Lack of CD2AP disrupts Glut4 trafficking and attenuates glucose uptake in podocytes

Insulin resistance is a primary defect contributing to the development of type 2 diabetes mellitus. It can be due to defects in the insulin signalling pathway, glucose transporter 4 (Glut4, also known as SLC2A4) trafficking, or both (Kahn et al., 1992; Maianu et al., 2001; Xiong et al., 2010). Glut4 is the principal insulin-responsive glucose transporter and thereby a key regulator of whole-body glucose homeostasis. In the absence of insulin, the majority of Glut4 resides in intracellular vesicles, but, upon insulin stimulation, Glut4 translocates to the plasma membrane, allowing facilitative transport of glucose into cells (Wang et al., 1999; Coward et al., 2005; van Dam et al., 2005).

Insulin resistance is associated with diabetic nephropathy, a serious complication of diabetes, in patients with type 1 and type 2 diabetes mellitus (Orchard et al., 2002; Parvanova et al., 2006), and correlates with the severity of albuminuria (Orchard et al., 2002; Ekstrand et al., 1998). Interestingly, glomerular visceral epithelial cells, also called podocytes, express Glut4 and Glut1 (also known as SLC2A1) and are insulin sensitive (Coward et al., 2005). Podocytes of diabetic db/db mice develop insulin resistance and fail to phosphorylate Akt in response to insulin stimulation (Tejada et al., 2008). Furthermore, podocyte-specific knockout of insulin receptor leads to glomerular lesions similar to those observed in diabetic nephropathy and development of albuminuria in the absence of hyperglycaemia (Welsh et al., 2010). This highlights the importance of insulin signalling for normal glomerular function.

We found previously that lipid phosphatase SHIP2, a negative regulator of insulin signalling, is expressed at a high level in the glomeruli of obese Zucker rats and diabetic db/db mice (Hyvönen et al., 2010). In line with this, overexpression of SHIP2 (also known as INPPL1) in cultured human podocytes diminished the effect of insulin on Akt phosphorylation. We also found that septin 7, a small GTPase, plays an important role in Glut4 trafficking in podocytes (Wasik et al., 2012). Interestingly, both SHIP2 and septin 7 form a complex with CD2-associated protein (CD2AP) (Hyvönen et al., 2010; Wasik et al., 2012), suggesting that CD2AP might regulate insulin signalling or glucose transporter trafficking.

CD2AP is a ubiquitously expressed cytoplasmic protein that functions as a scaffolding protein in different signalling and vesicle trafficking pathways (Huber et al., 2003; Schiffer et al., 2004; Kobayashi et al., 2004; Havrylov et al., 2008; Wasik et al., 2012). It is necessary for the formation of slit diaphragms and glomerular ultrafiltration, as CD2AP-knockout mice die at the age of 6 weeks due to renal failure (Shih et al., 1999). The mice exhibit a defect in the formation of multivesicular bodies, which implies that CD2AP is important for the degradation pathway (Kim et al., 2003). CD2AP also interacts with the active form of small GTPase Rab4 (Cormont et al., 2003; McCaffrey et al., 2001), which is involved in insulin-induced exocytosis of Glut4 (Imamura et al., 2003). Welsch et al. (2005) showed that CD2AP is present in a specific late endosomal compartment in podocytes and that it is involved in endosomal sorting and/or trafficking by regulating the assembly of actin on vesicles. CD2AP has also been shown to colocalize with coat protein I (COPI) vesicles and clathrin (Havrylov et al., 2008). The above data suggest that CD2AP has a role for in vesicle trafficking, and the association of CD2AP with proteins involved in insulin signalling and Glut4 trafficking led us to investigate the role of CD2AP in the regulation of glucose homeostasis.

In this study, we show that lack of CD2AP attenuates basal glucose uptake in podocytes and diminishes the effect of insulin in inducing glucose uptake. We used live-cell imaging and proximity ligation assays (PLAs) to elucidate the mechanisms causing the defect, and found that insulin-responsive Glut4 storage vesicle (GSV) trafficking is impaired in CD2AP^−/− podocytes. We found that CD2AP associates with GGA2, a clathrin adaptor and a protein, which sorts Glut4 to GSVs. We also found that CD2AP forms a complex with clathrin, suggesting that it has a role in regulating recycling of clathrin from the GSVs back to the trans-Golgi network. These findings reveal a new role for CD2AP in the formation of GSVs and as a regulator of GSV trafficking.

To analyse whether CD2AP had an effect on glucose transport into podocytes, we performed a [³H]2-deoxy-D-glucose (H³-2-DOG) uptake assay on cultured CD2AP^−/− and wild-type (WT) podocytes. The level of glucose uptake was 32% lower in CD2AP^−/− podocytes than in WT podocytes in the basal state (Fig. 1A). In response to insulin stimulation, glucose uptake in WT podocytes increased 19%, whereas CD2AP^−/− podocytes failed to increase their glucose uptake activity (Fig. 1B). Next, we analysed whether CD2AP regulates glucose uptake in muscle cells, the cell type responsible for the majority of insulin-stimulated glucose uptake and disposal. Knockdown of CD2AP in L6 myoblasts with siRNA led to a 50% reduction in CD2AP expression compared with control siRNA-transfected cells (Fig. S1A,B). This lowered glucose uptake by 20% in the basal state but did not affect insulin-stimulated glucose uptake, possibly due to the low efficiency of siRNA knockdown of CD2AP (Fig. S1C,D). These data indicate that glucose uptake was reduced in the basal state in CD2AP-knockout podocytes and in muscle cells expressing reduced levels of CD2AP, and that CD2AP^−/− podocytes fail to increase their glucose uptake in response to insulin.

