Endoplasmic reticulum–retained podocin mutants are massively degraded by the proteasome

Podocin is a key component of the slit diaphragm in the glomerular filtration barrier, and mutations in the podocin-encoding gene NPHS2 are a common cause of hereditary steroid-resistant nephrotic syndrome. A mutant allele encoding podocin with a p.R138Q amino acid substitution is the most frequent pathogenic variant in European and North American children, and the corresponding mutant protein is poorly expressed and retained in the endoplasmic reticulum both in vitro and in vivo. To better understand the defective trafficking and degradation of this mutant, we generated human podocyte cell lines stably expressing podocinwt or podocinR138Q. Although it has been proposed that podocin has a hairpin topology, we present evidence for podocinR138Q N-glycosylation, suggesting that most of the protein has a transmembrane topology. We find that N-glycosylated podocinR138Q has a longer half-life than non-glycosylated podocinR138Q and that the latter is far more rapidly degraded than podocinwt. Consistent with its rapid degradation, podocinR138Q is exclusively degraded by the proteasome, whereas podocinwt is degraded by both the proteasomal and the lysosomal proteolytic machineries. In addition, we demonstrate an enhanced interaction of podocinR138Q with calnexin as the mechanism of endoplasmic reticulum retention. Calnexin knockdown enriches the podocinR138Q non-glycosylated fraction, whereas preventing exit from the calnexin cycle increases the glycosylated fraction. Altogether, we propose a model in which hairpin podocinR138Q is rapidly degraded by the proteasome, whereas transmembrane podocinR138Q degradation is delayed due to entry into the calnexin cycle.

Nephrotic syndrome is clinically characterized by proteinuria, edema, hypoalbuminemia, and hyperlipidemia, and is a consequence of glomerular filtration barrier (GFB) 2 dysfunction. The prognosis of steroid-resistant nephrotic syndrome (SRNS) is poor, with a high proportion of patients rapidly developing end-stage renal disease, requiring dialysis or transplantation (1,2). The GFB is comprised principally of podocytes, specialized epithelial cells that interdigitate at junctions known as slit diaphragms (SDs). Mutations in the NPHS2 gene, encoding the slit diaphragm (SD) protein podocin, are the most frequent monogenic cause of SRNS in childhood (3,4). Importantly, the missense mutation NPSH2: R138Q (p.R138Q) accounts for 20% of all SRNS-causing alleles in Europe and North America (4), and is associated with an early-onset and rapidly progressing form of the disease (5). In accordance with the severe clinical phenotype, podocin p.R138Q (Pod R138Q ) is retained in the ER of podocytes and does not reach the SD, thus impairing correct functioning of the GFB (6,7).
Podocin belongs to the stomatin and prohibitin homology domain (PHB) protein family and is specifically expressed in podocytes. It has been proposed that podocin acts as a molecular scaffold for other SD proteins in lipid raft membrane subdomains. For example, podocin interacts with both nephrin and CD2AP through its carboxyl terminus, and participates in various signaling events at the SD (8 -10). Podocin has a predicted hairpin-loop topology, with both the N and C termini facing the cytoplasm, and its hydrophobic domain anchored either to the inner leaflet of the plasma membrane (PM) or to the outer leaflet of the ER-membrane (7). However, an alternative transmembrane topology has been described for stomatin and podocin, particularly when a conserved proline residue, critical for the kink of the hairpin topology, found at position 118 in podocin, is mutated, and has been linked to N-glycosylated forms of these proteins (11,12). In addition, although the stability and degradation of podocin has been associated with a short internalization motif located in its C terminus and to an interaction with the ubiquitin ligase Ubr4 (13,14), the mecha- cro ARTICLE nisms of Pod R138Q degradation are still unexplored. In this study, we report the degradation pathways followed by Pod R138Q , which might aid to establish new therapeutic strategies.

