Graefe's Arch Clin Exp Ophthalmol (1994) 232: 608-613 © Springer-Verlag 1994

Walter Noske B6atrice Levarlet Klaus Martin Kreusel Michael Fromm Michel Hirsch

Tight junctions and paraceUular permeability in cultured bovine corneal endothelial cells

Received: 4 January 1994 Accepted: 11 April 1994

The results reported here were presented in part at the 91st Congress of the German Ophthalmological Society, Mannheim, 19-22 September 1993 and published in abstract form [22].

W. Noske - K.M. Kreusel Augenklinik, Klinikum Steglitz, Freie Universit~it Berlin, D-12200 Berlin, Germany

B. Levarlet • M. Hirsch ([~) INSERM U86, H6tel Dieu, 1 place du Parvis Notre Dame, F-75181 Paris Cedex 04, France

K.M. Kreusel • M. Fromm Institut ftir Klinische Physiologie, Klinikum Steglitz, Freie Universit/it Berlin, D-12200 Berlin, Germany

Abstract Intramembrane special- izations of cultured bovine corneal endothelial cells were studied with thin section and freeze-fracture electron microscopy and related to the paracellular permeability and the transendothelial resistance (R t) of the monolayers. The following intercellular junctions were found: single and discontinuous networks of tight junctions (TJ) which girdle the apico-lateral cell perimeter in- completely, gap junctions, and membrane undulations suggesting intermediate junctions. The macro- molecular tracer ruthenium red penetrated into the lateral intercel- lular space beyond the level of the incomplete belt of TJ. R t of these monolayers was 20.9 _+ 1.0 f~. cm 2.

Protamine induced a reversible in- crease of R t to 118 ± 5% of its con- trol value. We conclude that in- complete belts of TJ may be the morphological counterpart of the high paracellular permeability of this monolayer and functionally and morphologically resemble those of their native endothelium. Cultured corneal endothelial cells are an excellent model for studying the influence of incomplete belts of TJ on paracellular permeability of cells.

Introduction

The tight junctions (TJ) are specialized plasma mem- brane structures that girdle the apical cell perimeter of epithelial and endothelial cells. They are involved in the control of paracellular permeability and in the mainte- nance of apico-basal polarity of the plasma membrane (see [2] for references).

The corneal endothelial monolayer covers the poste- rior face of the cornea and separates the underhydrated environment of the avascular corneal stroma from the aqueous humor. Despite the incompleteness of its TJ belt [10, 26], the endothelial layer of the cornea functions as a barrier to excessive water flow from the anterior chamber into the corneal stroma and effectively pumps small solutes and water from the stromal side into the

aqueous humor, thus controlling corneal hydration and transparency [5, 19, 29].

Cultured corneal endothelial cells can reconstitute a confluent polarized monolayer of polygonal cells which secrete a thick basement membrane and reveal several characteristics of corneal endothelium in vivo [8, 9, 11, 14]. The role of the TJ in regulating the paracellular pathway in this important endothelial layer is not com- pletely understood. Therefore, we studied the morphol- ogy of intercellular junctions in cultured bovine corneal endothelium and related it to the barrier function to- wards ruthenium red and the transendothelial electrical resistance (R t) of the cultures. Furthermore, since it has been shown that protamine, a polycationic protein, in- creases R t in Necturus gallbladder [6], we investigated its effect on the corneal endothelial monolayer.

609

Materials and methods Cell culture

Bovine corneas were obtained from the slaughterhouse. The en- dothelial layer was scraped in phosphate-buffered saline and cul- tured in Dulbecco's modification of Eagle's minimum essential medium (DMEM) supplemented with 10% fetal calf serum, 100 U/ml of penicillin and 25 gg/ml of gentamycin at 37 ° C and gassed with 5% CO 2 in air on commercial cell culture dishes with microporous membranes (Millicell-HA filter, area 0.6 cm 2, Mil- lipore, Bedford, Mass.) or on plastic coverslips (Thermanox, Sig- ma, St. Louis, Mo.). All experiments were performed on 10- to 15-day-old confluent subcultures.

Thin-section electron microscopy

Monolayers were fixed in 2.5% glutaraldehyde in 0.1 M cacody- late buffer, pH 7.4, at room temperature, rinsed in the buffer and post-fixed in 1% osmium tetroxide in the same buffer. In some cultures tannic acid (1%) or ruthenium red (1%) was added to the fixative at the apical side of the monolayer. After dehydration in ethanol, the cells were embedded in Epon 812. Thin sections were contrasted with lead citrate and uranyl acetate and observed with a Philips EM 300 or EM 410 electron microscope.

