Endothelial Claudin: Claudin-5/Tmvcf Constitutes Tight Junction Strands in Endothelial Cells

Rat antioccludin mAb (MOC37) was described previously (Saitou et al. 1997). Anti–VE-cadherin mAb was a kind gift from Dr. N. Matsuyoshi (Kyoto University, Kyoto, Japan). Anti–platelet endothelial cell adhesion molecule (PECAM) polyclonal antibody (pAb) was purchased from PharMingen. Mouse L cells and their transfectants were cultured in DME supplemented with 10% FCS.

Two polypeptides, KYSAPRRPTANGDYDKKNYV and YSTSVPHSRGPSEYPTKNYV, which correspond to the COOH-terminal cytoplasmic domains of mouse claudin-5/TMVCF (aa 199–218) and claudin-6 (aa 200–219), respectively (a cysteine residue was added at their NH₂ termini), were synthesized and coupled via the cysteine residue to keyhole limpet hemocyanin. These peptides were injected into rabbits as antigens. Rabbit antisera against the former and latter peptides were affinity-purified on nitrocellulose membranes with glutathione S-transferase (GST) fusion proteins with claudin-5/TMVCF and claudin-6Δ to obtain anti–claudin-5/6 pAb and anti–claudin-6 pAb, respectively (see Fig. 1 A).

To construct claudin-5/TMVCF or claudin-6 expression vectors with or without a FLAG-tag at their COOH termini, EcoRI sites were introduced at 3′ ends of claudin cDNAs by PCR, and amplified fragments were subcloned into pBluescript SK(−)–FLAG or SK(−). The inserts were excised by SaII-XbaI digestion followed by blunting with T4 polymerase, and then introduced into pCAGGSneodelEcoRI (Niwa et al. 1991), provided by Dr. J. Miyazaki (Osaka University, Osaka, Japan).

Mouse L cells were used for transfection. Aliquots of 1 μg of each expression vector were introduced into L cells in 1 ml of DMEM using Lipofectamine Plus (GIBCO BRL). After 24- or 48-h incubation, cells were replated and cultured in DMEM containing 10% FCS and 500 μg/ml of Geneticin (GIBCO BRL) to select stable transfectants.

For whole-mount staining, mouse 12.5-d embryos were killed. Samples were pretreated by microwaving in PBS for 20 s and fixed in 4% paraformaldehyde/PBS for 30 min. They were dehydrated in methanol and bleached with 30% H₂O₂. Samples were then rehydrated, blocked with PBS-MT (0.2% Triton X-100 and 1% skimmed milk/PBS) and incubated overnight with primary antibodies followed by secondary antibodies. HRP-conjugated goat anti–rabbit Ig (Chemicon International, Inc.) was used as a secondary antibody. They were then washed with PBS-MT and PBS-T (0.2% Triton X-100/PBS), each for 5 h. Bound antibodies were visualized by incubating with 0.025% diaminobenzidene, 0.08% NiCl₂, and 30% H₂O₂ in PBS-T.

Mouse brain, lung, kidney, and intestine were frozen using liquid nitrogen. Frozen sections ∼6 μm thick were cut on a cryostat, mounted on glass slides, air-dried, and fixed in 95% ethanol at 4°C for 30 min followed by 100% acetone at room temperature for 1 min. They were then rinsed in PBS containing 0.2% Triton X-100 for 15 min, blocked with 1% BSA/PBS for 15 min, and incubated with primary antibodies. After washing with PBS three times, samples were incubated for 30 min with secondary antibodies. Cy3-conjugated goat anti–rat Ig (Amersham Pharmacia Biotech) and Cy2-conjugated goat anti–rabbit Ig (Jackson ImmunoResearch Laboratories, Inc.) were used as secondary antibodies. Samples were washed three times with PBS, then mounted in 90% glycerol/PBS containing 0.1% paraphenylenediamine and 1% n-propylgalate. Specimens were observed using a Zeiss Axiophot photomicroscope (Carl Zeiss, Inc.).

