Occludin (Ocln), a MARVEL-motif-containing protein, is found in all tight junctions. MARVEL motifs are comprised of four transmembrane helices associated with the localization to or formation of diverse membrane subdomains by interacting with the proximal lipid environment. The functions of the Ocln MARVEL motif are unknown. Bioinformatics sequence- and structure-based analyses demonstrated that the MARVEL domain of Ocln family proteins has distinct evolutionarily conserved sequence features that are consistent with its basolateral membrane localization. Live-cell microscopy, fluorescence resonance energy transfer (FRET) and bimolecular fluorescence complementation (BiFC) were used to analyze the intracellular distribution and self-association of fluorescent-protein-tagged full-length human Ocln or the Ocln MARVEL motif excluding the cytosolic C- and N-termini (amino acids 60–269, FP-MARVEL-Ocln). FP-MARVEL-Ocln efficiently arrived at the plasma membrane (PM) and was sorted to the basolateral PM in filter-grown polarized MDCK cells. A series of conserved aromatic amino acids within the MARVEL domain were found to be associated with Ocln dimerization using BiFC. FP-MARVEL-Ocln inhibited membrane pore growth during Triton-X-100-induced solubilization and was shown to increase the membrane-ordered state using Laurdan, a lipid dye. These data demonstrate that the Ocln MARVEL domain mediates self-association and correct sorting to the basolateral membrane.
Epithelial cells surround and stringently regulate the contents of a variety of lumens and cavities by monitoring the bidirectional flow of water and solutes through specifically adapted apical membranes (Martin-Belmonte and Mostov, 2008). The polarized epithelial morphology that constitutes the cellular basis for these functions depends on the physical separation between the apical and basolateral membrane poles, a central function of the tight junctions (TJs) (Tsukita, 1997). The apical membrane has a unique lipid and protein content that enables it to face biochemically or physically harsh extracellular environments (digestive tract, lungs, kidney, blood vessels) and at the same time, regulate the environment's composition. Many of the lipids and proteins that are enriched in the apical membranes of epithelia are also known to facilitate the formation of liquid-ordered membrane domains (Ikonen and Simons, 1998). These domains are considered to form the structural platform for many cellular events associated with transport and signaling processes (Coskun and Simons, 2010). It is now recognized that protein-lipid interactions are crucial in forming, stabilizing and shaping these membrane domains (Hancock, 2006).
The MARVEL domain is a putative protein-lipid-interacting motif containing four intramembrane helices. The human genome includes 28 MARVEL-containing open reading frames (Sánchez-Pulido et al., 2002; Aranda et al., 2011). Myelin and lymphocyte-associated protein (MAL) (Magal et al., 2009) is the first and best characterized MARVEL-domain protein (Millán et al., 1997; Puertollano et al., 1999; Sánchez-Pulido et al., 2002). MAL is a small (17 kDa) membrane protein that is an obligatory component of cellular processes such as the apical sorting machinery, the immunological synapse in activated T cells (Martín-Belmonte et al., 1998; Antón et al., 2008) and myelin formation and maintenance (Magyar et al., 1997). The MARVEL domain of MAL is associated with the generation and stabilization of functional membrane domains via its propensity for homo-oligomerization and its ability to attract apical membrane lipids. A plausible mechanism for MAL function is that its clustering and lipid attraction are brought about in response to its hydrophobic mismatching with the surrounding membrane. Hydrophobic mismatching is defined as the predicted exposure of hydrophobic amino acids to aqueous media due to the difference in length between intramembrane helices and the average thickness of the hydrophobic fatty-acyl chains of membrane lipids. This type of interaction has been shown to provide the thermodynamic drive for the formation of lipid and protein domains (Magal et al., 2009).
Ocln is a tetra-spanning protein found in all epithelial TJs (Furuse et al., 1993; Matter and Balda, 1999). Within the TJ complexes, Ocln has multiple interactions with numerous proteins via its cytosolic domains. These interactions have been extensively studied and shown to be associated with regulating TJ functions. The two extracellular loops of Ocln have been shown to homodimerize with Ocln from opposing cell membranes (Feldman et al., 2005). The four membrane-crossing helices of Ocln comprise the MARVEL motif. Recently, two additional TJ-associated MARVEL-domain proteins have been identified, MALD1 (Raleigh et al., 2010) and MALD3 (Steed et al., 2009). However, the function of the Ocln membrane tetra-spanning MARVEL motif has never been experimentally addressed. In light of the findings on MAL, we speculated that the MARVEL domain of Ocln may exhibit comparable characteristics that could be essential for Ocln targeting or function. Based on the hypothesis that sorting and transport of membrane proteins are propelled at least in part by hydrophobic mismatch-mediated protein clustering (Dukhovny et al., 2009; Schmidt and Weiss, 2010), we tested whether the Ocln MARVEL domain is associated with targeting to the basolateral membrane pole and TJs.