Podocin R138Q is predominantly N-glycosylated
We generated two cell lines stably expressing two human influenza hemagglutinin (HA) tags at the N terminus of wildtype (WT) podocin (Pod wt ) and Pod R138Q by lentiviral transduction of a human podocyte cell line. Although both cell lines expressed comparable levels of podocin mRNA (Fig. 1A), we found significantly higher levels of Pod wt protein in comparison to Pod R138Q (Fig. 1B). We then confirmed by immunofluorescence the expected subcellular localization at the PM for Pod wt (co-localization with the wheat germ agglutinin, WGA) and at the ER for Pod R138Q (a reticular pattern staining partially colocalizing with calnexin) (Fig. 1C) (7). We demonstrated that neither Pod wt nor Pod R138Q stable overexpression caused ER stress as evidenced by the lack of up-regulated levels of BiP protein (GRP78), an ER-chaperone widely used as an ER-stress indica-tor ( Fig. S1) (15)(16)(17). As positive controls of ER stress induction caused by misfolded mutant proteins, we generated human podocyte cell lines stably expressing V5-tagged nephrin wt (Neph wt ) or nephrin S366R (Neph S366R ), because this latter mutant has been previously shown to induce ER stress (Fig. S1A) (18).
Interestingly, we observed three specific bands for both Pod wt and Pod R138Q on podocin immunoblots. The fastest migrating band (lower band) was predominantly observed in Pod wt protein extracts, whereas the slowest migrating band (upper band) was predominantly found in Pod R138Q extracts (Fig. 1B). To determine whether the upper bands in the WT and mutant podocin triplets were N-glycosylated forms, we either inhibited N-glycosylation by treating cells with tunicamycin, or treated cell lysates with peptide:N-glycosidase F (PNGase F) to digest N-glycan groups. Upper and middle bands of the podocin triplets disappeared using both strategies (Fig. 1D) allowing us to define the upper, middle, and lower bands of the podocin triplets as double-glycosylated (dg), monoglycosylated (mg), and non-glycosylated (ng) podocin. To test whether the difference in protein levels of Pod wt and Pod R138Q was due to these , and immunofluorescence analysis (C) comparing podocin expression levels, protein band distribution, and subcellular localization, respectively, in 2HA-Pod wt and 2HA-Pod R138Q expressing podocyte cell lines. B, ng, non-glycosylated podocin; mg, mono-glycosylated podocin; dg, doubleglycosylated podocin. Quantification of total podocin (ng, mg and dg) from three independent experiments are shown as mean Ϯ S.D. *, p Ͻ 0.05. HA monoclonal antibody was used to identify podocin and ␤-actin served as loading control. C, polyclonal anti-podocin AP-P35 (Pod) and monoclonal calnexin AF18 (Cnx) were used as primary antibodies. Cell membrane and nuclei were labeled with WGA and Hoechst (H), respectively. Scale bar ϭ 30 m. D, cells were treated overnight with 10 g/ml of tunicamycin (Tm) to impair N-glycosylation (upper immunoblot). Alternatively, cell lysates were treated with PNGase F, an enzyme that digests N-glycans from glycoproteins (lower immunoblot). Polyclonal AP-P35 or monoclonal anti-HA were used to immunoblot podocin. E, immunoblots of different podocin mutants known to possess different subcellular localizations. ER, endoplasmic reticulum-retained podocin mutants; V, vesicular podocin mutants; PM, plasma membrane localized podocin mutants. Podocin R138Q quality control and ERAD differential glycosylation patterns, we also created podocyte cell lines stably expressing HA-tagged Pod wt and Pod R138Q bearing mutations in the amino acid residues required for N-glycosylation, Asn 199 and Asn 355 (Pod wt(N199Q,N355S) and Pod R138Q(N199Q,N355S) ). Immunoblots of protein extracts from these cells revealed that although the podocin triplet disappears with mutation of residues Asn 199 and Asn 355 (Fig.  S1B), glycosylation alone does not account for the differences in protein levels of Pod wt and Pod R138Q , as there remains increased amounts of Pod wt(N199Q,N355S) compared with Pod R138Q(N199Q,N355S) .
Our results showed that the mutant Pod R138Q was predominantly N-glycosylated, similarly to Pod P118L (12), another ERretained podocin mutant, suggesting that the majority of intracellular Pod R138Q has a transmembrane topology (11,12). However, a small part of the Pod wt pool was also glycosylated, consistent with the observations of WT podocin and stomatin by other authors (11,12). We tested by immunoblot the protein expression pattern of a series of podocin mutants that present different subcellular localizations, confirming that only ER-retained mutants, such as Pod R168C , were enriched in N-glycosylated forms (Fig. 1E).