Lysates of Escherichia coli expressing GST–claudin fusion proteins (Morita et al. 1999a) were subjected to one-dimensional SDS-PAGE (12.5%) according to the method of Laemmli 1970, and gels were stained with Coomassie brilliant blue R-250. For immunoblotting, proteins were electrophoretically transferred from gels onto nitrocellulose membranes, which were then incubated with the first antibody. Bound antibodies were detected with biotinylated secondary antibodies and streptavidin-conjugated alkaline phosphatase (Amersham Pharmacia Biotech). Nitroblue tetrazolium and bromochloroindolyl phosphate were used as substrates for detection of alkaline phosphatase.

Immunoelectron microscopy to examine freeze–fracture replicas was performed as described (Fujimoto 1995). The mouse lung was cut into small pieces and quickly frozen in high pressure liquid nitrogen with an HPM 010 High Pressure Freezer (BAL-TEC). The frozen samples were fractured at −110°C and platinum-shadowed unidirectionally at an angle of 45° kin Balzers Freeze–Etch System (BAF 060; BAL-TEC). The samples were immersed in a sample lysis buffer containing 2.5% SDS, 10 mM Tris-HCl, and 0.6 M sucrose (pH 8.2) for 12 h at room temperature, and then replicas floating off the samples were washed with PBS. Under these conditions, integral membrane proteins were captured by replicas, and their cytoplasmic domains were accessible to antibodies. The replicas were incubated with anti–claudin-5/6 pAb for 60 min, then washed with PBS several times. They were then incubated with goat anti–rabbit Ig coupled to 10 nm gold (Amersham Pharmacia Biotech). The samples were washed with PBS, picked up on formvar-filmed grids, and examined in a JEOL 1200EX electron microscope at an accelerating voltage of 100 kV.

The GST fusion protein with the cytoplasmic domain of claudin-5/TMVCF (see Fig. 1 A) was produced in E. coli and used as an antigen to generate specific pAbs in rabbits. Several pAbs that recognized claudin-5/TMVCF were obtained, but all of them cross-reacted with claudin-6 on immunoblotting (Fig. 1 B) as well as immunofluorescence microscopy (Fig. 1 C). As shown in Fig. 1 A, the COOH-terminal KNYV sequence was shared between claudin-5/TMVCF and claudin-6. The GST fusion protein with the cytoplasmic domain of claudin-6 lacking these four aa (GST–claudin-6Δ) was not detected by these pAbs (Fig. 1 B), indicating that they specifically recognized the COOH-terminal KNYV. These pAbs were then referred to as anti–claudin-5/6 pAbs. On the other hand, when the GST fusion protein with the cytoplasmic domain of claudin-6 was used as an antigen, several pAbs, which recognized claudin-6 but not claudin-5/TMVCF on immunoblotting (Fig. 1 B) as well as immunofluorescence microscopy (Fig. 1 C), were obtained (anti–claudin-6 pAb). As expected, these pAbs recognized GST–claudin-6Δ (Fig. 1 B). Therefore, to examine the expression and localization of claudin-5/TMVCF in various tissues, these anti–claudin-5/6 pAbs and anti–claudin-6 pAbs were used in combination; if some cells were anti–claudin-5/6 pAb-positive and anti–claudin-6 pAb-negative, we concluded that they expressed claudin-5/TMVCF. Fortunately, Northern blotting revealed that in most organs of adult mice the expression of claudin-6 was fairly restricted (Morita et al. 1999a).

We first examined the distribution of claudin-5/TMVCF in the brain, which does not contain epithelial cells but expressed claudin-5/TMVCF as detected by Northern blotting (Morita et al. 1999a). When 12.5-d mouse embryos were labeled by whole-mount immunostaining with anti–claudin-5/6 pAb, a characteristic tree-like staining pattern was detected in the head portion (Fig. 2, a and b). Anti–claudin-6 pAb gave no significant staining in the head portion (Fig. 2 c), and a pAb specific for PECAM, a specific marker for endothelial cells of blood vessels (Risau and Flamme 1995), showed a very similar staining pattern to anti–claudin-5/6 pAb (Fig. 2 d). These findings indicated that claudin-5/TMVCF was expressed exclusively in endothelial cells of all segments of blood vessels in the fetal brain.