Freeze-fracture electron microscopy

Glutaraldehyde-fixed samples were cryoprotected in 25% glycerol in 0.1 M cacodylate buffer for 1 h. The supports with the attached monolayers were rapidly frozen in solid-liquid nitrogen. Plat- inum-carbon replicas were produced in a freeze-fracture appara- tus (Cryofract CF 250, Leica, France) equipped with electron beam guns and a quartz thickness film monitor at a stage temper- ature of - 150 ° C and in a vacuum of at least 10 -7 torr. Cleaned replicas were mounted on 300 mesh copper grids and observed with a Philips EM 300 or EM 410 electron microscope.

Measurement of R t

In order to monitor R t, the filters with confluent monolayers were mounted into Ussing-type chambers [16]. The hemi-chambers were filled with Ringer's solution (in mmol/l: 151 Na +, 5 K +, 130.4 CI, 1.7 Ca 2+, 0.9 Mg 2+, 1 H2PO4-, 0.9 SO42-, 28 HCO3; pH 7.4, 37 ° C) and gassed with 95% 02 5% CO2 and stirred by a bubble lift. The monolayers were short-circuited by a computer- ized automatic voltage clamp device (CVC 6, Hard & Software, O. Fiebig, Berlin, Germany), allowing continuous monitoring of R t. Protamine (Hoffmann-La Roche, Grenzach-Whylen, Switzerland) was used from a 1000 IU/ml stock solution and added to the apical side of the monolayer to give a final concentration of 3 x 10 -5 M, and its effect was reversed by 6 x 10 5 M heparin (Sigma, Munich, Germany). Data are means+standard errors of the means (SEM), and significances were evaluated using the unpaired t-test.

Results

Cultured bovine corneal endothelial cells reconsti tuted confluent monolayers of polygonal cells. The apico-lat- eral p lasma membranes between adjacent cells interdig- itated, and in places the external leaflets of the two inter-

Fig. 1 Apical part of cultured endothelium fixed in the presence of tannic acid. At places (arrows) the external leaflets of the adja- cent plasma membranes seem to fuse, indicating the location of apico-lateral TJ. IS = intercellular space. Bar 0.2 gm

digitating p lasma membranes seemed to "fuse", indicat- ing the presence of TJ (Fig. 1). At these points no tannic acid was revealed between the two opposing mem- branes, and at times the intercellular space seemed to be bridged by tiny white spots.

With the freeze-fracture technique, large areas of the different p lasma membrane regions with their respective membrane specializations were exposed. In glutaralde- hyde-fixed cultured corneal endothelial cells, the TJ structures appeared as smooth strands on the proto- plasmic face (P-face) of the fracture (Fig. 2a) and as fur- rows on the exoplasmic face (E-face) of the fracture (Fig. 2b) in the apico-lateral p lasma membrane . The ge- ometrical pa t tern of the TJ network varied consider- ably, even within short distances along the cell perime- ter. TJ often consisted of most ly unbranched P-face strands or E-face furrows, often oriented in an apico- basal direction (Fig. 2a). Therefore, the apical and baso- lateral p lasma membrane domains often were in conti- nuity between these strands without interspersed junc- tional elements. At other positions, TJ strands or fur- rows formed more or less loose networks (fasciae or maculae occludentes) with varying numbers of anasto- moses (Figs. 2b, 3), forming a local "fence" between the two p lasma membrane domains. Very elaborated net- works of TJ strands or furrows girdling long distances of the apical perimeter of the lateral p lasma membrane were not revealed. Therefore, when sufficiently large ar- eas of the apical lateral p lasma membrane were ex- posed, the apical and lateral m e m b r a n e domain were

610

Fig. 2 Apico-basally oriented TJ strands on a fracture P- face (P) with few anastomoses and many free ends (a) and small anastomosing TJ fur- rows on the fracture E-face (E) forming maculae occluden- tes associated with gap junc- tions (b). Arrows indicate regions of the plasma mem- brane not barred with any TJ elements. GJ = gap junc- tion. (a) Bar 0.5 p~m; (b) Bar i Ixm

always in continuity at some areas without interposed junctional specializations.

