| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
* Leibniz-Institut für Molekulare Pharmakologie, Berlin, Germany; and
Institute of Pathology, University of Tübingen, Tübingen, Germany
3 Correspondence: Leibniz-Institut für Molekulare Pharmakologie, Robert-Rössle-Str. 10, 13125 Berlin, Germany. E-mail: gkrause{at}fmp-berlin.de
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Key Words: transmembrane protein • protein-protein interaction • cell-cell contacts • freeze-fracturing • FRET
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
.
The TJ-strands consist of the transmembrane proteins occludin, tricellulin, and claudins (2
3
4
5)
. The claudin family includes 
24 members with a tetraspan transmembrane topology. Transfection experiments with TJ-free fibroblasts have demonstrated that claudins—but not other TJ-proteins—reconstitute membranous strands similar to those observed in epithelial cells (6)
. This demonstrates that claudins are the major functional constituents of the TJs. The scaffolding proteins zonula occludens(ZO)-1 and -2 are essential for the spatial organization of the claudin-based strands in epithelial cells (7)
but not for basic strand formation in claudin-transfected unpolar cells (8)
.
It has been shown that claudins tighten the paracellular cleft, with selectivity for tissue, size and charge. Thus, studies on knock-out mice have demonstrated that claudin-1 and -5 (Cld5) are responsible for the tightness of the skin (9)
and blood-brain barrier (10)
, respectively. Claudin-2, -15, and -16 form paracellular cation pores (11
12
13)
. Claudin-4, -5, -8, -11, and -14 expression selectively decreases the permeability of cations through TJ (14
15
16
17)
. Two cysteines in extracellular loop one (ECL1) of Cld5 seem to be essential for paracellular tightening (17)
. Charges in the ECL1 of claudin-15 are essential for pore formation (18)
.
Claudins can interact in a homo- and heterophilic manner (19)
, i.e., between two molecules of the same claudin member or between those of different claudin members. However, it is unknown how claudins interact at the molecular level to seal the paracellular cleft. Claudins are assumed to interact in the plasma membrane of the same cell (cis-interaction) and between plasma membranes of opposing cells (trans-interaction), similarly as defined for cadherins (20)
. To clarify the molecular interaction and strand-forming mechanism we studied the exogenous expression of an individual claudin in TJ-free human embryonal kidney (HEK) cells. Earlier, we demonstrated direct Cld5 self-association, although the interaction mechanism remained unclear (21)
. We, therefore, investigated the homophilic interaction of Cld5 in detail. In particular, we focused on the very short ECL2 of Cld5, as this loop contains amino acids that are highly conserved among claudins and species, and as we already had observed, oligomerization of this loop (21)
. The HEK cells were transfected with Cld5-mutants and analyzed by live-cell imaging, fluorescence resonance energy transfer (FRET), and electron microscopy. The experimental data were combined with structural data resulting in the first molecular model of a trans-interaction motif for a TJ multispan membrane protein.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
. Antibodies: Rabbit anti-Cld5 (recognizing a cytosolic C-terminal sequence), HRP-conjugated anti-rabbit, Cy3-conjugated anti-rabbit (Invitrogen, Carlsbad, CA, USA), rabbit-anti-FLAG (Sigma, St. Louis, MO, USA).
Mammalian expression vectors, site-directed mutagenesis, and transfection
Expression vectors for murine Cld5 based on pECFP-N1/pEYFP-N1 (Clonetech, Palo Alto, CA, USA): CFP, cyano-fluorescence protein; YFP, yellow-fluorescence protein (21)
. Site-directed mutagenesis was performed with the Quickchange kit (Stratagene, La Jolla, CA, USA), transfection with Lipofectamine 2000 (Invitrogen) according to the supplier’s recommendations. Stable lines were selected by 1 mg/ml G418 (Calbiochem, San Diego, CA, USA). To avoid the development of clonal variations, pools containing different Cld5-YFPs expressing and nonexpressing colonies were used. In addition, cells were sorted with FACSVantage SE System (BD Biosciences, San Jose, CA, USA) to achieve cultures in which 
95% of the cells express the Cld5-YFP-construct.
