Multivalent nephrin–Nck interactions define a threshold for clustering and tyrosine-dependent nephrin endocytosis

Organization of plasma membrane signaling proteins into higher-order assemblies or clusters is essential to control cellular responses to external cues (Mayer and Yu, 2018). Phase separation has recently emerged as a mechanism of membrane protein clustering, and it is driven by interactions between multivalent proteins, such as those found within the nephrin–Nck–N-WASp complex (herein, Nck refers to both Nck1 and Nck2 except where specified; N-WASp is also known as WASL) (Mayer and Yu, 2018). Nephrin is a transmembrane adhesion molecule that contains three phosphorylated YDxV motifs on its intracellular tail, each of which can bind the SH2 domain of the Nck1 and Nck2 (Nck1/2) cytoplasmic adaptors (Jones et al., 2006). Nck proteins also contain three SH3 domains and they associate with multiple proline-rich motifs (PRMs) on N-WASp (Rohatgi et al., 2001), which in turn binds and activates the Arp2/3 complex to promote actin nucleation (Svitkina and Borisy, 1999). The componentry of nephrin–Nck and Nck–N-WASp signaling has been studied using synthetic reconstitution systems, wherein the valency of nephrin YDxV motifs, Nck SH3 domains and N-WASp PRMs correlates with phase separation and actin assembly in solution (Banjade and Rosen, 2014; Ditlev et al., 2012; Li et al., 2012). Moreover, recent studies have shown that the highly cross-linked nature of the nephrin–Nck–N-WASp complex increases the dwell time and specific activity of N-WASp and Arp2/3, further driving actin polymerization at sites of phase-separated clusters (Case et al., 2019). These findings support the notion that nephrin–Nck–N-WASp clustering leads to the formation of a distinct actin-associated signaling node, yet the functional implications of these multivalent interactions in a relevant physiological setting have yet to be revealed.

Nephrin serves as the principal component of the slit diaphragm, a unique intercellular junction bridging the foot processes that extend from adjacent podocytes in the kidney (Scott and Quaggin, 2015). At the cell surface, trans interactions between extracellular immunoglobulin domains of nephrin form the core of the slit diaphragm, allowing it to act as a size-selective blood filtration barrier (Fig. 1A). Inside the cell, nephrin functions as a signaling scaffold to bind proteins such as Nck (New et al., 2014), which are crucial for supporting the podocyte's specialized morphology (Jones et al., 2006, 2009). Mutation of all three YDxV motifs (hereafter referred to as the nephrin-Y3F mutation) disrupts interaction with Nck and its ability to induce actin assembly at nephrin clusters in fibroblasts (Blasutig et al., 2008; Jones et al., 2006). Furthermore, changes in YDxV phosphorylation are observed in human kidney diseases and related experimental models in rodents (Jones et al., 2006; Uchida et al., 2008), and mice harboring the nephrin-Y3F knock-in mutation develop filtration defects with age (New et al., 2016). Precise regulation of nephrin–Nck signaling is therefore a central determinant of podocyte health.

Nephrin mislocalization is commonly observed in kidney diseases of hereditary and acquired origins (Königshausen et al., 2016; Martin et al., 2018; Nishibori et al., 2004; Ohashi et al., 2010; Shimizu et al., 2002; Soda et al., 2012; Wernerson et al., 2003), suggesting that abnormal nephrin trafficking likely contributes to barrier dysfunction. Localization of nephrin within the slit diaphragm requires dynamic phosphorylation events and endocytic recycling (Soda and Ishibe, 2013), though the molecular mechanisms that control this interplay are not well understood. Along with Nck, several slit diaphragm-associated signaling proteins implicated in endocytosis, including N-WASp and the GTPase dynamin, which participates in vesicle scission (Ferguson and De Camilli, 2012), are essential to maintain podocyte structure (Jones et al., 2009; Schell et al., 2013; Soda et al., 2012). Moreover, dynamin-deficient podocytes show reduced nephrin internalization and an accumulation of arrested endocytic pits, which contain Arp2/3, alluding to a connection between the endocytosis machinery and the actin cytoskeleton in these cells (Soda et al., 2012). Intriguingly, emerging evidence suggests that phase separation can drive membrane invagination and endocytosis in several cell types (Hilgemann et al., 2020). In this article, we now establish a connection between clustering and endocytosis in podocytes in vitro and in vivo, which appears to be regulated by threshold levels of nephrin tyrosine phosphorylation and Nck SH3 domain signaling with N-WASp and dynamin. We also show that a transient rise in nephrin phosphorylation occurs upon podocyte injury and correlates with decreased surface nephrin, leading to filtration barrier leakage, and that mice lacking the Nck-binding sites on nephrin are resistant to this injury. Altogether, our data suggest that nephrin–Nck interactions provide a tunable system to regulate protein composition and membrane turnover at the slit diaphragm.

