ZO-1 Stabilizes the Tight Junction Solute Barrier through Coupling to the Perijunctional Cytoskeleton

Abstract

ZO-1 binds numerous transmembrane and cytoplasmic proteins and is required for assembly of both adherens and tight junctions, but its role in defining barrier properties of an established tight junction is unknown. We depleted ZO-1 in MDCK cells using siRNA methods and observed specific defects in the barrier for large solutes, even though flux through the small claudin pores was unaffected. This permeability increase was accompanied by morphological alterations and reorganization of apical actin and myosin. The permeability defect, and to a lesser extent morphological changes, could be rescued by reexpression of either full-length ZO-1 or an N-terminal construct containing the PDZ, SH3, and GUK domains. ZO-2 knockdown did not replicate either the permeability or morphological phenotypes seen in the ZO-1 knockdown, suggesting that ZO-1 and -2 are not functionally redundant for these functions. Wild-type and knockdown MDCK cells had differing physiological and morphological responses to pharmacologic interventions targeting myosin activity. Use of the ROCK inhibitor Y27632 or myosin inhibitor blebbistatin increased TER in wild-type cells, whereas ZO-1 knockdown monolayers were either unaffected or changed in the opposite direction; paracellular flux and myosin localization were also differentially affected. These studies are the first direct evidence that ZO-1 limits solute permeability in established tight junctions, perhaps by forming a stabilizing link between the barrier and perijunctional actomyosin.

INTRODUCTION

Tight junctions form size- and charge-selective barriers to the paracellular movement of solutes and ions across epithelia (Powell, 1981; Madara, 1998). More than 40 different cytoplasmic and transmembrane proteins are located at the junction (Gonzalez-Mariscal et al., 2003; Schneeberger and Lynch, 2004; Guillemot et al., 2008) yet there is limited knowledge of how most contribute to the junction's selective sealing properties. The cytoplasmic protein ZO-1 was the first tight junction protein to be identified (Stevenson et al., 1986). It is a member of the MAGUK family of membrane-associated scaffolding proteins (Funke et al., 2005) and contains protein-binding domains for all the major transmembrane barrier proteins: claudins (Furuse et al., 1998; Itoh et al., 1999), occludin (Furuse et al., 1994; Fanning et al., 1998), tricellulin (Ikenouchi et al., 2005; Riazuddin et al., 2006), JAM-A (Martin-Padura et al., 1998; Ebnet et al., 2000), and several components of the perijunctional network of actin and myosin (Tsukita et al., 2001; Hartsock and Nelson, 2008; Yamazaki et al., 2008). Although ZO-1 gene deletions are embryonic lethal in mice (Katsuno et al., 2008), the depletion of ZO-1 in cultured epithelial cell models (Umeda et al., 2004; McNeil et al., 2006) results in only a delay in barrier formation. Once tight junctions are formed in the ZO-1–depleted cells, TER and flux for large (40 kDa) solutes are reported to be essentially normal (Umeda et al., 2004). There is no direct evidence to date suggesting that ZO-1 normally limits solute permeation in an intact tight junction, despite a widely held view that ZO-1 is a functionally critical tight junction component.

Permeation of solutes through the intact tight junction can be described by two components. The first is transport through a system of ∼4-Å radius charge-selective claudin-based pores, most often quantified by transepithelial electrical resistance (TER; Powell, 1981); this is a nearly instantaneous electrical assessment (typically <0.25-several seconds) of ionic permeation through the tight junction. TER is determined by the pattern of claudins in the epithelium and their different electrostatic permeability characteristics (Van Itallie and Anderson, 2006; Angelow et al., 2008). The second component is a lower capacity pathway for solutes larger than ∼4 Å in radius, which shows no selectivity for ionic charge or size (Adson et al., 1994; Knipp et al., 1997). Although the structural and mechanistic basis of this pathway is unknown, one speculation is that it represents dynamic breaks and resealing in the junction's cell–cell contacts (Sasaki et al., 2003; Watson et al., 2005). This pathway must be measured over longer time periods (typically 60–180 min) by flux of tracers that are larger than the pores, often mannitol (4.2 Å radius; Schultz and Solomon, 1961) or much larger fluorescent dextrans. A more continuous assessment of permeability as a function of molecular radius can be made by determining the permeability for a graded series of noncharged polyethylene glycol oligomers (Watson et al., 2001; Van Itallie et al., 2008); this methods allows simultaneous quantification of flux through both the small claudin-based pores and the low-capacity break pathway.

Although measurement of TER and solute flux are often combined to describe a monolayer's “permeability,” they represent different structural and functional characteristics of the barrier, as is evidenced by observations that they do not necessarily change in parallel (Balda et al., 1996; McCarthy et al., 2000). The distinction between these two pathways is important. The first is the physiologically relevant pathway for transepithelial salt and water transport; its charge selectivity, electrical resistance, and level of permeability influences overall transepithelial transport (Diamond, 1977; Powell, 1981). The second pathway is enhanced by pathological insults allowing transepithelial movement of larger material such as antigens, proinflammatory bacterial fragments or even bacteria and viruses (Clayburgh et al., 2004; Mankertz and Schulzke, 2007). Thus, elucidation of the molecular basis for this second pathway is critical to our understanding of barrier disruption in pathological situations.