Podocin R138Q has a shorter half-life than Pod wt and is rapidly degraded by the proteasome
To investigate whether the difference in protein levels of Pod wt and Pod R138Q was due to a higher degradation rate of the mutant protein, we determined the half-life (t1 ⁄ 2 ) of both proteins using a cycloheximide time course experiment, quantifying separately the glycosylated and non-glycosylated forms (Fig. 2). No differences were observed between the t1 ⁄ 2 of the three Pod wt forms, whereas a drastically reduced t1 ⁄ 2 was detected for ng-Pod R138Q when compared with Pod wt (7-fold). Consistent with this finding, we also demonstrated that the t1 ⁄ 2 of the glycosylation mutant Pod R138Q(N199Q,N355S) was also reduced around 7-fold compared with Pod wt(N199Q,N355S) (Fig. S2). Interestingly, ng-Pod R138Q (i.e. hairpin-like to-pology) appeared to be degraded faster than the N-glycosylated forms of Pod R138Q (i.e. those with transmembrane topology), implying that the hairpin-like structure is more exposed to the intracellular degradative machinery ( Fig. 2 and Fig. S2).
Because ng-Pod R138Q is degraded very quickly, we hypothesized that the proteasome was mediating its clearance. We therefore used the reversible proteasome inhibitor bortezomib (Bz) to perform dose-response and time course experiments (Fig. 3). We observed that ng-Pod R138Q accumulated in a dosedependent manner, whereas levels of ng-Pod wt increased significantly only at the highest Bz dose of 1 M (Fig. 3A). Consistent with their longer t1 ⁄ 2 , both glycosylated Pod wt and glycosylated Pod R138Q were not increased at any dose of Bz after 2 h (Fig. 3A). Thus, we next performed an overnight time course (16 h) using the lowest effective dose of Bz (0.1 M) (Fig. 3B). We found that only ng-Pod R138Q accumulated with time, and, most interestingly, the balance between ng-and dg-Pod R138Q was inverted, thus after 16 h the proportions of each Pod R138Q band resembled the proportions of each Pod wt band in podocin immunoblots (Fig. 3B, colored graph). Levels of all forms of Pod wt were increased only at the longest time points. Similar findings were observed when we performed an overnight Bz time course on cells expressing the glycosylation mutants Pod wt(N199Q,N355S) and Pod R138Q(N199Q,N355S) , with protein levels of the WT protein only significantly increased at the longest time point of 16 h, and a trend for levels of Pod R138Q(N199Q,N355S) to accumulate in the presence of 0.1 M Bz (Fig. S3A). The increased amount of Pod wt and Pod R138Q after an overnight treatment with 0.1 M Bz was confirmed by immunofluorescence analysis (Fig. 3C). As expected, a sustained exposure to Bz led to ER stress, as indicated by BiP induction (Fig. 3C), but to a higher extent in cells expressing Pod R138Q (Fig. S3). Therefore, we concluded that the proteasome contributes to the degradation of both Pod wt and Pod R138Q , but more actively for ng-Pod R138Q .

Podocin R138Q quality control and ERAD Podocin R138Q is not degraded by lysosomes
Because protein degradation proceeds via two major routes: the proteasomal and the autophagic-lysosomal pathways, we investigated whether lysosomal degradation contributes to Pod wt and Pod R138Q proteolysis. We therefore treated the cells with ammonium chloride (NH 4 Cl), a weak base known to inhibit lysosomal proteases. NH 4 Cl treatment significantly increased levels of the ng-Pod wt protein fraction and levels of Pod wt(N199Q,N355S) , implying that the lysosomal machinery is involved in Pod wt degradation but not in Pod R138Q proteolysis; indeed, we observed levels of both Pod R138Q and Pod R138Q(N199S,N355S) to fall after a 16-h NH 4 Cl treatment ( Fig.  4A and Fig. S4B). Then, by confocal microscopy, we confirmed the absence of Pod R138Q in the late endosome/lysosome compartment labeled with CD63 (Fig. 4B). Conversely, Pod wt is pre-dominantly present in this compartment, as already described (13). Taken together, our data suggest that Pod wt is mainly degraded in lysosomes, in contrast to Pod R138Q , which is exclusively degraded by the proteasome.