A bioinformatics sequence analysis was carried out to compare Ocln with most known MARVEL-domain proteins. In addition, an amino-acid-sequence-based three-dimensional structural model was obtained for the intramembrane helices of MAL and Ocln. The bioinformatics data strongly concurred with the intracellular distribution of a fluorescently tagged Ocln MARVEL motif lacking its N and C termini. This chimera was correctly targeted to the cell surface or to the basolateral pole of polarized MDCK cells. Fluorescence resonance energy transfer (FRET) and Bimolecular fluorescence complementation (BiFC) (Kerppola, 2008) were used to demonstrate MARVEL-mediated self-association of Ocln. Mutagenesis of conserved aromatic amino acids to alanines mildly affected the oligomerization but resulted in increased aggregate formation. We propose that by interacting with its proximal lipid environment, the MARVEL domain of Ocln mediates its oligomerization and targeting to the basolateral membrane.
Bioinformatics analysis and structure prediction of Ocln MARVEL
A large body of information has been accumulated on the complex protein-protein interactions associated with Ocln function in TJs. Nevertheless, the role of the four transmembrane helices known as the MARVEL domain has never been addressed. Previously, the small MARVEL-containing protein MAL was characterized and features that associate it with membrane-domain formation or stabilization were demonstrated. Whereas MAL is an apical small protein associated with the apical sorting machinery, Ocln is a large multidomain protein that is localized to the basolateral membrane pole and is part of the TJs. The N and C cytosolic domains of Ocln (see scheme in Fig. 1A) interact with numerous proteins, such as ZO1-2-3, claudins and actin. Ocln also contains several phosphorylation sites. We theorized that characterizing the interaction of the Ocln MARVEL domain with its surrounding membrane lipids is essential for understanding its role in TJ function.
Fig. 1B–D summarizes the results of the bioinformatics analysis and sequence-based structure determination carried out to characterize and distinguish the MARVEL domain of Ocln from other known MARVEL-domain proteins. The phylogenetic tree in Fig. 1B demonstrates that the Ocln protein family forms a unique cluster that is different from other MARVEL clusters, such as the neuronal clusters including synaptogyrin and synaptophysin, and other clusters of MYADM, MAL, MALD and CKLF. The sequence of the Ocln-cluster proteins is shown in Fig. 1C, color-coded for hydrophobicity level. The high level of conservation within the Ocln cluster is evident, shown as a color-coded bar graph, especially within the first three transmembrane domains (TMDs). The predicted TMD length of Ocln ranges from 20 to 25 aa for all TMDs, fitting into the range of surface-expressed proteins. Based on these predicted TMDs, the MARVEL sequences of Ocln and MAL were processed to establish a model of the three-dimensional structure of the intramembrane helices using Rosetta software (see the Materials and Methods). This software calculates the lowest free energy conformation in a hydrophobic environment. The output was further refined using previously described algorithms (Fleishman and Ben-Tal, 2006). The results for both MAL and Ocln are shown for top and side views (Fig. 1D). The first observation is that the helical organizations of MAL and Ocln differ. MAL helices are in a 1-2-3-4 clockwise order and Ocln helices are in a 1-4-3-2 order. The MAL helix arrangement is tighter than that of Ocln. These differences may derive from the fact that Ocln has two extensive extracellular loops that are aligned with each other, whereas MAL has very short and tight extracellular loops. Both proteins have many aromatic amino acids. In MAL, these are pointed either inward or outward, possibly to facilitate inter-helix and intermolecular interactions, respectively. A major difference between MAL and Ocln is the relative vertical position of the helices. Whereas Ocln helices are parallel, causing them to occupy minimal vertical transmembrane width, MAL helices are echeloned, generating an overall thicker transmembrane outline. These features correspond to the intracellular distributions of the two MARVEL domains, as MAL localizes to the thicker apical membrane and Ocln to the basolateral one. The same structural analysis was carried out on plasmolipin a small MARVEL domain protein with a high sequence similarity to MAL (supplementary material Fig. S1). Plasmolipin localizes mainly to the sphingolipid and cholesterol enriches myelin membranes as well as the apical membranes in epithelia (Bosse et al., 2003). Although the molecular model of plasmolipin showed a different helix organization (supplementary material Fig. S1), the relative echeloned position of its helices was similar to those of MAL and matched its intracellular distribution.
