| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Laboratoire d'Immunologie, Institut National de la Sante et de la Recherche Medicale U600, Centre National de la Recherche Scientifique FRE2059, Hôpital de Ste-Marguerite, Marseille, France
Correspondence: Address reprint requests to Professor Pierre Bongrand, Tel.: +33-491-26-03-31; Fax: +33-491-75-73-28; E-mail: bongrand{at}marseille.inserm.fr.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
0.045 µm3 surrounding each integrin. This aggregation induced up to 2.7-fold increase of the average number of bonds. Flow cytometric analysis of fluorescent ligand binding showed that THP-1 cells displayed low-affinity ß-1 integrins with a dissociation constant in the micromolar range. It is concluded that the initial step of cell adhesion was mediated by multiple incomplete bonds rather than a single equilibrium-state ligand receptor association. This interpretation accounts for the functional importance of integrin clustering. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
Many strategies are currently used to control cell adhesion. These include: i), adaptation of the repertoire of adhesive receptors through membrane translocation of stored molecules (Jagels et al., 1995
), de novo biosynthesis (Gailit et al., 1996), or alteration of the rate of receptor removal (Condic and Letourneau, 1997); ii), modulation of the intrinsic affinity of membrane receptors through conformational changes (Ma et al., 2002), proteolytic events (Van de Winkel et al., 1989), or variation of glycosylation patterns (Skelton et al., 1998); and iii), alterations of the cell receptor relationship with the membrane environment. The latter processes that may be denominated as postreceptor (Danilov and Juliano, 1989) or perireceptor events may involve an increase of receptor accessibility through concentration into convex parts of the cell surface such as the tip of microvilli (Erlandsen et al., 1993), partial degradation of repulsive molecular structures composing the glycocalyx (Sabri et al., 2000), release of cytoskeletal constraints resulting in increased capacity to diffuse toward ligands on interacting surfaces (Kucik et al., 1996), or on the contrary increased strength of connection with surrounding membranes resulting in higher capacity to maintain adhesion in the presence of disrupting forces (Dwir et al., 2001; Leid et al., 2001).
Much recent evidence supported the view that receptor clustering is frequently involved in the regulation of adhesion. Whereas dimerization of selectins (Li et al., 1998), members of the immunoglobulin superfamily such as ICAM-1 (Miller et al., 1995), or mucin-like selectin ligands such as PSGL-1 (Sako et al., 1993) may be important to ensure the functional efficiency of these adhesion receptors, clustering seems a dynamic way of regulating adhesive interactions mediated by cadherins (Lino et al., 2001) and most importantly integrin receptors (van Kooyk and Figdor, 2000); thus, although it has long been known that integrin complement receptor CD11bCD18 was inactive on resting phagocytes, neutrophil exposure to suitable stimuli such as phorbol ester resulted in concomitant increase of cell capacity to bind complement-coated particles and clustering of receptors into small aggregates of 2–6 molecules (Detmers et al., 1987). Much evidence was also reported on the correlation between functional activation and membrane clustering of CD11aCD18/LFA-1, another ß-2 integrin (van Kooyk et al., 1994; Stewart et al., 1998). Other reports support the hypothesis that ß-1- (Grabovsky et al., 2000; Maheshwari et al., 2000) and ß-3- (Kasirer-Friede et al., 2002) integrin clustering also influences functional capacity.
Although there is ample evidence that receptor clustering per se can enhance adhesion (Hermanowski-Vosatka et al., 1988; Yap et al., 1997; Hato et al., 1998), the molecular mechanisms responsible for this enhancement remain to be determined. At least three different mechanisms may be considered: first, binding site formation might require receptor multimerization, as was hypothesized for cadherins (Shapiro et al., 1995). Second, receptor clustering might trigger intracellular signaling events, such as tyrosine phosphorylation (Hato et al., 1998; Stripack et al., 1999). Third, clustered receptors might form multivalent bonds with surfaces bearing a sufficient density of ligand molecules, and multivalency would be required to achieve significant attachment that might be conducive to subsequent strengthening.
Although there is little doubt that the association of multiple weak bonds is a common way of achieving significant molecular interactions (Takaki et al., 2002), there remains to assess the quantitative consequence of multivalent association on the lifetime and mechanical strength of attachment. This is by no means a trivial question, as demonstrated in a recent theoretical study (Seifert, 2000); indeed, in absence of force, the lifetime of an attachment mediated by multiple parallel molecular bonds may be comparable to the lifetime of a single bond or much higher depending on the possibility of rebinding during the detachment process. Also, if a disruptive force is applied, the lifetime of a multimolecular attachment will be strongly dependent on the sharing of this force between individual bonds. Further, direct experimental determination of the effect of receptor clustering on adhesive efficiency may yield fairly inconclusive data. Thus, Chen and Moy (2000) used atomic force microscopy to determine the force required to break attachment between concanavalin-A-coated surfaces and control cells or cells expressing concanavalin-A ligand cross-linked with glutaraldehyde. Although cross-linking resulted in significant increase of the unbinding force (from 82 to 125 pN), it is difficult to determine whether this increase was a direct consequence of multivalent binding rather than increased cell rigidity or strengthening of the anchoring of cell surface molecules.
