Selective cell adhesion of neuronal cell lines.

Cloned neural cell lines derived from ethylnitrosoureatreated rat embryos (Schubert, D., Heinemann, S., Carlisle, W., Tarikas, H., Kimes, B., Patrick, J., Steinbach, J. H., Culp, W., and Brand& B. L. (1974) Nature 249, 224-227) adhere preferentially to monolayers of cells obtained by dissociation of neural tissue of either chick or rat embryos. One such cloned line, B103, has been investigated in some detail. B103 cells will bind to cells obtained from any of the major regions of the embryonal nervous system. B103 cells will bind only poorly to chick embryo fibroblasts, Chinese hamster ovary cells, or embryonal liver cells from either the chick or the rat. A plasma membrane-enriched fraction prepared from B103 cells shows the same relative binding characteristics to embryonal neural and non-neural cells as intact B103 cells. Treatment of the membranes with trypsin at low concentrations or treatment of the target cells with low concentrations of glutaraldehyde or formaldehyde also abolishes binding. Binding does not take place at 0”. A very similar binding pattern to that of B103 cells and plasma membranes is shown by B50 and B65 cells and plasma membranes in that both of these cell lines bind preferentially to monolayers prepared from cells from embryonal nervous tissue. The plasma membranes from these cells however show significant differences in binding to other cultured neural cell lines. It is suggested that only a part of the cells adhesive components is retained on the isolated plasma membrane and that there may be several adhesive components on each cell type. The cloned neural cells appear to be a suitable model system for the study of selective cell adhesion.

Cloned neural cell lines derived from ethylnitrosoureatreated rat embryos (Schubert, D., Heinemann, S., Carlisle, W., Tarikas, H., Kimes, B., Patrick, J., Steinbach, J. H., Culp, W., and Brand& B. L. (1974) Nature 249, 224-227) adhere preferentially to monolayers of cells obtained by dissociation of neural tissue of either chick or rat embryos. One such cloned line, B103, has been investigated in some detail. B103 cells will bind to cells obtained from any of the major regions of the embryonal nervous system. B103 cells will bind only poorly to chick embryo fibroblasts, Chinese hamster ovary cells, or embryonal liver cells from either the chick or the rat.
A plasma membrane-enriched fraction prepared from B103 cells shows the same relative binding characteristics to embryonal neural and non-neural cells as intact B103 cells. Treatment of the membranes with trypsin at low concentrations or treatment of the target cells with low concentrations of glutaraldehyde or formaldehyde also abolishes binding. Binding does not take place at 0".
A very similar binding pattern to that of B103 cells and plasma membranes is shown by B50 and B65 cells and plasma membranes in that both of these cell lines bind preferentially to monolayers prepared from cells from embryonal nervous tissue. The plasma membranes from these cells however show significant differences in binding to other cultured neural cell lines. It is suggested that only a part of the cells adhesive components is retained on the isolated plasma membrane and that there may be several adhesive components on each cell type.
The cloned neural cells appear to be a suitable model system for the study of selective cell adhesion.
Tissue specific cell-cell adhesion is presumed to be of importance in early developmental events and in differentiation ( 2). Cell surface proteins which are involved in cell aggregation have been isolated from slime molds (3, 4) and from sponges (5-7). In both cases cell-cell recognition appears to involve the binding of a cell surface protein to a specific carbohydrate moiety on an adjacent cell in agreement with an earlier suggestion by Roseman (8).
Cell-cell recognition in vertebrates has been primarily studied using the reaggregation of dissociated embryonal cells derived most frequently from the nervous system. This work has resulted in the demonstration that cell adhesion shows tissue specificity (9). Proteins which enhance the size of aggregates and which also have regional specificity have been isolated by Hausmann and Moscona (10) as well as by McDonough and Lilien (11). We have previously shown that cells derived from the embryonal nervous system of the chick can specifically bind homologous plasma membranes, and that such binding prevents cell-cell aggregation (12,13). Using as an assay the inhibition of cell aggregation by homologous plasma membranes, we could show that each major region of the brain has binding characteristics which change with time of development (14,15).