There is considerable evidence that perijunctional actin and the activity of myosin influence the second, suprapore, pathway (Turner et al., 1997; Madara, 1998). For example, flux is enhanced by disruption of F-actin (Bentzel et al., 1980; Shen and Turner, 2005) or by activation of nonmuscle myosin 2 ATPase activity by myosin light-chain kinase (MLCK; Zolotarevsky et al., 2002; Wang et al., 2005; Su et al., 2009) downstream in the RhoA and Rho kinase (ROCK) signaling pathways (Samarin and Nusrat, 2009). Significantly, ZO-1 binds to several myosin-associated proteins, including cingulin (Cordenonsi et al., 1999) and Shroom2 (Etournay et al., 2007), to multiple proteins that regulate actin dynamics (Hartsock and Nelson, 2008) and to F-actin itself (Fanning et al., 2002). Furthermore, ZO-1 is one of the few cytoplasmic proteins known to bind directly to the transmembrane barrier proteins of the tight junction, a short list that also includes the multi-PDZ protein MUPP1 (Hamazaki et al., 2002) and the homologues ZO-2 (Fanning et al., 1998; Itoh et al., 1999) and ZO-3 (Haskins et al., 1998; Itoh et al., 1999). Thus, ZO-1 is an ideal candidate to provide a crucial link between perijunctional actomyosin activity and the transmembrane barrier proteins.

To address the role of ZO-1 in the intact tight junction we generated stable Madin-Darby canine kidney (MDCK) cell lines expressing ZO-1 short hairpin RNAs (shRNAs) and assessed the tight junction barrier with a variety of assays and after cytoskeletal provocations. ZO-1 depletion resulted in an increase in the permeability specifically for solutes larger than ∼4 Å and in morphological changes and reorganization of actin. The barrier in knockdown cells also showed responses different from those in wild-type cells to manipulations affecting actin polymerization and myosin activity. We conclude that ZO-1 is a critical link that stabilizes the barrier for larger solutes and couples it to the myosin–actin cytoskeleton.

MATERIALS AND METHODS

Cell Culture, Immunoblots, Immunofluorescence

Tet-off MDCK II cells (Clontech, Mountain View, CA) were cultured under standard conditions in DMEM (4.5 g/l glucose), 10% fetal bovine serum, and pen-strep. To produce ZO-1 and ZO-2 knockdown cell lines, tet-off MDCK II cells were cotransfected with pSVZeo (Invitrogen, Carlsbad, CA) and a mixture of three different ZO-1 or -2 shRNAs (Supplemental Figure S1) cloned into the BglII/HindIII sites of the pSuper vector (Oligoengine, Seattle, WA). All transfections were performed with Amaxa nucleofector technology (Lonza, Cologne, Germany) using the manufacturer's recommended conditions for MDCK II cells. Stable cell lines were selected in 1 mg/ml zeocin (InvivoGen, San Diego, CA). Antibiotic-resistant clonal cell lines were screened by immunoblotting and immunofluorescence; three separate clones were selected from each of two separate transfections. ZO-1 rescue constructs (see Figure 4A) were cloned into pTRE vector (BD Biosciences, San Jose, CA) and cotransfected into knockdown cells with pTK hyg (BD Biosciences) or pBlast49 (InvivoGen). Antibiotic-resistant clonal cell lines were selected with 200 μg/ml hygromycin (Roche, Nutley, NJ) or 10 μg/ml blastocidin (InvivoGen) and screened as above; at least three independent clones inducibly expressing each of the three rescue constructs were used in further analyses.

Figure 4.

Reexpression of full-length human ZO-1 and -1 fragments restores P_app to control levels. (A) Diagram of rescue constructs. Top, ZO1FL full-length human ZO-1; ΔABR deletes only the direct actin-binding region (ΔAA1183-1663; Fanning et al., 2002), Nterm contains amino acids 1-888. The ZO-1 KD cells were created in the MDCK Tet-off cell line, and all rescue constructs were inducible by removal of doxycycline. (B) Expression of FLZO1 in MDCK tet-off cells results in increased ZO-1 levels (135% of control); ZO-1 KD reduces endogenous ZO-1 to 3–5% of control levels. Induction of ZO1FL in KD cells increases ZO-1 levels 3–4-fold over KD values, but only to 15% of control levels. No treatment altered ZO-2 or occludin (ocln) protein levels. (C) Immunofluorescence localization of ZO-1 in control (uninduced, top left panel), MDCK II cells overexpressing human ZO-1 (induced, top right), KD cells (uninduced, bottom left and KD cells induced (bottom right) to express human ZO-1. All constructs show similar tight junction localization. (D) PEG profiling of KD and rescue cell lines. MDCK cells induced (I) or not (U) to express exogenous ZO-1 and KD cells induced (I) to express human ZO-1 all show indistinguishable flux profiles. Only KD cells not induced to express the full-length rescue construct have consistently higher permeability (p < 0.05 compared with controls, triplicate filters for each clone per experiment, three separate experiments). (E) ZO-1 rescue fragments (green) are inducible and do not alter ZO-2 levels (red). Endogenous ZO-1 expression levels are not affected by doxycycline induction in either MDCK II or KD cells. When KD cells were stably transfected with rescue constructs as above, transgene expression was not detectable in the uninduced cells (U) but was clearly induced by doxycycline removal (I). ZO-2 levels were not affected by transgene induction. (F) Immunofluorescence localization of induced transgenes demonstrates localization to apical junctional complex and colocalization with occludin. The anti-human guinea pig polyclonal antibody used in this experiment also recognized the central cilium; this is not seen with other occludin antibodies. (G) P_app of 3-kDa FD reveals significantly increased flux (p < 0.05) in all KD cells lines in the absence of expression of recue constructs; induction of all rescue constructs (one FLZO1 clone measured ± dox, three independent clones for ΔABR and Nterm measured ± dox on triplicate filters) returns P_app for FD to levels indistinguishable from control values. (H) Induction of all rescue constructs significantly decreases TER values in all KD cell lines (as above, p < 0.001).