To define the mechanism for reduced insulin-stimulated glucose uptake into CD2AP^−/− podocytes, we first analysed whether the insulin signalling pathway is affected in the absence of CD2AP. For this, we quantified the increase of serine 473 phosphorylation of Akt (also known as protein kinase B) and the increase of threonine 642 phosphorylation of AS160 (Akt substrate of 160 kDa, also known as TBC1D4) in response to insulin. Both CD2AP^−/− and WT podocytes responded to insulin dose-dependently, as visualized by phosphorylation of Akt and AS160 (Fig. S2A–D). We next investigated whether attenuated glucose uptake in the absence of CD2AP was due to reduced expression of endogenous glucose transporters Glut4 and Glut1, both shown to be insulin-responsive in podocytes (Coward et al., 2005). Immunoblotting indicated a 20% increase in the expression level of Glut4 in CD2AP^−/− podocytes relative to WT cells, whereas the level of Glut1 was not significantly changed (Fig. 1C,D). Collectively, the data show that reduced insulin-stimulated glucose uptake activity in CD2AP-depleted cells is apparently not due to reduced activity of the insulin signalling pathway or altered expression of endogenous glucose transporters.

To investigate whether absence of CD2AP disturbs the trafficking of Glut4, the main insulin-responsive glucose transporter, we produced CD2AP^−/− and WT podocyte cell lines stably expressing HA–Glut4–GFP. GFP was used to sort the cells with fluorescence-activated cell sorting according to their HA–Glut4–GFP expression levels to obtain homogenous populations of CD2AP^−/− and WT cells with similar expression levels of HA–Glut4–GFP (Fig. S2E–H). The HA tag is located in the extracellular domain of Glut4, allowing detection of Glut4 molecules incorporated into the plasma membrane by surface-labelling of unpermeabilized cells with an antibody against the HA tag. Interestingly, insulin stimulation appeared to increase the relative amount of HA signal on the cell surface in WT podocytes (Fig. 2A,B), whereas CD2AP^−/− podocytes failed to increase the amount of HA signal on the cell surface in response to insulin (Fig. 2C,D). To quantify the difference, we performed an On-Cell Western assay, which revealed that insulin stimulation increases the amount of HA–Glut4–GFP on the plasma membrane by 50% in WT podocytes (Fig. 2E). In CD2AP^−/− podocytes, however, the amount of HA signal on the plasma membrane appeared the same in starved and insulin-stimulated cells (Fig. 2E). This indicates that CD2AP is required for the trafficking of Glut4 to the plasma membrane in response to insulin.

To confirm the involvement of CD2AP in the regulation of Glut4 trafficking, we performed an HA uptake assay to quantify the amount of HA–Glut4–GFP fused with the plasma membrane and endocytosed into the cells in a time-dependent manner. For this, podocytes were stimulated with insulin and simultaneously allowed to bind to and internalize an antibody against HA tag for 0, 5, 15 and 30 min. The amount of anti-HA antibody internalized and bound to the plasma membrane was quantified, setting the value of the 0-min time-point to 1 for both CD2AP^−/− and WT podocytes, to allow comparison of the internalization of HA–Glut4–GFP between the different cell lines at different time-points. Even by 5 min, we observed a 17% increase in HA signal in WT podocytes relative to CD2AP^−/− podocytes. This difference increased to 42% after 15 min and to 72% after 30 min (Fig. 2F). This further confirms that CD2AP is essential for the proper trafficking of Glut4 to the plasma membrane in response to insulin stimulation.

To define which step of Glut4 trafficking is defective in the absence of CD2AP, we pulse-labelled HA–Glut4–GFP on the plasma membrane with an antibody against HA and followed the trafficking of the tagged transporter by live-cell imaging after insulin stimulation. In WT podocytes, HA–Glut4–GFP was endocytosed and trafficked dynamically in the basal state (time-points 0 min to 30 min) (Fig. 3A; Movies 1 and 2), indicating slow recycling of the transporter even in the absence of insulin stimulation. Dynamic trafficking was also observed in insulin-stimulated WT cells, as visualized by the change in the intensity of the signal for HA in the perinuclear region and the appearance and disappearance of the HA signal above the nucleus (time-points 32 min to 90 min) (Fig. 3A; Movies 1 and 2). CD2AP^−/− podocytes endocytosed HA–Glut4–GFP in the basal state (Fig. 3B, compare time-points 0 min and 15 min). However, HA–Glut4–GFP condensed upon internalization into clusters in the perinuclear region (Fig. 3B, arrow at time-point 15 min) and did not respond to insulin stimulation (time-points 32 min to 90 min) (Fig. 3B, Movies 3 and 4). The videos of CD2AP-knockout podocytes show slow, continuous movement of the HA signal, suggesting that the trafficking machinery is not totally destroyed even in the absence of CD2AP. This was confirmed by showing that transferrin is also endocytosed into CD2AP^−/− podocytes and traffics to the perinuclear region (Fig. S2I–L). Taken together, the results confirm that HA–Glut4–GFP trafficking is impaired in the absence of CD2AP and suggest that the defect lies in the intracellular trafficking of Glut4 but not in general endocytosis.

