The RING finger- and PDZ domain-containing protein PDZRN3 controls localization of the Mg2+ regulator claudin-16 in renal tube epithelial cells

Ion exchange in the renal tubules is fundamental to the maintenance of physiological ion levels. Claudin-16 (CLDN16) regulates the paracellular reabsorption of Mg2+ in the thick ascending limb of Henle's loop in the kidney, with dephosphorylation of CLDN16 increasing its intracellular distribution and decreasing paracellular Mg2+ permeability. CLDN16 is located in the tight junctions, but the mechanism regulating its localization is unclear. Using yeast two-hybrid systems, we found that CLDN16 binds to PDZRN3, a protein containing both RING-finger and PDZ domains. We also observed that the carboxyl terminus of the cytoplasmic CLDN16 region was required for PDZRN3 binding. PZDRN3 was mainly distributed in the cytosol of rat kidney cells and upon cell treatment with the protein kinase A inhibitor H-89, colocalized with CLDN16. H-89 also increased mono-ubiquitination and the association of CLDN16 with PDZRN3. Mono-ubiquitination levels of a K275A mutant were lower, and its association with PDZRN3 was reduced compared with wild-type (WT) CLDN16 and a K261A mutant, indicating that Lys-275 is the major ubiquitination site. An S217A mutant, a dephosphorylated form of CLDN16, localized to the cytosol along with PDZRN3 and the endosomal marker Rab7. PDZRN3 siRNA increased cell-surface localization of WT CLDN16 in H-89-treated cells or containing the S217A mutant and also suppressed CLDN16 endocytosis. Of note, H-89 decreased paracellular Mg2+ flux in WT CLDN16 cells, and PDZRN3 siRNA increased Mg2+ flux in the H-89-treated WT CLDN16 and S217A mutant cells. These results suggest that PDZRN3 mediates endocytosis of dephosphorylated CLDN16 and represents an important component of the CLDN16-trafficking machinery in the kidney.