Sometimes the exposed plasma membrane displayed linear or curved undulations around the apico-lateral cell perimeter. The density of in t ramembrane particles was lower in these regions than in the surrounding plas- ma membrane (Fig. 4). No typical TJ elements were present at these undulations, but small gap junction-like condensations of in t ramembranous particles were often associated with these membrane areas. These zones may represent intermediate junctions.

Frequently, typical gap junctions were intimately as- sociated with TJ, especially with the more elaborated networks (Figs. 2b, 3), whereas other gap junctions were found isolated in the lateral plasma membrane. Clusters of vesicle fusion sites were observed isolated on the api- cal plasma membrane (Fig. 5a), but were especially nu- merous on the basal plasma membrane, at times prefer-

Fig. 3 Anastomosing TJ network with gap junctions (GJ) not demonstrating discontinuities in the network. P = Fracture P-face, E=Fracture E-face. Bar 0.5 pm

611

Fig. 4 Fracture P-face (P) of the apico-lateral plasma membrane with low mem- brane undulations (arrow- heads) which are relatively poor in intramembrane parti- cles. GJ = gap junction. Bar 0.5 gm

Fig. 5 Fracture E-face (E) of the apical plasma membrane (a) and fracture P-face (P) of the basal plasma membrane (b) showing numerous vesicle fusion sites (thick arrows). The shallow circular elevations (thin arrow) probably represent coated pits. Bars 1 mm

entially oriented in linear rows (Fig. 5b). We did not find increased numbers of vesicle fusion sites in m e m b r a n e areas where TJ elements are normal ly located.

When the macromolecu la r t racer ru thenium red was

Fig. 6 The tracer ruthenium red applied to the apical side (ap) of the monolayer stains the entire intercellular space (arrowheads), indicating the leakiness of the apical TJ towards this electron- dense tracer. Bar 2 gm

added to the apical side in order to test the barrier char- acteristics of the endothelial monolayer , it was often found in the lateral intercellular space beyond the apical region of close m e m b r a n e contacts, indicating the leaki- ness of the paracellular pa thway for ru thenium red in this monolayer (Fig. 6).

R t was used as a measure of small solute permeabil i ty of the monolayers . R t of 10 day-old monolayers was

612

Table 1 Transendothelial resistance (R t) of cultured corneal en- dothelial monolayers and the effect of protamine. (Protamine (3 x 10 -5 M) was added after stabilization of R t. Protamine values were read 20 rain after its addition, just before addition of heparin (6 x 10 -s M). Heparin values were read after 10 rain when its effect was complete. Values represent mean_+ SEM of five monolayers.)

g t (~'~" c m 2) P vs control

Control 20.9 + 1.0 Protamine 24.6 _+ 1.0 < 0.05 Heparin 19.8 + 0.9 n.s.

30

25

E o

20

15

Protemine Heplrin

l ! I i i l

40 60 80 100 120 140

time (min)

Fig. 7 Increase in R t after addition of protamine (3 × 10-SM), rapidly reversed by the addition of heparin (6 x 10 5 M), in cul- tured corneal endothelial cells

20.9_ 1.0 f 2 - c m 2 and remained stable for at least 2 h (Table 1). In order to test the regulation of the paracellu- lar permeability by the T J, protamine (3 x 10 -5 M) was added to the monolayer. A rapid rise in R t to 118 _+ 5% of the control value was recorded. This effect of pro- tamine was reversed by heparin (6 x 10 -5 M) (Fig. 7).

Discussion

Despite the presence of apico-lateral membrane "fu- sion" sites, the apically applied macromolecular tracer ruthenium red penetrated the lateral intercellular space of cultured bovine corneal endothelial cells, indicating that the paracellular pathway is permeable to relatively large molecules. Passage of horseradish peroxidase (row 40 000) [15] and ruthenium red (mw 551) [4] across the apical TJ domain of corneal endothelial cells has also been found in vivo. This correlates well with the freeze- fracture images of the apico-lateral T J, since the TJ strands form incomplete belts around the cell perimeter in vivo [10, 22, 26] and in cultures [11]. The discontinu- ities in the TJ network may be the morphological corre-

late of the high permeability of the paracellular pathway in this monolayer. The relatively low R t values mea- sured in this study further support this interpretation. Taking into account the influences on R t of the support on which the cells grow [8, 20] and the additives to the medium, such as glucose and adenosine [5], the R t val- ues of our study correlate well with values found in oth- er studies on cultured corneal endothelium [8] and in whole mounted corneas [12, 17], but are lower than the exceptionally high R t values reported by Narula et al. [20].

