Adhesion among Neural Cells of the Chick Embryo I. AN IMMUNOLOGICAL ASSAY FOR MOLECULES INVOLVED IN CELL-CELL BINDING*

method for the quantitative assay of molecules involved in cell adhesion is described. Three The cell

method for the quantitative assay of molecules involved in cell adhesion is described. Three observations served as a basis for this assay: (a) cells obtained by trypsinization of retinal tissue aggregated rapidly, provided they had been allowed to recover in culture from the dissociation process; (b) treatment of chick retinal cells with Fab' fragments from rabbit antibodies against these cells prevented their aggregation; and (c) incubation of these antibody fragments with antigens released by retinal cells in culture neutralized their ability to inhibit aggregation.
The amount of neutralizing antigen was determined by measuring the rates of cell aggregation in the presence and absence of antibody and antigen using a particle counter. Although adhesion was inhibited by antiretinal cell antibodies, it was not affected by lectins or anti-carbohydrate antibodies that also were bound to the cell surface. Together, the results suggest that the inhibition involved blockade or inactivation of particular cell surface molecules and that the retinal cell antigens capable of neutralizing the antibodies represented these molecules or their fragments. In the accompanying paper, we describe the use of this assay for the purification from culture supernatants of a cell surface molecule involved in cell to cell adhesion.
Cell-cell recognition and adhesion appear to be important processes in tissue formation, and a variety of intercellular junctions have been described by morphological methods. Nevertheless, there are few developmental events in which the role played by these interactions has been defined precisely. The major difficulty has been that interactions among cells in tissues are complex, so that identification of the molecules involved and analysis of their functions requires the development of in vitro systems that are comparatively simple, yet retain relevance to normal embryogenesis. A frequently used approach has been to study the aggregation of cells obtained from embryonic tissues. Adhesion among these cells has been assessed by a variety of techniques: measurement of the size of aggregates formed over relatively long time periods (1); determination of the initial rate of aggregation in rotary cultures using a particle counter (2,3); and monitoring the binding of radioactive or fluorescent cells to cell aggregates (4) or monolayers (5-7). Using these assays, variations in binding among cells from different tissues have been observed (1, 4, 6-101, and several models have been proposed to account for the results (1,6,9,(11)(12)(13). To test these models, several workers have attempted to identify molecules involved in adhesion (6, 9, 14-17). Because of differences in the methods used, however, the results obtained have been difficult to compare, and little consensus has emerged concerning the mechanism or specificity of cell adhesion.
We have also examined the aggregation of cells from dissociated embryonic tissues and describe here an immunochemical approach to the identification of molecules involved in cell adhesion. First, cells dissociated by trypsinization (18) are cultured in suspension to regenerate surface proteins removed or damaged during the dissociation process. The adhesiveness of these cells is then measured in terms of their initial rate of binding (2). Substances involved in adhesion are obtained from cultures (19) of retinal tissue and detected by an assay based on the specific neutralization of antibodies that inhibit initial formation of cell-cell bonds. This paper describes the general procedures used to prepare the cells, antibodies, and antigens required for analysis of cell adhesion in neural tissues, as well as the rationale behind the methods for quantitative detection of molecules involved in cell adhesion. The accompanying paper (20) describes the use of these techniques to purify and characterize such a molecule from retinal cells, to prepare specific antibodies to it, and to demonstrate its presence on the cell surface. to prevent cell aggregation. The flasks were then placed in an ice bath and stirred at 400 rpm for 20 min. The cells were pelleted for 5 min at 2000 rpm at 4" in a clinical centrifuge, resuspended in 2 ml of SME medium, layered on top of a step gradient consisting of 10 ml of cold SME medium containing 35 mg/ml of bovine serum albumin and a 0.8-ml cushion of isotonic 350 mg/ml of albumin (35% bovine albumin, Miles Laboratories), and then pelleted for 5 min at 2000 rpm at 4". The cells that formed a layer on the cushion were removed with a pipette and washed three times in cold SME medium.