Cations and anions filtered in the renal glomeruli are reabsorbed in the renal tubules, which is divided into several segments. Each segment has different properties for reabsorption of ions. Filtered Mg 2ϩ is reabsorbed in the proximal tubule, thick ascending limb (TAL) 2 of Henle's loop, and distal tubule, and the percentages of reabsorption rate were ϳ30%, ϳ60%, and ϳ5%, respectively, and only 5% is excreted into the urine (1,2). In the distal tubule, Mg 2ϩ is reabsorbed via a transcellular pathway. Transient receptor potential melastatin 6 channel is expressed in the apical membrane of the distal tubule and transports Mg 2ϩ from the lumen into the cells (3). Then, Mg 2ϩ may be transported by cyclin M2 from the cells to blood vessel (4). On the other hand, Mg 2ϩ is passively reabsorbed via a paracellular pathway dependent on the magnitude of a lumen-positive transepithelial potential difference in the TAL.
The permeability to ions via paracellular pathway is controlled by tight junctions (TJs). The TJ barrier contains aqueous channels capable of discriminating charge and molecular size, and its electrical resistance varies among different epithelia (5,6). Claudins (CLDNs), a large family of transmembrane proteins, are major components of TJs and bear four transmembrane domains with cytoplasmic amino and carboxyl termini (7,8). The carboxyl terminus of most CLDNs has a PDZ-binding motif that can interact with the PDZ domain of scaffolding proteins including zonula occludens 1 (ZO-1), ZO-2, and ZO-3. ZO-1 and ZO-2 indirectly link CLDNs to the actin cytoskeleton, and these junctional complexes are necessary to maintain the characteristics of permeability. Familial hypomagnesemia with hypercalciuria and nephrocalcinosis is a renal Mg 2ϩ and Ca 2ϩ wasting disorder caused by mutation in the CLDN16 or CLDN19 gene, resulting in impaired renal function and renal failure (9,10). CLDN16 and CLDN19 can form homo-or hetero-oligomeric divalent cation-permeable pore and play a key role in the paracellular reabsorption of Mg 2ϩ in the TJs of TAL (11). Various mutants of CLDN16 dissociate from the TJs and are distributed in the Golgi apparatus, endoplasmic reticulum, or lysosome (12)(13)(14)(15). We previously reported that CLDN16 is dephosphorylated in Dahl salt-sensitive hypertensive rats (16) and the dephosphorylated mutant of CLDN16 is mainly distributed in the lysosome, resulting in little permeability to Mg 2ϩ (17). The mistargeting of CLDN16 must be a cause of hypomagnesemia; however, the regulatory mechanisms for trafficking of CLDN16 have not yet been clarified in detail. PDZ domains are structural motifs that bind to a consensus motif at the carboxyl terminus of target proteins. A member of PDZ domain containing RING finger (PDZRN) includes a RING finger domain at an amino terminus and two or four PDZ domains at the carboxyl terminus. PDZRN3 has been reported to play positive roles in the myoblast differentiation (18) and nephrogenesis in Xenopus laevis (19). In contrast, it has negative roles in the bone morphogenetic protein-2-induced osteoblast differentiation (20) and adipogenesis in mouse 3T3-L1 preadipocytes (21). However, the functional characterization and binding target proteins of PDZRN3 have not been clarified in the mammalian kidney.
In the present study we investigated the novel association protein, which can regulate trafficking of CLDN16. In yeast two-hybrid systems, PDZRN3 was identified to bind to CLDN16. Immunofluorescence and immunoprecipitation assays showed that PDZRN3 is expressed in the CLDN16-expressing cells and binds to CLDN16 in the rat renal tubule. N- ethyl]-5-isoquinolinesulfonamide (H-89), a protein kinase A (PKA) inhibitor, increased the association of CLND16 with PDZRN3 in Madin-Darby canine kidney (MDCK) cells expressing FLAG-tagged CLDN16. PDZRN3 siRNA suppressed endocytosis and increased the cellsurface expression of CLDN16 in the H-89-treated cells. Furthermore, PDZRN3 siRNA increased paracellular Mg 2ϩ flux from the apical to basal compartments. Similar results were observed in the cells expressing S217A mutant, a dephosphorylated form of CLDN16. Our present results indicate that PDZRN3 may be involved in the endocytosis of dephosphorylated CLDN16 from the TJs to intracellular compartments.

Association of CLDN16 with PDZRN3 in immunoprecipitation and pulldown assays
To identify novel CLDN16-binding proteins involved in the regulation of its trafficking, we performed yeast two-hybrid screening using a human kidney cDNA library with the carboxyl cytoplasmic region of CLDN16 as bait. We obtained 234 positive clones and determined the sequences as described previously (22). Among them, we chose PDZRN3 (3 clones) as a candidate protein involved in regulating the trafficking of CLDN16 in the present studies. CLDN16 was previously shown to be exclusively expressed in the TAL of Henle's loop in humans (9), mice (23), and bovines (24). In the rat kidney, CLDN16 was localized in the cells expressing Tamm-Horsfall glycoprotein (THP), a marker of the TAL (Fig. 1A). PDZRN3 was widely expressed in the renal tubules including CLDN16-positive cells. These results indicate that PDZRN3 may be associated with CLDN16 in the kidney. To clarify the association between CLDN16 and PDZRN3, we performed immunoprecipitation assay using the homogenates from rat renal cortex. PDZRN3 was immunoprecipitated with anti-CLDN16 antibody but not with rabbit IgG (Fig. 1B). To clarify the binding site of CLDN16 with PDZRN3, we made a glutathione S-transferase (GST) fusion protein to the carboxyl cytoplasmic region of wild type (WT) and ⌬PDZ mutant, a deletion form of the PDZ-binding motif (TRV at the carboxyl terminus) of CLDN16. In the pulldown assay, the WT CLDN16 was associated with PDZRN3, but ⌬PDZ mutant was not (Fig. 1C). These results indicate that CLDN16 binds to PDZRN3 through the PDZ-binding motif.

