Myosin light chain phosphorylation regulates barrier function by remodeling tight junction structure

Epithelia form barriers that define tissue compartments within higher organisms. Essential to the function of this barrier is the tight junction (TJ), which seals the paracellular space between individual epithelial cells against passive solute flux. We now know that at least one determinant of TJ ion selectivity is the specific pattern of claudin proteins expressed (Amasheh et al., 2002; Furuse et al., 1998a; Simon et al., 1999; Van Itallie et al., 2001). Modification of this pattern over extended intervals could explain disease-associated changes in TJ barrier function (Heller et al., 2005; Prasad et al., 2005). In the short term, TJ barrier function is acutely regulated by signal transduction cascades that do not depend on altered protein expression. In many cases these signaling cascades depend on the actin cytoskeleton, as they frequently require activation of rho proteins (Benais-Pont et al., 2003; Gopalakrishnan et al., 1998; Hirase et al., 2001; Jou et al., 1998; Nusrat et al., 1995; Walsh et al., 2001) or myosin light chain kinase (MLCK) (Berglund et al., 2001; Ma et al., 1999; Spitz et al., 1995; Turner et al., 1997; Wang et al., 2005; Zolotarevsky et al., 2002).

Myosin and actin form a dense ring that encircles the cell at the level of the adherens junction and TJ. Activation of actomyosin contraction, as assessed by phosphorylation of myosin II regulatory light chain (MLC), has been implicated in TJ regulation (Berglund et al., 2001; Ma et al., 1999; Spitz et al., 1995; Turner et al., 1997; Wang et al., 2005; Zolotarevsky et al., 2002). This frequently requires MLCK activity. For example, inhibition of MLCK can prevent or reverse TJ barrier losses induced by Na⁺-nutrient co-transport (Turner et al., 1997), bacterial and parasitic infection (Dahan et al., 2003; Philpott et al., 1998; Scott et al., 2002; Yuhan et al., 1997; Zolotarevsky et al., 2002) and pro-inflammatory cytokines (Clayburgh et al., 2005; Ma et al., 2005; Wang et al., 2005; Zolotarevsky et al., 2002). Thus, it is now clear that MLCK-driven MLC phosphorylation is involved in signal transduction pathways that effect reduced barrier function in response to diverse stimuli (Turner, 2000).

Although studies using pharmacological MLCK inhibitors indicate a necessary role for MLCK, they are limited by the nonspecific effects inherent to most pharmacological agents. Moreover, inhibitor-based studies fail to consider the possibility that other signaling events known to be activated by upstream stimuli (Savkovic et al., 2003; Shiue et al., 2005; Sugi et al., 2001; Turner and Black, 2001; Zhao et al., 2004) may coordinate with myosin light chain phosphorylation to effect barrier regulation. Thus, the question of whether myosin light chain phosphorylation alone is a sufficient signal to trigger acute barrier regulation remains unanswered.

Two previous studies have examined the role of myosin in epithelial TJ biogenesis and assembly (Gandhi et al., 1997; Hecht et al., 1996). These studies used epithelial cell lines stably transfected to express either a non-phosphorylatable MLC or a constitutively active truncated MLCK (tMLCK). This resulted in markedly reduced or increased MLC phosphorylation-dependent actomyosin contraction, respectively (Gandhi et al., 1997; Hecht et al., 1996). In both cases, epithelial cells lines with altered MLC phosphorylation were unable to develop significant barrier function, suggesting that TJ assembly was disrupted (Gandhi et al., 1997; Hecht et al., 1996). Unfortunately, structure of these incompletely developed TJs was not studied, probably reflecting the relatively limited understanding of TJ protein composition at the time. More importantly, these studies did not evaluate the response of assembled TJs to enhanced MLC phosphorylation, as only TJ assembly was studied. Thus, no available studies directly examine the effect of MLC phosphorylation on assembled TJs in mature epithelial monolayers: the physiologically relevant state.

The aim of these studies was to determine if MLCK-dependent MLC phosphorylation is sufficient to trigger TJ regulation in mature, assembled epithelial monolayers. Moreover, in the event that MLCK activation alone could trigger TJ regulation, we sought to determine morphological and biochemical changes in TJ structure that accompany MLC phosphorylation-induced barrier function decreases. We developed a Caco-2 cell line stably expressing an inducible transactivating element, Tet-off (Gossen and Bujard, 1992; Yu et al., 2001), that drives expression of a constitutively active truncated mutant of MLC kinase, tMLCK (Guerriero et al., 1986). By suppressing tMLCK expression until the Caco-2 cells formed fully differentiated polarized monolayers, we were able to test the role of MLCK in the acute regulation of TJ function.

Our data show that tMLCK-mediated MLC phosphorylation alone is sufficient to induce an increase in TJ permeability. This altered function is associated with structural redistribution of specific TJ proteins. Details of the MLC phosphorylation-induced functional, morphological and biophysical remodeling may explain the underlying mechanisms of the TJ regulation that occurs in response to physiologic and pathophysiologic stimuli.

