Proteomic Analysis of Proteins Surrounding Occludin and Claudin-4 Reveals Their Proximity to Signaling and Trafficking Networks

Tight junctions are complex membrane structures that regulate paracellular movement of material across epithelia and play a role in cell polarity, signaling and cytoskeletal organization. In order to expand knowledge of the tight junction proteome, we used biotin ligase (BioID) fused to occludin and claudin-4 to biotinylate their proximal proteins in cultured MDCK II epithelial cells. We then purified the biotinylated proteins on streptavidin resin and identified them by mass spectrometry. Proteins were ranked by relative abundance of recovery by mass spectrometry, placed in functional categories, and compared not only among the N- and C- termini of occludin and the N-terminus of claudin-4, but also with our published inventory of proteins proximal to the adherens junction protein E-cadherin and the tight junction protein ZO-1. When proteomic results were analyzed, the relative distribution among functional categories was similar between occludin and claudin-4 proximal proteins. Apart from already known tight junction- proteins, occludin and claudin-4 proximal proteins were enriched in signaling and trafficking proteins, especially endocytic trafficking proteins. However there were significant differences in the specific proteins comprising the functional categories near each of the tagging proteins, revealing spatial compartmentalization within the junction complex. Taken together, these results expand the inventory of known and unknown proteins at the tight junction to inform future studies of the organization and physiology of this complex structure.


Introduction
Tight junctions (TJs) are localized at the apical end of the lateral plasma membrane of epithelial cells and form charge-and size-selective barriers that regulate paracellular movement of ions and solutes between the apical-and basolateral side of the epithelial cell layers [1]. TJs also function in cell polarity [2] and cytoskeletal regulation [3]. About 40 proteins have been localized to the TJ to date [4], for example, the scaffolding proteins Zonula Occludens-1 (ZO-1), ZO-2 and ZO-3 [5], and the transmembrane barrier proteins occludin (Ocln) [6], and claudins [7][8][9]. However, the list of identified TJ-associated proteins is likely to be incomplete. To expand the inventory of TJ proteins, we recently used biotin ligase fusion proteins to identify proteins proximal to the N-or C-termini of ZO-1 [10]. The proteins identified in this analysis included numerous previously identified TJ proteins and in addition a variety of trafficking, signaling, cytoskeletal and polarity proteins. Although many proteins were found in proteomic analyses from both fusion proteins, some proteins were uniquely identified as proximal to either the N-or the C-terminus of ZO-1 [10]. Further, comparison of ZO-1 proximal proteins with a recently generated list of proteins proximal to the adherens junction (AJ) protein, E-cadherin, revealed relatively little overlap, suggesting that the biotin ligase tagging method has a high degree of spatial resolution [11]. Thus, to gain further insights into TJ architecture we applied this method to the transmembrane proteins Ocln and claudin-4 (Cldn4); with the goal of comparing their proximal proteomes with those of ZO-1 and E-cadherin.
Occludin, a 65 kDa tetraspan protein was the first transmembrane protein identified at the TJ more than twenty years ago by Furuse et al. [6]. Although Ocln is a nearly invariant constituent of TJ, its functional role at the TJ is still not fully understood. Overexpression of Ocln in MDCK II cells leads to increased transepithelial resistance (TER) [12], whereas Ocln KO mice display an almost normal phenotype [13]. By itself, Ocln does not form the fibrils that characterize the TJ in freeze fracture electron microscopy, however it does co-polymerize with claudins in these strands [7]. The C-terminus of Ocln has been shown to bind ZO-1, subsequently mediating its intracellular trafficking to the lateral plasma membrane and TJs [14]. Ocln phosphorylation has been associated with concentration at the TJ [15] and Ocln extracellular loops and one transmembrane domain have been shown to contribute to its TJ localization and stability [16][17][18]. Although the role of Ocln in paracellular barrier function is yet not fully understood, numerous studies implicated functions in junctional signaling [14,[19][20][21][22][23] and trafficking pathways [24][25][26][27]. Taken together, these previous findings suggested that proteomic analysis of proteins proximal to both the N-and the C-terminus of Ocln might help elucidate relevant junctional signaling, trafficking and cytoskeletal proteins.
The main barrier forming proteins of the TJ are the 24 members of the claudin family of proteins [28]. Claudins are the main structural elements of the TJ and varying claudin composition specifies the barrier properties of epithelia in different organs and tissues [28][29][30][31][32]. Like Ocln, claudins contain four transmembrane helices; however, claudins are much smaller with molecular masses between 21-28 kDa [29]. Overexpression of Cldn4 in MDCK II cells increases TER by selectively decreasing Na + permeability (P Na ) over Clpermeability (P Cl ), and also increases the number of freeze-fracture fibrils [33]. However, like many other claudins, Cldn4 distribution is not limited to the TJ but is also localized along the lateral membrane [34]. Proteomic analysis of proteins proximal to Cldn4 would thus be expected to reveal TJ and trafficking proteins, as well as relevant lateral membrane and cytoskeletal proteins. As a caveat, this method does not allow us to discriminate between proteins proximal to Ocln and Cldn4 at the TJ versus on the lateral membrane.