Podocin R138Q enhanced interaction with calnexin
Misfolded N-glycosylated proteins located in the ER membrane are submitted to a strict quality control (QC) by the socalled calnexin (Cnx) cycle, being retained by the 90-kDa chaperone Cnx until they reach their native conformation or are otherwise sent for ER-associated degradation (ERAD). We hypothesized that if Pod R138Q is mostly N-glycosylated and possesses a transmembrane topology, interaction with Cnx may be a mechanism of Pod R138Q ER retention, as has already been described for some nephrin mutants (18). Immunoblots of . Asterisks refer to non-stimulated cells (ns or 0 h). *, p Ͻ 0.05 and **, p Ͻ 0.01. C, podocin was detected by incubation with monoclonal HA primary antibody (in green). Hoechst nuclei labeling was included (in blue). All images were taken using the same confocal microscope settings to allow comparison of the intensity of fluorescence. Scale bar ϭ 40 m.

Podocin R138Q quality control and ERAD
HEK293T cells transfected with HA-tagged WT podocin and the two ER-retained mutants, Pod R138Q and Pod R168C (7), showed that similarly to our findings from podocyte cell lines stably expressing Pod wt and Pod R138Q podocin was present as three differentially glycosylated species (Fig. 5A). Co-immunoprecipitation studies revealed an interaction between Cnx and both WT and mutant podocin, with the mutant proteins interacting to a greater extent compared with the WT protein (Fig.  5A). That we found Cnx also interacted with Pod wt was not surprising, because HA immunoprecipitation enriched a certain proportion of N-glycosylated Pod wt (Fig. 5A, upper panel). Surprisingly, ng-podocin, WT, and mutants, also co-immunoprecipitated with Cnx, suggesting that hairpin podocin might interact with Cnx either indirectly, perhaps through oligomerization with the podocin transmembrane fraction via its N terminus (19), or directly, through membrane or cytosolic domains. To test whether impairing the interaction of Pod R138Q with Cnx would allow its ER exit and promote membrane localization, we first treated the cells with castanospermine (Cst), a specific inhibitor of glucosidases I and II that prevents the sugar trimming necessary for the recognition of the substrate protein N-glycan groups by the lectin domain of Cnx. Cst decreased Cnx interaction with Pod wt , but not with Pod R138Q (Fig. 5B), suggesting that Pod R138Q interaction with Cnx is N-glycan independent. Similar results were obtained when these experiments were performed using cells expressing glycosylation mutants Pod wt(N199Q,N355S) and Pod R138Q(N199Q,N355S) (Fig. 5C), supporting the idea that the interaction with Cnx is more lectin dependent for Pod wt than for Pod R138Q . Finally, we found that introduction of p.N199Q and p.N355S substitutions into the Pod R138Q mutant in podocytes did not prevent its ER retention, implying that suppression of the N-glycan-dependent interaction with Cnx is not sufficient to bring Pod R138Q to the PM (Fig.  S4A), and supporting the idea that podocin is capable of lectindependent binding to Cnx.

Glycosylated podocin R138Q enters the calnexin cycle
Because we observed Pod R138Q(N199Q,N355S) is still able to interact with Cnx, we next knocked-down Cnx in the two stable podocyte cell lines using siRNA (Fig. 6A and Fig. S5A). Inter-

Podocin R138Q quality control and ERAD
estingly, we observed that levels of N-glycosylated forms of Pod R138Q were decreased upon Cnx knockdown, in contrast to levels of ng-Pod R138Q , which were significantly increased (Fig. 6A, colored graph), suggesting that Cnx may play a role in stabilization of the transmembrane form of N-glycosylated Pod R138Q , or indeed, that interaction with Cnx promotes glycosylation of Pod R138Q . However, in support of the former statement, the total amount of Pod R138Q was slightly decreased after Cnx knockdown, suggesting that interaction with Cnx might delay Pod R138Q degradation (Fig. 6A, lower graph). Next, we treated cells with kifunensine (Kif), a drug that inhibits the activity of ␣-mannosidase I, and thus the mannose trimming that tags Cnx substrates for ERAD (Fig. 6B). Here, our results were precisely the opposite to those obtained upon Cnx knockdown, that is, levels of ng-Pod R138Q were significantly reduced, whereas levels of dg-Pod R138Q were increased (Fig. 6B, colored  graph). Furthermore, the total amount of Pod R138Q was augmented after 4 h of treatment (Fig. 6B, lower graph), demonstrating that Kif treatment partially prevented Pod R138Q degradation. These last data support the idea that only N-glycosylated, and thus transmembrane Pod R138Q , enters the Cnx cycle before being directed to ERAD. Finally, we observed by immunofluorescence that neither knockdown of Cnx nor blocking entry into the ERAD pathway with Kif were sufficient to target Pod R138Q to the PM (data not shown). No significant changes on Pod wt levels were observed upon Cnx knockdown, but 4 h of Kif treatment did increase the levels of dg-Pod wt , suggesting that glycosylated Pod wt may also enter the Cnx cycle (Fig. S5B).

