pv^f/f mice narrowed 50% over time, while the tubule diameter in the Hoxb7:α-pv^f/f mice did not change over the same time (Fig. 3, C and D).

Fig. 3. CD thinning defect in α-parvin–null mice is due to abnormal cell rearrangements.

(A) Sample preparation for live imaging of embryonic kidneys. (B) Confocal image stacks of E11.5 kidneys cultured ex vivo. h/hr, hours. (C and D) 3D reconstruction of time-lapse confocal microscopy images of UBs. The orange line depicts the plane of cross-sectional analysis (C). The relative change in cross-sectional area over 22 hours is compared. n = 3, (means ± SEM), **P < 0.01; two-way analysis of variance (ANOVA) followed by Tukey’s test (D). (E) A defined region within the trunk (orange box) was extracted from z-stack confocal images, rendered in 3D and analyzed for cell rearrangement. (F to H) Rosette-like cell clusters (pseudocolored orange) in CDs followed over 6 hours (F). The standard distance was used to quantify the extent of cell dispersion during rearrangement (G). Standard distance for the cell clusters was plotted against elapsed time with a linear regression fit applied to the data. The P value indicates the slope difference of the two linear regression fits using the analysis of covariance (ANCOVA). Four independent UB trunks, with a minimum of four cell clusters from each trunk were analyzed (means ± SEM) (H). (I and J) Particle intensity velocity (PIV) analysis of the 3D live images. Representative images of velocity vector magnitude and orientation (left) and quantification of velocity (right). Note reduced velocity and lack of global directionality of vectors [n = 5 (control), 4 (Hoxb7^f/f:α-pv^f/f)] (I). Representative images of motion similarity heatmaps (left) and quantification of motion coherence (right) from live imaging. Three equidistant parallel lines were drawn across the length of the tissue to determine the average and standard deviation of motion similarity. Note lack of persistent and coordinated global motion across the tissue in Hoxb7:α-pv^f/f relative to control n = 5 (control), 4 (Hoxb7:α-pv^f/f) (J).

Tubular elongation and narrowing occur either by oriented cell division, as found in late CD development (9), or by luminal mitosis like in UB branching tips (9, 32). To define which of these mechanisms occurred in the early elongating UB trunk, we followed dividing cells over time from our time-lapse imaging of the newly formed E12.5 trunks. We found that cell division was not oriented with the longitudinal axis of the tubule in either α-pv^f/f or Hoxb7:α-pv^f/f UB trunks (fig. S3, A to C). However, we did observe that the dividing cells underwent luminal mitosis (32), a process where cells delaminate from the epithelium and divide within the lumen, in both α-pv^f/f and Hoxb7:α-pv^f/f CDs, as occurs in UB tips (fig. S3, D and E). Thus, normal UB trunk elongation occurs by luminal mitosis, and this process is not impaired in Hoxb7:α-pv^f/f UB trunks. Furthermore, these results suggest that cell rearrangement plays a role during the thinning process.

Cell rearrangement, a conserved feature of tubular epithelial morphogenesis (33, 34), is required for UB development. This process is characterized by the formation and dispersal of multicellular rosette structures (35), where cells rapidly exchange their neighbors by performing controlled epithelial locomotion (36). To measure the cell rearrangement, we marked multicellular rosettes within the tubule trunk (Fig. 3E) and tracked the position of the individual cells in the rosettes over time (Fig. 3F and movies S5 and S6). Rosettes were defined as a configuration of at least five cells that shared a vertex, and spatial cell dispersal was measured and quantified (see Materials and Methods and Fig. 3, G and H). There was a significant decrease in cell motility and neighbor exchange in Hoxb7:α-pv^f/f mice over 6 hours (Fig. 3H).

The reduced neighbor exchanges suggested that the α-parvin–deficient tissue is in a less fluid-like state than the control tissue (37). To infer the mechanical phase state of the tissue, we quantified coarse-grained velocity fields across the tissue (38). The average magnitude of the velocity field indicated slower motility of Hoxb7:α-pv^f/f tissues compared to α-pv^f/f mice (Fig. 3I and movies S7 and S8). Probing the local directions and alignment of velocity vectors over time (motion similarity; Fig. 3I and movies S9 and S10) further revealed coordinated cellular motion in α-pv^f/f that persisted over time, characteristic of fluid-like behavior. In contrast, Hoxb7:α-pv^f/f tissues did not maintain persistent, coordinated motion across the length of the tissue, characteristic of a jammed or solid-like state and consistent with the slower growth-rate of the tissues (Fig. 3, I and J) (39). Together, these results indicate a decrease in neighbor exchanges and persistent motion of the Hoxb7:α-pv^f/f UB, resulting in decreased tubule thinning.

Deleting α-parvin results in excessive F-actin in CDs

The glial cell line–derived neurotrophic factor (GDNF)/Ret receptor signaling axis is required for early UB branching and elongation (40, 41). To address whether this signaling was perturbed in Hoxb7:α-pv^f/f mice, we measured Ret mRNA expression using RNA scope and found no difference in Ret expression between Hoxb7:α-pv^f/f and α-pv^f/f E15.5 kidneys. (fig. S4A). We next examined the integrity of Ret signaling by adding exogenous GDNF to E11.5 ex vivo kidney cultures. GDNF induced ectopic branching (denoted by arrows) in both α-pv^f/f and Hoxb7:α-pv^f/f kidneys, although as expected, UB branching was more robust in the wild-type (WT) kidneys (fig. S4B). Last, we examined mitogen-activated protein kinase (MAPK) activation to determine whether signaling downstream of growth factor receptors like Ret and fibroblast growth factor receptors was intact. p-ERK staining in the UB epithelium of both α-pv^f/f and Hoxb7:α-pv^f/f E15.5 kidneys was similar (fig. S4C). These results suggest that α-parvin–deficient UBs retain the ability to transduce GDNF-dependent signaling.

As α-parvin can regulate cell-cell junctions (42), we assessed whether there were polarity defects in the Hoxb7:α-pv^f/f CDs by examining ZO-1 (Fig. 4A) and E-cadherin (Fig. 4B) distribution using immunofluorescence. No abnormalities in localization or amount of these polarity markers were noted. In addition, the localization of the integrin β1 subunit (Fig. 4C) and the basement membrane as depicted by laminin staining (Fig. 4D) were normal. Thus, the phenotype was not caused by polarity defects. We next investigated the actin cytoskeleton by culturing isolated E11.5 kidneys, and whole mount E-cadherin staining was used to identify the UB (Fig. 4E, a and a′). The actin cytoskeleton was stained with rhodamine-conjugated phalloidin, after which the tissue was subjected to confocal microscopy (Fig. 4E, b to f and b′ to f′). Analysis of actin distribution was conducted using a 1-μm z-step across the entire UB tip. We reconstructed the images computationally in 3D to highlight the tubule outline (Fig. 4E, b and b′). An enlarged tip was evident at the end of a slender trunk in the α-pv^f/f UB. However, there was no discernible trunk, and the tip was notably larger in the Hoxb7:α-pv^f/f UB (Fig. 4E, c and c′). Actin filaments were observed on both the apical and basal-lateral borders of the UB cells of the α-pv^f/f and Hoxb7:α-pv^f/f mice, but they appeared thicker on the basal side of Hoxb7:α-pv^f/f CD cells compared to those of α-pv^f/f CD cells (Fig. 4E, d and d′). When we specifically examined the basal side of the CD cells, we found a clear circumferential actin belt in the α-pv^f/f UB cells (Fig. 4E, e and f). By contrast, pronounced F-actin bundles were observed on the basal side of Hoxb7:α-pv^f/f UB tip (Fig. 4E, e′ and f′). Thus, there appears to be excessive F-actin on the basal sides of UB cells, which suggests that α-parvin promotes actin depolymerization in these epithelial cells.

Fig. 4. Deleting α-parvin in the UB causes excessive F-actin formation on the basal side of CDs.

