Immunologically unique and common domains within a family of proteins related to the retina Ca²⁺-dependent cell adhesion molecule, NcalCAM

Ca²⁺-dependent cell-cell adhesion molecules have been implicated in a variety of events during development including compaction of the mouse embryo (Hyafil. Morello, Babinet & Jacob, 1980; Shirayoshi, Okada & Takeichi, 1983), tissue interactions at epithelial-mesenchymal interfaces (Chuong & Edelman. 1985) and axonal guidance (Bixby, Pratt, Lilien & Reichardt, 1987). Furthermore, these molecules have been shown to be expressed at specific spatial and temporal points in development (Thiery et al. 1984; Hatta & Takeichi, 1986; Nose & Takeichi, 1986). At the cellular level, calcium-dependent cellcell adhesion systems have been demonstrated on a variety of embryonic chick (Takeichi, Ozaki, Tokunaga & Okada, 1979; Grunwald, Geller & Lilien, 1980; Magnani, Thomas & Steinberg, 1981; Thomas et al. 1981), mouse (Atsumi & Uno, 1979; Takeichi et al. 1981 ; Hyafil, Babinet & Jacob, 1981; Ogou, Okada & Takeichi, 1982) and rat (Ocklind, Odin & Obrink, 1984) cells and on several in vitro cell lines (Takeichi et al. 1981; Ogou et al. 1982; Takeichi, 1977; Urushihara. Ueda, Okada & Takeichi, 1977). Prior work has strongly implied the existence of tissue-specific differences among Ca²⁺-dependent adhesive systems (Takeichi et al. 1979; Thomas et al. 1981; Takeichi et al. 1981; Urushihara et al. 1977; Thomas, Thomson. Magnani & Steinberg, 1981; Atsumi & Takeichi. 1980). More recently, monoclonal antibodies have defined two major groups of Ca²⁺-dependent adhesive systems; one restricted to nerve and muscle and a second to epithelial tissues (Atsumi & Takeichi. 1980; Hatta, Okada & Takeichi, 1985; Takeichi, Hatta & Nagafuchi, 1985). In the latter group, there appears to be a number of independently identified components all of which are strikingly similar and which may prove to be identical or interspecific homologues (Takeichi et al. 1981; Hyafil et al. 1981; Ogou et al. 1983; Hyafil etal. 1980; Damsky etal. 1983; Damsky, Wheelock, Damjanov & Buck, 1985; Behrens, Birchmeier, Goodman & Imhof, 1985; Yoshida-Noro, Suzuki & Takeichi, 1984; Yoshida & Takeichi, 1982; Bertolotti, Rutishauser & Edelman, 1980; Gallin, Edelman & Cunningham, 1983). Each is a cell surface glycoprotein of approximately M_r127000 which is protected from tryptic digestion by Ca²⁺. Additionally, in each case a fragment of approximately M_r80000 has been identified that is resistant to proteolysis in the presence of Ca²⁺.

In our laboratory, we have identified a closely related Ca²⁺-dependent adhesive component, gpl30, on embryonic chick neural retina cells (Grunwald, Pratt & Lilien, 1982; Cook & Lilien, 1982). In common with other Ca²⁺-dependent cell adhesion molecules, it is approximately M_r 130000 and is protected from tryptic digestion by Ca²⁺ (Cook & Lilien, 1982). Furthermore, it is endogenously cleaved to generate a fragment of approximately M_r90000 also protected from tryptic digestion by Ca²⁺ (Grunwald et al. 1982; Cook, Pratt & Lilien, 1984). We have shown that the presence of this component at the cell surface is positively correlated with the ability of the cells to form Ca²⁺-dependent adhesions. This is true whether the cells are dissociated from the tissue under conditions where gpl30 is initially present (Grunwald et al. 1982) or whether the cells must repair in culture to regain the competence to form Ca²⁺-dependent adhesion (Geller & Lilien, 1983).