Proteins identified in this study included many known TJ and AJ proteins. In addition, we also found signaling, trafficking, cell-adhesion, cytoskeletal, and polarity proteins that deserve further investigation for their putative roles in different aspects of junction regulation, including cytoskeletal organization, cell-cell and cell-matrix adhesions, cell migration and proliferation. A number of proteins were biotinylated exclusively or predominantly by biotin ligase fused to either the N-or C-terminus of Ocln and/or the N-terminus of Cldn4, indicating the spatial specificity of this method. This inventory of Ocln and Cldn4 neighboring proteins may lead to new discoveries and insights into the regulation and function of the TJ.

Mass Spectrometry, MASCOT Database Search and Data Analysis
Liquid chromatography tandem mass spectrometry was performed using an Eksigent nanoLC-Ultra 1D Plus system (Dublin, CA) coupled to an LTQ Orbitrap Elite mass spectrometer (Thermo Fisher Scientific) using collision-induced dissociation fragmentation. Peptides were first loaded onto a Zorbax 300SB-C18 trap column (Agilent, Palo Alto, CA) at a flow rate of 6 μl/min for 6 min and then separated on a reversed-phase PicoFrit analytical column (New Objective, Woburn, MA) using a 40-min linear gradient of 5-40% acetonitrile in 0.1% formic acid at a flow rate of 250 nl/min. LTQ-Orbitrap Elite settings were as follows: spray voltage, 1.5 kV; full MS mass range, m/z 300-2000. The LTQ-Orbitrap Elite was operated in a data-dependent mode (i.e. one MS1 high resolution (60,000) scan for precursor ions followed by six data-dependent MS2 scans for precursor ions above a threshold ion count of 500 with collision energy of 35%).
The raw file generated from the LTQ Orbitrap Elite was analyzed using Proteome Discoverer version 1.3 software (Thermo Fisher Scientific, LLC). Data was submitted to Mascot v2.4 (Matrix Sciences) search engine with the following search criteria: database, National Center for Biotechnology Information (NCBI) RefSeq taxonomy (Canis lupus familiaris, (dog)); enzyme, trypsin; miscleavages, 2; variable modifications, oxidation (M), deamidation (NQ), acetyl (protein N-Term), Biotin (N-term), Biotin (K); fixed modification, carbamidomethyl (C); MS peptide tolerance 20 ppm; MS/MS tolerance as 0.8 Da. Post-database search, the peptides were filtered for a false discovery rate of 1% (using target decoy database) and rank 1 peptides (unique to one protein).
All samples were analyzed in triplicates from three independent experiments. Protein inclusion criterion required a protein be present in at least two of the three experiments. After proteins were compiled, keratins, histones, and endogenously biotinylated carboxylases were discarded before calculating the total peptide spectrum match (PSM) for each individual experiment and the normalized PSM for each protein. The average normalized PSM/Observable Peptide Number (OPN) (av n-PSM/OPN) was then calculated as previously described [11]. Ribosomal proteins were removed after the total PSM and normalized PSM for individual proteins in each run was calculated (ribosomal proteins are reported below the other identified proteins for each biotin ligase construct in S2 and S3 Tables). As previously described we also removed all proteins that were less than three times enriched when labeled by biotin ligase Ocln or Cldn4, as compared to the biotin ligase alone, before further functional analysis [11]. The complete protein lists can be found in S2 Table and the enriched protein lists in S3 Table. Only the top 150 proteins enriched around Ocln and Cldn4 were included in further analysis (S4 Table). Primarily UniProt descriptors [39], but also primary literature searches were used to classify proteins into functional categories. The S2 and S3 Tables are organized with the most abundant protein at the top and then in descending order as calculated by the average normalized PSM/Observable Peptide Number. Tables 1-8 and S4 Table are organized relative to the proteomic rank order list generated by BL-Ocln. This means that proteins highly enriched in the Ocln-BL and/or BL-Cldn4 proteomes, but not in the BL-Ocln, are found below BL-Ocln in Tables 1-8 and S4 Table. Proteins enriched in the ZO-1 and E-cad proteomes, that were not present in lists from Ocln and Cldn4 biotin ligase constructs, are not listed.