Frozen sections of adult brain were then immunofluorescently double stained with anti–claudin-5/6 pAb and anti–VE-cadherin mAb. VE-cadherin was reported to be specifically expressed in endothelial cells (Lampugnani et al. 1995). As shown in Fig. 2e and Fig. f, all blood vessels were exclusively costained with these antibodies. These anti–VE-cadherin mAb-positive endothelial cells were negative for staining with anti–claudin-6 pAb (Fig. 2g and Fig. h). Furthermore, at higher magnification, both claudin-5/TMVCF and VE-cadherin were shown to be precisely coconcentrated at cell–cell borders of endothelial cells of blood vessels (Fig. 3). Taken all together, we concluded that claudin-5/TMVCF is expressed and localized at cell–cell contact sites of endothelial cells of all segments of blood vessels in the adult brain.

Next, we examined the distribution of claudin-5/TMVCF in the lungs, which contained both epithelial and endothelial cells and expressed fairly large amounts of claudin-5/TMVCF as detected by Northern blotting (Morita et al. 1999a). When frozen sections of the lung were double stained with anti–claudin-5/6 pAb and anti–VE-cadherin mAb, both signals completely overlapped (Fig. 4, a–c). In contrast, on double staining with anti–claudin-5/6 pAb and antioccludin mAb, the two signals did not overlap at all (Fig. 4, d–f). Furthermore, anti–claudin-6 pAb gave no detectable signals (data not shown). Considering that occludin was expressed in epithelial cells but not in most of the endothelial cells in nonneuronal tissues, these findings indicated that claudin-5/TMVCF was expressed in endothelial cells of all segments of blood vessels but not in epithelial cells delineating the alveolar space in the lung. Furthermore, immunoreplica analysis with anti–claudin-5/6 pAb revealed that claudin-5/TMVCF was localized on the TJ strands of the endothelial cells of the lung (Fig. 4 g).

In the kidney, claudin-5/TMVCF did not appear to be expressed in endothelial cells of all segments of the blood vessels (Fig. 5). When frozen sections of the kidney cortex were doubly-stained with anti–claudin-5/6 pAb and antioccludin mAb, some blood vessels running in the connective tissues surrounding renal tubules were stained with anti–claudin-5/6 pAb but not with antioccludin mAb (Fig. 5, a and b). Instead, antioccludin mAb labeled TJs of distal tubules intensely (Fig. 5 b). Anti–claudin-6 pAb showed no significant signals (Fig. 5 e), indicating that only claudin-5/TMVCF was detected in the kidney by the anti–claudin-5/6 pAb. Judging from the density/number of the claudin-5/TMVCF–positive blood vessels, only some selected vessels appeared to be stained with anti–claudin-5/6 pAb. Frozen sections of the kidney were double stained with anti–claudin-5/6 pAb and anti–VE-cadherin mAb (Fig. 5c and Fig. d). Anti–VE-cadherin mAb stained numerous capillaries surrounding renal tubules and in glomeruli. These intertubular and glomerular capillaries were not stained by anti–claudin-5/6 pAb. Interestingly, afferent as well as efferent arterioles of glomeruli were reproducibly stained with anti–claudin-5/6 pAb (Fig. 5c and Fig. d). Furthermore, thicker arteries (probably interlobular arteries) but not veins were also intensely stained with anti–claudin-5/6 pAb (Fig. 5g and Fig. h). Taken together, we concluded that in the kidney, claudin-5/TMVCF constituted TJs only in endothelial cells of the arteries.

The expression and distribution of claudin-5/TMVCF were also examined in other organs, among them the intestine, skeletal muscle, skin, liver, testis. In these tissues claudin-5/TMVCF was specifically detected in endothelial cells in some segments of blood vessels, but not in epithelial cells. The endothelial cell–specific localization of claudin-5/TMVCF in the intestine was presented in Fig. 5i and Fig. j.