Intracellular localization of fluorescently tagged Ocln
Human Ocln was tagged with a variety of fluorescent proteins (FPs) and expressed in either nonpolarized COS7 or polarized MDCK cells. Fig. 2A,B demonstrates the arrival of FP-Ocln to the cell surface. However, the cells contain large amounts of the protein in intracellular compartments. We propose that based on previous observations (Furuse et al., 1996) and the fact that Ocln may interact with keen molecules on the adjacent membrane via its luminal loops, Ocln may have the propensity to induce organized smooth ER (OSER)-like structures within the ER (multilamellar structures as opposed to aggregates). Fig. 2C,D shows three-dimensional reconstructed Z-stacks of living MDCK cells. In polarized MDCK cells, FP-tagged Ocln is localized to the TJs and is enriched at the basolateral surface. These data, together with the three-dimensional structure models, led us to propose that the MARVEL domain of Ocln is associated with targeting to the basolateral surface. To investigate this hypothesis, we generated a FP-tagged chimera of Ocln (FP-MARVEL-Ocln) that includes aa 60 to 269: thus it contains the MARVEL membrane-spanning domain and excludes the C- and N-terminal cytosolic domains.
Analysis of FP-MARVEL-Ocln
As shown in Fig. 3A, mCherry-MARVEL-Ocln is localized to the PM of COS7 cells which is co-labeled with the externally added Alexa-488-modified cholera toxin B subunit. mCherry-MARVEL-Ocln also localized to very bright intracellular structures that were reminiscent of previously reported OSER structures. We hypothesized that overexpression of the MARVEL motif, which includes two luminal sticky loops, may have led to OSER formation and mCherry-MARVEL-Ocln accumulation within these structures. Fluorescence recovery after photobleaching (FRAP) (supplementary material Fig. S2) demonstrated that these structures are not aggregates. Moreover, a 24 hour chase with cycloheximide resulted in their elimination and localization of mCherry-MARVEL-Ocln to the PM and to a Golgi-like perinuclear compartment (Fig. 3B). Stable MDCK cells expressing mCherry-MARVEL-Ocln and GPI-YFP (Fig. 3C) grown on Transwell filters, demonstrated that mCherry-MARVEL-Ocln was enriched at the basolateral membrane compared with the GPI-YFP that was sorted to the apical pole. An apparent exclusion from the apical membrane was observed in a 12-day-old polarized culture of an mCherry-MARVEL-Ocln-expressing MDCK clone (Fig. 3D).
To further establish that the Ocln-MARVEL domain facilitates targeting to the basolateral membrane we generated a chimeric mCherry tagged Ocln with its entire MARVEL motif (amino acids 60–269) replaced by MAL (mCherry-OccMAL, Fig. 4A). The 17 kDa MARVEL protein MAL is part of the apical sorting machinery and is targeted to the apical surface in polarized epithelia. In non-polarized cells, mCherry-OccMAL is localized to the PM (Fig. 4B). Moreover, PM-localized mCherry-OccMAL induced the redistribution of soluble ZO1-YFP to the PM suggesting that the C-terminal interacts with ZO1 (not shown). Overexpression of mCherry-OccMAL in polarized MDCK cells stably expressing GPI-YFP resulted in some mistargeting of the GPI-YFP to the Basolateral membranes (Fig. 4C). In polarized MDCK cells, a significant portion of OccMAL was localized to intracellular membranes and did not colocalize with the GPI-YFP at the apical membrane. We hypothesized that the intracellular distribution of OccMAL may be affected by the interaction of the Ocln C-terminal with other proteins particularly with the actin cytoskeleton via ZO1. To reduce the effect of the above-mentioned interactions, polarized MDCK cells stably expressing GPI-YFP and mCherry-OccMAL were treated with 5 µg/ml Latrunculin B for 15 minutes. Fig. 4D demonstrates that upon actin depolymerization, a subpopulation of mCherry-OccMAL appeared to colocalize with GPI-YFP at the apical PM. These data demonstrate that replacing the Ocln MARVEL motif with MAL facilitated at least partial localization of mCherry-OccMAL to the apical membrane. These data suggest that although the MARVEL motif is sufficient to facilitate the polarized transport, the final distribution of the protein is determined by additional interactions mediated by the cytosolic tails.
