Direct evidence that neural cell adhesion molecule (NCAM) polysialylation increases intermembrane repulsion and abrogates adhesion.

Molecular force measurements quantified the impact of polysialylation on the adhesive properties both of membrane-bound neural cell adhesion molecule (NCAM) and of other proteins on the same membrane. These results show quantitatively that NCAM polysialylation increases the range and magnitude of intermembrane repulsion. The repulsion is sufficient to overwhelm both homophilic NCAM and cadherin attraction at physiological ionic strength, and it abrogates the protein-mediated intermembrane adhesion. The steric repulsion is ionic strength dependent and decreases substantially at high monovalent salt concentrations with a concomitant increase in the intermembrane attraction. The magnitude of the repulsion also depends on the amount of polysialic acid (PSA) on the membranes, and the PSA-dependent attenuation of cadherin adhesion increases with increasing PSA-NCAM:cadherin ratios. These findings agree qualitatively with independent reports based on cell adhesion studies and reveal the likely molecular mechanism by which NCAM polysialylation regulates cell adhesion and intermembrane space.

Polysialic acid (PSA) 1 is a long, linear ␣2,8-linked carbohydrate composed of N-acetylneuraminic acid (Neu5Ac) residues (1). This carbohydrate is added post-translationally to the neural cell adhesion molecule (NCAM), which is responsible for a variety of functions, including axon pathfinding, synaptogenesis, and tissue formation in the central nervous system (2). The expression of the polysialylated form of NCAM (PSA-NCAM) peaks early in development and decreases with age. In some exceptions, such as the hippocampus, cells continue to express PSA-NCAM throughout the life of the organism. These regions of PSA expression are also associated with neural plasticity and the remodeling of neural connections (1,2). Aberrant expression of PSA-NCAM is associated with tumor malignancy and metastasis, and the expression of PSA-NCAM has been detected in small cell carcinoma, neuroblastomas, and Wilm's tumor (3).
Polysialic acid is thought to facilitate cell migration and plasticity by inhibiting cell adhesion to other cells and to the extracellular matrix, as a result of the large excluded volume of the polymer (4,5). Electron microscopy images showed that PSA expression increased intercellular spacing by 10 -15 nm (4). The latter could be because of the inactivation of adhesion proteins or to the increased inter-membrane repulsion resulting from the confinement of the carbohydrate chains. Light scattering studies demonstrated that NCAM polysialylation doubles the hydrodynamic radius of NCAM. However, the latter results were based on calculations, using light scattering data and the assumption that the rod-like proteins were spherical. While this indicates the approximate size of the protein, the hydrodynamic radius does not quantify the effect of the carbohydrates on NCAM-mediated adhesion. In one proposed mechanism, for example, the increased repulsive pressure between the membranes is hypothesized to push the cells apart (6). Such a shift in the force balance between cells from attractive to repulsive requires the increased intermembrane repulsion to be at least as large as the protein attraction at the membrane distance at which the proteins bind. For example, if NCAM bridges two membranes at a separation of 40 nm with an adhesion energy of ϳ1000 k B T/m 2 , where k B is the Boltzmann constant and T is the temperature, then NCAM polysialylation would have to increase the repulsion at 40 nm by at least this amount, to disrupt the adhesive junction. Testing this, however, requires determining both the magnitudes of the intermembrane forces and their range.
The impact of ionic strength on the adhesion between cells expressing PSA-NCAM further supports the view that PSA acts by increasing the repulsion between cells. The hydrodynamic volume of polyelectrolytes decreases with increasing monovalent salt concentrations (7), and this would in turn reduce the repulsion between two membranes with surfaceanchored chains. Consistent with this, an increase in the monovalent salt concentration from 0.15 to 0.5 M NaCl restored the adhesion between cells expressing PSA-NCAM (5).
Investigations of cell adhesion also suggested that the effects of PSA can be generalized to a diverse set of adhesion proteins, including NCAM, cadherin, L1, and integrins (8). Furthermore, this general abrogation of adhesion did not require NCAM domains beyond those minimally required for polysialylation. Recent structural studies utilizing both x-ray and neutron specular reflectivity show that the carbohydrate extends sig-* This work was supported by National Institutes of Health Grant 1RO1 GM33986. The Laboratory for Fluorescence Dynamics is a National Institute of Health National Resource and is also partially supported by the University of Illinois. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
In this study, the surface force apparatus (SFA), and fluorescence correlation spectroscopy (FCS) were used to investigate the molecular mechanism by which polysialic acid abrogates NCAM and C-cadherin-mediated intermembrane adhesion. The SFA quantifies the distance dependence of the force between extended surfaces such as membranes (10,11). These measurements are particularly relevant to gaining a mechanistic understanding of the biological activity of PSA. The configuration of grafted polyelectrolyte chains such as PSA, as well as the magnitude of the repulsive pressure they exert between membranes is a function of the chain density (7). In this case, single molecule measurements would not yield useful mechanistic information. The results reported here show that the increase in the intermembrane repulsion following NCAM polysialylation is sufficient to overwhelm both homophilic NCAM and cadherin attraction at physiological ionic strength. Some adhesion recovers at high monovalent salt concentrations, concomitant with a significant reduction in the PSA-dependent intermembrane repulsion. FCS provided further, independent evidence that the hydrodynamic radius of PSA exceeds that of the unmodified NCAM. Additional measurements with mixed monolayers of NCAM and C-cadherin show that the inhibition of trans adhesive interactions can be generalized to other cell adhesion molecules.