Bortezomib partially re-addresses Pod R138Q to the plasma membrane
While studying the proteasomal degradation of Pod R138Q , we observed by immunofluorescence that a fraction of Pod R138Q was localized, after Bz exposure, to thin filopodia PM protrusions, similarly to Pod wt in untreated cells, whereas Pod R138Q was completely absent from these structures before treatment (Fig. 7A). Targeting of Pod R138Q to filopodia, quantified as the percentage of WGA-positive filopodia that were also positive for podocin, revealed that Bz treatment increased plasma membrane targeting of Pod R138Q , as soon as 2 h after Bz treatment, although levels did not reach those of Pod wt (Fig. 7A, graph).

Podocin R138Q quality control and ERAD
We also observed a decreased interaction of Pod R138Q with Cnx in HEK293T cells treated with Bz, which may serve as an indicator of Pod R138Q accumulation outside the ER (Fig. 7B). Because we do not see Pod R138Q at the PM (Fig. 7A) and the glycosylated forms of Pod R138Q seem to be insensitive to Bz (Fig. 3), we hypothesize that only ng-Pod R138Q is reaching the PM upon Bz treatment. Taking into account all our data we propose a scheme depicting a dynamic interchange between the hairpin and the transmembrane topology at the different steps of ER QC and subsequent ERAD (Fig. 7C). Briefly, hairpin Pod R138Q is detected by the cytoplasmic QC machinery and rapidly degraded by the proteasome. In the meantime, N-glycosylated transmembrane Pod R138Q enters the Cnx cycle to be eventually sent for ERAD. Because proteasomal degradation takes place in the cytosol, transmembrane Pod R138Q would lose its N-glycan groups at some point between retrotranslocation and proteasomal degradation (20 -22), possibly adopting a hairpin topology when exposed to the cytosol. Bz treatment would then inhibit the degradation of ng-Pod R138Q , giving it the chance to follow the secretory pathway and reach the PM.

Discussion
Despite p.R138Q being the most common podocin mutation causing SRNS in European and North American children (4), little is known about how this mutation affects podocin stability and degradation. In this study, we have found that human Pod R138Q , when stably overexpressed in human podocyte cell lines, is clearly resolved as a triplet on immunoblots. The upper bands of these triplets correspond to N-glycosylated forms, which is particularly intriguing because it implies that the C terminus of most Pod R138Q is inside the ER lumen. N-Glycosylation, and evidence of transmembrane topology, has been already described for stomatin, for the short isoform of podocin and for podocin P118L , another ER-retained podocin mutant (11,12,23). Whereas Pro 118 , an amino acid that is highly conserved throughout the stomatin family, is located within the hydrophobic intramembrane region and is responsible for the kink of the hairpin topology (12), Arg 138 and Arg 168 , which are also very well conserved, are located within the PHB domain, far away from this region. Nevertheless, we hypothesize that missense mutations in the PHB region may also destabilize podocin hairpin topology, the folding of which is already intrinsically inefficient (11). Indeed, most of the missense mutations affecting the PHB domain typically result in ER-retention (7), and only podocin mutants that are known to be retained in the ER present N-glycosylation levels comparable with Pod R138Q . The switch in topology has important implications for the stability and the degradation of Pod R138Q . We found the glycosylated Pod R138Q forms to be more stable than ng-Pod R138Q . The latter, with a hairpin structure, can be rapidly degraded by the proteasome. In contrast, we found evidence that the transmembrane isoform of Pod R138Q enters the calnexin cycle, which may explain its longer t1 ⁄ 2 . Of course, it remains that glycosylation and deglycosylation events may be affecting the observed differences in stability of the different glycosylated podocin spe-  (2)) or in combination (siCnx (1 ϩ 2)). Luciferase siRNA (siLuc) served as control. Cnx protein amount was quantified to confirm Cnx knockdown (upper left graph). B, effect of kifunensine on podocin protein content. ns, non-stimulated cells. A and B, podocin was quantified as the relative percentage amount of each band within the triplet (color graph) or as total protein (the three bands altogether; lower graph). *, p Ͻ 0.05 and **, p Ͻ 0.01.