For all experiments unless otherwise indicated cells were cultured for 7–10 d on Transwell permeable supports (0.4-μm polyester membrane, 12-mm inserts, Corning, Corning, NY). Immunoblotting and immunofluorescence microscopy of the cultured cells were performed as described previously (Colegio et al., 2002; Van Itallie et al., 2003). Unless otherwise noted, all antibodies were from Invitrogen, Carlsbad, CA). The anti-ZO-1 rat monoclonal was a gift from Dr. Bruce Stevenson (Scripps, La Jolla, CA; Stevenson et al., 1986); the guinea pig anti-human occludin antibody used in Figure 3E was from Hycult Biotechnology (Uden, Netherlands); the myosin 2B (myo2B) antibody was purchased from Covance (Emeryville, CA); the cingulin antibody was a gift from Dr. Sandra Citi (University of Geneva, Switzerland). The anti-human guinea pig polyclonal antibody used in this experiment also recognized the central cilium; this is not seen with other occludin antibodies. Filamentous actin was visualized using rhodamine-phalloidin (Molecular Probes, Invitrogen, Eugene, OR). Affinity-purified Cy-2–, -3–, and -5–labeled secondary antibodies for immunofluorescence were from Jackson ImmunoResearch (West Grove, PA), and IR680/700 and 800 labeled secondary antibodies for immunoblotting were from Invitrogen and Rockland Immunochemicals (Gilbertsville, PA).

Figure 3.

Stable ZO-2 KD in MDCK II cells changes neither cell morphology nor solute permeability. (A) Immunoblot analysis of MDCK control cells and three independent stable ZO-2 KD demonstrate a large decrease in ZO-2 levels (>97%). ZO-2 KD has no effect on ZO-1 or occludin levels. (B) Unlike ZO-1 KDs, ZO-2 KD cells exhibit similar convoluted apical junctions as seen in MDCK control cells. (C) immunofluorescence localization for myosin 2B (myo2B) and actin are similar in MDCK and ZO-2 KD cells, but altered in ZO-1 KD cell lines (see rightmost panel, 4× enlargement of myo2B staining). (D) ZO-2 KD has no effect on P_app for any PEG sizes (three independent clones, measured on triplicate filters), nor does it alter TER measurements (not shown).

Wide-field images were acquired on a Nikon E800 microscope (Melville, NY) using 60× or 100× Plan Apo lenses and an Orca ER cooled CCD camera controlled with the Metamorph Imaging software package (ver. 6.0; Universal Imaging, West Chester, PA). Filter sets and dye combinations have been previously described (Fanning et al., 2002). Confocal images were acquired on a Zeiss LSM510 Meta using a 63× Plan Apo lens (Thornwood, NY). Confocal Stacks and image projections were generated with Zeiss LSM Image Browser (ver. 3.2). All confocal images are maximum density projections of three image planes representing 1.05-μm final depth. Contrast adjustment and montages were generated using Adobe Photoshop (ver. 7.0; San Jose, CA).

Barrier Assays: Polyethylene Glycol, FITC-Dextran and Mannitol Permeability, Transepithelial Electrical Resistance, and Dilution Potentials

Polyethylene glycol (PEG) profiling to determine the size dependence of permeability has been described recently (Van Itallie et al., 2008); to correct the P_app of the size-dependent first phase, the second phase is extended after linear regression and subtracted from the first phase (Van Itallie et al., 2008). Determination of [³H]mannitol flux and dilution potentials were performed as described previously (Van Itallie et al., 2001). Dilution potential experiments were performed on cells plated on removable permeable supports (Snapwell, Corning). Transepithelial electrical resistance was determined using plate electrodes placed on either side of the monolayer attached to an EVOM (WPI, Sarasota, FL). FITC-dextran permeability was measured after preincubation of monolayers in Hanks' buffered salt solution (HBSS) containing CaCl₂ and MgSO₄; fluorescein-conjugated dextran (FD), 3 or 10 kDa (Molecular Probes, Invitrogen), was added to the apical compartment at a concentration of 1 mg/ml, and samples were removed from the basolateral compartment at the indicated times. Dextran flux was identical in the apical-to-basal and basal-to-apical directions over the time course of the experiment, suggesting negligible contribution from transcytosis. Fluorescent-dextran concentrations were quantified using a Typhoon 8600 and Image Quant software (Amersham Pharmacia Biotech, Piscataway, NJ); experimental values were determined by extrapolation from a standard curve of known fluorescent-dextran concentrations using linear regression (GraphPad Prism, San Diego, CA). Apparent Permeabilities (P_app) were defined as (dQ/dt)/AC_o (Van Itallie et al., 2008). Statistical analysis (ANOVA, Dunett's and t tests) were performed using GraphPad software.

Pharmacologic Reagents

Cytochalasin D and InSolution Y-27632 were purchased from Calbiochem (EMD Biosciences, San Diego, CA); S-(−)-blebbistatin was purchased from Toronto Research Chemicals (North York, ON, Canada).

RESULTS

ZO-1 Knockdown in MDCK Monolayers Changes Cell Contact Morphology and the Organization of Apical Actin

Previous studies demonstrated the requirement for ZO-1 for normal assembly kinetics of the tight junction (Umeda et al., 2004; McNeil et al., 2006). We used a similar knockdown approach in MDCK cells, but asked whether depletion of ZO-1 had an effect on barrier properties at steady state, i.e., after the junction had sealed. We stably expressed specific shRNAs in the pSuper vector and isolated stable clonal cell lines on the basis of zeocin resistance. Of these, six clones (e.g., see Figure 1A, KD clones 1–3) showed significant reduction of ZO-1 protein levels ranging from 95 to 97% when compared with controls and were chosen for further analysis. The levels of other tight and adherens junction proteins were not consistently affected by knockdown of ZO-1, including ZO-2, occludin, and claudin-2 (Figure 1A). There were also no consistent differences between control and knockdown cells in the expression levels of claudins 1, 3, 4, 7, actin, tricellulin, cadherin, or β-catenin (not shown).

Figure 1.