(A to D) The CDs of E18.5 α-pv^f/f and Hoxb7:α-pv^f/f mice were labeled with DBA and stained for ZO-1 (A), E-cadherin (B), integrin β1 (C), or laminin (D) and imaged with confocal microscopy. ZO-1 images were shown as maximum intensity projections (MAXIP) to visualize apical junctions. A white dashed line is used to outline the basal side of the tubule. Images are representative of three mice per genotype analyzed. (E) E11.5 kidneys isolated from α-pv^f/f and Hoxb7:α-pv^f/f mice were cultured for 12 hours and labeled with E-cadherin (green, 3D rendering) (a and a′) and stained with rhodamine-phalloidin (white) (b and f and b′ and f′), after which they were subjected to confocal microscopy analysis. 3D reconstruction of one branching tip is shown (b and b′). The tip surface (outlined) was subjected to 3D rendering using Imaris 10.0 (c and c′). Apical (d and d′) and basal (e and e′) actin distribution is shown. Basal F-actin stress fibers are demonstrated at higher magnification (f and f′) in the Hoxb7:α-pv^f/f mice. Images are representative of three mice per genotype that were analyzed.

α-Parvin^−/− CD cells have a severe in vitro tubulogenesis defect

To investigate the cellular mechanisms whereby α-parvin regulates the actin cytoskeleton, we isolated α-pv^f/f CD cells from 8-week-old α-pv^f/f mice and deleted α-parvin in vitro using adenovirus-mediated delivery of a Cre recombinase as previously described (43). α-parvin deletion (α-pv^−/−) was confirmed by immunoblotting (Fig. 5A) and immunofluorescence (fig. S5A). As we previously showed that deleting ILK in CD cells results in diminished α-parvin but normal Pinch expression, we determined ILK and Pinch expression in these cells. In contrast to the ILK^−/− CD cells (24), there was no difference in levels of either ILK or Pinch in α-pv^f/f and α-pv^−/− CD cells and they were localized in a focal adhesion (FA)–like pattern at the peripheral end of actin fibers (fig. S5, B and C). We also assessed the effects of deleting α-parvin on the formation of the IPP complex in CD cells. Immunoprecipitation of α-parvin in confluent α-pv^f/f CD cells verified that the IPP complex was formed in control cells but as expected, ILK and Pinch were not pulled down in α-pv^−/− CD cells (Fig. 5B). When we performed ILK immunoprecipitation, Pinch was pulled down in both α-pv^f/f and α-pv^−/− CD cells and there was more Pinch binding to ILK in the α-pv^−/− CD cells (Fig. 5, C and D).

Fig. 5. α-pv^−/− CD cells have normal ILK-Pinch complex formation but defects in cell function due to increased F-actin formation.

(A) Immunoblots for α-parvin, ILK, and Pinch. Relative ratios shown (right). n = 5 means ± SEM. ****P < 0.0001; two-tailed t test. (B to D) Cell lysates immunoprecipitated with α-parvin (B) or ILK antibodies (C) blotted for α-parvin, ILK, and Pinch. Relative ratios are shown (D) Means ± SEM. n = 4, **P < 0.01; two-tailed t test. (E) α-pv^−/− and α-pv^f/f CD cells grown in 3D collagen-1/Matrigel gels and stained with rhodamine-phalloidin. (F) α-pv^−/− and α-pv^f/f CD cells subjected to a linear scratch and followed over time. Percentages of closure are shown. n = 4 (means ± SEM), *P < 0.05, **P < 0.01, ****P < 0.0001; two-way ANOVA and Tukey’s test. (G) CD cell adhesion on laminin-511. n = 3 (means ± SEM), **P < 0.01, ****P < 0.0001; two-way ANOVA and Tukey’s test. (H) CD cells spread for 1 hour on laminin-511 stained with rhodamine-phalloidin and DAPI. (I) CD cell area (left) and circularity (right). Means ± SEM of n > 100 cells per group, ****P < 0.0001; two-tailed t test. (J) CD cells that transmigrate through 8-μm transwells coated with laminin-511 (10 μg/ml) over 4 hours. Five high-power fields per group. n = 5 (means ± SEM), ***P < 0.001; two-tailed t test. (K) CD cell migration on laminin-511 (10 μg/ml)–coated slides tracked for 7 hours. Fifteen individual cells shown (Ieft) and used for velocity analysis (right). Means ± SEM, ****P < 0.0001; two-tailed t test. (L) CD cells grown to confluent monolayers on transwells, stained with rhodamine-phalloidin, and analyzed by confocal microscopy. (M) Representative Western blots for G-actin and F-actin content in CD cells. The ratio of G-actin relative to F-actin is shown. n = 4 (means ± SEM), ****P < 0.0001; two-tailed t test.

Next, we analyzed tubulogenesis in vitro by plating cells into collagen/Matrigel gels. The α-pv^f/f CD cells organized into branched tubules after 7 days of culturing; however, the α-pv^−/− CD cells remained as single cells (Fig. 5E). To explain this phenotype, we initially assessed the ability of α-pv^−/− CD cells to undergo collective epithelial migration, a requirement for tubulogenesis. They were only able to close 30% of a wound relative to α-pv^f/f CD cells over a 24-hour period (Fig. 5F). To examine the mechanisms for this poor wound closure, we investigated whether the α-pv^−/− CD cells had adhesion, spreading, or migration defects. Contrary to expectations, α-pv^−/− CD cells exhibited a marked increase in adhesion to laminin 511 (Fig. 5G) when compared to α-pv^f/f CD cells. They also spread significantly more, formed pronounced filopodia leading to decreased circularity (Fig. 5, H and I) and developed significantly larger FAs (fig. S5D). In agreement with the excessive adhesion data, α-pv^−/− CD cells were more attracted toward laminin 511 (Fig. 5J) in a transwell migration assay. However, consistent with the collective migration data, the α-pv^−/− CD cells plated on laminin 511–coated glass slides moved less (Fig. 5K). We also found a major proliferation defect in the α-pv^−/− CD cells (fig. S5E). Because of the abnormalities in cell spreading, adhesion, and motility, as well as excess basal F-actin formation in the in vivo kidney, we hypothesized that α-pv^−/− CD cells would have a defect in the actin cytoskeleton. We therefore stained F-actin in CD cells polarized on transwells with rhodamine phalloidin and visualized them by confocal microscopy. The α-pv^f/f cells formed normal actin belts at the cell-cell junction, while α-pv^−/− CD cells had excessive F-actin bundles (Fig. 5L). When we measured the relative amounts of G versus F actin, there was a decreased G-to-F-actin ratio in the α-pv^−/− CD cells (Fig. 5M). Together, these data indicate that α-parvin enhances epithelial tube formation by regulating actin turnover during cell motility.

Deleting α-parvin does not affect integrin-dependent signaling

Since integrins are required for cell adhesion, regulate adhesion-dependent signaling pathways, and bind to the IPP complex, we initially assessed whether deleting α-parvin altered integrin surface expression by performing flow cytometry of the β1 integrin subunit and its major binding partners α1, α2 (collagen binding receptors), and α6 (laminin binding receptor). We found no differences between α-pv^f/f and α-pv^−/− CD cells (fig. S6A). We next performed in vitro replating assays on laminin 511 to assess the activation of focal adhesion kinase (FAK), paxillin, Akt, ERK, and p38MAPK signaling in α-pv^−/− cells. Adhesion of both α-pv^f/f and α-pv^−/− CD cells induced a significant increase in phosphorylation of FAK, paxillin, Akt, and ERK; however, there were no major differences between the two genotypes (fig. S6, B and C). Unexpectedly, there was a higher p38 activity in α-pv^−/− CD cells at baseline and after adhesion. To test whether this increased p38 played any role in sheet migration, we exposed both α-pv^f/f and α-pv^−/− CD cells to a p38 inhibitor before performing the assay. It did not change the response of either of the cells, suggesting that it had little or no role in this cell function (fig. S6D). Thus, the increased cell adhesion, spreading, and migration, as well as the decreased random migration observed in the α-pv^−/− CD cells are not due to excessive integrin surface expression, activation, or integrin-dependent signaling abnormalities and suggest that the defects are primarily within the actin cytoskeleton.

