The Tight Junction-specific Protein Occludin Is a Functional Target of the E3 Ubiquitin-protein Ligase Itch*

Tight junctions create a highly selective diffusion barrier between epithelial and endothelial cells by preventing the free passage of molecules and ions across the paracellular pathway. Although the regulation of this barrier is still enigmatic, there is evidence that junctional transmembrane proteins are critically involved. Recent evidence confirms the notion thatoccludin, a four-pass integral plasma-membrane protein, is a functional component of the paracellular barrier. The overall hydrophilicity of occludin predicts two extracellular loops bounded by NH2- and COOH-terminal cytoplasmic domains. To date, the binding of the COOH terminus of occludin to intracellular proteins is well documented, but information concerning the function of the cytoplasmic NH2 terminus is still lacking. Using yeast two-hybrid screening we have identified a novel interaction between occludin and the E3 ubiquitin-protein ligase Itch, a member of the HECT domain-containing ubiquitin-protein ligases. We have found that the NH2-terminal portion of occludin binds specifically to a multidomain of Itch, consisting of four WW motifs. This interaction has been confirmed by our results from in vivoand in vitro co-immunoprecipitation experiments. In addition, we provide evidence that Itch is specifically involved in the ubiquitination of occludin in vivo, and that the degradation of occludin is sensitive to proteasome inhibition.

Tight junctions (TJs) 1 consist of different types of transmembrane proteins which interact directly or indirectly with junctional plaque proteins located at the cytoplasmic surface of TJs (1)(2)(3)(4)(5). These peripheral plaque proteins in turn link TJs to the cytoskeleton (6 -8). Characteristically, TJs are located at the interphase between the apical and basolateral membranes where they act as a "molecular fence" to prevent intermixing of apical and basolateral membrane components. In addition, TJs function as semipermeable barriers restricting the paracellular diffusion of solutes and molecules between epithelial and endothelial cells.
It is now known that TJs do not form an absolute seal, but instead contain discrete pores to permit a selective paracellular diffusion. Although it is still unclear how the restrictive or permissive action of the junctional pores is modulated, there is a general consensus that the transmembrane components of the TJ are critically involved in this process.
Occludin is a 65-kDa integral plasma-membrane protein located specifically at tight junctions (9,10). According to its overall hydrophilicity, occludin appears to span the plasma membrane four times, forming two extracellular loops and exposing its NH 2 and COOH terminus to the cytosol. Interaction of occludin with several cytoplasmic proteins of the junctional plaque has been found to occur via its COOH terminus (10), while the extracellular loops are supposed to be involved in the regulation of paracellular permeability and cell adhesion (11)(12)(13). It came as a surprise that occludin-deficient mouse embryos were viable and, moreover, did not show any gross morphological alterations of TJs (14). Similarly, TJs of embryoid bodies originating from occludin-deficient ES cells were morphologically indistinguishable from their wild-type counterparts (15). These findings not only led to the discovery of an additional family of occludin-unrelated TJ-specific transmembrane proteins (16), but also to the notion that occludin is obviously not an indispensable structural component of TJs.
Instead it seems to fulfill a complex modulatory action on epithelial barrier properties, the molecular basis of which is still elusive.
While the COOH-terminal attachment of occludin to intracellular proteins is well documented, information concerning protein-protein interaction at the cytoplasmic NH 2 terminus is still lacking. Observations reported recently indicated that expression of an NH 2 -terminal modified occludin led to the upregulation of neutrophil migration across stably transfected epithelial monolayers (17). This first clue toward a functional significance of the NH 2 -terminal domain of occludin initiated the quest for potential proteins binding to this region.
Using yeast two-hybrid screening we have identified a novel interaction between the NH 2 terminus of occludin and the WW domains of the E3 ubiquitin-protein ligase Itch, a protein required for the ubiquitin-dependent endocytosis of plasmamembrane proteins (18). The specificity of this interaction has been verified by in vitro and in vivo co-immunoprecipitation experiments. We have further shown that occludin is specifically ubiquitinated in vivo and in vitro and its turnover is sensitive to inhibition of the proteasome pathway.