The Biotin Ligase Occludin and Claudin-4 Fusion Proteins Localize to Tight Junctions and Lateral Plasma Membranes
In order to determine the spatial specificity of the labeling method we determined both the cellular localization of the fusion proteins and the subcellular patterns of biotinylated proteins. Unlike ZO-1 which is focused at the TJ, both claudins and Ocln also show variable localization to the lateral membrane [6,7,12,19,20,33,38,40]. As integral membrane proteins they are also expected to be near proteins in biosynthetic vesicular trafficking pathways [2,25,26,[41][42][43][44][45][46][47]. As expected, en face immunofluorescent images of the TJ protein ZO-1 (Fig. 1A, left panels) and BL-Ocln, Ocln-BL and BL-Cldn4 fusion proteins (middle panels) reveals colocalization at TJs (right panels), as reported by myc epitope staining. The biotin ligase fusion proteins are also found to a variable extent in intracellular compartments. In contrast, we have previously shown that myc-tagged biotin ligase alone is diffusely distributed throughout the cells including the nucleus [10]. Similar to endogenous Ocln [6,12,20], Ocln biotin ligase fusion proteins are concentrated at the TJ, but there is also considerable lateral distribution of the transgenes (Fig. 1B, center panels). This may in part result from their over-expression. However, Ocln normally traffics to the TJ via the lateral cell membrane [16] and endogenous Ocln can be detected at this site with antibodies that recognize unphosphorylated Ocln [15]. These findings suggest that the use of biotin ligase Ocln transgenes to report proximal proteins should provide physiologically relevant information, both at the TJ and at the lateral membrane. Endogenous Cldn4 is localized both to the TJ and the lateral cell membrane with comparable immunofluorescent signal in cultured epithelial cells [34] and tissues [40]; thus the distribution of this transgene approximates that of the endogenous protein (Fig. 1B, bottom center panel). Even though the transgenes in Fig. 1 looks similar by myc staining, the proteins identified with mass spectrometry (MS) differ between the Ocln and Cldn4 biotin ligase fusion proteins (S2 and S3 Tables), and also differ from those identified using the laterally distributed E-cad biotin ligase fusion protein [11]. This suggests that biotin ligase proximity tagging reveals greater spatial resolution than is detectable by immunofluorescent localization.
In order to verify that the biotinylated proteins were concentrated near the expressed fusion proteins, we incubated cell cultures expressing the biotin ligase fusion proteins with 50μM biotin for 16h, fixed the cells and stained them with fluorescent streptavidin. The results show similar, although slightly more diffuse distribution of streptavidin stained proteins (Fig. 2), as compared to the myc-fluorescent signal from BL-Ocln, Ocln-BL and Cldn4-BL (Fig. 1). We have previously demonstrated that fluorescent streptavidin stained cells expressing biotin ligase alone, after incubation with biotin, shows a diffuse staining pattern of biotinylated proteins all Numbers in the columns for biotin ligase constructs are average normalized PSM/OPNx1000 from 3-fold enriched proteins compared to the biotin ligase alone [11]. PSM is based on peptide fragmentation and subsequent sequencing by collision-induced dissociation (CID) where the same precursor mass can be sequence more than once. Numbers in parenthesis shows that the protein is enriched, however not in the top 150. Not detectable (ND) means that a protein is not enriched. Data from ZO-1 and E-cad are taken from previously published data [10,11]. If a reference is not listed in the far right column, UniProt is the source of protein localization/function. Italic proteins fall into more than one functional category, for example exocytosis and endocytosis.
doi:10.1371/journal.pone.0117074.t001 over the cell [10]. Given the overlapping distribution with ZO-1 as well as expression on the lateral cell membrane, we would expect proteins biotinylated by the fusion proteins to include TJ proteins also identified by ZO-1 as well as novel relevant lateral plasma membrane proteins and trafficking proteins.