We now report that rabbit polyclonal antibodies prepared to gp90 recognize gpl30 and inhibit Camdependent cell-cell adhesion. Of all the tissues tested, one of the antibody preparations recognizes a component of M_r 130000 on neural tissues only. The second recognizes a polypeptide of identical molecular weight in all tissues examined, including nerve, muscle and various epithelia. We put forth the hypothesis that Ca²⁺-dependent cell-cell adhesion is mediated by a family of closely related molecules with immunologically distinct subgroups and refer to these as cal CAMs, reflecting usage of the acronym CAM for cell adhesion molecule (Edelmann, 1984).

Gel filtration fractions containing gp90 were collected and pooled (fraction 2A) as described previously (Grunwald et al. 1982; Cook et al. 1984). This pooled material was made up to 80% saturation in ammonium sulphate. (enzyme grade. Schwarz-Mann), stirred slowly (1h; 4°C) and the precipitate collected by centrifugation (15 000g; 4°C; 20 min). The precipitate was redissolved in 10 mM-Hepes, 0–15 M-NaCl (HBS), 1 mM-CaCl₂, pH 7–0 and dialysed overnight against this buffer. The dialysed material was centrifuged (15 000g; 4°C; 1 h) and the supernatant fluid was made 80% v/v in ethanol and stored overnight at −20°C. The precipitate so formed was collected by centrifugation (15000g; 4°C; 1h) and redissolved in isoelectric focusing (IEF) buffer (Cook & Lilien, 1982).

Culture medium (8h) from surface ¹²⁵I-labelled (Fraker & Speck, 1978) 10-day chick retinas (ten) was prepared and calcium trypsinized as previously described. We have demonstrated that tissue culture conditioned medium prepared in this way contains gp90 as the major, labelled high molecular weight species (Cook et al. 1984). The ethanol (80%) precipitate of this material was dissolved in IEF buffer and added to the unlabelled ethanol precipitate (above) of fraction 2A redissolved in the same buffer. The combined sample was heated (100°C; 3 min) just prior to isoelectric focusing.

Isoelectric focusing of this mixture was performed on a slab of polyacrylamide gel (5% acrylamide, 0·17% bisacrylamide) containing 8M-urea (Schwarz-Mann), 4% v/v Nonidet P40, 2% w/v ampholytes (pH2-ll; AG 2-11 Servalytes; Accurate Chemical Scientific Corp.) and 10% w/v glycerol. Gels were prefocused (200V; 20min) with IEF sample buffer; after the prefocusing, the sample containing calcium-trypsinized culture medium was applied to the gel. The gel was focused (400 V, 17h then 1000 V, 4–7 h) until the current had fallen to a steady minimum.

After focusing, the gels were removed, washed (1h) in 20% v/v glycerol and dried under vacuum (80°C). The dried gels were autoradiographed (−70°C) using Kodak XAR-5 film with Cronex lightening plus intensifying screens.

Strips of gel corresponding to the most intense radioactive bands were excised, reswollen in SDS sample buffer and homogenized. The homogenized sample was heated (100°C; 3 min) and applied to a 3mm-thick preparative SDS-electrophoresis slab gel (7·5% acrylamide, 0·2% bisacrylamide). After electrophoresis, the gels were treated, dried and autoradiographed in the same manner as the IEF gels. The band at approximately M_r90000, as visualized in autoradiographs, was excised, rehydrated in HBS, 1mM-CaCl₂, pH 7 ·0 and homogenized. For injection into rabbits, the homogenized sample was made 50% v/v (2·5–3 ·0 ml total volume) with Freund’s complete adjuvant (GIBCO) for the first injection and Freund’s incomplete adjuvant for the two subsequent injections. The three preparations of gp90 were each equivalent to 61 of tissue culture medium, conditioned by 3360 10-day chick embryonic retinas.

Two adolescent, female, New Zealand White rabbits were used to raise the antisera to gp90. After a preimmune bleeding, each rabbit was challenged with antigen by subcutaneous injection at six locations along the back. Two additional subcutaneous injection series were made 4 weeks and 10 weeks after the primary injection.

Rabbits were bled by vein puncture of the marginal ear vein. Sera were prepared from preimmune bleedings, weekly bleedings between the second and third injection and weekly, for one month, after the last injection. Sera pooled from all bleedings prior to the third injection were used in these studies.