Preparation
This procedure yielded single cells with only a few small aggregates.
The cell viability was 85 f 10% as judged by trypan blue exclusion. To

RESULTS
The development of an assay for molecules involved in cell adhesion required: (a) use of a quantitative system for evaluating the adhesiveness of individual cells (2); (b) preparation of antibodies capable of inhibiting cell adhesion; (c) the demonstration that these antibodies do not inhibit adhesion simply by coating the cell surface; and (d) the preparation of cellular antigens that can neutralize the ability of the antibodies to inhibit adhesion.

Aggregation
of Retinal Cells in Suspension -In Fig. 1, a and b are shown lo-day-old retinal cells before (a) and after (6) rotation at 90 rpm in suspension for 30 min at 37". The aggregates formed were often irregular in shape, unlike the more rounded "histotypic" aggregates that are observed after 24 h of incubation (1). As the cells aggregated, the total number of particles in suspension decreased. As reported by other workers (2, 31, the rate of decrease in particle number is a convenient measure of aggregation and can be rapidly and precisely determined using a particle counter (Fig. 2).
To evaluate critically the specificity of cell adhesion with respect to tissue source and developmental age, we previously devised an assay that directly quantitates individual binding events between pairs of single cells of the same or different type (6). It was important to establish that the aggregation assay also quantitatively reflected the adhesiveness of single cells; this was done by comparison of the two assays (Table I).
As we previously reported, the adhesiveness of retinal cells varies with embryonic age. Using cells from 8-, lo-, 12-, and 14-day-old chick embryos, it was found that over a 4-fold range the number of cell-bound cells per mmz (in the two-cell assay) was proportional to the decrease in particle number (A% in the aggregation assay) (Table I). This indicates that the two methods are quantitatively equivalent within experimental error, and therefore that the rate of decrease in particle number accurately reflects the adhesiveness of individual cells.
The retina of a lo-day-old chick embryo contains several cell types. Nevertheless, within 1 h, essentially all the cells obtained from this tissue had become associated with aggregates (see Fig. lb). It has also been observed in the two-cell assay (6) that over 50% of the cells form pairs within 1 h. Together, these results suggest that adhesiveness is a general The arrow indicates the time at which the rate of aggregation was routinely measured in assays. Although cells that had been cultured after trypsinization proteins have reappeared during culture in suspension. Per cent aggregated rapidly, freshly trypsinized cells did not bind to aggregation (solid lines) and the relative amount of total radioiodieach other (Fig. 3). It is likely that this lack of adhesiveness nated cell surface protein (dashed lines) is indicated as a function of reflects the removal by trypsinization of proteins essential for culture time in two experiments (open and closed circles). The amount of protein removed by trypsinization, and the rates of cell-cell binding. When these cells were cultured in suspenincrease in adhesiveness and surface protein varied with different sion, their surface proteins were gradually regenerated and cell preparations.
their adhesiveness increased, reaching a plateau after about 12 h (Fig. 3). We therefore carried out all experiments with inhibited the aggregation of the retinal cells (Figs. lc and 2). cells that were cultured in suspension for 18 to 24 h.
Considerable variation in the amount of inhibition was ob-Inhibition of Aggregation by Anti-R10 Fab' Fragments-served with antibodies from different rabbits subjected to the Immunoglobulin produced in rabbits against retinal cells from same injection schedule (Fig. 2). lo-day-old chick embryos (anti-RlO) was found to agglutinate As shown in Fig. 4, aggregation was inversely proportional retinal cells in suspension. Monovalent Fab' fragments pre-to the logarithm of the Fab' concentration. This relationship pared from the antibodies did not agglutinate, but instead is consistent with a mechanism in which the inhibition bv  4. Effect of anti-R10 Fab' concentration on the aggregation of retinal cells from lo-day-old chick embryos. Per cent aggregation represents the per cent decrease in particle number after 20 min (Fig. 2).  Fab' involves a simple binding equilibrium between the monovalent antibody fragment and a cell surface antigen that is involved in the formation of cell-cell bonds.

