We measured binding kinetics of intact α4β1 on Jurkat cells to two fluorescently labeled ligands, a phenylureide derivative of Leu-Asp-Val-Pro (FITC-LDVP) and a fragment of vascular cell adhesion molecule (VCAM) containing its first two domains (Alexa488-VCAM D1D2) (Figure 2). Before adding ligands, cells were equilibrated with saturating concentrations of Fabs for 30 min at 22°C to stabilize specific conformational states (Li and Springer, 2018). Integrin extension, that is, the EC and EO states, was stabilized with 8µM 9EG7 Fab, which binds to the β1-subunit knee (Figure 2B). The EO conformation was stabilized with a combination of 4 µM 9EG7 Fab and 2 µM HUTS4 Fab; the latter binds near the interface between the βI andhybrid domains and stabilizes the EO conformation (Figure 2C). Ligand-binding was monitored as mean fluorescence intensity (MFI) by flow cytometry without washing (Figure 2). Beginning at about 10 min, a 500-fold higher concentration of unlabeled ligand was added to measure the kinetics of dissociation. Background MFI at each fluorescent ligand concentration, measured under identical conditions except in the presence of 10 mM EDTA, showed no significant difference at different time points during the association and dissociation measurements (Figure 2—figure supplement 1) and was averaged across different time points and subtracted to obtain specific binding.

Figure 2 with 2 supplements see all

Download asset Open asset

Figure 2—source data 1

Source data for Figure 2A.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig2-data1-v2.txt

Download elife-73359-fig2-data1-v2.txt

Figure 2—source data 2

Source data for Figure 2B.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig2-data2-v2.txt

Download elife-73359-fig2-data2-v2.txt

Figure 2—source data 3

Source data for Figure 2C.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig2-data3-v2.txt

Download elife-73359-fig2-data3-v2.txt

Figure 2—source data 4

Source data for Figure 2D.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig2-data4-v2.txt

Download elife-73359-fig2-data4-v2.txt

Figure 2—source data 5

Source data for Figure 2E.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig2-data5-v2.txt

Download elife-73359-fig2-data5-v2.txt

Under basal condition with all three integrin states present in the ensemble, binding of FITC-LDVP to Jurkat cells reached equilibrium within 3 min (Figure 2A). Upon addition of a 500-fold excess of LDVP, dissociation of FITC-LDVP was rapid and was 99.7% complete by 5 min (Figure 2A). In contrast, both binding and dissociation of FITC-LDVP were slower when only the extended conformations (EC and EO) were present on Jurkat cells (Figure 2B). Reaching steady state required ~5 min after addition of 20 nM FITC-LDVP, ~10 min with 10 nM ligand, and was not reached after 10 min with 5 nM ligand. After 10 min of dissociation, only 19.4% of ligand had dissociated (Figure 2B). Association and dissociation were even slower when only the EO conformation was present (Figure 2C). After 15 min of association, much less ligand had bound (Figure 2C) than when both EC and EO conformations were present (Figure 2B). Dissociation was also slower, with only 1.2% of bound ligand dissociating after 10 min (Figure 2C).

VCAM D1D2 binds with ~100-fold lower affinity than LDVP to α4β1 (Figure 5 in Li and Springer, 2018). As a result, binding to the basal ensemble was too low to measure over the noise from unbound ligand; however, we were able to measure binding kinetics to intact α4β1 stabilized in the extended (EC + EO) and EO states (Figure 2D and E). When the two extended conformations (EC and EO) were present, binding of all three concentrations of Alexa488-VCAM D1D2 (10 nM, 20 nM, and 30 nM) reached equilibria within 2 min. Upon addition of a large excess of LDVP, dissociation of Alexa488-VCAM D1D2 was also fast; 100% dissociated by 5 min (Figure 2D). Association and dissociation both became markedly slower when only the EO conformation of α4β1 was present (Figure 2E).