Cloned neural cell lines from the central nervous system obtained from ethylnitrosurea-treated rat embryos retain many morphological and chemical characteristics of neural tissues (16,17). These neural cell lines, which can be grown in tissue culture in relatively large quantities as a uniform cell population, would be an ideal starting material for the isolation of specific adhesion components, if they shared with embryonal cells one or more specific adhesive components.
We present evidence in this communication that some of these cell lines adhere preferentially to neural cells' from either chick or rat embryos. Plasma membrane-enriched fractions prepared from these cells retain at least in part these specific adhesive properties. To the best of our knowledge this is the first demonstration of the presence in cloned neural cell lines of "specific" cell-cell adhesion components.
Further, examination of the adhesive specificity of several neural cell lines suggests that many of these cell lines have multiple adhesive determinants on their cell surface. There are many reasons why a given cell type will adhere more rapidly to a monolayer prepared from one cell type as compared to a monolayer prepared from a different cell type. One such reason may be that the pair of cells which gives the fastest adhesion rates has a greater number of complementary sites of a given type than the pair that does not, or the pair of cells that gives the faster rate may have a different type of  (19 Ci/ mmol,AmershamlSearle) added to existing cultures 48 h prior to use. The radioactive leucine was therefore diluted by the nonradioactive leucine present in the medium and in the serum. The medium contained no added unlabeled glucosamine, but very small quantities of nonradioactive glucosamine may have been present in the serum. The cell lines used in this work were B103, B65, and B50, which are all classified as neuronal ceil lines (16,17). In addition we have used C-6, a methylnitrosourea-induced astrocytoma obtained from Dr. S. Pfeiffer from the University of Connecticut (19). All cell lines were harvested as confluent monolayers at densities from 1.0 to 8.0 x 105/cm2. Cells were removed from dishes using 4 ml/T-75 flask of Ca'+, Mg2+-free Hanks solution (Grand Island Biological Co.) buffered with 0.02 M 4-i2-hydroxyethylj-1-piperazineethanesulfonic acid, pH 7.4, at 37" by gentle washing with a Pasteur pipette. This technique produces a greater percentage of viable cells than scraping and is less likely to modify the cell surface than either trypsin or EDTA. Membrane Preparation -The isolated tissue culture cells (approximately 1 to 2 x 10" cells) were cooled to 4", pelleted at 150 x g for 5 min, and washed once by centrifugation at 150 x g for 5 min at 4" with 10 ml of CMF* containing 5 mg/ml of bovine serum albumin. The cells were resuspended in 4 ml of CMF-A plus 0.1 mglml of DNase (Sigma DN-100) at 4", and were disrupted by homogenization with a glass Dounce homogenizer with tight pestle. The homogenization was monitored with a phase microscope and continued until 90% of the cells were disrupted.
For B103 cells, 20 to 30 strokes were sufficient.
The homogenate was sedimented at 39,000 x g for 20 min at 4" in a Sorvall SS-34 rotor. The pellet was taken up in 1.5 ml of 60% (w/v) sucrose in CMF-A, and applied to the bottom of a discontinuous sucrose gradient with additional l-ml layers of 48%, 43%, 40%, and 25% sucrose in CMF-A.
The tubes were centrifuged at 45,000 rpm for 120 min at 4" in an SW50.1 rotor in a Beckman L5-50 ultracentrifuge.
The turbid membrane layer present at the 25 to 40% sucrose interface was removed and after dilution with 8 ml of CMF-A, the membranes were collected by sedimentation at 39,000 x g for 20 min. cells has no effect on their adhesive properties. The suspensions, diluted to 8 ml with CMF-CS at 4", were allowed to settle at unit gravity for 3 min, sedimented clumps were discarded and the supernatant fluid was centrifuged at 150 x g for 5 min. The pellet was washed once with 8 ml of CMF-A, and resuspended in 4 ml of CMF-A at 4". Cell counts were determined with a hemocytometer. The ability of the cells to exclude trypan blue was noted and all cells used in binding assays were at least 85 to 90% trypan bluenegative.