EXPERIMENTAL PROCEDURES
Cell Culture, DNA Transfection, Immunofluorescence-LLC-PK1 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. HEK 293 cells were kept in MEM with 10% horse serum. MDCK cells were cultured in MEM enriched with 10% fetal calf serum. For transfection experiments cells were grown to 70 -80% confluence and transfected with Effectene TM transfection reagent (Qiagen) according to the manufacturer's protocol.
For immunofluorescence, cells were grown on coverslips, washed in PBSϪ, and fixed with ethanol/acetic acid (95:5) for 20 min at Ϫ20°C. Immunolabeling was done according to standard procedures. Fluorescent images were taken using a Zeiss Axioplan TM microscope and the Zeiss AxioVision Imaging System TM .
For cloning of pCR TM 2.1-TOPO TM -Itch a full-length cDNA encoding mouse Itch (GenBank TM accession number NM_008395) was amplified from a cDNA template derived from mouse liver RNA. Primers for the PCR reaction were: 5Ј-cgtcgacccatgggtagtctgaccatga-3Ј and 5Ј-gtggatccactcttgtccaaatccttcagtttc-3 Ј, flanked by SalI and BamHI sites. The amplified product was cloned into the pCR TM 2.1-TOPO TM vector as recommended by the manufacturer (Invitrogen).
Yeast Two-hybrid Screening and ␤-Galactosidase Assay-The MATCHMAKER GAL4 Two-Hybrid System 3 (CLONTECH) was used to identify new interaction partners for the NH 2 terminus of occludin. First, the Saccharomyces cerevisiae strain AH109 was transformed with pGBKT7-N-Occl using a small scale lithium acetate/singlestranded carrier DNA/polyethylene glycol (LiAc/ss-DNA/PEG) transformation protocol (20). Selected transformants were transformed with a mouse liver MATCHMAKER cDNA library (in AD-vector pGAD10; CLONTECH) using a large scale LiAc/ss-DNA/PEG transformation protocol. Positive transformants were selected on SD/ϪAde/ϪHis/ ϪLeu/ϪTrp ϩ 5 mM 3-aminotriazole selection media. All yeast clones were restreaked onto SD/ϪTrp/ϪLeu media and assayed for activity of ␤-galactosidase by a ␤-galactosidase plate assay (21). DNA was isolated from all lacZ positive clones and the AD-library plasmids were rescued in Escherichia coli strain DH5␣. For further analysis, every isolated library plasmid was retransformed into the S. cerevisiae strain Y190 and additionally co-transformed with the following DNA-BD plasmids:  (1); both clones interacted with the NH 2 terminus of occludin in yeast strain Y190 (2) but no interaction with the DNA-BD expressed by pGBKT7 was found (3). Furthermore, WW Itch and WW Itch-Hect did not interact with human lamin C, an unrelated protein (4). C, additional yeast two-hybrid analysis for WW Itch. A frameshift mutation of WW Itch (position indicated by arrow) did not interact with N-Occl (upper row), but ␤-galactosidase activity was observed when N-Occl was shifted from the DNA-BD vector to the AD vector and vice versa for WW Itch (lower row). D, domain mapping for occludin-Itch interaction. Schematic diagrams of the various constructs used in the yeast two-hybrid assay are shown. ␤-Galactosidase activity was only detected for transformants expressing the NH 2 -terminal domain of occludin and the 2 nd or 4 th WW domain of Itch. The 1 st and 3 rd WW domain were not capable of interacting with N-Occl. E, mutation of the putative WW recognition motif from Pro 9 -Pro 10 -Tyr 11 -Pro 12 to Ala 9 -Ala 10 -Tyr 11 -Pro 12 within the NH 2 terminus of occludin (N-Occl-mut; upper row) leads to a disruption of the interaction with WW Itch. Furthermore, co-transformation of WW Itch with a NH 2 -terminal deleted occludin (⌬N-Occludin) abolishes ␤-galactosidase activity (lower row).