CD2AP is known to colocalize with the endoplasmic reticulum (ER) marker calnexin and the Golgi marker GM130 (also known as GOLGA2) in HeLa cells (Havrylov et al., 2008). This, together with the results of live-cell imaging, led us to investigate whether CD2AP is observed in the same sub-cellular compartments as Glut4 and Glut4 storage vesicle (GSV) markers in WT podocytes. As expected, we detected CD2AP in the ER-enriched, high-density microsomal fraction (HDM) and in the Golgi-enriched, low-density microsomal fraction (LDM) (Fig. 4A). We also probed the HDM and LDM fractions with antibodies against Glut4, IRAP (also known as LNPEP) and sortilin, known components of GSVs (Waters et al., 1997; Shi and Kandror, 2005), and found them in both fractions (Fig. 4A) as expected. Immunofluorescence microscopy revealed that, in serum-starved WT podocytes, IRAP localized in the perinuclear region, and after insulin stimulation it dispersed throughout the cells (Fig. 4B). In contrast, in CD2AP^−/− podocytes IRAP was detected in the perinuclear region in starved cells and remained localized in the perinuclear region after insulin stimulation (Fig. 4B), suggesting a defect in insulin-responsive GSV trafficking. Sortilin was randomly distributed throughout the cells in WT podocytes, but in CD2AP^−/− podocytes, the majority of sortilin condensed in the perinuclear region (Fig. 4C). Insulin did not induce clear translocation of sortilin in either cell type. Immunofluorescence staining also suggested that sortilin could be upregulated in the absence of CD2AP. In line with this, quantitative immunoblotting revealed a threefold increase in the expression level of sortilin in CD2AP^−/− podocytes compared with WT cells (Fig. 4D,E). These results imply that there is impairment in the insulin-stimulated translocation of insulin-responsive GSVs in the absence of CD2AP.

The presence of CD2AP in the same microsomal fractions as proteins essential for the formation of insulin-responsive GSVs led us to investigate whether CD2AP forms a complex with these proteins in podocytes and whether the interactions are regulated by insulin. We hypothesized that CD2AP could interact with GGA2, the only protein known to bind to the cytosolic domain of sortilin (Nielsen et al., 2001). Co-immunoprecipitation assay using three different antibodies directed against CD2AP revealed that, indeed, GGA2 forms a complex with CD2AP (Fig. 5A; Fig. S3A). GGA2 differs from the other two GGA family members (GGA1 and GGA3) by its properties and functions, as it does not bind to ubiquitylated cargo or auto-inhibit itself (Doray et al., 2002; McKay and Kahn, 2004; Shiba et al., 2004; Yogosawa et al., 2006). GGA1 did not co-precipitate with CD2AP, indicating that the interaction between CD2AP and GGA2 is specific (Fig. S3B). Duolink in situ proximity ligation assay (PLA) further confirmed the interaction between CD2AP and GGA2 in the basal state (Fig. 5B). As expected, the interaction could not be detected in CD2AP^−/− podocytes, confirming the specificity of the Duolink PLA assay (Fig. 5C). Furthermore, quantification of the PLA signal revealed a 35% increase in CD2AP–GGA2 interaction in response to insulin stimulation (Fig. 5D–F). The data suggest a role for CD2AP in the sorting of Glut4 to insulin-responsive GSVs.

As GGA2 is a clathrin adaptor (Puertollano et al., 2001), we next investigated, by using the Duolink PLA assay, whether insulin stimulation or lack of CD2AP affects the interaction between GGA2 and clathrin. We observed that in both starved and insulin-stimulated conditions, GGA2–clathrin interaction was increased in CD2AP^−/− podocytes compared with WT cells (Fig. 6A). In WT podocytes, the interaction between GGA2 and clathrin remained at a similar level in starved and insulin-stimulated cells (Fig. 6A). Immunofluorescence microscopy revealed that the GGA2–clathrin complex was scattered around the podocytes after insulin stimulation compared with starved cells (Fig. 6B,C). In contrast to WT podocytes, in CD2AP^−/− podocytes the clathrin–GGA2 interaction increased by 25% after insulin stimulation (Fig. 6A). Furthermore, in the absence of CD2AP, the GGA2–clathrin complex accumulated in the perinuclear area after insulin stimulation (Fig. 6D,E). Collectively, these data indicate that the lack of CD2AP increases the interaction between GGA2 and clathrin and leads to accumulation of the GGA2–clathrin complex in the perinuclear region, implying a trafficking defect or a defect in the uncoating of the clathrin-coated vesicles.