Coomassie-stained Protein Gels of Samples from MDCK II-cells Expressing Occludin and Claudin-4 Biotin Ligase Fusion Proteins Reveal Differences and Similarities in their Biotinylation Patterns
To determine whether there is specificity to the proteins labeled by the biotin ligase fusion proteins and purified on streptavidin resin, we first compared the pattern of purified proteins on SDS-PAGE gels from cells expressing biotin ligase alone (Fig. 3A, left panels) with proteins from cells expressing BL-Ocln (Fig. 3A, right panels). This analysis demonstrated that the pattern of recovered proteins was dependent on induction of the specific biotin ligase fusion protein and on the addition of biotin to the cell cultures. Protein bands from un-induced cells, and bands from cells in the absence of added biotin, likely reflect the presence of endogenously biotinylated proteins, including carboxylases [48]. The pattern of these bands of endogenously biotinylated proteins in all controls (un-induced cells without added biotin, induced cells without added biotin, and un-induced cells with added biotin) appear very similar (Fig. 3A). In contrast, the pattern of Coomassie-stained proteins from cells induced to express biotin ligase fusion proteins and incubated with exogenous biotin reveals novel protein patterns that varies among the different fusion proteins (for example, compare

Proteomic Analysis Reveals Both Differences and Similarities in Protein Functional Categories Among Occludin and Claudin-4 Biotin Ligase Fusion Proteins
Triplicate MS analyses of proteins labeled with biotin by the biotin ligase fusion proteins, and purified by streptavidin binding, resulted in the identification of a large number of both  expected and unexpected proteins. Abundantly tagged proteins by Ocln and Cldn4 biotin ligase fusion proteins included numerous TJ proteins including the coxsackievirus and adenovirus receptor homolog, ZO-1, partitioning defective 3 homolog and many claudins (Table 1). Other categories enriched around the Ocln-and Cldn4 biotin ligase fusion proteins were signaling, trafficking, membrane, cytoskeletal, cell-adhesion and transport proteins (Fig. 4). For comparison, Tables 1-8 present the most highly tagged proteins near the N-and C-termini of Ocln and the N-terminus of Cldn-4 along with previously published result for proteins near ZO-1 [10] and E-cad [11]. The full lists of Ocln and Cldn4 tagged proteins are presented in S2 Table. To approximate the relative abundance of proteins recovered in the MS analyses, and to correct for overall recovery between experiments, the PSM value for each protein was normalized by dividing it with the total PSMs for that experiment. This value was then averaged between experiments and corrected for protein size by dividing with the theoretical observable peptide numbers (OPN) in the size range detectable by MS analysis [49]. As expected, the PSM values for the same protein varied among the three experiments. The mean coefficient of variation for the five most highly enriched proteins was 0.4 with a range between 0.1-1 (S5 Table). Inspection of the proteomic results revealed that some proteins were recovered at higher average normalized PSM/OPN than others. Some proteins were heavily tagged both in cells expressing biotin ligase alone and in Ocln and Cldn4 biotin ligase fusion-protein expressing cells; examples include transgelin-2 and PDZ and LIM domain protein 5. To focus on the proteins that  [117] 73980368 Vesicle-associated membrane protein 8 Exocytosis. Involved in the targeting and/or fusion of transport vesicles to their target membrane. Involved in the homotypic fusion of early and late endosomes. (2.6) 2 (0.6) ND ND ND [119] 61316260 Caveolin-2 Trancytosis. Acts as an accessory protein in conjunction with CAV1 in targeting to lipid rafts and driving caveolae formation. May act as a scaffolding protein within caveolar membranes.  Signaling and Trafficking Networks Surround Occludin and Claudin-4 were specifically tagged by Ocln and Cldn4 biotin ligase fusion proteins, we first removed all proteins that were less than 3-fold enriched compared to cells expressing biotin ligase alone. The full lists of enriched proteins around Ocln and Cldn4 are presented in S2 Table. Graphing the top 150 individual proteins in this enriched set from most abundant to least (by averaged normalized PSM/OPN) revealed that although many proteins were identified by MS, there were large quantitative differences in their recovery (Fig. 5). These differences could not only be a result of variability in spatial proximity to the biotin ligase fusion proteins, but also due to the number of available lysines and abundance and stability of the target proteins. Because proteins recovered with the highest average normalized PSM/OPN were likely to be the most biologically relevant, we chose to focus functional analysis on the top 150 most enriched proteins in each group (Figs. 4 and 5, S4 Table). Excluding self-biotinylated occludin, the top 10 most tagged proteins proximal to Ocln and Cldn4 include TJ proteins, trafficking proteins, such as VAMP2, VAMP5 and synaptobrevin homolog YKT6 and membrane proteins such as plasmolipin (Fig. 5). Of potential significance many of the top 10 proteins tagged by the Ocln and Cldn4 biotin ligase constructs have not previously been identified as TJ interacting proteins, and therefore deserve further study for potential roles in regulating TJ function.
When the top 150 enriched proteins proximal to Ocln and Cldn4 were analyzed by division into functional categories, as previously reported for BL-ZO-1 [10] and the E-cad-BL [11], the most abundant categories were those including signaling and trafficking proteins (Fig. 4). Ocln, Cldn4 and E-cad are all integral membrane proteins and the amino terminus of ZO-1 is positioned closer to the membrane than is the carboxyl terminus. Thus the proteome surrounding the N-terminus of ZO-1 could be expected to be more similar to those of Ocln and Cldn4. This is in contrast to the enrichment of cytoskeletal and "other" proteins enriched proximal to the Cterminus of ZO-1 [10], which is known to bind actin and other actin binding proteins [50]. We therefore conclude that it seems likely that similarities and differences in functional categories surrounding the different biotin ligase fusion proteins could be related to membrane proximity.

