Reagents

Latrunculin A (LatA); Alexa Fluor 488-conjugated phalloidin; Alexa Fluor 350-, 488-, or 594-conjugated secondary antibodies; rabbit anti-green fluorescent protein (GFP) antibody; and Hoechst 33342 were from Molecular Probes (Eugene, OR). Mouse monoclonal anti-occludin, rabbit anti-claudin-1, mouse anti-ZO-1, and mouse anti-claudin-4 antibodies were from Zymed Laboratories (South San Francisco, CA). Rabbit anti-caveolin-1 antibodies were from Abcam (Cambridge, MA) and Santa Cruz Biotechnology (Santa Cruz, CA). Rabbit anti-clathrin heavy chain antibodies also were from Santa Cruz Biotechnology. Rabbit anti-dynamin II was from Calbiochem (San Diego, CA). Rabbit anti-EEA1 was from Affinity Bioreagents (Golden, CO). Rat anti-lysosome-associated membrane protein (LAMP)-1 monoclonal antibody (clone ID4B) was the generous gift of Dr. M. R. Clark (The University of Chicago, Chicago, IL). Rat anti-E-cadherin (clone DECMA-1), chlorpramazine, amiloride, filipin III, methyl-β-cyclodextrin, and water-soluble cholesterol were all from Sigma (St. Louis, MO). LY294002 was from Cell Signaling Technology (Beverly, MA). All other chemicals were of the highest grade available.

Plasmids

The enhanced green fluorescent protein (EGFP)-β-actin construct was generated by fusing EGFP (BD Biosciences Clontech, Palo Alto, CA) to the amino terminus of human β-actin (I.M.A.G.E. clone 769559; American Type Culture Collection, Manassas, VA) and cloning the construct into pcDNA6/TR (Invitrogen, Carlsbad, CA). Monomeric red fluorescent protein 1 (mRFP1), a gift from Roger Y. Tsien and Robert E. Campbell (University of California, San Diego, La Jolla, CA) (Campbell et al., 2002), was cloned into NheI-KpnI sites of pcDNA3.1 zeo (+) (Invitrogen) to form the pmRFP1-C vector. mRFP1-tagged occludin was generated by PCR amplification and cloning of the coding region of human occludin, a gift from Randall J. Mrsny (Cardiff University, Cardiff, United Kingdom) into KpnI-XbaI sites of pmRFP1-C vector in frame with mRFP1. mRFP1-ZO-1 was generated using the coding region of pSK-ZO-1, a kind gift from Alan S. Fanning and James M. Anderson (University of North Carolina, Chapel Hill, NC) by first inserting KpnI sites flanking the coding region and then cloning the coding region into the KpnI site of pmRFP1-C. EGFP-claudin-1 was generated by PCR amplification of coding region of human claudin-1 (I.M.A.G.E. clone 4500534; American Type Culture Collection) and cloning into the KpnI-XbaI sites of pEGFP-C1 (BD Biosciences Clontech). Integrity of the plasmids was verified by restriction digestion and direct sequencing of the expression construct and adjacent regions. Dynamin II-EGFP, dyamin II-K44A-EGFP, EGFP-dyamin I, and EGFP-dyamin I-K44E (Di et al., 2003) were graciously provided by Dr. H. C. Palfrey (The University of Chicago). Caveolin-1-EGFP and enhanced yellow fluorescent protein (EYFP)-clathrin light chain A1 (Massol et al., 2004) were kindly provided by Dr. T. Kirchhausen (Harvard Medical School and the CBR Institute for Biomedical Research, Boston, MA). EGFP-Eps15Δ95-295 (Nichols et al., 2001) was generously provided by Dr. J. Lippincott-Schwartz (National Institutes of Health, Bethesda, MD).

Cell Culture and Transfection

Madin-Darby canine kidney (MDCK) cells (American Type Culture Collection) were grown in DMEM with 1 g/l glucose (Mediatech, Herndon, VA) supplemented with 10% fetal bovine serum (Invitrogen) and 15 mM HEPES, pH 7.4. For growth as monolayers cells were plated at confluent density, media were replenished at 24 h, and monolayers were studied 48 h after plating. Stable transfectants were generated by transfection with Lipofectamine 2000 (Invitrogen) in 100-mm dishes, as described previously (Zhao et al., 2004). After growth in medium supplemented with 2.5 mg/ml G418 (Mediatech) or 15 μg/ml blasticidin S (Calbiochem, San Diego, CA) for 2 wk, cells were trypsinized and sorted into 96-well plates using a MoFlow cell sorter (DakoCytomation California, Carpinteria, CA). After expansion, clones with readily-detectable TJ fluorescence were selected for further study. Monolayers of stable transfectants were cultured in media with 7.5 mM sodium butyrate to augment transgene expression, from 24 to 48 h after plating and then transferred to media without sodium butyrate for 4 h before use. Transiently-transfected MDCK cells were plated directly onto Transwells and studied 2 d after transfection.