Deleting α-parvin increases RhoA and Cdc42 and decreases LIMK-dependent cofilin activity

Deleting α-parvin in vascular smooth muscle cells results in actin cytoskeleton defects due to excessive activation of RhoA GTPase (18). We therefore tested the activation status of the Rho family of GTPases (Rho, Cdc42 and Rac1) in α-pv^f/f and α-pv^−/− CD cells. We found increased activation of RhoA as well as increased expression and activation of Cdc42 (Fig. 6, A and B) in α-pv^−/− CD cells. Because the Rho family of GTPases regulates F-actin formation (44) by inhibiting the actin depolymerizing factor cofilin (45, 46), we next defined whether α-parvin controls actin dynamics via cofilin. Cofilin phosphorylation at Ser-3 results in inactivation of its actin-binding and depolymerization activity (47). We therefore measured cofilin phosphorylation of α-pv^f/f and α-pv^−/− CD cells at baseline and after they adhered to collagen I. There was increased cofilin phosphorylation in nonadherent α-pv^−/− compared to α-pv^f/f CD cells, and this difference was accentuated after the cells adhered to laminin 511 (Fig. 6C). Consistent with the data from CD cells, papilla from α-parvin–null mice demonstrated increased cofilin phosphorylation (fig. S7A).

Fig. 6. α-pv^−/− CD cells have increased RhoA and Cdc42 activity, which regulate cofilin phosphorylation.

(A and B) Total and guanosine triphosphate (GTP) bound RhoA and CdC42 of α-pv^f/f and α-pv^−/− CD cells analyzed by immunoblotting. The ratio is calculated using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) to verify loading. n = 3 (means ± SEM), *P < 0.05, ***P < 0.001; two-tailed t test. (C) Suspended and laminin-plated α-pv^f/f and α-pv^−/− CD cells were immunoblotted for phosphorylated (p-) and total cofilin. n = 3 means ± SEM (bottom). ***P < 0.0001, ****P < 0.0001; two-way ANOVA and Tukey’s test. (D) α-pv^−/− and α-pv^f/f CD cells subjected to a linear scratch in the presence of CT04 (2 μg/ml); Y-27632 (20 μM); ML141, (10 μM). Percentage wound closure over 24 hours shown. n = 4 (means ± SEM), **P < 0.01, ****P < 0.0001; two-way ANOVA and Tukey’s test. DMSO, dimethyl sulfoxide. (E) Transwell migration of α-pv^f/f and α-pv^−/− CD cells with inhibitors described in (D). n = 5 (means ± SEM), ****P < 0.0001; two-way ANOVA and Tukey’s test. (F) Effects of inhibitors described in (D) on CD cell spreading on laminin-511. Means ± SEM, n > 100 cells. **P < 0.01, ****P < 0.0001; two-way ANOVA and Tukey’s test. (G) Wound healing over 24 hours in the presence of CK666 (100 μM) and blebbistatin (10 μM). n = 4 (means ± SEM), ****P < 0.0001; two-way ANOVA and Tukey’s test. (H) Transwell migration of CD cells treated with inhibitors described in (G) or blebbistatin (20 μM). n = 5 (means ± SEM). *P < 0.05, ****P < 0.0001; two-way ANOVA and Tukey’s test. (I) Effect of inhibitors described in (H) on CD cell spreading on laminin 511. Spreading area and circularity are shown as means ± SEM of n > 100 cells per group. ****P < 0.0001; two-way ANOVA followed by Tukey’s test.

We assessed the functional relevance of the increased activity of these Rho GTPases by treating the α-pv^f/f and α-pv^−/− CD cells with the Rho inhibitor CT04 (C3 exoenzyme analog), the ROCK inhibitor (Y-27632), and the Cdc42 inhibitor (ML141). CT04 did not alter the ability of α-pv^−/− CD cells to close a wound (Fig. 6D) but decreased transwell migration (Fig. 6E and fig. S7B). It also partially rescued the increased cell spreading and decreased circularity (Fig. 6F). The ROCK inhibitor induced wound closure (Fig. 6D) and partially decreased transwell migration (Fig. 6E and fig. S7B), cell spreading, and circularity (Fig. 6F). By contrast, the Cdc42 inhibitor ML141 significantly increased wound closure (Fig. 6D), reduced transwell migration (Fig. 6E and fig. S7B) and cell spreading (Fig. 6F), and reverted the abnormal cell shape (Fig. 6F). Cofilin phosphorylation was diminished by decreasing the activity of Rho A (CT104) and ROCK (Y-27632), and this was especially evident when Cdc42 (ML-141) was inhibited (fig. S7, C to E). To examine whether there was cross-talk between the Rho and Cdc42 pathways in mediating their effects on α-pv^−/− CD cells, we inhibited both pathways by adding CT04 and ML141 together. There was no additive effect on cell spreading, circularity, or migration (fig. S7, F to J). As expected, adding both CT04 and ML141 caused a significant decrease in cofilin phosphorylation (fig. S7K). Thus, the Rho GTPases, especially Cdc42, play a major role in α-parvin–dependent regulation of cell shape and movement, and this activity is mediated by modulating cofilin-dependent actin-binding and depolymerization.

One of the key mechanisms whereby the Rho family GTPases regulates cofilin activation is via the serine-threonine LIM kinases, which directly phosphorylate and inactivate members of the cofilin family (45, 46, 48, 49). To test whether this mechanism accounted for the effects of the increased Rho GTPase signaling found in the α-pv^−/− CD cells, we blocked LIM kinases 1 and 2 with the LIMKi 3 inhibitor. This resulted in a decrease in cell area and circularity in both WT and α-pv^−/− CD cells; however, the difference was more pronounced in the mutant cells (fig. S8, A to C). A similar result was seen with respect to cell migration (fig. S8, D and E). As expected, the inhibitor markedly degreased cofilin phosphorylation in both α-pv^−/− CD cells and WT CD cells (fig. S8F). Together, these results suggest that the excessive activation of Rho GTPases in α-pv^−/− CD cells mediate their effects in a large part by regulating LIM kinase–dependent inactivation of cofilin.

α-Parvin–dependent inactivation of cofilin induces excessive actin polymerization

Our data thus far suggest that the cytoskeletal defects observed in the α-pv^−/− CD cells are mediated by RhoA and Cdc42 activation, which in turn affects cofilin-dependent actin depolymerization. However, it is possible that the cytoskeletal changes are mediated by abnormal actomyosin contractility (50, 51). To define which of these mechanisms accounted for the α-pv^−/− CD cell phenotype, we examined the effects of cytochalasin D (a cell-permeable potent inhibitor of actin polymerization), CK666 (an Arp2/3 complex inhibitor), and blebbistatin (a myosin II inhibitor) on wound healing, CD cell spreading, and transwell migration. The effects of cytochalasin D on wound healing was not performed as prolonged exposure to cells induced cell death. However, CK666 stopped all wound healing and blebbistatin did not alter the assay for both α-pv^−/− CD cells and WT CD cells (Fig. 6G). Cytochalasin D and CK666 decreased the abnormal α-pv^−/− CD cell migration (Fig. 6H and fig. S8G), decreased spreading, and increased circularity in α-pv^f/f CD cells (Fig. 6I). By contrast, blebbistatin did not reduce cell spreading or increase circularity (Fig. 6I) in α-pv^−/− CD cells, although it decreased the transwell migration (Fig. 6H and fig. S8G). Blebbistatin did not change cofilin phosphorylation in α-pv^f/f or α-pv^−/− CD cells (fig. S8H). These data further suggest that the α-pv^−/− CD cell abnormalities are primarily due to excessive actin polymerization and not increased myosin contractility.

