Desmosomal glycoproteins 2 and 3 (desmocollins) show N-terminal similarity to calcium-dependent cell-cell adhesion molecules

Desmosomal glycoproteins 2 and 3 (dg2/3), or desmocollins, are major components of desmosomes. These intercellular adhesive junctions are found in epithelia, cardiac muscle, follicular dendritic cells, and meninges. Components immunologically related to these glycoproteins have been demonstrated in all desmosome-bearing tissues examined from mammalian, avian, reptilian and anuran amphibian species (Cowin and Garrod, 1983; Cowin et al. 1984a; Suhrbier and Garrod, 1986). Inhibition of desmosome assembly in Madin Darby bovine kidney (MDBK) cells by univalent antibody fragments against these glycoproteins has implicated them directly in desmosomal adhesion (Cowin et al. 1984b).

dg2/3 can be purified from preparations of bovine nasal epidermal desmosomes (Skerrow and Matoltsy, 1974a,b) or glycoprotein-enriched desmosomal cores (Gorbsky and Steinberg, 1981). In the latter preparations, they form the major components, together with the other major desmosomal glycoprotein (dgl), or desmoglein. Cohen et al. (1983) showed that dgl can be resolved into a triplet of polypeptides by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS–PAGE) whereas dg2/3 may be resolved into about four bands. Size heterogeneity of these glycoproteins has been supported by a number of studies (Suhrbier and Garrod, 1986; Jones et al. 1987). We have recently provided strong evidence for the presence of three size variants of dg2/3 in Madin Darby canine kidney (MDCK) cells (Parrish et al. 1990). The various isoforms are difficult to resolve and for simplicity we shall therefore refer to dg2 and -3 as a pair of glycoproteins having relative molecular weights of 115000 and 107 000.

Immunochemical characterisation of desmosomal glycoproteins using monoclonal antibodies suggested that dg2 and -3 are related and form a family distinct from the dgl glycoprotein (Cohen et al. 1983). This was supported by one-dimensional peptide mapping and by amino acid analysis (Kapprell et al. 1985). While the distinction between these families of glycoproteins is generally accepted, it is our experience that polyclonal antisera raised against dg2 or dg3 commonly show some crossreactivity with the dgl group (Suhrbier and Garrod, 1986), suggesting the possibility of some common epitopes between the two families.

The major desmosomal glycoproteins have been localised within desmosomes by immuno-electron microscopy (Miller et al. 1987; Steinberg et al. 1987; Garrod et al. 1990). Using metabolic labelling, immunoprecipitation and whole-cell trypsinisation we have recently shown that the dg2/3 polypeptides in MDCK cells have a transmembrane organisation in which approximately one-fifth of the relative molecular mass is cytoplasmic. The cytoplasmic domains are heterogeneous in size and phosphorylation (Parrish et al. 1990).

Desmosomal adhesion is calcium-dependent (Hennings and Holbrook, 1983; Watt et al. 1984; Mattey and Garrod, 1986; Bologna et al. 1986) and the desmosomal glycoproteins bind calcium ions (Mattey et al. 1987; Steinberg et al. 1987). Calcium-dependent adhesion in other systems has been shown to be mediated by a family of closely related adhesion molecules, which includes the cadherins, uvomorulin, A-CAM, cell CAM 120/80 and L-CAM (reviewed by Takeichi, 1988). The amino acid sequences of these proteins have been determined from cDNA clones and reveal that they represent a homologous family of calcium-binding, transmembrane glycoproteins with a single membrane-spanning domain (references given in legend to Fig. 1). A polypeptide related to the epithelial cadherin (E-cadherin) has been shown to be a minor component of bovine nose and tongue desmosomes (Jones, 1988). This molecule has a relative molecular weight of 125000 in these tissues and is distinguished from dg2/3 both by its lower electrophoretic mobility and its lower abundance.

In this paper we demonstrate, by direct sequencing of the amino terminus of dg2/3 from bovine nasal epidermis, that these glycoproteins show strong N-terminal similarity with glycoproteins of the cadherin family. This suggests that there is a relationship between the desmosomal glycoproteins and other calcium-dependent adhesion molecules, and supports the view that dg2/3 are directly involved in desmosomal adhesion.