Analysis of Ocln oligomerization
A major feature of the MAL protein is its ability to form oligomers. Although the sequence of the Ocln MARVEL is significantly different from that of MAL, we asked whether it also mediates oligomerization (McCaffrey et al., 2009). We used two different methods to examine Ocln oligomerization: acceptor photobleaching to detect FRET and BiFC. Fig. 5 shows a typical experiment in which a cell expressing both mCFP- and mYFP-tagged Ocln was fixed in formaldehyde and a rectangle over the area of the cell was repeatedly photobleached to eliminate the acceptor (YFP) while monitoring the CFP channel. Depletion of the acceptor produced an up to 10% increase in the values of the donor (CFP) fluorescence. This increase was not observed outside the bleached box (dashed lines). Also, no FRET was detected in cells coexpressing mCFP-Ocln and YFP-tagged vesicular stomatitis virus G protein (not shown), MAL (not shown), or GPI-anchored YFP, where the YFP acceptor is on the opposite side of the membrane. The negative fluorescence values obtained in the cases where no FRET occurred were a result of an overall photobleaching of the CFP fluorescence due to image capturing. These data demonstrate that CFP- and YFP-tagged Ocln are in sufficient proximity to enable significant levels of FRET. Next, we used BiFC, whose principles are illustrated in Fig. 6A. Briefly, the proximity between two interacting molecules tagged with two complementary halves of a FP facilitates their joining and maturation to a functional FP, indicating dimer formation. The advantages of this method are its strong output signal compared to FRET, and the fact that experiments are carried out using intact cells and membranes. However, since FP formation and maturation are irreversible and occur with relatively slow kinetics, this method does not provide information on the binding dynamics. A weak binding propensity for the two FP halves requires proper controls in the form of mutations within the binding molecules. As shown in the images in Fig. 6A, the YN- and YC-tagged Ocln molecules efficiently form oligomers that fluoresce green at the PM, as well as in a perinuclear intracellular organelle. To estimate transfection efficiency, cells are co-transfected with cytosolic CFP. Our first goal was to verify that Ocln oligomerization is still observable at lower concentrations of the expressed proteins. Fig. 6B shows a western blot analysis of cells co-transfected with decreasing amounts of YN-Ocln and YC-Ocln. The total amounts of plasmid DNA in the transfection mixture was kept constant by replenishing the decreasing amounts of BiFC reactants with increasing amounts of the CFP plasmid. Ocln was identified in the western blot analysis by both anti-GFP and anti-Ocln antibodies. Fig. 6C shows representative images of cells from the experiments shown in Fig. 6B. BiFC fluorescence was detected in the cells transfected with as little as 0.05 µg YN-Ocln and YC-Ocln DNA per 21 cm2 of cells. These data show that Ocln dimerization is detectable with BiFC, even at low expression levels. Next, we asked to verify that the observed dimerization is mediated by the MARVEL motif. To this end, we carried out BiFC experiments with combinations of Ocln, MARVEL-Ocln, and MAL. The MARVEL-Ocln lacked the N- and C-terminal domains and MAL lacked the outer loops, which may potentially be associated with dimerization. Fig. 6D shows that fluorescent BiFC dimers were formed with all combinations. However, the Ocln homodimers generated the strongest binding signals compared to the MAL-Ocln heterodimers. Homodimers of MARVEL-Ocln were localized to the ER, while all others were localized to a perinuclear compartment and the PM. The fact that Ocln and MAL generated some BiFC dimers demonstrates that the hetero-oligomerization interactions are of low specificity.
The relative abundance of aromatic amino acids within the MARVEL domain of Ocln and their role in MAL dimerization led us to hypothesize that these may function in facilitating the dimerization. Fig. 7A lists a series of conserved aromatic amino acids, most of which are embedded within the MARVEL domain. Their relative location is also highlighted in the three-dimensional model generated by Rosetta. As already shown, BiFC dimers of MARVEL-Ocln are transport-incompetent and localized to the ER (Fig. 7B). However, when tagged with mCherry, all of the mutations efficiently arrived at the PM (Fig. 7C). Quantitative analysis of BiFC fluorescence of all mutations was carried out by analysis of BiFC pairs consisting of one of the mutants and the wild type. Representative single mutations, where the aromatic amino acid was replaced by alanine, are shown for F80A (1), F140A (2), Y172A (5) and Y193A (6). The intracellular distribution of the BiFC pairs (mutant-wild type) is shown in Fig. 7B. The intracellular distribution of the mCherry-tagged mutations is shown in Fig. 7C. The quantitative analysis was carried out using three cells expressing mutant-wt BiFC pairs for each mutation as compared with the wt-wt BiFC pairs. The analysis is shown for the 12-bit pixel value range of 100 to 800. This range excludes the background as well as the strong BiFC fluorescence from aggregates. In some of the single mutations, such as in 2, 5 and 6, there was a limited, yet significant effect on the level of BiFC pair formation. However in all mutations, there was an increase in the amount of aggregates and a decrease in the fluorescence that is evenly distributed throughout the ER. The increase in the levels of aggregate formation was also observed in the mCherry-tagged mutations. MARVEL-Ocln versions with multiple mutation combinations were generated as well. A representative selection is shown in supplementary material Fig. S3. There, as in Fig. 7, based on the decrease in BiFC fluorescence, dimerization was affected to a limited extent. Altogether, the FRET and BiFC data demonstrate that the MARVEL domain of Ocln mediates dimerization or oligomerization. This self-association is at least partially dependent on aromatic amino acids, as their replacement by alanine reduces the dimerization and causes an increase in the tendency to aggregate.