α-Parvin regulates UB branching morphogenesis by restricting Cdc42-dependent cell rearrangement and actin polymerization

To assess the functional significance of the excessive activation of the Rho GTPases in UB branching morphogenesis in Hoxb7:α-pv^f/f mice in vivo, we isolated E12.5 UBs from Hoxb7:α-pv^f/f and α-pv^f/f mice and grew them on transwells for 24 hours in the presence or absence of the Cdc42 inhibitor ML141 or the RhoA inhibitor CT104. ML141 partially reversed the branching defect in the Hoxb7:α-pv^f/f kidneys (Fig. 7A); however, there was no rescue of the RhoA inhibitor CT104 (Fig. 7B) and there was no additive effect of CT104 when added to ML141 (fig. S9A). Blebbistatin did not increase the branching of Hoxb7:α-pv^f/f kidneys (fig. S9B). This suggests that like in the CD cells, UB branching is facilitated by α-parvin–dependent Cdc42 inhibition.

Fig. 7. α-Parvin–dependent Cdc42 activity, epithelial cell movement, and F-actin turnover are required for UB branching morphogenesis.

(A and B) E11.5 kidneys were isolated and grown ex vivo in the presence of a Cdc42 inhibitor (ML141, 10 μM) or a Rho inhibitor (CT04, 2 μg/ml) for 24 hours and imaged by confocal microscopy. There is a partial rescue of the branching defect in the Hoxb7:mTmG^f/f:α-pv^f/f kidneys treated with ML141 (A) but not with CT04 (B). Branching points were counted (means ± SEM) shown on the right panel. n = 7 per group. **P < 0.001; paired two-tailed t test. ns, not significant. (C) Representative time-lapse images showing the dispersal of a rosette-like cell cluster in control and Hoxb7:mTmG^f/f:α-pv^f/f CDs over 2 hours. Cells within the cluster were pseudocolored to visualize cell rearrangement over time. (D) Track speed of UB tip cells was significantly reduced in Hoxb7:mTmG^f/f:α-pv^f/f kidneys compared to the control kidneys and was restored by ML141 treatment. Data are shown as means ± SEM. ****P < 0.0001; two-tailed t test. (E) Normalized standard distance for the cell clusters were plotted against elapsed time (solid line) with a second-order polynomial curve fit (dotted line). n = 3 UB tips per genotype with a minimum of 12 cell clusters from each tip. Data are shown as means ± SEM. **P < 0.01, *P < 0.05; Hotelling’s t² tests. (F) E11.5 kidneys isolated from control and Hoxb7:mTmG^f/f:α-pv^f/f mice were cultured ex vivo with either ML141 (10 μM) or CT04 (2 μg/ml) for 24 hours and imaged by confocal microscopy. UB trunks were visualized by GFP staining, and actin filaments were visualized with rhodamine-conjugated phalloidin. Panels from left to right show the 3D reconstruction of the UB trunk, apical section, basal section, and the higher magnification of the basal actin.

To determine the cellular functions regulated by Cdc42 during UB branching morphogenesis, we performed ex vivo cultures of E12.5 kidneys in the presence of the Cdc42 inhibitor ML141, followed by live imaging for 22 hours. Since ML141 significantly reduced membranous GFP intensity in the UB trunk, we limited our analysis to the UB tips, where GFP signal remained robust, likely due to active GFP turnover and synthesis in the proliferating cells (52). Because of the highly motile nature and frequent mitotic events in tip cells, we focused on a shorter, more informative 2-hour window, in contrast to the 8-hour window used for trunk cells (Fig. 3). We found that there was a significant reduction in cell migration speed in Hoxb7:α-pv^f/f compared to α-pv^f/f tips, and ML141 treatment restored cell migration speed to that of α-pv^f/f tips (Fig. 7, C and D). ML141 modestly decreased migration speed in α-pv^f/f kidneys, suggesting that precise regulation of Cdc42 activity is required for normal cell motility in the UB tip. To measure how groups of cells dispersed over time, we followed small clusters of cells in our time-lapse imaging data (see Materials and Methods) (Fig. 7C). We found that cell dispersal was significantly reduced in Hoxb7:α-pv^f/f compared to α-pv^f/f tips (Fig. 7, C and E). ML141 treatment restored cell dispersal in Hoxb7:α-pv^f/f kidneys back to the levels of α-pv^f/f mice. In contrast to its effect on migration speed, ML141 had minimal impact on cell dispersal in α-pv^f/f kidneys (Fig. 7, C and E), suggesting that cell migration and intercellular movement differ in their sensitivity to low-dose Cdc42 inhibition under normal conditions. Together, these results suggest that excess Cdc42 activity contributes to the migration defect in α-parvin–deficient cells, which results in abnormal dispersal dynamics in the α-parvin–deficient UB.

To investigate whether Cdc42 or RhoA activity contributes to the cytoskeletal disorganization observed in Hoxb7:α-pv^f/f UBs, we treated isolated E12.5 kidneys with the Cdc42 inhibitor ML141 or the Rho inhibitor CT04 for 24 hours and performed super-resolution confocal imaging (Fig. 7F). The whole explants were stained with phalloidin-647, and the actin was imaged on both the apical and basal side of the entire UB epithelium in 3D. As expected, Hoxb7:α-pv^f/f mice exhibited pronounced basal stress fiber formation, with bundled, intersecting F-actin structures visible at the basal side, while control kidneys displayed actin belts and no prominent stress fibers. Strikingly, treatment of α-parvin KO explants with either ML141 or CT04 led to a marked reduction in basal actin stress fiber formation, resulting in an actin architecture that resembled the WT phenotype. Neither inhibitor altered actin organization in control kidneys (fig. S9C). These results indicate that excessive activation of Rho GTPases, particularly Cdc42 and RhoA, contributes to the aberrant basal actin remodeling observed in the KO and suggest that α-parvin regulates UB cytoskeletal architecture through modulation of Rho GTPase activity.

DISCUSSION

In this study, we investigated how the scaffold protein α-parvin contributes to UB development, a complex morphological process that is dependent on coordinated cell rearrangement. We used ex vivo live imaging to show that α-parvin regulates UB cell reorganization by controlling F-actin depolymerization via the Rho GTPase–dependent LIMK/cofilin pathway. This function contrasts with that of ILK (53), the obligatory binding partner of α-parvin, as well as integrins (54), both of which promote actin polymerization (Fig. 8). These results highlight that in addition to morphogen-induced signaling, adhesion protein–dependent regulation of actin turnover that controls epithelial cell movement, is critical for branching morphogenesis.

Fig. 8. Proposed mechanism of α-parvin function in UB branching morphogenesis.

α-Parvin inhibits RhoA and Cdc42 GTPase activity to maintain LIMK/cofilin-dependent actin turnover. This promotes epithelial cell dispersal, a process required for tubule narrowing and subsequent branching morphogenesis.

One of the key obstacles to understanding the cellular mechanisms of UB development is the limited live imaging studies performed on this intricate process. By combining an ex vivo kidney explant culture system with a visible membrane tethered GFP and low phototoxicity Airyscan confocal microscopy, we were able to visualize individual cell dynamics in ex vivo branching UB tubules over an extended period. We show that a rosette-based cell intercalation process occurs in early UB tubule elongation in control mice, like that observed in embryogenesis such as the Drosophila germ band epithelium elongation (35) and organogenesis such as Xenopus kidney development (34). We show that growth and thinning of the UB requires cell movement between neighbors and intercalation along an axis that is perpendicular to the direction of elongation, and loss of α-parvin in the developing UB impaired this process leading to less branch formation and decreased tubular narrowing. These defects were due to slower cellular motility and an inability to maintain persistent, coordinated motion across the length of the tissue. This is characteristic of a jammed or solid-like state and is consistent with the slower growth rate of the UB in the α-parvin–null mice. Tissue fluidity governs its response to internally generated forces, and fluid-like behavior is required for dynamic changes in tissue shape (55, 56). Thus, this change in tissue mechanical properties, in combination with the defects in cell morphology and behavior, is sufficient to explain the defects in morphogenesis observed in the α-parvin–deficient UB. The inability of early α-parvin–null UB trunks to narrow was due to actin cytoskeleton abnormalities rather than a loss of oriented cell division, which is required in later stages of development (9). This finding is consistent with studies conducted in kidneys of mice where both cofilin I and destrin were deleted in the UB using the same Hoxb7 mice (57), and this increased F-actin polymerization together with increased adhesive properties most likely also explains the solidification of the tissue (37). Together, our observations suggest that actin-dependent cell movement is a key requirement for early UB morphogenesis.