To develop a cell model for the inducible activation of MLCK, we first established an inducible expression (`Tet-off') system in Caco-2 cells (Yu et al., 2001). These cells were then stably transfected with an inducible construct to express a truncated myosin light chain kinase (tMLCK) lacking the calmodulin-binding inhibitory domain (Ito et al., 1991). The inducible expression approach used makes it possible for monolayers to differentiate and assemble TJs in the absence of transgene expression. Upon gene activation in mature, fully differentiated monolayers transgene protein activity and expression are detectable within 2 hours and continue to increase until reaching a steady-state plateau after 8-16 hours. RT-PCR analysis of tMLCK expression shows that tMLCK mRNA is not detectable in the absence of induction, indicating successful repression of tMLCK transcription. Induction readily induces tMLCK mRNA transcription (Fig. 1A) and increases cellular MLC kinase activity 4.7±0.4-fold (Fig. 1B). As expected for tMLCK, the increase in MLC kinase activity did not depend on the presence of divalent cations. Thus, this Caco-2 cell model allows inducible tMLCK mRNA transcription and MLC kinase activation in mature, fully differentiated monolayers.

Although tMLCK expression is clearly regulated by this inducible expression system, it is crucial to determine if the increase in MLC kinase activity detected in cell lysates (Fig. 1B) reflects increases in intracellular MLC phosphorylation. For example, it is possible that compensatory increases in MLC phosphatase or other counter-regulatory activities occur. Thus, we assessed MLC phosphorylation by induction of tMLCK expression in fully differentiated Caco-2 monolayers with completely assembled TJs. tMLCK expression resulted in increased intracellular MLC phosphorylation, with maximal MLC phosphorylation equal to 150±11% of levels in the control achieved within 8 hours of induction (Fig. 1C). This increase is comparable in magnitude to the MLC phosphorylation that occurs in response to physiological stimuli in cultured intestinal epithelial monolayers and isolated mucosae (Table 1) (Berglund et al., 2001; Turner et al., 1997). Similar results were obtained with at least five independently generated clones. This Caco-2-inducible tMLCK model is therefore suitable for studies of the effects of MLCK activation in fully differentiated intestinal epithelial monolayers.

Many studies have linked MLC phosphorylation to increased actin polymerization and stress fiber assembly (Goeckeler and Wysolmerski, 1995; Katoh et al., 2001). Since both actin and myosin are intimately associated with the TJ, we therefore asked if actin assembly was altered by tMLCK expression. F-actin content was assessed by quantitative analysis of phalloidin binding (Howard and Oresajo, 1985a; Howard and Oresajo, 1985b; Shen and Turner, 2005). Within 4 hours of tMLCK induction, phalloidin binding increased by 18±1% (Fig. 1D), consistent with increased microfilament number or length. This was not due to increased actin synthesis, because SDS-PAGE immunoblot analyses showed no change in total actin content. Inhibition of MLCK activity completely prevented increases in phalloidin binding, suggesting that the enzymatic activity of tMLCK was responsible for the observed changes. Morphological assessment of F-actin labeled with fluorescent phalloidin conjugates showed typical perijunctional actin rings in control monolayers (Fig. 1E). Within 4 hours of tMLCK induction, phalloidin labeling of perijunctional actin rings was irregularly intensified, with brighter foci alternating with areas of reduced or unchanged labeling intensity (Fig. 1F).

The data above show that induction of tMLCK expression in mature, fully differentiated Caco-2 monolayers with assembled TJs results in MLC phosphorylation and reorganization of perijunctional F-actin. To evaluate the effect of tMLCK expression on TJ barrier function we measured transepithelial electrical resistance (TER) before and after tMLCK induction. Progressive decreases in TER occurred during induction of tMLCK expression, with TER loss apparent within 2 hours. TER fell by 18±2% at 8 hours after doxycycline washout (Fig. 2A). This was maximal, as the TER reductions regularly observed after 24-48 hours of induction ranged from 17% to 26% in individual experiments using the clone shown and five other independently generated clones. The magnitude of this TER loss was remarkably similar to that observed after initiation of Na⁺-nutrient co-transport in cultured monolayers and isolated mucosae (Table 1) (Atisook et al., 1990; Berglund et al., 2001; Turner et al., 1997). Moreover, these TER reductions occurred with similar kinetics to increases in MLC phosphorylation (r=0.94). To determine if the observed effects on TER required increased MLC kinase activity, kinase activity was inhibited in monolayers expressing tMLCK. Preliminary experiments verified that this treatment was sufficient to normalize MLC phosphorylation levels in monolayers expressing tMLCK. This MLCK inhibition, of both tMLCK and endogenous MLCK (Clayburgh et al., 2004; Ikebe et al., 1987; Ito et al., 1991), rescued the TER drop recorded in monolayers expressing tMLCK (Fig. 2B). Thus, the regulation of TJ barrier function that occurs following induction of tMLCK expression requires MLCK activity.