To address the generality of these results, we studied another integrin and cell type by measuring binding of a fluorescently labeled two-domain fragment of fibronectin (Alexa488-Fn3_9-10) to intact α5β1 integrin on K562 cells (Figure 3). The BC conformation of α5β1 integrin on K562 cells is more stable than that of α4β1 integrin on Jurkat cells (Li and Springer, 2018). Therefore, to assure that the extended states (EC + EO) were saturably populated, they were stabilized with a combination of two Fabs, 6 µM 9EG7 Fab and 2 µM SNAKA51 Fab (Figure 3A, schematic). The EO state of α5β1 was stabilized with the same combination of Fabs as used for α4β1 (Figure 3B, schematic). Although binding affinity was too low to measure kinetics of the basal ensemble (Li et al., 2017), we were able to measure Alexa488-Fn3_9-10 kinetics with the EC + EO and EO ensembles of intact α5β1 (Figure 3). When α5β1 was stabilized in the EO conformation, Alexa488-Fn3_9-10 bound and dissociated significantly more slowly than when both the EC and EO states of α5β1 were present in the ensemble (Figure 3A and B). Faster binding and dissociation of Alexa488-Fn3_9-10 from the EC + EO ensemble than EO showed that the EC state of α5β1 binds and dissociates faster than the EO state, just as found for α4β1.

Figure 3

Download asset Open asset

Figure 3—source data 1

Source data for Figure 3A.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig3-data1-v2.txt

Download elife-73359-fig3-data1-v2.txt

Figure 3—source data 2

Source data for Figure 3B.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig3-data2-v2.txt

Download elife-73359-fig3-data2-v2.txt

To quantify the binding kinetics of intact α4β1 and α5β1 under each condition, we globally fit specific binding in both association and dissociation phases at each concentration of fluorescently labeled ligand to the 1 vs. 1 Langmuir binding model (lines in Figure 2A-E and Figure 3A, B). Apparent on- and off-rates, ${\displaystyle k_{\mathrm{o}\mathrm{n}}^{\mathrm{a}\mathrm{p}\mathrm{p}}}$ and ${\displaystyle k_{\mathrm{o}\mathrm{f}\mathrm{f}}^{\mathrm{a}\mathrm{p}\mathrm{p}}}$, were well fit, with low fitting errors relative to the measured values (Figure 2F and Figure 3 C). The ratio of the apparent off- and on-rates, ${\displaystyle k_{\mathrm{o}\mathrm{f}\mathrm{f}}^{\mathrm{a}\mathrm{p}\mathrm{p}}/k_{\mathrm{o}\mathrm{n}}^{\mathrm{a}\mathrm{p}\mathrm{p}}}$ , agrees reasonably well with the previously determined equilibrium dissociation constants shown as K_d in Figure 2F and Figure 3C (Li and Springer, 2018; Li et al., 2017). These agreements suggest that the 1 vs. 1 Langmuir binding model can reasonably fit the kinetic data; that is, the kinetics of conformational change between states is sufficiently fast relative to the kinetics of ligand binding to not perturb ligand-binding kinetics. Thus, the relative populations of unbound states in the ensemble were not perturbed by removal by binding to ligand.

Overall, these results show that ligand binds to and dissociates from the EO conformation more slowly than from the BC and EC conformations. The kinetics measured here for the basal and EC + EO ensembles are apparent because they include contributions from distinct conformational states present in these ensembles. In contrast, EO state kinetics are measured exactly because EO is the only state present in the EO ensemble. In the final section of ‘Results,’ we will use previous measurements of the populations of the states in each ensemble to calculate the on- and off-rates for conformations within mixtures of states.