Podocin R138Q quality control and ERAD
cies, and undoubtedly these processes are also contributing to our results. That said, based on our experiments with Pod wt(N199Q,N355S) and Pod R138SQ(N199Q,N355S) , which show these proteins behave in the same way as ng-Pod wt and ng-Pod R138Q , we believe that glycosylation of ng-podocin is not contributing to the observed increased stability of the glycosylated species, and in fact the longer t1 ⁄ 2 of mg-and dg-Pod R138Q are due to decreased exposure to the intracellular degradative machinery. According to our results, transmembrane Pod R138Q would be exclusively located in the ER as an intermediary of Pod R138Q QC, and is there stabilized by interaction with Cnx.
Indeed, in contrast to what has been shown for Pod P118L (12), we did not find Pod R138Q to be localized at the PM by immunofluorescence. This same model may apply for misfolded Pod wt , because it also interacts with Cnx and its glycosylated forms are enriched after Kif treatment. Nevertheless, the increased levels of Pod wt following NH 4 Cl addition, together with the lack of response to Bz at short times, suggest that Pod wt is mainly degraded in the endosome/lysosome compartment, a finding in accordance with results from other authors (13).
Cnx recognizes a complex code of glucose and mannose trimming in the N-glycan groups of its substrate proteins (21, . Pod wt co-localization with WGA is included as a positive control. Hoechst nuclei labeling (H) is also shown (blue). Scale bar ϭ 10 m. Image analysis of the WGA co-localization with podocin only at filopodia was obtained through the quantification of regions of interest (ROIs) that carefully delimit cell perimeters. Graph corresponds to the quantification of one representative experiment. One-way analysis of variance followed by Dunnett's post-test was used as statistical analysis. **, p Ͻ 0.01 and ***, p Ͻ 0.001. B, co-immunoprecipitation (IP HA) of HA-podocin and Cnx after Bz addition (16 h at 0.1 M). The graph represents the densitometry quantification of Cnx when podocin is immunoprecipitated. Data are normalized to total immunoprecipitated podocin (HA triplet) and represent to at least three independent experiments. ***, p Ͻ 0.001. C, schematic summarizing the influence of different treatments on Pod R138Q topology and subcellular localization. Red and green arrows indicate a dynamic change to the transmembrane "wrong" or hairpin "right" topology, respectively. Numbers circled in yellow: (1) there is an imbalance of hairpin Pod R138Q toward a transmembrane topology, (2) decreasing Pod R138Q interaction with Cnx through siCnx transfection favors the hairpin topology, (3) stabilization of Pod R138Q interaction with Cnx, through inhibition of Cnx cycle exit with Kif, enhances podocin transmembrane topology, and (4) inhibition of Pod R138Q proteasomal degradation with Bz increases the proportion of hairpin Pod R138Q and is the only treatment that allows partial relocalization to the PM. EC, extracellular matrix; PM, plasma membrane; C, cytosol. Podocin R138Q quality control and ERAD sensitive to mannose trimming implies that most probably there is an interaction with Cnx through its N-glycan groups. This result apparently contradicts the data obtained by the specific inhibition of the N-glycan-dependent interaction of Cnx with Pod R138Q , because the interaction with Cnx was not decreased. Nevertheless, there is an increasing amount of literature suggesting alternative sites of interaction to the lectin domain of Cnx (i.e. the Cnx transmembrane domain), especially in the case of mutant proteins (27)(28)(29). An interesting proposal is that there are at least two different types of interaction occurring sequentially or simultaneously; one carbohydrate based, through the N-glycan groups, and one peptide based, outside the Cnx lectin domain (30 -32). This would explain why inhibiting lectin binding is insufficient to impair Pod R138Q interaction with Cnx. In contrast, Cnx interaction with Pod wt seemed to be more dependent on the N-glycan groups, even though Pod wt is predominantly non-glycosylated. However, we cannot rule out a second Cnx interaction site also for Pod wt , because deletion of N-glycan sites did not completely impair the interaction with Cnx and ng-Pod wt also co-immunoprecipitated with Cnx. Finally, it is tempting to speculate that N-glycosylation and interaction with Cnx may be part of normal podocin biosynthesis and initial folding, and not only a mechanism of podocin quality control. In fact, Cnx is part of the ribosome-translocon complex and its initial association to nascent glycoproteins is usually co-translational (33,34).
The targeting of Pod R138Q to the cell membrane using chemical chaperones as glycerol, trimethylamine-N-oxide, and DMSO has already been reported in transiently transfected human podocytes (35). Nevertheless, there is still no treatment available for SRNS patients carrying the p.R138Q mutation. Here, we report the therapeutic potential of Bz (Velcade), a reversible inhibitor of the 26S proteasome that has been approved and successfully used in multiple myeloma therapy (36,37). Pod R140Q , the mouse equivalent to Pod R138Q , is poorly expressed in the glomeruli of the constitutive Nphs2 R140Q/R140Q knock-in mice (6), similarly to our findings in podocyte cell lines. By studying Pod R138Q t1 ⁄ 2 and the degradation pathways it follows, we have found that low intracellular levels of Pod R138Q are due to rapid proteasomal degradation. Bz not only increases Pod R138Q protein levels, but allows a significant amount of Pod R138Q to reach the PM, at least at filopodia. This could be simply caused by an overflow from the ER, because 2 h of Bz treatment already quadruples Pod R138Q protein levels. Indeed, proteasomal inhibition has been found to inhibit the retrotranslocation of most ERAD substrates, thus giving them the chance to remain in the secretory pathway (21,38,39). Alternatively, because de-glycosylation by cytoplasmic peptide:Nglycanases (PNGases) is an additional step in glycoprotein ERAD (40), Bz would specifically rescue the ng-Pod R138Q already targeted for proteasomal degradation. In support of this idea, a promising effect of Bz is that it enriches the ng-Pod R138Q fraction corresponding to the potentially functional hairpin topology of Pod R138Q at the SD. A successful therapy would need not only to bring Pod R138Q to the PM, but also promote its hairpin topology. Interestingly, Bz has been reported to rescue other misfolded mutant proteins back to the PM in vitro and in vivo (42). Additionally, second generation proteasomal inhibitors are currently being tested in clinical trials with less severe side effects than those of Bz, such as neurotoxicity or lymphopenia (43). Altogether, although we lack in vivo studies, we propose that increasing ng-Pod R138Q protein levels through the inhibition of the proteasomal degradation may be a reasonable strategy to treat patients with p.R138Q mutation and possibly with other ER-retained podocin mutations.