CD2AP has a clathrin-binding FxDxF motif (Brett et al., 2002), but their interaction has not been biologically verified. To investigate whether CD2AP forms a complex with clathrin, we performed a PLA assay. The results confirmed that CD2AP interacts with clathrin heavy chain and also revealed that the interaction is regulated by insulin (Fig. 7A–C); we detected 40% less PLA signal for CD2AP and clathrin after insulin treatment relative to the starved state (Fig. 7A). In both starved and insulin-stimulated states, the CD2AP–clathrin complex was scattered throughout cells (Fig. 7B,C). Collectively, these data indicate that CD2AP is associated with clathrin.

To define the mechanism by which CD2AP regulates GSV trafficking, we first analysed whether vesicles containing Glut4 bud properly from the donor membranes located in the trans-Golgi network (TNG) to form the GSVs. For this, we performed a 16,000 g fractionation that pellets the heavier donor membranes and leaves the lighter vesicles, including GSVs, in the supernatant. The ratio of Glut4 in donor membranes versus GSVs was the same in WT and CD2AP^−/− podocytes in the starved state. Insulin stimulation decreased the amount of Glut4 in the donor membrane fraction in a similar fashion in both WT and CD2AP^−/− podocytes, indicated by the decreased proportion of Glut4 in donor membranes versus GSVs (Fig. 7D,E). This indicates that Glut4-containing vesicles bud normally from the donor membranes in the absence of CD2AP.

As CD2AP has been shown to regulate the assembly of actin on vesicles (Carreno et al., 2004), we next investigated whether CD2AP links clathrin to actin by PLA assay. This revealed that indeed, clathrin heavy chain forms a complex with actin, and that lack of CD2AP decreases the PLA signal between actin and clathrin heavy chain by 38% (Fig. 7F–J). Insulin treatment decreased the signal by 24% in WT podocytes but had no effect on the complex formation between actin and clathrin in CD2AP^−/− podocytes (Fig. 7F). The PLA signal localized perinuclearly in both cell types (Fig. 7G–J). In all cases, the signal was fairly weak suggesting a transient interaction between clathrin and actin, as described previously (Carreno et al., 2004). We also probed clathrin from the 16,000 g fractions, and found that during starvation, clathrin is weakly expressed in the donor membrane fractions (Fig. 7K,L). In WT podocytes, insulin increased the proportion of clathrin in donor membranes versus GSVs, indicating that insulin increases recycling of clathrin to the donor membranes. This, however, did not occur in the absence of CD2AP, suggesting that delivery of free clathrin to the donor membranes is defective in CD2AP^−/− podocytes (Fig. 7K,L).

We next investigated whether CD2AP plays a role in glucose uptake in vivo using zebrafish as a model organism. Knockdown of CD2APL, the zebrafish orthologue of mammalian CD2AP, with a translation-blocking morpholino antisense oligonucleotide (TBMO) led to pericardial and yolk sac oedema (Fig. 8A,B), as shown previously (Hentschel et al., 2007). To analyse the role of CD2AP in glucose uptake, we injected a fluorescent 2-deoxyglucose analogue into the yolk sacs of 24-h post-fertilization (hpf) CD2APL-knockdown and control embryos. CD2APL-knockdown embryos showed a 42% reduction in glucose uptake after 10 min compared with the control embryos (n=21 per group) (Fig. 8C–E). This indicates a defect in glucose uptake in CD2APL morphants similar to that observed in cultured CD2AP^−/− podocytes.

In this study, we demonstrate that lack of CD2AP attenuates glucose uptake in basal and insulin-stimulated conditions in podocytes. We further found that knocking down CD2APL, a zebrafish orthologue of mammalian CD2AP, in zebrafish in vivo reduces glucose uptake from the yolk sack into the embryos, in a similar fashion to knockdown of Glut1 (Jensen et al., 2006). Live-cell imaging of cultured podocytes expressing HA–Glut4–GFP revealed that insulin-stimulated GSV trafficking is disturbed in the absence of CD2AP. Taken together, these findings point to a new role for CD2AP in glucose transporter trafficking, and thereby, in the regulation of glucose uptake into cells. We further show that CD2AP is associated with the clathrin adapter GGA2 and participates in sorting of Glut4 to insulin-responsive GSVs in podocytes.

CD2AP has previously been shown to bind, together with nephrin and podocin, to the p85 regulatory subunit of PI3K and to activate PI3K–Akt signalling in podocytes (Huber et al., 2003). CD2AP is not, however, essential for Akt activation, as we observed no difference in Akt phosphorylation between WT and CD2AP^−/− podocytes, consistent with a previous report (Schiffer et al., 2004). Neither could we observe differences in AS160 phosphorylation, which releases the Glut4-containing vesicles from the insulin-responsive vesicle compartment in response to insulin (Larance et al., 2005). The reduction in glucose uptake was also not due to reduced expression of glucose transporters, as we observed no difference in the expression of Glut1 in the absence of CD2AP, but the level of Glut4 was increased. This suggested that reduced glucose uptake could stem from a defect in the trafficking of glucose transporters that occurs during endocytosis of the transporter, during sorting in the vesicular compartment or during exocytosis. In line with this, CD2AP has previously been shown to play a role in regulating endocytosis of epidermal growth factor receptor (EGFR) in MDA-MB-231 human breast cancer cells and HeLa cells (Lynch et al., 2003) and in controlling endosome morphology and trafficking in CHO cells (Cormont et al., 2003). However, as assessed by HA uptake assays and live-cell imaging, we observed that HA–Glut4–GFP is also endocytosed into podocytes in the absence of CD2AP. In addition, transferrin was endocytosed into CD2AP^−/− podocytes. This implies that endocytosis of Glut4 is not regulated by CD2AP but that the trafficking defect lies in a later step. Accordingly, live-cell imaging illustrated that the majority of endocytosed Glut4 does not traffic back to the plasma membrane, instead remaining in a perinuclear cluster in insulin-stimulated CD2AP^−/− podocytes.