pGBKT7-N-Occl, empty pGBKT7, and pGBKT7-LAM (CLONTECH; encodes human lamin C and provides a control for a fortuitous interaction). Transformation reactions were plated onto appropriate minimal synthetic dropout media, followed by a ␤-galactosidase plate assay. Furthermore, the library insert was transferred from the pGAD10-AD vector into the BamHI site of the pGBKT7-BD vector, and N-Occl (bait) was subcloned from the pGBKT7-BD vector into the NdeI/XhoI sites of the pGADT7-AD vector, followed by a two-hybrid assay in yeast strain Y190.
Additionally, a frameshift mutation just upstream of the cDNA insert in the activation domain vector pGAD10 was created by digesting the plasmid with MluI, filling in the overhangs, and then religating the library plasmid. The frameshifted prey was then co-transformed with pGBKT7-N-Occl into yeast strain Y190 and transformants were assayed for ␤-galactosidase activity as described above.
For domain mapping, the appropriate expression constructs (see constructs) were co-transformed into the yeast strain Y190 using a small scale LiAc/ss-DNA/PEG transformation protocol. Transformants were plated onto appropriate minimal synthetic dropout media and tested for ␤-galactosidase activity.
Antibodies-A mouse monoclonal anti-c-Myc antibody was pur-chased from Santa Cruz Biotechnology Inc. Rabbit anti-occludin and anti-claudin-1 antibody was from Zymed Laboratories Inc. Alexa Fluor TM 568 goat anti-rabbit IgG was used from Molecular Probes. A rabbit anti-Itch polyclonal antibody was generated as previously described (19). Monoclonal anti-HA was kindly provided by M. Gimona (Institute of Molecular Biology, Salzburg, Austria).
In Vitro Binding Assay-WW Itch was subcloned from pGAD10 into the BamHI site of the AD-vector pGADT7. Using the TNT TM T7 Coupled Reticulocyte Lysate System (Promega), WW Itch and the unmodified and modified versions of occludin (wt-Occl, Occl-⌬N, Occl-⌬N1/2, and Occl-mut) were in vitro transcribed/translated from the T7 promoter, which lies downstream of the GAL4 coding sequences. Equal amounts of the 35 S-labeled occludins and WW Itch were pooled and incubated at 30°C for 1 1 ⁄2 h. Then, 500 l of co-immunoprecipitation buffer (0.5% Triton X-100, 20 mM Tris-Cl (pH 7.5), 150 mM NaCl, 1 mM CaCl 2, 5 mM Na 3 VO 4 , 10 mM NaF, 5 g/ml aprotinin, 5 g/ml leupeptin, 2 g/ml pepstatin A, 1 mM phenylmethylsulfonyl fluoride, 10% glycerol) and 2 g of mouse monoclonal anti-c-Myc (Santa Cruz Biotechnology Inc.) were added. The samples were rotated for 2 h at 4°C followed by the addition of 20 l of 1:1 slurry of Protein G-Sepharose TM 4 Fast Flow (Amersham Bioscience Inc.). Collection of the immunocomplexes was FIG. 2. WW Itch binds a putative WW recognition motif in the N terminus of occludin. A, schematic diagram illustrating the wild-type occludin and the truncated/mutated occludin constructs used in this in vitro binding study. Each construct was COOH-terminal Myctagged. B, the various Myc-tagged occludins and WW Itch were synthesized in vitro in the presence of [ 35 S]methionine. After combining equal amounts of radiolabeled protein, Myc-tagged fusion proteins were immobilized to Protein G-Sepharose TM using a monoclonal antic-Myc antibody. Only wt-Occludin was capable of binding WW Itch (lysate ϭ stoichiometric amount of WW Itch). C, stoichiometric amounts of the translated occludin-constructs were separated on a 10% polyacrylamide-gel. carried out at 4°C for 3 h. The samples were centrifuged at 1600 ϫ g for 30 s and washed 4 times with 1 ml of co-immunoprecipitation buffer. The samples were then resuspended in 15 l of SDS sample buffer, boiled for 5 min, and analyzed on 10% SDS-polyacrylamide gels. After electrophoresis, the gel was fixed for 30 min in 30% methanol, 10% acetic acid, soaked for 30 min in Amplify TM (Amersham Biosciences Inc.), and dried under vacuum at 80°C for 30 min. Bound 35 S-labeled proteins were analyzed by fluorography.