Plasmids, cell culture, and establishment of lentiviral cell lines
Human podocin-coding constructs were generated as described by Tory et al. (44). Human wildtype NPHS1 cDNA, encoding nephrin, was amplified from the construct described by Philippe et al. (6) and subcloned into NotI and SpeI sites of LentiORF pLEX-MCS (Open Biosystems). Thus, the encoded proteins consist of podocin with two hemagglutinin tags (2HA) fused to its N terminus, and nephrin with a V5 tag fused to its N terminus. Site-directed mutagenesis (QuikChange kit, Stratagene) was used to generate the missense mutations used in this study: p.R138Q, p.R168C, p.N199Q, p.N355S for podocin, and p.S366R for nephrin. All constructs were verified by Sanger sequencing. A human immortalized podocyte cell line (AB8/ 13), obtained by transfection of the temperature-sensitive mutant tsA58 of the SV40-T-antigen-encoding gene, was kindly provided by M. A. Saleem (45). Stable podocyte cell lines were obtained by transduction of the above cell line with lentiviral vectors expressing either podocin (wildtype or mutants) or nephrin (wildtype or mutants) at a multiplicity of infection of 1, and subsequently selected by puromycin (2 g/ml). Podocytes were cultured at 33°C with 7% CO 2 in RPMI 1640 medium supplemented with 10% fetal bovine serum, insulin/transferrin/ selenium, glutamine, and penicillin/streptomycin (all from Life Technologies). At this growth-permissive temperature, podocytes are proliferating and undifferentiated, and do not express either endogenous podocin or nephrin. HEK293T cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, glutamine, and penicillin/ streptomycin (all from Life Technologies). Cell lines used in this study were tested mycoplasma-free.