Desmosomes were prepared from bovine nasal epidermis using the method of Skerrow and Matoltsy (1974a). Desmosomal components were separated by sodium dodecyl sulphate-poly-acrylamide gel electrophoresis (SDS–PAGE) under reducing conditions (Laemmli, 1970). The dg2/3 polypeptides were excised together and electro-eluted (Jacobs and Clad, 1986) in SDS/Tris–glycine buffer, dialysed against 0.05% (w/v) SDS in water and lyophilised. N-terminal sequencing was performed by automated solid-phase Edman degradation (Findlay et al. 1989) using the microsequencing facility in the Department of Biochemistry, University of Leeds.

Peptides were synthesised by the t-Boc solid-phase method (Merrifield, 1986). Peptide 1 comprised the N-terminal 23 amino acids of dg2/3 (Fig. 1). To enhance the immune response peptide 1 was conjugated to keyhole limpet haemocyanin (KLH) (Green et al. 1982). An alternative strategy to increase the immune response to small peptides is to synthesise a peptide polymer. To do this peptide 2 was synthesised as peptide 1 with the addition of an N-chloroacetyl-derivatized alanine residue at the N terminus. The cysteine residue at position 19 provides a free sulphydryl group for reaction with the chloroacetyl group. Peptide 2 was dissolved in 0.1m NaHCOg and allowed to polymerise at room temperature (Lindner and Robey, 1987). Peptide 2 was also conjugated to KLH and used as a third immunogen.

Monoclonal antibody 07–4D was raised by immunising Balb/c mice with two intraperitoneal injections of desmosomal cores prepared by the method of Gorbsky and Steinberg (1981). The first injection was in Freund’s complete adjuvant (FCA), the second in Freund’s incomplete adjuvant (FIA). This was followed by one booster injection of gel-purified dg2/3 in FIA. Spleen cells were fused with NS-1 myeloma cells. Clones with activity against desmosomal cores were identified by ELISA (Cohen et al. 1983) and cloned by limiting dilution as described by Parrish et al. (1987):

Polyclonal rabbit antisera were raised by sub-cutaneous injections of 200 μg KLH–peptide 1, KLH–peptide 2 or polymerised peptide 2 in FCA and boosted with 200 μg of peptide-KLH or peptide polymer in FIA at intervals of three weeks. Test bleeds were performed after the second and third injections and the rabbits bled out 8 weeks after the initial immunisation.

Mouse monoclonal antibody 52–3D against dg2/3 has been described elsewhere (Parrish et al. 1990).

Polyclonal antisera were raised in guinea-pigs according to Cowin and Garrod (1983).

Polyclonal antisera were screened for activity against KLH or peptide 1 by ELISA. Peptide 1 or KLH (1 pg/well in 100 pl of 15mM Na₂CO₃, 35 mM NaHCO₃, 0.02% (w/v) NaN₃, pH9.6) was allowed to adsorb to 96-well microtitre plates at 4 °C overnight. Plates were washed five times in phosphate-buffered saline (PBS)+0.05% (v/v) Tween 20 and then blocked with 1% (w/v) bovine serum albumin in 150 mM NaCl, 50 mM Tris-HCl, 0.05% (v/v) Tween 20, pH 7.4 (TB) for lh at 37 °C.

Serial 1:2 or 1:5 dilutions of antisera in TB were made across the microtitre plate and incubated for 1 h at 37 °C before washing five times with PBS/Tween. Plates were then incubated for 1 h at 37°C with 100 μl/well peroxidase-conjugated anti-rabbit immunoglobulins (Amersham International) diluted 1:1000 in TB, and washed five times with PBS/Tween. The plates were then incubated with 3,3’,5,5’-tetramethyl benzidine (TMB) (150 μl/well) as follows. A 15 mM TMB stock solution in dimethyl sulphoxide was diluted to 0.15 mM in 100 mM sodium acetate, (pH 6.0)+0.003% (v/v) H₂O₂. The reaction was halted by addition of 50 μl/well of 2 m H₂SO₄ and the absorbance at 450 nm read using an ELISA plate reader (Dynatech).