Effect of CLDN16 expression on the endogenous expression of junctional proteins
We previously established stable cells expressing FLAGtagged CLDN16 using MDCK/Tet-off cell line (25). The induction of CLDN16 by removal of doxycycline was confirmed by immunoblotting using anti-FLAG antibody ( Fig. 2A). The expression levels of endogenous junctional protein including ZO-1, PDZRN3, E-cadherin, and occludin were not changed by CLDN16 expression. CLDN16 was not associated with PDZRN3 under control conditions, but H-89, a PKA inhibitor, increased the association between CLDN16 and PDZRN3 (Fig.  2B). The expression levels of PDZRN3 were not changed by the expression of mock, WT, and ⌬PDZ mutant CLDN16 in the presence of H-89 (Fig. 2C). WT CLDN16 was associated with PDZRN3 in the presence of H-89, which was inhibited by the deletion of the PDZ-binding motif of CLDN16. These results indicate that the PDZ-binding motif within the carboxyl cytoplasmic region of CLDN16 is involved in the interaction with PDZRN3, consistent with the data of the pulldown assay.

Effects of H-89 and PDZRN3 siRNA on intracellular distribution of CLDN16
Immunofluorescence assay showed that CLDN16 is colocalized with ZO-1 in the TJs, whereas it is not colocalized with Rab7, a late endosome marker, in the cytosol under control conditions (Fig. 3A). PDZRN3 was mainly distributed in the cytosol and slightly in the cell-cell contact area. H-89 increased the cytosolic distribution of CLDN16 and the colocalization of CLDN16 with PDZRN3 or Rab7 (Fig. 3, B and F). The colocalization of early endosomal antigen 1 (EEA1), an early endosome marker, with CLDN16 was slightly increased by H-89, whereas that with PDZRN3 was not (Fig. 3, C and D). The colocalization of CLDN16 with Rab7 caused by H-89 was inhibited by the introduction of PDZRN3 siRNA (Fig. 3, E and F). These results indicate that PDZRN3 may be involved in the regulation of intracellular trafficking of dephosphorylated CLDN16.

Effects of H-89 and PDZRN3 siRNA on ubiquitination and cellsurface localization of CLDN16
Western blotting showed that the bands of FLAG-tagged CLDN16 are detected around 28 kDa in the short-exposure point (Fig. 4). In the long-exposure point, another band was detected around 36 kDa. The total levels of PDZRN3 and upper bands of CLDN16 were decreased by the introduction of PDZRN3 siRNA, whereas those of CLDN1 and the lower bands of CLDN16 were not. The cell-surface biotinylation assay has been used to examine the cell-surface localization of CLDN1, CLDN2, and CLDN4 in MDCK cells (26 -28). The level of cellsurface localization of CLDN16 in the H-89-treated cells was significantly lower than that in the control cells. PDZRN3 siRNA inhibited the H-89-induced reduction of cell-surface localization of CLDN16. In contrast, the cell-surface localization of CLDN1 was not changed by H-89 and PDZRN3 siRNA. Next, we examined the effect of PDZRN3 on ubiquitination of CLDN16. H-89 increased mono-ubiquitination level of

PDZRN3 regulates endocytosis of claudin-16
CLDN16 and the association of CLDN16 with PDZRN3, which were inhibited by the introduction of PDZRN3 siRNA. The bands of CLDN16 were detected around 36 kDa, which is similar to the size of ubiquitin-conjugated CLDN16. In contrast, CLDN1 was not associated with PDZRN3 in the control, H-89treated, and PDZRN3 siRNA-transfected cells.