In this article, we describe a quantitative study of the effect of receptor clustering on the initial step of adhesion. We chose as a model the interaction of human monocytic THP-1 cells with fibronectin-coated surfaces; binding is mediated by low-affinity interactions involving 
4ß1 and 5ß1 integrins (Faull et al., 1993). Three sets of measurements were performed. First, the surface topography of cell surface integrins was manipulated by cross-linking with a monoclonal antibody that did not alter integrin affinity state, and receptor clustering was studied quantitatively by combining immunofluorescence, confocal microscopy, and image analysis. Second, receptor binding affinity was controlled with a modification of standard Scatchard analysis. Third, the effect of receptor clustering on adhesive efficiency was assessed by monitoring transient attachment events displayed by cells driven along fibronectin-coated surfaces in a laminar flow chamber. This chamber was operated under very low shear stress so that a single molecular bond might provoke a detectable arrest. This procedure was previously demonstrated to allow quantitative analysis of initial binding events, independently of postreceptor phenomena (Pierres et al., 1994) and it yielded a quantitative estimate of the lifetime of bonds formed by selectins (Kaplanski et al., 1993) or integrins (Masson-Gadais et al., 1999) borne by living cells. Here we found that fairly extensive receptor clustering resulted in significant change of the lifetime of initial cell-surface binding events. Detailed analysis of the duration of these binding events under diverse experimental conditions strongly supported the hypothesis that initial cell arrests were mediated by short-lived transient molecular interactions. This might explain the need for simultaneous formation of multiple ligand-receptor interactions to achieve sufficient avidity.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
). Cells were maintained as previously described (Sabri et al., 2000) in RPMI-1640 medium (Invitrogen, Cergy Pontoise, France) supplemented with 10% fetal calf serum (Invitrogen), 2 mM L-glutamine, 50 U/ml penicillin, and 50 µg/ml streptomycin. As checked with flow cytometry (not shown), these cells express CD49dCD29 (VLA-4) and CD49eCD29 (VLA-5) ß-1 integrins. Sterile plastic coverslips (10.5 x 22 mm2, Thermanox Ref 179434, Nalge Nunc International, Rochester, NY) were incubated for 2 h in pH 7.2 phosphate buffered saline buffer containing 10 µg/ml, 1 µg/ml, or 0.1 µg/ml human fibronectin (ref F6277; Sigma Aldrich France, St. Quentin Fallavier, France). They were washed extensively before use.
In some experiments, microfilaments were altered by treating cells for 15 min before adhesion tests with 10 µg/ml cytochalasin D (Sigma Aldrich) or 0.3 µM latrunculin A (Calbiochem, Merck Eurolab, Darmstadt, Germany)
Receptor aggregation and labeling
In some experiments, cells were treated with monoclonal antibodies specific for integrin ß-1 chain: 12G10 (a mouse IgG1 supplied by Serotec, Cergy Saint-Christophe, France) is known as a function-enhancing antibody whereas K20 is a neutral mouse IgG2a supplied by Beckman-Coulter-Immunotech (Marseille, France). Cell pellets were incubated in 25 µl of antibody solution for 30 min in RPMI supplemented with 1 g/l bovine albumin in ice bath. Integrin aggregation was induced by adding 25 µl of goat anti-mouse immunoglobulin (Immunotech, Marseille, France).
Deaggregated fluorescent antibodies were obtained by 45-min centrifugation (80,000 g) in a L7 Beckmann ultracentrifuge. Pellets were discarded and protein concentration was determined by measuring absorbance at 280 nm wavelength and using 1.46 as the optical density of 1 mg/ml immunoglobulin solutions (Mishell and Shiigi, 1980).