¹²⁵I-labelled cell surface components or ¹²⁵I-labelled fraction 2A (Grumwald et al. 1982) were immunoprecipitated using fixed Staphylococcus aureus (Pansorbin, Calbiochem-Behring Corp.) or protein A-Sepharose (Sigma Chemical Co.). Aliquots (100μ1) of immune serum or preimmune serum (PIS) were incubated with bound protein A (40μl; 4h; ice) and washed three times in ‘blotto’. These aliquots were stored at −20°C until use. Cells from trypsin-dissociated tissue were allowed to repair (5 h) in rotation culture prior to labelling by lactoperoxidase catalysed iodination (Geller & Lilien, 1983). Repaired cells (approx. 5×10⁸) were homogenized in 1ml HST (20mM-Hepes, pH7·2, 150mM-NaCl, 1% v/v Triton X-100) containing: 1mM-MgCl₂, 1 mM-CaCl₂, 10mM-iodoacetamide, 500μg BSA, 100v μ g DNase and 50μg each of antipain, leupeptin and chymostatin. After an incubation (30min; ice), the insoluble residue was pelleted (9000g; 2min). The supernatant fluid (500μl) was made 1% v/v in SDS, heated (3 min; 100°C) and diluted with HST (500 μ 1). Fraction 2A (Grunwald et al. 1982) or cell homogenate was incubated with antisera or preimmune sera bound to protein A for 3 h on ice with gentle agitation. The bound protein A-serum-antigen complex was washed four times in ‘blotto’ followed by two washes in HST containing 0 · 25 % SDS and 50 μ g ml⁻¹ antipain. Bound antigen was eluted in IEF sample buffer and analysed by two-dimensional gel electrophoresis.

Tissue extracts were subjected to SDS-PAGE on 7·5 % gels. Proteins were transferred (4h; 60 V) to nitrocellulose sheets (Bio-Rad Labs) in transfer buffer (Towbin, Staelin & Gordon, 1979) containing 0·1% w/v SDS. The nitrocellulose sheet was treated with ‘blotto’ (Johnson, Gautsch, Sportsman & Eldler, 1984) for 1 h at 37°C, incubated (overnight; 4°C) with the rabbit antiserum diluted (1:250) in ‘blotto’, washed five times in ‘blotto’ and incubated with either ¹²⁵I-labelled goat anti-rabbit IgG or horseradish peroxidase-goat anti-rabbit IgG diluted 1:10³ in ‘blotto’. The sheets were washed in ‘blotto’ followed by five washes in 10 mM-Tris, 150mM-NaCl, pH 7·4 containing 0·3% v/v Tween-20 and 01 % w/v SDS. Finally the nitrocellulose sheets were autoradiographed. The HRP blots were washed in ‘blotto’, and developed with HRP colour development reagent (4-chloro-l-naphthol; BioRad).

The IgG fraction from rabbit serum (4 ml) was prepared from DEAE Affi-Gel Blue (Bio-Rad Labs), as described by the manufacturer’s protocol and lyophilized.

Fab’ fragments were prepared from the IgG preparation by the method of Brackenbury, Thiery, Rutishauser & Edelman (1977) as modified by Grunwald et al. (1982).

Cells from 10-day embryonic chick neural retina tissue were trypsinized in the presence of 1 mM-Ca²⁺ as previously described (Grunwald et al. 1980). These cells (10⁷ml⁻¹) were incubated (30 min; 0°C) in HBS containing 2 mg ml⁻¹ glucose, 3ntM-KCl, 1 mM-CaCl₂, antipain (5μgml⁻¹), cycloheximide (5μgml⁻¹) and serum, IgG or Fab’ fragments. When serum or IgG was used, the cells were pelleted and resuspended in HBS (5 × 10⁶ cells ml⁻¹) containing Fab’ fragments of goat anti-rabbit serum (500gml⁻¹) directed against the Fab’ portion of rabbit IgG (Springer & Barondes, 1980) and incubated (30 min, 0°C). The cell suspension was aliquoted into a 96-well tissue culture plate (Falcon microtiter III; 0·25×10⁰ cells per well, 50μl total volume per well) and incubated (37°C; 1 h) in rotation culture (175 rev min⁻¹). Each well had been previously coated with agar (50 μ l; 1 · 5% w/v in HBS). At selected time points, the cultures were examined microscopically and/or the number of single cells remaining in suspension counted on a Coulter counter. Inhibition of adhesion was calculated as %I = Ce — Cc/Co × 100, where Ce is the number of single cells remaining in suspension in the presence of Fab’ and Cc is the number of single cells remaining in suspension in the presence of 1 mg ml⁻¹ control Fab’ and Co is the number of input single cells. The concentration of RR1 or RR2 Fab’s varied from 0 to 1 mg ml⁻¹. Each well received control Fab’ to bring the total concentration of Fab’ to 1 mg ml⁻¹.