To probe the downstream consequences of nephrin–Nck clustering in podocytes, we first monitored the subcellular distribution of nephrin and the Y3F variant, which harbors Y-to-F substitutions in all three of the Nck-binding YDxV motifs to prevent their phosphorylation (Fig. 1A). To facilitate these studies in cultured podocytes, which lack robust expression of endogenous nephrin, we generated a series of adenoviruses containing full-length Myc-tagged wild-type (WT) or Y3F nephrin, or chimeric nephrin proteins in which the intracellular (IC) regions of GFP-tagged WT or Y3F nephrin are fused to the extracellular domain of CD16 (FCGR3A) and the transmembrane domain of CD7 (denoted CD16/7) (Fig. 1B). We and others have previously reported that clustering of CD16/7-nephrin(WT)^IC is induced upon addition of anti-CD16 cross-linking antibodies, that the clusters persist for several hours, and that this chimera is a suitable surrogate for full-length nephrin in signaling assays (Blasutig et al., 2008; Jones et al., 2006; Verma et al., 2003). Similar to findings in mouse embryonic fibroblasts (MEFs) (Blasutig et al., 2008; Jones et al., 2006; Verma et al., 2003), cross-linking of CD16/7-nephrin(WT)^IC in mouse podocytes with CD16 antibodies for 30 min induced the formation of punctate clusters throughout the cell (Fig. 1C, left). Intriguingly, CD16/7-nephrin(Y3F)^IC organization within podocytes was heterogeneous, displaying clusters of varying sizes, with a predominance of clusters with an elongated morphology (Fig. 1C, left). To investigate the localization of these structures, we performed surface immunolabeling of nephrin in podocytes expressing CD16/7-nephrin^IC-GFP, in which total nephrin was labeled with green and surface nephrin was labeled with red. Under these conditions, CD16/7-nephrin(WT)^IC clusters appear as both green and yellow in the merged panel (Fig. 1C, center and right), indicating that they are found both in the cytosol and on the cell surface, as previously reported (Blasutig et al., 2008; Qin et al., 2009). The clusters that accumulate with the CD16/7-nephrin(Y3F)^IC protein also appear with both total (green) and surface (red) labeling (yellow in merge) (Fig. 1C, center and right). However, quantification of these structures revealed that over 86% of CD16/7-nephrin(Y3F)^IC-expressing cells contained elongated clusters, compared with 22% of CD16/7-nephrin(WT)^IC-expressing cells (Fig. 1C). We further examined the membrane expression profiles of the CD16/7-nephrin(WT)^IC or CD16/7-nephrin(Y3F)^IC proteins in transduced podocytes using cell surface biotinylation assays. Consistent with our imaging observations, CD16/7-nephrin(Y3F)^IC is enriched over 2-fold on the cell surface compared with CD16/7-nephrin(WT)^IC (Fig. 1D,E). To ensure that formation of these structures was intrinsic to nephrin, we repeated these experiments in MEFs, and again observed the appearance of elongated clusters coincident with an increase in surface nephrin accumulation with CD16/7-nephrin(Y3F)^IC (Fig. 1F-H). In addition, we transduced the corresponding full-length nephrin constructs into podocytes, and likewise noted the induction of elongated clusters and heightened surface levels of nephrin-Y3F compared with nephrin-WT (Fig. 1I-K). Biotinylation assays on HEK cells expressing full-length nephrin showed similar results (Fig. S1A,B). Lastly, to verify the link between nephrin tyrosine phosphorylation and surface dynamics, we expressed full-length nephrin-WT and nephrin-Y3F with and without active Fyn kinase in HEK cells and performed surface biotinylation assays. We found that Fyn induced a decrease in surface levels of nephrin and that this induction was significantly greater for nephrin-WT compared with nephrin-Y3F, supporting a role for these particular tyrosines in nephrin endocytosis (Fig. S1C,D). Altogether, our results implicate nephrin phosphorylation on the Nck-binding sites in regulating cluster formation and nephrin surface expression in podocytes.

To complement our in vitro findings, we next monitored nephrin localization in our gene-targeted nephrin-Y3F mice, which develop progressive proteinuria and podocyte dysmorphology as they age (New et al., 2016). Compared with WT control animals, aged nephrin-Y3F mice show mislocalization of nephrin along with its binding partner podocin in immunostained glomerular tissue sections, with both proteins present in focal aggregates (Fig. 2A). To examine this in greater detail, we utilized super-resolution structured illumination microscopy (SR-SIM) to image nephrin-immunostained kidney sections prepared from WT and Y3F mice. Again we observed aggregates of nephrin in Y3F mice, whereas nephrin showed fine linear staining along the capillary wall in WT mice (Fig. 2B). Lastly, ex vivo biotinylation of intact glomeruli isolated from kidneys of WT and mutant animals revealed a significant 2-fold increase in surface nephrin levels in nephrin-Y3F mice compared with WT controls (Fig. 2C,D). Altogether, these results support the notion that loss of nephrin phosphorylation on the YDxV motifs impairs nephrin localization in podocytes.