In the leaky epithelium of Necturus gallbladder, pro- tamine has been shown to cause a rapid increase in R t without affecting apical membrane permeability, selec- tively altering the function of the TJ [7]. The observed rapid and reversible increase in R t after addition of pro- tamine in cultured monolayers of corneal endothelial cells supports the interpretation that the incomplete TJ belts limit, at least in part, the paracellular permeability in this monolayer. However, the protamine-induced in- crease in R t (18%) is relatively low compared to the 70-90% increase in R t in Necturus gallbladder with complete belts of TJ [1, 6] and the 84% increase in cul- tured human non-pigmented ciliary epithelial cells with incomplete belts of TJ [21, 27]. Therefore, the contribu- tion of the incomplete TJ network to the total paracellu- lar resistance in corneal endothelial monolayers may be less than in other systems. Furthermore, the relatively high number of vesicle fusion sites in the plasma mem- branes may indicate a very active transcellular route for bulk transport of charged or uncharged molecules across the endothelial layer [24]. This may have to be considered when interpreting transendothelial trans- port and R t of these monolayers.

Incomplete TJ belts have also been encountered in developing [3] and cultured [21] epithelia that normally express complete TJ belts and in cultured cerebral en- dothelia in the absence of astrocyte-derived factors (see [25] for review). Interestingly, discontinuous TJ belts are also found in native corneal endothelium [10, 26]. It is interesting to note that tight vascular endothelia may be formed in the anterior chamber in the presence of astro- cytes [13]. Thus, the abundance of incomplete TJ in the absence of complete TJ belts in cultured corneal en- dothelial cells is probably a reflection of some inherent properties of corneal endothelium rather than due to the in vitro conditions.

The membrane undulations sometimes observed in the apico-lateral areas of the plasma membrane which are poor in intramembranous particles and on which, at times, aggregates of intramembranous particles may be found may correspond to intermediate junctions [18, 23, 28]. These are believed to be a prerequisite of TJ forma- tion and may indicate continuous formation of TJ in 10- to 15-day-old confluent monolayers of bovine corneal endothelial cells.

613

Our results show that cultured corneal endothelial cells morphologically and functionally closely resemble native corneal endothelium. Thus, cultured corneal en- dothelium represents a model for studying incomplete TJ belts and their influence on the paracellular pathway and cellular polarization.

Acknowledgements We would like to thank Drs. M. Wiederholt, S. Berweck and A. Lepple-Wienhues and Mrs. A. Krolik of the Institut ffir Klinische Physiologie, Universit/itsklinikum Steglitz, Germany, and Jacqueline Tassin of Ul18, INSERM, France, for providing cultures of corneal endothelial cells. We further want to thank Ursula Lempart, Frangoise Dagonet, Dani61e Raison and Genevieve Prenant for skilled technical assistance. This research was supported by grants from the French Institut National de la Sant~ et de la Recherche M6dicale.

References

1. Bentzel C J, Fromm M, Palant CE, Hegel U (1987) Protamine alters structure and conductance of Nectu- rus gallbladder tight junctions without major electrical effects on the apical cell membrane. J Membr Biol 95:9-20

2. Cereijido M (ed) (1992) Tight junc- tions. CRC Press, Boca Raton, pp 1- 367

3. Dermietzel R, Meller K, Tetzlaff W, Waelsch M (1977) In vivo and in vitro formation of the junctional complex in choroid epithelium. A freeze-etch- ing study. Cell Tissue Res 181:427- 441

4. Edelhauser HF, Van Horn DL, Records RE (1982) Cornea and sclera. In: Duane TD, Jaeger EA (eds) Biomedical foundations of ophthal- mology, vol 1. Harper & Row, Philadelphia, pp 1-26

5. Fischbarg J, Lim JJ (1984) Fluid and electrolyte transports across corneal endothelium. In: Zadunaisky JA, Davson H (eds) Current topics in eye research. Academic Press, Orlando, pp 201-223

6. Fromm M, Palant CE, Bentzel C J, Hegel U (1985) Protamine reversibly decreases paracellular cation perme- ability in Necturus gallbladder. J Membr Biol 87:141-150

