A novel cell-cell junction system: the cortex adhaerens mosaic of lens fiber cells

The functions of a specific tissue generally depend on its architecture, which is usually based on dual-function cell-cell adhesion elements. These morphologically distinct plasma membrane domains serve on the one hand as position-specific cell-cell attachment structures and, on the other, provide cytoplasmic anchorage plates (`plaques') for cytoskeletal filament bundles. Here, the `adhering junctions' (e.g. Farquhar and Palade, 1963; Staehelin, 1974), which share some molecular principles because they comprise clusters of glycoproteins of the larger cadherin multigene family (Koch et al., 1990; Takeichi, 1988; Takeichi, 1991) and possess the common arm-repeat plaque protein plakoglobin (Cowin et al., 1986), are usually divided into two major categories with specific, mutually exclusive constituents. (1) The `adherens junctions', pleiomorphic as zonulae adhaerentes, fasciae adhaerentes or puncta adhaerentia, are formed by `classic' cadherins, which on their cytoplasmic side are complexed with a set of plaque proteins dominated by α- and β-catenins, and anchor actin microfilaments (Nagafuchi and Takeichi, 1989; Ozawa et al., 1989). (2) By contrast, the desmosomes (maculae adhaerentes) are assemblies of two special cadherin subgroups (the desmogleins and the desmocollins), with a plaque characterized by desmoplakin and members of the plakophilin subfamily of arm-repeat proteins, which anchors bundles of intermediate-sized filaments (IFs) (cf. Franke et al., 1981; Schmidt et al., 1994; Schmidt et al., 1999; Hatzfeld, 1999; Green and Gaudry, 2000).

It has, however, become increasingly clear over the past decade that there are diverse adhering junctions that cannot be subsumed under these two major categories but represent special structures sui generis. Examples include the complexus adhaerentes described in special vascular endothelia, notably the retothelial cells of lymph node sinus (Schmelz and Franke, 1993; Valiron et al., 1996), the M-/N-cadherin containing contactus adhaerentes connecting the cells of the granule layer of the cerebellum (Rose et al., 1995; Hollnagel et al., 2002), the area composita in the intercalated disks connecting cardiomyocytes (e.g. C. M. Borrmann, Molekulare Charakterisierung der Adhärens-Zellverbindungen des Herzens: Identifizierung einer neuen Art, der Area composita, PhD Thesis, University of Heidelberg, Germany, 2000) and the heterotypic adhering junctions connecting the photoreceptor and Mueller glia cells of the retina, which characteristically contain the arm-protein neurojungin (Paffenholz et al., 1999).

The eye lens contains a central mass of densely packed anucleate fiber cells surrounded by a layer of cells with epithelioid features, often referred to as `epithelium'; this layer is in turn surrounded by a capsule of extracellular matrix material (for reviews, see Maisel et al., 1981; Rafferty, 1985). The physical laws of the vision process require transparency and homogeneity of the lens body. Therefore, the tight package of the lens fibers, with frequent and often regularly spaced interdigitations, as well as the high concentration and homogenous distribution of cytoplasmic proteins, including loosely arranged cytoskeletal filaments, are crucial for lens function (Ramaekers et al., 1980; Benedetti et al., 1981; Maisel et al., 1981; Ramaekers and Bloemendal, 1981). It is thus not surprising that diverse disturbances of the composition and distribution of lens proteins all lead to cataract formation (e.g. Capetanaki et al., 1989; Duncan et al., 2000; He and Li, 2000; Jakobs et al., 2000; Krutovskikh and Yamasaki, 2000).

Previously, we have noted that the cell-cell interactive structure of the lens fibers represents a relatively thin but extended cortex around the entire cell, obviously a cell-type-specific junctional complex (Schmidt et al., 1994). This cortex comprises the plasma membrane proper and a subjacent plaque-equivalent layer that only rarely shows distinct substructures (cf. Franke et al., 1987; Lo et al., 1997; Lo et al., 2000) but is generally associated with actin microfilaments, actin-binding proteins (ABPs) and adhering junction proteins including plakoglobin (Franke et al., 1987), α- and β-catenin (Bassnett et al., 1999; Duncan et al., 2000; Bagchi et al., 2002), α-actinin and vinculin (Geiger et al., 1985; Beebe et al., 2001). N-Cadherin has been reported to be the most prominent, if not the only, cadherin present (e.g. Hatta and Takeichi, 1986; Atreya et al., 1989; Citi et al., 1994; Lo et al., 2000; Bagchi et al., 2002). More recently, however, we have noticed that this cortex is highly complex and heterogenous, and comprises a set of proteins hitherto not shown – or even expected – in the lens.