Protein Expression and Purification-The cDNA of chicken NCAM120, encoding the seven extracellular domains of NCAM, was truncated at the onset of the transmembrane segment and fused to a sequence encoding a hexahistidine tag and a 3-amino acid linker, as described (13). This sequence was cloned into a PEE14 expression vector (gift from B. Gumbiner), and the soluble protein was expressed in Chinese hamster ovary cells under methotrexate selection. The resulting protein consisted of the seven extracellular domains fused to a C-terminal polyhistidine tail (13). The polyhistidine tail allows both the purification of the protein and its immobilization on fluid NTA-TRIG-DLGE lipid monolayers (13). PSA-NCAM expression was achieved by co-transfecting Chinese hamster ovary cells with a plasmid encoding the NCAM ectodomain and a second plasmid encoding the sialyltransferase enzyme PST (8,13). Both NCAM and PSA-NCAM were purified from the cell culture supernatants in a three-step process (13). Briefly, the supernatant was first loaded onto a Ni-NTA-Sepharose column (Qiagen), and the protein was eluted with a buffer consisting of 20 mM HEPES, 100 mM NaCl, and 250 mM imidazole (pH 7.5). Following the concentration of the protein eluant with a Centricon YM-30 concentrator (Millipore, Bedford, MA), the protein was further purified on a HiTrap Q anion exchange column (Amersham Biosciences). The PSA-NCAM fractions were collected and concentrated. Finally, the protein was purified to homogeneity by gel filtration (Sephacryl S2000, Amersham Biosciences). The PSA-NCAM fractions were concentrated in a buffer containing 20 mM HEPES and 150 mM NaCl at pH 7.5. Optical densities at 280 nm and an extinction coefficient of 99,030 cm Ϫ1 M Ϫ1 were used to determine the NCAM concentration. The expression and purification of histidine-tagged NCAM and C-cadherin have been described previously (13,15). The extinction coefficient used to determine the C-cadherin concentration at 280 nm was 69,040 cm Ϫ1 M Ϫ1 (16).
Protein Monolayers-Force measurements were carried out with proteins immobilized on opposing, supported lipid bilayers (Fig. 1a). The bilayers were prepared by Langmuir-Blodgett deposition (13,(15)(16)(17). Lipid monolayers were prepared by spreading a chloroform solution of the lipid mixture at the air-water interface of a Langmuir trough (Nima, Coventry, England). The lipid monolayers were compressed to a FIG. 1. Protein configurations in direct force measurements. a, the immobilized PSA-NCAM configuration on lipid bilayers in the direct force measurements. The distance, D, refers to the distance between bilayers, and is determined by subtracting twice the thickness of the NTA-TRIG-DLGE layer from the measured distance between the 1,2-di-palmitoyl-sn-glycero-3-phosphoethanolamine layers on the opposite surfaces: D ϭ T 1 -2 ϫ T NTA-TRIG-DLGE . b, the proposed binding configurations of opposing NCAM extracellular domains at 31 and 39 nm. c, the proposed binding configurations between proteins on opposite, mixed monolayers of cadherin and NCAM, at membrane separations of 25, 31, and 40 nm. desired surface pressure at room temperature, and deposited onto the surface of a mica sheet (15,16). The first lipid monolayer was gel phase 1,2-di-palmitoyl-sn-glycero-3-phosphoethanolamine, deposited onto the mica from the air-water interface at a density of 43 Å 2 /lipid. The second, outer fluid phase lipid layer, containing 75 mol % of NTA-TRIG-DLGE and 25 mol % of 1,2-ditridecanoyl-sn-glycero-3-phosphocholine was deposited at 36 mN/m, which corresponds to ϳ60 Å 2 /lipid. The supported bilayers were then incubated for 1 h at room temperature in a buffer containing 20 mM HEPES, 150 mM NaNO 3 , 15 M NiSO 4 (pH 7.7), and micromolar concentrations of the polyhistidine-tagged protein. The protein was adsorbed to the NTA lipid head group via the histidine tag. After 1 h, the lipid bilayers were washed with buffer to remove nonspecifically adsorbed protein, and the supported bilayers were then mounted in the surface force apparatus. For mixed protein monolayers, the supported bilayers were incubated either with C-cadherin and PSA-NCAM or with NCAM and C-cadherin at different molar ratios, to vary the PSA-NCAM:cadherin and NCAM:cadherin surface densities.