Cell suspensions at this point were essentially free of debris. Cell viability was unchanged when cells were maintained in suspension at 0" for up to 2 h, or up to 1 h at 37". Cell alzd Membrane Binding Assays-Cell to cell binding was measured by the binding of dissociated labeled cells to a confluent cell monolayer (23) on derivatized (1.7 cm21 glass vials as described by Gottlieb and Glaser (24) and Gottlieb et al. (15). Suspension of cloned neural cells obtained as described above could be substituted without alteration in technique for the dissociated embryonal cells in Gottlieb et al. (15). For most cell lines 1.5 to 2.0 x 10" cells were used to form the monolayer; 1.0 x lo" probe cells were used per assay which had between 0.3 to 0.5 dpm/cell. In the data reported below, all assays were carried out with confluent monolayers at 37 and 120 rpm except the data in Fig. 5  The reaction mixture was then gently pipetted with a Pasteur pipette into a 5-ml conical centrifuge tube containing 1 ml of CMF-A at 0". The cells and adhering plasma membranes were sedimented at 150 x g for 5 min. The pellet, rinsed without stirring with 1 ml of CMF-A, was solubilized in 1 ml of 1% Triton X-100. The radioactivity was determined in a scintillation counter using 3a70 (Research Products International) as a counting fluid. In all membrane to cell binding experiments, a control without cells was included and the counts sedimented in this control were subtracted from the experimental values obtained in the presence of the cells, such a control remained constant for all incubation times examined.

Binding
Characteristics of Cloned Neuronal Cell Lines-As an assay for cell-cell adhesion we have used the binding of radioactive probe cells to a target cell monolayer (23,24).
Using this assay we have previously shown the presence of an adhesive gradient in the chick neural retina (15). The data in Fig. 1  The binding of 5 x 10' B103 cells labeled with [3Hlleucine to monolayers prepared from 3 x 10" S-day-old tectal cells or g-day-old telencephalic cells was measured.
B, binding of B103 cells labeled with [3H]leucine to the indicated monolayers was measured using 4 x 10" g-day-old tectal cells, 1 x 10" Chinese hamster ovary cells (CHO), and 1 x 10" chick embryo fibroblasts. C, binding of B103 cells labeled with ["Hlleucine to the indicated monolayers was measured.
In this experiment, liver cells were allowed to aggregate in rotating culture. The aggregates were separated from the nonaggregating single cells and monolayers were prepared from the aggregates after mechanical dispersion of the cells. Monolayers used were prepared with 4 x 10" cells obtained either from S-day-old chick retina or liver.
of binding of B103 cells to liver monolayers is not simply an indication that the liver cells are nonadhesive cells, but is an indication of binding selectivity. It would appear from the data in Fig. 1 that B103 has one or more adhesive components on its surface which allow these cells to bind selectively to monolayers prepared from cells derived from the embryonal nervous system. The data in Fig. 1 were obtained in a heterologous system where B103, a rat cell line, is binding to chick neural cells. In Fig. 2, we show that this cell line also binds selectively to a monolayer prepared from cells obtained from the rat cerebral cortex but not to a monolayer prepared from rat liver cells. Organ adhesive specificity across species barriers has previously been shown by Yaffe and Feldman (251,Roth (91,and Garber and Moscona (26). B103 cells bind to neural cells from both chick and rat, but in a number of experiments we have failed to see any indication that this binding shows either spatial or temporal specificity within the nervous system. The possible significance of these observations will be discussed below.
We have also examined the binding characteristics of several other cell lines, B50 and B65, which were also isolated by Schubert et al. (16) and are classified as neuronal and C-6, a methylnitrosourea-induced rat astrocytoma (19).  -, binding to tectal monolayers; --, binding to liver monolayers. will bind to a monolayer prepared from C-6 cells only slightly better than to a liver monolayer (data not shown   as determined by ultraviolet absorption of an aliquot dissolved in sodium dodecyl sulfate. Crystalline trypsin was added to a final concentration of 0.01 mglmg of membrane protein (0) and 0.1 mg/mg of membrane protein (A). After incubation at 3'7" for 10 min the reaction was stopped with 0.1 mg/ml of soybean trypsin inhibitor.