Ubiquitination of Lys-275 of CLDN16 by H-89
Ubiquitination sites are predicted at Lys-261 and Lys-275 of human CLDN16 by the bioinformatics tool of CKSAAP_Ub-Site. Alanine substitution assay showed that the high molecular weight band of CLDN16 is decreased, and the low molecular weight band is increased by the mutation at K275A in the presence of H-89 (Fig. 5A). Total levels of PDZRN3 and CLDN16 were not changed by the mutation at K261A and K275A. The ubiquitination and association of PDZRN3 with CLDN16 were significantly inhibited by the mutation at Lys-275 but not at Lys-261 (Fig. 5B). These results indicate that Lys-275 of CLDN16 may be a major ubiquitination site.

Effect of PDZRN3 siRNA on endocytosis and stability of CLDN16 protein
The association between CLDN16 and PDZRN3 may increase the trafficking of CLDN16 from the TJs to subcellular compartments. Therefore, we examined the effect of PDZRN3 siRNA on the endocytosis of CLDN16. CLDN16 was time-dependently internalized in the presence of H-89, which was suppressed by the knockdown of PDZRN3 (Fig. 6A). These results indicate that PDZRN3 may be involved in the endocytosis of dephosphorylated CLDN16 from the TJs to subcellular com- partments. Next, we examined the effect of PDZRN3 on the protein stability of CLDN16. The levels of WT CLDN16 were time-dependently decreased in the presence of cycloheximide, an inhibitor of protein synthesis, and H-89 (Fig. 6B). The rate of decrease was significantly inhibited by the introduction of PDZRN3 siRNA. These results indicate that PDZRN3 has an effect on the protein stability of CLDN16 in the presence of H-89.

Effect of PDZRN3 siRNA on the intracellular localization and stability of S217A mutant CLDN16
We previously reported that the tight junctional localization of CLDN16 is regulated by its phosphorylation at Ser-217 (17). S217A mutant, a dephosphorylated form of CLDN16, was mainly distributed in the cytosol and colocalized with PDZRN3 or Rab7 (Fig. 7, A and C). The association levels of CLDN16 with PDZRN3 and mono-ubiquitination of S217A mutant were higher than those of WT. In contrast, the level of cell-surface localization of S217A mutant was lower than that of WT. These results are similar to those in H-89-treated cells (Figs. 3 and 4). PDZRN3 siRNA increased the tight junctional localization of S217A mutant and colocalization between CLDN16 and ZO-1, whereas it decreased the co-localization between CLDN16 and Rab7 (Fig. 7, B and C). PDZRN3 siRNA increased the cell-sur-face localization of CLDN16, whereas it decreased mono-ubiquitination and the association with CLDN16. These results indicate that PDZRN3 is involved in the cytosolic retention of dephosphorylated CLDN16. Next, we examined the effect of PDZRN3 siRNA on the protein stability of S217A mutant CLDN16. The levels of S217A mutant CLDN16 were time-dependently decreased in the presence of cycloheximide, which were significantly inhibited by the introduction of PDZRN3 siRNA (Fig. 7D). These results are similar to those in the WT CLDN16expressing cells treated with H-89.

Effect of PDZRN3 siRNA on paracellular permeability
Induction of WT CLDN16 expression by removing doxycycline caused increases in transepithelial electrical resistance (TER) and paracellular Mg 2ϩ flux (Fig. 8, A and B). H-89 significantly inhibited the CLDN16-induced increase in TER and Mg 2ϩ flux, which was rescued by PDZRN3 siRNA. Similarly, TER and Mg 2ϩ flux were increased by PDZRN3 siRNA in the cells expressing S217A mutant CLDN16 (Fig. 8, C and D).
These results indicate that dephosphorylation causes a loss of function of CLDN16 mediated by the mislocalization, which is rescued by PDZRN3 siRNA.