Bovine and porcine eyes were obtained freshly from a local slaughterhouse, murine (mouse and rat) eyes from animals of the animal house of the German Cancer Research Center. For immunohistochemistry, eyes were enucleated and lenses routinely snap-frozen in isopentane cooled with liquid nitrogen to a temperature of about –130°C. For biochemical experiments, lenses were separated into the outermost `epithelioid' cell layer attached to the capsule, the cortex, and the `nucleus' and the dissected tissue portions were used either directly or frozen in liquid nitrogen and kept at –80°C until needed. Cultured epithelial cells of the human lines HaCaT, PLC (ATCC CRL-8024) and CaCo-2 (ATCC HTB-37), the canine line MDCK (ATCC CCL-34), the bovine line MDBK (ATCC CCL-22) and calf lens cells were grown as described (Ramaekers et al., 1980; Peitsch et al., 1999).

Antibodies used included mouse monoclonal antibodies (mAbs) against: ezrin (3C12), N-cadherin, vinculin (hvin-1), actin and tropomyosin from Sigma (St Louis, MO, USA); fodrin/spectrin (MAB 1822) from Chemicon (Hofheim/Taunus, Germany); drebrin from MoBiTec (Göttingen, Germany); E-, N-, P- and R-cadherin, cadherin-5, moesin, α- and β-catenin, and p120^ctn from Transduction Laboratories (Lexington, KY, USA); protein p0071 (mAb 6D-1-10) (Hatzfeld et al., 2003); and cadherin-11 from Zymed (South San Francisco, CA, USA). In addition, we used a series of mAbs from Progen Biotechnik (Heidelberg, Germany): desmoplakins (DP I/II, 2.15, 2.17 and 2.20) (Cowin et al., 1985), desmogleins Dsg1-Dsg3 (e.g. Dsg 1&2: 3.10.), desmocollins Dsc1-Dsc3 (e.g. mAbs U100 and U114), plakoglobin [Pg 5.10 (Cowin et al., 1986); 11E4], vimentin (3B4 and V9) (Hermann et al., 1989), plakophilins PKP1-PKP3 (Mertens et al., 1996; Schmidt et al., 1999) and an antibody to neurojungin (J 19.97) that in lens tissue also reacts with phakinin (cf. Paffenholz et al., 1999). Rabbit antibodies routinely used were against α-catenin, β-catenin, pan-cadherin, tropomyosin, α-actinin, l/s-afadin, l-afadin (Sigma) (for a review, see Takai and Nakanishi, 2003), protein ZO-1, connexin Cx 43, ponsin, claudin-1, occludin, cadherin-11 (from Zymed), non-muscle myosin heavy chain (Biotrend, Cologne, Germany), merlin/NF-2 from Santa Cruz Biotechnology (Heidelberg, Germany), CD44 (generous gift from M. Zöller, German Cancer Research Center) and protein p0071 (Hatzfeld et al., 2003). Specific guinea-pig antisera against plectin were also used (P2, from H. Herrmann, German Cancer Research Center) (cf. Schröder et al., 1999).

Monoclonal antibodies against desmoyokin (Dy 2.4. and Dy 47.27.5) obtained in this laboratory were systematically compared with desmoyokin rabbit antisera kindly provided by T. Hashimoto (Department of Dermatology, Keio University of Medicine, Tokyo, Japan) (cf. Hashimoto et al., 1993). To generate further antibodies specific for human desmoyokin, the synthetic peptides D1 (amino acids 2038-2056: PDVKIPKFKKPKFGFGPKS; `AHNAK fragment', accession number A45259), D2 (amino acids 2792-2812: PKGKGGVTGSPEASISGSKGD) and D4 (amino acids 298-320: PNLEGTLTGPRLGSPSGKTGT; all peptides used were from Peptide Specialty Laboratories, Heidelberg, Germany) were coupled to KLH and used to immunize guinea pigs after dissolution in Freund's complete adjuvant (Sigma). After three booster injections using Freund's incomplete adjuvant, the animals were anesthetized and blood was collected by heart puncture.

