Design of mutations in D1D2 affecting IpaA-induced trimer formation

Structural models indicate that aVBD triggers a 30% angle displacement of the major axis of the D1 and D2 subdomains relative to the apo D1D2 or D1D2 in complex with IpaA VBS1-2 only (Fig 1B; Valencia-Gallardo et al, 2023). As opposed to the closed D1D2 conformer, this open D1D2 conformer is associated with D1D2-mediated trimerization of vinculin (Valencia-Gallardo et al, 2023). We hypothesized that IpaA VBS3 induced allosteric changes leading to D1D2 trimerization and that mutations in D1D2 residues interfacing IpaA VBS3 should affect trimer formation. We, therefore, scrutinized the interface residues between IpaA VBS3 and D1D2 in the D1D2:aVBD complex.

As shown in Fig 1C, in the closed D1D2:aVBD complex, IpaA VBS3 mainly interacts with the H10 helix of D2. A set of putative polar interactions and salt bridges can be identified, where IpaA residues T492, E495, K499, and S503 interact with vinculin residues N393, K139, D389, and K386, respectively (Fig 1C). In the open complex, IpaA VBS3 mainly interacts with the H4 helix of D1, where IpaA K499 interacts with vinculin E143 (Fig 1D). Of note, in this open complex, an electrostatic clash between vinculin K139 and IpaA K499 may contribute to the dynamics of the IpaA VBS3 during its interaction with different allosteric conformers leading to D1D2 oligomerization (Fig 1D). All identified vinculin residues were substituted for a charged residue to introduce a charge inversion or disrupt polar interactions using site-directed mutagenesis (see the Materials and Methods section; Table 1). In this rationale, mutations affecting IpaA VBS3’s interface with the close D1D2 conformer are expected to alter its initial docking of IpaA VBS3 on D1D2, whereas the E143K charge inversion would destabilize the open conformer and alter subsequent allosteric changes leading to trimer formation.

Independent of the aVBD-D1D2 interface, the hidden Markov model-based algorithm MARCOIL predicts the presence of a coiled-coil domain in the H10 helix of D2 between vinculin residues 348–393, buried into the D2 helix bundle and adjacent to the IpaA VBS3 interaction sites in the close D1D2 conformer (Delorenzi & Speed, 2002; Fig S1). This putative coiled-coil domain contains the classical a-g heptad sequence at residues 367–373, with the V367 and A373 corresponding to the “a” and “d” hydrophobic residues, respectively, presumed to intersperse their non-polar side chains at the α-helices interfaces during oligomerization. To test the role of this domain in D1D2 trimerization, we introduced the V367D substitution predicted to disrupt supercoiled helix packing (Fig S1A–D).

Quantitative analysis of D1D2 trimer formation using CN-PAGE

In previous works, we showed that aVBD induced the formation of D1D2:aVBD 1:1 and dimeric complexes and trimeric D1D2 complexes rapidly associating and dissociating an aVBD molecule in SEC-MALS (Size Exclusion Chromatography-Multi-Angle Light Scattering) experiments (Valencia-Gallardo et al, 2023). Here, we studied the effects of increasing aVBD molar ratio using SEC and CN (Clear Native)-PAGE. As shown in Fig 2, at a D1D2:aVBD 1:1 M ratio, the major peaks with similar amplitude corresponded to the 1:0 and likely the D1D2 trimeric complexes (Fig 2A, peaks A and C). At a D1D2:aVBD 1:1.5 M ratio, the D1D2 trimeric complexes represented the major peak (Fig 2A, peak A). Upon increasing of IpaA molar ratio, a major shifted band was observed in CN-PAGE with apo D1D2 rapidly disappearing and the formation of intermediate shifted bands (Fig 2B). The similarity between the distribution of the bands with increasing molar ratio of aVBD in CN-PAGE and the SEC peaks suggested that the A and A′ peaks corresponded to trimeric D1D2 complexes. To confirm this, we dissected the A′ band form CN-PAGE and analyzed it in a second dimension using regular SDS–PAGE, followed by Coomassie staining and scanning densitometry (see the Materials and Methods section). As shown in Fig 2C and D, the relative amounts of D1D2 and AVBD in A′ indicated a D1D2:aVBD integrated density ratio of 10.1 ± 0.36 corresponding to a molar ratio of 3.2 ± 0.09, consistent with a mixture of D1D2:aVBD 3:0 and 3:1 complexes inferred from the SEC-MALS analysis (Valencia-Gallardo et al, 2023).