Force Measurements-The forces were measured between two protein monolayers as a function of the intermembrane distance, D, with a Mark II SFA. The SFA quantifies the force between the surfaces of two crossed, hemi-cylindrical discs of geometric average radius (10,11).
From deflections in the spring supporting the lower disk, the force between the curved surfaces F c , normalized by the radius, is determined as a function of the intersurface distance with a resolution of Ϯ1 nN/m ϭ mJ/m 2 (10,18). Using multiple beam interferometry, SFA measurements quantify the absolute distances between the surfaces with a resolution of Ϯ0.1 nm (18). In this work, the distance, D, refers to the distance between the dehydrated surfaces of the bilayers (Fig. 1a) (13,(15)(16)(17). The radius R is also quantified directly from the interference fringes (18).
An important aspect of this approach is that for geometries where the radius is significantly greater than the range of interaction, R Ͼ Ͼ D, the total force between the macroscopic curved surfaces F c is directly related to the interaction energy between two equivalent flat surfaces E f by the Derjaguin approximation, F c /R ϭ 2E f (19). Importantly, the curved geometry only introduces a 2R factor, which scales the magnitude of the force. The normalized force F c /2R is therefore a direct measure of the intersurface potential. This well established result is described in several standard textbooks (20). The adhesion energy per area is related to the force to detach the surfaces, or the pull-off force, F po , by the Johnson-Kendall-Roberts theory, according to E f ϭ (F po )/ (1.5R) (21).
Quantification of Protein Surface Densities-To determine the surface density of proteins immobilized on the NTA-TRIG-DLGE supported bilayers, we measured the amount of bound 125 I-labeled protein, according to procedures described previously (22,23). 125 I measurements were conducted with pure NCAM and PSA-NCAM monolayers, as well as with mixed NCAM:cadherin and PSA-NCAM:cadherin monolayers at ϳ1:1 molar ratios. For the measurements with the mixed protein monolayers, one of the protein components was labeled with 125 I, and both proteins were incubated with the supported bilayer. Thus, measurements with mixed monolayers were conducted with labeled NCAM and then with labeled cadherin.
To quantify the surface densities of immobilized NCAM, PSA-NCAM and cadherin prepared from solutions of two proteins at molar ratios other than 1:1, we quantified the fluorescence intensities on the supported membranes. By fluorescently labeling one of the proteins and measuring the fluorescence intensity from monolayers prepared from the different protein mixtures, we determined the relative change in the surface density from the measured changes in the fluorescence. The surface densities of proteins adsorbed from 3:1 NCAM:cadherin and 1:3 PSA-NCAM:cadherin were then scaled by the change in fluorescent intensity relative to protein monolayers adsorbed from equimolar protein solutions. For example, a 3-fold increase in the fluorescence intensity from labeled NCAM would signal a 3-fold greater NCAM density. The rhodamine green (RG) dye used does not self-quench, and the measured intensities were below the saturation limit of the detector. Therefore, the intensity is assumed to scale linearly with the protein coverage.
Because the fluorescence measurements quantify relative differences in fluorescence intensities, we calibrated the fluorescence intensity against the amount of 125 I-labeled protein immobilized under identical conditions. For example, if the surface density of NCAM adsorbed from a 1:1 NCAM:cadherin solution is 2.0 Ϯ 0.6 ϫ 10 4 /m 2 , then a 2-fold increase in the fluorescence intensity would correspond to a surface density of 4 ϫ 10 4 /m 2 .
The proteins were fluorescently tagged with rhodamine green carboxylic acid, succinimidyl ester (Molecular Probes, Eugene, OR), according to the manufacturer's instructions. The molar ratio of dye to protein was kept at a stoichiometric ratio of 3:1 dye:protein. This yielded a labeling ratio of 1.1 for NCAM, 0.9 for PSA-NCAM, and 1.8 for cadherin.
Fluorescence intensity images were taken using a ϫ100 UPLANFI oil objective (Olympus, Tokyo, Japan), on a BX60 optical microscope from Olympus (Hamburg, Germany), which was interfaced to a digital camera (Diagnostics Instruments, Sterling, MI). For these measurements, a supported bilayer displaying a mixed protein monolayer, prepared as described above, was placed in a fluid cell and mounted on the microscope stage. The fluorescence intensity was determined by averaging the intensities from a minimum of 10 images taken from different regions of the supported bilayer.  where G() is the autocorrelation amplitude and ␦F(t) is the fluctuation in the number of molecules in the excitation volume at time t (25,26). If applied to the translational diffusion of point-like particles, Equation 2 can be expressed as the following.
Here, w and z are the radii of the beam in the xy and z planes, respectively, N is the number of proteins, and d is the diffusion coefficient of the protein. We can thus accurately quantify differences in the hydrodynamic radii of NCAM and PSA-NCAM from their relative diffusion coefficients.