Adhesion under flow: data acquisition
Our procedure was previously described (Masson-Gadais et al., 1999). Briefly, fibronectin-coated coverslips were glued to the bottom of a plexiglass block bearing a rectangular cavity of 17 x 6 x 1 mm3 forming the chamber. Cells were suspended (1 million per ml) in RPMI 1640 medium supplemented with 10% fetal calf serum. They were then driven into the chamber with a 2-ml syringe mounted on an electric syringe holder (Razel Scientific Instruments, supplied by Bioblock, Illkirch, France). The wall shear rate G was calculated according to the standard formula: G = 6 Q/h2w, where Q is the flow rate (in mm3/s), whereas h and w are the chamber height (1 mm) and width (6 mm), respectively. A standard value of G = 4 s–1 was used throughout all experiments. The advantage of using such a low flow rate is that the hydrodynamic drag exerted on a cell bound to the surface is 5 pN, as obtained by modeling the cell as a sphere of 6-µm radius and using standard results from fluid mechanics (Pierres et al., 1996). Therefore, a single ligand-receptor bond should be able to maintain a cell at rest during a detectable amount of time, thus allowing direct visualization of individual molecular interactions (Pierres et al., 2001).
The chamber was set on the stage of an inverted microscope (Olympus IX, Olympus, Melville, NY) equipped with a charge-coupled-device video camera (Model SPT M208CE, Sony, Clichy, France). Observation was performed with a 10x objective and 1.5x additional magnification. In a typical experiment, the flow was recorded for 10 min in a selected microscope field with a VHS tape recorder for delayed analysis. The recorder output was connected to a PCVision+ digitizer (Imaging Technology, Bedford, MA) mounted on an IBM-compatible desk computer. This allowed real-time digitization yielding 512 x 512 pixel images with 256 gray levels. A custom-made software allowed tracking of the front edge of moving cells with a sampling frequency of 20/s (Kaplanski et al., 1993; Masson-Gadais et al., 1999). Pixel size was 0.6 µm along the trajectory axis. A typical trajectory is displayed on Fig. 1. Each videotape was replayed a sufficient number of times to monitor the motion of every cell passing through the microscope field during the recording period of time.
|
Binding frequency
The binding frequency P was calculated as the ratio between the number N of (transient or durable) arrests and total displacement L (P = N/L). The accuracy of determination of this parameter was quantified by using N1/2/L as an estimate of the standard deviation 
P, based on Poisson law (Masson-Gadais et al., 1999).
Dissociation rate
The durations of individual arrests detected in a given set of experiments were ordered, yielding a sequence such as:
 | (1) |
 | (2) |
However, as exemplified on Fig. 2, experimental dissociation plots of ln(F(t)) versus time were not straight lines. These plots were found to be adequately described with two parameters:
 | (3) |
d was estimated as previously described by assuming Poisson distribution for the numbers A and B of arrests with duration within time intervals [0.15 s, 0.45 s] and [0.45 s, 
], respectively (Pierres et al., 2002), yielding the formula:
 | (4) |
:
 | (5) |
|
 | (6) |
Note that the limiting values of 
2 with one and two degrees of freedom are, respectively, 3.84 and 5.99 for a confidence level of 0.05.
Comparison of experimental dissociation plots with theoretical models
As previously reported (Kaplanski et al., 1993; Pierres et al., 1995; Zhu et al., 2002), the interpretation of dissociation plots is not straightforward due to the possible occurrence of two main phenomena.
First, when a first bond triggered a cell arrest, sequential bond formation and dissociation events may occur. The simplest model allowing one to account for this possibility consists of defining the rates of bond formation (kon) and dissociation (koff) and writing the following set of equations (Cozens-Roberts et al., 1990; Kaplanski et al., 1993):
 | (7) |
1i, where is Kronecker's symbol.
It was recently found (Pierres et al., 2002) that better fit between experimental and calculated plots might be obtained with the so-called multiple-bond model with instantaneous bond formation: following this model, kon is set equal to zero. Further, according to Poisson law:
 | (8) |
represents an average bond number that will be referred to as Poisson parameter. The underlying assumption is that several bonds may be formed within less than a fraction of a second if several ligand-receptor couples are at binding distance when cell-surface contact occurs.
Second, it is now well documented that a single ligand-receptor complex may undergo different states with different lifetimes (Pierres et al., 1995, 2002; Merkel et al., 1999). The simplest way of accounting for this possibility consisted of assuming that each cell arrest resulted from the formation of a transient complex with dissociation rate koff and rate of transition kt toward a stable state (i.e., a binding state with a lifetime much higher than 1 s). The basic equations can be solved analytically when there is a single bond (Pierres et al., 1995). A set of master equations resembling Eq. 9 can be obtained by defining as Pi,j(t) the probability that a cell be bound by i transient-state interactions and j stabilized interactions at time t. The standard equation is:
 | (9) |
The simplest two-parameter model was obtained by considering only the off rate of the transient state (koff) and the transition rate kt from intermediate to stabilized state. In some cases, it was found necessary to account for the reversion rate k2m, i.e., the rate of passage from stabilized to transient interaction.
Numerical solution of this set of equations was straightforward, with proper truncation by limiting at 10 the maximum number of cell-substrate bonds during the first second after arrest.