We utilized biolayer interferometry (BLI) (Wallner et al., 2013) to measure the kinetics of binding of an ectodomain fragment of α5β1 to the biotin-labeled Fn3_9-10 fragment of fibronectin immobilized on streptavidin biosensors (Figure 4). The ectodomain was truncated just prior to the transmembrane domains of the α5 and β1 subunits and was expressed in a cell line containing a glycan processing mutation so that it had high-mannose rather than complex-type N-glycans. Truncation of α5β1 and high mannose glycoforms substantially raises the basal population of the EO relative to that of the EO state of intact α5β1 on cell surfaces ( Figures 4 and 7E in Li et al., 2017), enabling measurement here of basal ensemble Fn3_9-10 binding kinetics.

Figure 4

Download asset Open asset

Figure 4—source data 1

Source data for Figure 4A.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig4-data1-v2.txt

Download elife-73359-fig4-data1-v2.txt

Figure 4—source data 2

Source data for Figure 4B.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig4-data2-v2.txt

Download elife-73359-fig4-data2-v2.txt

Figure 4—source data 3

Source data for Figure 4C.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig4-data3-v2.txt

Download elife-73359-fig4-data3-v2.txt

Figure 4—source data 4

Source data for Figure 4D.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig4-data4-v2.txt

Download elife-73359-fig4-data4-v2.txt

Binding kinetics were measured by transferring Fn3_9-10 biosensors to wells containing the α5β1 ectodomain in the absence or presence of conformation-stabilizing Fabs. Dissociation kinetics were measured by transfer of sensors to wells lacking the integrin but containing identical Fab concentrations (Figure 4A–D). Equilibrium K_d values were previously shown to be independent of the Fab used to stabilize a particular state (Figure 3 in Li et al., 2017). However, we were concerned that binding of Fabs, particularly those that bind near ligand-binding sites, might slow kinetics and therefore tested this by varying the Fabs used to stabilize the EO state.

The kinetic curves showed that the α5β1 ectodomain EO state associated more slowly than the mixtures with the closed states and also dissociated more slowly (Figure 4A-D,F). Overall, these differences among ensembles resembled those found for the EC + EO ensemble and EO state of intact α5β1 on K562 cells and extended measurements to the basal α5β1 ensemble. The on- and off-rates of the EO state for Fn3_9-10 determined in the presence of 12G10 Fab were fourfold and twofold lower, respectively, than those determined in the presence of 9EG7&HUTS4 Fabs (Figure 4C, D and F). As 12G10 Fab binds close to the ligand-binding site in the β1 domain (Figure 1A), we use ${\displaystyle k_{\text{off}}^{\text{app}}}$ and ${\displaystyle k_{\text{on}}^{\text{app}}}$ kinetics determined with the 9EG7, 8E3, 9EG7&HUTS4 Fabs, which bind far from the ligand-binding site, for calculating true (${\displaystyle k_{\text{off}}}$ and ${\displaystyle k_{\text{on}}^{}}$) kinetic rates for each state in the final section of ‘Results.’

Due to the low affinities of the closed states there was too little binding to directly measure k_on or k_off in the presence of saturating closure-stabilizing Fabs. We therefore used another approach. We first allowed ligand binding to integrins to reach steady state in the absence of a closure-stabilizing Fab. We then added different concentrations of closure-stabilizing Fab mAb13 and measured dissociation kinetics (Figure 5 and Figure 6). Dissociation of the ligand from the EO state is very slow as shown above and is negligible in our experimental time scale. At high Fab mAb13 concentrations, when the EO ligand-bound state (EO•L) converts to either BC•L or EC•L (grouped together here as C•L), mAb13 Fab binds and prevents back-conversion to EO•L (Figure 5A and B). After saturating concentrations of Fab mAb13 are added to basal or EO+ EC ensembles pre-equilibrated with ligand, the effective off-rate is contributed by two steps, the conformational change from EO•L to C•L and the dissociation of ligand from mAb13-bound C•L (mAb13•C•L) (Figure 5A and B). Thus, the observed off-rate at saturating concentration of mAb13 Fab is contributed by the rates of both steps and permits the determination of the lower limit of ${\displaystyle k_{\text{off}}^{\text{C}}}$ .