Periplakin antibodies used included murine mAbs [AE11 (Ma and Sun, 1986); clone IIb (Simon and Green, 1984); generous gifts from T. T. Sun (Department of Dermatology, New York University Medical Center, NY) and M. Simon (School of Dental Medicine, SUNY, Stony Brooks, NY, USA)] as well as guinea pig antibodies generated against three synthetic peptides, P1 (amino acids 7-24: KRNKGKYSPTVQTRSISN; accession number AAC 17738), P2 (amino acids 336-356: LRKVDSDLNQKYGPDFKDRYQ) and P3 (amino acids 815-835: ENGRSSHVSKRARLQSPATKV) as described above for desmoyokin.

Lens cryosections of ∼5 μm thickness were air dried for several hours and fixed for 5 minutes in methanol, followed by 5 minutes in acetone both at –20°C. Because of the fragile character of lens, the incubation protocols were optimized for relatively brief periods of incubation time, and the specimens were usually incubated with the primary antibodies diluted in PBS for 30 minutes, followed by two repeated washes in PBS for 5 minutes or less each, incubation with the secondary antibodies for 30 minutes, and two subsequent 5-minute washes in PBS. The sections were then rinsed in distilled water, fixed for 5 minutes in ethanol and mounted in Fluoromount G (Biozol Diagnostica, Eching, Germany). To enhance the accessibility of certain large proteins such as desmoyokin and periplakin to immunoglobulins, the specimens were initially exposed to PBS containing either 0.1% Triton X-100 or 0.1% saponin and then washed twice in PBS for 5 minutes each. Immunofluorescence of cell cultures and of other tissues was performed as described (Peitsch et al., 1999).

Epifluorescence was observed and documented with a Zeiss Axiophot photomicroscope, confocal laser-scanning immunofluorescence microscopy was done with a Zeiss LSM 510 (Zeiss, Jena, Germany).

Cortex preparations from several lenses were homogenized with either a Dounce or a Potter-Elvehjem homogenizer (Braun, Melsungen, Germany) at very high volume-to-tissue mass ratios in low salt buffer (50 mM Tris-HCl, pH 7.4, 5 mM MgCl₂, 1 mM DTT) containing the protease inhibitor phenylmethylsulfonylfluoride (PMSF) at 1 mM or Pefablock SC (Roche Diagnostics, Mannheim, Germany). Pellets obtained after centrifugation at 4°C at 18,000 g (in a Beckman centrifuge Optima XL-70, München, Germany) for 30 minutes contained the `water-insoluble' particle fraction (WIF), primarily the cytoskeleton and membranous structures, whereas the supernatant contained the water-soluble particle fraction (WSF) including the crystallins (Alcala et al., 1975). This procedure was repeated twice with the WIF to minimize residual crystallins. Tissues other than lens and cell cultures were directly homogenized in SDS sample buffer as described (Peitsch et al., 1999).

Gel electrophoresis and immunoblotting SDS-PAGE, and two-dimensional gel electrophoresis involving either non-equilibrium pH-gradient electrophoresis (NEPHGE) or isoelectric focusing (IEF) were performed as described (Achtstätter et al., 1986). For SDS-PAGE, samples were suspended in electrophoresis buffer (250 mM Tris-HCl, pH 6.8, 20% SDS, 25% glycerol, 125 mM DTT), often with the addition of benzonase (1:1000; Merck, Darmstadt, Germany). For NEPHGE or IEF, protein samples were precipitated with methanol and chloroform, and solubilized in lysis buffer containing 9.5 M urea, 2.0% NP-40, 2.0% ampholine and 20 mM DTT.

Immunoblotting was performed using PVDF membranes (Millipore, Bedford, MA, USA). After blocking with 10% non-fat dry milk in Tris-buffered saline containing 0.1% Tween (TBST) for at least 1 hour, blots were incubated with the primary antibodies in PBS for 1 hour, followed by three washes in TBST for 30 minutes each. Horseradish peroxidase (HRP)-conjugated antibodies to rabbit, mouse or guinea pig IgG (diluted 1:10000 in TBST) were applied for 30 minutes, followed by a 30 minute wash in TBST and enhanced chemiluminescence (ECL; Amersham Biosciences, Freiburg, Germany).