Immunoblot and immunofluorescent characterization of ZO-1 knockdown (KD) MDCK monolayers. (A) Immunoblots of ZO-1 and -2, occludin (ocln), and claudin-2 (cldn2) in control MDCK monolayers and in three independent stable KD clones (1, 2, and 3). KD lines show a 95–97% reduction in ZO-1 levels as quantified by densitometry. KD does not alter the levels of the other proteins. (B) Coimmunolocalization of ZO-1 and claudin-2, and ZO-1 and actin in MDCK control monolayers (top panels) show normally convoluted apical junctional staining. Similar colocalization in KD cells (bottom panels) demonstrates straighter cell junctions and striking redistribution of actin localization. Bar, 10 μm.

Immunofluorescence microscopic analysis of ZO-1 knockdown cells revealed a striking alteration in cell contact morphology compared with controls. Confluent MDCK II cells typically have convoluted cell–cell contacts at the apical junctional region, as revealed by localization of any of the tight junction proteins, such as ZO-1 or claudin-2 (Figure 1B). Actin is distributed uniformly throughout the cell cortex of MDCK cells and notably is concentrated in basal stress fibers (not shown). In addition, a thin band of cortical actin staining colocalizes with ZO-1 and other cell junction proteins at the apical junctional complex (Figure 1B). When ZO-1 levels are reduced, there is a conversion from tortuous into linear contacts as revealed by the immunolocalization of claudin-2 and the residual ZO-1 (Figure 1B). More dramatically, actin staining is now enriched at both the apical junctional complex and in intense apical dots (Figure 1B). All knockdown cell lines showed this altered junctional morphology and actin reorganization. A conversion from convoluted to straight cell contacts was previously reported after overexpression of Shroom2, a ZO-1–binding protein that also binds actin and myosin VIIIa (Hildebrand, 2005).

Depletion of ZO-1 Destabilizes the Barrier for Large Solutes But Does Not Disrupt the Claudin-based Pores

Solute permeation through the tight junction can be described by two pathways; solutes and ions smaller than ∼4 Å radius pass through claudin-based pores; although the physical pathway for larger solutes is unclear, it may represent transient breaks in the barrier (Sasaki et al., 2003; Watson et al., 2005). To determine the role of ZO-1 in regulating these two components of paracellular permeability, we profiled the apparent permeability (P_app) of a graded series of PEG oligomers in control and ZO-1 knockdown monolayers. This method allows us to determine simultaneous changes in P_app at different molecular radii (Van Itallie et al., 2008). Comparison of MDCK parental cell lines to four separate knockdown clones demonstrated no difference in P_app for PEG oligomers below ∼4 Å but about a twofold increase in P_app for PEG oligomers larger than ∼4 Å (Figure 2A). This twofold increase in P_app was confirmed by two other flux markers (Figure 2B); [³H]mannitol, which has a hydrated radius of 4.2 Å (Schultz and Solomon, 1961) similar to the hydrated radius of PEG8 (4.3 Å) and 3-kDa FD, which has a molecular radius of ∼14Å (Kondoh et al., 2005). Both tracers show the same twofold increase in permeability in the ZO-1 knockdown cells when compared with controls, even though the absolute value of P_app for the FD is only 3% that of mannitol. This lower baseline flux of the 3-kDa FD is likely a function of its lower diffusion coefficient. The increased flux of larger compounds in ZO-1 knockdowns is unlikely to represent the formation of a new size-restrictive pore, because there is no evidence for a second inflection point in the permeability data. Instead there is a proportional increase in the flux of all solutes that are larger than the claudin pores, at least for molecules up to 14 Å in radius. When flux of a much larger 10-kDa FD was measured, there was a small but not statistically significant increase in flux in knockdown cells compared with control cells (not shown).

Figure 2.

ZO-1 KD cell lines show increased permeability for solutes larger than 3.7Å. (A) P_app as a function of PEG radius in MDCK II cells reveals no significant change in flux for small PEGs (3.2- and 3.5-Å radius) but a significant increase in flux for all PEGs 3.7–7.0 Å (n = 4 independent KD cell lines, measured on triplicate filters, p < 0.01). (B) Significant (p < 0.01) increase in flux for larger PEGs was verified by comparison with other methods. P_app for PEG with 4.3-Å radius was identical to flux determined using [³H]mannitol (4.2-Å hydrated radius; Schultz and Solomon, 1961), whereas there was a similar increase in P_app using 3-kDa FD, both measured using three independent clones, on triplicate filters. (C) In spite of increased flux there was a slight but significant increase in TER in KD cell lines (n = 6, three filters each); expression of full-length human ZO-1 (FLZO1) in KD cell line (measure in triplicate) resulted in TER indistinguishable from control.

The lack of effect on the flux of molecules through the claudin-based pores suggests that depletion of ZO-1 does not change the biochemical components of this barrier; this is consistent with the lack of effect on the levels and subcellular localization of the tight junction barrier proteins. It is further supported by the finding that ZO-1 knockdown results in a slight but significant increase in TER (Figure 2C), which is the opposite of what would be expected if pores were disrupted. The additional finding that there is no change in the paracellular dilution potential (not shown) is also consistent with the lack of change in the profile of claudins lining the pores in the knockdown cells. The apparent increase in P_app in ZO-1–depleted cells for PEGs, which are <4 Å, Figure 2A, does not represent a real increase in permeability of the pores. The smallest PEGs can also pass through the leakier large pathway, so that subtraction of the contribution of the leak pathway from the values for the small-pore pathway reveals that the P_app for the smallest PEGs is identical between wild-type and knockdown cells.