The control membranes were incubated with a mixture of trypsin and trypsin inhibitor (0). The three membrane preparations were then sedimented at 39,000 x g for 20 min at 4", the pellets were resuspended in the original volume, and tested for binding to tectal cells. The results were identical whether the membranes were tested directly or after removal of soluble material. In this experiment, membranes were tested directly, and fraction bound was based on trichloroacetic acid (TCA )-precipitable counts. 100% values for (0) = 9.7 x lo4 cpm, for (0) = 8.5 x lo4 cpm, and for (A) = 6.8 x lo4 cpm, the control values in absence of cells varied between 5 and 7% in all cases. Similar data have been obtained with membranes labeled with ["Hlleucine. B, B103 cells were grown to confluency in T75 flasks and treated for 15 min at 37" with 5 ml of 1 mg/ml of crystalline trypsin in CMF/T75 flask. The reaction was stopped by the addition of 1 mg/ml of soybean trypsin inhibitor. Control cells were incubated for the same time period with a mixture of trypsin and trypsin inhibitor. The binding of normal B103 membranes to trypsin-treated cells (0) and to normal cells (0) is shown, experiments were done in triplicate, and the range is given by the bars. The same results were obtained with trypsin concentrations between 0.1 to 5 mg/ml and for up to 30 min of incubation at 37" with 1 mg/ml of trypsin. 100% = 8.3 x lo4 dpm and control in the absence of cells was 1 x lo4 dpm. experiments identical with those illustrated in Fig. 14, and similar experiments carried out with membranes prepared from C-6 cells, which adhere preferentially to tectal cells and B65 cells and very poorly to B103, C-6 cells, or liver. The membranes prepared from each of these cell lines have distinct binding properties which are significantly different from the binding properties shown by intact cells. The data in Fig. 15 illustrate the results of membrane adhesive experiments with all possible combinations of these four cell lines.

DISCUSSION
The cloned neural lines exhibit many neuronal characteristics (16) and these cells have also been shown by Stallcup and Cohn (17) to have neuronal-specific cell antigens on their surface.
The binding properties of B103 cells are analogous to the binding properties of dissociated embryonal neural cells in that they show selectivity for neural cells. B103 cells bind preferentially to neural cells from heterologous as well as homologous species but cannot distinguish among different neural cell populations with respect to region or embryonic age.
The selective binding observed in the cell to cell adhesion Leucine-labeled B103 cells were treated for 10 min at 37" with 2 mg/ml of trypsin in 5 ml of CMF/T75 flask. Trypsin inhibitor was then added and membranes were prepared by standard method. As control membranes were prepared from duplicate flasks incubated with trypsin-trypsin inhibitor. @, binding of control membranes to cells; 0, binding of membranes prepared from trypsintreated cells. A shows the binding to tectal cells and B to B103 cells. Different membrane preparations are shown in A and 23. For control membranes in A, 100% = 8.4 x 10' dpm and the control in absence of cells is 1.2 x 10" dpm for membranes from trypsinized cells. 100% = 3.5 x lo4 dpm and the control in the absence of cells = 1.4 x 103. In B for normal membranes 100% = 6.3 x IO4 dpm and control in absence of cells 6.6 x lo3 dpm. For the membranes from trypsintreated cells, 100% = 2.3 x lo4 dpm and the control in the absence of cells = 1.8 x 10" dpm. assays is usually considered an indication of specific cell binding. We have used the term preferential binding in this paper to emphasize the fact that such measurements do not always allow a distinction to be made between quantitative and qualitative differences in cell surface adhesive components.
An initial step in the characterization of the components involved in cell adhesion is to demonstrate the presence of some of the adhesive components in plasma membranes. The experiments reported here serve to define adhesive characteristics of a plasma membrane fraction, and also to define the minimum number of complementary cell surface molecules which are responsible for cell to cell and membrane to cell adhesion in these systems.
The identification of the binding fraction used in membrane to cell adhesion assays as cell surface membranes is based on the enrichment of cell surface markers in this fraction and in particular on the fact that 1251-labeled plasma membranes can be successfully used in the membrane binding assay. Since the lZ51 labeling was carried out on intact cells by the lactoperoxidase method, the lz51 would be expected to be preferentially localized on protein originally exposed on the cell surface. The fact that membranes prepared from trypsin-treated cells show a substantial decrease in binding ability is also in agreement with the surface location of these membranes.