LatA Treatment

Monolayers were equilibrated in Hank's balanced saline (HBSS) with 15 mM HEPES, pH 7.4, for 1 h at 37°C before study. The basolateral media were then exchanged for HBSS containing LatA. For studies at different temperatures, monolayers were equilibrated at indicated temperatures for 30 min (after equilibration in HBSS at 37°C) before LatA treatment. To synchronize vesicle formation and internalization monolayers were cooled to 14°C before and during 30 min of LatA treatment to allow actin depolymerization. The temperature was then raised to 37°C to allow synchronous endocytosis.

Inhibitor Studies

For inhibitor studies, monolayers were transferred to HBSS as described above. Appropriate inhibitors were added bilaterally and monolayers incubated for 1 h at 37°C before basolateral LatA addition. For cholesterol depletion/repletion experiments monolayers were treated with methyl-β-cyclodextrin (MBCD) for 45 min followed by transfer to media containing MBCD or 1:10 cholesterol–MBCD complexes (Christian et al., 1997) for an additional 45 min. Monolayers were then treated with LatA as described above.

Transepithelial Electrical Resistance (TER) Measurement

TER was measured using electrodes placed on either side of the monolayer. These were attached to a current clamp (University of Iowa Bioengineering, Iowa City, IA) or EVOM (WPI, Saratosa, FL), as described previously (Turner et al., 1997; Wang et al., 2005).

Determination of F-Actin Content

Confluent MDCK monolayers were washed, equilibrated in HBSS for 1 h at 37°C, and transferred to HBSS at indicated temperatures for 30 min. Monolayers were then transferred to HBSS of the same temperature with LatA and fixed in 3% paraformaldehyde in HBSS at indicated times. After washing with HBSS with 50 mM NH₄Cl and permeabilization in HBSS containing 3% bovine serum albumin (BSA) and 0.5% saponin monolayers were incubated with 0.5 μM Alexa Fluor 488-conjugated phalloidin and 0.2 μg/ml Hoescht 33342 in HBSS with 3% BSA and 0.5% saponin for 1 h at room temperature. After five washes in HBSS with 3% BSA and 0.5% saponin and one wash with HBSS, fluorescent intensity was measured (Synergy HT; Bio-Tek Instruments, Winooski, VT). Alexa Fluor 488 fluorescence was normalized to Hoechst 33342 fluorescence to correct for cell number.

Immunofluorescence Staining

Monolayers were fixed with 1% paraformaldehyde in phosphate-buffered saline (PBS), pH 7.4, with 1 mM CaCl₂ for 30 min at room temperature. After three washes in PBS and a 15-min incubation in PBS with 50 mM NH₄Cl cells were permeablized in PBS with 3% BSA and 0.05% saponin. Monolayers were then incubated with primary antibody in PBS with 3% BSA and 0.05% saponin overnight at 4°C; washed five times with PBS, 3% BSA, and 0.05% saponin; and incubated with the appropriate Alexa Fluor 350-, 488-, or 594-conjugated secondary antibodies. After five washes monolayers were rinsed in water and mounted in SlowFade (Molecular Probes).

To retain intracellular vesicles, monolayers were briefly washed with ice-cold HBSS and fixed in methanol overnight at –20°C. After fixation monolayers were air-dried, rehydrated with 100 μM bis(sulfosuccinimidyl)suberate (BS³) in PBS with 0.1% n-octyl-glutaraldehyde (PBS+) for 30 min, washed in PBS+, quenched in 100 mM ethylenediamine, pH 7.5, and washed once more in PBS+. Monolayers were then blocked in 1% nonfat dry milk, 1% fish gelatin, and 1% normal donkey serum in PBS+ for 1 h; incubated with primary antibodies for 2 h; washed; and incubated with the appropriate Alexa Fluor 350-, 488-, or 594-conjugated secondary antibodies for 1 h. After five washes, monolayers were rinsed in water and mounted in SlowFade.