We next set out to demonstrate that presence of the MARVEL domain of Ocln can influence the physical characteristics of the cell membrane. Previously, we found that overexpression of MAL increases cell membrane resistance to solubilization by cold 1% Triton X-100 (Dukhovny et al., 2006). This resistance was exhibited by the tendency of MAL-containing membranes to restrain coalescence of detergent-mediated membrane pores. The detergent-mediated increase in pore formation is demonstrated in Fig. 8A, where the solubilization is followed using cells coexpressing either GPI-anchored FP or VSVG-FP and a soluble cytosolic protein. The outflow of cytosolic protein marks the first instance of the detergent hitting the membrane and generating pores. Rapid percolation of the pores ensues. In the cells expressing VSVG, the protein disappears as it partitions to the detergent-soluble membrane fraction. As previously reported in the presence of MAL, the detergent-mediated percolation was significantly limited (Fig. 8B). In cells expressing MARVEL-Ocln, the detergent-mediated solubilization did not involve any apparent pore percolation and growth. Rather, the solubilization proceeded without any increase in pore diameter. These data are consistent with the idea that MARVEL-mediated protein-protein and protein-lipid interactions may increase membrane rigidity. Next, we asked to obtain direct information on the effect of MARVEL-Ocln on the degree of membrane order. To this end we used multi-photon confocal analysis of cells labeled with the Laurdan dye. The Laurdan analysis is based on changes of emission spectra that reflect the degree of membrane ordered state. The analysis is based on obtaining the generalized polarization (GP) values (Gaus et al., 2006). A statistically significant increase in GP values was measured for both Ocln (0.58±0.01, P<0.0001) and MARVEL-Ocln (0.55±0.02, P = 0.0005) compared with control untransfected cells (0.47±0.01). These data demonstrate that the Ocln MARVEL motif can induce an increase in the degree of order of its proximal membrane. Thus, although a basolateral targeting motif, MARVEL-Ocln can affect the properties of its proximal membrane lipids. This may play a role both in its targeting as well as its function as a key modulator of tight junctions.
In summary, the MARVEL domain of Ocln is shown to comprise of a distinctive cluster among the MARVEL family proteins. Its predicted structure was consistent with its localization to the epithelial basolateral membrane. The MARVEL domain of Ocln was efficiently targeted to the basolateral cell surface in polarized cells. FRET and BiFC demonstrated its propensity to self-associate. Mutagenesis of conserved aromatic amino acids resulted in a mild decrease in dimer formation but an increase in aggregate formation. Finally the Ocln MARVEL domain can induce an increase in the resistance of the membrane to detergent mediated pore formation conceivably via increasing the membrane ordered state.
To date, three MARVEL domain-containing proteins have been found associated with TJs. However, little is known about the purpose of this domain in the context of TJ function. Our bioinformatics analysis yielded several interesting observations: the first was that the Ocln MARVEL domain constitutes a unique cluster within the MARVEL-domain protein family. There is little sequence similarity between Ocln and other MARVEL-containing proteins, while Ocln itself remains highly conserved throughout evolution. The predicted hydrophobic helices of the Ocln TMDs are long. However, the three-dimensional structure obtained from the Rosetta algorithm showed that the peptides are aligned such that the overall thickness occupied by the MARVEL domain is equivalent to the thickness of each helix. On the other hand, according to the same algorithm applied to the sequence of the MARVEL-containing proteins MAL and plasmolipin, the transmembrane helices are echeloned, thus generating a thicker overall profile compared to the average thickness of the individual helices. These characteristics of the obtained structures are consistent with the intracellular localization of the proteins. MAL and plasmolipin, with their extended thickness, are localized to the myelin and epithelial apical membranes, which are the thickest membranes in the cell. Ocln is localized to the much thinner basolateral membrane (Mitra et al., 2004). Although both Ocln and MARVEL-Ocln exhibited efficient transport to the PM, a significant portion accumulated in intracellular membranes and was found to be associated with the ER based on colocalization with FP-tagged ER markers (data not shown). These structures are generated by the outer loops since they were missing in cells expressing the mCherry-OccMAL chimera. Thus, we established that the MARVEL motif of Ocln and MAL contains all of the information necessary to travel through the secretory pathway to the cell surface. The MARVEL-Ocln specifically targets to the basolateral membrane domain. A potential mechanism for its cell-surface targeting in the absence of its cytosolic domains is its tendency to partition into or induce the formation of domains enriched with a distinct set of lipids and proteins. This is supported by the ability of both