Pelleted fractions were suspended in immunoprecipitation buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA or 0.5 mM CaCl₂, 1% Triton X-100, 1 mM DTT, 1 mM PMSF or Pefablock SC) and centrifuged for 15 minutes at 14,900 g (in an Eppendorf centrifuge 5414, Hamburg, Germany) and 4°C. The supernatant obtained was then precleared with protein-A- or protein-G-coupled Sepharose for several hours, and the supernatant obtained after centrifugation was reacted overnight with protein A and/or protein G beads coated with the specific antibody in 50 mM Tris-HCl, pH 7.5. The pellet obtained was solubilized in 20-40 μl sample buffer, and the immunoprecipitate was separated using SDS-PAGE. Protein gels stained with colloidal Coomassie Blue (Novex, Frankfurt, Germany) were used to analyse unknown bands by peptide mass fingerprinting [matrix-assisted laser desorption ionization mass spectrometry (MALDI-MS)] and amino acid sequence analysis (see below).

Protein bands were excised from the gel and cut into 1×1 mm pieces that were washed twice with deionized water, 50% acetonitrile/water (1:1) and acetonitrile. Proteins were digested overnight with sequencing-grade modified trypsin (Promega) in 40 mM ammonium bicarbonate at 37°C. The reaction was stopped by freezing.

MALDI mass spectra were recorded in positive ion reflector mode with delayed extraction on a Reflex II time-of-flight instrument (Bruker-Daltonik, Bremen, Germany) equipped with a SCOUT multiprobe inlet and a 337-nm nitrogen laser. The ion-acceleration voltage was set to 20.0 kV, the reflector voltage to 21.5 kV and the first extraction plate to 15.4 kV. Mass spectra were obtained by averaging 50-200 individual laser shots. Calibration of the spectra was performed internally by a two-point linear fit using the autolysis products of trypsin at mass:charge ratios of 842.50 and 2211.10.

For the mass spectrometric analysis of tryptic digests, MALDI samples were prepared on thin film spots (Jensen et al., 1996). Briefly, 0.3 μl aliquots of a nitrocellulose-containing saturated solution of α-cyano-4-hydroxycinnamic acid (Sigma-Aldrich) in acetone were deposited onto individual spots on the target. Subsequently, 0.8 μl 10% formic acid and 0.4 μl of the digest sample were loaded on top of the thin film spots and allowed to dry slowly at ambient temperature. To remove salts from the digestion buffer, the spots were washed with 1% formic acid and with water.

Post-source-decay (PSD) analysis was performed in positive ion reflector mode with delayed extraction by setting an ion gate width of 40 Da around the ion of interest. Data were acquired in 14 segments by decreasing the reflector voltage in a stepwise fashion. For each segment, 200 individual laser shots were accumulated. The fragment ion spectrum was obtained by pasting together all segments to a single spectrum using the FAST software (Bruker). Fragment ion calibration was performed externally with the fragment masses of the adrenocorticotropic hormone (ACTH) 18-39 clip.

Sample preparation for PSD analysis was achieved by cocrystallization of matrix with samples concentrated using Zip Tip C18 (Millipore, Schwalbach, Germany). Briefly, the peptides in the supernatant of the in-gel digestion were absorbed to a prewashed (50% acetonitrile/water) and equilibrated (0.1% trifluoroacetic acid/water) Zip Tip C18 by repetitive pipetting steps. Following washing of the Zip Tip C18 by equilibration buffer, the peptides were eluted from the Zip Tip with 1 μl of matrix (α-cyano-4-hydroxycinnamic acid saturated in 50% acetonitrile/water).

Singly charged monoisotopic peptide masses were used for database searching. Searches were performed against the NCBInr database using the ProFound search algorithm (http://129.85.19.192/prowl-cgi/ProFound.exe) and the Protein prospector software developed at the University of California, San Francisco (http://prospector.ucsf.edu/), with an IEP range of 0-14 and the oxidation of methionine as a possible modification. Up to one missed tryptic cleavage was considered, and the mass tolerance for the monoisotopic peptide masses was set to ±100 ppm or ±0.1 Da.

Searches with fragment masses from PSD experiments were performed against the NCBInr database using the MS-Tag search algorithm provided by Protein prospector. Parent mass tolerance was set to ±100 ppm and fragment ion tolerance was set to ±1500 ppm.