CD2AP has been shown to colocalize in the perinuclear region with key proteins that participate in the formation of vesicles, including COPI, clathrin and Rab5 (Havrylov et al., 2008). We found that CD2AP co-fractionates in the same subcellular fractions as the GSV markers IRAP, sortilin and Glut4 itself. Interestingly, sortilin, a key component in the formation of GSVs (Shi and Kandror, 2005, 2007), was overexpressed in CD2AP^−/− podocytes. Upregulation of sortilin is rare and has been reported only during differentiation of osteoblasts (Maeda et al., 2002) or in neurons upon accumulation of amyloid β_1–42 (Saadipour et al., 2013). In contrast, diabetes and obesity have been demonstrated to downregulate sortilin expression in adipose tissue and muscle (Kaddai et al., 2009) and in a state of insulin resistance in liver (Li et al., 2015). Sortilin overexpression stabilizes Glut4 (Shi and Kandror, 2005), which could explain the increased protein level of Glut4 that we observed in the CD2AP^−/− podocytes. As a transmembrane protein, sortilin controls the formation of GSVs by interacting with Glut4 and IRAP in the lumen of vesicles (Shi and Kandror, 2007) and by binding the clathrin adaptor GGA2 (Bilodeau et al., 2004) through its cytoplasmic tail (Takatsu et al., 2001). GGA2 then recruits clathrin and adaptors, facilitating the formation of insulin-responsive GSVs (Li and Kandror, 2005). Sortilin overexpression has been shown to increase the efficiency of the formation of GSVs and to induce glucose uptake into adipocytes (Shi and Kandror, 2005) and myocytes (Ariga et al., 2008). Despite upregulation of Glut4 and sortilin in the absence of CD2AP, glucose uptake into podocytes was reduced, and overexpressed HA–Glut4–GFP was not efficiently trafficked to the plasma membrane in response to insulin. This suggests that increased expression of endogenous Glut4 and sortilin in CD2AP^−/− podocytes is a compensatory mechanism, possibly caused by a defect in GSV formation (Shi and Kandror, 2005) or trafficking, or that the mechanisms differ between cell types.

We found that CD2AP interacts with GGA2 in podocytes and that the interaction is induced by insulin. GGA2 (also known as Vear) is specifically expressed in podocytes and not in other glomerular cell types (Poussu et al., 2001). We also verified that CD2AP interacts with clathrin, as expected based on the presence of the clathrin-binding FxDxF motif in CD2AP (Brett et al., 2002), and showed that the interaction is reduced upon insulin stimulation. This rules out the possibility of competitive binding of CD2AP and GGA2 with clathrin. After clathrin coats disassemble, clathrin and its adaptors are recycled back to donor membranes to form new clathrin coats (Jiang et al., 2000). Besides being a clathrin adaptor, GGA2 is involved in sorting of the newly synthesized Glut4 to GSVs and regeneration of GSVs from Glut4 transporters internalized from the plasma membrane (Li and Kandror, 2005). Increased interaction of CD2AP and GGA2 after insulin stimulation, together with the live-cell imaging data, implies a role for CD2AP in sorting the internalized Glut4 back to GSVs. We also quantified GGA2-clathrin interaction before and after insulin stimulation in WT and CD2AP^−/− podocytes. There was no difference in the complex formation between clathrin and GGA2 after insulin stimulation in WT cells, apparently because clathrin coats assemble and disassemble constantly when clathrin-coated vesicles form and circulate normally (Newmyer and Schmid, 2001). Lack of CD2AP, however, increased the interaction between GGA2 and clathrin, and insulin stimulation increased the interaction further. This, together with the finding that vesicles containing Glut4 bud properly from the donor membranes at the TNG, suggests that the clathrin-coated pits form and bud normally, but the vesicles fail to traffic properly in the absence of CD2AP, leading to their accumulation in the perinuclear region. This is further supported by concentration of the GSV proteins IRAP and sortilin in the perinuclear region in CD2AP^−/− podocytes after insulin stimulation.