molecules to increase membrane order (Fig. 8C). Notably, Ocln and MARVEL-Ocln were predominantly excluded from the apical pole of polarized MDCK cells. Ocln strongly labeled the TJs and was localized to the basolateral membrane. It is not clear whether this distribution is a result of Ocln overexpression or reflects a dynamic steady state of two actively exchanging pools. MARVEL-Ocln, on the other hand, localized to the basolateral membrane but did not label TJs. This finding lend support to the premise that the Ocln-MARVEL domain is involved in targeting to the TJs as well as interacting with the proximal TJ membrane lipids, but is not involved in associating with other protein components therein. Substituting the Ocln MARVEL with MAL (OccMAL) resulted in targeting to the apical membrane domain only after actin depolymerization with Latrunculin B. This can be attributed to the fact that after arrival at the PM, the cytosolic domain may determine the final distribution of the protein via interactions with other TJ proteins as well as with actin binding proteins such as ZO1. Moreover, evidence is shown that the OccMAL chimera has a disruptive effect on epithelial polarity (Fig. 4C). We also observed that mCherry-OccMAL overexpression in polarized cells resulted in abnormal elongation of the lateral membrane in the z-axis (not shown).
Our working hypothesis regarding oligomerization as a defining feature of the MARVEL domain was tested for Ocln, which exhibits a distinctly different amino acid sequence compared to MAL. The two independent methods FRET and BiFC demonstrated MARVEL-mediated Ocln self-association. The Ocln MARVEL motif was identified as the mediator of this oligomerization. Mutagenesis of conserved aromatic amino acids was carried out to analyze their contribution to oligomer formation. The replacement of conserved aromatic amino acids had a negative effect on dimerization. This is consistent with our previous findings in MAL (Magal et al., 2009) as well as in Plasmolipin (not shown). In addition, aromatic-amino-acid-lacking mutants of MARVEL-Ocln had an increased propensity to aggregate. A conceivable explanation for this result is that the outward extended aromatic amino acids serve to facilitate MARVEL oligomer formation by establishing intermolecular π-stacking interactions. This type of interaction may prevent the formation of tightly associated aggregate-like structures due to irreversible direct binding of the hydrophobic membrane-crossing helices.
The BiFC dimers of the full-length Ocln were transport-competent and their Golgi-to-PM trafficking could be observed in time-lapse images. However, the BiFC dimers of MARVEL-Ocln were transport-incompetent and thus accumulated in the ER. This observation is an indication of the intricate relationship between oligomerization of proteins and their export from the ER. It has been reported that oligomerization of the Cosmc chaperone mediates its retention in the ER (Sun et al., 2011). Here, the BiFC-stable oligomers were retained in the ER while their mCherry counterparts were fully transport-competent.
Finally, the effect of the Ocln MARVEL domain on its proximal lipid environment was demonstrated by its ability to sequester detergent-mediated pore growth: Normally, detergent induces the formation of submicroscopic pores that rapidly coalesce and grow until the system arrives at a steady state in which only the insoluble membrane components remain (Dukhovny et al., 2006). In the presence of Ocln, no pore expansion was apparent. We previously demonstrated that MAL inhibits and restricts pore formation in such a system. Thus, it is conceivable that the interaction of Ocln with membrane lipids shifts the membrane to an ordered state that does not allow any pore growth. In the context of the role of MARVEL in Ocln localization and function in the TJs, it is possible that regardless of its interaction with numerous proteins, Ocln, via its MARVEL domain, plays a crucial role in mediating the interaction of the TJ with membrane lipids. Its ability to oligomerize and prevent detergent-mediated pore expansion may translate, in vivo, to its contribution to the membrane fence function, separating apical from lateral membrane content. Indeed, TJs are found in Ocln-knockout mice (Saitou et al., 2000). However, their phenotype, including multiple histological abnormalities indicating epithelial dysfunctions, underscores the significance of Ocln.
Materials and Methods
Reagents and constructs
All reagents were purchased from Sigma-Aldrich Chemical Co. unless otherwise specified. CTXB488 was purchased from Invitrogen. GPI-mCytrin and GPI-mCFP were prepared as described previously (Glebov and Nichols, 2004). VSVGtsO45-YFP was prepared as described elsewhere (Ward et al., 2001). Human Ocln cDNA (BC029886, Human MGC Verified FL cDNA) was purchased from Open Biosystems and was amplified by PCR using the primers listed in supplementary material Table S1. The Ocln cDNA was subcloned into pmCherry-C1, pECFP-C1 or pEYFP-C1 (Clontech), using XhoI and SacII restriction sites, and subsequently verified by sequencing. For western blot analysis, rabbit anti-Ocln polyclonal antibody was purchased from Invitrogen and mouse monoclonal anti-GFP antibody was from Roche.