Proteins Identified with Biotin Ligase Occludin and Claudin-4 Fusion Proteins Localize to Tight Junctions and Lateral Plasma Membranes
To verify that proteins biotinylated by the biotin ligase fusion proteins and identified with mass spectrometry co-localize with endogenous occludin and claudin-4, we used immunofluorescent techniques to visualize several novel proteins in cultured MDCK II cells, namely plasmolipin, RNtre, FLRT2 and Mark3. The multi-pass transmembrane protein plasmolipin (PLLP) was highly enriched around both Ocln and Cldn4. GFP-tagged PLLP co-localizes with Ocln and Cldn4 at the TJ and along the basolateral plasma membrane in MDCK II cells (Fig. 6A, S1 Fig.). GFP-PLLP is also diffusely distributed in the cytoplasm (Fig. 6A, S1 Fig.). The trafficking protein USP6 N-terminal-like protein, also called: related to the N-terminus of tre (RNtre) was identified with mass spectrometry in the neighborhoods surrounding Ocln, Cldn4, ZO-1 and E-cad. GFP-RNtre co-localizes with both Ocln and Cldn4 at the lateral plasma membrane and TJ (Fig. 6B, S2 Fig.). GFP-tagged Leucine-rich repeat transmembrane protein FLRT2 (FLRT2), a protein involved in cell adhesion and or receptor signaling, was localized diffusely in the cytoplasm as well as co-localized with Ocln and Cldn4 along the basolateral plasma membrane and the apical TJ (Fig. 6C, S3 Fig.). Rabbit antibody to MAP/microtubule affinity-regulating kinase 3 (Mark3), the most highly enriched kinase proximal to the N-terminus of Ocln and also enriched around Cldn4, ZO-1 and E-cad, revealed diffuse cytoplasmic as well as distinct apical TJ staining in MDCKII cells. Mark3 was co-localized with Ocln at the apical region of the lateral cell membrane and with both Ocln and Cldn4 at the basolateral membrane below the TJ (Fig. 6D, S4 Fig.). Signaling proteins, including kinases, phosphatases, signaling adapters/scaffolds and membrane receptors were the most highly enriched group around both the N-and the C-terminus of Ocln and second largest group enriched around the N-terminus of Cldn4 (Fig. 4, Table 2 and Table 3). Many of the signaling proteins found in this study have been shown to play important roles in regulating cytoskeleton reorganization, cell polarity, cell adhesion and cell fate (e.g. differentiation, proliferation and apoptosis), processes which all previously has been shown to be of important relevance to the TJ [51][52][53][54]. Some of these proteins were also identified in proteomic screens using biotin ligase fused to ZO-1 and/or E-cad, but many others were found enriched uniquely in the Ocln and/or the Cldn4 proteomes. For example, among the most abundant proteins identified proximal to Ocln was adapter molecule CRK, enriched both around the N-and the C-terminus (av n-PSM/OPN of 24.6 and 71.7 respectively). CRK was also previously shown to be enriched in the neighborhoods of the N-and C-terminus of ZO-1 (av n-PSM/OPN of 17.6 and 51.6 respectively) and the C-terminus of E-cad (av n-PSM/OPN of 22.1), however it is not enriched around Cldn4 (Table 2). CRK is reported to interact with  mitogen-activated protein kinase kinase kinase kinase 5 [55], which was also highly enriched in the BL-Ocln proteome (av n-PSM/OPN of 23.3) but not the Ocln-BL and BL-Cldn4 (Table 3). Another example of difference in the biotin ligase fusion protein proteomes is the finding that all three members of the adaptor protein family DVL-1, -2 and-3 are identified as proximal proteins to ZO-1, E-cad and Ocln at comparable abundances, but was not enriched in the Cldn4 proteome (Table 2). DVL-1 has previously been associated with cell-cell junctions [56]. In contrast, some signaling proteins were identified as proximal to both Ocln and Cldn4 but were not found in the ZO-1 proteome. For example, Eph/Ephrin signaling proteins, involved in bidirectional signaling responsible for modulation of cell adhesion and developmental processes [57], were enriched around Ocln and Cldn4 but not ZO-1 (Table 2 and  Table 3). The interaction between Cldn4 and Eph-A2 has previously been shown to lead to tyrosine phosphorylation of Cldn4, in turn resulting in increased paracellular permeability [58]. In addition, Cldn4 has also been shown to interact with ephrin-B1, leading to tyrosine phosphorylation of ephrin-B1 which affected intercellular adhesion [59]. Ephrin-B1 was enriched around both ends of Ocln and was found at the highest abundance at the N-terminus of Cldn4 (Table 2), whereas ephrin type-A receptor 1 (EPHA1) and EPHA2 were only enriched around Cldn4 (Table 3). Ephrin-B1 and EPHA1 were previously shown to be enriched around E-cad [11], although at lower abundances than cldn4, whereas no members of this family were detected in the enriched lists of ZO-1 [10] (Table 2 and Table 3). Similarly, members of the src family of protein tyrosine kinases including src, lyn and yes were enriched at the highest abundance in the Cldn4 proteome. They were also enriched in the Ocln, but not in the ZO-1 and E-cad, proteomes (Table 3). Yes and src have both been previously associated with Ocln [60][61][62].