Co-immunoprecipitation and Immunoblotting-Co-immunoprecipitation of endogenous Itch with occludin or claudin-1 was examined using E13 whole mouse embryo homogenate. In short, embryonic tissue was homogenized in 2 ml of lysis buffer (0.5% Triton X-100, 0.5% Nonidet P-40, 20 mM Tris-Cl (pH 7.5), 250 mM NaCl, 0.2 mM EDTA, 5 mM Na 3 VO 4 , 25 mM NaF, 5 g/ml aprotinin, 5 g/ml leupeptin, 2 g/ml pepstatin A, 1 mM phenylmethylsulfonyl fluoride) using a tight fitting glass/Teflon homogenizer. The homogenate was kept on ice for 30 min, centrifuged at 16,000 ϫ g for 10 min and the supernatant was split into 4 tubes. 2.5 g of anti-occludin or anti-claudin-1 antibody, 2 g of anti-c-Myc (an unrelated antibody as a control), and no antibody (negative control) were added to the samples, and incubated for 2 h at 4°C. 30 l of 1:1 slurry of Protein G-Sepharose TM 4 Fast Flow were added to each sample and immunoprecipitations were carried out at 4°C overnight. The beads were recovered by centrifugation at 1,600 ϫ g for 30 s and washed four times with 1 ml of lysis buffer. Bound proteins were eluted by boiling in SDS sample buffer and analyzed on 4 -12% SDS-polyacrylamide gels.
Fractionated proteins were transferred to Fluorotrans Transfer Membrane (PALL) by electroblotting and immunoprobed with antioccludin, anti-claudin-1, or anti-Itch Ab. For visualization of detected proteins immunoblots were analyzed using an ECL Western blot detection kit.
Pulse-Chase Experiments-LLC-PK1 cells were starved in methionine/cysteine-deficient MEM (Invitrogen) for 45 min. The medium was replaced by 1.5 ml of methionine/cysteine-deficient MEM containing 0.1 mCi/ml ARS 110-Met-label [ 35 S] (American Radiolabeled Chemicals). After pulse labeling for 1 h, the cells were washed 3 times with nonradioactive Dulbecco's modified Eagle's medium and then were chased in 1.5 ml of methionine/cysteine-deficient MEM supplemented with nonradioactive L-cysteine (0.24 mg/ml) and L-methionine (0.15 mg/ml) for 0 -3 h. 80 M MG-132 (dissolved in DMSO; Calbiochem) or DMSO as a vehicle control was added to the starving, pulse labeling, and chase medium.
At appropriate time intervals, the cells were washed 3 times with ice-cold phosphate-buffered saline, resuspended in 300 l of PCL buffer (1% Triton X-100, 0.75% Nonidet P-40, 0.1% SDS, 20 mM Tris-Cl (pH 7.5), 150 mM NaCl, 5 mM EDTA, 5 mM Na 3 VO 4 , 25 mM NaF, 5 g/ml aprotinin, 5 g/ml leupeptin, 2 g/ml pepstatin A, 1 mM phenylmeth- Endogenous occludin was immunoprecipitated from equal amounts of total protein and collected immunocomplexes were subjected to an immunoblot analysis using monoclonal anti-HA antibody. As a negative control the assay was performed with untransfected cells. HA-ubiquitin conjugates of occludin are highlighted by a bracket. B, Triton X-100-soluble (sol) and -insoluble (insol) fractions of HEK 293E cells expressing HA-tagged ubiquitin were immunoprecipitated using anti-occludin antibody after treatment with MG-132 (50 M, 1 h). Bound proteins were analyzed as described above. Polyubiquitinated occludin could be detected in the Triton X-100-soluble and -insoluble fraction.