Immunoblotting

MDCK monolayers prepared as described above were washed once with ice-cold HBSS and lysed directly in nonreducing SDS-PAGE sample buffer. After boiling at 95°C for 5 min, aliquots were resolved on SDS-PAGE gels (Bio-Rad, Hercules, CA) and transferred to polyvinylidene difluoride membranes. Membranes were blocked in 5% nonfat dry milk in Tris-buffered saline (TBS), washed in TBS with 0.05% Tween 20 (TBS-T), and incubated with primary antibodies in TBS with 1% nonfat dry milk overnight at room temperature. Membranes were washed with TBS-T, incubated with appropriate horseradish peroxidase-conjugated secondary antibodies (Cell Signaling Technology), and washed with TBS-T before detection by enhanced chemiluminescence (SuperSignal; Pierce Chemical, Rockford, IL).

Subcellular Fractionation

Monolayers were washed with 37°C HBSS and then scraped into ice-cold HBSS. After recovery by centrifugation at 4°C, cells were resuspended in TBS containing 1 mM 4-(2-aminoethyl)benzenesulfonyl fluoride, 0.8 mM aprotinin, 0.5 mM bestatin, 15 μM E-64, 20 μM leupeptin, and 10 μM pepstatin A and lysed by 60 passes through a bent 25-gauge needle. Cell disruption was monitored microscopically. Lysates (4 ml) were loaded above 8-ml 10–60% continuous sucrose gradients (in TBS) and centrifuged at 280,000 × g for 18 h at 4°C. Then, 0.5-ml fractions were collected and analyzed for sucrose concentration by refractometry and distribution of specific proteins by SDS-PAGE and immunoblot.

Fluorescence Microscopy with Simultaneous TER Measurement

Monolayers were mounted in a Petri dish in a custom-designed temperature-controlled stage (Brook Industries, Lake Villa, IL). TER was measured using an EVOM and Ag-AgCl electrodes embedded in the Petri dish base and fastened to the side of the microscope objective. Multidimensional imaging was performed by using an epifluorescence microscope (DMLB; Leica Microsystems, Bannockburn, IL), a 63× HCX PL APO L U-V-I aqueous immersion objective, and a 51022 filter set optimized for EGFP/mRFP1 (Chroma Technology, Rockingham, VT). The z-stacks were collected in 0.2–0.5 μm steps using motorized excitation filter wheels, z-motor, and shutter (Ludl, Hawthorne, NY) and a Retiga EXi camera (Q Imaging, Burnaby, British Columbia, Canada) controlled by MetaMorph 6 (Universal Imaging, Downingtown, PA).

For fixed cells, stained monolayers were observed using the same microscope equipped with an 88000 filter set (Chroma Technology) and 63× or 100× PL APO oil immersion objectives. The z-stacks were collected at 0.2-μm intervals. Image stacks were deconvolved using AutoDeblur 9.3 (Autoquant, Watervliet, NY).

Morphometric Analysis

Deconvolved z-stacks were merged, after pseudocolor assignment, using MetaMorph. Vesicles were defined as round or oval structures present in three or more z-planes. The number of vesicles in a single cell was counted over the full height of the cell. Signals were considered to be colocalized if there was ≥80% overlap between channels. For each measurement, 20 average-shaped and -sized cells chosen randomly were counted.

Statistical Analysis

All data are presented as average ± SE. All experiments were performed with triplicate or greater samples, and data shown are representative of three or more independent studies. P value was determined by two-tailed Student's t test and was considered to be significant if ≤0.05.

F-Actin Depolymerization Leads to Functional and Morphological TJ Disruption

Alteration of epithelial barrier function by cytochalasins is well established (Bentzel et al., 1980) and has been reported to exhibit an unusual biphasic dose dependence: low doses cause mild increases in TER, whereas higher doses cause TER to fall (Stevenson and Begg, 1994; Turner, 2000) (Figure 1A). The mechanism of this biphasic response has not been defined, but it may reflect the multiple mechanisms by which cytochalasins disrupt actin filaments, because cytochalasins cap barbed ends at low doses, whereas at higher concentrations they also sever microfilaments (MacLean-Fletcher and Pollard, 1980; Cooper, 1987; Spector et al., 1989). We therefore compared the effects of cytochalasin D to those of LatA, which causes microfilament depolymerization by a single mechanism, binding to and sequestration of G-actin monomers (Spector et al., 1983; Morton et al., 2000). Like cytochalasin D, LatA caused TER to fall at 0.2 μM (Figure 1A). However, unlike cytochalasin D, low doses of LatA did not cause TER to increase. Because of this simpler monophasic TER response to LatA treatment, LatA was used throughout these studies aimed at defining the mechanisms by which actin depolymerization disrupts TJ structure and function. Kinetic analyses of the effects of LatA on TER and F-actin content showed that LatA reduced F-actin content by 17 ± 0.4% and TER by 17 ± 0.6% within 5 min (Figure 1B). Both F-actin content and TER continued to decrease over time, falling by 42 ± 1.4 and 82 ± 1.1%, respectively, at 30 min. Thus, TER decreases induced by LatA correlated temporally with actin depolymerization.