Some signaling proteins were enriched only around Cldn4. One example is tumor-associated calcium signal transducer 2 (TROP-2; Table 2), a single-pass transmembrane glycoprotein belonging to the EPCAM family. Loss of TROP2 function is associated with corneal dystrophy  Relative abundance of proteins tagged by biotin ligase fusion proteins identified by mass spectrometry. The y-axis is proportional to the amount of protein recovered and was calculated as follows: PSMs from each of the three isolations were normalized (PSM for each protein/total PSMs for that isolation), these normalized PSMs were averaged between the three runs and then divided by the number of theoretical observable peptide number falling in the size range detectable by MS and this value multiplied by 1000. and knockdown of TROP2 in epithelial cells has been shown to lead to deceased TER and possibly to prevent proper TJ localization of claudins 1 and 7 [63].
In general, Ocln is near the largest number of signaling proteins, however some of these signaling proteins were also found in the neighborhood around Cldn4, ZO-1, and E-cad suggesting functional overlaps ( Table 2 and Table 3), [10,11]. GFP-PLLP localizes diffusely in the cytoplasm and at cell-cell contacts (two middle panels). Co-localization with Ocln and Cldn4 appears to be at cell-cell contacts (two right panels). B. GFP-RNtre predominantly localizes to cell-cell contacts (two middle panels) where the co-localization with Ocln and Cldn4 occurs (two right panels). C. GFP-FLRT2 localizes diffusely in the cytoplasm and at cell-cell contacts (two middle panels).Co-localization with Ocln and Cldn4 appears to be at cell-cell contacts (two right panels). D. The majority of Mark3 localizes to cell-cell contacts but is also present in punctate structures in the cytoplasm (two middle panels). Ocln and Cldn4 co-localize with Mark3 at cell-cell contacts (two right panels). Bar, 20 microns (x63 oil objective). Trafficking proteins was the largest functional group of proteins enriched around the N-terminus of Cldn4 and second largest around both the N-and the C-termini of Ocln (Fig. 5). The trafficking proteins were further sub-divided into four separate groups, endocytosis, clathrin-dependent endocytosis, ER/Golgi and exocytosis/trancytosis (Table 4 and Table 5). The majority of trafficking proteins were found to be within the endocytosis subgroup, followed by ER/Golgi, exocytosis/transcytosis and clathrin-dependent endocytosis. The most highly enriched trafficking proteins around Ocln and Cldn4 were the endocytic SNARE proteins synaptobrevin homolog YKT6-like (YKT6), clathrin interactor 1 (CLINT1), vesicleassociated membrane protein 2 (VAMP2) and VAMP5 (Table 4 and Table 5). A VAMP2 interacting protein, syntaxin 6 [64], was also enriched around Ocln and Cldn4 although at lower abundances (Table 4). VAMP5 was previously shown to be enriched around E-cad and the N-terminus of ZO-1, potentially indicating a more general role for VAMP5 in trafficking of TJ associated proteins (Table 4). YKT6 has not only been shown to have a function in endocytosis [65], but has also been linked to ER-Golgi transport [66] which may mean that the high biotinylation of this protein could have occurred during protein synthesis of the biotin ligase fusion proteins.
Many members of the Rab GTPase family were enriched around Ocln-and Cldn4 biotin ligase fusion proteins. The most highly tagged Rab GTPases by Ocln and Cldn4 biotin ligase fusion proteins were Rab-5b, Rab-7a, Rab-8a, Rab-10 and Rab-23 (Table 4). Only Rab-7a, a Rab that controls vesicular membrane traffic to late endosomes and lysosomes as well as maturation of phagosomes and autophagic vacuoles, was also enriched around E-cad [11], but at a significantly lower abundance (Table 4). Rab7 has previously been shown to co-localize with Cldn4 and Ocln in internalized vesicular structures [42,67].