Recent studies have shown dissociation between TER and TJ permeability to macromolecular tracers following transgenic expression of selected TJ proteins (Balda et al., 1996; McCarthy et al., 1996; Van Itallie et al., 2001). By contrast, in vivo and in vitro physiological TER regulation associated with MLC phosphorylation causes size-selective increases in TJ permeability (Fihn et al., 2000; Turner et al., 1997). We therefore asked if, like physiological MLCK-dependent TJ regulation, tMLCK expression causes size-selective increases in TJ permeability. As probes we used radiolabeled mannitol and inulin, which have hydrodynamic radii of 3.6 Å and 11.5 Å, respectively. Previous studies have shown that physiological MLCK-dependent TJ regulation increases permeability to mannitol, but not inulin, in cultured monolayers and isolated mucosae (Fihn et al., 2000; Turner et al., 1997). Induction of tMLCK expression in Caco-2 cell monolayers caused a 40±4% increase in transepithelial flux of mannitol (Fig. 2C). By contrast, the transepithelial flux of inulin was not changed by tMLCK expression (Fig. 2C). Thus, tMLCK-mediated MLC phosphorylation effects size-selective increases in TJ permeability that recapitulate physiological TJ regulation in vitro and in vivo (Table 1) (Fihn et al., 2000; Turner et al., 1997).

TJ membranes have biochemical characteristics of lipid microdomains that are operationally defined as detergent-insoluble, low-density, glycolipid- and cholesterol-rich membranes (Nusrat et al., 2000). These detergent-insoluble membranes are enriched in TJ proteins (Nusrat et al., 2000) and the TJ proteins within these fractions are thought to represent those specifically associated with the TJ (Nusrat et al., 2000). For example, more than 90% of hyperphosphorylated occludin, which is thought to represent the tight junction-associated pool, is present in these detergent-insoluble fractions (Nusrat et al., 2000). This represents between 5% and 15% of total cellular occludin (Bruewer et al., 2003; Clayburgh et al., 2005; Nusrat et al., 2000; Nusrat et al., 2001; Wang et al., 2005). Disruption of these cholesterol-rich membrane domains by extraction of plasma membrane cholesterol disrupts TJ barrier properties (Francis et al., 1999) and TJ proteins are lost from detergent-insoluble membrane fractions when TJ barrier function is disrupted by pathophysiological stimuli (Bruewer et al., 2003; Clayburgh et al., 2005; Nusrat et al., 2001; Wang et al., 2005). We therefore evaluated the effect of tMLCK expression on targeting of TJ proteins to detergent-insoluble membrane fractions. We initially asked if MLC phosphorylation might alter the proportion of these proteins in the detergent-insoluble fractions. Neither the proportion of total occludin, claudin-1 or claudin-2 present in these fractions, as opposed to detergent-soluble membranes, nor total cellular content of these proteins was affected by tMLCK expression. However, tMLCK expression resulted in a shift of occludin to higher-density fractions within the detergent-insoluble pool. The peak fractions of occludin shifted from a density of 1.081 g/ml to 1.090 g/ml after induction of tMLCK expression (Fig. 3). This shift corresponded to a marked shift in the overall distribution of occludin within the detergent-insoluble membranes. Although the peak fraction of claudin-1 was also shifted by tMLCK expression, the overall distribution of claudin-1 within the gradient did not change significantly (Fig. 3). The density of claudin-2-containing membranes was also unchanged after tMLCK expression. Despite the unchanged density of claudin-containing detergent-insoluble membranes after tMLCK expression, we considered the possibility that the density of a large subset of low-density, glycolipid- and cholesterol-rich membranes was changed by tMLCK expression. To evaluate this we used a general marker of low-density glycolipid-rich membranes, ganglioside GM1 (Wolf et al., 1998). Like claudin-1, and claudin-2, ganglioside GM1 was present in the detergent-insoluble low-density membrane fractions and the density of the GM1-containing fractions was not changed by tMLCK expression (Fig. 3). Thus, a global increase in the density of detergent-insoluble low-density membranes did occur after induction of tMLCK expression. This indicates that the observed shift in the density at which occludin separates represents a specific effect on occludin-containing detergent-insoluble membranes.

Previous studies have reported the presence of dilatations within the TJ in response to activation of Na⁺-glucose co-transport in isolated intestinal mucosae (Atisook et al., 1990; Atisook and Madara, 1991; Madara and Pappenheimer, 1987). Although recognizable by transmission electron microscopy, these were especially impressive when examined by freeze-fracture electron microscopy (Madara and Pappenheimer, 1987). The latter revealed that the dilatations consisted of expansions of intrajunctional compartments bounded by TJ strands (Madara and Pappenheimer, 1987). To determine whether these changes were associated with MLC phosphorylation in the absence of stimulated Na⁺-glucose co-transport, TJ ultrastructure was examined in freeze-fracture replicas of monolayers with or without tMLCK expression (Fig. 4A,B). Multiple replicas were examined and no significant difference was detected in the morphology or organization of the TJ strands. Most notably, intrajunctional dilatations were not detected. Morphometric analysis of strand number revealed no detectable differences induced by tMLCK expression. TJs in monolayers with or without tMLCK expression were predominantly composed of four strands, with a range of two to nine strands (Fig. 4C). Further morphometry showed that TJ height was 0.195±0.078 μm in the absence of tMLCK expression (181 profiles with a total length of 35.3 μm examined) and increased marginally, to 0.208±0.066 μm height (194 profiles with a total length of 40.3 μm examined), in monolayers expressing tMLCK (P=0.043). Although statistically significant, we interpret these changes as small and of minimal biological significance. These data therefore show that the ultrastructural TJ reorganization, primarily intrajunctional dilatations, previously observed following activation of Na⁺-glucose cotransport in isolated intestinal mucosae cannot be reproduced by MLC phosphorylation alone.