ylsulfonyl fluoride), and kept on ice for 30 min. The lysate was centrifuged at 16,000 ϫ g for 10 min at 4°C, the supernatant was collected, and the protein content was determined using a BCA TM -200 Protein Assay Kit (Pierce). 2.25 g of anti-occludin or anti-claudin-1 antibody was added to equal amounts of total protein (500 g) and immunoprecipitations were carried out as described above. Bound proteins were eluted by boiling in SDS sample buffer and were analyzed by SDS-PAGE and fluorography. Densitometric measurements were performed using E.A.S.Y. Win32 software (Herolab).
In vivo ubiquitination assays were performed as previously described (19). In short, HEK 293 cells were transiently co-transfected with full-length Myc-occludin cloned into pcDNA3.   , left figure); MG-132 and DMSO were present during the starving, the pulse, and the chase period. Occludin immunoprecipitation was performed as described above and densitometric measurements were carried out using E.A.S.Y. Win32 software (Herolab). In the presence of MG-132 ϳ30% of initially labeled occludin could be detected after a 3-h chase period, whereas only ϳ1% was left in the control experiment.
ti-HA antibody. Generation of Triton X-100-soluble and -insoluble fractions was performed as previously described (23).
Transepithelial Electrical Resistance-To measure transepithelial electrical resistance, MDCK cells were plated onto Falcon R cell culture inserts (Becton Dickinson) at a density of about 2 ϫ 10 5 cells/cm 2 and maintained in serum-containing medium until TER values reached a plateau at maximum levels (150 -300 ⍀⅐cm 2 ). The resistance of the monolayers was determined using an EVOM voltohmmeter (World Precision Instruments, Sarasota). TER values were multiplied by the surface area of the filter (4.2 cm 2 ) after subtraction of background values.

RESULTS AND DISCUSSION
To test for candidate proteins interacting with the NH 2terminal domain of occludin the entire NH 2 -terminal region (N-Occl) was used as "bait" to screen a MATCHMAKER TM mouse liver cDNA expression library. DNA from two putative positive yeast clones (WW Itch and WW Itch-Hect) (Fig. 1B) was isolated and plasmid DNA was rescued in E. coli strain DH5-␣. Following plasmid purification from E. coli, inserts of both clones were sequenced and compared with the sequence entries in GenBank TM and EMBL using a BLAST2 homology search. The isolated cDNAs showed 100% identity to the coding sequence of Itch, a mouse E3 ubiquitin-protein ligase (Gen-Bank TM accession number AF037454) (18).
Itch is a member of the HECT domain-containing subfamily of E3 ubiquitin-protein ligases (24). Fig. 1A shows the functional domains of mouse Itch. The HECT (homologous to the E6-associated protein carboxyl terminus) domain has been suggested to be responsible for the ubiquitinating activity of the E3 ligases, whereas the binding specificity for target proteins is mediated by sequences within the NH 2 terminus. Disruption of Itch was found to induce a spectrum of immunological diseases in mutant mice, including inflammations of the large intestine with infiltrates consisting of mainly neutrophils (18). Whether occludin is affected in this mutant phenotype is unknown.
The cDNA inserts of the clones WW Itch and WW-Itch Hect comprise 851 and 1620 bp of mouse Itch cDNA, extending from nucleotides 851-1636 and 494 -2114, respectively (Fig. 1B). Interestingly, both clones encode an overlapping region containing all four WW domains of Itch.