Immunocytochemistry and live-cell imaging
Immunocytochemistry was performed as described (21)
. For live-cell imaging, transfected cells were transferred to 1 ml DMEM, 10 mM N-(2-hydroxyethyl)piperazine-N'(2-ethanesulfonic acid) pH 7.5 without phenol red. The plasma membrane was visualized by addition of 20 µl trypan blue, 0.05% in PBS (phosphate buffered saline) (22)
. For membrane staining we also used the membrane marker FM1–43 (Molecular Probes, Eugene, OR, USA) and obtained the same results. Cells were examined with a LSM 510 META system, using an Axiovert 135 microscope equipped with a PlanNeofluar 100x/1.3 objective (Zeiss, Oberkochen, Germany) (21)
. YFP was investigated at 
exc. 488/em. 505–550 nm, trypan blue exc. 543/em.>590 nm, DAPI exc. 364/em. 385–470 nm. The thickness of optical sections was <0.9 µm. To quantify Cld5 enrichment at contacts between two Cld5-YFP-expressing cells (Cld5-contacts), intensity profiles of confocal images of living cells were analyzed
(Fig. 2)
. Contacts between two cells were identified by the trypan blue fluorescence peaks. For each cell pair, 5 intensity profiles were quantified. To investigate Cld5-YFP targeting to the plasma membrane, colocalization of the fluorescence of the Cld5-YFP constructs with that of trypan blue in confocal images was analyzed. If no colocalization of a YFP and trypan blue peak was detectable, Cld5 constructs were considered not to be present in the plasma membrane as the wild-type.
|
|
FRET analysis
Cells were cotransfected with plasmids encoding Cld5wt-CFP and YFP-fusion proteins of Cld5wt, Cld5 mutants, or corticotropin releasing factor receptor 1 (CRFR1). CRFR1-YFP was used as control, as this receptor was found to colocalize with Cld5 at cell-cell contacts but does not interact with Cld5 (21)
. FRET acceptor photobleaching was performed three days after transfection as described (21)
. Briefly, CFP and YFP were excited at 458 and 514 nm and detected from 463–495 and 527–634 nm, respectively. Photobleaching of YFP at the area of cell-cell contacts was performed by using 30–40 pulses of the 514 nm argon laser line at 100% intensity. FRET efficiency (EF) was calculated as EF=(IA–IB)x100/IA, where IB and IA refer to the CFP intensity before and after photobleaching. In each experiment, the pair Cld5wt-CFP and Cld5wt-YFP was the internal standard. The relative EF was calculated as the EF of a distinct Cld5wt-CFP/YFP-protein (Cldmutant or CRFR1) pair divided by EF of Cld5wt-CFP/Cld5wt-YFP. For quantitation, the laser and detector settings were kept constant, and only cells with CFP and YFP intensities similar to the intensities of the internal standard (Cld5wt-CFP/Cld5wt-YFP) were used.
Cell-surface biotinylation
Cells were biotinylated as described (22)
, with minor modifications. Briefly, cells were incubated for 50 min at 4°C with 0.4 mg/ml EZ-link-NHS-SS-biotin (Pierce, Rockford, IL, USA) in PBS with Ca/Mg; 50 mM glycine in PBS with Ca/Mg was used for quenching. Lysis was performed in 50 mM Tris/HCl, pH 7.5; 150 mM NaCl; 1 mM ethylendiaminetetraacetate; 1% Nonidet-P40 (Calbiochem); 0.5% Na-deoxycholate, 0.1% Na-dodecylsulfate and Complete Protease Inhibitor Cocktail (Roche, Indianapolis, IN, USA). For each Cld5-YFP construct, the intensities of the immunoreactive bands were normalized to the respective intensity for Cld5wt-YFP detected in parallel. Surface biotinylation rate was calculated by dividing the normalized intensity in the cell surface fraction by the normalized intensity in the lysate.
Freeze-fracture electron microscopy
Transfected cells were cultured in poly-L-lysine coated tissue culture flasks (Greiner, Flacht, Germany) until confluence, washed with PBS with Ca/Mg, fixed for 2 h with 2.5% glutaraldehyde (electron microscopy grade, Sigma) in PBS with Ca/Mg, washed, and processed for freeze-fracture electron microscopy (23)
.
Structural bioinformatics and molecular modeling
To generate a homology model for mouse Cld5-ECL2, a three-step strategy was used: 1) selection of suitable structural templates, by sequence similarity search (FASTA-program) at the Protein Data Bank (http://www.rcsb.org/pdb/). The revealed templates and predictions of secondary structure (program jpred) (24)
indicated a helix/turn/helix motif for ECL2. Since the transmembrane (TM) regions were also predicted as helices (Swiss-Prot database) (25)
, the conjunctions between ECL2 and the TM helices 3 and 4 (Cld5135–140, 162–165) were built on the template in the same manner. 2) The monomer template models were evaluated and ranked to transport/folding-defect phenotypes of all ECL2-mutants according the consistency of intramolecular side chain interactions. 3) A dimer model was generated in that two molecules of the best monomer model were systematically placed next to each other to simulate all possible interfaces of the hypothesized trans-interaction of the loop. As score, the number of intermolecular interactions of side chain pairs was used. In particular, the consistency to the disjunction types of mutants was considered. All manual reciprocal dockings, manipulations, and optimizations of ECL2 models were performed with the program Sybyl 6.9 (Tripos Inc., St. Louis, MO, USA). The models were energetically minimized, using the AMBER95 force field. The geometrical quality of the final model was evaluated by the PROCHECK program (26)
.