Previous studies have shown that phosphorylation of nephrin triggers its internalization into endosomal compartments using both clathrin-dependent and -independent mechanisms (Jeon et al., 2012; Lahdenperä et al., 2003; Qin et al., 2009). To investigate whether nephrin clusters represent endocytic structures, we first examined our well-characterized fibroblast model for the presence of nephrin and clathrin at clusters formed after 30 min of CD16 antibody treatment in cells expressing CD16/7-nephrin(WT)^IC or CD16/7-nephrin(Y3F)^IC (Fig. 3A,B). We observed nephrin and clathrin in close proximity in most of the clusters in CD16/7-nephrin(WT)^IC cells (Fig. 3A, arrows). Interestingly, in CD16/7-nephrin(Y3F)^IC cells, clathrin was also localized to clusters but only observed close to the tips of the elongated clusters rather than along the entire structure (Fig. 3B, arrows). Next, we performed a time-course analysis of CD16/7-nephrin(WT)^IC cluster formation and surface expression in mouse podocytes, and also monitored nephrin colocalization with clathrin (Fig. 3C-E). In line with our previous kinetic analysis in MEFs (Blasutig et al., 2008), confocal imaging revealed surface cluster formation of nephrin-WT within 15 min of the addition of CD16 antibodies, some of which colocalized with clathrin (Fig. 3D, arrow), demonstrating that endocytic machinery is recruited to nephrin signaling clusters. At 30 min, colocalization between nephrin-WT and clathrin continued to be detected; however, the nephrin-clathrin clusters were mostly cytosolic (Fig. 3E, arrow). Lastly, we examined clathrin co-staining in podocytes expressing the CD16/7-nephrin(Y3F)^IC mutant after 30 min of CD16 antibody clustering (Fig. 3F). Surface nephrin-Y3F clusters also shared some localization with clathrin, but only at the termini of the elongated clusters (Fig. 3F, arrow). Altogether, our results support the conclusion that the elongated clusters seen with nephrin-Y3F represent nephrin-enriched endocytic structures.

The three nephrin YDxV motifs are arranged in tandem and our previous studies have shown that a single YDxV motif is sufficient to recruit Nck and promote actin assembly (Blasutig et al., 2008; New et al., 2013; Verma et al., 2003), suggesting that they might be functionally redundant. To determine the relative contributions of each YDxV motif in cluster assembly, we expressed chimeric nephrin proteins with Y-to-F point mutations in one, two or all three YDxV motifs (Fig. 4A) alongside CD16/7-nephrin(WT)^IC in MEFs, performed CD16 antibody cross-linking for 30 min, and monitored nephrin distribution by surface immunolabeling. Unexpectedly, we noted that disruption of a single YDxV motif caused elongated clusters to amass (Y1F series), and this phenotype was enhanced when any two motifs were mutated (Y2F series) (Fig. 4A). Quantification of the percentage of elongated clusters revealed that all mutants were significantly different from WT (Fig. 4B). Notably, a gradient or dose-dependent effect existed between Y1F single mutants with all Y2F double mutants, and between all Y2F mutants and the triple Y3F mutant (Fig. 4B). Interestingly, of the single mutants, Y1217F (denoted YYF) induced more elongated clusters than either Y1176F (denoted FYY) or Y1193F (denoted YFY) (Fig. 4B), which may reflect its preferential binding to Nck (Jones et al., 2006). Collectively, our findings suggest that the efficiency of nephrin clustering correlates with the number of nephrin YDxV-based phosphorylation sites.

The ability of Nck to associate with all YDxV motifs and its central role in actin assembly prompted us to explore the consequence of Nck disruption on nephrin clustering and surface localization. To interfere with Nck signaling specifically, we employed a FLAG-tagged Nck2 variant with point mutations in all three tandem SH3 domains (SH3*x3) to prevent binding to the proline-rich sequences of downstream effectors, including N-WASp (Fig. 5A). Nck2-SH3*x3 and Nck2-WT were co-transfected with GFP-tagged CD16/7-nephrin(WT)^IC into MEFs lacking Nck1/2 expression. Following addition of CD16 antibodies for 30 min to promote cross-linking, cells were fixed and permeabilized to allow immunolabeling of Nck2 (FLAG), as well as contrast-stained with Texas Red phalloidin to visualize F-actin. Expression of Nck2-WT stimulated development of discrete clusters containing colocalized nephrin and Nck, and they were associated with actin comet tails (Fig. 5B, top panel), consistent with our previous findings (Jones et al., 2006). By contrast, Nck2-SH3*x3 maintained colocalization with nephrin, but both proteins showed a prominence of elongated clusters, similar to observations with nephrin Y-to-F mutations, and actin polymerization did not occur in the presence of the Nck mutant (Fig. 5B, middle panel, and 5C), in line with our previous findings (Blasutig et al., 2008). A similar defect was observed with Nck1-SH3*x3 but not Nck1-WT (Fig. S2). To examine further the nature of the elongated nephrin clusters associated with the Nck SH3*x3 mutants, we performed biotinylation assays in HEK cells transfected with full-length nephrin-WT. Co-expression of Nck2-WT but not Nck2-SH3*x3 caused a significant reduction in surface nephrin levels (Fig. 5D,E), and was correlated with a rise in nephrin phosphorylation (Fig. 5D). We have previously shown that Nck mediates recruitment of Fyn kinase (New et al., 2013), and indeed, expression of Fyn alone with nephrin-WT produced a similar effect on nephrin surface removal (Fig. S1C,D). Together, these studies suggest a link between nephrin cluster formation and nephrin surface expression that is dependent on Nck SH3 domain signaling.