To rule out the possibility that the observed interaction was artifactual, WW Itch and WW Itch-Hect were tested for autoactivation of the E. coli lacZ reporter gene, for interaction with the GAL4 activation domain, and for interaction with an unrelated protein (human lamin C) in a two-hybrid assay. Moreover, a frameshift mutation was generated, shifting the codon usage of WW Itch by one nucleotide. As summarized in Fig. 1B and C, no interaction was found in control experiments. Additionally, interaction of WW Itch with N-Occl was still traceable when cloning vectors were switched by transferring the library insert from the AD to the DNA-BD vector and vice versa (Fig. 1C).
WW domains are protein-protein interaction modules of about 35-45 residues found in signaling and regulatory proteins (25). These domains are named after a pair of highly conserved tryptophans and a strictly conserved proline. Like SH3 domains WW domains recognize proline-rich sequences although with highly diverse sequence preferences (for review, see Ref. 26). Thus, we have tested whether the Pro 9 -Pro 10 -Tyr 11 -Pro 12 motif located in the NH 2 -terminal domain of occludin serves as a possible recognition site for WW Itch. Using co-immunoprecipitation and a two-hybrid assay we have found that the interaction of WW Itch with occludin is abolished when the putative WW recognition motif of occludin is mutated by substituting Pro 9 and Pro 10 for Ala 9 and Ala 10 ( Fig. 1E and Fig. 2B).
We have further analyzed which of the WW domains present in WW Itch and WW Itch-Hect are responsible for the binding of Itch to the NH 2 terminus of occludin (Fig. 1D). As a result, the presence of WW2 and WW4 was found to be critical for the association of Itch with occludin. Neither WW1 nor WW3 alone were able to bind to the occludin bait. The partial HECT domain which is included in WW Itch-Hect does not seem to be of importance for the interaction, since occludin-Itch interaction has also been found with WW Itch, lacking the HECT domain.
Additional evidence for the association of Itch and occludin was provided by in vitro pull-down assays. 35 S-Labeled modified and unmodified versions of Myc-occludin and WW Itch protein were synthesized using a rabbit reticulocyte lysate system. WW Itch was found to co-precipitate with wild-type occludin but not with NH 2 -terminal truncated occludins (occl-⌬N, occl-⌬N1/2) (Fig. 2B). This was also confirmed by our

FIG. 8. Proteasome inhibition influences TER in MDCK cells.
A, MDCK cells were grown to confluence on filter supports in normal medium and were subsequently changed to low calcium medium for 12 h. After low calcium incubation, medium was switched to normal MEM and TER was monitored in the presence or absence of MG-132. The time point of medium exchange was considered as t ϭ 0. Addition of 50 M MG-132 markedly accelerated TER increase within the first hour. Incubation times Ͼ3 h appeared to be toxic to the cells. B, confluent MDCK cells, exhibiting maximum TER values, were switched to low calcium medium with or without 25 M MG-132 and TER was measured at various time intervals. TER decrease was inhibited by 31% at the time point of 150 min. Again, the time point of medium exchange was considered as t ϭ 0. Filters were used in triplicate. Three independent experiments were performed and means of one representative experiment are shown in each panel. Variations were below statistical significance (error bars not shown).
results from a two-hybrid assay, showing that the NH 2 terminus of occludin is indispensable for the interaction of WW Itch with occludin (Fig. 1E).
To further substantiate the interaction found, endogenous Itch was co-precipitated with endogenous occludin from homogenized embryonic mouse tissue (Fig. 3). As controls, immunoprecipitations were carried out either in the absence of an antibody or in the presence of an unrelated antibody (mouse monoclonal anti-c-Myc). Itch specifically co-precipitated with occludin (Fig. 3, lane 2) but not with claudin-1 (Fig. 3, lane 6). Only minimal unspecific binding of Itch to the Sepharose beads could be detected.