To identify morphological changes in TJ and perijunctional actin structure that correlate with loss of barrier function, monolayers were fixed using standard paraformaldehyde protein cross-linking, and specific TJ proteins or F-actin was labeled (Figure 1, C–G). Unlike decreases in TER and F-actin content, morphological changes were not detected until at least 20 min after LatA addition, long after substantial TER loss had occurred. The morphological changes detected included disruption of perijunctional actin, with alternating actin-rich and actin-poor zones, loss of microvillus F-actin, formation of apical actin aggregates (Figure 1C), loss of basal stress fibers (Figure 1D), and disruption of TJ-localized occludin (Figure 1E), ZO-1 (Figure 1F), and claudin-4 (Figure 1G). Although consistent with previous reports of actin and ZO-1 redistribution after actin disruption by cytochalasin D, these data show that this approach only detects morphological changes well after LatA-induced barrier function loss has occurred.

Development and Characterization of Fluorescent TJ Fusion Proteins

The data mentioned above demonstrate that traditional immunofluorescence approaches are unsuitable for identification of specific morphological changes that correlate with LatA-induced loss of barrier function. We therefore developed fusion proteins of β-actin, occludin, ZO-1, and claudin-1 using two fluorochromes: EGFP and mRFP1 (Campbell et al., 2002). EGFP and mRFP1 fluoresce with high quantum yield at wavelengths that are resolvable from one another, thus allowing their use for double labeling. These fluorescent proteins were linked to the amino termini of β-actin and ZO-1 because similar constructs have been shown to be effective (Choidas et al., 1998; Riesen et al., 2002). For claudins, both amino and carboxy-terminus fusions constructs have been reported (Sasaki et al., 2003; Matsuda et al., 2004), but, because the carboxy terminus is required for interaction with PDZ domains, e.g., of ZO-1, the fusion protein was attached to the amino terminus of claudin-1. Similarly, occludin was tagged at the amino terminus, because the carboxy terminus has been reported to be important for interactions with other proteins (Nusrat et al., 2000a; Peng et al., 2003). These fusion proteins were stably expressed in green (EGFP)–red (mRFP1) pairs in MDCK monolayers. Compared with the corresponding endogenous protein, each transfected fusion protein represented no more than one-half of the total of that protein (Figure 2, A–C). Expression of β-actin, ZO-1, and claudin-1 fusion proteins did not alter TER of transfected MDCK monolayers (our unpublished data). Expression of occludin increased TER by 21 ± 4%. This effect was similar, but less pronounced, than that reported after expression of chicken occludin in MDCK monolayers (McCarthy et al., 1996), suggesting that the fusion construct is a functional occludin protein. All transfected lines were sensitive to LatA with dose-responses and kinetics similar to nontransfected MDCK monolayers (our unpublished data). Morphological analyses of the distributions of these proteins showed that, in each case, the tagged fusion protein colocalized completely with the endogenous protein (Figure 2, D–F), suggesting that the tagged proteins are correctly targeted. Subcellular fractionation studies provide further evidence for correct targeting of tagged proteins, because both mRFP1-occludin and EGFP-claudin-1 comigrated with endogenous proteins in subcellular fractions in transfected cells and were of the same density as endogenous occludin or claudin-1 in nontransfected cells (Figure 2, G and H). Thus, the fusion proteins do not disrupt TJ assembly, LatA response, or targeting of endogenous TJ proteins and are themselves appropriately targeted, as assessed morphologically and biochemically. Therefore, these fusion proteins are suitable tools for analysis of TJ dynamics in live cells.

Live Cell Imaging Allows the Identification of Structural Changes That Are Temporally Correlated with LatA-induced Loss of Barrier Function

We analyzed polarized MDCK monolayers expressing fluorescent fusion proteins in an apparatus that allows simultaneous imaging and electrophysiological analysis (Figure 3A). Preliminary experiments showed that both protein localization and TER were generally stable during imaging of these monolayers, although subtle movements of individual TJ proteins within the TJ and minor cell shape changes could be detected (Figure 3B and Supplemental Movie S1). Thus, this apparatus is suitable for real-time analyses of TJ protein distribution and TER.