We next looked for a potential Nck binding partner with a known role in nephrin endocytosis. The protein dynamin-2 has been shown previously to influence nephrin's phospho-dependent trafficking in pancreatic beta cells through an unknown intermediate (Fornoni et al., 2010; Jeon et al., 2012), and it contains a proline-rich region required for vesicle scission (Ferguson et al., 2009). Given this potential connection, we investigated the ability of the Nck SH3 domains to bind dynamin-2. We initially determined that both Nck1 and Nck2 could associate with dynamin-2 upon co-expression in HEK cells (Fig. 6A). Notably, the strengths of these interactions were much weaker than those observed with N-WASp, suggesting that dynamin-2 may represent a low-affinity binding partner for Nck. Next, we verified that interaction of dynamin-2 with Nck2 requires intact SH3 but not SH2 domains (Fig. 6B), consistent with earlier reports using recombinant SH3 domains in vitro (Wunderlich et al., 1999). To confirm that dynamin-2 can alter surface nephrin levels, we co-expressed full-length nephrin-WT along with either dynamin2-WT or mutant dynamin2-K44A, a dominant negative unable to sever membranes (Ferguson and De Camilli, 2012) (Fig. S3A,B). We found that dynamin2-K44A co-expression resulted in a significant increase in surface nephrin levels, consistent with previous reports of reduced nephrin endocytosis when dynamin function is impaired (Lahdenperä et al., 2003; Qin et al., 2009).

Having identified a role for the Nck SH3 domains in nephrin clustering and localizing endocytic machinery to phosphorylated nephrin, we reasoned that transplantation of this region to the defective nephrin-Y3F protein should rewire the nephrin-Nck signal to bypass the YDxV motifs and allow for clustering and endocytosis to proceed. To this end, we generated a construct in which the three SH3 domains of Nck2 were cloned in-frame downstream of the intracellular tail of CD16/7-nephrin(Y3F)^IC, creating the CD16/7-nephrin(Y3F)^IC-Nck2(SH3x3)-GFP fusion protein (Fig. 6C) (Keyvani Chahi et al., 2016). We first confirmed that interaction of dynamin-2 was reduced with nephrin-Y3F compared with nephrin-WT, and that the fusion protein could strongly associate with both transfected and endogenous dynamin in HEK cells (Fig. 6D). Next, CD16/7-nephrin(WT)^IC, CD16/7-nephrin(Y3F)^IC or CD16/7-nephrin(Y3F)^IC-Nck2(SH3x3) were independently expressed in Nck1/2-null MEFs, followed by CD16 cross-linking for 30 min and surface immunolabeling. Cells expressing the fusion protein showed nephrin clusters on the cell surface, which appeared similar albeit smaller than those with WT nephrin, and in sharp contrast to the elongated nephrin clusters induced by the Y3F mutant (Fig. 6E,F). Biotinylation experiments in HEK cells verify that surface nephrin levels are normalized by fusion of the Nck2 SH3 domains to nephrin-Y3F (Fig. 6G,H). In further support of this concept, addition of the TIR YDxV sequence, which recruits Nck to CD16/7-nephrin(Y3F)^IC, restored nephrin clusters to a punctate shape (Fig. S4), and we have shown previously that this protein/peptide fusion also restores actin polymerization (Blasutig et al., 2008). Together, these findings imply that nephrin-Nck interaction serves to localize SH3 effectors, including N-WASp and dynamin-2, which drive cluster formation and endocytosis.

Lastly, we set out to determine whether the interplay between nephrin phosphorylation and endocytosis seen in cultured cells might have physiological implications. To examine this, we employed a mouse model of reversible podocyte injury caused by injection of nephrotoxic serum (NTS), which correlates with transient permeability of the glomerular filtration barrier (proteinuria) and nephrin endocytosis (Haase et al., 2017; Königshausen et al., 2016). As we have reported previously (Keyvani Chahi et al., 2016), proteinuria peaked at 24 h post-NTS injection (Fig. 7A), although it could be detected as early as 6 h post-NTS injection (Fig. 7A). To characterize the kinetics of nephrin tyrosine phosphorylation and endocytosis at the onset of injury, mice were injected with NTS, and glomeruli were isolated from kidneys at 0 (baseline), 2, 4, 6 or 12 h post-injection for use in ex vivo biotinylation assays. Nephrin immunoblotting of avidin-precipitated samples demonstrates a significant loss of nephrin from the cell surface at 2 and 4 h post-injection, with a return to near baseline at 6 and 12 h post-injection (Fig. 7B,C). An inverse relationship was observed with nephrin phosphorylation on the Nck-binding sites, with an increase above baseline at 2 and 4 h post-injection, and a decrease at 6 and 12 h post-injection (Fig. 7B,D), which appeared to persist at 24 h post-injection, as reported previously (Keyvani Chahi et al., 2016). Given that dynamic changes in phosphorylation on the YDxV motifs appear to accompany filtration barrier remodeling, we reasoned that nephrin-Y3F mutant mice might show an altered response to NTS challenge. For these studies, we used young nephrin-Y3F animals, which do not show evidence of proteinuria (New et al., 2016). WT and nephrin-Y3F mice were injected with NTS or control IgG, and urine was collected at baseline or at 24 h post-injection. Remarkably, nephrin-Y3F mice did not develop significant NTS-induced proteinuria compared with WT counterparts (Fig. 7E,F), and they retained their foot process ultrastructure (Fig. 7G), unlike WT mice, which showed evidence of retraction. Altogether, these data demonstrate that hyperphosphorylation of nephrin YDxV motifs leads to rapid induction of nephrin endocytosis and subsequent remodeling of the filtration barrier.