A third model was based on the hypothesis that cell attachment might be mediated by two different kinds of complexes with dissociation rates koff,A and koff,B, respectively. After straightforward calculations, the binding probability at time t is:
 | (10) |
Assessment of receptor aggregation with confocal microscopy
Data acquisition
Receptor aggregation was assessed semiquantitatively by taking advantage of the exquisite sensitivity of confocal microscopy. Indeed, this device was found to detect a few or even single fluorescent molecules (Nie et al., 1994). Under standard conditions, cells were labeled in the cold with fluorescein-conjugated anti-CD29 mouse monoclonal antibodies as previously described, with or without a second layer of unlabeled polyclonal goat anti-mouse immunoglobulin. They were then fixed with 1% paraformaldehyde and examined rapidly with a confocal laser fluorescence microscopy (Leica CLSM, Leica Microsystems, Heidelberg, Germany), using an Argon/Krypton laser (Omnichrome, Leica) and a 40x dry objective. Pixel size was thus 245 x 245 nm2 with a vertical resolution of order of 700 nm. Typically, a given cell was represented as a series of about six 512 x 512 pixel images (8-bit depth) with 2-µm separation between sequential section planes. It was found convenient to set detection photomultiplier at a moderate value of 600 volts so that the background level is sufficiently low to avoid any confusion between random noise and fluorescence of volume elements containing a few labeled molecules.
To relate the fluorescence intensities of individual pixels to local density of fluorescent molecules (Pierres et al., 1994), 10 µl aliquots of sequential dilutions of fluorescent antibodies were deposited on glass slides and covered with 22 x 22 mm2 coverslips, thus forming a liquid layer of 20-µm thickness that was examined with the same microscope settings as cell samples.
Data processing
Images were transferred to desktop computers with a local network and processed with a custom-made image analysis software (André et al., 1990). Conventional observation of cell images would convey a misleading grasp of molecule aggregation because visible bright points represented large molecular aggregates representing only a small fraction of the total amount of receptors. Small cluster aggregation was assessed by calculating a median fluorescence aggregation index defined as the fluorescence intensity f such that 50% of fluorescent molecules are located in pixels with a brightness lower than f. This may be viewed as a weighted median fluorescence because the following equation holds:
 | (11) |
Study of receptor affinity with soluble ligand molecules
To achieve correct interpretation of experimental data, it was important to check the effect of receptor aggregation treatment on their affinity. Thus, we studied the binding of two custom-made fluorescent integrin ligands by THP-1 cells: fluorescein-AEILDVPST contains the LDVPS sequence recognized by VLA-4 in the CS-1 domain of fibronectin (Mould and Humphries, 1991). Fluorescein-AGRGDSPK contains the RGD sequence found in the fibronectin region recognized by VLA-5 (Wayner et al., 1989). These molecules were supplied by Neosystem (Strasbourg, France).
Experiments were conducted by incubating cell pellets in 50 µl of sequential twofold dilutions of peptide ligand with concentration ranging between 0.38 µg/ml and 200 µg/ml. They were then examined with a Coulter EPICS/XL flow cytometer (Beckmann-Coulter France, Villepinte) for determination of relative median fluorescence.
Data analysis was performed with a modification of Scatchard procedure (Scatchard, 1949): the basic assumption is that the mean fluorescence F of a cell incubated in a solution of fluorescent ligand of concentration [L] is:
 | (12) |
The first term on the right-hand side of Eq. 12 is cell autofluorescence, the second term represents nonspecific binding that is assumed to be linearly dependent on ligand concentration, and the third term represents specific binding, defining T as the total number of binding sites on the cell, f as the intrinsic fluorescence of a ligand molecule, and Ka the affinity constant of ligand receptors. Equation 12 may be written:
 | (13) |
Because the receptor affinity was low enough to make the amount of bound ligand negligible with respect to the total amount of fluorescent ligand, [L] was readily derived from the known concentration of fluorescent ligand solution. Further, F – F° was calculated by mere subtraction of the fluorescence of unlabeled cells samples. Finally, constant was approximated as the limit of (F – F°)/[L] for high-ligand concentration: this limit was determined by calculating the average value of (F – F°)/[L] corresponding to three to four values of [L] ranging between 50 µg/ml and 300 µg/ml (i.e., 5 10–5 and 3 10–4 M), which was much higher than the dissociation constant 1/Ka. Parameters Ka and Tf were then determined with the Scatchard procedure by plotting 1/{(F – F°)/[L] – } versus [L] and determining the equations of the regression line in the ligand concentration range 
Determination of fibronectin surface density
Mouse anti-human fibronectin monoclonal antibody HFN 7.1 was purified from hybridoma culture supernatants (American Type Culture Collection CRL-1606) and labeled with the Alexa Fluor protein labeling kit provided by Molecular Probes (Eugene, OR) as indicated by the supplier. Fibronectin-coated surfaces were treated with Alexa-Fluor 488-conjugated antibodies and fluorescence surface density was determined with confocal microscopy. Calibration was achieved by measuring the fluorescence of 5-µl samples of serial dilutions of labeled antibody deposited on a glass slide and covered with 10.2 x 22 mm2 coverslips.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
|
It is concluded that adhesion was essentially mediated by CD49eCD29 (5ß1) integrin under our experimental conditions.