Distributions of fluorescent fusion proteins were initially imaged by collecting z-stacks at 1-min intervals before and after addition of LatA. TER was measured simultaneously. In monolayers expressing EGFP-β-actin and mRFP1-occludin, the latter was uniformly distributed along the TJ before LatA addition. Within 2 min after LatA addition, when TER decreases were first detected, distribution of mRFP1-occludin at the TJ became nonuniform, with alternating zones relatively enriched in or depleted of mRFP1-occludin (Figure 3C and Supplemental Movie S2). This persisted throughout the z-stack and was therefore not due to vertical redistribution or an artifact of imaging. EGFP-β-actin localization changed similarly and regions with decreased mRFP1-occludin generally also had reduced EGFP-β-actin fluorescence, suggesting a coordinated redistribution of EGFP-β-actin and mRFP1-occludin. The small zones enriched in mRFP1-occludin frequently formed small buds and then separated from the TJ in a process that was repeated frequently over the first 15 min after LatA addition (Figure 3D). To further characterize this process of occludin removal from the TJ, imaging was performed at higher rates. These images clearly show that occludin is first concentrated in a small zone within the TJ, followed by local invagination, rounding, and detachment (Figure 3E and Supplemental Movie S3). These data are therefore consistent with endocytic removal of occludin after LatA addition. Analysis of xz reconstructions of fixed monolayers immunostained for occludin, as shown in Figure 3F, demonstrated budding and detached structures near the lower portion of the TJ, i.e., adjacent to the border of the TJ and lateral membrane, in LatA-treated monolayers. Occludin internalization continued as TER fell and was accompanied by progressive increases in numbers of intracellular occludin-containing vesicles. Loss of TER and mRFP1-occludin redistribution were accompanied by EGFP-β-actin reorganization, including appearance of EGFP-β-actin aggregates, and changes in cell shape (Figure 3C and Supplemental Movie S2). At later times after LatA addition, residual TJ-associated mRFP1-occludin was only present in dot like structures that also contained EGFP-β-actin. Thus, endocytic removal from the TJ is the primary morphological change in occludin distribution that correlates with LatA-induced loss of barrier function.

Similar studies were performed using monolayers expressing EGFP-β-actin and mRFP1-ZO-1 (Figure 4A and Supplemental Movie S4). LatA-induced redistribution of EGFP-β-actin in these monolayers was similar to that in monolayers expressing EGFP-β-actin and mRFP1-occludin, thus providing an internal control for comparisons. Although mRFP1-ZO-1 fluorescent intensity did vary along the TJ during first 20 min after LatA addition, this variability was similar to that in control monolayers without LatA addition. As in the fixed immunostained preparations, significant mRFP1-ZO-1 redistribution was only detected >20 min after LatA addition, long after the majority of TER losses had occurred. At these times, mRFP1-ZO-1 was lost from large areas of the TJ. Like mRFP1-occludin, the mRFP1-ZO-1 that remained at the TJ colocalized with the EGFP-β-actin. However, in contrast to mRFP1-occludin, new intracellular pools of mRFP1-ZO-1 were not identified. Thus, the mRFP1-ZO-1 redistribution detected occurs after the bulk of TER loss and is not associated with occludin internalization.

The effect of LatA on claudin-1 distribution was studied using MDCK cells stably transfected with EGFP-claudin-1 and mRFP1-occludin (Figure 4B and Supplemental Movie S5). As in cells expressing EGFP-β-actin and mRFP1-occludin, LatA induced redistribution and internalization of mRFP1-occludin. In contrast, EGFP-claudin-1 was not removed from the TJ during the first 12 minutes after LatA addition. At later time points, the distribution of EGFP-claudin-1 became discontinuous at the TJ. At these times EGFP-claudin-1 at the TJ was present primarily in areas containing mRFP1-occludin. Although claudin internalization has been reported in steady-state epithelial monolayers (Matsuda et al., 2004), in response to calcium depletion (Ivanov et al., 2004), and after cytokine treatment (Wang et al., 2005), increased numbers of EGFP-claudin-1-containing vesicles were not identified after LatA treatment. Moreover, the newly formed mRFP1-occludin intracellular pools did not colocalize with intracellular foci of EGFP-claudin-1 that were present with or without LatA treatment. Thus, it seems that the EGFP-claudin-1 redistribution observed represents a late event that occurs after the bulk of TER loss and is not associated with occludin internalization. Although these data cannot exclude changes in claudin-1 distribution or biochemical interactions that may accompany LatA-induced loss of barrier function, they do show that changes in claudin-1 distribution detectable by fluorescence microscopy do not accompany such TER loss.