We used scanning densitometry to determine the rates of D1D2 trimer formation and D1D2 monomer disappearance normalized to the initial amounts of D1D2 (see the Materials and Methods section). As shown in Fig 2E, the appearance of D1D2 trimers and disappearance of apo D1D2 as a function of increasing aVBD molar ratio could be nicely adjusted to linear fits with a Pearson coefficient R² > 0.95. The CN-PAGE assay was used to determine a rate of D1D2 trimer appearance of 1.51 AU⁻¹ ± 0.3 (SEM) and D1D2 monomer disappearance of −1.06 AU⁻¹ ± 0.05 (SEM) (Table 1).

Characterization of D1D2 mutations affecting trimer formation

We estimated that the CN-PAGE assay was sufficiently robust and reproducible to determine potential differences in rates of trimer formation in the various D1D2 variants. All mutations were introduced in D1D2 and the corresponding variants were purified to homogeneity (Fig S2). Samples were incubated with increasing molar ratio of the aVBD vinculin-binding domain (aVBD) and analyzed by CN-PAGE followed by Coomassie staining (see the Materials and Methods section). As shown in Figs 3 and S3, most of the mutants showed decrease rates of trimer formation. The rates of D1D2 trimer formation and monomer disappearance were inferred from linear fits, following quantification by scanning densitometry (Table 1).

As shown in Table 1, all mutations showed reduced rates of D1D2 trimer and could be distinguished in two classes based on to their effects on the rates of monomer disappearance. Class 1 mutations D389R, E143K and V367D showed no difference in D1D2 monomer disappearance, whereas class 2 mutations K386D and K139E showed increased rates in D1D2 monomer disappearance (Table 1 and Figs 3 and S3). Mutation N393D showed effects similar to the latter mutations, with a slightly reduced rate of trimer formation that appeared not significant, but increased rates in monomer disappearance (Fig 3A–C). These classes generally correlated with the mutational design based on structural modeling because class 2 and class 1 mutations were expected to target formation of the closed and open D1D2:aVBD 1:1 complex, respectively (Fig 1C and D). The D389R mutation represents an exception, because it was expected to target the D1D2:aVBD 1:1 close complex, but showed no difference in the rate of monomer disappearance (Table 1 and Fig 3A–C).

D1D2 mutations targeting the close conformer prevent Shigella invasion

During Shigella invasion, targeting of vinculin by IpaA leads to the cytoskeletal reorganization allowing bacterial attachment to host cells and productive bacterial internalization. In addition, IpaA-mediated vinculin supra-activation triggers the formation of focal adhesions (FAs) distal to the invasion sites that likely contribute to strengthen the adhesion of infected cells even in the absence of mechanotransduction (Valencia-Gallardo et al, 2023). To analyze the role of D1D2 head-domain oligomerization on Shigella invasion, D1D2 mutations were transferred to full-length vinculin fused to mCherry (HV-mCherry) cloned in a eukaryotic expression vector and transfected into MEF vinculin −/− cells (see the Materials and Methods section). In control experiments, Western-blot analysis showed that all constructs were expressed at the expected molecular size at levels comparable with that of WT vinculin-mCherry construct (Fig S4). Transfected cells were challenged with bacteria and bacterial association and internalization were quantified in cells expressing comparable levels of HV-mCherry constructs (see the Materials and Methods section).

As shown in Fig 4, the profiles of bacterial association and invasion were generally similar, consistent with a prominent role in bacterial attachment to the cell during Shigella invasion. Consistent for the previously described role for vinculin in Shigella invasion, cells complemented with HV-mCherry showed a ca 8-fold and 30-fold increase in bacterial association and invasion compared with mCherry transfected control cells, respectively (Fig 4A–C, HV and mCherry). Similar differences in bacterial association and invasion were observed when comparing with HV-mCherry transfected cells challenged with a non-invasive isogenic mutant Shigella strain (Fig 4B and C, HV and BS176). Strikingly, mutations D389R and N393D resulted in a drastic reduction in bacterial association and invasion of host cells, to levels comparable with the negative controls (Fig 4B and C). In contrast, the other tested mutations had modest or non-significant effects. These results suggest a key role for polar and charged interactions involving the D389 and N393 at the N-terminal extremity of the H10 helix of D2. The absence of effects associated with the cysteine clamp (CC) and E143K mutations targeting the open D1D2 conformer suggested that early steps linked to vinculin supra-activation but not later steps linked to vinculin oligomerization are required for Shigella invasion.