Redistribution of specific TJ proteins during MLC phosphorylation-dependent TJ regulation has not been described in detail. One ultrastructural study reported ZO-1 dissociation from TJ fibrils following activation of Na⁺-glucose co-transport in isolated intestinal mucosae (Madara et al., 1993). More complex stimuli associated with MLC phosphorylation, including bacterial infection and cytokine treatment, have been associated with in vitro and in vivo TJ protein reorganization (Clayburgh et al., 2005; Hopkins et al., 2003; Ivanov et al., 2004a; Ivanov et al., 2004b; Ma et al., 2005; Shifflett et al., 2005; Simonovic et al., 2000; Tomson et al., 2004; Wang et al., 2005). However, it is clear that these stimuli activate many signaling pathways in addition to MLC kinase (Bruewer et al., 2003; Bruewer et al., 2005; Edens et al., 2002; Hopkins et al., 2003; Ivanov et al., 2004a; Ma et al., 2004; Tomson et al., 2004). We therefore asked whether specific alterations in TJ protein distribution could be induced by tMLCK-mediated MLC phosphorylation. When viewed en face, in the x-y plane, ZO-1 profiles in control monolayers are generally smooth arc-like structures (Fig. 5A). Occasional gentle undulations impart a focal waviness to the profiles. Within 4 hours of tMLCK induction this appearance was markedly modified; smooth arc-like ZO-1 profiles were difficult to appreciate. Almost all areas showed the gentle undulations that were focally present in control monolayers (Fig. 5B). Moreover, in many areas, ZO-1 profiles were dramatically rearranged into an irregular series of fine and course undulations (Fig. 5B). Similar ZO-1 reorganization occurred in five independently generated clones with inducible tMLCK expression.

Subjective semi-quantitative analyses showed that the number of TJ segments, i.e. the interval between two adjacent tricellular zones, with ZO-1 undulations of any sort increased from 16±4% before to 55±3% after tMLCK expression (n=878; P<0.001). To characterize these changes in a completely quantitative fashion the actual ZO-1 profile length between adjacent tricellular zones was measured. To correct for normal segment length variation, linear distance between adjacent tricellular zones was used to normalize ZO-1 profile length for each segment measured. The ZO-1 profile length was increased from 107±0.5% to 120±2.4% of linear length by tMLCK expression (n=40 per condition; P<0.001) (Fig. 5C).

To better define the reorganization of perijunctional F-actin and ZO-1 proteins that was induced by tMLCK expression, we analyzed double-labeled preparations at higher magnification. Perijunctional F-actin was distorted into undulating profiles by tMLCK expression (Fig. 5D). However, this was less prominent than the changes in ZO-1 and quantitative analysis of F-actin profile length did not show a significant difference after tMLCK expression (Fig. 7). This quantitative difference and the merged images, which show a greater proportion of the ZO-1 profile without prominent F-actin colocalization after tMLCK expression, suggest the presence of focal separations of F-actin from ZO-1 (Fig. 5D, also see Fig. 8 for a very high magnification view).

The morphological changes described above correlate with the decreased TER induced by tMLCK expression. In the case of decreased TER, the effects of tMLCK expression could be completely prevented by inhibition of kinase activity (Fig. 2B). We therefore asked whether kinase inhibition can prevent the ZO-1 undulations that were induced by tMLCK expression. Examination of monolayers expressing tMLCK after kinase inhibition showed that this treatment blocked redistribution of ZO-1 into the undulating profiles seen in monolayers expressing tMLCK (Fig. 5D). Quantitative analysis confirmed that the increase in ZO-1 profile length induced by tMLCK expression was completely blocked; ZO-1 profile length in tMLCK-expressing monolayers was reduced from 120±2.4% to 108±0.5% of linear length by MLCK inhibition (n=40 per condition; P<0.001) (Fig. 5D). This length after MLCK inhibition (108%) is no different to the ZO-1 profile length of 107% measured in control monolayers not expressing tMLCK (n=40 per condition; P>0.1). Thus, the ZO-1 redistribution induced by tMLCK expression requires kinase activity.

To define whether the ZO-1 redistribution induced by tMLCK expression represented a wholesale TJ reorganization, we also evaluated the distributions of occludin, claudin-1, and claudin-2 before and after tMLCK expression. Like ZO-1, occludin distribution in the x-y plane was reorganized from smooth arcs into a series of fine undulations (Fig. 6A). Quantitative analysis of occludin profile length showed that this was increased, from 106±0.6% to 116±2.0% of linear length by tMLCK expression (P<0.001) (Fig. 7). These data suggest that the reorganization of occludin and ZO-1 may be linked. To assess this we directly compared ZO-1 and occludin distributions after tMLCK expression (Fig. 6A). These proteins remained colocalized, even at the sites of undulations (see Fig. 8 for a very high magnification view). Therefore, despite marked redistribution, ZO-1 and occludin maintain their spatial association after tMLCK expression.