Several members of the Rab 11 family interacting proteins (RABFIP) were highly associated with both the N-and the C-termini of Ocln and only RAB11FIP1 was also associated with Cldn4 (Table 4). More specifically, RAB11FIP1, RAB11FIP2 and RAB11FIP5 were highly associated with both Ocln biotin ligase fusion proteins. RAB11FIP2 phosphorylation has previously been shown to regulate polarity and localization of TJ proteins in MDCKII cells [68]. Both the phosphomimetic and WT RAB11FIP2 overexpression resulted in recruitment of claudin-1 and claudin-2 to TJ whereas the phosphorylation mutant failed to recruit Cldn4 and Ocln. The enrichment of RAB11FIP around Ocln, supports the idea that Ocln delivery and recycling is important to maintain and regulate epithelial paracellular barrier function both during steady state and epithelial wound healing [24,47].
Taken together the trafficking proteins identified in our proteomic study of Ocln and Cldn4 neighboring proteins, combined with previously published ZO-1 and E-cad data [10,11], indicate that the transmembrane barrier sealing proteins are more highly associated with trafficking proteins than the intracellular TJ scaffold ZO-1. This finding could possibly mean that the regulation of these transmembrane proteins is more dependent upon efficient turnover than ZO-1, e.g. that they are being delivered, removed and recycled to the plasma membrane (or degraded in lysosomes) at higher rates. Of note for future studies, none of the most highly enriched trafficking proteins found in this study has so far been described in the TJ literature.

Cell Adhesion Proteins are Enriched Around Occludin and Claudin-4
Complex cell-cell and cell-matrix interactions play crucial roles in mediating and regulating many processes, including cell migration, tissue homeostasis, wound healing, and tumorigenesis. CD44 antigen precursor, a protein that has been shown to play a role in both cell-cell and cell-matrix adhesion and to regulate TJ assembly and barrier function [69], was the most highly enriched within the cell adhesion functional category surrounding both Ocln and Cldn4, with the strongest association at the N-terminus of Cldn4 (Table 6). In the cell-matrix adhesion category integrin β1and α2 were enriched around both Ocln and Cldn4, and β3 only around Cldn4. Overall, the integrins were more highly enriched in the Cldn4 neighborhood as compared to Ocln (for example compare av n-PSM/OPN of 43.6 at Cldn4 N-terminus to 6.6 and 2.1 at the N-and the C-terminus of Ocln). Although no studies thus far have shown direct interactions between Cldn4 and integrins, a number of other claudins have. For example, β1-integrin-mediated adhesion of brain endothelial cells to the surrounding ECM is critical for stabilizing claudin-5 at blood brain barrier (BBB) TJ, and to maintain BBB integrity [70]. Complexes of claudin-7, integrin α2, and claudin-1 have also been shown to be of importance for normal epithelial basolateral compartments of intestines [71].
Two members of the Leucine-rich repeat transmembrane protein (FLRT) FLRT2 and FLRT3, believed to be involved in cell adhesion and/or cell signaling [72][73][74][75], were both enriched around Cldn4 and FLRT2 was also enriched around the N-and the C-terminus of Ocln (Table 6). FLIRT2 knockout has been shown to lead disruptions to the epicardial cell layer preventing fully formed cell-cell junctions [76].
Cytoskeletal, Membrane, Transport, Other, and Unknown-Proteins are also Enriched in the Neighborhoods around Occludin and Claudin-4 Apart from the predominant functional categories, e.g. signaling and trafficking proteins, several other groups of proteins were found in Ocln and Cldn4 proteomes including cytoskeletal, membrane, transport, other proteins and proteins of unknown function.
TJ proteins are connected to the actin-cytoskeleton via ZO-1 and other scaffolding proteins such as spectrin and erythrocyte membrane protein band 4.1 [77][78][79] (S2 Table). There are also other proteins interacting with the TJ that regulate cytoskeleton reorganization through intracellular signaling pathways and transcription regulation such as CDC42 and BAI1associated protein 2 [80][81][82][83], (Table 7). Interestingly, even though the percentage of total enriched proteins identified as cytoskeletal around ZO-1 was higher, especially the C-terminus of ZO-1 [10], only 5 of the 19 cytoskeletal proteins found around Ocln and/or Cldn4 in this study were also identified around ZO-1 (Table 7), indicating different neighboring cytoskeletal partners. In addition, the percentage of cytoskeletal proteins enriched around E-cad was similar to that of both Ocln and Cldn4, however only three proteins were identical [11]. Future studies are needed to understand the protein interactions regulating the interplay between the TJ proteins and the actin cytoskeleton.