For high-performance liquid chromatography (HPLC) separation the tryptic digest was extracted twice with 0.1% trifluoroacetic acid (TFA) in 60% acetonitrile. After concentration on a SpeedVac, the extracted tryptic peptides were separated on a HPLC system equipped with a 140B solvent delivery system (Applied Biosystems, Weiterstadt, Germany), Acurate splitter (LC Packings, Idstein, Germany), UV absorbance detector 759A (Applied Biosystems), U-Z capillary flow cell (LC Packings) and a Probot fraction collector (BAI, Lautertal, Germany) using a reversed-phase column (Hypersil C18 BDS, mean particle diameter 3 μm, 0.3×150 mm) and a linear gradient from 12% acetonitrile in 0.1%TFA to 64% acetonitrile in 0.08% TFA in 90 minutes, with a flow rate of 4 μl minute^–1 at room temperature.

Peptide elution was monitored at 214 nm and individual fractions from the HPLC separation were re-analysed by MALDI-MS. Sequence analysis of selected peptide-containing fractions was performed on a Procise Protein Sequencer 494 cLC using standard programs supplied by Applied Biosystems.

In previous immunohistochemical studies of mammalian lenses and in biochemical analyses of cytoskeleton fractions obtained therefrom, we had noted that the border structures of the anucleate lens fiber cells were compositionally and structurally different from other kinds of junctions and collectively referred to this large complex as the cortex adhaerens (Franke et al., 1987; Schmidt et al., 1994). Moreover, because we found unexpectedly many proteins in lens cortical structures, we decided to examine the composition of the cortex adhaerens more systematically.

When total proteins from mammalian lens fiber tissue, without the capsule and the outermost `epithelioid' cell layer, were extracted with relatively large amounts of `low salt buffer', a residual (`cytoskeletal') fraction was obtained that revealed, on SDS-PAGE or two-dimensional gel electrophoresis, a remarkably complex protein pattern. In addition to the IF proteins vimentin, phakinin and filensin (e.g. Lieska et al., 1980; Ramaekers et al., 1980; Merdes et al., 1991; Merdes et al., 1993; Perng et al., 1999) (U. Haus, Zur molekularbiologischen Charakterisierung von Cytoskelett-Proteinen der Rinderlinse. Diploma Thesis, University of Cologne, Germany, 1990), some typical junctional proteins were consistently seen in all three species studied. These included the transmembrane glycoproteins N-cadherin, cadherin-11 and – in much lower amounts – E-cadherin (Fig. 1, lanes 1-6), which seemed to be restricted to the outer layers, as well as the plaque proteins α- and β-catenin (Fig. 1, lanes 7-10), p120^ctn (see below) and, in amounts markedly differing in different species, plakoglobin. Although the finding of an N-cadherin-based ensemble of adhering junction proteins essentially confirmed earlier reports (e.g. Geiger et al., 1985; Cowin et al., 1986; Hatta and Takeichi, 1986; Franke et al., 1987; Atreya et al., 1989; Bassnett et al., 1999; Ferreira-Cornwell et al., 2000; Leong et al., 2000; Bagchi et al., 2002), the recognition that the cortices adhaerentes of lens fibers contained the type-II cadherin-11 in similar large amounts was novel. On the one hand, it was compatible with the general widespread occurrence of cadherin-11 in diverse kinds of mesenchymal and other mesodermally derived cells but, on the other hand, it was surprising because it had not been detected in previous studies of lens tissue (e.g. Hoffmann and Balling, 1995; Simonneau et al., 1995; Hadeball et al., 1998; Simonneau and Thiery, 1998). The other cadherins examined (see Materials and Methods) were not detected. Among the cytoskeletal proteins, we regularly detected vinculin (Fig. 1, lanes 15 and 16), α-actinin, actin, tropomyosin, myosin, spectrin, ankyrin and plectin (not shown), all of which had previously been reported to occur in lens-fiber cortices (e.g. Kibbelaar et al., 1979; Repasky et al., 1982; Allen et al., 1987; Franke et al., 1987; Weitzer and Wiche, 1987; Lee et al., 2000).