Activity of sera against desmosomal cores was measured in the same way except that desmosomal cores were diluted to 20 μg ml⁻¹ in PBS (pH 11.0) for adsorption (Cohen et al. 1983).

Western blotting was performed as described by Suhrbier and Garrod (1986) using ¹²⁵°I-labelled protein G or ¹²⁵I-labelled antimouse immunoglobulins (Amersham International).

Rabbit antiserum was diluted either 1:3000 in TB and adsorbed with peptide 1 (1 mg ml⁻¹) or 1:10000 with TB and adsorbed with KLH (1 mg ml’¹). Activity against KLH and peptide 1 before and after adsorption was assessed by ELISA. Rabbit antiserum was diluted 1:3000 with 0.25% (w/v) gelatin in 150 mM NaCl, 50 mM Tris-HCl, 0.5% (v/v) Tween-20, (pH 7.4) and adsorbed with either peptide 1 (1 mg ml⁻¹) or KLH (1 mg ml⁻) prior to blotting on desmosomes prepared from bovine nasal epidermis.

Bovine nasal epidermis was fixed in 4% formaldehyde in 200 nw Hepes (pH7.4). Blocks were then infused for 15min with 2.1m sucrose in PBS and frozen in liquid nitrogen. Frozen sections (lOOnm) were cut at –90°C using an RMC MT6000-XL ultramicrotome with cryo-chamber attachment. Immunolabelling of thawed cryo-sections was performed using protein A conjugated to 9nm gold particles for rabbit antisera or rabbit anti-mouse immunoglobulins (DAKO) followed by protein A–9 nm gold conjugate for monoclonal antibodies (Tokuyasu, 1980; Griffiths et al. 1984). Protein A–gold was kindly provided by Dr G. Griffiths, EMBL, Heidelberg, FRG.

Staining of 5 fan frozen sections was performed as described by Mattey and Garrod (1986).

Peptide mapping was performed by the method of Cleveland et al. (1977) with some modifications. Isolated desmosomes from bovine nasal epidermis were separated on a 7.5% polyacrylamide gel, the dg2 and dg3 bands were cut from the unstained gel and incubated in 125 him Tris–HCl, 0.1% (w/v) SDS, ImM EDTA, 0.4% (w/v) 2-mercaptoethanol (pH 6.8) (Buffer A) for lh at room temperature. Three pieces of gel were inserted into each well of a 15% polyacrylamide gel with 1 mM EDTA added to the gel solutions before polymerisation. The gel slices were overlaid with Staphyloccus aureus V8 protease (5μg/well in 10 μl of Buffer A+0.0001% (w/v) Bromophenol Blue and 10% (v/v) glycerol). Electrophoresis was performed at 125 V until the proteins had stacked and then interrupted for 30 min to allow digestion to take place at room temperature. Protein bands were visualised by Coomassie Blue staining or transferred to nitrocellulose for immunoblotting.

Release 13 of the Swiss-prot protein sequence databank was searched with the N-terminal sequence of dg2/3 using FASTA, a modification of FASTP (Lipman and Pearson, 1985). The database and supporting programs have been mounted on the Southampton University IBM 3090–150VF mainframe computer with the aid of the Southampton University Computing Project Fund.

Secondary structural predictions were carried out by the method of Chou and Fasman (1974).

Sequencing a mixture of gel-purified dg2 and dg3 yielded a sequence for the first 23 amino acids. Repeat sequencing runs for the two proteins together or individually yielded similar results. The sequence obtained showed significant identity with members of the cadherin–uvomorulin–L-CAM family of cell adhesion molecules, but no significant similarity to the sequences of other adhesion molecules including N-CAM and integrins.

The highest identity was with N-cadherin, ten (43%) of the amino acids being identical between the two molecules (Fig. 1). The identity between the dg2/3 sequence and the cadherins is much less than between the members of the family themselves (Table 1). Several other interesting features arise from the sequence comparisons. Five amino acids at positions 2,6,11,12 and 15 are conserved among all of the proteins including dg2/3 (Fig. 1). In addition, three conservative substitutions exist between dg2/3 and N-cadherin at positions 3,8 and 21 (marked by dots in Fig. 1). These correspond to conserved residues in the cadherin family. Structural analysis by the method of Chou and Fasman (1974) suggests a high probability of a turn in positions 11–17 in the dg2/3 sequence (overlined in Fig. 1).