ZO-2 Depletion Does Not Replicate the Morphological and Barrier Phenotypes of ZO-1 Depletion

ZO-2 binds to ZO-1, colocalizes at the junction, and shares several key protein-binding domains (Roh and Margolis, 2003; Utepbergenov et al., 2006), suggesting it, like ZO-1, might also regulate solute flux. To test this hypothesis, we used shRNA methods to generate stable clonal MDCK cells lines depleted of ZO-2 (Figure 3). Similar to what was reported in Eph4 cell ZO-2 knockdown cells (Umeda et al., 2006), in MDCK II cells stable ZO-2 depletion did not alter the levels of ZO-1 or occludin (Figure 3A). Immunofluorescent analysis of ZO-2 knockdowns revealed that, unlike ZO-1 knockdowns, ZO-2–depleted cells still showed the convoluted cell–cell contact morphology characteristic of wild-type MDCK II cells (Figure 3B). Also unlike ZO-1 knockdowns, localization of apical myo2B and actin was comparable to that seen in wild-type MDCK cells (Figure 3C). The enlargement in the rightmost panel (Figure 3C, right) reveals that apical myo2B staining appears mostly continuous in both the wild-type and ZO-2 knockdown MDCK cells, but is more punctuate in the ZO-1 knockdowns. Analysis of paracellular solute permeability using PEG size profiling revealed no difference in the P_app for any PEG size either below or above the size of claudin pores (Figure 3D); in addition, ZO-2 knockdown did not affect TER (not shown). Thus, unlike ZO-1 the knockdown of ZO-2 had no effects on morphology or barrier physiology. The apparent cell size difference in Figure 3B between ZO-1 knockdown cells and controls or ZO-2 knockdowns was not a consistent finding.

ZO-1 or the N-Terminal Half of ZO-1 Rescues the Barrier Defects of ZO-1 Depletion

To determine if reexpression of full-length ZO-1 can rescue the barrier defect caused by depletion of ZO-1, we introduced the ZO-1 knockdown vectors into an MDCK cell line that had been previously engineered to express an inducible human ZO-1 (Fanning et al., 2007). Knockdown vectors were designed to interfere with expression of canine but not human ZO-1. The levels of endogenous canine ZO-1 in these rescue cell lines was significantly reduced (to <5% of that seen in the parental line), comparable to that seen in the knockdown lines described above. Induction of the recombinant full-length human ZO-1 (FLZO1, diagrammed in Figure 4A) in ZO-1 knockdown cell lines restored ZO-1 to levels that were ∼12–17% of those seen in the parental lines. Thus, although expression was lower than in wild-type MDCK cells, levels are induced at least threefold over the protein levels in knockdown cells (Figure 4B). Immunofluorescent analysis in ZO-1 knockdown and rescue cells demonstrated that FLZO1 is appropriately localized to tight junctions in induced cells (Figure 4C) in either the presence or absence of the endogenous canine ZO-1 (Figure 4C), although there is a small increase in cytosolic ZO-1 staining when the transgene is induced in ZO-1 knockdown cells. Notably, there is some increase in tortuosity of the junction contacts when FLZO1 is restored in knockdown cells.

Despite low expression levels, FLZO1 clearly rescues the enhanced permeability seen in ZO1 knockdown cells as revealed by PEG permeability profiling (Figure 4D). Similar to experiments described above, in the absence of FLZO1 expression (□, Figure 4D) the ZO1 knockdown lines demonstrated increased flux of PEG oligomers of 4 Å and above, whereas induction of FLZO1 in these cells restored P_app for the larger PEGs to control levels (■, Figure 4D). There was no significant difference in permeability for the PEG oligomers <4 Å in any of the knockdown or rescue cell lines. As expected, expression of FL ZO1 in the parental cell line (which already expresses endogenous ZO-1) had no discernable effect on P_app for the small or large PEG oligomers.

To determine critical regions of ZO-1 responsible for rescuing the barrier defect seen in knockdown cells, we transfected ZO-1 knockdown cells with tet-inducible human ZO-1 constructs missing either the actin-binding region (ΔABR, Figure 4A) or encoding only the N-terminal 888 amino acids (Nterm, Figure 4A). Immunoblot analysis demonstrates that both ΔABR and Nterm are efficiently induced and expressed at levels comparable to FLZO1, and neither endogenous ZO-1 nor knockdown ZO-1 levels are affected by the presence (U, uninduced) or absence (I, induced) of doxycycline (the suppressing agent). Furthermore, ZO-2 levels are unchanged by doxycycline treatment (Figure 4E). Immunofluorescent analysis of induced and uninduced cells demonstrates that both deletion constructs localize to the apical junctional complex (Figure 4F). Expression of the ΔABR causes some slight return of the convoluted tight junction morphology seen in control cells, but without a reliable method for quantification, the degree of this rescue is difficult to assess. There is no apparent rescue of the convoluted tight junction morphology with expression of the Nterm construct.

Permeability of 3-kDa FD was significantly increased in all of the MDCK rescue lines before induction of rescue transgenes (Figure 4G), although the level of increase was variable in different clonal cell lines. Nevertheless, induction of any of the transgenes (FL, ΔABR, or Nterm) completely reverses the P_app increase to control levels (Figure 4G). Similarly, all of the knockdown cell lines show increased TER relative to MDCK II controls (Figure 4H); in each case, induction of the rescue construct significantly reduces TER to levels similar to MDCK controls or below. These findings suggest that regions sufficient for physiological stabilization of the barrier lie within the amino terminal half of ZO-1.

ZO-1–depleted Cell Lines Demonstrate Increased Sensitivity to Both Cytochalasin D and Ca²⁺ Removal

ZO-1 has been demonstrated to bind F-actin directly (Fanning et al., 1998) and indirectly (reviewed in Hartsock and Nelson, 2008). The phenotype of MDCK knockdown monolayers is consistent with the possibility that ZO-1 is required to maintain the normal perijunctional actin cytoskeleton and stabilize the barrier. Because disruption of the actin cytoskeleton with cytochalasin D is known to disrupt tight junction barrier function (Bentzel et al., 1980; Stevenson and Begg, 1994; Shen and Turner, 2005), we asked whether ZO-1 knockdowns might demonstrate increased sensitivity to cytochalasin D. After culture for 7–10 d on filters, control MDCK II cells are remarkably resistant to cytochalasin D treatment as measured by TER (Figure 5A). In contrast, cytochalasin D treatment of ZO-1 knockdown cells results in a significant time-dependent drop in TER. These results are consistent with a role for ZO-1 in stabilizing normal actin organization required to maintain the barrier.