We also assessed the effects of tMLCK expression on the claudin distribution by examining the two principal claudin isoforms expressed in these monolayers: claudin-1 and claudin-2. To avoid confusion with the large lateral membrane pools of claudin-1 in Caco-2 cell monolayers, the TJ was defined as the z plane with maximal ZO-1 staining. Claudin-1 was only minimally redistributed by tMLCK expression (Fig. 6B). High-magnification images comparing the distribution of claudin-1 to ZO-1 and F-actin in tMLCK-expressing monolayers confirms that, at sites of ZO-1 undulation, claudin-1 distribution is generally unaffected (Fig. 6D) or, in some cases, shows a focal reduction in staining intensity (Fig. 6E). Quantitative analysis confirmed that, unlike ZO-1 and occludin, claudin-1 profile length was unchanged after tMLCK expression; claudin-1 profile lengths before and after tMLCK expression were 106±2.0% and 110±0.9%, respectively (Fig. 7). Although only small amounts of claudin-2 were present at the lateral membranes of Caco-2 monolayers, the TJ was defined as the z plane with maximal ZO-1 staining to avoid misinterpretation owing to extra-TJ claudin-2. Like claudin-1, claudin-2 distribution was also unaffected by tMLCK expression (Fig. 6C). Claudin-2 profile lengths before and after tMLCK expression were 109±0.8% and 111±0.7%, respectively (Fig. 7). Thus, unlike ZO-1 and occludin, claudin-1 and claudin-2 distributions were not significantly modified by tMLCK expression.

The qualitative and quantitative data suggest that ZO-1 and occludin are reorganized into irregular undulating profiles by tMLCK expression. F-actin is clearly reorganized qualitatively, but in a manner different to that of ZO-1 and occludin. This too is supported by the quantitative analyses and suggests that focal morphological separation of ZO-1 and occludin from F-actin may be induced by tMLCK expression. To investigate this possibility, the distributions of these proteins were examined at high magnification in double-labeled preparations. Even at sites of extreme undulations, ZO-1 and occludin remained precisely colocalized (Fig. 8). By contrast, F-actin was frequently separated from ZO-1 and occludin at sites of undulations, such that in some focal planes it appeared that F-actin was absent from all or part of the irregular ZO-1 or occludin profiles (Fig. 8).

Inhibitor studies have indicated that MLCK activation, as assessed by MLC phosphorylation, is necessary for TJ regulation secondary to diverse physiological and pathophysiological stimuli (Hopkins et al., 2003; Scott et al., 2002; Turner et al., 2000; Turner et al., 1997; Yuhan et al., 1997; Zolotarevsky et al., 2002). However, the nonspecific nature of pharmacologic agents, the complexity of the stimuli used, and the absence of appropriate experimental models has made it impossible to determine if MLC phosphorylation alone is sufficient to trigger acute downstream regulation of assembled TJs. To address this problem, we developed a model of inducible tMLCK expression in fully differentiated epithelial monolayers, thereby overcoming limitations of previous studies (Gandhi et al., 1997; Hecht et al., 1996). Our data show that induction of tMLCK expression in mature monolayers causes MLC phosphorylation and is sufficient to cause increases in TJ permeability. These changes in TJ function are accompanied by morphological and biochemical TJ reorganization. Thus, these data are the first to show that MLC phosphorylation, in the absence of other more proximal signals, is sufficient to trigger the rapid regulation of assembled TJs.

Physiological TJ regulation is associated with size-selective increases in paracellular permeability (Fihn et al., 2000; Turner et al., 1997). We found that tMLCK-mediated MLC phosphorylation caused similar size-selective regulation of paracellular permeability. This increase in permeability requires tMLCK enzymatic activity, because MLCK inhibition reversed the effects of tMLCK expression. Moreover, the magnitude of MLC phosphorylation increases and TER decreases and the size selectivity of these changes induced by tMLCK expression were comparable to those following Na⁺-nutrient co-transport (Table 1) (Atisook et al., 1990; Berglund et al., 2001; Fihn et al., 2000; Madara and Pappenheimer, 1987; Turner et al., 1997). Thus, TJ barrier regulation induced by MLC phosphorylation alone is similar to that induced following Na⁺-nutrient co-transport.

Claudin protein expression determines TJ ion-permeability characteristics, but has not been shown to affect size selectivity (Furuse et al., 2001; Van Itallie et al., 2001). Moreover, changes in the pattern of claudin isoform expression have been reported in chronic disease (Heller et al., 2005; Prasad et al., 2005). The size-selective increases in TJ permeability to small hydrophilic solutes induced by tMLCK expression as well as the rapid onset and reversibility of these changes suggest that this mechanism of TJ regulation does not involve changes in claudin isoform expression. This is consistent with absence of apparent changes in synthesis or degradation of TJ proteins after tMLCK expression, as no change in expression of claudin-1, claudin-2, ZO-1 or occludin was detected. Thus, rather than the quantitatively large, ion-selective and long-lasting changes in permeability characteristics induced by changes in claudin isoform expression, MLC phosphorylation appears to represent a quantitatively small, size-selective and rapidly reversible form of TJ regulation. From a physiological perspective, this is an ideal mechanism to fine-tune TJ permeability in response to nutrient co-transport or pathogen exposure.