The duration of binding events display wide variations, ranging between a fraction of a second and more than several minutes
As exemplified in Fig. 1, monocytic THP-1 cells driven along fibronectin-coated surfaces with low hydrodynamic forces exhibited periods of motion with fairly uniform velocity interspersed with arrests of varying duration. The distribution of arrest durations was determined for several experimental conditions. The detachment plot obtained after recording 448 arrests on a series of 1226 control THP-1 cells moving along surfaces pretreated by 10 µg/ml fibronectin is shown on Fig. 2, revealing three sequential parts:
0.15 s. This was felt difficult to analyze because the reliability of experimental data was hampered by potential artifacts related to cell shape irregularities, as discussed below. 0.45 s after arrest, corresponding to first-order detachment kinetics. Therefore, it seemed reasonable to estimate the actual number of binding events by extrapolation to time zero of the tangent to the detachment curve in the linear region (Fig. 2). The difference between the extrapolated and experimental number of arrests ranged between 13% and 40% following experimental conditions as described below. It is concluded that three quantitative parameters could be derived with sufficient robustness from experimental analysis of cell attachment and detachment: i), binding frequency, defined as the mean number of binding events per µm displacement, ii), initial detachment rate, defined as the slope of detachment curves in the time interval [0.15 s, 0.45 s], and iii), fraction of cells remaining bound 1 s after arrest. The experimental values of these parameters were determined under various conditions affecting surface ligand density or cell receptor and cytoskeleton status. Parameters obtained after monitoring 8582 individual trajectories are summarized in Table 2 and full detachment curves are displayed in Fig. 3. A total number of 2136 arrests were observed, and 52% of them were shorter than 1 s.
|
|
). This question was addressed by modifying the surface density of fibronectin on coverslips. When the fibronectin concentration used to derivatize adhesive surfaces was decreased from 10 µg/ml to 1 µg/ml, the initial detachment rate displayed about twofold increase (Table 2). However, when fibronectin concentration was further decreased from 1 µg/ml to 0.1 µg/ml, binding frequency exhibited nearly fourfold decrease, whereas detachment curves were identical on a time period of 5 s (Fig. 3 A). Thus, arrests observed with the lower two fibronectin concentrations reflected minimal binding events, and they were assumed to represent single molecular bonds.
Detachment curves obtained with low-ligand density for time values lower than 1 s may be accounted for by assuming the existence of a transient ligand-receptor complex with a dissociation rate of 3.6 s–1 and a transition frequency of 1.30 s–1
At least two nonexclusive mechanisms might account for the upward curvature of detachment curves (Fig. 3):
Results displayed on Fig. 3 A suggest that no bond formation occurred after arrest at low-fibronectin density, because detachment curves found on surfaces treated with 1 µg/ml and 0.1 µg/ml fibronectin were identical. To check this hypothesis, we determined the numerical parameters yielding minimal value of the 2 difference (as calculated with Eq. 6) between theoretical curves and the detachment curve obtained with control cells interacting with surfaces coated with low (0.1 µg/ml) fibronectin concentration. The best fits obtained with both models are shown in Fig. 4. Clearly, a much better agreement was obtained with the bond-stabilization model (2 = 0.003) than with the multiple-bond model (2 = 4.3; P < 0.05). The fitted parameters were koff = 3.6 s–1 and kt = 1.30 s–1.
|
). This was called the immediate bond-formation model. As shown on Fig. 5, the immediate bond-formation model yielded slightly better 2 value (1.34; P 
0.25) than the continuous bond-formation model (2 = 3.05; P < 0.10). It may also be noticed that only the immediate bond-formation model yielded downward concavity of the initial part of detachment curves.