Figure 5.

ZO-1 KDs are more sensitive to cytochalasin D treatment and Ca²⁺ removal than are control MDCK cells. (A) MDCK and KD cells cultured for 7 d on Transwell filters were preincubated in HBSS containing Ca²⁺ and Mg^{2+ for} 20 min and TER was measured. Monolayers were then changed into fresh media containing either DMSO or 10 μg/ml cytochalasin D in DMSO. TER was measured on triplicate filters at the indicated times; * p < 0.05 or less, similar results were obtained with a second KD cell line. (B) MDCK and KD cells were cultured as above; TER was measured and cells were washed four times with Ca²⁺-free SMEM and then incubated in SMEM supplemented with 10% dialyzed (against EGTA) fetal bovine serum and 5 μM CaCl₂. TER was measured at the indicated times, measurements made on triplicate filters; * p < 0.05 or less; similar results were obtained with a second KD cell line.

We also tested the role of ZO-1 in stabilizing the barrier after Ca²⁺ removal. Transferring confluent MDCK II cell monolayers into low Ca²⁺ media (5 μM) results in a decrease in TER that reaches a stable level at 60 min (Figure 5B). Ca²⁺ removal from ZO-1 knockdown cells results in a significantly faster decrease in TER, reaching a new plateau level at 15 min. As previously reported by others, we also observed a delay in the reformation of tight junctions after Ca²⁺ replacement in the knockdown cells relative to control cells (not shown). Although it is not known precisely how Ca²⁺ removal leads to barrier loss, our results are again consistent with a role for ZO-1 in tight junction stabilization.

The Barrier Is Functionally Uncoupled from Myosin in ZO-1–depleted Monolayers

Many studies have demonstrated a role for myosin in the regulation of paracellular permeability (reviewed in Turner, 2006; Ivanov, 2008), but the functional links between the tight junction and actomyosin have not been established. To determine whether ZO-1 is such a link, we asked whether depletion of ZO-1 changes the sensitivity of the barrier to agents that alter myosin ATPase activity, namely the ROCK inhibitor Y27632 and myosin inhibitor blebbistatin.

ROCK is thought to activate nonmuscle myosin 2 ATPase through both direct phosphorylation of the regulatory light chain and by inhibition of myosin light chain (MLC) dephosphorylation (Amano et al., 1996; Kimura et al., 1996; Nakai et al., 1997). There is no single model for how myosin controls the barrier. For example, administration of the ROCK inhibitor Y27632 increases basal TER in confluent MDCK cells (Fujita et al., 2000) yet has the opposite effect in T84 cells (Walsh et al., 2001). These differing responses may represent different levels of myosin activation (Ivanov, 2008) or the activity of other downstream targets of ROCK (Riento and Ridley, 2003). With this caveat we tested whether depletion of ZO-1 altered coupling between myosin activity and the barrier. As previously reported, Y27632 increased TER in our wild-type MDCK II cells in a dose-dependent manner (Figure 6A). This increase was similar at 2, 6, and 24 h; the dose dependence at 24 h is shown in Figure 6A. In contrast, ZO-1–depleted monolayers were strikingly insensitive to Y27632 at any dose or time.

Figure 6.

MDCK II control and ZO-1 KD cells show differential sensitivity to the Rho kinase (ROCK) inhibitor Y27632. (A) Cells were cultured on filters for 7 d and then changed into media containing vehicle or the indicated doses of Y27632. TER was measured at 0, 2, 6, and 20 h (shown); there was little difference in effects on TER at any time point. MDCK control but not KD cells demonstrated a dose-dependent (triplicate filters, p < 0.001) increase in TER. (B) Cells treated with 30 μM Y27632 for 20 h as above were used for flux measurement; there was a significant increase in P_app for 3-kDa FD after treatment with ROCK inhibitor but no further elevation in flux in KDs (triplicate filters, * p < 0.05 compared with vehicle, ** p < 0.001 compared with vehicle-treated MDCK cells, and ^# no significant difference with vehicle-treated KD cells.) (C) Immunofluorescence analysis of vehicle and Y27632-treated MDCK and KD cells lines. Ethanol-fixed MDCK monolayers cultured on filters and treated with 30 μM Y27632 for 20 h as above were stained for ZO-1 and Myo2B; Y27632 treatment increased apical junctional immunolocalization of Myo2B in control cells but not in KDs, but increased apical spots in KD cells. Bar, 5 μm.

Solute permeability showed a similar insensitivity to Y27632. In control MDCK cells 24 h of treatment with Y27632 increased permeability for both 3-kDa FD (Figure 6B) and mannitol (not shown). In contrast, ZO-1 knockdown cells, which have higher basal dextran permeability, failed to show further permeability increases with Y27632 administration. The different sensitivity of these cells to Y27362 correlates with differential response in myosin immunolocalization in control and knockdown cells after Y27632 treatment. myo2B immunofluorescence at apical junctional contacts follows the convoluted cell contacts characteristic of wild-type MDCK cells (Figure 6C), whereas in the ZO-1–depleted cells myo2B follows the straighter cell borders that characterize the knockdowns, with a more punctuate, less continuous distribution (see Figure 3C, 4× magnified images). Treatment of wild-type cells with the ROCK inhibitors decreases the membrane tortuosity of wild-type cells and slightly increases the intensity of myo2B staining at apical cell borders. Administration of the ROCK inhibitor to ZO-1–depleted cells does not increase the intensity of myosin 2B at apical cell contacts, but dramatically increases the amount of myo2B concentrated in scattered apical spots.