Since α-parvin is part of the IPP complex that interacts with integrins (19), we expected the α-parvin–UB-null mice, ILK-UB–null mice, and integrin β1-UB–null mice to have similar phenotypes. While all three mice had a branching morphogenesis defect, there were significant and unexpected differences. The β1 integrin–null mice had the most severe branching morphogenesis defect but normal sized tubules (58). The phenotype was due to a severe adhesion defect and a loss of major anchorage-dependent signaling pathways such as Erk and Akt, which are intact in both ILK (24) and α-parvin–null UBs. The ILK-UB–null mice have a moderate branching defect and marked tubular obstruction (24) due to CD proliferation within the tubular lumens caused by a loss of p38/MAPK-dependent contact inhibition. The α-parvin–UB-null mice also had a moderate UB branching defect; however, the CDs in these mice were broadened because of the inability of epithelial cells to reorganize during UB branching and thinning. These phenotypic differences in the developing UB contrast with studies in keratinocytes (15, 16), vascular smooth muscle cells (17, 18), and muscle attachment in Drosophila (59, 60) and C. elegans (61, 62) where the mechanisms of action of ILK and α-parvin were shown to be similar. This demonstrates the cell type–specific functions of these proteins in different developmental contexts.

It is unclear how the IPP complex is formed and regulates cell function. Deleting ILK results in diminished levels of α-parvin and Pinch, and the IPP complex cannot be formed in multiple cell types (13, 19), including CD cells (24, 25). By contrast, we demonstrate that deleting α-parvin does not change the levels of ILK and results in increased interactions between ILK and Pinch. This suggests that α-parvin is not an obligatory binding partner of ILK and Pinch, and it might negatively regulate ILK binding to Pinch in CD cells.

One of the key effects of deleting α-parvin in CD cells was an increase in RhoA and Cdc42 GTPase activity. A possible mechanism to explain these data is that the increased Pinch-ILK binding induces Rho GTPase activity through Pinch as Pinch2 was demonstrated to activate RhoA and Cdc42 in central nervous system myelination by controlling assembly of the septin protein scaffold (63). It is also conceivable, however, that α-parvin reduces Rho GTPase activity in CD cells by recruiting other Rho-regulating scaffold proteins such as cdGAP (64) or ARHGEF6 (23), which have been shown to bind to α-parvin.

In contrast to the increase in RhoA and Cdc42 GTPase activities in α-parvin in CD cells, ILK-null CD cells have decreased RhoA and unchanged Cdc42 and Rac1 activity, suggesting that α-parvin and ILK may have antagonistic functions. Consistent with this, the α-Parvin–null CD cells exhibit increased cell spreading, adhesion, and migration, whereas ILK-null CD cells displayed reduced adhesion, spreading, and motility (24). Furthermore, p38 phosphorylation is elevated in α-parvin–null cells but diminished in ILK-null cells (24). We attempted to test the genetic interaction between ILK and α-parvin by generating UB-specific double knockouts of ILK and α-parvin; however, deleting a single allele of ILK in the α-parvin–null background results in embryonic or early postnatal lethality (fig. S9D). Thus, although we cannot answer whether α-parvin negatively regulates ILK function within the IPP complex, ILK and α-parvin appear to have critical functions that are independent of these interactions.

Another key finding is that the increased adhesion and spreading in the α-parvin CD cells contrasted with findings in vascular smooth muscle cells, where there was decreased cell adhesion and spreading (18). Although both cell types have increased RhoA activity (18), there was also excessive Cdc42 activation in CD cells. The defect in vascular smooth muscle cells was caused by Rho-dependent elevated myosin light chain phosphorylation, resulting in increased actomyosin contractility (18). In α-parvin–null CD cells, however, the excess Rho GTPase activation led to decreased actin turnover due to the inhibition of the actin severing function of cofilin. We demonstrate that LIM-kinases partially mediate α-parvin–dependent regulation of cofilin in CD cells. Of note, although the Cdc42 inhibitors completely reconstituted the spreading and migration defects in the CD cells, inhibiting Cdc42 was only partially effective in the organ cultures. This suggests that in addition to the Cdc42-cofillin axis, α-parvin regulates pathways like WASP-mediated actin branching via the Arp2/3 complex (65, 66) and/or formin family proteins (67) in the developing UB.

In conclusion, we demonstrated that cellular rearrangement is critical for UB branching morphogenesis. This process is dependent on dynamic regulation of the actin cytoskeleton by the integrin-binding scaffold protein, α-parvin, which highlights the importance of understanding the complexity of the role of cell-ECM receptors in tubule formation.

MATERIALS AND METHODS

Mice

α-Parvin^flox/flox mice (α-pv^f/f) were generated as previously described (18) and were crossed with HoxB7 promoter–driven Cre recombinase transgenic mice (27). All mice are on a C57/B6 background. Female and male mice were used in equal ratios, and aged-matched littermates (α-pv^f/f) mice were used as controls. For live imaging, α-pv^f/f mice were crossed with mT/mG mice. Mice were bred and maintained under standard conditions in accordance with the Vanderbilt University Institutional Animal Use and Care Committee and conducted in Association for Assessment and Accreditation of Laboratory Animal Care–accredited facilities. The animal protocol number is V2200066.

Antibodies and reagents

Antibodies and chemical used in this study are listed in Table 1.

Table 1. List of antibodies and chemical used in the study.

Listed by name in the left column, manufacturer in middle column, and catalog number in left column.