We reported previously that claudin-1 and -2 reconstituted TJ strands between adjacent transfectants when introduced into L fibroblasts (Furuse et al. 1998b). Then, we examined the ability of claudin-5/TMVCF to reconstitute TJ strands in L fibroblasts. When stable L transfectants expressing claudin-5/TMVCF were stained with anti–claudin-5/6 pAb, claudin-5/TMVCF was shown to be concentrated at cell–cell borders as planes, similar to claudin-1 and -2 (Fig. 6, a and b). Conventional freeze–fracture electron microscopy of glutaraldehyde-fixed L transfectants revealed that huge well-developed networks of TJ strands were formed in the cell–cell contact planes (Fig. 6 c). As shown previously in L transfectants expressing claudin-1 and -2 (Furuse et al. 1998b), claudin-1–induced strands were largely associated with the P-face as mostly continuous structures with vacant grooves at the E-face (P-face–associated TJs), whereas claudin-2–induced strands were discontinuous at the P-face with complementary grooves at the E-face that were occupied by chains of particles (E-face–associated TJs). As shown in Fig. 6 c, the claudin-5/TMVCF–induced TJs were an extreme case of the E-face–associated TJs; almost all particles were associated with the E-face, leaving particle-free ridges on the P-face (Fig. 6 c, insets).

TJs are thought to be involved in the barrier and fence functions in both epithelial and endothelial cells, but to date no differences have been reported in the molecular architecture of TJs between epithelial and endothelial cells (Rubin 1992; Schneeberger and Lynch 1992; Bowman et al. 1992; Goodenough 1999). In this study, we found that claudin-5/TMVCF was expressed specifically in endothelial cells, not in epithelial cells, and that this molecule constituted TJ strands in endothelial cells. The human TMVCF gene was originally found to be localized to chromosome 22q11, which is frequently deleted in velo-cardio-facial/DiGeorge syndrome patients (Sirotkin et al. 1997), and we proposed to designate it as claudin-5/TMVCF since it showed significant similarity to claudin-1 and -2 (Morita et al. 1999a). This gene was also identified as a membrane protein named MBEC1 that was expressed in the brain endothelial cells (Chen et al. 1998). In both studies, the concentration of the product of TMVCF/MBEC1 at TJs was not examined, probably due to the difficulty generating antibodies specific for this protein. We avoided this technical difficulty by generating and using anti–claudin-5/6 pAb and anti–claudin-6 pAb in combination. Immunofluorescence microscopy with these pAbs revealed that in the brain and the lung, the endothelial cells of all segments of blood vessels, which were positive in the anti–VE-cadherin mAb staining, expressed significant amounts of claudin-5/TMVCF. In contrast, in the kidney, veins and capillaries lacked the expression of claudin-5/TMVCF, and its expression was restricted to arteries. In addition, in all other tissues we examined, the expression of claudin-5/TMVCF was restricted to endothelial cells of some segments of blood vessels. Therefore, we concluded that claudin-5/TMVCF is an endothelial cell–specific component of TJ strands, but that in contrast to VE-cadherin (Lampugnani et al. 1995), claudin-5/TMVCF is not expressed in all types of endothelial cells.

Since it is widely accepted that TJs play central roles in the regulation of vascular permeability, the structure of endothelial TJs has been extensively examined, mainly by freeze–fracture electron microscopy (Dempsey and Bullivant 1973; Simionescu et al. 1975, Simionescu et al. 1976; Schneeberger 1982; Bowman et al. 1992). In general, in glutaraldehyde-fixed endothelial cells in nonneuronal tissues, TJ strands in endothelial cells were characterized by particle-free ridges on the P-face and continuous grooves at the E-face, which were densely occupied by chains of particles (E-face–associated TJs) (Dempsey and Bullivant 1973). Interestingly, the claudin-5/TMVCF–based TJ strands reconstituted in L transfectants showed the same morphological characteristics (see Fig. 6), favoring the notion that claudin-5/TMVCF is a major constituent of endothelial TJ strands of nonneuronal tissues in situ. Furthermore, in previous freeze–fracture studies, TJs were shown to be developed also in veins and capillaries in various tissues (Simionescu et al. 1975, Simionescu et al. 1976; Schneeberger 1982; Bowman et al. 1992). Therefore, some species of claudins other than claudin-5/TMVCF must be involved in the formation of TJ strands in veins and capillaries. To date, in addition to pAbs used in this study, pAbs specific for claudin-2, -3, -4, -8, -11, and -14 were available. Preliminary immunofluorescence microscopy with these pAbs showed that these claudins were not detected in endothelial cells in all of the organs we examined. Among the other claudin species (claudin-1, -7, -9, -10, -12, -13, and -15), claudin-1 is expected to be expressed in endothelial cells, since this claudin species is expressed rather ubiquitously, even in organs lacking epithelial tissues as shown by Northern blotting (Morita et al. 1999a). However, in contrast to claudin-5/TMVCF, claudin-1 appeared to be expressed also in epithelial cells, because this claudin species was first identified from hepatocytes and was expressed in large amounts in the liver and the kidney (Furuse et al. 1998a). Further identification and characterization of claudin species expressed in endothelial cells will be important in future studies.