Mutations affecting D1D2 trimerization impair focal adhesions’ formation

Vinculin oligomerization has been mainly studied in vitro and reported to occur through the Vt-tail domain (Thompson et al, 2013). In cells, the role of vinculin oligomerization remains unclear. We therefore tested the effects of D1D2 mutations of FA formation, a process relevant for bacterial infection of host cells because it controls the adhesion/detachment of Shigella-infected cells from the extracellular matrix.

As shown in Fig 5, all tested mutations induced a decrease either in FA area or FA number per cell (frequency) or both, although to variable extent (Figs 5 and S5A–C). The K386D mutation showed a decrease in FA area and frequency but the difference relative to parental HV was not statistically significative, as opposed to the other mutations (Fig 5B and C). As previously reported, the CC mutation altered the size and frequency of FAs (Fig 5; Valencia-Gallardo et al, 2023). The K139E mutation also significantly affected both FA size and frequency. The E143K, D389R and N393D mutations affected the size of FAs, whereas the difference in FA number per cell relative to parental HV was not significant because of the large dispersion of values (Fig 5B and C). Interestingly, the V367D variant also did not show a significant difference in the average FA size compared with control cells expressing parental mCherry-vinculin (Fig 5B and C), despite a significant reduction in the number of FAs per cell. The FAs in the V367D variant, although similar in size, looked different than those in control cells, being mostly at the cell periphery and particularly elongated (Fig 5A). This qualitative difference was confirmed by quantification of the AR index, showing a pronounced difference for the V367D variant compared with the other samples (Fig S5B and C).

Together, these results support a role for vinculin supra-activation and head-domain–mediated oligomerization in the formation of focal adhesions independent of Shigella invasion.

IpaA VBSs prevents cell motility

Our findings suggested that during Shigella invasion, IpaA-mediated vinculin head-domain oligomerization played mostly a role in up-regulating cell adhesion rather than bacterial invasion. Vinculin, however, is paradoxically described as a prognostic marker favoring the migration of cancer cells or as a tumor suppressor stimulating cell anchorage (Mierke et al, 2010; Hamidi & Ivaska, 2018). These contradictory findings reflect its complex and poorly understood regulation, as well as different roles in 2D or 3D systems (Gulvady et al, 2018).

We therefore tested the effects of IpaA on the motility and invasion of melanoma cells. In time-lapse microscopy experiments in 2D-chambers, GFP-aVBS1-2 inhibited melanocyte motility compared with control cells, with a rate of Root Median Square Displacement (rMSD) of 3.16 and 15.6 μm.min⁻¹, respectively (Fig S6A and B). An even stronger inhibition was observed for GFP-aVBD transfected cells (rMSD = 2.3 μm.min⁻¹) (Fig S6A and B). Transmigration of melanocytes in 3D-matrigels was similarly inhibited by aVBS1-2 and aVBD (Fig S6C).

These results suggest that IpaA-mediated vinculin “canonical” activation induced by aVBS1-2, as well as “supra-activation” mediate by aVBD do not stimulate cell motility but rather strengthen cell adhesion, consistent with their described effects of FAs (Valencia-Gallardo et al, 2023).