The preferential binding of plasma membranes to cells that we have reported in this paper is consistent with the notion that initial cell adhesion is a specific event as previously suggested by studies in this and other laboratories (9,15,(27)(28)(29)(30) Cell to cell adhesion and cell to membrane adhesion is  Table I. At least one additional adhesive component must be present in the cells to account for the binding of intact cells to liver, or the binding of C-6 cells to C-6 cells. This component also is not detected during the membrane binding assay. It should be clear that if an adhesive component is not expressed in the plasma membrane under our assay conditions, this could be a kinetic phenomenon and the relevant molecules may actually be present in the membrane preparation.
We had previously suggested that one of the complementary adhesive components was not functional in membranes prepared from embryonal cells in order to account for the inhibition by these membranes of cell aggregation (12,14).
To explain our observation of asymmetry in membrane to cell binding with various cell lines, we have postulated that B103 cells contain one adhesive component "b" but not its complementary structure "B." These observations strongly suggest that specific binding in this system results from the recognition of two dissimilar molecules and not from dimerization of a single molecule as has been suggested in other systems (32). The presence of multiple adhesive components on the surface of the neuronal cell lines, which may also be present on the surface of embryonal cells, is of interest because it suggests the possibility that a large variety of adhesive specificities could be generated with only a few adhesive components. For example, 16 different cell surface specificities can arise if a cell can have on its surface one or more of the components designated as A, a, B, b; in general 2" different cell surfaces can be generated in this way from "n" cell surface components. It should be clear that membranes prepared from B65 cells could contain small quantities of the ligand "a," all that our assays indicate is that this ligand is not present in large enough quantity to allow B65 membranes to bind to B103 cells. membranes to neuronal cell lines since in the absorption experiments these sets of adhesive specificities would only be separated in the unlikely event that they segregated into different populations of membrane vesicles.
The apparent loss of surface adhesive components from the plasma membrane fractions indicates that the binding of membranes to cells can only be used to prove the presence of certain adhesive components on the surface of the cell from which the membranes were prepared, but cannot be used to prove the absence of an adhesive component in the surface of the original cell.
The observations presented in this paper should be useful as a guide to the isolation and characterization of cell surface adhesive components. Our demonstration of the selective adhesion of neural cell lines to the cells derived from the nervous tissue and our evidence for multiple adhesive components on a single cell surface have potential implications for the way in which tumor cells grow in the animal as well as for developmental organization.
Since the original submission of this manuscript, two papers have appeared which are relevant to our work. Stallcup has independently reached the conclusion that adhesion between different neural cell lines reflects the presence of multiple pairs of adhesive components upon the cell surface of these cells. This conclusion is based on the differential effects of trypsin, temperature, and antibodies on the binding of B50 cells to various cell monolayers (33). Obrink et al. (34) have shown that plasma membranes prepared from chick and rat liver cells retain the adhesive specificities of the parental cells, thus demonstrating that membranes will also be useful for the study of cell adhesive specificity in non-neural systems. Acknowledgments -We are extremely grateful to Dr. David Schubert of the Salk Institute for making his cell lines available to us, and to Mr. L. Andrews for examining the membrane preparation in the electron microscope. There are major differences between the membrane binding experiments and the cell to cell binding experiments. In cell to cell binding experiments, all cell lines show a significant rate of binding to liver, and the rate of binding to monolayers prepared from the cultured neuronal cell lines is always less than the rate of binding to tectal monolayers. In membrane to cell binding experiments many of the cell lines bind as well or better to suspensions of cultured neuronal cells as to optic tectum, and do not bind well to liver. This discrepancy is most notable with B65 cells, but is also true with B50 cells and in part with B103 cells. These differences serve to emphasize two points, the first is that the binding of cells to liver and to neuronal cells must be due to different adhesive components. The second is that the rate-limiting step in the cell to cell adhesion assay and in the membrane to cell adhesion assay may be different. In addition, it is possible as we have emphasized previously (15), that under the conditions used to measure cell to cell adhesion many of the adhesive sites on the cells in the monolayer have already been occupied by interaction with neighboring cells. This would not be true in a membrane to cell adhesion assay.
In adsorption experiments we have not been able to separate the plasma membrane fraction from B103 cells that binds to tectal cells from that which binds to B103 cells. However, the preferential binding shown by the neuronal cell lines and membranes to cells obtained from the embryonal nervous system may be determined by different adhesive molecules than those involved in the binding of the same cell and