NCAM and PSA-NCAM were fluorescently labeled with RG as described above. The protein was separated from excess dye by NTA affinity chromatography as described above, and was stored at 4°C until use. The instrumentation for the measurement and the software for the analysis of FCS data are at the Laboratory for Fluorescence Dynamics at the University of Illinois at Urbana-Champaign (24,26). To excite the sample, a mode-locked Ti:sapphire laser (Mira 900, Coherent, Palo Alto, CA) pumped by an intracavity doubled Nd:YVO4 vanadate laser (Verdi, Coherent Inc., Santa Clara, CA) was used as a photon source. Photon counts were collected using an Avalanche photodiode detector (model SPCM-AQ-151, EG&G) and the output was directed toward a data acquisition card. A Zeiss Axiovert 135 TV microscope (Thornwood, NY) was used with an oil immersion objective (NA ϭ 1.4). The PSA-NCAM and NCAM concentrations used in the FCS experiments were ϳ10 Ϫ8 M. Data were analyzed to determine the diffusion coefficient by fitting the experimental results to an autocorrelation function that assumes a Gaussian-Lorentzian intensity profile (25,26). The Gaussian-Lorentzian function depends on the experimental setup and therefore must be calibrated. A solution of 1 ϫ 10 Ϫ8 M fluorescein in 50 mM Tris was used both for calibration and as a standard for determining the size of the beam waist.  Table I. In studies performed with 125 Ilabeled His 6 PSA-NCAM, the measured surface density was somewhat higher at 4.2 Ϯ 0.7 ϫ 10 4 PSA-NCAM/m 2 . In mixed monolayers adsorbed from a 1:1 solution of NCAM:cadherin, the NCAM surface density was 2 Ϯ 0.6 ϫ 10 4 NCAM/m 2 , whereas the cadherin surface density was 2.4 Ϯ 0.7 ϫ 10 4 cadherin/m 2 (Table I). 125 I measurements of monolayers adsorbed from a 1:1 PSA-NCAM:cadherin mixture yielded surface densities of 1 Ϯ 0.5 ϫ 10 4 PSA-NCAM/m 2 and 2.5 Ϯ 0.6 ϫ 10 4 cadherin/m 2 . Thus, in a solution of equal protein concentrations, the ratios of the adsorbed proteins differ somewhat from those in solution.

Fluorescence Correlation Spectroscopy-To
Fluorescence intensities were used to quantify the protein coverage at ratios other than 1:1. A comparison of the average intensities from (RG)NCAM on monolayers adsorbed from (RG)NCAM:cadherin solutions shows a 68 Ϯ 3% increase in the fluorescence intensity when the solution composition was changed from 1:1 to 3:1 NCAM:cadherin. Scaling the intensity by the NCAM surface density determined from radiolabeling measurements (Table I), we calculate an NCAM density in the 3:1 monolayer of 3.3 Ϯ 0.6 NCAM/m 2 . The measurement was repeated with labeled cadherin. In this case the average fluorescence intensity decreased by 59 Ϯ 3% when the relative amount of cadherin in solution decreased from 50 to 25 mol %. Using the known cadherin density on monolayers prepared from the 1:1 mixture, the calculated cadherin surface density is 1.1 Ϯ 0.5 cadherin/m 2 (Table I).
Fluorescence measurements with 1:1 and 1:3 (RG)PSA-NCAM:cadherin ratios show that the intensity decreased 34% when (RG)PSA-NCAM was adsorbed at the lower concentration (Table I). The calculated surface density on the resulting monolayer was 0.7 Ϯ 0.5 PSA-NCAM/m 2 . The fluorescence intensity changed little when comparing the intensities of labeled cadherin adsorbed from solutions with 1:1 and 1:3 cadherin:PSA-NCAM ratios. The surface density increased by only 2%. The results are summarized in Table I. PSA Modification Inhibits trans NCAM Adhesion-Previous direct force measurements of the homophilic NCAM adhesion revealed that NCAM binds homophilically through two separate, spatially distinct binding configurations (13). As seen in Fig. 3a, at bilayer distances greater than 48 nm, there is no force between the NCAM monolayers. However, the onset of the repulsive force (F/R Ͼ 0) is approximately D Ͻ48 nm. As the distance, D, decreases further, the steric repulsion between proteins increases, until a steep repulsive wall is reached at D Ͻ22 nm.