Neural retinas from 9- or 10-day chick embryos were digested with trypsin in the presence of 1 ntM-Ca²⁺ and the resulting cells iodinated as previously described (Cook & Lilien, 1982). Iodinated cells were homogenized in ice-cold 10mM-Hepes buffer (pH 7 · 5) containing lmM-MgCl₂, 100 μ g ml⁻¹ DNase, 50 μ g ml⁻¹ antipain, 50 μ g ml⁻¹ chymostatin and 50 μ g ml⁻¹ leupeptin (approx. 10^scells ml⁻¹) by repeated passage (50 times) through a 22-gauge syringe needle. The homogenate was centrifuged (20000g; 30min; 4°C) and the pellet washed by homogenization and centrifugation as above. For two-dimensional Western blots cell membranes were prepared from intact neural retina.

Crude membranes from iodinated cells or their Nonidet P40 insoluble residue and ¹²⁵I-labelled fraction 2A were fractionated by two-dimensional gel electrophoresis and the required spots excised as described previously (Cook et al. 1984). Where specified, the excised spots were reiodinated and reisolated by SDS-PAGE (Cook et al. 1984).

Each spot was equilibrated three times (30 min) with aliquots (1ml) of 50mM-ammonium bicarbonate (pH 8-0) containing 0 · 17% w/v SDS and then homogenized in the same buffer (500 μ l). After 1 h at room temperature, the homogenate was centrifuged and the supernatant fluid removed. The pellet was extracted again as above and the combined supernatants filtered through glass wool. The eluted material was concentrated and freed of ammonium bicarbonate by repeated lyophilization. Recoveries were generally greater than 80%.

Complete tryptic digests of reiodinated NcalCAM and gp70 were prepared and compared by two-dimensional peptide mapping based on the procedure of Elder, Picket, Hampton & Lerner (1977) as modified in our laboratory (Cook et al. 1984).

Endoglycosidase digestions were performed on NcalCAM and gp70 prepared from iodinated cells and reiodinated gp90 from turnover medium (Oli.u.ml⁻¹ of endoglycosidase D, 0 · 1i.u.ml⁻¹ of endoglycosidase H or 10i.u.ml⁻¹ of endoglycosidase F). All digestions were performed for 10 h at 37 °C in 0 · 1 M-Tris-HCl buffer (pH 6 · 5 for D; pH5 · 5 for H [Miles Scientific]; pH 6-1 for F [New England Nuclear]) containing 0 · 1% w/v SDS, 1% v/v Nonidet P40 and 1 % v/v 2-mercaptoethanol. The protease inhibitors antipain, chymostatin and leupeptin (50 μ g ml⁻¹ each) were employed in digestions with endoglycosidases D and H; 50mM-EDTA was used for this purpose in endoglycosidase F digestions. The digested samples were resolved by SDS-PAGE on 7-5 % polyacrylamide slabs. In order to maximize the separations, the slabs were run until the bovine serum albumin marker had neared the bottom of the gel. The autoradiographs were projected on a screen, traced and changes in mobility calculated from the centre of density of each band relative to internal bovine serum albumin markers.