We hypothesized that CD2AP could link clathrin to actin cytoskeleton, as CD2AP binds actin directly (Lehtonen et al., 2002) and has been shown to regulate the assembly of actin on vesicular structures (Welsch et al., 2005). Indeed, we found the actin–clathrin complex to be diminished in CD2AP^−/− podocytes. Actin has been suggested to help clathrin-coated vesicles to bud from the donor membranes (Carreno et al., 2004), but we did not observe a defect in the budding of Glut4 vesicles. The phenotype we observed in the absence of CD2AP, with clathrin remaining associated with its adaptor GGA2, and the majority of the GSVs staying in the perinuclear region despite insulin stimulation, could be due to (1) an uncoating defect, which could be caused by defective association of clathrin with actin, or (2) a defect in linking clathrin-coated vesicles to the actin cytoskeleton and thus initiating the trafficking of the GSVs. The first hypothesis is further supported by the fact that we observed clathrin recycling back to donor membranes after insulin stimulation only in podocytes expressing CD2AP. A phenotype resembling this, caused by an uncoating defect, has been reported recently (Pechstein et al., 2015). Still more research is required to elucidate the exact functions of the protein complexes including CD2AP, clathrin and GGA2.

In humans, CD2AP haploinsufficiency (Kim et al., 2003) or a homozygous mutation in CD2AP (R612Stop) results in focal segmental glomerulosclerosis (FSGS) (Löwik et al., 2007). In the case of R612Stop, the expression of CD2AP is lost (Löwik et al., 2007), and in the case of haploinsufficiency the expression is lowered by 80% (Kim et al., 2003). This raises an interesting question about the role of altered glucose uptake and metabolism in disease progression in these patients. A recent metabolomics study revealed an increased amount of glucose in urine of patients with FSGS as one of the factors distinguishing these patients from patients with three other types of glomerulopathies and healthy controls (Hao et al., 2013). It has also been suggested that metabolic disorders contribute to podocyte injury during development of obesity-related glomerulopathy (Chen et al., 2006). Altered glucose metabolism, specifically insulin resistance, is associated with the development of diabetic nephropathy in patients with type 1 and type 2 diabetes mellitus and might precede development of albuminuria by several years (Martin et al., 1992). Interestingly, a recent study has revealed that diabetic conditions downregulate CD2AP in murine podocytes in vitro and in an experimental model of diabetes in vivo (Ha et al., 2015). We have also previously shown that single-nucleotide polymorphisms in the CD2AP gene are associated with end-stage renal disease in patients with type 1 diabetes mellitus (Hyvönen et al., 2013). Here, we propose that a reduced level of CD2AP contributes to disturbances in glucose metabolism, as absence or a reduced level of CD2AP reduces glucose uptake into podocytes without affecting the activity of the insulin signalling pathway and might thus contribute to podocyte injury. The role of altered glucose metabolism in podocyte dysfunction is supported by recent studies indicating that glucose uptake into podocytes affects the nutrient-sensing pathways in podocytes and contributes to the development of diabetic nephropathy (reviewed in Kume et al., 2012). Furthermore, a reduction or lack of CD2AP also affects glucose uptake in other tissues, as shown here in muscle cells and in zebrafish embryos in vivo.

Collectively, we show that CD2AP is associated with clathrin and the clathrin adaptor GGA2, suggesting that CD2AP controls sorting of Glut4, and thus, regeneration of insulin-responsive GSVs. We further show that absence of CD2AP reduces trafficking of GSVs from the perinuclear region to the plasma membrane in podocytes and attenuates glucose uptake, possibly due to a defect in uncoating of clathrin-coated vesicles. The data suggest a new mechanism by which CD2AP regulates Glut4 trafficking.

Rabbit anti-CD2AP 209 and 211 were raised against amino acids 331–637 and 1–330 (Lehtonen et al., 2008), and rabbit anti-CD2AP 1764 against amino acids 6–574 (Lehtonen et al., 2000) of mouse CD2AP, respectively. Rabbit anti-GGA2 (sc-30103) and mouse anti-GGA2 (sc-133147) were from Santa Cruz Biotechnology). Rabbit anti-β actin (ab75186) and anti-sortilin IgG (ab-16640) were from Abcam (Cambridge, UK). Rabbit anti-phospho-Akt (Ser473) (9271) and rabbit anti-IRAP (3808) IgGs were from Cell Signaling Technology and mouse anti-pan Akt (MAB2055) IgG from R&D Systems. Mouse anti-α-tubulin (T6199) IgG was from Sigma-Aldrich. Rabbit anti-Glut4 (07-1404) and rabbit anti-Glut1 (07-1401) and anti-phospho-AS160 (Thr642) (ABS271) IgGs were from Millipore. Mouse anti-HA.11 (MMS-101R) and mouse anti-HA.11 Alexa-Fluor-594-labelled (A594-101L) IgGs was from Covance (Princeton, NJ), and mouse anti-clathrin heavy-chain IgG (MA1065) was from Thermo Fisher Scientific (Waltham, MA).

Conditionally immortalized mouse CD2AP^−/− and WT podocytes were maintained in high-glucose Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), ultraglutamine, penicillin, streptomycin and 10 U/ml interferon-γ at +33°C (Schiffer et al., 2004). All cell culture reagents were from Sigma-Aldrich, except ultraglutamine, which was from Lonza (Basel, Switzerland).

Glucose uptake was measured using 50 µmol/l (1 µCi/ml) 2-deoxy-D-[(1, 2-[³H](N)]-glucose (PerkinElmer), as previously described (Wasik et al., 2012).