One recent study has suggested that immature adherens junctions can be disassembled after latrunculin treatment (Ivanov et al., 2005). To determine whether the loss of TER after LatA treatment could be due to loss of adherens junctions, E-cadherin and occludin were imaged simultaneously in immunostained monolayers. At 20 min after LatA addition, when TER loss was complete, occludin was largely removed from the TJ, whereas the distribution of E-cadherin was unaffected (Figure 4C). Thus, LatA treatment does not cause gross disassembly of adherens junctions in these monolayers under the conditions studied here.

Disruption of Epithelial Barrier Function by LatA Requires Membrane Traffic

The data show that, along with decreased F-actin content, the most prominent early change that accompanies LatA-induced loss of barrier function is the endocytic removal of occludin from the TJ. To test the role of endocytosis in LatA-induced TER loss, we blocked membrane traffic by reducing temperature. Reducing temperature from 32 to 18°C did not affect LatA-induced loss of barrier function. In contrast, LatA-induced TER loss was completely prevented at ≤14°C (Figure 5A). Similar results were obtained with nontransfected and transfected MDCK monolayers, verifying that this was not an artifact of fusion protein expression. The protective effect was also not due to inhibition of actin depolymerization at 14°C (Figure 5A). This abrupt blockade of LatA-induced TER loss at 14°C suggests that the protective effect is due to inhibition of membrane traffic, rather than effects on protein–protein interactions or enzymatic activity. To determine whether cooling to 14°C also prevented LatA-induced internalization of occludin, we studied monolayers expressing EGFP-β-actin and mRFP1-occludin. At 14°C, LatA addition did not cause redistribution or endocytosis of mRFP1-occludin (Figure 5B and Supplemental Movie S6). Thus, the LatA-dependent actin depolymerization is dissociated from effects of LatA on TER and occludin internalization at 14°C. These data suggest that endocytosis may be an important step in LatA-induced disruption of TJ barrier function.

To further test the dependence of LatA-induced TER loss on endocytosis, vesicular traffic was inhibited at 37°C using hypertonic media. This treatment inhibits multiple pathways of vesicular traffic, including transport from the trans-Golgi network to the plasma membrane, endocytosis, and retrograde transport of plasma membrane proteins into the Golgi complex (Docherty and Snider, 1991). Incubation of monolayers in media made hypertonic with 0.4 M sucrose caused TER to fall in the absence of LatA. However, no further loss of TER was induced by addition of LatA (Figure 5C), consistent with a requirement for vesicular transport in LatA-induced barrier dysfunction.

To perform quantitative analysis of the effects of lowered temperature and hypertonic buffer on occludin internalization, we sought to quantify intracellular occludin pools after LatA treatment in fixed preparations. However, the number of occludin-containing intracellular vesicles detected in paraformaldehyde-fixed monolayers, either using anti-occludin immunolabeling or mRFP1-occludin fluorescence, was always less than the number detected by mRFP1-occludin fluorescence in live cells. We therefore considered the possibility that intracellular vesicles were damaged during paraformaldehyde fixation and subsequent processing. A protocol designed to prevent such damage by fixation in –20°C methanol followed by protein cross-linking (Hammond and Glick, 2000) successfully preserved intracellular occludin vesicles during fixation and subsequent immunostaining. This fixation protocol was used to assess intracellular occludin pools induced by LatA treatment in HBSS at 37 or 14°C or at 37°C in hypertonic buffer (Figure 5D). At 37°C, LatA treatment induced the accumulation of intracellular occludin-containing vesicles in fixed cells in a manner identical to that seen during live cell imaging, with an increase from 28 ± 1.2 to 43 ± 1.6 vesicles per cell (Figure 5, D and E; p < 0.01). In monolayers cooled to 14°C, LatA did not cause an increase in occludin-containing intracellular vesicles (Figure 5, D and E). Similarly, LatA did not cause occludin internalization in monolayers incubated in hypertonic buffer (Figure 5, D and E). Thus, LatA-induced TER loss and occludin internalization were both blocked when membrane traffic was inhibited at 14°C or with hypertonic media. These data show that LatA-induced TJ disruption requires membrane traffic that correlates with occludin internalization.