Blebbistatin, unlike Y27632, is a direct inhibitor of myosin 2 activity; its administration to confluent MDCK II cell monolayers results in a dose-dependent increase in TER, although this increase is less pronounced than that seen after ROCK inhibition (Figure 7A). However, unlike the effect of the ROCK inhibitor on knockdown cells blebbistatin treatment causes a significant dose-dependent decrease in TER in the ZO-1–depleted cells (Figure 7A). Blebbistatin administration has no effect on permeability in the control cells, whereas in knockdown cells it further elevated 3-kDa FD flux (Figure 7B). In addition, blebbistatin administration resulted in a more pronounced relocalization of myo2B in knockdown cells when compared with control cells (Figure 7C). The reasons for the changes in morphology are unclear, but they are consistent with an influence of ZO-1 on myosin organization, perhaps through its interaction with actin. One important caveat is that only a fraction of the cellular actin and myosin are localized to the tight junction, and the ROCK inhibitor and blebbistatin have multiple cellular actions.

Figure 7.

MDCK II control and ZO-1 KD cells show differential sensitivity to the myosin ATPase inhibitor blebbistatin. (A) Cells cultured on filters for 7 d were washed once and preincubated for 20 min in HBSS containing Ca²⁺ and Mg²⁺. Filters were then changed into fresh HBSS with vehicle or blebbistatin at the indicated doses. TER was measured after 2 h of blebbistatin treatment. MDCK control demonstrated a dose-dependent (* p < 0.01) increase in TER; KD cells demonstrate a dose-dependent (^# p < 0.05) decrease in TER (triplicate filters). (B) Cells treated with 50 μM blebbistatin or vehicle were used for determination of FD flux; there was no change in MDCK control cells with blebbistatin, but blebbistatin treatment resulted in a further significant elevation of P_app for 3-kDa FD (* p < 0.05 compared with controls, ** p < 0.001 compared with vehicle-treated KD cells, triplicate filters). (C) Immunofluorescence analysis of vehicle and blebbistatin-treated MDCK and KD cells lines. Ethanol-fixed MDCK monolayers cultured on filters and treated with vehicle 50 μM blebbistatin for 2 h as above were stained for ZO-1 and myo2B; blebbistatin treatment resulted in greater disorganization of myo2B staining in KD cells than in control cells. Bar, 5 μm.

DISCUSSION

In the current study we demonstrate that ZO-1 depletion in MDCK cells results in a size-selective increase in permeability for solutes that are larger than the claudin-based pores. In addition, ZO-1 depletion results in uncoupling of the normal barrier response to inhibition of myosin activity. The barrier defect does not involve a disruption of the small claudin-based pores as measured by the P_app of PEGs <4 Å or by TER. In fact, TER, a measure of instantaneous ionic permeability, is slightly but significantly increased in knockdown cells relative to controls. The increased permeability for larger solutes is rescued by expression of both full-length ZO-1 and a truncated construct containing amino-terminal domains that, among other things, binds to all of the know tight junction transmembrane proteins (Figure 8). ZO-1 depletion also results in enhanced sensitivity to Ca²⁺-depletion and actin disruption with cytochalasin D. Taken together this is the first direct evidence that ZO-1 normally serves to stabilize the barrier and functionally link it to the actin–myosin cytoskeleton.

Figure 8.

Diagram of reported ZO-1–binding proteins and their binding sites. Original references are cited in Fanning and Anderson (2009). Many ZO-1–binding proteins also bind actin including AF-6, cingulin, ZO-2 and -3, α-catenin, Shroom 2, Protein 4.1, and cortactin. These in turn bind many others that regulate actin dynamics.

Although ZO-1 is a universal component of tight junctions and interacts with multiple other tight junction proteins, two recent studies (Umeda et al., 2004; McNeil et al., 2006) reported limited physiological consequences after ZO-1 knockout or knockdown in epithelial cells. In contrast, in the current study we demonstrate a requirement for ZO-1 in normal barrier function in MDCK cells and show that ZO-1–depleted MDCK cells have an altered morphology characterized by a loss of tortuosity at the apical junction contacts and changes in apical actin localization. The reasons for the differences between previously published studies and our findings are likely a function of both cell type and methodology. Umeda et al. (2004) used homologous recombination to knockout ZO-1 expression in Eph4 cells and reported no change in flux for 40-kDa FD; we report a consistent increase in flux for a smaller (3 kDa) FD, but like Umeda et al. have also found that flux of larger 10-kDa FD was not significantly altered. Modest permeability changes (in our case up to twofold) are more difficult to detect as the tracer size increases, except cases where the barrier is catastrophically opened, for example, by Ca²⁺ depletion. The lack of effect on cell morphology may be related to cell type, because Eph4 cells normally lack the convoluted tight junction staining pattern that characterizes MDCK II cells. Macara and coworkers (McNeil et al., 2006), on the other hand, did not measure flux in their ZO-1 knockdown cell lines; that they did not note a change in morphology is likely to be a function of the fact that they used transient transfection to deplete ZO-1 rather than generating stable cell lines as we have done. A third group (Aijaz et al., 2007) has also generated stable MDCK ZO-1 knockdown cell lines. In contrast to our finding, they did not find an effect on paracellular flux. However, this group reported a decrease in ZO-1 levels only to 25% of control levels, compared with 2–5% remaining in our cell lines. We suspect the residual level of ZO-1 in their cells is the reason for their failure to see change in flux, because this level of ZO-1 expression in our ”rescued“ cell lines is sufficient to return flux to normal levels.