Statistics
Data are means ± SEM; the statistical analysis was one-way ANOVA with subsequent unpaired t test, P < 0.01 was taken as significant. Supplemental information is available on the FASEB Journal website.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
A, S2). To investigate the role of the ECL2 of Cld5 in trans-interaction, systematic exchange of single amino acids was performed (position 145 to 160 of mouse Cld5; predicted ECL2, Swiss-Prot). The mutants were transfected and the subcellular distribution was analyzed by immunocytochemistry and live-cell imaging. The Cld5-mutants exhibited three phenotypes: Mutants of the TJ type showed enrichment at Cld5-contacts, which resembled Cld5wt (Fig. 1B
). This indicates that the amino acid substitution did not alter the ability to form trans-interactions. Mutants of the disjunction type were homogenously distributed throughout the plasma membrane and not enriched at Cld5-contacts (Fig. 1C
). This indicates that the amino acid substitutions abolished the trans-interaction. Mutants of the intracellular type were mainly accumulated intracellularly, plasma membrane levels of Cld5 were drastically decreased (Fig. 1D
). This indicates that the mutation disturbs folding and/or transport. For an overview of the mutants, see
Fig. 8
.
|
|
|
|
|
|
A distinct subset of amino acid substitutions in the extracellular loop 2 of claudin-5 blocks trans-interaction
The enrichment of Cld5-YFP mutants at Cld5-contacts was quantified by confocal microscopy of living cells (Fig. 2
). The intensity ratio (I) between the Cld5 signal intensity at contacts between two Cld5-expressing cells and the intensity at contacts between a Cld5-expressing and -nonexpressing cell was calculated for each Cld5-YFP mutant. Thus, two groups were distinguished (Fig. 2
, bottom panel): Group A (R145A, R145Q, E146N, D149N, P150A, T151A, V154A, S155A, Q156A) gave I 2.0. These values were not significantly different from the wild-type (I=3.2±0.5) and reflected strong enrichment at Cld5-contacts (TJ type, black columns). In Group B (E146A, F147A, Y148A, Y148L, D149A, V152A, P153A, Q156E, K157A, K157M, Y158A, E159A, E159Q, L160A), the I values were 1.0 ± 0.3. This was highly significantly lower than for the wild-type and indicated that there was no enrichment at Cld5-contacts. The lack of enrichment at the contact site for the group B could be either due to a specific effect on the trans-interaction (disjunction type, white columns) or due to low levels of Cld5 at the plasma membrane (intracellular type, gray columns).
To analyze plasma membrane targeting, colocalization of all Cld5-YFP mutants with the plasma membrane, visualized with trypan blue, was investigated by confocal microscopy (examples Fig. 1
). In an independent biochemical approach, cell surface biotinylation was applied to quantify the level of mutated Cld5 at the plasma membrane (Fig. 3
). Comparison of the different Cld5-YFP mutants revealed considerable differences in cell surface targeting (Fig. 3B
). The plasma membrane targeting data demonstrate that substitutions R145A, R145Q, E146N, F147A, Y148A, D149N, P150A, T151A, V154A, S155A, Q156A, Q156E, Y158A, and E159Q did not critically decrease the amount of Cld5 in the plasma membrane. In contrast, E146A, Y146L, D149A, V152A, P153A, K157A, K157M, E159A, and L160A reduced the level of Cld5 in the plasma membrane to less than 50% of that of Cld5wt.
Intracellular type of extracellular loop 2 mutants leads to retention of claudin-5 in the endoplasmic reticulum
Intracellular accumulation could be due to misfolding of Cld5 and retention in the endoplasmic reticulum (ER). To clarify this, HEK cells were cotransfected with Cld5-YFP constructs and CFP-ER (ER-marker, Clonetech), and colocalization was analyzed. Cld5wt was mainly found at Cld5-contacts and exhibited weak colocalization with CFP-ER (Fig. 4
A). Cld5E159Q, an example of the disjunction type, was mainly found throughout the plasma membrane and also exhibited weak colocalization with the ER-marker (Fig. 4B
). However, the intracellular type Cld5V152A was not detectable in the plasma membrane; it mainly colocalized with CFP-ER (Fig. 4C
). This demonstrated that the intracellular type, but not the disjunction type, is restrained in the ER. Cld5P150A exhibited an intermediate phenotype: the majority of the transfected cells exhibited intracellular colocalization with CFP-ER and no or weak plasma membrane localization (Fig. 4D
). However, in a subpopulation of cells, wild type-like enrichment at Cld5-contacts was detected (Fig. 4E
). Cld5Y158A represented another intermediate phenotype: intracellular colocalization with CFP-ER and homogenous plasma membrane localization in most cells (Fig. 4F
), wild-type-like enrichment at Cld5-contacts in a subpopulation (Fig. 4G
). In general, the ER colocalization studies indicate that the intracellular phenotype of Cld5 mutants is likely to be due to ER retention.