Effect on Cell Adhesion of Other Molecules that Bind to
Cell Surface-To test whether anti-R10 Fab' inhibited adhesion by nonspecifically coating the cell surface or by binding to particular cell surface molecules, succinyl concanavalin A, a lectin that binds to cells but does not agglutinate them (26) and Fab' fragments of antibodies that bind to cell surface carbohydrates (24) were also assayed for inhibitory activity (Table II). Although all three reagents bound to the surface of retinal cells giving bright staining in immunofluorescence studies, only anti-R10 Fab' caused a significant decrease in cell aggregation.
The specific assay to detect molecules involved in cell adhesion rests on their ability to neutralize anti-R10 Fab', thus permitting aggregation to occur. Lilien and Moscona (19) showed previously that retinal cells in monolayer cultures release proteins into the medium, and we found that supernatants from 24-h cultures of intact lo-day-old retinal tissue were a convenient and rich source of neutralizing antigens. An example of the results obtained in a typical assay is presented in Table III. Although the retinal cell molecules present in the tissue culture supernatants did not by themselves block or enhance aggregation, they were able to neutralize the ability of anti-R10 Fab' to inhibit cell adhesion.
The relationship between neutralization and the amount of culture supernatant added is complicated by the fact that, in general, antibodies display a wide range of affinities for antigen. The addition of a small amount of antigen would preferentially neutralize those Fab' fragments having a high " See "Materials and Methods" for details of assay conditions b The standard deviation in triplicate assays is shown.
affinity, and because higher affinity antibodies are likely to be more effective in blocking adhesion, their removal should result in a relatively large increase in the rate of adhesion. Consequently, as more and more antigen is added, the effect on aggregation would be expected to decrease gradually. The amount of neutralization, expressed as the per cent decrease in inhibition (Table III), approximated a linear relationship with respect to the logarithm of the supernatant volume added (Fig. 5). Using this relationship, it was possible to estimate the relative amount of Fab'-neutralizing antigen in a sample by its effect on the aggregation of cells in the presence of a constant amount of Fab'. Because of the logarithmic relationships involved, the most reliable estimates of antigen were obtained when enough anti-R10 Fab' was used to inhibit aggregation by 40 to 60% (Fig. 4), and when the neutralizing antigen reduced this inhibition by 25 to 75% (Fig. 5). Because the anti-R10 Fab' from different rabbits varied considerably in its ability to inhibit adhesion (Fig. 2), Fab' from unimmunized rabbits was used to dilute more potent samples so that addition of a total of 1 mg of Fab' caused a 50% decrease in the rate of aggregation. For purposes of quantitation, we have defined 1 unit of neutralizing activity as the amount of antigen required to cause a 25% decrease in the inhibition of adhesion produced by 1 mg of Fab' fragment. The standard deviation in the measurement of activity units (calculated from three aggregation assays, Table III)  Neutralization is expressed as the per cent decrease in the amount of inhibition of aggregation produced by the Fab' (see Table III). One unit of activity is defined as the amount of the 50 times concentrated culture supernatant required to decrease the inhibition by 25%.