Antibodies
Rabbit anti–α-parvin (D7F9)	Cell Signaling Technology	8190
Mouse anti–α-parvin (clone 1D4)	MilliporeSigma	MABT1441
Monoclonal anti-cytokeratin, pan (mouse)	Sigma-Aldrich	C2562
Rat WT1 polyclonal antibody	BiCell Scientific	00180
Purified mouse E-cadherin (clone 36)	BD Bioscience	610182
Rabbit polyclonal anti-SIX2	Proteintech	11562-1-AP
ZO-1	Cell Signaling Technology	8193
Rabbit monoclonal anti–E-cadherin (24E10)	Cell Signaling Technology	3195
Rat monoclonal anti–β1-integrin (clone MB1.2)	MilliporeSigma	MAB1997
Laminin	MilliporeSigma	L9393
Rhodamine-conjugated phalloidin	Thermo Fisher Scientific	R415
DAPI	Cell Signaling Technology	4083
Rabbit anti-Ki67	abcam	ab15580
Rat anti-mouse CD31 (clone MEC13.3 RUO)	BD Pharmingen	533370
Rabbit anti-PDGFRB monoclonal antibody (G.290.3)	Invitrogen	MA5-15143
Rabbit monoclonal anti–phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) (D13.14.4E)	Cell Signaling Technology	4370
Rabbit anti–α-parvin Sepharose beads	Cell Signaling Technology	12591
Rabbit monoclonal anti-ILK (4G9)	Cell Signaling Technology	3856
Mouse monoclonal anti-pinch	BD Biosciences	612711
Mouse monoclonal anti-paxillin	BD Biosciences	610568
Rabbit monoclonal anti-paxillin (E6R6Z)	Cell Signaling Technology	50195
Rabbit monoclonal anti–phospho-p44/42	Cell Signaling Technology	4370
Rabbit monoclonal anti-p44/42	Cell Signaling Technology	4695
Rabbit monoclonal anti–phospho-p38	Cell Signaling Technology	4511
Rabbit monoclonal anti-p38	Cell Signaling Technology	8690
Rabbit monoclonal anti–phospho-Akt	Cell Signaling Technology	4060
Rabbit monoclonal anti-Akt (pan)	Cell Signaling Technology	9272
Mouse monoclonal anti-GAPDH	Cell Signaling Technology	97166
Rabbit monoclonal anti–β-actin	Cell Signaling Technology	4970
Mouse monoclonal anti–β-actin	Cell Signaling Technology	3700
Rabbit monoclonal anti-RhoA	Cell Signaling Technology	2117
Rabbit polyclonal anti-Cdc42	Cell Signaling Technology	2462
Rabbit monoclonal anti–phospho-cofilin	Cell Signaling Technology	3313
Rabbit monoclonal anti-cofilin	Cell Signaling Technology	5175
Dolichos biflorus agglutinin (DBA, FITC-conjugated)	Vector Laboratories	FL-1031
Lotus tetragonolobus (asparagus pea) lectin (LTL), fluorescein (FITC)	Invitrogen	L32480
Alexa Fluor 488 goat anti-rabbit	Thermo Fisher Scientific	A-11008
Alexa Fluor 488 goat anti-mouse	Thermo Fisher Scientific	A-11001
Alexa Fluor 594 goat anti-mouse	Thermo Fisher Scientific	A-11005
Alexa Fluor 594 goat anti-rabbit	Thermo Fisher Scientific	A-11037
Alexa Fluor 647 goat anti-rabbit	Thermo Fisher Scientific	A-21245
Alexa Fluor 647 goat anti-mouse	Thermo Fisher Scientific	A-21235
Alexa Fluor 647 goat anti-rat	Thermo Fisher Scientific	A-21247
680RD goat anti-rabbit	LI-COR	926-68071
800CW goat anti-rabbit	LI-COR	926-32211
680RD goat anti-mouse	LI-COR	926-68070
800CW goat anti-mouse	LI-COR	926-32210
Reagents
Prolong gold anti-fade without DAPI	Thermo Fisher Scientific	P10144
Human GDNF recombinant protein	Invitrogen	RP-8602
M-PER mammalian protein extraction reagent	Thermo Fisher Scientific	78501
RIPA	Sigma-Aldrich	R0278
Rho inhibitor I	Cytoskeleton	CT04
ROCK inhibitor (Y-27632)	Cell Signaling Technology	13624
ML141	Selleck Chemicals	CID-2950007
LIMKi 3	Tocris	4745
Cytochalasin D	Millipore Sigma	C8273
Arp2/3 inhibitor CK666	Millipore Sigma	182515
Blebbistatin	Millipore Sigma	B0560

Kidney function

Kidney function was determined by measuring serum BUN using the QuantiChrom Urea Assay Kit (BioAssay Systems, DIUR-100) according to the manufacturer’s protocol. Briefly, serum was mixed with a chromogenic reagent that was measured at 520 nm and the urea concentration deduced from a standard curve.

Histological analysis

For morphological analysis, kidneys were removed at different stages of development and at the indicated ages in adults. They were fixed in 10% formaldehyde or paraffin and embedded. Paraffin tissue sections (5 μm) were stained with hematoxylin and eosin for morphological evaluation by light microscopy. In all experiments, noon of the day on which the mating plug was observed was designated E0.5.

Kidney area was measured in FIJI using cross-sectional sections where the papilla was visible. Kidney length was determined by measuring the longest linear axis across the kidney section.

For immunofluorescence, frozen kidney sections were prepared as previously described (68). In brief, kidneys were fixed in 4% PFA for 1 hour at room temperature and incubated in 30% sucrose overnight, after which they were frozen in OCT for optimal cutting. Sections were blocked with 8% normal horse serum, 0.3% Triton X-100, and 0.1% Tween in phosphate-buffered saline (PBS) for 1 hour and subjected to primary antibodies diluted in the same blocking buffer overnight at 4°C. Samples were washed, incubated with Dolichos Biflorus agglutinin (DBA) and secondary antibodies, and mounted. Sections were imaged by confocal microscopy (Zeiss LSM 980, Airyscan 2).

For whole mounts, samples were prepared as previously described (41, 69–71). Briefly, E10.5 to E12.5 kidneys were collected, fixed in 4% PFA or −20°C methanol at room temperature for 1 hour, washed in PBS, and blocked in blocking buffer [PBS-BB: PBS with 1% bovine serum albumin (BSA), 0.2% nonfat dry milk powder, and 0.3% Triton X-100] (70). Kidneys were incubated with primary antibodies diluted in PBS-BB for 1 to 3 days (depending on the embryonic stage of the kidneys) at 4°C and washed overnight with PBST (0.1% Tween) at 4°C. Kidneys were then exposed to secondary antibodies diluted in PBS-BB for 1 to 3 days at 4°C. Kidneys were washed as described above and examined by confocal microscopy (Zeiss LSM 980, Airyscan 2). Sections were scanned at 1-μm steps, and Z-stacks were used to conduct 3D reconstruction using FIJI or Imaris. Branching tips were quantified using FIJI software with the following standardized pipeline: Images acquired from confocal microscopy were first processed by maximum intensity projection to create a 2D representation. After background subtraction and Gaussian blur filtering to smooth the images, a threshold was applied to generate a binary mask of kidney structures. This mask was then skeletonized, producing a simplified structural representation used to objectively identify branching points and tips. Each terminal end point identified in the skeletonized image was counted as one tip.

For tubular diameter analysis, 8-μm-thick frozen sections were used. CDs were labeled with DBA, and the entire section was imaged using confocal microscopy (Zeiss LSM980, AiryScan 2). Only diameters of the tubules, where the lumen was visible, were measured. Medians were calculated, and significance was determined by the Mann-Whitney U test.

For cell height and number measurement, 40-μm-thick frozen sections were used to obtain Z-stack images of the DBA-labeled collecting tubules. Cortical and medullary CDs were analyzed. The cell number was manually counted in cross sections that were vertically aligned with the tubule’s longitudinal axis, using 4′,6-diamidino-2-phenylindole (DAPI) as the marker for the nuclei. Cell height was measured using FIJI.

RNAscope analyses

RNAscope 2.5 HD Detection Kit-BROWN (Advanced Cell Diagnostics, catalog no. 322330) was used to determine Ret expression pattern in E15.5 kidney. Briefly, E15.5 embryos were fixed in 10% formaldehyde for 24 hours at room temperature, paraffin embedded, cut at 5 μm, and transferred onto superfrost glass slides. Sample preparation, pretreatment, and RNAscope assays were performed according to Formalin-Fixed Paraffin-Embedded Sample Preparation Pretreatment Guide User Manual, Part 1 (catalog no. 322452-USM) and RNAscope 2.5 HD Detection Kit (BROWN) User Manual, Part 2 (catalog no. 322310-USM). Probe for mouse Ret (catalog no. 431791) and negative probe (catalog no. 310043) were used as controls.

Ex vivo kidney cultures with GDNF

E11.5 kidneys carrying the mT/mG transgene were isolated at E11.5 and cultured ex vivo on Corning Transwell inserts (#3801) in Dulbecco’s modified Eagle’s medium (DMEM)/F12 supplemented with 10% fetal bovine serum (FBS) at 37°C and 5% CO₂. Explants were treated with recombinant GDNF (20 ng/ml) and cultured for 72 hours. The UB was visualized at 0 hours by GFP fluorescence using the Nikon spinning-disk confocal microscope (SoRa) with a 10× objective. At 72 hours, explants were fixed in 4% PFA and immunostained for Six2; z-stacks were acquired at 5-μm steps and maximum-intensity projections were generated in FIJI for visualization.