Calcium-dependent cell-cell adhesion among chick neural retina cells is mediated by a cell surface glycoprotein termed gp!30 (Grunwald et al. 1982; Cook et al. 1984) and here referred to as NcalCAM for neural, calcium-dependent, cell adhesion molecule. This molecule is turned over into tissue culture medium as a fragment of M_r90000 (Cook etal. 1984). We have purified gp90 by two-dimensional gel electrophoresis and used this highly purified preparation to prepare two polyclonal antisera referred to as RR1 and RR2. The purity of the immunogen was assessed by iodination of a sample and analysis by 2-D SDS-PAGE. Only one component at M_r90000, pl 4 · 8 was seen (data not shown).

Calcium-dependent adhesion among retina cells is inhibited by both antisera directed against gp90, a fragment of NcalCAM (Cook et al. 1984). Fig. 1 shows representative photomicrographs of 3h cultures in the presence of 2 mg ml⁻¹ RR1 Fab’. Fig. 2 shows the effect of antibody dosage on inhibition for RR2 Fab’. Cultures containing 1 mg ml⁻¹ of antibody are visually (not shown) and quantitatively indistinguishable from those containing EGTA. In the presence of EGTA, no Ca²⁺-dependent cell-cell adhesion can occur; however, small clusters are formed (Grunwald et al. 1980) resulting in a value of approximately 30% for maximal inhibition.

The specificity of the two antisera for the immunogen (gp90) and cell surface NcalCAM was tested by immunoprecipitation. Antisera from both rabbits immunoprecipitated only gp90 from ^l25I-labelled fraction 2A derived from tissue culture conditioned medium (Fig. 3). In addition, the antisera immunoblot a M_r90000 component from fraction 2A. Like isolated gp90 (Cook et al. 1984) the immunoprecipitated component is protected from typsinization by both Ca²⁺ and Mn²⁺ but not Mg²⁺ (data not shown).

Immunoprecipitates of ^l25I-labelled cells by RR1 were separated by 2-D SDS-PAGE and contained two surface-labelled components, corresponding in molecular weight and isoelectric point to NcalCAM (pl 4 · 8) and an M_r 70000 component also with a pl of 4 · 8 (Fig. 4). We have previously suggested that the M_r70000 component is generated from NcalCAM (Grunwald etal. 1982; Cook & Lilien, 1982). This was based on the fact that there are only two heavily labelled species in iodinated intact tissues one of which is NcalCAM; on digestion with trypsin in the presence of Ca²⁺ NcalCAM is reduced and the M_r 70000 component (gp70) appears.

Immunoblots of 2-D SDS-PAGE of membranes prepared from retina tissue reveal that each antiserum binds solely to NcalCAM (Fig. 5). gp70 (pl 4 · 8) does not appear on immunoblots of 2-D SDS-PAGE of retina membranes or tissue, consistent with our prior suggestion that it is generated during the trypsin treatment (Cook & Lilien, 1982). Some additional spots are sometimes seen; these, however, vary from blot to blot. The blot shown in Fig. 5 has three weak spots at M_r65 000, pl 6 – 6 · 5 which were only seen in this set.

The tissue specificity of each antiserum was tested by immunoblotting a series of embryonic chick tissue homogenates after 1-D SDS-PAGE (Fig. 6). Of all the tissues tested, RR1 recognizes a component identical in molecular weight to NcalCAM only in retina, cerebral lobes, optic lobes, hind brain and spinal cord (Fig. 6A). The components other than the M_r 130000 recognized in whole tissue homogenates are highly consistent and also limited to neural tissues with the exception of a band at approximately M_r 97 000 in kidney. As they do not appear in the 2-D SDS-PAGE immunoblots of membranes (Fig. 5) or immunoprecipitates of surface iodinated retinas (Fig. 4) they apparently represent a non-membrane neural-specific epitope. In contrast to RR1, RR2 immunoblots a component equivalent in molecular weight to NcalCAM in all tissues tested, which include liver, kidney, lung, stomach, skin, heart, leg and breast muscle, cerebral lobes, optic lobes, hind brain, spinal cord, neural retina and pigmented retina. As with RR1, when whole tissue homogenates are used additional components are recognized. Most consistent is a band at M_r 70 (XX). This component has a pl of approximately 5-5 and can easily be distinguished from gp70/4 · 8. As this component is not seen in immunoblots of membrane, it is also apparently a non-membranous component containing an epitope crossreactive with NcalCAM.