Endothelial cells in the adult brain, which form the blood–brain barrier (BBB), are coupled by TJs of extremely low permeability that are more like those of epithelial barriers (Reese and Karnovsky 1967; Brightman and Reese 1969; Risau and Wolburg 1990; Rubin 1992). Despite the large numbers of in vivo studies that have been performed, the developmental timing of the formation and maturation of BBB in vivo is still controversial, but the TJs of brain endothelial cells are thought to be leaky in embryos (Møllgård and Saunders 1975; Farrell and Risau 1994). As shown in this study, claudin-5/TMVCF was clearly detected in endothelial cells of both the fetal and adult brain (see Fig. 2), suggesting that the formation and maturation of BBB is not attributable to the developmental changes in the expression level of claudin-5/TMVCF. TJ strands of endothelial cells in the fetal brain are E-face–associated similarly to those in nonneuronal tissues as discussed above, whereas those in the adult brain are P-face–associated (Wolburg et al. 1994; Kniesel et al. 1996). Therefore, it is tempting to speculate that during development the claudin-5/TMVCF–based, E-face–associated TJ strands are modified by addition of other claudin species or by some inside-out signaling, to be switched to the P-face–associated forms.

Although occludin and claudins were identified as components of epithelial TJ strands (Furuse et al. 1993; Ando-Akatsuka et al. 1996; Furuse et al. 1998a; Morita et al. 1999a), our knowledge of the molecular architecture of endothelial TJs is limited. Occludin was reported to be expressed in large amounts in brain endothelial cells, but was undetectable in most endothelial cells in nonneuronal tissues (Saitou et al. 1997; Hirase et al. 1997). Now that claudin-5/TMVCF has been identified as a specific component of endothelial TJ strands, it will be possible to examine and modulate the mechanism of regulation of vascular permeability in molecular terms. For example, vascular endothelial cell growth factor has been shown to elevate vascular permeability through binding to its tyrosine kinase–type receptors flt-1 and flk-1 (Risau et al. 1998; Gale and Yancopoulos 1999), and to cause disorganization of interendothelial junctions (Kevil et al. 1998). Thus, it would be intriguing to examine whether and how activation of these receptors modulates claudin-5/TMVCF. Furthermore, it will be important to analyze the possible involvement of claudin-5/TMVCF in vasogenic edema and inflammation in various pathological states. Further detailed analyses of claudin-5/TMVCF along these lines will lead to a better understanding of the molecular mechanisms behind vascular permeability.

We thank Drs. H. Yoshida and S.I. Nishikawa for their technical help with whole-mount immunostaining of mouse embryos. Our thanks are also due to all the members of our laboratory (Department of Cell Biology, Faculty of Medicine, Kyoto University) for helpful discussions. K. Morita thanks Prof. Y. Miyachi (Department of Dermatology, Faculty of Medicine, Kyoto University) for the opportunity to work in the Department of Cell Biology.

This study was supported in part by a Grant-in-Aid for Cancer Research and a Grant-in-Aid for Scientific Research (A) from the Ministry of Education, Science and Culture of Japan.

Note Added in Proof. Positional cloning has identified a new member of the claudin family (paracellin-1), mutations of which cause hereditary renal hypomagnesemia in humans (Simon, D.B., Y. Lu, K.A. Choate, H. Velazquez, E. Al-Sabban, M. Praga, G. Casari, A. Bettinelli, G. Colussi, J. Rodriguez-Soriano, et al. 1999. Paracellin-1, a renal tight junction protein required for paracellular Mg²⁺ resorption. Science. 285:103–106). This new claudin species would be called claudin-16.