In this article, we uncover a threshold effect of nephrin YDxV motifs and Nck SH3 domain signaling in controlling the assembly of membrane clusters and regulating endocytosis. We also show that dynamic changes in nephrin phosphorylation on the Nck-binding sites in vivo correlate with nephrin turnover in the slit diaphragm and filtration barrier remodeling. Taken together, our genetic and cell-based studies are the first to uncover a physiological function for this high valency signaling system.

The number of YDxV motifs on nephrin has been shown to correlate with phase separation in solution, wherein the presence of a tri- versus diphosphorylated nephrin tail lowers the concentrations of Nck and N-WASp required to reach the transition boundary for droplet formation (Banjade and Rosen, 2014; Li et al., 2012). Nephrin multivalency also appears to be necessary to increase the local density of Nck molecules within the phase-separated clusters, and in turn the stoichiometry-dependent dwell time and specific activity of associated effector proteins (Hilgemann et al., 2020; Oh et al., 2012; Rosendale et al., 2019). Our previous studies have demonstrated that a single nephrin YDxV motif is sufficient to bind Nck and induce actin polymerization (Blasutig et al., 2008); however, our present findings now indicate that phosphorylation of multiple motifs is required to drive endocytosis. We hypothesize that these distinct cellular outputs might reflect the dynamic composition of signaling effectors in nephrin clusters. Indeed, our data indicate that higher amounts of N-WASp than dynamin-2 are bound to Nck, suggesting that more nephrin/Nck clustering may be required to increase the dwell time and localized activity of dynamin-2. Furthermore, mutation of Y1217 alone, which is the YDxV motif most strongly associated with Nck (Jones et al., 2006), leads to a greater accumulation of irregular clusters than the Y1176 or Y1193 mutations. Lastly, the engineered CD16/7-nephrin(Y3F)^IC-Nck2(SH3x3)-GFP fusion protein shows smaller puncta than CD16/7-nephrin(WT)^IC, yet it is still able to bypass the clustering defect, likely because there is no dissociation between nephrin and Nck, and the dwell time of Nck-associated effector proteins is therefore increased. A similar rescue effect is seen with CD16/7-nephrin(Y3F)^IC-TIR-GFP, which inducibly binds Nck with notably higher affinity than nephrin YDxV sites (Blasutig et al., 2008). Intriguingly, recent studies support a link between phase separation and endocytosis, wherein droplet formation generates a force that drives invagination of the plasma membrane (Bergeron-Sandoval et al., 2018, preprint). We therefore suggest that threshold levels of nephrin/Nck clustering are required to facilitate endocytosis.

Our findings further indicate that multisite phosphorylation of nephrin modulates the potency of Nck signaling, as has been shown for the T-cell receptor and its signaling effectors (James, 2018; Su et al., 2016). Several previously identified feed-forward mechanisms may enhance the available valency within nephrin-Nck clusters. Notably, Nck expression can augment nephrin tyrosine phosphorylation, and this is dependent on Nck binding to nephrin via its SH2 domain while binding to Fyn kinase via its SH3 domains (New et al., 2013). In addition, Nck and N-WASp act cooperatively to relieve their respective intramolecular inhibitory conformations (Fig. 8). First, N-WASp relies on an Nck SH3 interdomain linker region for its ability to recruit and activate Arp2/3 (Okrut et al., 2015), and this linker also contributes to phase separation (Banjade et al., 2015). Next, the high-affinity binding of N-WASp to Nck disrupts the SH3-mediated autoinhibition of Nck (Takeuchi et al., 2010), which may in turn allow subsequent binding of low-affinity interactors, such as dynamin.

The association of nephrin phosphorylation and clustering with actin assembly has been established; however, their connection with endocytosis is less well understood (Swiatecka-Urban, 2017). Several studies have shown that nephrin phosphorylation regulates its trafficking (Espiritu et al., 2019; Martin et al., 2018; Qin et al., 2009; Quack et al., 2006), and that dynamin is required for both clathrin-dependent and raft-mediated nephrin endocytosis (Espiritu et al., 2019; Qin et al., 2009). Consistent with these findings, we show that Fyn induces removal of nephrin from the cell surface, and that this effect is blocked with both the nephrin-Y3F mutant and the dynamin K44A mutant. Nephrin can be tethered by podocin into raft-like membrane clusters (Schwarz et al., 2001; Simons et al., 2001), and raft-mediated endocytosis has been proposed as a mechanism of slow internalization of phosphorylated nephrin following signal activation (Lahdenperä et al., 2003; Qin et al., 2009). By contrast, clathrin-mediated endocytosis is more rapid and may instead regulate steady-state recycling of nephrin (Espiritu et al., 2019; Satoh et al., 2014; Soda et al., 2012). We demonstrate that clathrin associates with nephrin clusters, and a recent study highlights a putative interaction between nephrin and clathrin (Espiritu et al., 2019). Altogether, our results support the cooperation of nephrin phosphorylation and endocytosis in slit diaphragm remodeling, and suggest that Nck functions at the interface of these signaling pathways.