Vinculin supra-activation catalyzed by IpaA strengthens cell adhesion

We next investigated the role of vinculin supra-activation by comparing the effects of aVBS1-2 and aVBD in dynamic cell adhesion experiments. First, we performed cell replating experiments, where resuspended transfected cells were allowed to attach to fibronectin-coated surfaces for controlled time periods. As shown in Fig 6A, aVBD induced higher yields of adherent cells when replating was performed with short kinetics, with a fivefold increase over control cells or cells transfected with aVBS1-2 after 10 min replating. By contrast, little difference in adhesion yield was detected between samples after 15 min replating suggesting that aVBD predominantly affected the early dynamics rather than increasing cell adhesion (Fig 6A). To extend these findings, we measured cell adhesion strength using controlled shear stress in a microfluidic chamber (Fig 6B and Video 1). Consistent with replating experiments, when cells were allowed to adhere to fibronectin-coated surfaces for 30 min, little difference in resistance to shear stress could be detected between GFP-aVBD and GFP-transfected cells samples (Fig 6C, small solid circles). However, when shear stress was applied 15 min after cell incubation, GFP-aVBD-transfected cells showed significantly higher resistance to shear stress up to 22.2 dyn.cm⁻² compared with GFP-aVBS1-2 or GFP-transfected cells, with 1.7 ± 0.2- and 0.9 ± 0.14-fold enrichment ± SD of adherent cells for GFP-aVBD and GFP-aVBS1-2 transfected cells, respectively (Fig 6C, large solid circles, Video 1). Consistent with a role for vinculin oligomerization in adhesion, cells transfected with the cysteine-clamped vinculin variant showed adhesion levels under shear stress comparable with cells depleted for vinculin by siRNA treatment that were decreased threefold compared with WT vinculin-transfected cells (Figs 6D and S7). These results are in full agreement with effects observed on adhesion structures and suggest that aVBD-mediated vinculin supra-activation accelerates endogenous processes occurring during mechanotransduction to promote strong adhesion.

Bacterial strains, cells, and plasmids

The bacterial strain used for the purification of the D1D2 construct is E. coli BL21 (DE3) from Invitrogen. E. coli DH5-α F– endA1 glnV44 thi-1 recA1 relA1 gyrA96 deoR nupG purB20 φ80dlacZΔM15 Δ(lacZYA-argF) U169, hsdR17(rK–mK+), λ– was used for the purification of aVBD. The GFP-expressing WT serotype V Shigella strain M90T and isogenic plasmid-cured non-invasive derivative BS176 were described previously (Valencia-Gallardo et al, 2019).

MEF and MEF vinculin null cells (Humphries et al, 2007) were grown in DMEM 1g/liter glucose containing 10% FCS in a 37°C incubator containing 10% CO₂.

The pGEX-4T2-aVBD and the pET15b-D1D2 plasmids were described previously. The mutations in D1D2 were introduced by site-directed mutagenesis using pET15b-D1D2 as a matrix and the primer pairs indicated in Table 2. The pmCherry-N1-human vinculin (HV-mCherry) and was from Addgene. Mutations in D1D2 were transferred in mCherry-HV by exchanging the NheI-PspXI fragment with the corresponding XbaI-PspXI fragment of pET15b-D1D2. The CC mutation corresponds to substitutions of glutamine 69 and Alanine 386 of human vinculin by cysteine residues preventing the formation of IpaA-induced vinculin oligomers but not F-actin binding upon vinculin activation (Valencia-Gallardo et al, 2023).

MEF vinculin null cells (Humphries et al, 2007) were routinely grown in DMEM 1g/liter glucose containing 10% FCS in a 37°C incubator containing 10% CO₂.

The 1205Lu melanoma cell line, a gift from Dr. M. Herlyn (Wistar Institute, Philadelphia, PA), was cultured in RPMI supplemented with 10% FCS as described (Javelaud et al, 2005).

Protein purification

BL21 (DE3) competent E. coli was transformed with the pET15b-D1D2 variant constructs. D1D2 were purified as described (Park et al, 2011b). For the IpaA derivatives, DH5-a competent E. coli was transformed with pGEX-4T2-aVBD. Bacteria were grown at 37°C with shaking until OD_600nm = 1.0 were induced with 1 mM IPTG and incubated for another 2 h. Bacteria were pelleted and washed in ice-cold lysis buffer containing 25 mM Tris PH 7.4, 100 mM NaCl and 1 mM beta-mercaptoethanol, containing complete protease inhibitor. All subsequent steps were performed at 4°C. Bacterial pellets were resuspended in 1/20th of the original culture volume and lyzed using a high pressure homogeneizer (LM20, Microfluidics Corp.). Cell debris were pelleted by centrifugation at 8,000 g for 20 min. Clarified lysates were subjected to affinity chromatography using a GSTrap HP affinity column (GE Healthcare). Briefly, after incubation with the clarified lysates, the column was washed with five column volumes before incubation in PBS containing 100 μg/ml Thrombin (ref 27084601; Cytiva) for 16 h at 21°C. aVBD was then eluted in PBS and further subjected to purification using SEC using a Superdex 200 10/300 GL (Ge Healthcare). Samples were stored aliquoted at −80°C at concentrations ranging from 1 to 10 mg/ml.