Occludin Is Internalized by a Dynamin II-dependent Process

To further understand the process by which occludin is internalized, potential pathways of occludin endocytosis were evaluated. Dynamin II plays an important role in caveolae- and clathrin-mediated endocytosis from basolateral membranes (van der Bliek et al., 1993; Altschuler et al., 1998; Oh et al., 1998). In control monolayers, 6.8 ± 0.6 vesicles per cell labeled for both occludin and dynamin II. This represented 22 ± 1.4% of the 31 ± 1.7 occludin-containing vesicles present in each cell. LatA treatment increased the number of occludin-containing vesicles that also contained dynamin II to 16 ± 0.9 per cell (p < 0.01), or 36 ± 1.3% of all occludin-containing vesicles (Figure 6, A and B). Dynamin II localization at the TJ also was occasionally seen in LatA-treated cells, but not in the absence of LatA treatment. These data suggest that dynamin II may be involved in the occludin endocytosis that occurs after LatA treatment.

To directly test the role of dynamin II in occludin endocytosis, MDCK cells were transiently transfected with dynamin II-EGFP or dynamin II K44A-EGFP. The latter is a dominant negative dynamin II that lacks GTPase activity and blocks both caveolae- and clathrin-mediated endocytosis (van der Bliek et al., 1993; Oh et al., 1998). In cells expressing dynamin II-EGFP, occludin was delivered to the TJ normally (Figure 6C), and LatA treatment caused the number of intracellular occludin vesicles to increase from 27 ± 2.4 to 45 ± 2.2 per cell (Figure 6D; p < 0.01). Monolayers expressing dynamin II K44A-EGFP had a reduced baseline TER that was 71 ± 3% of those expressing dynamin II-EGFP. Dynamin II K44A-EGFP-expressing cells displayed discontinuous occludin staining at the TJ (Figure 6E), with decreased intensity of TJ occludin staining and increased occludin-containing intracellular vesicles to 48 ± 2.5 per cell in the absence of LatA treatment (Figure 6, D and E). These data suggest that dynamin II may be important for basal delivery and recycling of occludin. When cells expressing dynamin II K44A-EGFP were treated with LatA there was no significant increase in the number of occludin-containing intracellular vesicles (Figure 6, D and E). In contrast, expression of dominant negative dynamin I was unable to prevent occludin internalization (our unpublished observations). Because dominant negative dynamin I inhibits apical but not basolateral endocytic events in polarized epithelia (Altschuler et al., 1998), this observation is consistent with the apparent origin of occludin-containing vesicles from the lower portion of the TJ adjacent to the border with the lateral membrane. Thus, these data demonstrate that LatA-induced occludin endocytosis occurs by a process that requires dynamin II function.

Occludin Is Internalized into Caveolin-1- but Not Clathrin-containing Vesicles

Dynamin II participates in both caveolae- and clathrin-mediated endocytosis (van der Bliek et al., 1993; Oh et al., 1998). Intracellular occludin has been identified in association with both caveolin-1 (Nusrat et al., 2000b) and clathrin (Ivanov et al., 2004). Thus, although the data show that dynamin II-mediated endocytosis is involved in LatA-induced occludin internalization and TER loss, they do not differentiate clathrin- and caveolae-mediated endocytic pathways. To evaluate the association of clathrin or caveolin with occludin during LatA-induced internalization, monolayers expressing mRFP1-occludin and either EYFP-clathrin light chain A1 or caveolin-1-EGFP were analyzed. mRFP1-occludin internalization was frequently observed in EYFP-clathrin light chain A1-expressing cells, but clathrin light chain A1 never colocalized with mRFP1-occludin during internalization (Figure 7A). In contrast, caveolin-1-EGFP colocalized with mRFP1-occludin in surface membrane invaginations and detaching vesicles in nearly all observed instances of LatA-induced occludin internalization (Figure 7B and Supplemental Movie S7). At later time points, some mRFP1-occludin could be seen to segregate away from caveolin-1-EGFP (Figure 7C and Supplemental Movie S8). Initially, a rim of occludin devoid of caveolin-1 was noticeable. Later, a separate vesicle containing occludin, but not caveolin-1, seemed to detach from the initial vesicle, resulting in two vesicles; one containing mRFP1-occludin and caveolin-1-EGFP and a second containing only mRFP1-occludin (Figure 7C).

To map these pathways quantitatively, it was necessary to synchronize occludin endocytosis. To accomplish this, we took advantage of the observation that occludin endocytosis did not occur at 14°C, although actin was effectively depolymerized at this temperature. Thus, monolayers were treated with LatA for 30 min at 14°C before being acutely warmed to 37°C. Before warming only 13% of occludin-containing vesicles were positive for caveolin-1 (Figures 8A and 9). Within 10 min after warming, this number increased to 55% (p < 0.001). However, this increase was transient, because 10 min later, at 20 min after warming, only 37% of occludin-containing vesicles were positive for caveolin-1 (Figures 8A and 9). These data are consistent with trafficking of internalized occludin through a caveolin-1-containing compartment.