L6 cells (CRL-1458; ATCC) were maintained in same medium as described above but without interferon-γ and at +37°C. Cells were transfected with 100 nmol ON-TARGET plus SMARTpool rat CD2AP siRNA (L-098575-02), or siCONTROL Non-Targeting Pool#2 siRNA (D-001206-14-05) (Dharmacon) using Lipofectamine 2000 (Thermo Fisher Scientific). Cells were used for experiments after 72 h.

HA–Glut4–GFP was subcloned from pHA-GLUT4-GFP (Dawson et al., 2001) into a lentiviral pSIN18.cppt.hEF1αp.WPRE vector (Gropp et al., 2003) to generate hEF1-HA-Glut4-GFP. HEK293FT cells were co-transfected with hEF1-HA-Glut4-GFP and packaging plasmids KS pCMVΔ8.9 and pHCMV using Lipofectamine 2000 (Invitrogen). Media were collected after 3 days, filtrated through a 0.45 µm filter and ultracentrifuged at 85,000 g at +4°C for 90 min to concentrate the viruses. CD2AP^−/− and WT podocytes were infected with viruses resuspended in PBS. Cells expressing HA–Glut4–GFP were selected with 0.625 µg/ml (WT) and 2 µg/ml (CD2AP^−/−) puromycin for 14 days, and the cells were sorted by fluorescence-activated cell sorting (FACS) based on GFP fluorescence. Cells expressing HA–Glut4–GFP at similar mean fluorescence intensity were used for experiments.

CD2AP^−/− and WT podocytes expressing HA–Glut4–GFP were grown on 96-well plates, starved for 20 h, or starved for 20 h and stimulated with 20 nM insulin (Actrapid, Novo Nordisk, Denmark) for 15 min, and fixed with 2% paraformaldehyde (PFA) for 20 min. Cells were incubated with an antibody against HA diluted 1:100 in Odyssey Blocking buffer (LI-COR Biosciences, Lincoln, NE) for 1 h followed by IRDye 800 donkey anti-mouse IgG and nuclear marker DRAQ-5 (Thermo Fisher Scientific), which was used for normalization. Detection and quantification were performed with Odyssey Infrared Imager (LI-COR Biosciences).

Cells were cultured as for On-Cell Western. HA antibody was diluted 1:100 in serum-free medium with or without 20 mM insulin. Cells were rinsed with PBS and incubated with the antibody-containing medium for 30, 15 or 5 min at +33°C. Thereafter, the plate was transferred on ice and the antibody-containing medium was added to the 0 min wells to label HA on the plasma membrane. The plate was incubated on ice for 15 min, after which the cells were rinsed with PBS, fixed with 2% PFA for 20 min and permeabilized with 0.1% Triton X-100 in PBS for 10 min. Cells were incubated with IRDye 800 donkey anti-mouse IgG and DRAQ-5 (Thermo Fisher Scientific). Detection and quantification were performed with Odyssey Infrared Imager (LI-COR Biosciences).

CD2AP^−/− and WT podocytes stably expressing HA–Glut4–GFP were cultured on Nunc Lab-Tek II Chamber Slides (Thermo Fisher Scientific), washed with ice-cold PBS and incubated on ice with Alexa Fluor 594 anti-HA.11 IgG (Covance) diluted 1:100 for 20 min. Before imaging, the cells were pre-warmed at +33°C for 10 min. After 30 min of imaging, the cells were stimulated with 20 nM insulin and imaging was continued for 1 h. Time-lapse video imaging was performed with a Leica TCS CARS SP8 confocal microscope. 3D-rendering and creation of the time-lapse videos were performed using Imaris 7.5 software (Bitplane Scientific Software, Zurich, Switzerland).

WT podocytes were washed three times with ice-cold HES buffer (20 mM Hepes pH 7.4, 1 mM EDTA, 225 mM sucrose) and homogenized with ice-cold HES buffer supplemented with 1× Complete™ proteinase inhibitor cocktail (Roche), 50 mM NaF and 1 mM Na₃VO₄. Nuclei were removed by centrifugation (1000 g for 5 min). Membrane and cytosolic fractions were separated by centrifugation of the post-nuclear supernatant (PNS) at 19,000 g for 22 min. The cytosolic fraction was centrifuged at 41,000 g for 22 min to pellet the high-density microsomal fraction (HDM). The obtained supernatant was centrifuged at 180,000 g for 77 min to pellet the low-density microsomal fraction (LDM). All centrifugation steps were performed at +4°C. The HDM and LDM pellets were re-suspended in HES buffer supplemented with phosphatase and proteinase inhibitors and analysed by immunoblotting as described below.

CD2AP^−/− and WT podocytes were washed and homogenized as above in ice-cold PBS with 1× Complete™ proteinase inhibitor cocktail, 50 mM NaF and 1 mM Na₃VO₄. Nuclei were removed by centrifugation (500 g at +4°C for 10 min). Donor membranes and the GSV vesicle fraction were separated by centrifugation at 16,000 g at +4°C for 20 min as previously described (Lamb et al., 2010). Supernatant was collected into a separate tube and the pellet was re-suspended in PBS. The volume of the pellet fraction loaded to the gel was five times the volume of the supernatant.