Cell culture and transfection
COS7 (African green monkey) and Madin-Darby canine kidney (MDCK) cells were grown at 37°C in a 5% CO2 humidified atmosphere. Cell cultures were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (v/v) fetal calf serum (FCS) and penicillin/streptomycin (Biological Industries, Bet-Haemek, Israel). FuGENE-6 reagent (Roche) and lipofectamine 2000 reagent (Invitrogen) were used for plasmid DNA trasnsfections of subconfluent COS7 and MDCK cells, respectively. To produce stably expressing mCherry-Ocln, mYFP-Ocln and mCFP-Ocln clones, the Neo resistance-containing expression constructs were transfected using 10 µg DNA and 20 µl lipofectamine 2000 to produce clones stably expressing Ocln. Selection medium containing 800 µgG418/ml was added 48 hours after transfection to select stably expressing clones. Confocal laser-scanning microscopy (LSM) experiments were carried out from 18 to 24 hours after transfection. Where level of expressed protein was low, cells were incubated with sodium butyrate for 2 to 6 hours prior to the experiment (Shen et al., 2008). MDCK cells stably expressing mYFP-Ocln were grown on Transwell filters (Corning) for 4 to 12 days.
We collected 422 homologous MARVEL family proteins from the Pfam database (release 23.0), 94 Swissprot sequences were extracted and aligned using MAFFT (with default parameters), and a phylogeny tree for the MARVEL-containing sequences was built using the RAxML webserver. The best tree was scored by the maximum likelihood method after 500 bootstraps. Trees were visualized using FigTree. Transmembrane helix predictions for the human Ocln sequence were obtained from four different webservers described in Fig. 1C, and a consensus of the transmembrane prediction was obtained. For the model structure prediction of human MAL, plasmolipin and Ocln, ab-initio model structures were built using the Rosetta membrane ab-initio protocol (version 3.0). To this end, sequences of the two proteins and the consensus of transmembrane predictions were used. The final models were selected by their ConQuass and Rosetta scores.
Generation of the FP-tagged Ocln MARVEL motif and the FP-tagged Occ MAL
By deleting the cytosolic C- and N-terminal segments of Ocln, we generated a series of fluorescently tagged (mCherry, mYFP) chimeras of Ocln MARVEL, termed FP-MARVEL-Ocln, using a fragment containing aa 60 to 269. Subcloning the MARVEL-Ocln chimera into the FP vectors (Clontech, Palo Alto, CA) was performed using XhoI and SacII restriction sites. The forward primers sequences used for the PCR reaction are listed in supplementary material Table S1.
OccMAL was generated by using mutagenesis to create KpnI and HindIII restriction sites in MAL. The mutated MAL was ligated to the Ocln molecule after cutting the MARVEL with KpnI and HindIII. Mutagenesis was then applied to eliminate the MAL mutations. Sequence analysis was carried out at all stages.
Confocal LSM, time-lapse imaging, FRAP analysis and image processing
Cells were imaged in DMEM without Phenol Red but with supplements, including 20 mM HEPES, pH 7.4. Transfection and imaging were carried out in Lab-Tek chambers (Nunc). Fluorescence images were obtained using a confocal microscope (LSM model PASCAL or 510 META with an Axiovert 200 microscope; Carl Zeiss MicroImaging). Fluorescence emissions resulting from Ar 458 nm, 488 nm, 514 nm and 543 nm laser lines for ECFP, EGFP, EYFP and DiHcRED or mCherry, respectively, were detected using filter sets supplied by the manufacturer. The confocal and time-lapse images were captured using a Plan-Apochromat 63× NA 1.4 objective (Carl Zeiss MicroImaging). Image capture was carried out using the standard time-series option (Carl Zeiss MicroImaging). Temperature on the microscope stage was monitored during time-lapse sessions using an electronic temperature-controlled airstream incubator. Images and movies were generated and analyzed using the Zeiss LSM software, NIH Image and ImageJ software (W. Rasband, NIH, Bethesda, MD).