SEC analysis

D1D2 and aVBD were at the indicated molar ratio, with D1D2 at a final concentration of 20 μM, for 60 min at 21°C. 200 μl of the protein mixtures were analyzed by SEC on a Superdex 200 10/300 GL (GE Healthcare) using a GE ÄKTA FPLC (Fast Protein Liquid Chromatograph, GMI) and a collection volume of 200 μl per fraction and 20 ml of total collected volume. The SEC buffer was 25 mM Tris–HCl pH 7.2, 100 mM NaCl.

CN-PAGE and densitometry analysis

The D1D2 variants and aVBD were at the indicated molar ratio, with D1D2 at a final concentration of 20 μM, for 60 min at 21°C. Protein complex formation was visualized by PAGE under non-denaturing conditions using à 7.5% polycrylamide gel, followed by Coomassie blue staining, as described previously (Wittig & Schägger, 2005).

Gel raw images were aquired with a ChemiDoc Imaging System (Bio-Rad Laboratories, Inc.) using the Coomassie staining settings. Band intensities corresponding to the monomer, higher order oligomer or intermediate species were measured with Fiji-ImageJ using a fixed rectangular area adjusted with the max width for the monomer and max high order oligomer for the height (Schindelin et al, 2012). A reference area outside the lanes used for the running reactions was included to subtract the background. Values were plotted and then normalized to the value obtained for the corresponding apo construct.

Cell transfection

For transfection experiments, cells were seeded at 1 × 10⁴ cells on 25 mm-diameter coverslips coated with fibronectin at a concentration of 20 μg/ml. Cells were transfected with 1 μg of the pGEX-4T2-D1D2 construct and 4 μls JetPEI transfection reagent (Polyplus) for 16 h following the manufacturer’s recommendations.

Bacterial association and invasion assays

GFP-expressing bacteria were inoculated from pre-cultures at a dilution of 1:100 in trypticase soy broth containing chloramphenicol at 15 μg/ml and grown in a shaking incubator at 37°C until reaching an optical density at 600 nm (OD_{600 nm}) of 0.4. Bacterial cultures were centrifuged at 7,500g for 10 min and the bacterial pellet was resuspended in EM buffer (120 mM NaCl, 7 mM KCl, 1.8 mM CaCl₂, 0.8 mM MgCl₂, 5 mM glucose, and 25 mM Hepes, pH = 7.3) at a final OD_{600 nm} of 0.1. Cells were incubated with the bacterial suspension for 60 min at 37°C before fixing and processing for fluorescence microscopy analysis. Samples were fixed in PBS containing 3.7% paraformaldehyde for 645 min at 21°C, before processing for fluorescence staining of F-actin using Phalloidin-Alexa 488. Extracellular and internalized bacteria were scored using an “inside-out” staining procedure as previously described (Valencia-Gallardo et al, 2023). Briefly, samples were incubated for 45 min at RT with a rabbit anti-LPS antibody diluted 1:5,000 in PBS in the absence of permeabilization, followed by incubation with an anti-rabbit IgG antibody coupled to Alexa Fluor 635 diluted 1:200 to label extracellular bacteria. Samples were then permeabilized by incubation with PBS containing 0.1% Triton X-100 for 4 min, before incubation with Phalloidin-Alexa 350 diluted 1:5,000 to label F-actin. Samples were mounted in DAKO mounting media. Bacterial association with cells was determined by scoring all GFP-expressing bacteria. Bacterial invasion was determined by scoring GFP-expressing bacteria that did not label with Alexa Fluor 650. Bacterial counts were normalized to the cell number.

Fluorescence confocal microscopy analysis

For Shigella invasion assays, samples were analyzed using a Leica DMIRBe microscope equipped with a 63/1.25 HCX PL APO immersion objective, and a CoolSnap HQ2 camera controlled by the Metamorph software 7.7 from Roper Scientific Instruments. For each field, 24–30 Z-planes spaced by 400 nm were acquired in stream mode for each wavelength. The bacterial association rates were determined by scoring GFP-expressing bacteria in contact with cells. The bacterial invasion rates were determined by scoring GFP-expressing bacteria that were not labeled with the anti-LPS antibody. For focal adhesion, samples were analyzed using an Eclipse Ti inverted microscope (Nikon) equipped with a 60 x objective APO TIRF oil immersion (NA: 1.49), a CSU-X1 spinning disk confocal head (Yokogawa), and a Prime 95B CMOS camera (Photometrics) controlled by the Metamorph 7.7 software.