Much to our surprise, however, we also noted among the major lens junction proteins a series of plaque components such as ezrin (Fig. 1, lanes 11 and 12), periplakin (Fig. 1, lanes 13 and 14), periaxin (see also below) and desmoyokin (Figs 2, 4), which so far had only been reported from other kinds of cells [ezrin (Bretscher, 1983; Bretscher et al., 1997); periplakin (Simon and Green, 1984); periaxin (Gillespie et al., 1994); desmoyokin (cf. Hieda et al., 1989; Shtivelman et al., 1992; Hashimoto et al., 1993)]. In addition, we found considerable amounts of moesin in our immunoblots of total lens fiber proteins, but no significant signals for merlin [see below, however, for reports of the occurrence of merlin in outer (i.e. epithelioid) cells of lenses] (see Claudio et al., 1995; Claudio et al., 1997; Huynh et al., 1996).

To identify the protein complexes containing these lens cortex proteins, we performed immunoprecipitations, subjected the pelleted proteins to SDS-PAGE, excised the bands under question and analysed them by MALDI-MS, PSD and amino acid sequencing. The results allowed us to determine the complement of proteins associated with the specific antigen. For example, N-cadherin immunoprecipitates consistently contained not only the associated plaque proteins α- and β-catenin, plakoglobin and p120^ctn, but also remarkable amounts of cadherin-11 (Fig. 3). Conversely, considerable proportions of N-cadherin were identified in the immunoprecipitates obtained with antibodies to cadherin-11. In lens tissue material also containing outer cortical cell layers, junctional plaque proteins were detected in combinations with both N- and E-cadherin as well as with cadherin-11 (data not shown). These results also showed for the first time the existence of such heterotypic cadherin complexes with common plaque proteins but did not yet allow to distinguish between lateral heterocomplexes in the same membrane from transcellular heterotypic complexes of cadherins (e.g. Volk et al., 1987); that is, from the `heterocadherins' in the sense used by Duguay et al. (Duguay et al., 2003) [for the controversial literature on cadherin organization see Shapiro et al., and others (Shapiro et al., 1995; Leckband and Sivasankar, 2000; Boggon et al., 2002; Ahrens et al., 2003)]. By contrast, ezrin immunoprecipitates contained desmoyokin, periplakin and periaxin (Table 1; Fig. 4a), and, when further probed with specific antibodies, positive reactions were also seen for spectrin (Fig. 4b), plectin and moesin (data not shown). These results left no doubt that desmoyokin was a major protein of lens fiber cells, where it mostly occurred in ezrin complexes, apparently together with periplakin and periaxin (Fig. 4a), suggesting the existence of a special category of large EPPD plaque complexes. For reasons not yet clarified, however, we detected little if any ezrin and moesin in the reciprocal immunoprecipitates of desmoyokin and periplakin (data not shown).

In similar biochemical experiments, we failed to detect in lens fibers significant amounts of desmosome-specific proteins such as desmoplakins, desmogleins, desmocollins and plakophilins, or of the plaque proteins neurojungin, merlin, afadin, ponsin and drebrin (results not shown). The negative results obtained for the transmembrane glycoprotein CD44 were surprising because this protein has been reported in previous studies of lens material by other authors (Nishi et al., 1997; Saika et al., 1998), but we have not yet excluded all the diverse CD44 variants.

Neither in our electron micrographs nor by immunofluorescence microscopy did we find, in sections through lens fibers, any indication of the presence of tight junctions, including negative reactions for occludin and several claudins (cf. Langbein et al., 2002), and of desmosomes (see also Franke et al., 1987). By contrast, various sizes of gap junctions were regularly found by electron microscopy and with antibodies to both connexins and protein ZO-1, especially in the central region of the long cell sides, confirming previous reports (Giepmans and Molenaar, 1998; Toyofuku et al., 1998; Nielsen et al., 2001). The cytoplasm displayed the notorious intense positivity for IF proteins such as vimentin (cf. Ramaekers et al., 1980) and phakinin (e.g. Fig. 6B) (cf. Merdes et al., 1991; Merdes et al., 1993), whereas actin and the ABPs examined appeared to be generally enriched in the cortical zone (cf. Kibbelaar et al., 1979; Lo, 1988; Lo et al., 1997).