Dynamin localization at actin tails requires its proline-rich region (Lee and De Camilli, 2002), and the arrested endocytic pits in dynamin-deficient MEFs are enriched for N-WASp, Arp2/3 and actin (Ferguson et al., 2009). In podocytes, Nck has been proposed as a linker protein between dynamin and nephrin at clathrin-coated pits (Soda et al., 2012), and an intermediary was similarly proposed to facilitate indirect interactions between tyrosine phosphorylated nephrin and dynamin in pancreatic β cells (Fornoni et al., 2010; Jeon et al., 2012). Our data suggest that the Nck SH3 domains couple dynamin to nephrin. Specifically, we demonstrate that expression of Nck SH3 domain mutants, which disrupt binding to N-WASp and dynamin-2, lead to defects in actin assembly and nephrin clustering, as we observed with nephrin-Y3F (Blasutig et al., 2008). As well, we show that direct fusion of the Nck SH3 domains or TIR YDxV sequence to nephrin-Y3F can restore dynamin-2 recruitment to nephrin and/or revert the clustering defects associated with mutant nephrin. It is tempting to speculate that the irregular clusters seen with the nephrin tyrosine mutants and Nck SH3 domain mutants, which have a tubular appearance, are continuous with the cell membrane, and colocalize with clathrin, in fact represent endocytic intermediates that fail to undergo dynamin-mediated fission.

Finally, our findings support the notion that a physiological relationship exists between the actin cytoskeleton and endocytosis in podocytes (Soda et al., 2012). Though the role of actin in endocytosis in mammalian cells is variable and thus debated (Aghamohammadzadeh and Ayscough, 2009; Dai and Sheetz, 1999), it is essential to counteract high membrane tension induced by stretch or at sites of strong adhesions, where actin filament assembly is proposed to provide the additional constriction force required to close the budding vesicle (Boulant et al., 2011). The lack of a uniform requirement for actin at endocytic sites may reflect multiple and possibly redundant mechanisms for the formation and fission of endosomes, as well as the local abundance of actin regulatory proteins, including N-WASp (Mooren et al., 2012). At sites where a dense meshwork of actin is present, such as focal adhesions, membrane deformation and movement may be restricted, and higher Arp2/3 activity (and possibly phase separation) may be required to generate and transmit sufficient force to pull the membrane inward (Mooren et al., 2012). This mechanism is likely of particular importance in the podocyte owing to the significant mechanical stretch and shear strain imposed on the slit diaphragm (Kriz and Lemley, 2017a,b), as well as the additional force needed to interrupt trans nephrin interactions (Endlich et al., 2017). Indeed, our findings in the NTS challenge model reveal a transient rise in nephrin phosphorylation on the YDxV motifs coincident with decreased nephrin surface expression and filtration barrier remodeling. We hypothesize that this change in nephrin valency alters the composition and function of the slit diaphragm, to allow dynamic switching between nephrin stabilization and nephrin turnover (Fig. 8). Altogether, we posit that multivalent nephrin–Nck interactions serve to regulate the density of signaling effectors at the slit diaphragm to tune downstream signaling responses.

Human embryonic kidney (HEK) 293T cells (from Tony Pawson, Lunenfeld-Tanenbaum Research Institute, Toronto) and Nck1/2 null MEFs (Jones et al., 2006) were grown in Dulbecco's high glucose modified eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) (both HyClone), 200 units/ml penicillin and 200 μg/ml streptomycin (Invitrogen) and maintained at 37°C with 5% CO₂. HEK cells were authenticated with GenePrint10 (Promega). Mouse podocyte cells (MPCs) were generated by isolating podocytes from Immortomouse [Charles River, official designation CBA;B10-Tg(H2Kb-tsA58)6Kio/Crl] according to standardized protocols (Shankland et al., 2007). MPCs were grown under permissive conditions in DMEF/12 (Hyclone) with 10% FBS, 200 units/ml penicillin, 200 μg/ml streptomycin (Invitrogen) and 20 units IFN-γ (Peprotech) at 33°C and 5% CO₂, then thermoswitched to 37°C with 2% FBS-containing medium for 10-14 days to induce differentiation. Cell lines were tested for the presence of Mycoplasma.