DISCUSSION
The major purpose of this report is to describe an assay system developed for detecting molecules mediating cell adhesion. This discussion therefore focuses largely on the evidence supporting the hypothesis that cell adhesion involves particular cell surface molecules, and the evidence that the immunological assay provides a basis for the purification and characterization of cell adhesion molecules.
In no case has rigorous and convincing evidence been reported that a particular molecule is actually involved in adhesion. Several workers have attempted to identify the ligands involved in cell adhesion on the assumption that their addition to cells should enhance or inhibit aggregation, often over relatively long time periods (1,8,9,17). We have not adopted this approach in our current studies for three reasons: (a) substances irrelevant to adhesion may affect aggregation; (b) addition of an external ligand to cells will not necessarily enhance their natural adhesiveness; and (c) when released from the cell surface, potential ligands may not retain their binding activity. Consequently, we have focused our efforts on the development of a specific immunological assay in which initial adhesion is mediated by molecules already present on the surface of the cells. This adhesion was inhibited by antibodies raised against the cells, and antigens released by retinal cells in culture could therefore be tested for their ability to neutralize the inhibitory effect of these antibodies. In order to be detected, a molecule involved in adhesion is required only to retain appropriate portions of its antigenic structure when released from the cell. It should be noted that quantitation of the antigen by this procedure requires an extremely precise determination of cell adhesiveness, in that three measurements of aggregation rates must be made in order to calculate the amount of neutralizing activity. To achieve this precision, we have relied on the excellent reproducibility (21%) of a particle counter. A particular assay is not likely to define all of the mechanisms of cell adhesion, but may help to suggest useful working hypotheses. Three classes of hypotheses have been proposed for the molecular basis of adhesion: (a) formation of cell-cell bonds through different tissue-specific ligands (1,4,9,16); (b) a single molecular mechanism for initial adhesion of cells in which differences in adhesiveness reflect different amounts of control of this process (6, 14, 27); and (c) binding through relatively nonspecific forces (electrostatic, hydrophobic, van der Waal's interactions) between large areas of the cell surface (12). The results of the present and previous (6, 27) studies have a bearing on each of these hypotheses.
Using cells that have recovered from trypsinization, we have previously evaluated the specificity of initial binding among individual retinal and brain cells from embryos of different ages (6, 27). These studies revealed that the majority of retinal and brain cells from embryos of the appropriate ages could bind just as well to cells of different tissue type (i.e. retina to brain) as to cells of their own tissue type. Furthermore, antibodies prepared against a purified fraction of retinal proteins were able to block completely both the homologous and heterologous binding among cells from the two tissues. Together, the results argued against the presence of tissue-specific ligands for the mediation of these events (1,9); instead they suggested that the mechanism of adhesion is the same for most, if not all, brain and retinal cells.
The assumption that cell adhesion is nevertheless mediated by particular molecules underlies the immunochemical approach adopted in the present studies. It was therefore important to rule out the alternative possibility that adhesion involves weak interactions among a large ensemble of cell surface structures. We have shown that succinyl concanavalin A as well as Fab' fragments of anti-carbohydrate antibodies, which bind to the surfaces of retinal cells in amounts comparable to anti-R10 Fab', do not inhibit adhesion. This suggests that the inhibition caused by anti-R10 Fab' results from a specific inactivation of particular cell surface molecules, and not simply from a blockade of the cell surface as a whole. Further support for this conclusion is presented in the accompanying paper (20).
Although in these studies the activity was found to be associated with a particular molecule, it should be noted that immunoassays potentially can detect any cell surface component that is essential to adhesion. If the formation of cell-cell bonds involved bridging (R-L-R) of membrane-bound receptors (R) by a divalent linker molecule (L), or if it involved binding between two nonidentical receptors (RI-R?), then antibodies to any or all of these components could inhibit adhesion. Consequently, complete neutralization of the inhibition might require the presence of more than one antigen. It is possible, however, that even an antiserum prepared by multiple injections with whole retinal cells might not contain antibodies against all such components. Therefore, although the association of activity with a single surface molecule would be consistent with the conclusion that it is the only cell surface component essential to formation of cell-cell bonds, it would not prove that conclusion.
As with many in vitro phenomena, the direct relevance of cell aggregation to physiological mechanisms has not yet been established. Cells do not collide with each other at 90 rpm in an embryo, nor is it likely that they destroy and regenerate all their cell surface proteins 1 day prior to making contact with each other. Certain observations suggest, however, that the ability of cells to aggregate represents an important parameter during embryonic development: (a) adhesiveness follows an age dependence that differs for retinal and brain cells (6, 27) and this property is shared by a large proportion of the cells obtained from these tissues; (b) maxi-in adhesion, and specific antibodies have now been prepared to this molecule. More direct tests of the physiological role of adhesion in viuo can therefore be carried out by using these reagents to perturb processes of cellular interaction and to localize cell adhesion molecules in tissues.