Eight antisera were raised against the N-terminal peptide antigens as described in Materials and methods. All antisera showed similar properties. Data for one serum raised against peptide 2–KLH conjugate (designated 486) will be presented here. Titre and specificity were assayed by ELISA. The serum showed significant activity against both the peptide and KLH (Fig. 2). The titres (dilution of antiserum giving half-maximal binding under the conditions specified in Materials and methods), were 1:58000 against the N-terminal peptide and 1:360 000 against KLH. The activity against the peptide was blocked by the peptide but not by KLH and the activity against KLH was blocked by KLH but not by the peptide (Fig. 2). The antibody also showed binding to desmosomal cores (data not shown).

On Western blots of whole desmosomes from bovine nasal epidermis the anti-peptide serum showed reactivity with two bands of 115 000 and 107 000 corresponding to dg2/3 (Fig. 3). No other desmosomal components were recognised by the antiserum. Identical bands were recognised by two monoclonal antibodies to dg2/3, antibody 07–4D raised in the present study and antibody 52–3D, which reacts with the cytoplasmic domains of dg2/3 as described by Parrish et al. (1990). The same polypeptides were also recognised by a guinea-pig polyclonal anti-dg2/3 serum (Fig. 3). The anti-dg2/3 activity of the anti-peptide serum was completely adsorbed by the peptide but was not blocked by KLH (Fig. 4).

The anti-peptide serum gave fluorescent staining of bovine nasal epidermis, human skin and bovine cardiac muscle. In the epithelia the staining was at the cell borders appearing linear in some places and punctate in others (Fig. 5). This pattern is similar to that obtained in previous studies with a variety of other polyclonal and monoclonal antibodies to desmosomal glycoproteins. In the heart the staining was confined to the intercalated discs (Fig. 5). Staining was also found in bovine corneal, lingual and oesophageal epithelia (not shown), and in two bovine simple epithelia: kidney tubule and liver.

Staining of bovine epidermis was found predominantly in the desmosome-rich regions of cell boundaries, with negligible staining over the cytoplasm and nucleus. In desmosomes cut transversely, the staining was located predominantly in the intercellular space (Fig. 6). Quantitative analysis of 71 labelled desmosomes as described by Miller et al. (1987) showed that 68% of the label was in the intercellular space, 17% on the membrane and 15% in the plaque or satellite zones of the desmosome. The maximum distance between the antigen molecule and the gold particle with this immunolabelling technique has been estimated as 12 nm (Tolson et al. 1981) to 15 nm (Steinberg et al. 1987). Our data therefore firmly locate the N-termini of the dg2/3 glycoproteins within the desmosomal intercellular space, which is approximately 33 nm wide. A similar distribution of staining was found with monoclonal antibody 07–4D. Monoclonal antibody 52–3D did not react with formaldehyde-fixed ultra-thin frozen sections.

Reactivity of the anti-peptide serum and the monoclonal antibodies was compared in more detail by blotting onedimensional peptide maps of dg2 and dg3 obtained by V8 protease digestion. Coomassie Blue staining showed eight major fragments of relative molecular masses 21000, 19 000, 18000, 17 000, 15400, 14 000, 12 000 and 11500. Fig. 7, lane 1 shows data for dg3. The M_r values of the fragments derived from dg2 and dg3 were identical.

Blotting of the maps with a guinea-pig polyclonal antiserum identified six major bands for each protein digest (Fig. 7, lane 2). A polypeptide of 23 000 M_r is seen in some digests of both dg2 and dg3 (Fig. 7, lane 1, arrow head), when present it is identified by the guinea-pig antiserum, it may represent a product of incomplete digestion. The M_r values of these bands were the same as those found by Coomassie Blue staining except that the 17 000 and 11 500 bands were absent. The anti-peptide serum recognised a 19000 band in dg2 and dg3 digests (Fig. 7, lane 3). Monoclonal antibody 07–4D also recognised the same 19000 band and showed slight reactivity with the 18000 band in digests of both molecules (Fig. 7, lane 4). Monoclonal antibody 52–3D reacted with the 12000 band in dg3 digests (Fig. 7, lane 5), while in dg2 digests it reacted with the same band but showed additional reactivity with bands of 17 000 and 11500 (result not shown). None of these antibodies showed any reactivity with V8 protease alone (result not shown). Monoclonal antibody 07–4D did not show binding to the N-terminal peptide on ELISA.