Discussion
A dysfunction of CLDN16 or 19 is involved in the pathogenesis of hypomagnesemia. More than 40 and 10 mutations, respectively, have been currently identified in CLDN16 and CLDN19 genes (29). Mutations in CLDN16 genes affect Ͼ28 different amino acids, which mainly occur in the extracellular loop and transmembrane domains. The mutants of CDLN16 distributed in the TJs have full or partial function after being transfected into LLC-PK 1 (15) and MDCK-C7 cells (12). In contrast, the mutants mislocalized to the intracellular compartments including the endoplasmic reticulum, Golgi apparatus, and lysosome lose their function. Therefore, the subcellular trafficking step of CLDN16 from intracellular compartments to the TJs plays an important role in the controlling Mg 2ϩ reabsorption under physiological conditions. The intracellular localization of several CLDNs is regulated by phosphorylation. Phosphorylated CLDN1 and -2 are localized at the TJs in MDCK II cells (30,31). CLDN4 phosphorylated by atypical protein kinase C (PKC) is localized at the TJs in human keratinocyte cells (32). Phosphorylation of CLDNs may be also involved in the pathogenesis. The phosphorylation status of CLDN4 -7 was individually changed during the course of colitis caused by lipopolysaccharide in T84 colonic cells (33). Mislocalization of CLDN1 in the nuclei is caused by the PKCdependent phosphorylation and increases metastasis in osteosarcoma cells (34). In contrast, nuclear localization of CLDN2 is increased by dephosphorylation and enhances proliferation in lung adenocarcinoma cells (35). Phosphorylation levels of CLDN16 in Dahl salt-sensitive hypertensive rats is lower than that in normotensive rats, and dephosphorylated CLDN16 caused by H-89 treatment is mainly distributed in the cytosol (16,25). Furthermore, the elevation of extracellular Mg 2ϩ concentration increased the intracellular localization of CLDN16 mediated by the activation of the MEK/ERK pathway. The reg-