Our ultrastructural analysis of TJ architecture failed to reveal significant changes in strand organization that correlated with tMLCK-induced TJ regulation. Together with the absence of changes in claudin expression or distribution, this is consistent with the observation that claudin proteins are primary constituents of TJ strands (Furuse et al., 1998b). However, the distributions of two TJ proteins, ZO-1 and occludin, that interact directly with F-actin (Fanning et al., 1998; Wittchen et al., 1999) and with each other (Furuse et al., 1994) are markedly altered by tMLCK-mediated MLC phosphorylation. The appearance of these proteins at the TJ assumes an undulating profile after tMLCK expression that can be reversed by MLCK inhibition. Moreover, the density of occludin-containing, detergent-insoluble membranes, but not membranes containing claudin-1 and claudin-2, was significantly increased by tMLCK expression, suggesting a change in the protein:lipid ratio of these fractions. The discordance between the observed occludin and ZO-1 redistribution and the lack of such changes in claudin-1 and claudin-2 distribution after tMLCK expression was not expected, particularly in light of the effect on barrier function. Several potential explanations for this difference can be considered. First, it is notable that ZO-1 and occludin, but not claudins, can interact directly with F-actin. Thus, it could be that this direct binding to actomyosin filaments is necessary for the observed TJ protein reorganization. Alternatively, homotypic interactions between claudin molecules on adjacent cells may serve to neutralize opposing actomyosin tension in adjacent cells: thereby preventing claudin proteins from being redistributed into undulated profiles. Biochemical analysis of TJ protein interactions in live epithelial cells within intact monolayers will be necessary to resolve these and other potential mechanisms. Nonetheless, it is clear from both morphological and biochemical data that tMLCK expression causes focal TJ reorganization.

Size-selective increases in paracellular permeability have been suggested to reflect an increase in the number of small pores within the TJ (Fihn et al., 2000). However, the structural correlate to this increase in pore number has not been defined. Some have speculated that the intrajunctional dilatations that appear following Na⁺-nutrient co-transport are the site of these pores (Atisook and Madara, 1991). However, although tMLCK-mediated MLC phosphorylation did cause functional changes similar to those induced by Na⁺-glucose co-transport, it did not induce the appearance of intrajunctional dilatations. Thus, these dilatations are not required for size-selective increases in TJ permeability and may not be related to MLC phosphorylation. This is consistent with our unpublished observations that MLCK inhibition, which prevents increased paracellular permeability following activation of Na⁺-glucose co-transport in isolated intestinal mucosae (Turner et al., 1997), does not prevent formation of intrajunctional dilatations (J.R.T. and J. L. Madara, unpublished observations). It may therefore be that intrajunctional dilatations form in response to the osmotic gradient and water flux that follow transepithelial Na⁺-glucose co-transport. Alternatively, the intrajunctional dilatations may require activation of one of the other signaling pathways triggered by Na⁺-glucose co-transport (Shiue et al., 2005; Turner, 2000; Turner and Black, 2001; Turner et al., 2000; Zhao et al., 2004). Nevertheless, the structural correlate of the increases in permeability, i.e. the small size-selective pores induced by MLC phosphorylation, remains unknown. Although our data do not reveal the identity of these pores or their mechanism of assembly, it is tempting to speculate that the local disruptions in TJ architecture represented by ZO-1 and occludin undulations may be the sites of these pores.

In summary, the results of these studies show that tMLCK-mediated MLC phosphorylation is sufficient to trigger regulation of TJ barrier function in fully differentiated epithelial monolayers. This regulation is associated with biochemical and morphological reorganization of TJ proteins. Based on these results, we conclude that such reorganization secondary to MLC phosphorylation explains the reversible changes in TJ function. This general mechanism of TJ reorganization may underlie the acute TJ regulation seen in vitro and in vivo.

The Tet-off inducible expression system (Gossen and Bujard, 1992) was established in stably transfected Caco-2BBe cell clones (Peterson and Mooseker, 1992; Yu et al., 2001). These cell lines were then stably cotransfected with a vector derived from pBI-L (Clontech, Palo Alto, CA) that expresses luciferase and truncated myosin light chain kinase (tMLCK) from a single Tet-transactivator transcriptional unit (Baron et al., 1995) and pTK hygromycin (Clontech) and selected using hygromycin.

Cells were grown as confluent monolayers on Transwell supports (Corning, Acton, MA), as described previously (Turner et al., 1997). Transwells inserts were maintained in media supplemented with 5 ng/ml of doxycycline, to prevent tMLCK expression during maturation, and studied 14-21 days post confluence. Induction of tMLCK was accomplished by culture in media without doxycycline. Complete suppression of tMLCK expression was ensured during terminal experiments with 25 ng/ml doxycycline.