Using immunohistochemistry, N-cadherin and cadherin-11 were the only cadherins that consistently reacted at the contacts of lens fiber cells, very intensely at the short apical sides and rather weakly, sometimes hardly visible at all, along the long lateral sides (Fig. 5A-C). By contrast, immunostaining for E-cadherin was weakly present in the outer cortical layers but diminished centripetally in a steep gradient (not shown). Particularly at the short sides, both N-cadherin and cadherin-11 colocalized with both α-catenin and the major arm-repeat proteins β-catenin (Fig. 5A-C″), plakoglobin and p120^ctn (results not shown), as well as with actin and vinculin (cf. Volk and Geiger, 1984; Geiger et al., 1985; Franke et al., 1987). Specifically in bovine lens, the plakoglobin reaction was relatively strong in the outermost cell layers, but was practically undetectable in the lens interior [for other species, see Franke et al. (Franke et al., 1987)]. Again, the immunoreaction of all these proteins was very intense at the short sides and weak on the long sides.

This immunohistochemical reaction of typical proteins of the cortex adhaerens on the short sides was somewhat different from the pattern of other junction-associated proteins, in particular several ABPs such as ezrin (Fig. 5D-D″, Fig. 6A,A′), moesin (not shown) and plectin (Fig. 6B), all of which showed immunostaining on both the short and the long sides, although mostly again more intensely on the short sides. By stark contrast, gap junction proteins such as connexin Cx 43 and protein ZO-1 appeared in clusters of punctate reaction sites in the central region of the long sides (Fig. 6C-C″). Clearly, however, the newly discovered large lens proteins desmoyokin (Fig. 7A-A″) and periplakin (Fig. 7B-B″) showed, like ezrin and moesin, intense immunostaining on the short sides, whereas their reactions on the long sides were in most regions weak and appeared to be interrupted in places. Moreover, at higher resolution, the cortical proteins enriched on the short sides differed in their distribution patterns and displayed a mutually exclusive patchwork (Fig. 8) – although the typical junctional plaque proteins (such as α- and β-catenin) colocalized with N-cadherin (Fig. 8A) and cadherin-11 (Fig. 5C-C″), ezrin, moesin, periplakin and desmoyokin were concentrated in interspersed regions that did not react with N-cadherin (Fig. 8B) or any other cadherin (not shown), although there were also regions of overlap indicative of colocalization (Fig. 8B; see also Fig. 5D).

The immunofluorescence reaction pattern of merlin [another cortical protein related to ezrin and moesin (for a review, see Bretscher et al., 2002)] was surprising in two ways. It was generally weak, often negative, in the more cortical regions of the lens fiber mass and stronger in the deeper regions, opposite to what has been reported for mouse and chicken lens (e.g. Claudio et al., 1995; Claudio et al., 1997; Huynh et al., 1996); however, where positive, it often appeared with nearly equal intensity on both the long and the short sides.

Immunolocalization reactions for neurojungin, afadins, ponsin, drebrin and protein CD44 were negative.

The extended structure which so tightly connects the plasma membranes of the anucleate fiber cells of the lens represents one large, mostly homogeneous-looking cortex suggestive of a continuous adhering junction complex. This giant cortex adhaerens is interrupted only sparsely by locally densified adherens plaques (e.g. Rafferty, 1985; Franke et al., 1987; Lo et al., 1997; Lo et al., 2000) and by gap junctions containing a specific set of connexins to which certain cytoplasmic plaque proteins such as ZO-1 are attached. Although this cortex adhaerens appears structurally rather uniform under the electron microscope (e.g. Ramaekers et al., 1980; Rafferty, 1985; Franke et al., 1987; Lo, 1988; Lo et al., 2000), our detailed studies have shown marked biochemical complexity and mosaicism.

Taken together, the present results show that the cortex adhaerens hexagons, notably the short sides, are characterized by a typical adhaerens junction ensemble, dominated by N-cadherin and cadherin-11 as major transmembrane glycoproteins and a plaque comprising not only α- and β-catenin, but also plakoglobin, p120^ctn and vinculin. The results of our N-cadherin/cadherin-11 cross-immunoprecipitation experiments have also directly demonstrated intimate complexes of different type-I and type-II cadherins with the plaque proteins mentioned, and experiments are under way to decide whether these are ipso- or heterocellular cadherin complexes. The immunocytochemical results further suggest that the long-side cortex of the fiber cells, at least in the species examined, contains much less of the adherens junction components but is relatively rich in ABPs characteristic of other microfilament-anchorage complexes, including band 4 proteins (Aster et al., 1984; Allen et al., 1987), spectrin (Nelson et al., 1983; Green and Maisel, 1984; Thomas, 2001) and plectin (Weitzer and Wiche, 1987).