Figure 5

Download asset Open asset

Figure 5—source data 1

Source data for Figure 5A.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig5-data1-v2.txt

Download elife-73359-fig5-data1-v2.txt

Figure 5—source data 2

Source data for Figure 5B.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig5-data2-v2.txt

Download elife-73359-fig5-data2-v2.txt

Figure 6

Download asset Open asset

Figure 6—source data 1

Source data for Figure 6A.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig6-data1-v2.txt

Download elife-73359-fig6-data1-v2.txt

Figure 6—source data 2

Source data for Figure 6B.

https://cdn.elifesciences.org/articles/73359/elife-73359-fig6-data2-v2.txt

Download elife-73359-fig6-data2-v2.txt

We measured FITC-LDVP dissociation from basal or extended ensembles of α4β1 on Jurkat cells after addition of a range of mAb13 Fab concentrations (Figure 5A and B). Saturable binding of mAb13 Fab to nascent cell surface C•L was evident from the approach to a plateau of ${\displaystyle k_{\text{off}}^{\text{max}}}$ (Figure 5A–C). The ${\displaystyle k_{\text{off}}^{\text{max}}}$ values measured for LDVP dissociation from basal and extended α4β1 ensembles on Jurkat cells were similar and within error of one another, with an average of ~120 * 10^–3 /s (Figure 5C).

Similarly, we measured Fn3_9-10 dissociation from basal or extended ensembles of the α₅β₁ ectodomain (Figure 6). The effect of mAb13 Fab on increasing k_off was saturable, as shown by approach to a plateau (Figure 6A–C). The fit to a saturation dose-response curve yielded ${\displaystyle k_{\text{off}}^{\text{max}}}$ values for the basal and extended ensembles of (1600 ± 100) * 10^–3 /s and (1900 ± 100) * 10^–3/s, respectively (Figure 6C).

We directly measured the ligand-binding and dissociation kinetics for the EO state of α4β1 and α5β1 (Figure 2C,E, Figure 3B, Figure 4C ). In contrast, kinetics for the BC and EC states were only measured within ensembles. Their kinetics are convoluted in two respects. First, measurements on ensembles contain kinetics contributed by all states within the ensemble. Second, apparent association and dissociation kinetics may each contain a contribution from the kinetics of conformational change (Figure 1B). Figure 1B (left) shows apparent on- and off-rates, and Figure 1B (right) shows all the actual pathways by which ligand binding and dissociation can occur, which include all known integrin conformational states and the kinetics of conformational change between them. Furthermore, after ligand binding to the closed states, rapid conformational change to the EO state occurs and is responsible for our ability to measure the kinetics of binding as a result of accumulation of ligand-bound integrin in the EO state.

The underlying assumption for deconvoluting the kinetics of the closed states is that if integrin conformational transition kinetics (BC ⇌ EO, EC ⇌ EO, BC ⇌ EC, BC•L ⇌ EO•L, EC•L ⇌ EO•L, BC•L ⇌ EC•L) are sufficiently fast so that the populations of the three integrin states do not deviate significantly during our experiments from the equilibrium values of the populations, then measured kinetics will not be significantly limited by conformational transition kinetics. In this case, both free integrins and ligand-bound integrins can be considered as readily equilibrated among their conformational states, and ligand binding coupled with integrin conformational changes can be approximated by the apparent 1 vs. 1 reaction between integrin and ligand (this allows the double tildes in Eqs. 1–4 in Figure 7A to be treated as equal signs). All on- and off-rates measured here were well fit with the 1 vs. 1 Langmuir binding model (Figure 2A-E, Figure 3A,B, Figure 4A,B,C,D, Figure 5A, B, Figure 6A, B), supporting this assumption. Moreover, reasonable agreement between the ratios of the apparent off- and on-rates, ${\displaystyle k_{\text{off}}^{\text{app}}/k_{\text{on}}^{\text{app}}}$ , and previously determined equilibrium dissociation constants, K_d, (Figure 2F, Figure 3C, Figure 4F), validates the assumption that the apparent on- and off-rates (${\displaystyle k_{\text{on}}^{\text{app}}}$ and ${\displaystyle k_{\text{off}}^{\text{app}}}$) for each defined ensemble can be approximated by the on- and off-rates of each state weighted by its population in the ensemble (Figure 7A, Eqs. 1–4). The population of the integrin states in the absence of ligand (BC, EC, and EO) and in the presence of saturating concentrations of ligand (BC•L, EC•L, and EO•L) was calculated based on the previously determined population and ligand-binding affinity of each state (Figure 7—figure supplement 1B, Eqs. S5–S10) in the respective integrin α4β1 and α5β1 preparations (Li and Springer, 2018; Li et al., 2017) and are shown in Figure 7B.