|
Increasing complexity may be addressed by considering time values comprised between 0.15 and 5 s after initial arrest
As shown in Fig. 3 B, our two-parameter model (koff = 3.6 s–1 and kt = 1.3 s–1; thick line) did not satisfactorily account for single-bond rupture curves during the first 5 s after arrest. The most reasonable way of dealing with this discrepancy was to allow for the reversibility of bond strengthening by adding a transition frequency k2m for the return from "stable" to transient state in single bonds. As shown in Fig. 3 B, excellent fit was obtained with k2m = 0.19 s–1. Introducing this additional constant also allowed satisfactory account of detachment curves at high fibronectin surface density (Fig. 3 A). Interestingly, this change did not significantly affect the analyses performed on a narrower time interval of [0.15 s, 1 s]. However, because <10% of all arrests lasted between 1 and 5 s, k2m could be estimated with only low accuracy. Thus, because we were essentially interested in the initial step of cell-surface interaction, it was found safer to restrict analyses to the shorter time interval [0.15 s, 1 s], to avoid interpretation problems due to the possibility of fitting curves with multiple combinations of parameters.
However, the above interpretation was not unique. Detachment curves could also be accounted for by a three-parameter model assuming that cell attachment might be mediated by two different complexes with dissociation rates koff,A and koff,B (Eq. 9). Taking for koff,B the average slope of the detachment curve (Fig. 3 B) on the time interval [1 s, 5 s], i.e., 1.37 s–1, koff,A, and the proportion r of type-A complex were fitted on time interval [0 s, 1 s] as described above. The dissociation rate koff,A was estimated at 5.1 s–1 and the proportion of arrests mediated by a type-A bond was 0.66, with a 2 parameter of 1.28, yielding a satisfactory agreement between experimental and theoretical values on the entire [0 s, 5 s] interval (Fig. 3 B, dotted line).
Clustering surface ß-1 integrins on THP-1 cells with anti-mouse immunoglobulin increases binding frequency and duration of cell arrests on fibronectin-treated surfaces under flow
Integrin activity is difficult to study because it is known to be dependent on both molecular topography and conformation, and both parameters may be modulated by cell activation. We attempted to assess the specific consequences of topographical changes by selective manipulation of this parameter with neutral K20 antiintegrin antibody.
When K20-treated THP-1 cells were made to interact with surfaces coated with 1 µg/ml fibronectin (Fig. 3 D), experimental detachment curves were quite close to theoretical curves corresponding to the two-parameter model (koff = 3.6 s–1, kt = 1.3 s–1) because 2 was 0.93, and binding frequency displayed 40% decrease as compared to control cells (Table 2).
When K20-treated THP-1 cells interacted with surfaces coated with 10 µg/ml fibronectin, binding frequency was decreased by 50% as compared to controls, and the experimental detachment curve was slightly but significantly different from the aforementioned two-parameter single-bond model (2 = 4.88, as compared to 2 = 171 for control THP-1 cells). The most straightforward interpretation of these experiments was that K20 antibody slightly decreased integrin accessibility under flow, but that it did not affect kinetic parameters, as expected.
In another set of experiments, THP-1 cells were first treated with K20 antibody, then exposed in the cold to polyclonal anti-mouse immunoglobulin to induce clustering of K20-treated integrins, and finally allowed to interact with fibronectin-coated surfaces. As shown in Table 2, this treatment resulted in both increase of binding frequency (2.6-fold; P < 0.001) and decrease of initial detachment rate (44%; P < 0.05) of THP-1 cells on surfaces treated with low fibronectin concentrations (1 µg/ml). When surfaces pretreated with higher fibronectin concentrations (10 µg/ml) were used, the same trend was observed, although neither binding frequency increase nor detachment rate decrease were significant.
Using the initial bond-formation model, the best fits between experimental and theoretical curves were obtained by assuming a mean initial bond number of 2.5 and 2.7 with aggregated integrins (on surfaces treated with 10 µg/ml and 1 µg/ml fibronectin, respectively), and 1.5 and 1 in the absence of integrin aggregation (on surfaces treated with 10 µg/ml and 1 µg/ml fibronectin, respectively).
Binding frequency and duration under flow are increased by function-enhancing 12G10 antibody
It was interesting to determine the influence on our experimental parameters of agents known to modify integrin conformation; as shown in Table 2, function-enhancing 12G10 antibody increased both binding frequency (twofold) and binding duration (2.8-fold) on surfaces coated with 1 µg/ml fibronectin, whereas no significant effect was found on surfaces coated with 10 µg/ml fibronectin.
As shown on Fig. 3 E, detachment curves determined on 12G10-treated cells were similar on surfaces coated with 10 µg/ml and 1 µg/ml fibronectin. Assuming that arrests were mediated by single bonds, optimal fit with theoretical curves was obtained with a koff value of 0.915 s–1, using unchanged kt = 1.3 s–1 (2 = 0.57). Thus, the effect of 12G10 treatment might be accounted for by a nearly fourfold reduction of the dissociation rate of transient state. However, several different parameter combinations might yield similar results, and more studies would be required to determine the precise effect of 12G10 treatment on receptor-binding parameters.