To find out whether occludin is a substrate for ubiquitination, we performed in vitro and in vivo ubiquitination experiments. For in vitro studies we used rabbit reticulocyte lysate as a source of ubiquitination enzymes. 35 S-Labeled occludin was incubated with the reticulocyte lysate in the presence of His 6x -ubiquitin and reaction products were purified by metal chelation chromatography. In parallel, in vitro translated radiolabeled claudin-1, ZO-1, ZO-2, and lamin C (control) were tested for ubiquitination. Strong ubiquitination of occludin was observed as evidenced by a high molecular weight smear (Ͼ60,000), representing mono-and polyubiquitinated forms of occludin ( Fig. 4 lane 2). In contrast to occludin, claudin-1, ZO-1, and ZO-2 were not ubiquitinated in vitro, nor was the junctionunrelated protein lamin C (Fig. 4, lanes 5, 9, 12, and 15).
We have also demonstrated that ubiquitination of occludin occurs in vivo. For these experiments cell lysates from HEK 293 cells overexpressing HA-ubiquitin were immunoprecipitated with anti-occludin antibody and Western blots were probed with anti-HA antibody. We found that short-term treatment with the proteasome inhibitor MG-132 significantly led to an accumulation of occludin-ubiquitin conjugates in these cells (Fig. 5A).
To further confirm that Itch is critically involved in the ubiquitination of occludin, HEK 293 cells were co-transfected with Myc-occludin and HA-ubiquitin plasmids in the absence or presence of an Itch expression construct. Following immunoprecipitation of cell lysates with anti-Myc and immunoblotting with anti-HA antibody, ubiquitination of occludin was found to be specifically induced in cells overexpressing Itch (Fig. 6). Cells containing only endogenous Itch (detectable upon overexposure of immunoblots) exhibited substantially lower levels of ubiquitinated occludin. Again, MG-132 increased the levels of occludin-ubiquitin conjugates in HA-ubiquitin overexpressing cells (Fig. 6).
To address the important question which pool of occludin is ubiquitinated, occludin-ubiquitin conjugates were determined in Triton X-100-extracted fractions of HEK 293 cells overexpressing HA-ubiquitin. The major portion of ubiquitinated occludin was observed in the soluble fraction (Fig. 5B, lane 1). However, substantial amounts of occludin-ubiquitin conjugates were also detectable in the insoluble fraction (Fig. 5B, lane 2), suggesting that Itch is capable of associating with junctional and lateral occludin as well.
Since ubiquitination of proteins is usually associated with their rapid turnover, we initially investigated the turnover rate of occludin using pulse-chase experiments. Fig. 7A shows SDS-PAGE patterns of occludin immunoprecipitates from pulselabeled LLC-PK1 cells. Our results indicate that occludin is a short-lived protein with a t1 ⁄2 of about 1.5 h. After a 3-h chase, only 1.2% of 35 S-labeled occludin was still detectable in cell lysates of LLC-PK1 cells, while the amount of 35 S-labeled claudin-1 was unchanged. When the proteasome inhibitor MG-132 was added during a 1-h pulse and the subsequent chase period, degradation of occludin was reduced by 30% (Fig. 7B). This is in line with previous reports showing that proteasome inhibition led to the stabilization of junctional proteins such that experimentally induced scattering of epithelial cell monolayers could be inhibited (27). Similarly, the half-life of ␤-catenin was increased by about 3-fold upon proteasome inhibition (28), and reducing connexin degradation with inhibitors of the proteasome was found to induce assembly and function of gap junctions (29).
Although the rapid turnover of occludin is reminiscent of degradation rates reported for connexins (29,30), it appears rather unusual compared with the relatively long half-lives shown for E-cadherin (27) and desmosomal proteins (31). A first attempt to study the degradation of occludin in cultured cells was made by Wong and Gumbiner (32), showing that a synthetic peptide, corresponding to the second extracellular loop of occludin, enhances the degradation of occludin in steady-state 35 S-labeled Xenopus A6 epithelial cells (32). In contrast to our findings, occludin appeared remarkably stable in this experiment, showing a first significant decrease only after 12 h of peptide treatment. It has to be mentioned that the properties of various types of epithelia and endothelia differ substantially as reflected by TER values ranging from 5 ⍀⅐cm 2 to 8,000 ⍀⅐cm 2 . Thus, the slow turnover of occludin reported by Wong and Gumbiner (32) may be explained by the fact that exceptionally tight A6 monolayers exhibiting TER values of up to 8,000 ⍀⅐cm 2 were used for their study. In our experiments, maximum TER values of LLC-PK1 and MDCK cells leveled at 150 -300 ⍀⅐cm 2 which is consistent with most reports found in the literature.