Quantitative real-time PCR
Total RNA was extracted from podocyte cell lines using the Qiagen extraction RNeasy kit (Qiagen) and treated with DNase I. One g of total RNA was reverse-transcribed using SuperScript II according to the manufacturer's protocol (Invitrogen). The relative expression levels of the studied mRNAs were determined by quantitative real-time PCR using Absolute SYBR Green ROX Mix (ABgene) with the following specific primers for human NPHS2: forward (929F) 5Ј-GGCTGA-AGCGCAAAGACAAG-3Ј and reverse (988R) 5Ј-GCAGC-CTTTTCCGCTTCTG-3Ј. Human hypoxanthine phosphoribosyltransferase (Hprt) was used as an internal standard. Data were analyzed with the 2 Ϫ⌬⌬Ct method (47).

Immunofluorescence
Podocytes were cultured on type I collagen-coated coverslips and either transiently transfected using FuGENE HD (Promega) or treated with different drugs. When appropriate, cells were washed once with cold PBS, incubated with Alexa Fluor 555-conjugated WGA (1:600) for 15 min at 4°C, and then fixed with ice-cold ethanol for 5 min. Fixed cells were blocked with 1ϫ PBS, 1% BSA for 30 min before incubation with mouse anti-HA primary antibodies (1:500) or rabbit anti-podocin AP-P35 (1:300) when combined with mouse anti-calnexin (1:500) or mouse anti-CD63 (1:200), followed by Alexa Fluor 488-or 647-conjugated secondary antibodies (1:200). Confocal images were captured using a ϫ40 oil objective attached to a Leica SP8 confocal microscope. At least five random fields, with the only condition that filopodia were well visualized in the WGA channel, were considered for the quantification of the percentage area of WGA colocalization with podocin using ImageJ 1.48i software. Masks were carefully created to specifically quantify the labeling of the filopodia. Confocal settings and Image J thresholds were kept the same.

Immunoblotting
Proteins from podocyte cell lines were extracted in lysis buffer containing 150 mM NaCl, 50 mM Tris-HCl, pH 7, 0.5% Triton X-100 with Complete TM protease inhibitors (Roche Applied Science). Protein dosage was then performed using the BCA protein assay kit (Thermo Scientific). Fifty micrograms of protein were loaded on a 10% SDS-PAGE gel and transferred to a nitrocellulose membrane. Membranes were blocked in 5% skimmed milk in 1ϫ Tris-buffered saline, 0.1% Tween 20 (TBST) for 1 h and incubated with primary antibodies at a 1:1,000 dilution. After washing, membranes were incubated with HRP-conjugated secondary antibodies diluted 1:10,000 in TBST, 5% milk for 1 h at room temperature. Signals were detected using ECL reagents (Amersham Biosciences) and acquired in a Fusion Fx7 darkroom (Vilber Lourmat). Densitometry quantification was performed using Bio-1D software.

Immunoprecipitation
HEK293T cells were transiently transfected with HA-tagged WT podocin and mutants using calcium phosphate. Fortyeight h post-transfection, cells were lysed in 150 mM NaCl, 25 mM Tris-HCl, pH 8, 0.5% Triton X-100 with protease inhibitors, and HA-tagged podocin was immunoprecipitated using the MACS TM Epitope Tag Protein Isolation Kit (Miltenyi Biotec). Briefly, fresh lysates (1-1.5 mg of protein) were incubated either with mouse anti-calnexin antibodies, followed by a 30-min incubation with magnetic beads-coupled to protein A, or directly with magnetic beads coupled to a HA antibody. A rat IgG2a isotype control was included to discard unspecific interactions. Immunoprecipitated proteins were isolated using MACS Separation Columns in a magnetic MACS separator and subsequently eluted with 1ϫ Laemmli buffer. Lysates and immunoprecipitated samples were subjected to immunoblot.

Statistical analyses
All immunoblots were normalized to ␤-actin and then to the corresponding control group or the immunoprecipitated protein in control conditions. Statistical analysis of at least three independent experiments was done using the one-sample twotailed t test (41) or a one-way analysis of variance, p values: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. GraphPad Prism 5 software was used to perform all statistical analyses (mean Ϯ S.D.).