Image analysis

Focal adhesions were analyzed using the ImageJ 2.1.0/1.53c software. Cells expressing similar levels of average fluorescence intensity of the HV-mCherry constructs were analyzed. For each set of experiments, the confocal plane corresponding to the basal plane was subjected to thresholding using strictly identical parameters between samples. Adhesion clusters were detected using the “Analyze particle” plug-in, setting a minimal size of 3.5 μm².

Statistical analysis

The number of adhesions was analyzed using Dunn’s multiple comparisons test. The median area was compared using Mann-Whitney test. Differences in the rates of D1D2 trimer formation and monomer disappearance based were analyzed using an ANCOVA test.

Live cell tracking

1205Lu melanoma cells were transfected with IpaA constructs or GFP alone (control) and transferred in microscopy chamber on a 37°C 5%-CO₂ stage in RPMI1640 medium, containing 25 mM Hepes. For cell tracking, samples were analyzed using and inverted Leica DMIRBe microscope and a 20 X phase contrast objective. Image acquisitions were performed every 3 min for 200 h. The mean velocity of migration was measured for all tracks followed for at least 5 h. The root square of MdSD over time was plotted over time and fitted by linear regression. The slopes of the linear fit were compared using an ANCOVA test (linear model). The median cell surface was quantified as the mean of the surface for three time points (25%, 50%, and 75%) of the whole cell track and dispersion measured by the median absolute dispersion.

Cell invasion assays

Tissue culture Transwell inserts (8 μm pore size; Falcon) were coated for 3 h with 10 μg of Matrigel following the manufacturer’s instructions (Biocoat, BD Biosciences). Inserts were placed into 24-well dishes containing 500 μl of RPMI medium supplemented with 1% FCS. 5 × 10⁴ melanoma cells were added to the upper chamber in 250 μls of serum-free RPMI medium. After 24 h, transmigrated cells were scored by bright field microscopy. Experiments were performed at least three times, each with duplicate samples.

Microfluidics cell adhesion assay

Analysis of cell detachment under shear stress was based on previous works (Ghannam et al, 2011). 1205Lu melanocytes were transfected with the indicated constructs, then labeled with 2 μM calcein-AM (Life Technologies) in serum-free DMEM for 20 min. Cells were detached by incubation with 2 μM Cytochalasin D (Sigma-Aldrich) for 40 min to disassemble FAs, followed by incubation in PBS containing 10 mM EDTA for 20 min. Cells were washed in EM buffer (120 mM NaCl, 7 mM KCl, 1.8 mM CaCl₂, 0.8 mM MgCl₂, 5 mM glucose and 25 mM Hepes at pH 7.3) by centrifugation and resuspended in the same buffer at a density of 1.5 × 10⁶ cells/ml. Calcein-labeled transfected cells and control unlabeled cells were mixed at a 1:1 ratio and perfused onto a 25 mm-diameter glass coverslips (Marienfeld) previously coated with 20 μg/ml fibronectin and blocked with PBS containing 2% BSA (Sigma-Aldrich) in a microfluidic chamber on a microscope stage at 37°C. We used a commercial microfluidic setup (Flow chamber system 1C, Provitro) and a Miniplus3 peristaltic pump (Gilson) to adjust the flow rate in the chamber. Microscopy analysis was performed using a LEICA DMRIBe inverted microscope equipped with a Cascade 512B camera and LED source lights (Roper Instruments), driven by the Metamorph 7.7 software (Universal imaging). Cells were allowed to settle for the indicated time before application of a 4 ml/min, flow corresponding to a wall shear stress of 22.2 dyn/cm² (2.22 Pa). Acquisition was performed using a 20 X objective using phase contrast and fluorescence illumination (excitation 480 ± 20 nm, emission 527 ± 30 nm). Fluorescent images were acquired before and after flushing to differentiate between target and control cells. Phase contrast images were acquired every 200 ms. Fold enrichment was defined as the ratio between of attached labeled and unlabeled cells.