Cells were lysed with 1% NP-40 buffer (1% Nonidet P-40, 20 mM Tris-HCl, pH 7.4, 150 mM NaCl) supplemented with 1× Complete™ proteinase inhibitor cocktail (Roche), 50 mM NaF and 1 mM Na₃VO₄. Proteins were separated by SDS-PAGE and transferred to PVDF-FL membranes (Millipore), blocked with low-fat milk in Tris-buffered saline (TBS; 20 mM Tris-HCl, pH 7.4, 150 mM NaCl), and incubated with primary antibodies (CD2AP 1764, 1:6000; tublin, 1:3000; p-Akt, 1:1000, panAkt, 1:1000; p-AS160, 1:1000; clathrin, 1:1000, Glut1 and Glut4, 1:1000; IRAP, 1:1000; sortilin, 1:1000 and GGA2, 1:500) followed by Alexa-Fluor-680-labelled anti-rabbit- or anti-mouse-IgGs (Molecular Probes), IRDye-800-labelled donkey anti-rabbit- or anti-mouse-IgGs, or True Blot-HRP-conjugated anti-rabbit-IgG (Rockland Immunochemicals, Gilbertsville, PA). Detection and quantification were performed with an Odyssey Infrared Imager (Li-COR Biotechnology) or by enhanced chemiluminescence using SuperSignal West Pico reagent (Thermo Fisher Scientific).

CD2AP^−/− and WT podocytes were stimulated with 20 nM insulin as described above, fixed with 2% PFA, permeabilized with 0.1% Triton X-100 and blocked with CAS-block (Zymed). Cells were incubated with primary antibodies (sortilin, 1:300; IRAP, 1:50) diluted in ChemMate (Dako) for 1 h followed by Alexa-Fluor-488-, -555- or -594-labelled goat-anti-mouse or goat anti-rabbit IgGs diluted in ChemMate for 1 h. Samples for light microscopy were mounted in Moviol and examined with a Zeiss Axiophot 2 microscope (Carl Zeiss Microscopy GmbH, Jena, Germany). Samples for confocal microscopy were mounted in Vectashield (Vector, Burlingame, CA) and examined with a Leica TCS CARS SP8 confocal microscope.

Mouse podocytes were lysed as described above. Lysates were precleared with TrueBlot anti-rabbit Ig IP beads (eBioscience, San Diego, CA) at +4°C for 1 h and incubated with 5 µl of anti-CD2AP 1764 or normal rabbit serum at +4°C for 16–20 h. Immune complexes were bound to TrueBlot beads at +4°C for 1 h, washed three times with lysis buffer, boiled in Laemmli sample buffer and immunoblotted as described above.

CD2AP^−/− and WT podocytes were treated with insulin as described above and fixed with 2% PFA. Dilutions of primary antibodies were: anti-CD2AP 1764, 1:600; anti-mouse GGA2, 1:100; anti-mouse clathrin, 1:100; and anti-rabbit actin, 1:300. Duolink in situ proximity ligation assay was carried out according to the manufacturer's instructions (Olink Biosciences, Sigma-Aldrich). Imaging was performed with a Zeiss Axiophot 2 microscope and quantification with the Duolink ImageTool software (Olink Biosciences), as described previously (Madhusudhan et al., 2015) with minor modifications. Signal threshold was set to 150 and size threshold to 5 pixels. The size of cytoplasm was set to 200 pixels, and cells at the borders of the image were excluded. Only signal inside the defined area of the cytoplasm was included, and the averages of the signal per cell per image were counted. For each experiment, at least 50 cells were counted per cell type and condition.

The wild-type Turku zebrafish line was maintained and raised as described previously (Whitlock and Westerfield, 2000). All animal experiments were approved by the National Animal Experiment Board and were performed according to approved guidelines. Embryos were staged according to somite number or hours or days post-fertilization (Kimmel et al., 1995). Morpholino antisense oligonucleotide (MO), which blocks translation of CD2APL (5′-CATACTCCACCACCACCTCAACCAT-3′) (Hentschel et al., 2007), and a standard control MO (5′-CCTCTTACCTCAGTTACAATTTATA-3′) were obtained from Gene Tools (LLC, Philomath, OR). MOs (4 ng) were injected into fertilized eggs using a PLI-90 microinjector (Harvard Apparatus, Cambridge, MA).

24-hpf control and CD2APL morphants were injected in the yolk with 5 mg/ml 2-(N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl) amino)-2-deoxyglucose (2-NBDG), a fluorescent glucose analogue (Molecular Probes) as described in Jensen et al., 2010. After 10 min, the embryos were fixed with 4% PFA and imaged with a Zeiss Axioplan2 microscope. The mean fluorescence intensity of the embryos, excluding the yolk, was measured with constant settings using ImageJ 1.43u software (n=21 per group).

We thank Dr A. Shaw (Washington University School of Medicine, St Louis, MO, USA) for kindly providing CD2AP^−/− and WT podocytes, and Dr S. Cushman (National Institutes of Health, Bethesda, MD, USA) for the pHA–Glut4–GFP construct. Dr N. Peitsaro (Biomedicum FACS Core Facility, University of Helsinki, Finland) is thanked for help with the sorting of HA–Glut4–GFP cell line populations, and Dr A. Isomäki (Biomedicum Imaging Unit, University of Helsinki, Finland) is acknowledged for help with live-cell imaging.