Live-organ imaging of kidney ex vivo

HoxB7^Cre (Hoxb7) or Hoxb7:α-pv^f/f were crossed with mT/mG mice (the Jackson Laboratory, #007676). Kidney explants from E12 to E12.5 HoxB7^Cre:mT/mG (Hoxb7), Parvin^fl/wt:mT/mG:HoxB7^Cre (Hoxb7:α-pv^f/wt), and Parvin^fl/fl:mT/mG:HoxB7^Cre (Hoxb7:α-pv^f/f) were isolated as mentioned above (69, 70). Ex vivo tissue culture set-up for live imaging was adapted from the FiZD method (64) and salivary gland live imaging (63). Briefly, double-adhesive imaging spacers (Grace Bio-Labs, 654008) were sterilized by soaking in 70% ethanol for 3 min and attached to the glass bottoms of 35-mm ibidi dishes (ibidi, 81218-200). Kidneys were carefully transferred to the center of ibidi dish using a 200-μl pipette. A maximum of three kidneys were included in one dish. Ten microliters of Matrigel (Corning, 356230, 5 μg/μl) was dropped onto each kidney to prevent tissues from moving. The transwell filter of the Snapwell (Corning, #3801, 0.4-μm pore size) was stuck to the image spacer so that the kidneys were sandwiched between the glass bottom and the filter. Five hundred microliters of organ culture medium (DMEM/F12, 10% FBS, with penicillin-amoxicillin) was then added to the well. The kidneys were incubated at 37°C with 5% CO₂ for 16 hours before imaging. For inhibitor experiments, kidneys were submerged in inhibitors for 1 hour at 37°C with 5% CO₂ before being transferred to the ibidi dish. A Zeiss LD LCI Plan-Apochromat 25×, 0.8 numerical aperture (NA), water-immersion objective on a Zeiss LSM980 AiryScan 2 Confocal Microscope controlled by Zeiss Zen Blue software was used for image acquisition. Images were acquired at 1-μm z intervals at ∼50-μm thickness and 10-min intervals for 22 hours using CO-8Y mode. The laser was used at 5.5% for excitation of membrane-GFP.

Live video analysis

3D surface reconstruction of indicated time points was performed by drawing polylines along the epithelial surface at every other z plane (2-μm interval) on an x-y view using Imaris software. To study the change in tubular size during elongation, z-stack images were rotated so that the longitudinal axis of the tubule was aligned with the x axis in FIJI. The size of the tubule was assessed by measuring the cross-sectional area in the orthogonal view, accomplished by manually drawing polylines around the tubule at indicated time points during an 11-hour time span.

To measure the orientation of cell division, UB trunk cells (at least 50 to 100 μm from the tip edge) were manually identified when they were in either telophase or cytokinesis, as observed in the time-lapsed movies. The division angle was measured in FIJI using the “angle” tool as previously described (72). Briefly, a first line was drawn connecting the centroids of the dividing cells and the angle was defined by drawing a second line parallel to the longitudinal axis of the tubular lumen.

The measurement of cell rearrangement was adapted from Packard et al. (73). Briefly, cells were first segmented using Cellpose (cyto3 model) (74), followed by manual correction. Cell trajectories were then tracked using TrackMate (75) in FIJI. For cell dispersal analysis, four to eight cells that shared the same convergent point (defined as located within 15 pixels of one another) were randomly picked and identified as a cell cluster. Those cell clusters were then followed over indicated time. Dispersal was quantified as the standard distance or the geometric mean of the distance of each cell from the group centroid over time. Standard distance (SD) was calculated using the following function SD = $\sqrt{\frac{{\displaystyle \sum _{\mathrm{i}=1}^n}{(x_{\mathrm{i}}-\overline{X})}^2}{n}+\frac{{\displaystyle \sum _{\mathrm{i}=1}^n}{(y_{\mathrm{i}}-\overline{Y})}^2}{n}}$, where ($x_{\mathrm{i}},y_{\mathrm{i}}$) is the pixel coordinate and $(\overline{X}, \overline{Y})$ is the mean center of the cells.

To evaluate whether the trend of cell dispersal differed between experimental groups, we analyzed UB trunk and UB tip regions separately because of their distinct dispersal dynamics. For UB trunks, the standard distance over time was fit into a linear regression equation y = at + b, where t is time. The slope (a) and intercept (b) were extracted as descriptors of dispersal behavior. To statistically compare dispersal trends between groups, we performed analysis of covariance on the regression slopes. Dispersal behavior at the UB tips followed a nonlinear pattern and was better captured using a second-order polynomial fit. For each cell cluster, the standard distance over time was fitted to the equation y = at² + bt + c. The resulting coefficients (a, b, and c) were treated as multivariate representations of dispersal dynamics. To compare these across experimental groups, we performed pairwise Hotelling’s t² tests, a multivariate extension of the t test. P values were adjusted for multiple comparisons using the false discovery rate method.

Method for coarse grained velocity field estimation

Coarse-grained particle image velocimetry was performed using quickPIV—an opensource script written in Julia Language. The time series of the 3D image stacks were processed in Jupyter notebooks with a window size of (32, 32, and 7) pixels in the (x, y, and z) directions, respectively. On the basis of a constant height of the image stacks, the z dimension of the window was set to 7 pixels. Window overlaps of (8, 8, and 3) pixels were specified unless explicitly stated with a search margin set to (10, 10, and 3) pixels to account for the boundaries. To compute the motion similarity maps, a scalar product between the raw velocity vector and its 4-neighbors were used. The visualizations of the velocity vectors and the motion similarity maps were rendered in Paraview 5.10.

Cell culture

Primary CD cells were isolated from kidneys of α-pv^f/f mice and immortalized with SV40 as previously described (58). The PARVA gene was deleted by infecting with adenoCre virus, after which null-CD cells were cloned. CD cells were cultured in DMEM/F12 supplemented with 10% fetal calf serum and antibiotics. Deletion of α-parvin was verified by Western blot analysis. Numerous clones were tested, and they behaved similarly. For cell treatments, all drugs were dissolved for stock solutions and used according to the manufacturer’s instructions. Solvents for stock solutions were used as negative control treatments in respective experiments.

Immunoprecipitation

For immunoprecipitation, confluent cells were lysed with M-PER and protein extracts (0.5 mg for α-Parvin immunoprecipitation and immunoblotting for α-Parvin, Pinch, and ILK, and 1.0 mg for ILK immunoprecipitation and immunoblotting for ILK and Pinch) were preconjugated beads or precleared with protein A/G beads and incubated with the specific primary antibody or overnight at 4°C with rotating, respectively. The antibody-protein complex was washed twice with TBS-0.1%Tween and twice with TBS-1% Triton X-100 and collected by centrifugation. The immune complex was eluted by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer and subject to Western blot analysis.

Tubulogenesis assay

Tubulogenesis of CD cells was performed in 3D ECM gels composed of rat tail collagen I (1.5 mg/ml; Corning, 354236) and Matrigel (0.25 mg/ml) in DMEM containing 20 mM Hepes. CD cells (1.5 × 10³) were mixed with the collagen/Matrigel mixture, transferred into a 96-well plate, and returned to tissue incubator (37°C and 5% CO₂). After the cell/collagen/Matrigel mixture solidified, 100 μl of DMEM/F12 with 10% FBS and antibiotics were added to each well to prevent evaporation. The cells were allowed to grow for 7 days. The gels were stained with rhodamine-conjugated phalloidin, and the tubules were imaged using a Zeiss Axio 510 confocal microscope (400×).