To test further the cell-type specificity of RR2, liver was separated into fibroblastic and parenchymal cells by a 2 h incubation in tissue culture dishes followed by transfer of the less adherent population to new tissue culture dishes. Cultures of aortic endothelium were also prepared. Immunoblots of these separate populations using RR2 show that both contain a component equivalent in molecular weight to NcalCAM (data not shown).

The M_t 130000 component recognized by each antiserum is the same component; homogenates of retinal tissue immunoprecipitated by RR1, separated by SDS-PAGE and immunoblotted with RR2 show a single major component at M_r130000 (not shown). All of these data taken together indicate that a molecule related to retina NcalCAM exists in all tissues and that, in neural tissue, this molecule has at least one domain that serves to identify the molecule as neural in origin.

We have previously shown by 2-D peptide mapping that gp90 is a fragment of NcalCAM (Cook el al. 1984). Both our previous data, mentioned above (Grunwald et al. 1982; Cook & Lilien, 1982), and the immunoprecipitations presented here suggest that gp70 is a fragment of gpl30. To reinforce this relationship the two polypeptides were obtained from two-dimensional gels, reiodinated and subjected to two-dimensional tryptic peptide mapping. As can be seen in Fig. 7 the two polypeptides are very similar. As might be expected there are several distinct peptides in each. However, out of a total of 22 iodinatable peptides 14 (labelled A to O) clearly migrate to the same position, indicating a close relationship.

Endoglycosidases D, H and F were used alone and in combination on each ¹²⁵I-labelled component, NcalCAM, gp90 and gp70. Changes in apparent molecular weight were observed and quantified from 1-D SDS-PAGE autoradiographs. A representative set of digestions for NcalCAM is shown in Fig. 8. Minor bands seen in lanes 3, 4 and 8 are due to protease contaminants in endoglycosidase H. Table 1A records changes in apparent molecular weight on digestion with each endoglycosidase alone and in combination. Endoglycosidase F (F) removes all N-linked oligosaccharides, both high mannose and complex (Elder & Alexander, 1982). Endoglycosidase H (H) removes only the high mannose, AMinked oligosaccharides (Tarantino & Maley, 1978) and, in our case, endoglycosidase D (D) (Muramatsu, 1978) removes a subset of the high mannose chains. On the basis of these specificities, it is possible to define three classes of N-linked oligosaccharides on NcalCAM and its fragments. Class I, defined by sensitivity to D, H and F, represents the subset of high mannose chains sensitive to D. Class II represents the high mannose chains not cleaved by D and class 111 represents complex N-linked chains. Table IB records the amount, × 10⁻³, of chain class present. Since the apparent mobility of gp70 is altered by only 1000 on removal of class III chains while NcalCAM and gp90 are each altered by 8000, NcalCAM must have at least two class 111 chains shared with gp90, one of which is missing from gp70. NcalCAM thus contains a minimum of four N-asparagme-linked oligosaccharide chains: two class 111, one class II and one class I. gp90 contains 1000 less of class II chains than NcalCAM; this could represent a portion of a class II oligosaccharide chain cleaved during its generation or it could represent a completely missing chain and therefore represent a portion of NcalCAM not present in gp90. In the latter case, NcalCAM would have a minimum of five oligosaccharide chains. Like gp90, gp70 lacks a small amount of class II chains but is also lacking all class I chains and most of the class III chains. A pictorial summary of the data in Table 1 along with the approximate lineal relationship of the three polypeptides is shown in Fig. 9.

Two rabbit polyclonal antisera were raised to gp90, a fragment of NcalCAM (gpl30) purified from chick retina tissue conditioned medium. Both of these antisera inhibit Ca²⁺-dependent cell-cell adhesion and recognize NcalCAM on tissue and cells, gp90 in turnover medium and gp70 on cells prepared by trypsin treatment of retina tissue. When tested by immunoblotting on a panel of embryonic chick tissues, RR1 recognizes a component corresponding to the molecular weight of retinal NcalCAM (M_R 130000) in all neural tissues but in no other tissues tested. In contrast, RR2 recognizes the M_r 130000 component in all tissues tested.