7. Fromm M, Tykocinski M, Schulzke J- D, Hegel U, Bentzel CJ (1990) pH dependence of protamine action on apical membrane permeability in Necturus gallbladder epithelium. Biochim Biophys Acta 1027:179-184

8. Geroski DH, Hadley A (1992) Char- acterization of corneal endothelium cell cultured on microporous mem- brane filters. Curr Eye Res 11:61-72

9. Gospodarowicz D, Greenburg G, Vlo- davsky I, Alvarado J, Johnson LK (1979) The identification and localiza- tion of fibronectin in cultured corneal endothelial cells: cell surface polarity and physiological implications. Exp Eye Res 29:485-509

10. Hirsch M, Renard G, Faure J-P, Pouliquen Y (1977) Study of the ultra- structure of the rabbit corneal en- dothelium by the freeze-fracture tech- nique: apical and lateral junctions. Exp Eye Res 25:277-288

11. Hirsch M, Tassin J, Noske W, Cour- tois Y (1989) Filipin-induced deforma- tions in plasma membranes of cul- tured bovine corneal endothelial cells with incomplete belts of tight junc- tions. Exp Eye Res 49:717-727

12. Hodson S, Wigham C (1983) The per- meability of rabbit and human corneal endothelium. J Physiol (Lond) 342:409M19

13. Janzer RC, Raft MC (t987) Astro- cytes induce blood-brain barrier prop- erties in endothelial cells. Nature 325:253-257

14. Jentsch TJ, Keller SK, Wiederholt M (1985) Ion transport mechanisms in cultured bovine endothelial cells. Curr Eye Res 4:361-369

15. Kaye GI, Sibley RC, Hoefle FB (1973) Recent studies on the nature and function of the corneal endothelial barrier. Exp Eye Res 15:585-610

16. Kreusel K-M, Fromm M, Schulzke J- D, Hegel U (1991) CI secretion in ep- ithelial monolayers of mucus-forming human colon cells (HT-29/B6). Am J Physiol 261:C574-C582

17. Lira JJ, Fischbarg J (1981) Electrical properties of rabbit corneal endotheli- um as determined from impedance measurements. Biophys J 36:677-695

18. Mancel E, Hirsch M (1989) Develop- ment of tight junctions in the human ciliary epithelium. Exp Eye Res 48: 87-97

19. Maurice DM (1984) The cornea and sclera. In: Davson H (ed) The eye. Academic Press, Orlando, pp 1-158

20. Narula P, Xu M, Kuang K, Akiyama R, Fischbarg J (1992) Fluid transport across cultured bovine corneal en- dothelial cell monolayers. Am J Phys- iol 262:C98-C103

21. Noske W, Straug O, Wiederholt M, Hirsch M (1992) Tight junctions of cultured ciliary epithelium. Exp Eye Res 55 [Suppl 1]: 202

22. Noske W, Kreusel KM, Fromm M, Hirsch M (1993) Tight junctions of the corneal endothelium: morphologi- cal and electrophysiological features. German J Opththalmol 2:270

23. Polak-Charcon S, Shoham J, Ben- Shaul Y (1978) Junction formation in trypsinized cells of human adenocar- cinoma cell line. Exp Cell Res 116:1- 13

24. Raviola G, Butler JM (1985) Morpho- logical evidence for the transfer of an- ionic macromelecules from the interi- or of the eye to the blood stream. Curr Eye Res 4:503-516

25. Shivers RR, Arthur FE, Bowman PD (1988) Induction of gap junctions and brain endothelium-like tight junctions in cultured bovine endothelial cells: local control of cell specialization. J Submicrosc Cytol Pathol 20:1-14

26. Stiemke MM, McCartney MD, Can- tu-Crouch D, Edelhauser HF (1991) Maturation of the corneal endothelial tight junction. Invest Ophthalmol Vis Sci 32:2757-2765

27. Straug O, Wiederholt M (1991) Transepithelial resistance of ciliary ep- ithelial cells in culture: functional modification by protamine and extra- cellular calcium. Comp Biochem Physiol 100A: 987-993

28. Townes-Anderson E, Raviola G (1981) The formation and distribution of intercellular junctions in the rhesus monkey optic cup: the early develop- ment of the cilio-iridic and sensory retinas. Dev Biol 85:209-232

29. Waring GO, Bourne WM, Edelhauser HF, Kenyon KR (1982) The corneal endothelium. Normal and pathologic structure and function. Ophthalmolo- gy 89:531-590