PDZRN3 regulates endocytosis of claudin-16
ulatory factors and the effect of phosphorylation on the localization of CLDNs may be different in each CLDN and tissue.
We recently reported that syntaxin-8, a target soluble Nethylmaleimide-sensitive factor attachment protein receptor, binds to CLDN16 (22). The association of CLDN16 with syntaxin-8 and the cell-surface localization of CLDN16 are decreased by H-89 treatment. Furthermore, the association of S217A-mutant CLDN16 with syntaxin-8 is lower than that of WT, and the mutant was mainly localized in the cytosol. Therefore, we suggested that trafficking of CLDN16 from the intracellular compartments to the TJs is up-regulated by syntaxin-8. Here, we found that PDZRN3 binds to CLDN16 and the association of CLDN16 with PDZRN3 is increased by H-89 treatment. The endocytosis of CLDN16 was suppressed by PDZRN3 siRNA (Fig. 6A). These data indicate that the tight junctional localization of CLDN16 is inversely regulated by syntaxin-8 and PDZRN3. The late endosome marker Rab7 was colocalized with CLDN16 in the cells treated with H-89 or expressing S217A mutant. Therefore, we suggest that the endocytosis of dephosphorylated CLDN16 from the TJs to endosomes is upregulated by PDZRN3. This is the first report showing that the intracellular trafficking of CLDNs is regulated by PDZRN3. Urinary cAMP excretion and the renal adenylate cyclase response to parathyroid hormone are lower in Dahl-saltsensitive hypertensive rats than in Dahl-salt-resistant normotensive rats (36). Urinary Mg 2ϩ excretion is increased in hypertensive rats (37). We suggest that the reduction of cAMP content induces dephosphorylation of CLDN16 and PDZRN3-dependent endocytosis, resulting in the decrease in Mg 2ϩ reabsorption.
Four types of PDZRN (PDZRN1-4) have been cloned (38,39), but the function of PDZRN3 and -4 is largely unknown. PDZRN2 (also known as ligand-of-Numb protein X1, LNX1) is a multidomain protein with E3 ubiquitin (Ub) ligase activity (40). Protein ubiquitination is mediated by three enzymes, Ubactivating (E1), Ub-conjugated (E2), and Ub-ligating (E3). E3 Ub ligases directly bind substrate and render substrate specificity. There are Ͼ600 putative E3 ligases with respect to individual protein substrates. Single or poly-Ub chain is covalently attached to lysine residues of target proteins. The Ub-interacting motifs are compiled into databases such as UbiProt, CKSAAP, and SysPTM. Using the CKSAAP database, we predicted the ubiquitination site of CLDN16 and found that Lys-275 is ubiquitinated by PDZRN3. PDZRN2 binds to junctional adhesion molecule (JAM4), a tight junctional one transmembrane protein, and regulates endocytosis of JAM4 (41). LNX1p80 binds to CLDN1 and induces endocytosis of CLDN1 in MDCK cells (42). An E3 Ub ligase complex Kelchlike 3 and Cullin 3 binds to CLDN8, resulting in ubiquitination and degradation of CLDN8 (43). Our data indicate that PDZRN3 binds to CLDN16 but not to CLDN1 (Fig. 4). We suggest that each CLDN can bind to different E3 Ub ligases. The E3 Ub ligase activity of PDZRN3 has not been investigated. In addition, PDZRN3 contains two PDZ domain, but it is unknown which domain is needed for the association with CLDN16. ZO-1 and ZO-2, which include three PDZ domains, are expressed in the TJ of TAL (44). We need further study to clarify the regulatory mechanism of tight junctional localization of CLDN16 by PDZ domain-containing proteins in the kidney. The interaction between PDZ protein and target proteins is often regulated by phosphorylation of PDZ-binding motif. Phosphorylation of PDZ-binding motif down-regulates the interaction of the NR2B subunit of NMDA receptor with postsynaptic density 95 (PSD95) (45), membrane-associated guanylate kinase with corticotropin-releasing hormone receptor type 1 (46), GluR2 with GRIP1 (47), and Kir2.3 potassium channel with PSD95 (48). Canonical PDZ domains are composed of six ␤-strands (␤A-␤F) and two ␣-helixes (␣A and ␣B) (49). The Ser or Thr residue in the PDZ ligands may form hydrogen bonds with the side chain of the His residue at ␣B-1 or ␣B-5, respectively. Therefore, the phosphorylation of Ser or Thr residue may interfere with the interaction between these proteins (50). In contrast, phosphorylation of the PDZ-binding motif up-regulates the interaction of the myotilin and FATZ (calsarcin/myozenin) family proteins with ZASP/Cypher (51) and Yes-associated protein with Partitioning-defective 3 (52). Furthermore, dephosphorylation of PDZ-binding motif up-regulates the interaction of syndecan-1 with syntenin-1 (53). These reports indicate that the phosphorylation/dephosphorylation of PDZ-binding motif is involved in the interaction with target proteins. In contrast, there is no report showing that the PDZ-binding motif of CLDNs is phosphorylated. Interestingly, our data indicate that PDZRN3 binds to the PDZ-binding motif of CLDN16, and the interaction is up-regulated by the dephosphorylation of Ser-217, which is located at a different site from the PDZ-binding motif. The interaction between PDZRN3 and CLDN16 may be regulated by both the PDZ domain and the spatial structure near the phosphorylation site (Ser-217) of CLDN16.
Taken together, the results of the present study demonstrated that PDZRN3 directly binds to CLDN16 and is involved in the trafficking pathway of CLDN16 in renal tubular epithelial cells. This pathway was found to be sensitive to phosphorylation of CLDN16 at Ser-217 and mono-ubiquitination at Lys-275. Dephosphorylation of CLDN16 induced abnormal cyto- The cell-surface-biotinylated proteins were precipitated using streptavidin-beads and immunoblotted with anti-FLAG or CLDN1 antibody. The cell lysates were immunoprecipitation with anti-PDZRN3 antibody, and the immune pellets were immunoblotted with anti-FLAG antibody. n ϭ 3-4. C, the co-localization of CLDN16 (FLAG) with ZO-1, PDZRN3, or Rab7 is shown as the percentage (n ϭ 12-16 cells from Ͼ3 separate experiments). D, the cells expressing S217A mutant CLDN16 were transfected with negative control or PDZRN3 siRNA. 48 h after transfection, the cells were incubated with 5 M cycloheximide (CHX) for the indicated periods. The cell lysates were immunoblotted with anti-FLAG antibody. The band density of FLAG is represented as relative to the values at 0 h. n ϭ 3. **, p Ͻ 0.01; *, p Ͻ 0.05.