A similar approach was used to evaluate the potential role of clathrin-mediated occludin endocytosis. In contrast to caveolin, LatA induced only a small increase in the number of occludin-containing vesicles positive for clathrin heavy chain, and this number was similar at both 10 and 20 min after warming (Figures 8B and 9). Trafficking of occludin to early and late endosomes also was assessed using EEA1 (Figure 8C) and LAMP-1 (Figure 8D), respectively, as markers of these compartments. In each case, increased labeling of occludin-positive vesicles was only apparent 20 min after warming (Figure 9), at which time colocalization of caveolin-1 and occludin was decreasing. These data suggest that clathrin is not involved in the initial internalization of occludin and that this initial internalization does not involve EEA1-positive early endosomes. The observed separation of occludin from caveolin-1 and small increases in colocalization of occludin with EEA1 and LAMP-1 at later times raises the possibility that, like cholera toxin (Pelkmans et al., 2004), a fraction of occludin internalized through caveolae can be subsequently directed to other endocytic compartments.

LatA-induced Barrier Dysfunction Is Not Prevented by Inhibition of Clathrin-mediated Endocytosis or Macropinocytosis

To functionally define the role of clathrin-mediated endocytosis and macropinocytosis in LatA-induced TER loss, we assessed the effect of inhibitors of these pathways. Macropinocytosis was prevented with amiloride, which inhibits Na⁺/H⁺ exchange and blocks macropinocytosis without affecting clathrin-mediated endocytosis (West et al., 1989). LY294002 inhibits phosphoinositide 3-kinase and blocks maintenance, maturation, and translocation of macropinocytic vesicles (Araki et al., 1996; Murray et al., 2000). Neither of these drugs prevented LatA-induced barrier dysfunction (Figure 10A). Thus, although each of these drugs has additional effects unrelated to macropinocytosis, the failure of both to block LatA-induced barrier loss suggests that macropinocytosis does not play an important role in this process.

Chlorpromazine inhibits assembly of the clathrin adapter protein AP2, thereby blocking clathrin-mediated endocytosis (Petris et al., 2003). Although chlorpromazine increased TER to 175 ± 6.6% of control monolayers in the absence of LatA, it did not prevent LatA-induced TER loss (Figure 10B), suggesting that clathrin-mediated endocytosis is not involved in LatA-induced TJ disruption. Moreover, specific inhibition of clathrin-mediated endocytosis by expression of EGFP-Eps15Δ95-295 (Benmerah et al., 1999; Nichols et al., 2001) failed to block LatA-induced occludin internalization (Figure 10C). Thus, both pharmacological and dominant negative inhibition of clathrin-mediated endocytosis failed to prevent LatA-induced TER loss and occludin internalization, suggesting that this pathway is not essential for LatA-induced TJ disruption.

LatA-induced Barrier Dysfunction Is Prevented by Cholesterol Extraction

The data suggest that dynamin II-dependent occludin internalization is responsible for LatA-induced TER loss. The morphological and morphometric data suggest that this occludin internalization may occur via caveolae-mediated endocytosis. Caveolae-mediated endocytosis depends on the specialized lipid composition of caveolae. To functionally define the role of caveolae-mediated endocytosis in LatA-induced TER loss, we extracted membrane cholesterol using either MBCD or filipin III, both of which have been shown to block caveolae-mediated endocytosis (Pike and Casey, 2002; Shigematsu et al., 2003). MBCD and filipin III caused TER to decrease by 12 ± 0.6 and 25 ± 0.4%, respectively, consistent with a previous report (Francis et al., 1999). However, further TER decreases were not induced by LatA in MBCD- or filipin III-treated monolayers (Figure 11A). Repletion of cholesterol restored the sensitivity of barrier function to LatA treatment, confirming that the effect of MBCD was due to cholesterol extraction (Figure 11B). Morphological analysis confirmed that MBCD prevented LatA-induced occludin internalization and that cholesterol repletion restored LatA-induced occludin internalization (Figure 11C). Moreover, morphometric analysis showed that although MBCD treatment alone caused the number of occludin-containing vesicles to increase slightly, from 28 ± 1.2 to 33 ± 1.7 vesicles per cell, there was no further increase upon LatA addition (Figure 11D). Thus, disruption of caveolae-mediated endocytosis prevents both LatA-induced barrier dysfunction and occludin internalization. Therefore, the functional and morphological data both indicate a critical role for caveolae-mediated endocytosis of TJ components, including occludin, in LatA-induced barrier disruption.