As a next step we examined whether a reduction in the rate of occludin degradation influences functional properties of TJs. To this end, MDCK cells were grown on filter supports and TER measurements were performed in the presence or absence of the proteasome inhibitor MG-132. In a first experiment, confluent MDCK cells were grown in low calcium medium (LC) for 12 h to disrupt pre-existing junctional complexes. Subsequently, the culture medium was switched to normal calcium (NC) levels with or without MG-132 and TER was measured at various time intervals. We found that MG-132 significantly accelerated the TER increase compared with control cells (Fig.  8A). On the other hand, MG-132 diminished TER decrease in MDCK monolayers switched to LC medium by about 30% (Fig.  8B). Although it is reasonable to assume that MG-132 leads to the accumulation/stabilization of any junctional protein normally degraded by the proteasome, the exceptionally rapid increase in TER indicates that preferably short-lived junctional proteins may have been affected.
The molecular mechanism underlying the "stabilizing" effect FIG. 9. Localization of Itch-EGFP and occludin in HEK 293 cells. Subconfluent HEK 293 cells transfected with Itch-EGFP were grown in 30-mm plastic Petri dishes, fixed in ethanol/acetic acid, and stained for occludin using Alexa Fluor TM 568 IgG as secondary antibody. A, Itch-EGFP is uniformly distributed in the cells, which was also confirmed by staining with a polyclonal anti-Itch antibody (not shown). B, arrows indicate localization of occludin at plasma membranes, which appears unchanged in cells overexpressing Itch. C, multichannel image of A and B (original magnification ϫ400). Immunofluorescence was visualized using a Zeiss Axioplan microscope and the Zeiss Axiovision TM imaging system. of proteasome inhibition on TJs is still unclear. However, taking into account that deubiquitinating enzymes are active at junction sites (33,34), it is tempting to speculate that ubiquitinated junctional proteins, which accumulate upon proteasome inhibition, may well serve as a pool for rapid (re)integration into TJs following deubiquitination. In this way, spontaneous alterations of TJ properties could be explained.
Taken together, our results provide evidence to suggest that occludin directly interacts with the E3 ubiquitin-protein ligase Itch, a member of the HECT domain-containing ubiquitinprotein ligases. Our molecular studies have shown that occludin is ubiquitinated in vivo and in vitro and its turnover is slowed down by the proteasome inhibitor MG-132. Overexpression of Itch substantially increased the amount of occludinubiquitin conjugates predominantly present in the soluble fraction in cultured epithelial cells. Interestingly, overexpression of Itch did not alter the total amount of occludin in HEK 293 cells (not shown) nor did it perturb the localization of occludin at plasma membranes (Fig. 9).
Since ubiquitination of proteins is considered to be sufficient to induce their internalization, E3 ubiquitin-protein ligases exert a crucial function in the endocytosis of plasma-membrane proteins. So far, several receptors and one ion channel have been shown to undergo ubiquitination at the plasma membrane (35)(36)(37)(38). In addition, cytoplasmic junctional components such as ␤-catenin and AF-6 were found to serve as targets for the ubiquitin-proteasome pathway (28,33,34).
Interestingly, claudin-1 did not interact with Itch nor was it ubiquitinated in HEK 293 cells. Together with our results showing that turnover rates of occludin and claudin differed markedly in pulse-chase experiments, this finding supports the notion that occludin and claudins exert physiologically distinct functions at TJs and thus are subjected to different degradation pathways. Previous reports have already suggested that TJ function can be up-regulated at the level of occludin turnover (32). We now provide further molecular evidence to understand the regulatory steps underlying this dynamic process of TJ modulation.