Mutations in the extracellular loop 2 do not affect cis-interaction at the plasma membrane
To analyze the contribution of the ECL2 to the cis-interaction of Cld5, we used a FRET assay at Cld5-contacts. For this purpose, HEK cells were cotransfected with Cld5wt-CFP and Cld5-YFP mutants to enable the analysis of the mutants in TJ-strands that are formed by the wild-type. Cotransfection of Cld5wt-CFP with Cld5wt-YFP caused a 10-fold increase in FRET-efficiency compared to the CRFR1-YFP control (Fig. 5
, relative FRET-efficiency 1.00±0.05 and 0.09±0.06, respectively). This confirmed the cis-interaction for Cld5wt, as reported (21)
. Coexpression of Cld5wt-CFP with any mutant of TJ or disjunction type (Cld5-YFP) gave relative FRET-efficiencies (>0.86), which were not significantly different from that with Cld5wt-YFP (Fig. 5
, black/white columns). This indicates that the substitutions R145A, R145Q, E146N, F147A, Y148A, D149N, P150A, S155A, Q156A, Q156E, and E159Q do not alter cis-interaction in TJ strands.
In contrast, coexpression of Cld5wt-CFP with mutants of the intracellular type gave significantly lower relative FRET-efficiencies (
0.65). But for the intracellular type Cld5E146A, Y148L, D149A, and E159A, the FRET efficiency was still higher than that of the control (0.48, 0.65, 0.41, and 0.32, respectively). However, for the intracellular types and the intermediate type Cld5Y158A the levels of Cld5 in the plasma membrane were much lower than for the wild-, TJ-, or disjunction types, as judged by the measurement of the fluorescence intensities at Cld5-contacts (data not shown). Consequently, at the plasma membrane, the FRET efficiencies of these Cld5 mutants can only be compared in a semiquantitative manner (further details, see Materials and Methods).
Mutations of the disjunction type and intracellular type inhibit formation of membranous claudin-5 strands
In stably transfected HEK cells, the formation and morphology of TJ strands were examined by freeze-fracturing. Cld5wt-YFP formed extended networks of discontinuous strands at the exoplasmatic (E)-face (Fig. 6
A). These were similar to those formed by nontagged Cld5wt (Supplemental Fig. S2D), indicating that the YFP-tag does not influence strand formation. The mutants of the TJ type Cld5R145A (not shown), Cld5E146N (Fig. 6B
), Cld5D149N (Fig. 6C
), or Cld5Q156A (not shown) formed E-face associated, discontinuous strands similar to Cld5wt. The data demonstrate that these amino acid substitutions did not alter either the capability of Cld5 to form strands or strand morphology. In contrast, no strands were found for cells transfected with intracellular type Cld5V152A (not shown). Strikingly, cells transfected with mutants of the disjunction type (F147A, not shown; Y148A, Fig. 6E
; E159Q, Fig. 6D
) exhibited no or much less strand formation. In addition, some amino acid substitutions altered the degree of association of particles with the P- or E-face. Thus, the very few strands found for the intracellular type Cld5P153A exhibited equal E-/P-face association of particles (Fig. 6G, H
). Finally, the very few and simplex strands noted for Cld5Y148A or Cld5E159Q (disjunction type) displayed pronounced P-face association (Fig. 6E, D
). In the case of the intermediate type Cld5Y158A, some of the strands had double- or triple-line morphology (Fig. 6F
).