Long time-lapse image sequences were captured using the autofocusing function integrated into the ‘advanced time series’ macro set (Carl Zeiss MicroImaging, Inc.). For quantitative FRAP measurements, a 63× 1.4 NA Plan-Apochromat objective was used. Fluorescence recovery in the bleached region during the time series was quantified using Zeiss LSM software. For presentation purposes, confocal images were exported in TIFF and their contrast and brightness optimized in Adobe Photoshop. For FRET analysis, cells were washed three times in PBS, fixed in 2% paraformaldehyde (Merck) in PBS for 15 minutes and then washed twice with imaging buffer.
For acceptor photobleaching FRET, cells grown on glass coverslips were fixed in 2% formaldehyde in PBS for 20 minutes at room temperature, washed with DMEM, and mounted onto the microscope. One or two regions of interest (ROIs) within a field were acceptor-photobleached. The 458-nm laser line was used for imaging CFP. The 514 nm laser line was used for the acceptor-photobleaching.
Mutagenesis was carried out on both mCherry-Ocln plasmid and BiFC pairs of Ocln as well as MARVEL-Ocln plasmids using the QuickChange (Stratagene) mutagenesis kit. All of the forward primer sequences are listed in supplementary material Table S1.
BiFC is based on the ability of interacting proteins to fuse to non-fluorescent fragments of the yellow fluorescence protein (YFP) such that when the proteins interact, the YFP fragments complement each other, reconstructing a functional YFP (Hu and Kerppola, 2003). The YFP fragments (YN and YC) were fused to Ocln and to MARVEL-Ocln. The N and C plasmids were a kind gift from Chang-Deng Hu's lab at the Department of Medicinal Chemistry and Molecular Pharmacology at Purdue University, College of Pharmacy, West Lafayette, IN. They generated a split between aa 174 and 175 (N-…Ile-Glu-aSp174/Gly175-Ser-Val…-C) such that the tested Ocln protein was fused to either the N-terminal part of EYFP (nEYFP) or the C-terminal part of EYFP (cEYFP). The Ocln protein, as well as the MARVEL-Ocln chimera, were excised from pEYFP-C1 using XhoI and SacII restriction enzymes and then ligated to nEYFP and cEYFP, using T4 ligase (Roche). Prior to visualization, floating cells were washed to remove dead and dying cells that often show nonspecific fluorescence. Cells were then incubated at 30°C for 2 hours for fluorophore maturation. For negative controls, including the irreversible spontaneous association of the YC and YN moieties, we used the mutations of MARVEL-Ocln protein.
Western blot analysis
COS7 cells were plated at 60% confluence on a 60 mm tissue culture dish (Corning). Cells were transfected with FP-Ocln and cGFP using FuGENE6. COS7 cells expressing Ocln-FP and cGFP were lysed with 1% (v/v) Triton X-100 (TX-100) in PBS. After 15 minutes incubation on ice, the supernatant was collected, and lysates were separated by 7.5% SDS-PAGE and transferred to a nitrocellulose membrane (Bio-Rad Laboratories). Membranes were blocked and probed with primary rabbit anti-Ocln polyclonal and primary mouse monoclonal anti-GFP antibodies, and then with secondary horseradish peroxidase-conjugated anti-rabbit antibody and sheep anti-mouse IgG (NA 931, Amersham Pharmacia Biotech). Detection was carried out using the ECL SuperSignal kit (Pierce).
Membrane labeling with cholera toxin B subunit (CTXB488)
Cells at 60% confluence in Lab-Tek chambered glass coverslips (NalgeNunc International) were transfected with mYFP-Ocln using FuGENE6 and after 18 to 24 hours, labeled with 1 µg/ml CTXB488.
Differential extraction using Cold TX-100
In COS7 cells, detergent extraction was carried out on the microscope stage at 16 to 17°C. Cells were treated with a solution containing 2% TX-100 dissolved in DMEM. All cells expressed soluble CFP to identify the initial contact between detergent and membranes.
Fluorescence images were acquired using a scanning confocal microscope (TCS SP5, Leica Microsystems) equipped with a pulsed Titanium Sapphire laser (Mai-Tai, Spectra-Physics) for multi-photon excitation and a 63× 1.4 NA water-immersion objective lens. The mCherry channel was excited at 561 nm and fluorescence collected in the range 580–700 nm. Laurdan was imaged using multi-photon excitation at 800 nm and fluorescence emission was collected at 400–460 nm (ordered membranes) and 470–530 nm (disordered membranes). The mCherry and Laurdan images were acquired sequentially. GP images were generated as previously described (Gaus et al., 2006). GP values were quantified in regions of interest around the cell plasma membrane.
Many thanks to Chang-Deng Hu, Purdue University, College of Pharmacy, West Lafayette, IN for generously sharing his BiFC reagents and protocols.
This work is funded by the German-Israeli Foundation for Scientific Research and Development [grant number 137/2010].