Upon separating the bilayers, the force curve drops below the advancing force curve, because of the attractive force between the proteins. At the force minimum, which corresponds to the maximum gradient in the intersurface potential, adhesive failure occurs, and the two surfaces jump out of contact. The minimum is at D ϭ 31 Ϯ 1 nm, which corresponds to full, antiparallel overlap between the Ig1-5 segments of opposing NCAM ectodomains (Fig. 1b) (13). The adhesion is Ϫ0.30 Ϯ 0.05 mN/m (Table II). Using the JKR theory (20, 21), we cal-culate an adhesion energy per area of 1.6 ϫ 10 4 k B T/m 2 at this NCAM density. A second bound state is also detected at D ϭ 39 Ϯ 1 nm, when the surfaces are brought to distances that allowed only partial protein overlap. This bound state is mediated by the outer two Ig domains of NCAM (Fig. 1b) (13). The adhesive strength at 39 nm is Ϫ0.15 Ϯ 0.07 mN/m (Table II), which corresponds to 7.5 ϫ 10 3 k B T/m 2 .
In force measurements between identical PSA-NCAM layers at a surface density of 4.2 Ϯ 0.7 ϫ 10 4 PSA-NCAM/m 2 , the addition of the PSA substantially increases the magnitude of the intersurface repulsion. Fig. 3b compares the distance dependence of the repulsive force between approaching PSA-NCAM monolayers with that measured between bare NCAM. With PSA-NCAM, the initial onset of the intersurface repulsion is at D Ͻ 48 Ϯ 2 nm. The magnitude of the repulsion increases with decreasing separation, and then increases steeply at D Ͻ 28 Ϯ 1 nm. The repulsive force is larger than between bare NCAM at all distances, and the steep increase in the repulsion is shifted out to 28 nm (Fig. 3b). For example, at 30 nm, the repulsive force between PSA-NCAM monolayers is 3 mN/m compared with 1 mN/m between NCAM monolayers at the same distance. Using the Derjaguin approximation (19), this difference corresponds to an increase in the intersurface repulsive energy per area by ⌬E f ϭ ⌬F c /2R ϭ 0.32 mJ/m 2 or 8 ϫ 10 4 k B T/m 2 . This is five times the magnitude of the NCAM adhesion energy at this same distance.
To quantify the adhesion between the PSA-NCAM monolayers, the discs were brought to membrane separations D Ͻ 31 nm and 31 nm Ͻ D Ͻ 39 nm, before separating the proteins. At these distances, we would detect any residual NCAM binding (cf. Fig. 3a). There was no measured adhesion upon separation (Fig. 3c), although there was some hysteresis in the receding curve near 31 nm (Table II).
It is important to point out that, to abolish trans NCAM adhesion merely by opposing the NCAM attraction with a nonspecific repulsive force, the repulsion would have to increase by at least 0.3 mN/m at 31 nm (cf. Fig. 3a and Table II), or ϳ0.6 mN/m if we take into account the higher PSA-NCAM surface density (Table I). With PSA-NCAM, the repulsion at 30 nm increased from ϳ1 mN/m between bare NCAM to ϳ3 mN/m. This difference of 2 mN/m, which is because of the osmotic repulsion between the carbohydrate chains, is more than sufficient to overwhelm the NCAM-mediated attraction and abrogate adhesion, without invoking the disruption of inplane interactions.
The pretreatment of PSA-NCAM with endo-N before immobilizing the protein established that the increased repulsion and consequent inhibition of adhesion is because of the carbohydrate. Endo-N removes polysialic acid by cleaving PSA randomly along the carbohydrate chain (27). Because endo-N requires a minimum of 5-8 sialic acid residues for cleavage, even after complete digestion, a small glycan core remains (27,28). Fig. 3d shows the force measurements conducted with PSA-NCAM after endo-N treatment. These force curves are similar to those measured with unmodified NCAM (cf. Fig. 3a). Specifically, the magnitude of the repulsion was significantly smaller, and the stiff steric repulsion at 28 nm shifted inward to ϳ23 nm. Perhaps the most significant feature is the recovery of NCAM adhesion at both 31 Ϯ 1 and 39 Ϯ 1 nm. We note that the magnitude of adhesion at 31 nm is slightly lower than between non-sialylated NCAM ectodomains (13). The adhesion at 31 nm was Ϫ0.3 Ϯ 0.05 and Ϫ0.15 Ϯ 0.07 mN/m, for bare and endo-N-treated NCAM, respectively (Table II). There was not a statistically significant difference between the bound states at 39 nm (Table II). SDS-PAGE showed that the molecular weight of the treated PSA-NCAM was the same as the unmodified protein, within error, indicating essentially complete PSA removal. The reduced adhesion at 31 nm is therefore attributed to the inability of endo-N remove the glycan core (27,28). PSA is anionic, and theory predicts that the dimensions of polyelectrolytes in good solvent decrease when the charges are screened by electrolyte (7). The dependence of the extension of surface-anchored polymers, and hence the range of the osmotic repulsion between grafted polyelectrolytes, scales roughly with C s Ϫ1/3 , where C s is the concentration of monovalent electrolyte (7). Increasing the ionic strength would therefore reduce the hydrodynamic radius of PSA and correspondingly reduce the range and magnitude of the intermembrane repulsion. If the repulsive and attractive forces between the NCAM ectodomains are additive, then the intermembrane adhesion will increase with the decreasing repulsion.