The amino-terminal amino acid sequence of desmosomal glycoproteins 2 and 3 has been directly determined. Antiserum raised against a peptide corresponding to the N-terminal sequence reacts with both the dg2 and dg3 molecules. This confirms that the sequence relates to these molecules and shows that the two proteins are closely related.

Although the data are based on protein sequence analysis and minor variations may be expected when the DNA sequences become available, a striking comparison with the known sequences of cadherins has emerged. The N-terminal sequence of dg2/3 has 43% identity with N-cadherin. In addition there are three conservative substitutions at positions that are highly conserved throughout the cadherin family. The similarity between the N-terminal sequence of dg2/3 and the cadherins is intriguing, given that both are mediators of calciumdependent adhesion.

The similarity between dg2 and dg3 is stressed by the apparent correspondence of the one-dimensional peptide maps obtained by V8 protease digestion. Our results thus support the significant similarities between the two molecules suggested by the immunochemical studies of Cohen et al. (1983) and the biochemical studies of Kapprell et al. (1985), although we differ from these two groups of workers in that we find no differences between the peptide maps of the two proteins.

Our immuno-electron microscopical studies demonstrate that the amino termini of the dg2 and -3 molecules are located in the intercellular space of the desmosome, i.e. on the extracellular face of the plasma membrane. This view is reinforced by the demonstration that monoclonal antibody 07–4D, which also localises to the intercellular space, recognises the same 19000M_r V8 protease fragment as the N-terminal peptide antiserum. The antipeptide serum and 07–4D appear to react with different regions of the amino-terminal fragment, as the monoclonal antibody does not react with the amino-terminal peptide. The monoclonal antibody 52–3D (Parrish et al. 1990), which recognises the cytoplasmic domains of dg2 and 3, reacts with a different V8 fragment.

Jones (1988) has demonstrated the presence of E-cadherin as a minor component of the desmosomes of bovine nasal epidermis and lingual epithelium. The E-cadherin in these preparations is quite distinct from dg2/3 in two important respects: first, it is much less abundant; and second, it is present as a single band or a closely running doublet of higher relative molecular weight than dg2. Thus, there appears to be no possibility of confusion between E-cadherin and dg2/3. The latter have been clearly identified in the present study by two monoclonal antibodies and one antiserum, as well as by their peptide maps, which are both identical to each other and similar to those previously reported (Cohen et al. 1983; Kapprell et al. 1985). There is no evidence from our work that our anti-peptide serum recognises a polypeptide corresponding to E-cadherin.

In conclusion, these results support the view that dg2/3 are directly involved in desmosomal adhesion (Cowin et al. 1984b). It will be of interest to discover the extent to which the similarity with members of the cadherin family persists as more of the dg2 and dg3 protein sequences becomes available. Two further similarities that may be expected are that the dg2/3 molecules will emerge as transmembrane proteins with single membrane-spanning domains (Parrish et al. 1990), and that calcium-binding domains, as found in uvomorulin (Ringwald et al. 1987), may be expected in dg2/3 because of their calcium-binding properties (Mattey et al. 1987; Steinberg et al. 1987). It may be speculated that similarities between dg2/3 and the cadherins may reflect their calcium-dependent adhesive function while differences may reside in regions involved in interactions with different cytoskeletal and junctional proteins.

We thank John Findlay for advice on N-terminal protein sequence determination, Mike Calder for performing protein structure predictions, Ian Giles for advice on protein sequence comparison, Janice Baker for technical assistance, Derek Mattey for desmosomal antibodies and Bridget Warland for typing the manuscript. The work was supported by the Agriculture and Food Research Council and the Cancer Research Campaign.