Monomer model reflects intramolecular interactions of the extracellular loop 2 of claudin-5
Different structural templates (PDB-ID: 1AD6, 1DEE, 1YC0, 2ANU, 2BDV) were identified with homologous sequences to the Cld5-ECL2. They were used to generate various models (β-hairpin, 
-helix-turn-helix). The best model based on 2BDV (hypothetical protein BB2244, Bordetella bronchiseptica) was selected according to the highest number of relevant intramolecular side chain interactions, consistent with the mutations that caused a folding/transport defect (6 out of 9 intracellular type mutants). This template comprises a helix-turn-helix motif, has >80% sequence similarity to Cld5143–160 and matches seven identical residues (40%) of claudin’s ECL2, especially including two prolines (corresponding to P150, P153) and their flanking residues, which are very probably responsible for the turn conformation between the two helices of the loop (Fig. 7
A). The fold defective phenotype of mutants D149A, P150A, V152A, P153A, K157A, K157M, and L160A is due to the loss of stabilization of the turn conformation by their wild-type residues (Fig. 7B
). In detail, the hydrophilic residues D149 and K157 form hydrogen bonds across the loop/turn conformation, whereas D149 resembles the helix capping conformation as in the template structure. Apart from the two proline residues, which cause a bend in the backbone and are involved in the helix capping and turn conformations; the mutants V152A and L160A are also folding/transport defect. Their wild-type residues form a hydrophobic core between the turn and the two helices. E146 (Fig. 7B
) and E159 (Fig. 7C
) protrude from the loop conformation in our monomer model. They would be capable of interacting with other parts of Cld5, which would explain the observed folding/transport defect for E146A and E159A.
Intermolecular interactions across the paracellular space according a dimer model of the extracellular loop 2 of claudin-5
Amino acids, which are necessary for trans-interaction (corresponding to substitutions of the disjunction type), point toward two opposing sides in the monomer model (Fig. 7C
). The hydrophobic residues F147, Y148, and Y158 form an aromatic interface in a row at one side of the loop model, while the hydrophilic residues Q156 and E159 are situated at the opposite side. The aromatic core formed by an interface of two sets of F147, Y148, and Y158, (F147', Y148', and Y158') are responsible for the dimer formation in the model (Fig. 7D, E
). This model was selected out of eight possible trans orientations on the basis of the highest score of intermolecular side chain interactions.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
). The designated TJ type of substitutions does not considerably disturb trans-interaction, cis-interaction, strand formation, and -morphology. The intracellular type exhibits intracellular accumulation, probably due to misfolding of Cld5. Consequently, interactions at the plasma membrane cannot be assessed. The disjunction type specifically impairs the trans-interaction and strand formation but does not affect cis-interaction or plasma membrane targeting.
So far, other studies have focused on the effect of overexpression or knockout of claudins on paracellular permeability (27)
. In these cases, the coexistence of different claudins made it difficult to look at defined modes of claudin-claudin interaction and oligomerization. In this study, homophilic Cld5 interaction was analyzed by transfection of TJ-free HEK cells with Cld5-YFP/CFP fusion constructs, enabling live-cell imaging and FRET analysis. YFP or CFP is fused to the C-terminus of Cld5. The free C-terminus could bind to a PDZ-domain of ZO-1 (28)
. However, the C-terminus when blocked by YFP/CFP cannot bind to ZO-1, since neither claudin-1-GFP (8)
nor Cld5-YFP (not shown) colocalize with ZO-1 in TJ-free cells. Consequently, in our system the Cld5-Cld5 interaction is not influenced by an interaction with ZO-1. In the case of the TJ and disjunction type, the amount of Cld5 in the plasma membrane is similar to that of the wild-type, as proven by colocalization with a plasma membrane marker and cell-surface biotinylation. Consequently, the lack of enrichment at Cld5-contacts should be directly due to a failure in trans-interaction caused by the substitutions of the disjunction type.
Previously we have demonstrated cis-interaction of Cld5wt by FRET at Cld5-contacts (21)
. Here, we determine the ability of TJ and disjunction type mutants for cis-interaction. Neither type influenced Cld5 cis-interaction at Cld5-contacts. For the intracellular type mutants, the level at Cld5-contacts was much lower than that of the wild-, TJ-, or disjunction types. Thus, the FRET efficiencies of Cld5wt and intracellular type mutants cannot be compared quantitatively. Nevertheless, at Cld5-contacts, most mutants of the intracellular type still exhibit higher relative FRET efficiency than the control. Taken together, these results indicate that the ECL2 is not primarily involved in the cis-interaction.
Intracellular accumulation, as found for some Cld5 mutants, has also been reported for ECL mutations of other claudins (29)
or connexins that form gap junctions (30)
. Predominant colocalization with an ER-marker was found for the intracellular type and partly for the intermediate types (Cld5P150A, Cld5Y158A) but not for the TJ- and disjunction type. This strongly suggests that the respective amino acid substitutions cause misfolding of Cld5, since ER retention is a common feature of misfolded proteins (31)
.