Transient transfection of HEK293Ts was performed using polyethylenimine (PEI) for 24-48 h, whereas transfection of MEFs was performed using Lipofectamine 3000 (Invitrogen) as per manufacturer's specifications. Constructs encoding human FLAG-tagged NCK1 and NCK2, human Myc-tagged full-length and green fluorescent protein (GFP)-CD16/7-nephrin and fusions and mutants thereof were generated previously (Blasutig et al., 2008; Jones et al., 2006; Keyvani Chahi et al., 2016; Li et al., 2004). GFP-dynamin2-WT and GFP-dynamin2-K44A was provided by Dr Costin Antonescu (Ryerson University, Toronto, Canada) and GFP-N-WASp was provided by Dr Samantha Gruenheid (McGill University, Montreal, Canada). The AdEasy XL adenoviral system (Agilent Technologies) was utilized to generate all viruses in this study for transient expression in MPCs (Bi et al., 2014). Differentiated MPCs were infected with the indicated viruses at a dilution of 1:100 in 2% FBS-containing DMEF/12 over 15 h.

For CD16 clustering experiments, transfected or transduced cells were serum-starved (HEK293Ts or MEFs) or placed in 2% serum (MPCs) for 15 h before stimulation with 200 ng anti-CD16 antibody (Santa Cruz Biotechnology, sc-19620, clone 3G8) per ml medium containing 10% FBS for 30 min at 37°C. Cells were then placed on ice and lysed in cold PLC+ lysis buffer (PLC+) (50 mM HEPES pH 7.5, 150 mM NaCl, 10% glycerol, 1% Triton X-100, 15 mM MgCl₂, 1 mM EGTA, 10 mM NaPPi, 100 mM NaF) supplemented with fresh protease and phosphatase inhibitors (1 mM PMSF, 1 mM sodium orthovanadate, 10 μg/ml aprotinin and 10 μg/ml leupeptin) by sonicating on ice for 10 s, incubating on ice for 5 min, and centrifuging at 14,000 g for 10 min at 4°C. Protein concentrations in supernatants were determined using a bicinchoninic acid (BCA) Protein Assay (Pierce). Supernatants were mixed with appropriate amounts of SDS sample buffer and incubated at 100°C for 5 min.

Immunoprecipitation (IP) of GFP (Abcam, 290) was performed using protein A beads for 1 h at 4°C with rotation. IP of CD16 (Santa Cruz Biotechnology, sc-19620, clone 3G8) was performed using mouse beads overnight at 4°C with rotation. IP incubations were followed by washing three times with PLC+ and protein complexes were eluted from the beads in 2× SDS loading buffer by boiling at 100°C for 2 min prior to proceeding to immunoblotting (IB). For IB, PVDF membranes were incubated with the following primary antibodies overnight at 4°C: goat anti-dynamin2 (Santa Cruz Biotechnology, sc-6400, 1:1000); rabbit anti-nephrin (a gift from Dr Tomoko Takano, McGill University, Montreal, Canada; 1:1000), rabbit anti-Nck (a gift from Dr Louise Larose, McGill University, Montreal, Canada; 1:1000) or anti-pY1176/1193 (Abcam, ab80299, 1:1000), mouse anti-FLAG M2 (Sigma-Aldrich, F3165, 1:2000), anti-myc 9E10 (Millipore, 05-419, 1:1000), anti-GFP B2 (Santa Cruz Biotechnology, sc-9996, 1:1000) or anti-pY 4G10 (Upstate Biotechnology, 16-101, 1:1000). After washing, secondary horseradish peroxidase (HRP)-conjugated donkey anti-goat (sc-2020, Santa Cruz Biotechnology), goat anti-rabbit (170-6515, Bio-Rad) or goat anti-mouse (170-6516, Bio-Rad), antibodies were used at 1:10,000 in TBST for 1 h at room temperature. Signals were detected using ECL (Pierce) and imaged using a ChemiDoc XRS+ (Bio-Rad) or exposed to film (Pierce). Densitometry was performed using ImageLab version 2.0 analysis software (Bio-Rad).

Nephrin-Y3F knock-in mice were generated by homologous recombination targeting mutation of mouse Y1191, Y1208 and Y1232 (corresponding to human Y1176, Y1193 and Y1217) to phenylalanine (F) as described previously (New et al., 2016) and back-crossed onto the CD-1 (official designation CD-1.B6;129Nphs1,Tm1.1./Njns) or C57BL/6N backgrounds. Both male and female animals were used unless otherwise noted. No statistical method was used to predetermine the sample size. The investigators were not blinded to animal genotypes or assignments, and no animals were excluded from the study. All animal studies were carried out in accordance with Canadian Council on Animal Care protocols and approved by the University of Guelph Animal Care Committee (Animal Use Protocol 3291).

Non-proteinuric male WT or nephrin-Y3F mice (both C57BL/6N) aged 8-10 weeks were injected with 0.04 mg/g of IgG γ2 subclass sheep anti-rat nephrotoxic serum (a gift from Dr David Salant, Boston University School of Medicine, MA, USA) or an equivalent dose of normal sheep IgG serum as a control (Capralogics, SG0010) via tail vein injection. Spot urine samples were collected prior to and at several time points after injection. Mice were euthanized with CO₂ at the indicated time points and kidneys were removed and immediately processed for analysis.

Mice were placed into a metabolic cage until they urinated freely, and spot urines were collected. For Coomassie urine gels, urine samples (2 or 5 μl) were diluted in sodium dodecyl sulfate (SDS) sample buffer, separated by 10% SDS-polyacrylamide gel electrophoresis (PAGE), and stained with Coomassie Brilliant Blue R. The urinary albumin/creatinine ratio was determined using the Albuwell M enzyme-linked immunosorbent assay (ELISA) (Exocell) and Creatinine Companion (Exocell) kits according to the manufacturer's instructions.