In spite of many studies which have reported changes in ZO-1 localization in association with alterations in paracellular permeability (Nusrat et al., 2000; Wang et al., 2005), the finding that ZO-1 depletion results in increased flux in the size-independent pathway is the first demonstration of a clear physiological role for this much studied protein. The observed increase in flux is modest, but it occurs in a background of some residual ZO-1 and minor or no changes in the distribution and levels of most other tight junction proteins. The increased flux is not associated with a decrease in TER or with a change in dilution potential; these outcomes are consistent with the immunolocalization and immunoblot findings that the tight junction transmembrane proteins such as claudins, occludin, and tricellulin are unchanged. Along with the change in flux, the observed alteration in actin localization in the knockdown cells provides clear evidence that ZO-1 is a crucial scaffolding protein, linking a physiologically relevant actin pool to the tight junction barrier. Evidence for a similar role for ZO-2 is lacking, because in our hands, ZO-2 knockdown had no observable physiological effect. This result is different from that reported elsewhere (Hernandez et al., 2007), but could be due to the fact that Hernandez and coworkers tested transiently transfected cells, whereas we analyzed flux in stable knockdowns.

Increased flux for solutes which are larger than the claudin pores in the absence of decreased TER implies that the long-term dynamic characteristics of the barrier are altered by ZO-1 depletion without a compromise in the instantaneous electrical barrier. The tight junction is arranged as a series of continuous claudin-based cell–cell contacts. Multiple barriers presumably provide a fail-safe in case individual contacts transiently break. We speculate that the selective effect of ZO-1 depletion could result from more frequent transient breaks in the contacts but without the simultaneous loss of all of the contacts in series. This would increase permeability for solutes measured over long times but not affect a fast assessment of the barrier with TER. Future studies will test whether ZO-1–depleted junctions are more dynamic.

Previous studies have demonstrated that a ring of filamentous actin normally encircles the apical junctional complex and have indicated that contraction of the ring is associated with changes in cell shape and tight junction permeability (Madara, 1987). Additionally, ultrastructural studies have detected direct contact between actin filaments and the tight junction, although the functional significance of these contacts is unknown (Hirokawa et al., 1983; Madara, 1987). Our studies indicate that F-actin localization is significantly altered in the ZO-1 knockdown cells, with an increase in actin staining at apical junctional complex and into scattered apical dots. These observations suggest that ZO-1 has very specific effects on actin dynamics at the apical junctional complex, perhaps by localizing the activity of cytoskeletal proteins [such as α actinin; Chen et al., 2006), α catenin (Itoh et al., 1997), or shroom2 (Etournay et al., 2007)] or signaling pathways [such as the RhoGEF TUBA (Otani et al., 2006) and Gα₁₂ (Meyer et al., 2002)] that regulate actin dynamics.

How these changes in actin dynamics or cortical F-actin architecture affect permeability is a matter of speculation. One possibility is that there is a specific cortical actin network that is associated with the barrier and that the assembly or plasticity of this network is organized by ZO-1 (via contacts with cytoskeletal proteins outlined in Figure 8). In the absence of ZO-1, the network does not maintain a normal barrier. It seems unlikely that ZO-1 acts as a direct transducer between the physical force generated by actomyosin activity and components of the barrier. The C-terminus of ZO-1 binds directly to F-actin (outlined in Figure 8); however, because the increased flux is rescued by the Nterm construct, which lacks the direct actin-binding site, direct interactions between ZO-1 and F-actin do not appear essential for barrier stabilization. Instead, our data suggest that indirect contacts via other actin-binding and/or -signaling proteins that bind to the Nterm construct are sufficient.

Whatever the mechanism, our observations suggest that one critical factor is the ability of ZO-1 to link the activity of myosin 2 to the state of the barrier. Perhaps surprisingly, the localization of myosin within the apical junctional complex is only slightly altered in ZO-1 knockdown cells. There is a tendency for it to become less continuous and more punctuate in staining pattern. This pattern, most evident in Figure 7C, is again consistent with loss of normal interactions. Many studies (Nusrat et al., 1995; Jou et al., 1998; Walsh et al., 2001; Benais-Pont et al., 2003; Shen et al., 2006) have implicated myosin activation in changes in paracellular flux downstream of RhoA and MLCK. Despite the subtle changes in myosin 2B localization, there is a clear difference between control and ZO-1 knockdown cells in their responses to pharmacologic inhibition of ROCK or myosin 2 ATPase. This suggests that ZO-1 does not localize myosin 2B, but may still act as a scaffolding protein indirectly regulating myosin activity. For example, ZO-1 knockdown decreases cingulin localization (Umeda et al., 2004; Supplementary Figure S3). Cingulin (and the related protein, paracingulin, Guillemot et al., 2008) have been demonstrated to bind the guanidine nucleotide exchange factor, GEF-H1, which inhibits Rho signaling (Benais-Pont et al., 2003; Aijaz et al., 2005). Decreased tight junction cingulin might increase Rho activation and thus myosin phosphorylation; however, as previously reported, neither cingulin knockout (Guillemot et al., 2004) nor knockdown (Guillemot and Citi, 2006) altered permeability. The role of paracingulin in flux regulation has not yet been tested. Another possibility is that ZO-1 might normally act as a sink for Gα₁₂, which can bind to the SH3 domain of ZO-1 (Meyer et al., 2002; Sabath et al., 2008; Figure 8) and that ZO-1 depletion would lead to increased activation of Rho A through an increase in accessible Gα₁₂.

In summary, this is the first direct experimental evidence of a role for ZO-1 in control of paracellular permeability through coupling it to perijunctional actin and myosin. However, providing the evidence that ZO-1 does in fact form a link between the barrier proteins and the cortical cytoskeleton is just a small piece of the puzzle. A cursory inspection of Figure 8 reveals just some of the other proteins that might be important links in this interaction. Unraveling the most relevant of these interactions in the physiological regulation of the tight junction will aid in understanding how barrier integrity is maintained and what components are likely to be involved when the barrier is pathologically compromised.