Monolayers were scraped into Trizol (Invitrogen, Carlsbad, CA) and total RNA was purified, treated with DNase I (Invitrogen), and first strand cDNA synthesis carried out using Superscript II RT (Invitrogen) and oligo (dT)_12-18 primers. PCR amplification of tMLCK transcripts was for 30 cycles (forward primer: 5′-AAGAAAGACTGGGATCGGGGAA-3′; reverse primer: 5′-CCGAAAGAGGGAAGTACC-3′). The identity of the 584 base pair PCR product was confirmed by restriction digestion with NsiI, which generated fragments of 300 and 284 base pairs.

Confluent monolayers were washed and scraped into lysis buffer (20 mM MOPS pH 7.4, 2 mM MgCl₂, 0.2 mM CaCl₂, 1 mM DTT, 0.5% NP-40, 0.5% Triton X-100, 0.5 mM AEBSF, 0.4 mM aprotinin, 25 μM bestatin, 7.5 μM E-64, 10 μM leupeptin, 5 μM pepstatin, 2 mM imidazole, 1 mM sodium fluoride, 1.15 mM sodium molybdate, 1 mM sodium orthovanadate and 4 mM sodium tartrate). Total protein was measured (BCA assay, Pierce, Rockford, IL) and 0.15 μg lysate was incubated in a final reaction mixture with 10 μM recombinant purified intestinal epithelial MLC, 0.5 mM [γ-³²P]ATP, in 20 mM MOPS pH 7.4, 0.2 μM calmodulin, 0.2 mM CaCl₂ and 2 mM MgCl₂ at 30°C (Zolotarevsky et al., 2002). Parallel assays were also performed with 0.5 mM EDTA in place of calmodulin and divalent cations. After 30 minutes, reactions were stopped by addition of SDS-PAGE sample buffer and boiled. After SDS-PAGE, autoradiographs were scanned and analyzed using Sigma gel software (SPSS, Chicago, IL).

For analysis of intracellular (endogenous) MLC phosphorylation confluent monolayers were grown on 4.5 cm² Transwell inserts (Corning). Monolayers were scraped into lysis buffer (10 mM Tris pH 7.5, 50 mM NaCl, 0.5% Triton X-100, 0.1% SDS, 0.1% sodium deoxycholate, 0.5 mM AEBSF, 0.4 mM aprotinin, 25 μM bestatin, 7.5 μM E-64, 10 μM leupeptin and 5 μM pepstatin) and boiled. Samples were diluted with 3× SDS-PAGE sample buffer and analyzed by SDS-PAGE and immunoblot using antisera specific for Ser19 monophosphorylated MLC (Berglund et al., 2001). Chemiluminescent images were collected and analyzed using a Fluor-S-Max MultiIamger system (Bio-Rad, Hercules, CA).

Quantitative analysis of phalloidin binding was used to determine F-actin content based on the method of Howard and Oresajo (Howard and Oresajo, 1985a; Howard and Oresajo, 1985b). Cells were grown to confluence in media supplemented with 5ng/ml doxycycline and then transferred to media with or without 25 ng/ml doxycycline. Monolayers were fixed with 3% paraformaldehyde in HBSS at indicated times after tMLCK induction. After fixation the cell monolayers were washed twice with HBSS, permeabilized for 15 minutes in HBSS with 0.5% saponin (w/v), and incubated with 0.5 mM Alexa Fluor 488-conjugated phalloidin (Invitrogen) for 1 hour at room temperature. Unbound phalloidin was removed by five washes in HBSS. Bound Alexa Fluor 488-conjugated phalloidin was then extracted quantitatively by incubation for 18 hours at 4°C (in the dark) in 100% methanol. The fluorescent intensity of the extracted Alexa Fluor 488-conjugated phalloidin was measured on an Ultrospec 3300 fluorescent plate reader (Molecular Dynamics, Sunnyvale, CA). All data were then normalized to the F-actin content of control monolayers incubated in medium with 25 ng/ml doxycycline.

TER of monolayers grown on 0.33 cm² Transwell supports was measured using agar bridges and Ag-AgCl calomel electrodes, as described previously (Turner et al., 1997). Paracellular permeability was assessed in monolayers grown on collagen-coated Snapwells (Corning) mounted in HBSS and placed in modified Üssing chambers (Physiologic Instruments, San Diego, CA), as described previously (Zolotarevsky et al., 2002).

Monolayers were fixed in 1% paraformaldehyde in PBS and stained as described previously (Zolotarevsky et al., 2002). Rabbit polyclonal antibodies against ZO-1, occludin, claudin-1, claudin-2 and monoclonal anti-occludin were from Zymed (South San Francisco, CA). Alexa Fluor 488- and Alexa Fluor 594-conjugated goat anti-rabbit and goat anti-mouse antibodies, Alexa Fluor 488-conjugated phalloidin and Hoechst 33342 were from Invitrogen. Stained monolayers were mounted in Slowfade (Invitrogen) and images collected using 63× N.A. 1.32 or 100× N.A. 1.35 PLAN APO objectives mounted on a Leica DMLB microscope equipped with an 88000 filter set (Chroma Technology, Brattleboro, VT), motorized excitation wheels and z-motor (Ludl Electronic Products, Hawthorne, NY), and Roper Coolsnap HQ camera controlled by Metamorph 6 (Universal Imaging Corporation, Downingtown PA). Z stacks were collected at 0.2 μm intervals.