The effect of microfilament inhibition may be accounted for by assuming decrease of receptor accessibility without any change of dissociation rate
Microfilament inhibitors have often been used to alter integrin function by modulating topographic distribution and mobility on the cell membrane. It was therefore interesting to quantitate the effect of such treatment on binding parameters measured under flow. As shown in Table 2, cytochalasin D and latrunculin A, respectively, decreased binding frequency by 47% and 26%. Both detachment curves matched our simple two-parameter single-bond model with 2 = 1.38 and 1.68, respectively, without any need for a change of parameters kon and kt (Fig. 3 F). The simplest interpretation of these results was that microfilament inhibition decreased receptor accessibility without any conformational change.
Receptor aggregation may be estimated with confocal microscopy
Experiments were performed to assess the capacity of confocal microscopy to monitor the aggregation of fluorescent molecules. Serially diluted solutions of fluorescent anti-CD29 antibodies were examined with a confocal microscope as liquid sheets of 20-µm thickness on a glass slide. As shown in Fig. 6, the average field brightness was linearly dependent on antibody concentration, and a 5-µg/ml concentration increase resulted in an increase of 1 unit on the 256 level intensity scale. Because the pixel size was 0.245 µm and vertical resolution was 0.7 µm, the average number of molecules per pixel was 0.8 for a 5-µg/ml solution. Thus, with our microscope settings, an intensity unit represented about one molecule of labeled immunoglobulin (i.e., two to three fluorescein groups) in a volume element (i.e., a voxel). Note that a 5-µg/ml fluorescent antibody solution appeared as completely dark on the microscope screen with standard color palette.
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
To clarify the significance of our findings, some methodological points deserve to be discussed. Although the use of flow chambers became more and more common in biological laboratories during the last 10 years, most authors used wall shear rates of the order of 100 s–1 or more (Lawrence and Springer, 1991) to mimic rheological conditions occurring in blood flow. In contrast, we operate the flow chamber at 100-fold lower shear rates, leading to several interesting features.
) or some integrins (Grabovsky et al., 2000) can withstand the higher forces usually exerted in flow chambers after an interaction shorter than a few milliseconds. Even in these cases there remains a possibility that reported first-order dissociation kinetics (Alon et al., 1995) be mediated by bond clusters rather than single bonds (Evans et al., 2001; Zhu et al., 2002). In any case, our conditions are certainly required to allow direct observation of the initial step of integrin-ligand interactions that would not be detected at higher shear rate (Lawrence and Springer, 1991). ) that may mimic adhesion-related arrests, and the cell boundary may not be determined with the same accuracy as the center of gravity of a sphere. Thus, it was deemed unwarranted to analyze the initial part (t < 0. 15 s) of detachment curves. Our analysis relies on the basic assumption that during the first second after arrest, binding events involving control cells interacting with lower ligand densities were mediated by single bonds. This assumption is supported by the following two arguments: i), it seems reasonable to assume that our methodology allowed single-bond detection; and ii), if more than one bond occurred during the first second after arrest, the average bond number would be significantly lower on surfaces coated with 0.1 µg/ml fibronectin than on surfaces coated with 1 µg/ml fibronectin, where binding frequency was nearly fourfold higher, and detachment kinetics should be different.
As shown in Fig. 3 B, at least two models might be considered to account for these results. First, fibronectin-integrin interaction might be initiated by the formation of a transient complex with a dissociation rate of 3.6 s–1 and 1.3 s–1 frequency of transition toward a more stable state with a lifetime much higher than 1 s. Alternatively, attachment might involve two complexes with respective dissociation rates of 5.1 s–1 and 0.137 s–1, and respective proportions of 66% and 34%. In any case, the most frequent event after integrin/fibronectin encounter seems to be the formation of a transient complex with a dissociation rate close to 4–5 s–1. This is in contrast with recent studies made on complexes formed between truncated 5ß1 and wild-type or mutated fibronectin: the dissociation rate was 0.0065 s–1 and 0.0113 s–1, respectively, when fibronectin bore or was deprived of an active synergy site (Takagi et al., 2003). Also, the force-free separation of fibronectin and 5ß1 integrin was, respectively, estimated at 0.13 s–1 and 0.012 s–1 with high- or low-affinity state integrin (Li et al., 2003). The simplest interpretation of this apparent discrepancy would be that the transient binding events we reported could not be detected with other methods based on surface plasmon resonance (Takagi et al., 2003) or atomic force microscopy (Li et al., 2003). Indeed, the contact time between interacting molecules is of the order of 1 ms under our experimental conditions (i.e., bond range of 10 nm divided by the relative membrane-to-surface velocity of 10 µm/s). In contrast, contact duration in the aforementioned experiments might range between several hundreds of seconds (Takagi et al., 2003) and 50 ms or more (Li et al., 2003). It may be emphasized that a similar disclosure of previously undetected short interactions was recently reported by Smith et al. (1999) when they studied selectin-ligand interaction under flow with a rapid video camera.