Ultrastructural implications for strand formation by claudin-5
The freeze-fracture experiments demonstrate that only mutants of the TJ type, but not those of the disjunction or intracellular types, can form wild type-like strands with respect to number, interconnection, and E-face association. We demonstrate that in the ECL2 of Cld5, in particular amino acids F147, Y148, Y158, and E159 are involved in strand formation. The same residues are involved in trans-interaction also. Consequently, trans-interaction is a prerequisite for strand formation, as illustrated in Fig. 9
. It was reported earlier that ECLs of claudin-2 and -4 do not contribute to the morphology of heteromeric strands in transfected TJ-containing MDCK cells (18)
. Similarly, we find no major effect of ECL2 substitutions on the morphology of homomeric claudin-5 strands in TJ-free HEK cells, if trans-interaction, folding, and membrane targeting are maintained.
|
The appearance of TJ strands in freeze-fracture replica depends on the nature of the tissue and the claudin. We find that Cld5wt and TJ type mutants form discontinuous rows of particles associated with the E-face in HEK cells, similar as observed in Cld5-transfected fibroblasts (32)
. In contrast, claudin-1, e.g., forms continuous strands associated with the P-face (6)
. The molecular base of these difference is unclear. However, continuity and P-face association (PFA) of TJ strands may correlate with the barrier integrity (33
, 34)
. It has also been suggested that continuity and PFA are favored by heterophilic claudin interactions (35)
. Interestingly, we observe differences in the PFA, too. Mutations, such as Y148A or P153A, affect trans-interaction either directly or indirectly by destabilizing the loop structure (see below) and increase PFA of the Cld5 strands. In detail, the very few and uncomplex strands found for the disjunction type, e.g., Y148A, or intracellular type P153A exhibit strong PFA. This, for the first time, indicates that association of Cld5 particles with the membrane faces could be related to the strength of the trans-interaction. On the other hand, it is assumed that a strong PFA is connected to a strong anchorage to the cytoskeleton (34)
. However, in our system, Cld5 cannot be linked to the cytoskeleton via ZO-1 as discussed above. Consequently, the cytoskeleton is most likely not responsible for the observed difference in PFA.
Molecular model of the homophilic interaction of the extracellular loop 2 of claudin-5
Molecular modeling of the Cld5-ECL2 monomer based on the high sequence homology to the protein structure 2BDV resulted in a helix-turn-helix structural motif. The turn conformation between the two helices is stabilized by a network of hydrogen bonds, two prolines (P150, P153) and hydrophobic side chain interactions (V152, L160). These stabilizing patterns in the monomer model are consistent with folding defect phenotype of alanine mutations at exactly these residues. Due to the cyclic structure of the proline side chain, this amino acid possesses a restricted backbone conformation, which is suited to form turn structures (36)
. In contrast to the intracellular type mutant D149A, D149N results in a wild-type-like phenotype, which is consistent with the model in that it maintains the hydrogen bridge pattern. This is supported by the fact that asparagine occurs at this position in several other claudin-subtypes. This conserved H-bridge between the side chain of D149 and backbone -NH of V152 could also explain why an alanine mutation at position P150 gives a phenotype with only slightly abnormal folding, since an additional stabilizing feature, the H-bridge, bypasses P150 across the first half of the loop. The intracellular phenotype of mutations K157A and K157M can be explained by the loss of the H-bridge of lysine 157 toward backbone of D149 (Fig. 7B
).
Based on our data, the best evaluated trans-dimer model consists of an aromatic core of six residues, which is stabilized by F147, Y148, and Y158 at ECL2 of Cld5 from either side of opposing cells. The postulated interaction via aromatic residues is often described in protein-protein interactions (37)
. Mutations of these aromatic residues gave phenotypes with disturbed trans-interaction. Thus, three of five mutants with disturbed trans-interactions can directly be explained by such an ECL2 dimer model. The remaining two mutants, Q156E and E159Q, are located in our model at the opposite side of the aromatic core as is E146. We suggest that this side of the ECL2 model interacts with other parts of claudin. This idea is supported by the fact that E146 and E159 are highly conserved in the majority of claudins, and alanine mutants of the two residues lead to intracellular phenotypes probably due to disturbed folding.
The claudin family can be divided into two groups on the basis of their amino acid pattern in ECL2. Claudin-1 to -10, -14, -15, -17 and -19 represent the classic group with highly conserved amino acid positions (% similarity): E146 (85%), F147 (100%), Y148 (95%), P150 (98%), K157 (98%), E159 (100%), L160 (100%). Interestingly, all these highly conserved amino acids are shown in our mutation study to be important for folding and/or trans-interactions. This indicates that our findings are important for all members of this classic claudin group, where the ECL2s most probably possess both a common structural fold and a common molecular mechanism to recognize and attach ECL2 of claudins at opposite cells across the TJ. In contrast, claudin-11, -12, -13, -16, -18, and -20 to -24 show high variability in amino acid composition and loop length, indicating different folds from those described here for Cld5 in the classic group. For claudin-16, diverse pathological mutants have been described with effects on folding/transport or on function (Mg2+pore) in different parts of the molecule, including ECL2 (29)
. However, as the ECL2-sequence is so different, mutations of claudin-16 cannot be considered for Cld5.