PDZRN3 regulates endocytosis of claudin-16
plasmic localization and loss of Mg 2ϩ transport, which were rescued by knockdown of PDZRN3. The physiological reabsorption of Mg 2ϩ in the TAL may be dynamically associated with the syntaxin 8-and PDZRN3-regulated expression of CLDN16 in the TJs.

Materials
Rabbit anti-ZO-1 antibody was obtained from Zymed Laboratories Inc. (South San Francisco, CA). Rabbit anti-FLAG antibody was from Medical & Biological Laboratories Co. (Nagoya, Japan). Mouse anti-FLAG antibody and sodium 2-mercaptoethanesulfonate (MESNA) were from Wako Pure Chemical Industries (Osaka, Japan). Rabbit anti-THP antibody was from Santa Cruz Biotechnology (Dallas, TX). Rabbit anti-Rab7 antibody was from Cell Signaling Technology (Beverly, MA). Rabbit anti-EEA1 antibody was from Affinity BioReagents (Golden, CO). H-89 was from LKT laboratories (St. Paul, MN). Lipofectamine 2000 was from Invitrogen. All other reagents were of the highest grade of purity available.

Yeast two-hybrid screening
The interacting protein of carboxyl cytoplasmic region of CLDN16 was screened using the Matchmaker Gold Yeast Twohybrid System (BD Biosciences Clontech, Mountain View, CA) as described previously (22).

Pulldown assay
The carboxyl cytoplasmic region of human CLDN16 was subcloned into the pGEX4T1 vector (GE Healthcare). The vec-tor was introduced into Escherichia coli BL21 and grown in the overnight express. GST-fused proteins were purified with glutathione-Sepharose 4B beads. The beads were incubated in buffer composed of 10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.1% Nonidet P-40, 2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and protease inhibitor mixture for 12 h at 4°C. Bound proteins were then eluted with a sample buffer (2% SDS, 10% glycerol, 5% 2-mercaptoethanol, 0.2% bromphenol blue, and 0.2 mM Tris-HCl, pH6.8) and applied to the sodium dodecylsulfate (SDS)-polyacrylamide gel. Proteins were blotted onto a PVDF membrane and incubated with each primary antibody followed by a peroxidase-conjugated secondary antibody. The blots were visualized as described in the immunoblotting section.

Plasmid DNA construction
Human CLDN16 cDNA was amplified by reverse transcriptase-polymerase chain reaction using a set of forward and reverse primers containing custom EcoRV and SalI restriction sites, respectively ( Table 1). The cDNA was cloned into the mammalian expression vector, pCMV-Tag, containing the FLAG epitope, and subcloned into pTRE2hyg vector. The mutants of CLDN16 (K261A and K275A) were generated using a KOD-plus-mutagenesis kit (Toyobo Life Science, Osaka, Japan). Sequencing was performed by Bio Matrix Research (Chiba, Japan).

Cell culture and transfection
The MDCK Tet-OFF cell line was obtained from BD Biosciences Clontech. Cells expressing FLAG-tagged WT and S217A mutant CLDN16 were generated previously (25). Cells were grown in Dulbecco's modified Eagle's medium (Sigma) supplemented with 5% fetal bovine serum, 0.07 mg/ml penicillin-G potassium, 0.14 mg/ml streptomycin sulfate, 0.1 mg/ml G418, and 0.1 mg/ml hygromycin B in a 5% CO 2 atmosphere at 37°C. Plasmid DNA, PDZRN3 siRNA, and negative control siRNA (Sigma) were transfected into cells using Lipofectamine 2000 as recommended by the manufacturer.