Images were processed post-acquisition using Autodeblur 9 (AutoQuant Imaging, Watervliet, NY). Widefield fluorescence, adaptive blind, three-dimensional deconvolution was performed for ten iterations using low noise settings and an iteratively reconstructed point spread function. Each channel was deconvolved independently. Direct examination verified that the deconvolution process did not create new morphological features; as the images were similar, but with less out-of-plane haze, to those viewed directly through the widefield oculars. Moreover, images with essentially identical morphological features were obtained by confocal imaging using a Leica SP2 AOBS system over a range of pinhole settings. For presentation of 3D projections, deconvolved stacks were imported into AutoVisualize 9 (AutoQuant Imaging) and 3D hardware-generated maximum volume projections were created using matched settings for corresponding images. For x-y plane images, single deconvolved planes at the level of the TJ were selected based on the peak of ZO-1 labeling. The peak of occludin labeling was used to define the z plane of the TJ when ZO-1 was not labeled.

To quantify TJ profile undulations, well-orientated deconvolved z stacks were used. For semi-quantitative subjective analysis, the TJ profile between two adjacent tricellular zones was categorized as straight or undulated using coded images by a blinded observer. Six randomly chosen fields were counted for each condition. All TJ segments in each field counted were examined and the percentage of with undulations calculated.

For quantitative analysis we reasoned that undulations should increase junction length in en face profiles. To measure length, images were skeletonized using Metamorph 6 (supplementary material Fig. S1A,B) and actual linear junction length between two adjacent tricellular zones was determined using the fiber length measuring tool (supplementary material Fig. S1C). A straight line connecting the two tricellular zones was then drawn and hinged at up to two sites to maximize alignment over the junction (supplementary material Fig. S1D). The length of the hinged line was also measured. The actual linear junction length was normalized to the length of the hinged line for each TJ segment. The tight junction profile length is defined as the ratio between actual linear junction length and corresponding hinged line length. Measurements were performed on randomly chosen fields by a blinded observer. Forty segments were measured for each condition examined.

Confluent monolayers grown on 44 cm² Transwell supports were induced as described above and fixed in 2% glutaraldehyde in PBS for 20 minutes at 4°C. After rinsing in PBS, monolayers were scraped from the filter and infiltrated with 25% glycerol in 0.1 M cacodylate pH 7.3 for 60 minutes at 4°C. Cell pellets were frozen in liquid nitrogen slush and then freeze fractured at -115°C in a Balzers 400 freeze-fracture unit (Balzers, Liechtenstein). After cleaning with sodium hypochlorite, replicas were examined using a Philips 301 electron microscope. TJ height and number of parallel strands within the TJs was counted at 1 cm intervals on coded electron micrographs at a magnification of 62,500× (Lynch et al., 1993).

Detergent-insoluble membrane fractions were prepared by scraping monolayers grown on 44 cm² Transwell inserts into 4°C lysis buffer (150 mM NaCl, 20 mM Tris-HCl pH 7.4, 1% Triton X-100, containing 0.5 mM AEBSF, 0.4 mM aprotinin, 25 μM bestatin, 7.5 μM E-64, 10 μM leupeptin, 5 μM pepstatin, 2 mM imidazole, 1 mM sodium fluoride, 1.15 mM sodium molybdate, 1 mM sodium orthovanadate and 4 mM sodium tartrate). Nuclei and cellular debris were removed by a low speed centrifugation (1000 g for 10 minutes). The supernatant was diluted with an equal volume of 80% sucrose, layered beneath a preformed 4-30% sucrose (w/v) gradient, and centrifuged at 39,000 g for 18 hours. Fractions (0.5 ml) were collected from the top of the gradient and the sucrose concentration in each fraction was measured by refractometry. Equal volumes of the fractions were combined 1:1 with 2× Laemmli SDS-PAGE sample buffer, separated by 4-20% gradient SDS-PAGE and transferred to PVDF membranes for immunoblotting. Primary antibodies were incubated overnight in blocking buffer. Rabbit polyclonal antibodies against actin, occludin, claudin-1 and claudin-2 were used at 0.5 μg/ml (Zymed). GM1 ganglioside was detected with peroxidase-conjugated cholera toxin B subunit (0.5 nM, Sigma). Membranes were developed using chemiluminsecent detection reagents (Pierce) and images were collected and analyzed using a Fluor-S-Max MultiImager (Bio-Rad).

Data are presented as mean ± s.e.m. All experiments were performed with triplicate or greater samples, and data shown are representative of three or more independent studies. P values were determined by two-tailed Student's t-test and were considered to be significant if P<0.05.

We are pleased to acknowledge the expert technical assistance of Sara Palkon and Yingmin Wang. This work was supported by the National Institutes of Health (DK61931 and DK68271 to J.R.T., DK53056, DK48106, and DK57827 to W.I.L., and HL25822 and HL36781 to E.E.S.), the Crohn's Colitis Foundation of America (to J.R.T.), the University of Chicago Digestive Disease Center (DK42086), and the University of Chicago Cancer Center (CA14599).