Measuring the forces between PSA-NCAM monolayers in 1 M NaNO 3 tested the relationship between the PSA-dependent repulsion, the abrogation of adhesion, and the ionic strength. Fig. 4a compares the steric repulsion between PSA-NCAM bathed in 1 M electrolyte relative to the repulsive force measured with 150 mM NaNO 3 . In 1 M NaNO 3 both the range and magnitude of the repulsion are lower, and the steep repulsive wall shifted inward from 28 Ϯ 1 to 25 Ϯ 1 nm. Additionally, during the separation of the proteins (Fig. 4b), the NCAM-dependent adhesion reappeared at 31 Ϯ 1 nm, with a magnitude of Ϫ0.29 Ϯ 0.07 mN/m (Table II). This reduction in the repulsion, coupled with the recovery in the adhesion at high ionic strength, further supports the view that PSA functions by increasing the nonspecific intermembrane repulsion (1,6,8).
Unlike the studies with endo-N-treated PSA-NCAM, the second adhesive bond at 39 nm was not detected in 1 M NaNO 3 (Table II). Thus, although the excluded volume of the carbohydrate is much smaller in 1 M NaNO 3 , the polymer still contributes a large enough repulsion (ϳ0.2 mN/m) at 39 nm to overwhelm the second weaker bond.
Additional studies were also carried out with 20 mM NaNO 3 in the bathing medium (Fig. 4c). Under these conditions, the range of the repulsion shifted outward to D ϳ 60 nm, and the position of the steep repulsion increased to ϳ30 nm. The overall magnitude of the repulsive force is also greater at the lower ionic strength. Additionally, in contrast to measurements with 150 mM NaNO 3 (Fig. 3c), there was no hysteresis in the receding curves (Table II), that is, the curves measured during approach and separation overlap.
Adhesion Mediated by Mixed Monolayers of NCAM and Cadherin-The organization of intercellular space is often governed by more than one type of adhesion molecule, and NCAM polysialylation influences the adhesive function of other proteins on cell membranes (8). Force measurements between membranes displaying mixed monolayers of PSA-NCAM and C-cadherin were carried out to investigate the molecular basis of the impact of PSA on the function of cadherin. Experiments were also conducted with mixed monolayers of NCAM and cadherin, for comparison. Fig. 5a shows the force-distance profiles between mixed pro- tein monolayers containing NCAM and cadherin at a ratio of 1:1.2 (Table I). Upon approach, the repulsion increases smoothly at D Ͻ 40 nm, and then increases more steeply at D Ͻ 27 nm. This profile differs from advancing profiles between NCAM (Fig. 3a) (13) or cadherin (15) alone.
Upon separation from distances D Ͻ 26 nm, the surfaces adhered at D ϭ 25 Ϯ 1 nm (Fig. 5a). This distance matches one of the three bound states measured between C-cadherin ( Fig.  1c) (15, 16). The magnitude of the adhesion was Ϫ0.7 Ϯ 0.05 mN/m (Table III). Because NCAM binds at 31 and 39 nm, the adhesion at 25 nm is attributed solely to homophilic cadherin binding.
To test for adhesion at other intermembrane spacings, retreating force profiles were measured from distances D Ͼ 25 nm, where the ectodomains of both proteins only partially overlap. Using this method, which is described extensively in earlier publications (13,15,16), a second, weaker adhesive minimum was detected at 30 Ϯ 1 nm, with a strength of Ϫ0.35 Ϯ 0.05 mN/m (Table III). At this intermembrane distance, both NCAM and cadherin adhere (Fig. 1c) (13, 15). Retreating force profiles measured at distances D Ͼ 31 nm detected a third adhesive interaction at D ϭ 41 Ϯ 1 nm (Table  III). This distance is, within error, the same as the positions of the outer bound states of both cadherin (D ϭ 40 Ϯ 1 nm) and NCAM ectodomains (D ϭ 39 Ϯ 1 nm) (Fig. 1c). The adhesion at 41 nm is therefore attributed to the combined contributions of both proteins. The magnitude of the adhesion was Ϫ0.21 Ϯ 0.06 mN/m. No other adhesive interactions were detected.
The dependence of the adhesion on the protein densities was determined in a second set of force measurements with an NCAM:cadherin ratio of 3:1 (Table I). As in Fig. 5a, the advancing profile in this case differed from those of either NCAM or cadherin (not shown). The retreating force profile similarly detected three bound states at the same three distances. However, in this case, the adhesion at 25 nm decreased from Ϫ0.7 Ϯ 0.05 to Ϫ0.35 Ϯ 0.07 mN/m, because of the reduced cadherin density (Table III). The adhesion at 30 Ϯ 1 and 41 Ϯ 1 nm changed only slightly (Table III), because the reduction in cadherin-mediated adhesion is offset by the increased NCAM density, and hence by the larger number of NCAM bonds.