Many membrane proteins were biotinylated by Ocln and Cldn4 biotin ligase fusion proteins (Table 8). Among the most highly enriched membrane proteins around the N-terminus of Ocln was plasmolipin (PLLP) (av n-PSM/OPN of 37.9). PLLP was also enriched, but at lower abundance, at the C-terminus of Ocln and the N-terminus of Cldn4. PLLP is a MARVELdomain containing tetraspan protein with sequence similarities with Ocln, tricellulin and marvel D3 [84]. PLLP has been localized both to apical and basolateral plasma membranes in epithelial cells in a variety of tissues [85]. The most highly enriched membrane protein around Cldn4 was basigin (CD147; av n-PSM/OPN of 28.2). Basigin is a transmembrane glycoprotein involved in embryonic development [86], extracellular matrix metalloproteinase (MMP) induction [87] and promotion of epithelial-mesenchymal transition (EMT) [88]. Basigin has been shown to have a basolateral membrane localization in thyroid epithelial cells (FRT) and various basigin mutants transfected into MDCK II cells also localize to the basolateral membrane [89], indicating a potential co-localization and a possible functional interaction with other lateral membrane proteins such as Cldn4. Basigin interaction with Cldn4 has not been described, but clustering of basigin with galectin-3 results in MMP9 release initiating cell-cell disassembly and redistribution of Ocln through its N-terminal domain in corneal epithelial cells [90].
The group of proteins designated to "other" function in Fig. 4 and S4 Table was in general low on the top 150 enriched proteins lists surrounding Ocln and Cldn4. A couple of exceptions included adipose most abundant gene transcript 2 protein (APM2) and breast carcinoma-amplified sequence 1 (BCAS1). BCAS1, a protein that is increased in for example breast cancer and is found in cytoplasmic vesicular structures [91], was enriched around both Ocln biotin ligase fusion proteins and Cldn4, with the highest abundance at the N-termini of Ocln and Cldn4 (S4 Table). AMP2, also implicated to have a role in cancer [92], was highly enriched around the C-terminus of Ocln (Fig. 5, S4 Table), but also at the N-terminus of Ocln.
Two members of the unknown protein group were enriched around many of the biotin ligase constructs tested in our lab; these are sickle tail protein homolog (SKT) and protein FAM83F. SKT was the most highly enriched around E-cad (av n-PSM/OPN of 31.7), but was also enriched around Ocln and ZO-1 (S4 Table). FAM83F was present with the strongest abundance at the N-terminus of Cldn4, but it was also enriched in the neighborhoods of Ocln, ZO-1 and E-cad (S4 Table).
Taken together, even though most proteins identified in the Ocln and Cldn4 proteomes were signaling, trafficking and known TJ/AJ interacting proteins, our data shows that some proteins assigned to other functional categories were also present at high abundance and deserve further investigation for a role in junction regulation.

Conclusion
The proteins identified by the Ocln and Cldn4 biotin ligase fusion proteins in this study should provide a resource for further understanding the organization and function of tight junctions. When prioritizing proteins for further study it seems appropriate to start with those tagged at the highest level. Alternatively, proteins falling in functional categories highly enriched around Ocln and Cldn4 for example, signaling or endocytic proteins could provide new insights into these functions near tight junctions. Although the many signaling, trafficking and cytoskeletal proteins identified are unlikely to be unique to tight junctions, their identification in this screen suggests that they could play important roles associated with this complex structure. Finally, comparison between proteins tagged by biotin ligase fusion proteins of Ocln and Cldn4, and those identified in our previous studies of E-cad and ZO-1 [10,11], should allow identification of sets of tight-and adherens junction proteins and their compartmentalization.  Table. All proteins identified around BL-Ocln, Ocln-BL and BL-Cldn4. Numbers in the columns for biotin ligase constructs are average normalized PSM/OPNx1000. Ribosomal proteins can be found below each column respectively. (XLSX) S3 Table. Enriched proteins identified around BL-Ocln, Ocln-BL and BL-Cldn4. Numbers in the columns for biotin ligase constructs are average normalized PSM/OPNx1000 from 3-fold enriched proteins compared to the biotin ligase alone [11]. Ribosomal proteins can be found below each column respectively. (XLSX) S4 Table. Functional categories of enriched proteins identified around BL-Ocln, Ocln-BL and BL-Cldn4. Numbers in the columns for biotin ligase constructs are average normalized PSM/OPNx1000 from 3-fold enriched proteins (S3 Table) compared to the biotin ligase alone [11]. Numbers in parenthesis shows that the protein is enriched, however not in the top 150. Not detectable (ND) means that a protein is not enriched. Data from ZO-1 and E-cad are taken from previously published data [10,11]. If a reference is not listed in the far right column, UniProt is the source of protein localization/function. Italic proteins fall into more than one functional category, for example exocytosis and endocytosis. and Jason Hoffert (Systems Biology Center, NHLBI, National Institutes of Health) for invaluable help.