These data suggest to us that there is a family of closely related, yet immunologically distinct components related to the chick neural retina Ca²⁺-dependent adhesive component, NcalCAM, and that these components are involved in calcium-dependent cell-cell adhesion. This is consistent with the recent work of Hatta and colleagues (Hatta & Takeichi, 1986; Nose & Takeichi, 1986; Hatta et al. 1985; Takeichi et al. 1985). They have prepared three monoclonal antibodies that exhibit tissue-specific inhibition of Ca²⁺-dependent cell-cell adhesions. One antibody recognizes a calcium-dependent adhesion molecule present in nerve, striated muscle and cardiac muscle, termed N-cadherin. The second antibody recognizes a similar molecule, termed E-cadherin, in a variety of epithelial tissues. N-terminal amino acid sequence data reveal complete identity to seven residues for the two cadherins (Shirayoshi et al. 1986). A third cadherin, P-cadherin, has most recently been identified although its relation to the other two is unknown (Nose & Takeichi, 1986). On the basis of the similarity in molecular characteristics and tissue distribution of the mouse-derived molecules E-cadherin, uvomorulin (Hyafil et al. 1980) cell CAM 120/80 (Damsky et al. 1983) and chick L-CAM (Gallin et al. 1983) Hatta & Takeichi (1986) proposed that they are equivalent. The neural specificity of RR1 demonstrates that there are domains unique to certain tissue types within the broader classes of cadherin proposed by Hatta et al. 1985 and Takeichi et al. 1985 and thus that additional categories of Camdependent adhesive components may exist. Additionally, our data make explicit the assumption that the molecules in all these classes are closely related both immunologically and structurally.

The Ca²⁺-dependent adhesive components identified from a variety of sources share certain properties. All are protected from typsinization by Ca²⁺ and are cleaved to an M_r 80 000-90 000 fragment (uvomorulin, Hyafil el al. 1980; cell CAM 120/80, Damsky et al. 1983, 1985; L-CAM, Gallin et al. 1983; Cunningham et al. 1984; NcalCAM, Cook & Lilien, 1982; Cook et al. 1984; E-cadherin, Ogou et al. 1983). Furthermore, the structural analyses reported here on NcalCAM show striking similarity to those done on L-CAM (Cunningham et al. 1984).

The results of our analysis of NcalCAM and its fragments are summarized in Fig. 9. We have previously demonstrated that gp90 is a fragment of NcalCAM. Gp70 was originally shown to be generated during trypsinization of intact tissue, most probably from NcalCAM (Cook & Lilien, 1982). We now further strengthen this relationship showing immunological cross reactivity and structural similarity. gp90 contains most of the N-asparagine-linked carbohydrate and is released from cells; this polypeptide is thus most likely a distal fragment. Conversely, gp70 remains associated with cells after trypsinization and lacks most of the N-asparagine-linked sugar and thus is most likely a proximal, or cell surface-associated, fragment. Since we cannot exactly position these peptides relative to each other we have indicated ambiguity by incorporating dashed lines at the ends of gp70 and 90. Like LCAM, NcalCAM has a minimum of four N-asparagine-linked oligosaccharide chains which are present on a distal N-terminal fragment released from cells during culture. We are at present performing a detailed comparison of NcalCAM and L-CAM to elucidate further the relationship between these molecules.

Based on the immunological reactivities reported here and the data of Hatta and colleagues (Hatta & Takeichi, 1986; Hatta et al. 1985; Takeichi et al. 1985) we propose that there is a hierarchy of immunologically specific domains within the members of a family of structurally similar Ca²⁺-dependent adhesive components ranging from pan-specific to tissue-specific. This is exemplified by the pan-specificity of RR2, the nerve-muscle specificity of the antibody reported by Hatta and colleagues (Hatta & Takeichi, 1986; Hatta et al. 1985; Takeichi et al. 1985) and the neural specificity of RR1. We further propose that these specific domains will be important in determining the specificity of cellular interactions during development.

This work was supported by a grant from the National Science Foundation. S.L.C. is supported by a NIH Training Centre Grant, HD07118.