PSA-NCAM Inhibits Cadherin Adhesion-Direct force measurements between monolayers of PSA-NCAM and C-cadherin at a 1:2.5 molar ratio are shown in Fig. 5b. The bathing medium contained 150 mM NaNO 3 . The range and magnitude of the repulsion are much greater than between the NCAM:cadherin monolayers (Fig. 5a). The onset of repulsion occurs at D Ͻ 50 Ϯ 2 nm, and the steep repulsive force is at D ϳ 27 Ϯ 1 nm. There is also no adhesion at any distance (Table III). The magnitude of the force at 25, 30, and 40 nm is 3.69 Ϯ 0.06, 1.22 Ϯ 0.08, and 0.43 Ϯ 0.08 mN/m, respectively. At all three distances, the repulsive force exceeds the combined attractive forces of both cadherin and NCAM (Table II), and is sufficient to abrogate the protein-mediated adhesion.
If the loss of adhesion is because of the additive effect of steric repulsion and protein attraction, then lower PSA-NCAM densities should correspondingly reduce the magnitude of the repulsion and increase the attraction. Force measurements with monolayers containing a 1:3.7 PSA-NCAM:cadherin ratio tested this. Decreasing the PSA-NCAM surface density decreased the repulsive force, as expected, but the magnitude of the remaining repulsion still exceeded the protein attractive forces, and there was no adhesion (not shown). There was hysteresis between the advancing and receding force curves, and we attribute this to some protein binding. However, between monolayers adsorbed from a 1:8 PSA-NCAM:cadherin solution, the range and magnitude of the repulsion were lower, and the protein monolayers adhered at D ϭ 31 Ϯ 1 and 41 Ϯ 1 nm (Fig. 5c). The magnitude of adhesion at D ϭ 31 nm was Ϫ0.29 Ϯ 0.08 mN/m, and the magnitude of the adhesion at D ϭ 41 nm was Ϫ0.13 Ϯ 0.06 mN/m. There was no adhesion at distances D Ͻ 31 nm (Table III).
To assess the relationship between ionic strength and PSA in controlling cadherin adhesion, force measurements were conducted with solutions containing 1 M NaNO 3 (Fig. 5d). In comparison to the measurements with mixed PSA-NCAM:cadherin monolayers in 150 mM NaNO 3 (Fig. 5b), the range and magnitude of the repulsive force are reduced (Fig. 5d). Specifically, the onset of the repulsion shifted inward from 50 to 42 Ϯ 1 nm, and the steep repulsive force shifted inward from 27 Ϯ 1 to 20 Ϯ 1 nm.
Upon separation, we measured two adhesive interactions: namely, Ϫ0.31 Ϯ 0.08 mN/m at 31 Ϯ 1 nm and Ϫ0.2 Ϯ 0.07 mN/m at 41 Ϯ 2 nm (Table III). There was no adhesion at 25 nm. Control measurements with NCAM and cadherin in 1 M NaNO 3 showed that high monovalent salt concentrations do not affect either protein.
Relative Contributions of Cadherin and NCAM-To demonstrate that the adhesive interactions at 25, 30, and 40 nm are because of the combined effects of both proteins, the cadherin was inactivated by decreasing the calcium concentration to 50 M (9, 29). Cadherin adhesion involving the outer N-terminal domain is abolished at this calcium concentration (9). Between cadherin monolayers (1.9 ϫ 10 4 cadherin/m 2 ) in 50 M Ca(NO 3 ) 2 , there was no adhesion at 32 and 40 nm. However, the proteins still bound, albeit more weakly, at 25 nm, with an adhesion of Ϫ0.3 Ϯ 0.1 mN/m (Table III).
Between mixed NCAM:cadherin (1:2.5) monolayers in a buffered solution containing 50 M calcium, the adhesion at 31 nm was significantly reduced to Ϫ0.15 Ϯ 0.08 mN/m (Table III). In the absence of cadherin adhesion, we attribute this to binding between NCAM Ig1-5 domains (c.f. Fig. 1b) (13). The adhesion at 40 nm also decreased to Ϫ0.1 Ϯ 0.06 mN/m, and is similarly attributed solely to homophilic NCAM binding. In similar measurements between mixed PSA-NCAM:cadherin (1:2.5) monolayers in 1 M NaNO 3 and 50 M CaNO 3 , the hysteresis at ϳ25 nm was absent. The adhesion at D ϭ 30 nm decreased to Ϫ0.1 Ϯ 0.06 mN/m, and the only evidence for binding at 40 Ϯ 1 nm was the hysteresis in the receding curve. These findings confirm that, at the higher calcium concentrations, both cadherin and NCAM contribute to the adhesion measured at 31 and 40 nm.
A control measurement of the force profile between PSA-NCAM in 3 mM calcium tested whether the divalent calcium altered the PSA properties. The CaNO 3 salt did not significantly alter the ionic strength, because NaNO 3 is the dominant electrolyte. The calcium had no effect on the force curves.