Preparation of renal homogenates, cell lysates, and immunoprecipitation
The homogenates of the rat renal cortex were prepared as described previously (16). Confluent MDCK cells were scraped into cold PBS and precipitated by centrifugation. The cells were then lysed in radioimmune precipitation assay buffer containing 150 mM NaCl, 0.5 mM EDTA, 1% Triton X-100, 50 mM Tris-HCl (pH 8.0), protease inhibitor mixture (Sigma), and 1 mM phenylmethylsulfonyl fluoride and were then sonicated for 20 s. After centrifugation at 1000 ϫ g for 5 min, the supernatant

PDZRN3 regulates endocytosis of claudin-16
was collected (cell lysates). In the immunoprecipitation assay, cell lysates were incubated with protein G-Sepharose beads and anti-FLAG antibody at 4°C for 16 h with gentle rocking. After centrifugation at 6000 ϫ g for 1 min, the pellet was washed 3 times with the radioimmune precipitation assay buffer. In the biotinylation assay, cell-surface proteins were biotinylated as described previously (26). In the endocytosis assay, cell-surface proteins were biotinylated as described previously (26). After incubation of the cells for the period indicated, MESNA was used to cleave biotin from all cell-surface protein. The cell lysates, immunoprecipitants, and biotinylated proteins were solubilized in a sample buffer for SDS-polyacrylamide gel electrophoresis. Protein concentrations were measured by a protein assay kit (Bio-Rad) in which bovine serum albumin was used as a standard.

SDS-polyacrylamide gel electrophoresis and immunoblotting
SDS-polyacrylamide gel electrophoresis was performed as described previously (54). Briefly, cell lysates or immunoprecipitates were applied to the SDS-polyacrylamide gel. Proteins were blotted onto a PVDF membrane and incubated with each primary antibody followed by a peroxidase-conjugated secondary antibody. Finally, the blots were incubated in EzWestLumi plus (ATTO Corp., Tokyo, Japan) and scanned with a C-DiGit Blot Scanner (LI-COR Biotechnology, Lincoln, NE). Band density was quantified with ImageJ software (National Institute of Health software). ␤-Actin was used for normalization.

Measurement of TER and paracellular permeability
MDCK cells expressing FLAG-tagged CLDN16 were plated at confluent densities on transwells with polyester membrane inserts (Corning Inc.-Life Sciences, Acton, MA). TER was measured using Millicell-ERS volt-ohmmeter (Millipore, Bedford, MA). TER values (ohm⅐cm 2 ) were normalized by the area of the monolayer and calculated by subtracting the blank values (ϳ35 ohm⅐cm 2 ) from the filter and bathing medium. Paracellular permeabilities to Mg 2ϩ were measured as described previously (55).

Confocal microscopy
Rat kidney slices and MDCK cells expressing FLAG-tagged CLDN16 were immunostained as described previously (22). Immunolabeled cells were visualized on an LSM 700 confocal microscope (Carl Zeiss, Jena, Germany) set with a filter appropriate for DAPI, Alexa Fluor 488, and Alexa Fluor 546. Fluorescence intensities of CLDN16, Rab7, and PDZRN3 were determined by measuring the mean pixel density of staining area using ImageJ software (National Institute of Health, Bethesda, MD). The image area containing the signal from the TJ and cytosol was manually marked using ImageJ. After subtraction of background, the merged intensities of CLDN16 with PDZRN3 or Rab7 were shown as percentage of total intensities of CLDN16.

Statistics
Results are presented as the mean Ϯ S.E. Differences between groups were analyzed using one-way analysis of variance, and corrections for multiple comparisons were made using Tukey's multiple comparison test. Comparisons between two groups were made using Student's t test. Significant differences were assumed at p Ͻ 0.05.
Author contributions-K. M., C. F., and N. F. performed the experiments and analyzed the data. T. K., T. F., T. M., S. E., H. H., N. A., M. Y., and Y. Y. contributed to the experiment plan and discussion of the manuscript. A. I. contributed to supervision of the project, interpretation of the data, and writing the paper.