Isolated glomeruli or adherent cells were subjected to biotinylation and avidin precipitation as described previously (Martin et al., 2018). Where indicated, cells were stimulated with CD16 antibody at 37°C for 30 min unless otherwise noted as described in ‘Cell stimulation and lysis’ above. Subsequently, cell-surface proteins were biotinylated at 4°C for 45 min before lysis and avidin precipitation as described previously (Martin et al., 2018). Quantification of surface nephrin in each sample was determined by dividing the density of band observed in immunoblotting of the streptavidin-precipitated sample by the density of band observed in the input sample. The ratio of this value in test samples to the value obtained in control samples was calculated and represents the amount of surface nephrin in the indicated test groups relative to control samples.

Where indicated, cells were stimulated with CD16 antibody at 37°C for 30 min unless otherwise noted as described in ‘Cell stimulation and lysis’ above. For indirect immunofluorescence experiments, cells were first gently washed three times in PBS to remove residual CD16 antibody prior to further processing. Slides were then fixed for 10 min in 4% paraformaldehyde before being permeabilized in 1% Triton X-100 for 10 min to stain surface and cytosolic proteins or left unpermeabilized to label surface proteins. Samples were then blocked for 1 h in 10% goat serum, incubated with the indicated primary antibody and/or Texas Red phalloidin for 1 h followed by appropriate secondary antibodies (1:400) for 1 h and finally mounted using Prolong Gold mounting medium (Invitrogen), with intermediary washes in PBS. To view surface nephrin, cells transfected with various GFP-CD16/7-nephrin constructs (visible in the 488 wavelength) were stimulated with mouse anti-CD16 antibody at 37°C for 30 min, then cold-stained with anti-mouse Alexa Fluor 594 (to label surface nephrin in the 594 wavelength) prior to fixation and mounting. Where applicable, cells were also stained with rabbit anti-clathrin (Cell Signaling, 4796) and anti-rabbit Alex Fluor 647. In the merged immunofluorescent image, surface nephrin, which appears in both the 488 (total, green) and 594 (surface alone, red) channels, appears yellow, whereas endogenous nephrin, protected from anti-mouse Alexa Fluor 594 labeling, remains green. Epifluoresence images were obtained using Volocity software version 5.3.2 (Improvision) on a DMIRE2 microscope (Leica) using a 40× objective. Stacks were captured at 0.2 µm z-intervals, then deconvoluted using iterative restoration (95% confidence with 15 iterations) in Volocity. Confocal images were obtained using the 60× objective (oil immersion) with EZ-C1 software on a Nikon Eclipse Ti microscope.

For quantification of cluster morphology, cells containing non-spherical or irregular-shaped clusters were scored as positive for elongated clusters. The percentage of total cells scored as positive for elongated clusters was determined for each condition. A minimum of 100 cells per replicate were scored for MEFs and 50 cells per replicate for podocytes.

For SR-SIM imaging, kidneys were fixed in 4% paraformaldehyde and embedded in paraffin prior to sectioning. Prior to staining, slides were blocked for 1 h in 10% goat serum, then incubated with respective antibodies for 1 h. After three washes in PBS, slides were incubated with anti-rabbit secondary antibodies (1:400) for 1 h. Slides were washed, then mounted using Prolong Gold anti-fade mounting medium (Invitrogen). SR-SIM images were obtained using an Elyra LSM 880 microscope (Zeiss) and processed with ZEN software (Zeiss) to obtain the SIM.

For electron microscopy (EM), kidney tissue was fixed in 0.1 M sodium cacodylate buffer containing 4% paraformaldehyde and 2% glutaraldehyde (Electron Microscopy Sciences), post-fixed in 1% OsO₄, and dehydrated through graded ethanols. Transmission EM samples were embedded in Quetol–Spurr resin. Ultrathin sections were cut and stained with uranyl acetate and lead citrate and viewed using an FEI CM100 TEM. For scanning EM, samples were critical point dried and sputter coated with gold. Samples were viewed using a Hitachi S-540 SEM.

Data are reported as arithmetic mean±s.e.m. Data analysis was performed using GraphPad Prism version 5 and SAS version 9.4. For all analyses, P<0.05 was considered statistically significant. Two groups were compared using unpaired Student's t-test when data followed a Gaussian distribution; when it did not, data were analyzed using Mann–Whitney test where variances differed. When data were normalized so that all control values were identical, the one-sample t-test was used. Three or more groups were analyzed using ANOVA (analysis of variance) with post-hoc Tukey's HSD test when data followed a Gaussian distribution; where it did not, or when values were normalized to the control, data were analyzed using the non-parametric ANOVA Kruskal–Wallis test with post-hoc Dunn's Multiple Comparison test (Curtis et al., 2018).

We gratefully acknowledge C. Antonescu, S. Gruenheid, L. Larose and D. Salant for providing reagents; M. Krendel for training in viral construction; P. Lu for technical support; and J. McGlade for critical review of the manuscript.