DISCUSSION
The force measurements described here provide direct support for the proposed "push-pull" mechanism for the inhibition of NCAM-mediated adhesion by PSA (6). This model postulates that the additive contribution of attractive protein forces and repulsive steric forces determines the equilibrium separation between cells. For PSA to block NCAM adhesion by increasing the nonspecific repulsive force between cells, the force balance requires that the magnitude of the additional repulsion exceed the NCAM attraction at both membrane separations at which NCAM binds. These measurements show quantitatively that the post-translational modification of NCAM substantially increases the magnitude of the repulsive pressure between membranes, sufficient to abrogate both NCAM-and cadherin-mediated adhesion. The increased excluded volume of PSA-modified NCAM was confirmed by both FCS and direct force measurements, but the latter directly demonstrated the functional impact of this modification on protein-mediated intermembrane adhesion. Furthermore, PSA reportedly has a general effect on several adhesion proteins, including cadherin. For PSA to similarly influence other adhesion molecules, the magnitude of the steric repulsion would also have to be large enough to abrogate, for example, cadherin binding at all three cadherin binding distances. The quantified magnitude of the steric repulsion at 25, 31, and 40 nm, the cadherin binding distances, again bears this out. Even with a 3-fold excess of cadherin, the PSA-associated repulsion exceeded the cadherin attraction, and abolished all adhesion.
The incomplete recovery of NCAM adhesion at 31 nm following endo-N digestion raises the question of the efficiency of the PSA removal. However, it is important to consider that the effects of glycosylation can be local, because of short-ranged steric repulsion between proteins, or long-ranged due to polymer repulsion between membranes. Because the NCAM adhesion at 31 nm requires full Ig1-5 overlap, the bulky glycan remaining on Ig5 would likely impede this association by shortranged steric repulsion. However, the short-ranged force would not affect the outer bond, and this is the observed behavior. Nevertheless, if this attenuation is because of long-ranged repulsion from uncleaved chains, then polyelectrolyte theory estimates the reduction in PSA chain lengths that would decrease the measured repulsion from 2 to ϳ0.15 mN/m. For monodisperse chains at a uniform grafting density in polyelectrolyte, the repulsion scales with the square of the degree of polymerization N 2 (7), so that, where F 1 and F 2 are the magnitudes of the repulsive forces at 31 nm before and after PSA cleavage, respectively. The average degree of PSA polymerization N 1 is estimated to be greater than 55 (14) and SDS-PAGE suggests that it could be as much as 200. 3 If we assume 100, then the degree of polymerization of the shortened chains would be. Although this is a very rough estimation, the endo-N-treated PSA-NCAM does not exhibit a molecular weight range consistent with this amount of residual PSA. This suggests that the reduced adhesion is most likely because of local steric repulsion by the residual glycan core. It is important to point out that this PSA-dependent repulsion is a general property of the polymer, and is independent of the identities of other proteins on the membrane. For this reason, the repulsive pressure between the PSA chains on adjacent membranes will have a general effect on the ability of other proteins to support intermembrane adhesion at membrane separations less than 45 nm, the measured range of the steric repulsion.
The ionic strength dependence of the range and magnitude of the PSA-mediated intermembrane repulsion provides further direct evidence that PSA impacts cell adhesion primarily through its large excluded volume. These data show directly that increasing the ionic strength decreases the steric dimensions of PSA. This in turn reduces the intermembrane repulsion, and restores the intercellular adhesion.
The repulsion between grafted brushes also scales with the grafting density of the chains (7). The influence of PSA on homophilic cadherin adhesion similarly depends on the ratio of PSA-NCAM to cadherin on the membranes; and hence on the balance between the steric repulsion and cadherin attraction. Again, this is a graded response, with the magnitude of cadherin adhesion increasing with decreasing PSA-NCAM coverage and with the corresponding decrease in the intermembrane repulsion. At the 1:2.5 PSA-NCAM:cadherin ratio, the cadherin-mediated adhesion, which was distinguished by binding at 25 nm, was abolished. At a 1:3.7 protein ratio, any residual attraction was only apparent as hysteresis. Cadherin adhesion finally reemerged at a ratio of ϳ1:8. This correlation between the PSA-NCAM density, the magnitude of the repulsion, and the adhesion further supports the hypothesis that the impact of PSA is directly linked to the range and magnitude of the associated nonspecific, intermembrane repulsion.
These findings show quantitatively that the magnitude of PSA-dependent repulsion correlates directly with the decreased trans homophilic adhesion by NCAM and cadherin. It is important to point out that the polymer occupies a threedimensional volume. Although there are no data showing that PSA affects any cis interactions, known or otherwise, the steric barrier generated by the excluded volume could also affect other, lateral NCAM interactions. This issue will be addressed in future studies.