Our surprising finding of a totally novel group of actin filament anchorage proteins in the lens now adds another ensemble of cytoskeletal proteins to the cortex adhaerens that have hitherto been reported only from diverse other cells. These include ezrin and moesin [cortical ABPs of various epithelial and certain non-epithelial cells (e.g. Bretscher, 1983; Bretscher et al., 1997; Yonemura et al., 1998; Bretscher et al., 2002)], periplakin [a desmoplakin-related protein so far found only in epidermis and other stratified epithelia (Ruhrberg et al., 1997; DiColandrea et al., 2000; Karashima and Watt, 2002) [for periplakin gene transcripts in certain other cells, see also Aho et al. (Aho et al., 1998)] and desmoyokin [a large protein with a hotly debated location (e.g. Hieda et al., 1989; Shtivelman et al., 1992; Hashimoto et al., 1993; Shtivelman and Bishop, 1993; Masanuga et al., 1995)]. In addition, in such complexes, we have detected periaxin, a protein originally identified in myelinating Schwann cells, where it is enriched at plasma membranes (e.g. Gillespie et al., 1994; Melendez-Vasquez et al., 2001). Our immunoprecipitation results have further shown that lens-fiber ezrin occurs in the same junctional plaque complexes as periplakin, periaxin and desmoyokin, often also in association with lens spectrin(s), which suggests but does not yet prove that all these proteins can co-assemble into a giant cortical EPPD complex.

The constitutive occurrence of desmoyokin and periaxin in the cortex adhaerens of lens fibers is especially noteworthy because both proteins have been reported as `dual location proteins' that occur in certain plasma membrane regions as well as in the nucleoplasm of a broad variety of cell types (e.g. Shtivelman et al., 1992; Shtivelman and Bishop, 1993; Masanuga et al., 1995; Sherman and Brophy, 1999; Nie et al., 2000; Sussmann et al., 2001). The mere absence of nuclei in the lens fiber cells now also demonstrates that desmoyokin and periaxin are indeed major, stable components of adhering junctions that can occur in a long-lasting form in the absence of a nucleus. In this context, it is also worth mentioning that desmoyokin has recently been shown in the area composita plaques of cardiac intercalated disks (e.g. Hohaus et al., 2002).

At present, the transmembrane protein(s) to which EPPD complexes are attached are still elusive. Although we cannot formally exclude some contribution of cadherins to these cortical complexes, their relatively low concentrations along the long sides of the fiber cell cortices (see Results) (Bassnett et al., 1999; Lo et al., 2000; Beebe et al., 2001) and their absence in our EPPD immunoprecipitates tend to suggest an involvement of other transmembrane proteins. Because the known ezrin-binding protein CD44 (Tsukita et al., 1994; Yonemura et al., 1998) has not been detected in our ezrin immunoprecipitates, we will have systematically to examine the series of known candidates of possible transmembrane partners in the EPPD complex (for a review, see Bretscher et al., 2002). It will also be interesting to investigate the possible existence of junctional plaque complexes of the EPPD category in other cell types.

Our high-resolution double-label immunofluorescence microscopy also revealed a mosaicism of the cortex adhaerens, in particular on the short sides of hexagons, where puncta-adhaerentia-type complexes comprising cadherins and catenins alternate with junctional structures containing EPPD complexes. Whether this regular patchwork pattern is a general characteristic of the junction system and how it is formed in the development of lens fibers from the `epithelioid' cells of the lens surface, remain to be studied. Considering the frequency and sensitivity with which alterations of protein composition of lens fibers result in cataract formations, including changes of junctional components such as gap junction connexins (Martinez-Wittingham et al., 2003), we expect that gene abrogation experiments will probably help to elucidate the functional importance of the molecular complexity and pattern arrangement of the cortex adhaerens.

We thank C. Grund and S. Winter-Simanowski for excellent technical help, J. Osterholt for expert photographic work, and E. Gundel for typing the manuscript. The work has been supported in part by a grant of the Deutsche Forschungsgemeinschaft (DFG).