Conceptual model for formation of claudin-5 strands
We propose a two-step scenario of Cld5 polymerization (Fig. 9)
: 1) Due to cis-interaction, Cld5 monomers oligomerize (brackets, n). However, Cld5 does not polymerize into continuous strands within one membrane. The oligomers are of limited size, as they appear as small intramembranous particles in freeze fracture replica. Furthermore, this size is consistent with Cld5 hexamers found in lysates of eukaryotic cells (35)
. 2) At TJs, the ECL2 essentially contributes to trans-interaction between oligomers in the two opposing plasma membranes. This trans-interaction triggers the formation of polymeric strands. Moreover, this mechanism would prevent uncontrolled polymerization of Cld5 in intracellular compartments. On the other hand, it could explain the discontinuous appearance of Cld5 strands on the E-face in freeze-fracture replica. If the polymerization takes place in the paracellular space due to trans-interaction rather than cis-interaction, as suggested earlier (19)
, the continuity of the strands would be visible only from the paracellular side. Since in freeze-fracture replica the Cld5 strands can be observed only from the intramembranous side (black arrows, Fig. 9
), the paracellular continuity is hidden and, therefore, only discontinuous rows of particles are observed. In addition, since an association of Cld5 with ZO-1 is not possible in our system, it is unlikely that the integration of Cld5 oligomers into strands is caused by this scaffolding protein. Taken together, we suggest that the polymerization takes place in the paracellular space via trans-interactions.
Clarification of the oligomerization of Cld5 is of pharmacological relevance, as knockout mice demonstrate that Cld5 specifically tightens the blood-brain barrier for molecules <800 Da (10)
. This includes the size of most drugs in clinical use. Consequently, elucidation of the interaction mechanism has the potential for the development of specific TJ-modulators, which could be of interest for the improvement of drug delivery to brain tumors or the treatment of neurodegenerative diseases.
Based on our mutation data and supported by the antiparallel homodimer model of ECL2, for the first time, we have specified intermolecular interaction patterns and molecular determinants of how two claudins hold onto each other across the paracellular space. Four of five residues, which are crucial for trans-interaction, are highly conserved in the classic claudin group. Hence, our findings are of general importance for classical claudins and present a conceptual advance in the molecular and structural understanding of how claudins of opposing cells work together to form TJs.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
2 Present address: Wellcome Trust Centre for Cell Biology, University of Edinburgh, Edinburgh, UK.
Received for publication February 12, 2007. Accepted for publication July 19, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. Burek, P. A. Arias-Loza, N. Roewer, and C. Y. Forster Claudin-5 as a Novel Estrogen Target in Vascular Endothelium Arterioscler Thromb Vasc Biol, February 1, 2010; 30(2): 298 - 304. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Zhang, J. Piontek, H. Wolburg, C. Piehl, M. Liss, C. Otten, A. Christ, T. E. Willnow, I. E. Blasig, and S. Abdelilah-Seyfried Establishment of a neuroepithelial barrier by Claudin5a is essential for zebrafish brain ventricular lumen expansion PNAS, January 26, 2010; 107(4): 1425 - 1430. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. L. Robertson, J. G. Smedley III, and B. A. McClane Identification of a Claudin-4 Residue Important for Mediating the Host Cell Binding and Action of Clostridium perfringens Enterotoxin Infect. Immun., January 1, 2010; 78(1): 505 - 517. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Kimura, H. Abe, S. Kamitani, H. Toshima, A. Fukui, M. Miyake, Y. Kamata, Y. Sugita-Konishi, S. Yamamoto, and Y. Horiguchi Clostridium perfringens Enterotoxin Interacts with Claudins via Electrostatic Attraction J. Biol. Chem., January 1, 2010; 285(1): 401 - 408. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Tatum, Y. Zhang, K. Salleng, Z. Lu, J.-J. Lin, Q. Lu, B. G. Jeansonne, L. Ding, and Y.-H. Chen Renal salt wasting and chronic dehydration in claudin-7-deficient mice Am J Physiol Renal Physiol, January 1, 2010; 298(1): F24 - F34. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. M.K. Morrow, G. Tyagi, L. Simon, K. Carnes, K. M. Murphy, P. S. Cooke, M.-C. C. Hofmann, and R. A. Hess Claudin 5 Expression in Mouse Seminiferous Epithelium Is Dependent upon the Transcription Factor Ets Variant 5 and Contributes to Blood-Testis Barrier Function Biol Reprod, November 1, 2009; 81(5): 871 - 879. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Campbell, A. T. H. Nguyen, A.-S. Kiang, L. C. S. Tam, O. L. Gobbo, C. Kerskens, S. N. Dhubhghaill, M. M. Humphries, G.-J. Farrar, P. F. Kenna, et al. An experimental platform for systemic drug delivery to the retina PNAS, October 20, 2009; 106(42): 17817 - 17822. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Angelow and A. S. L. Yu Structure-Function Studies of Claudin Extracellular Domains by Cysteine-scanning Mutagenesis J. Biol. Chem., October 16, 2009; 284(42): 29205 - 29217. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Hou, A. Renigunta, A. S. Gomes, M. Hou, D. L. Paul, S. Waldegger, and D. A. Goodenough Claudin-16 and claudin-19 interaction is required for their assembly into tight junctions and for renal reabsorption of magnesium PNAS, September 8, 2009; 106(36): 15350 - 15355. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. M. Krug, S. Amasheh, J. F. Richter, S. Milatz, D. Gunzel, J. K. Westphal, O. Huber, J. D. Schulzke, and M. Fromm Tricellulin Forms a Barrier to Macromolecules in Tricellular Tight Junctions without Affecting Ion Permeability Mol. Biol. Cell, August 15, 2009; 20(16): 3713 - 3724. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Winkler, C. Gehring, A. Wenzel, S. L. Muller, C. Piehl, G. Krause, I. E. Blasig, and J. Piontek Molecular Determinants of the Interaction between Clostridium perfringens Enterotoxin Fragments and Claudin-3 J. Biol. Chem., July 10, 2009; 284(28): 18863 - 18872. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. J. Mee, H. J. Harris, M. J. Farquhar, G. Wilson, G. Reynolds, C. Davis, S. C. D. van IJzendoorn, P. Balfe, and J. A. McKeating Polarization Restricts Hepatitis C Virus Entry into HepG2 Hepatoma Cells J. Virol., June 15, 2009; 83(12): 6211 - 6221. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Yoshida, Y. Ban, and S. Kinoshita Tight Junction Transmembrane Protein Claudin Subtype Expression and Distribution in Human Corneal and Conjunctival Epithelium Invest. Ophthalmol. Vis. Sci., May 1, 2009; 50(5): 2103 - 2108. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Wolburg, S. Noell, K. Wolburg-Buchholz, A. Mack, and P. Fallier-Becker Agrin, Aquaporin-4, and Astrocyte Polarity as an Important Feature of the Blood-Brain Barrier Neuroscientist, April 1, 2009; 15(2): 180 - 193. [Abstract] [PDF]
|
|
|
|
|
|
S. Liu, W. Yang, L. Shen, J. R. Turner, C. B. Coyne, and T. Wang Tight Junction Proteins Claudin-1 and Occludin Control Hepatitis C Virus Entry and Are Downregulated during Infection To Prevent Superinfection J. Virol., February 15, 2009; 83(4): 2011 - 2014. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J M Timpe and J A McKeating Hepatitis C virus entry: possible targets for therapy Gut, December 1, 2008; 57(12): 1728 - 1737. [Full Text] [PDF]
|
|
|
|
|
|
D. B. N. Lee, N. Jamgotchian, S. G. Allen, M. B. Abeles, and H. J. Ward A lipid-protein hybrid model for tight junction Am J Physiol Renal Physiol, December 1, 2008; 295(6): F1601 - F1612. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Angelow, R. Ahlstrom, and A. S. L. Yu Biology of claudins Am J Physiol Renal Physiol, October 1, 2008; 295(4): F867 - F876. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. W. McColl, N. J. Rothwell, and S. M. Allan Systemic Inflammation Alters the Kinetics of Cerebrovascular Tight Junction Disruption after Experimental Stroke in Mice J. Neurosci., September 17, 2008; 28(38): 9451 - 9462. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. J. Harris, M. J. Farquhar, C. J. Mee, C. Davis, G. M. Reynolds, A. Jennings, K. Hu, F. Yuan, H. Deng, S. G. Hubscher, et al. CD81 and Claudin 1 Coreceptor Association: Role in Hepatitis C Virus Entry J. Virol., May 15, 2008; 82(10): 5007 - 5020. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. J. Mrsny, G. T. Brown, K. Gerner-Smidt, A. G. Buret, J. B. Meddings, C. Quan, M. Koval, and A. Nusrat A Key Claudin Extracellular Loop Domain is Critical for Epithelial Barrier Integrity Am. J. Pathol., April 1, 2008; 172(4): 905 - 915. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