Scratch/wound healing assay

Confluent monolayers were washed, and the serum concentration was decreased to 2% to minimize proliferation-dependent contributions to gap closure, and a linear scratch was generated using a sterile 10-μl pipette tip. Where indicated, inhibitors were added at the specified concentrations 30 min before imaging (CT04, 2 μg/ml; Y-27632, 20 μM; ML141, 10 μM; CK666, 100 μM; blebbistatin, 10 μM; p38 inhibitor, 10 μM). Time-lapse images were acquired for 24 hours on a Keyence BZ-X800 microscope using a 10× objective and phase-contrast microscopy. To analyze wound closure, acquired time-lapse images were segmented in FIJI using Trainable Weka Segmentation, and the resulting classified image was converted to an 8-bit mask by thresholding the wound class to generate a binary wound region. Residual wound area (A_residual) was measured from the mask at each time point and normalized to the initial wound area (A₀). Percent wound closure was calculated as % wound closure = (1 − A_residual/A₀) × 100.

Cell spreading and FA analysis

Eight-well chamber slides coated with laminin 511 (10 μg/ml) were blocked with 1% heat-denatured BSA for 1 hour. CD cells were then plated in serum-free medium with indicated treatments onto the slides. Cells were allowed to spread for 1 hour, fixed with 4% PFA, and stained with rhodamine-conjugated phalloidin. Images were collected with fluorescence microscope (Olympus, 20×/1.25 NA), and the cell area and circularity were determined using FIJI software. For FA analysis, cells were allowed to spread for 4 hours. FAs were labeled by immunostaining of paxillin, and the number/cell area and FA size/cell area were quantified using FIJI.

Cell adhesion, migration, and proliferation

For the cell adhesion assay, 96-well round-bottom plates (Corning, #38018) were coated with different concentrations of laminin 511 and blocked with 1% heat-denatured BSA. Cells (10⁵) in serum-free DMEM/F12 were placed in each well for 1 hour; nonadherent cells were removed, and the remaining cells were fixed with 4% PFA, stained with 1% crystal violet, and solubilized in 20% glacial acid. The plates were measured for absorbance at 540 nm.

For the transwell migration assay, transwells with 8-μm pore size (Corning, #3422) were coated with laminin 511 (10 μg/ml) on the lower side of the filter, and 10⁵ cells were added to the upper well in serum-free medium. For inhibitor treatments, cells were pretreated with Y-27632 (ROCK inhibitor) and ML141 (Cdc42 inhibitor) for 45 min before seeding. CT04 (Rho inhibitor), LIMKi 3 (LIMK inhibitor), blebbistatin, cytochalasin D, and CK666 were mixed with the cell suspension at seeding. Cells that migrated through the filter after 4 hours were stained with 1% crystal violet (Sigma-Aldrich, C3886) and hematoxylin Gill #3 (Sigma-Aldrich, GHS332) for cell nuclei. Images were collected with microscope (Olympus, 20×/1.25 NA), and five random fields were analyzed per group. Cells were counted using FIJI.

For the random migration assay, the cells were sparsely seeded on laminin 511 (10 μg/ml)–coated 35-mm glass-bottom ibidi dishes (ibidi, 81218-200) in serum-free culture medium. Time-lapse differential interference contrast (DIC) imaging was conducted at 10-min intervals for 24 hours using a 10× objective lens on a Zeiss LSM980 confocal microscope. DIC images were acquired with a laser power of 7%, using a photomultiplier tube gain of 700 V. Cell tracking was performed using TrackMate (75) in FIJI, measuring cell velocity as mentioned above.

Cell proliferation was assessed by measuring incorporation of 5-bromodeoxyuridin (BrdU) in an enzyme-linked immunosorbent assay according to the manufacturer’s manual (Exalpha, X1327K). Briefly, 96-well plates were either left uncoated or coated with laminin 511 (10 μg/ml) and blocked with 1% BSA. A total of 1500 cells were plated in medium (DMEM/F12 with 2% FBS) in each well, allowed to adhere for 6 hours, and then incubated with BrdU for 16 hours. BrdU incorporation was quantified by a change of absorbance (optical density) at 450 nm.

Integrin-related signaling pathway and cofilin activation analysis

For adhesion-dependent integrin signaling and the cofilin activation assay, serum-starved CD cells were plated in serum-free medium on laminin 511 (10 μg/ml) for the indicated time points. For inhibitor treatments, cells were pretreated with Y-27632 (ROCK inhibitor) and ML141 (Cdc42 inhibitor) for 45 min before seeding. CT04 (Rho inhibitor), LIMKi 3 (LIMK inhibitor), blebbistatin, cytochalasin D, and CK666 were mixed with the cell suspension at seeding. Nonadherent cells were removed by gently washing with PBS. Adherent cells were lysed using M-PER reagent with protease and phosphatase inhibitor cocktails. Protein extracts (20 to 40 μg) were subject to Western blot analysis.

Flow cytometry

Flow cytometry was performed as previously described (76). CD cells were incubated with the indicated antibodies followed by fluorescein isothiocyanate–conjugated secondary antibodies.

Rho GTPases pull-down and activity assays

The activation status of RhoA, Cdc42, and Rac1 was assessed using pull-down assays specific for the guanosine triphosphate (GTP)–bound form of the Rho GTPases. We used a Rho Activation Assay kit (Millipore Sigma, 17-294) according to manufacturer’s protocol. The pull downs were quantified by Western blots.

G-to-F-actin ratio assay

G-actin/F-actin ratio analysis was performed using the G-actin/F-actin in vivo assay Beachem kit (Cytoskeleton, BK037). Briefly, equal amount of protein was determined by BCA protein quantification before the G-to-F-actin ratio assay. Confluent cells were lysed in F-actin stabilization buffer containing 1 mM adenosine triphosphate, followed by centrifugation to separate the F-actin (pellet) and G-actin (supernatant) pools. The F-actin pellet was then dissolved in F-actin depolymerization buffer that contains 8 M urea. The resulting fractions were subject to SDS-PAGE, and actin was quantified by Western blot.

In vitro immunofluorescence

For cellular immunostaining, cells were grown on eight-well Milli cell EZ chamber slides (Millipore, PEZZGS0816) coated with laminin 511 (10 μg/ml) and allowed to adhere and spread for the times indicated. Cells were fixed in 4% PFA, rinsed with PBS, and permeabilized and blocked with blocking buffer (0.3% Triton X-100 and 8% normal horse serum in TBS). Subsequently, cells were incubated with the indicated primary antibodies followed by secondary antibodies. The slides were then washed and mounted using prolong gold anti-fade mounting media. Fluorescent images were acquired using a Zeiss laser scanning confocal microscope (LSM980, 63×/1.5 NA oil Olan Apochromat oil). For polarized cell staining, cells were grown on transwell inserts until they formed a monolayer, fixed in 4% PFA, permeabilized, and blocked as mentioned above. The cells were then stained with rhodamine-conjugated phalloidin. Z-stack images were acquired at 1-μm intervals. Line scan was performed using the plot profile tool of FIJI software.

Ex vivo kidney cultures with inhibitors

E12.5 kidney explants with mT/mG transgene were collected as described above. Kidneys were submerged in inhibitors 1 hour before transferring to a Transwell (Corning, #3460). Kidneys were cultured with complete medium (DMEM/F12 and 10% FBS, antibiotics) with inhibitors for 24 hours and fixed. Images were examined by confocal microscopy (Zeiss LSM 980, Airyscan 2). Sections were scanned at 5-μm steps, and maximal intensity projection was used for visualization and analysis in FIJI. Branching was assessed by quantifying the number of tips from images.

Statistical analysis

Data are shown as means ± SEM. Unpaired two-tailed t test was used to evaluate statistically significant differences (*P < 0.05) between two groups. Two-way analysis of variance (ANOVA) was used to test statistical significance (*P < 0.05) among multiple groups. Post hoc comparisons of ANOVA were corrected with the method of Tukey. For fold-change calculations (e.g., Western blot densitometry), the control number was normalized to 1 for each comparison. Data distribution was assumed to be normal, but this was not formally tested. GraphPad Prism software was used for statistical analysis.

Supplementary Materials

The PDF file includes:

Figs. S1 to S9

Legends for movies S1 to S10

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S10