Because the "bond-stabilization" and the "two-complex" model can yield quite similar detachment curves after proper parameter fitting, it seems very difficult to find a formal way of discriminating between these models. However, we tend to favor the bond-stabilization model because treatments supposed to change integrin accessibility without altering their conformation did not modify the shape of detachment curves, whereas they might be expected to change the proportion of independent complexes. Indeed, exposure to neutral K20 antibody or microfilament inhibition, resulted in a decrease of binding efficiency without any alteration of detachment kinetics. The decrease of binding frequency after K20 treatment (Table 2) may be ascribed to a steric hindrance that should be particularly effective under flow conditions, due to the briefness of molecular encounter. It may seem surprising that cytochalasin D and latrunculin A decreased binding frequency. Because detachment curves were not altered by these drugs, the simplest interpretation might be that receptor accessibility be altered, possibly by a change of cell shape and/or receptor repartition between cell surface protrusions and other membrane regions, because such variations were suggested to alter integrin function (Erlandsen et al., 1993).
The effect of 12G10 antibody is more difficult to understand: indeed, in view of previous evidence, function-activating antiintegrin antibodies might be expected to increase the rate of bond formation (Cai and Wright, 1995; Li et al., 2003; Takagi et al., 2003), as well as decrease the rate of bond dissociation (Chigaev et al., 2001; Li et al., 2003). Therefore, it was surprising that this antibody did not promote multiple-bond formation at least on surfaces coated with high-fibronectin density. However, this conclusion was supported by two independent arguments, i.e., the lack of increase of bond frequency (Table 2, first and fourth data rows) and analysis of detachment curves. This might be due to a steric effect precluding multiple-bond formation. Indeed, due to the short duration of molecular encounter, steric hindrance may be more apparent under flow than under conditions used by other authors, in accordance with the inhibitory effect of supposedly neutral K20 antibody on bond formation (Table 2).
It seems, therefore, warranted to ascribe the slower release of cells bound to surfaces coated with higher fibronectin densities to the formation of multiple bonds. Because the slightly better fit obtained with the initial bond-formation model, as compared to continuous bond-formation model (Fig. 5) is not sufficient to rule out the latter model, it is useful to emphasize additional support for our preference to the initial bond-formation model; the delay required for bond formation when a fibronectin molecule is within binding distance of the cell membrane is certainly much less than one second. Indeed, because the size of a receptor is of the order of 10 nm, the local fibronectin concentration near the receptor would be of the order of 1/[N x (10 nm)3] = 0.0017 M, where N is the Avogadro-Loschmidt number. Because the kinetic constant of the integrin-ligand association is of the order of 106 M–1 s–1 (Chen et al., 1999), bond formation would require <1 ms. Note that although the aforementioned estimates were obtained on soluble molecules, similar amounts of time are expected to be involved in bond formation between free and surface-attached molecules provided free receptor rotation is allowed (Pierres et al., 1998). The only consequence of receptor attachment to a surface, as compared to soluble molecules, would be a possible reduction of the maximum number of bonds due to limitation of the number of available molecular orientations. Note that this initial bond-formation model was also found to hold in a recent study of the microkinetics of streptavidin-biotin association (Pierres et al., 2002).
Given that K20 antibody did not change the intrinsic fibronectin-integrin interaction, as showed by dynamic
), 1 µg/ml (crosses), or 0.1 µg/ml (
) fibronectin. The thin line was obtained with a three-parameter model (koff = 3.6 s–1, kt = 1.3 s–1, k2m = 0.19 s–1). The thick line was obtained with a five-parameter model derived from the previous one by allowing bond formation with kinetic rate kon = 1.44 s–1 and average initial number of bonds 2.16. (B) Control cells interacting with surfaces coated with 0.1 µg/ml fibronectin. The thin line was obtained with the same three-parameter model as shown in A, the thick line was obtained with the simplest two-parameter model (single intermediate complex with off rate koff = 3.6 s–1 and transition rate to more stable conformation kt = 1.3 s–1). The dotted line was obtained with another three-parameter model, assuming that attachment might be mediated by two complexes with dissociation rates of 5.1 s–1 (66%) or 0.137 s–1 (34%). (C) Cells were treated with neutral K20 antibodies without (
) or subtraction of autofluorescence and nonspecific binding assumed to be linearly dependent on [L] (crosses). The latter curve is linear, in accordance with