Figure 7 with 1 supplement see all

Download asset Open asset

On- and off-rates for each α4β1 and α5β1 integrin state on intact cells and for the purified α5β1 ectodomain are summarized in Figure 7C. Values are best determined, that is, with the lowest errors, for the directly determined EO state on- and off-rates. Errors were higher for the BC and EC states, particularly for k_off. Therefore, k_off values for each state were also calculated from ${\displaystyle k_{\text{off}}=K_{\text{d}}\ast k_{\text{on}}}$ (Figure 7C), where ${\displaystyle K_{\text{d}}}$ is from equilibrium measurements (Li and Springer, 2018; Li et al., 2017). The ${\displaystyle k_{\text{off}}}$ values of each state determined from these two strategies agree well with one another for each integrin-ligand pair. We do not report in Figure 7C the lower limit of ${\displaystyle k_{\text{off}}^{\text{C}}}$ and ${\displaystyle k_{\text{off}}^{\text{EC}}}$ approached by measuring dissociation in the presence of a closure-stabilizing Fab in Figure 5 and Figure 6; however, these values were comparable to those calculated from measurements on ensembles as described in this section or calculated from ${\displaystyle k_{\text{off}}=K_{\text{d}}\ast k_{\text{on}}}$ .

Employing conformation-specific Fabs against the integrin β1 subunit to stabilize integrins into defined ensembles, we determined the on- and off-rates of each integrin conformational state. The conformational specificities of these Fabs were determined with negative stain EM (nsEM) with the integrin α5β1 ectodomain (Su et al., 2016). All antibodies used here were 100% conformation-specific; this conformation was present in all EM class averages of identifiable integrin conformation with bound Fab. Specificity was further verified by ligand-binding affinity measurements in the presence of saturating concentrations of Fab (Li et al., 2017). In the latter study, we used between 2 and 4 independent antibodies or antibody combinations, often binding to distinct domains, to stabilize α5β1 ectodomain ensembles containing the EO, EC, EO + EC, and EC + BC states. In all cases, independent antibodies that stabilized the same conformational state(s) as determined by EM yielded similar ensemble K_d values (Figure 3 in Li and Springer, 2018; Li et al., 2017), validating stabilization of the same state(s), and supporting the assumption that these states resembled those in the absence of Fab. Quantitatively, the antibodies must be highly state-specific in order to give large shifts in affinities, to give consistent intrinsic affinities on constructs with large differences in basal affinities, and to give similar intrinsic affinities using Fabs to distinct epitopes. The conformational specificity of these antibodies and their use at sufficient concentration to saturate these states are important for our previous thermodynamic studies ( Li and Springer, 2018; Li et al., 2017) and the current kinetic study.

It is common in cryoEM and nsEM to form complexes in a large excess of one or more components, and then separate the complex from free components by gel filtration. During gel filtration and up until cryo-cooling or fixation by negative stain, complexes will dissociate because the law of mass action requires co-existing free components to stabilize complex formation, as defined by [I•L] = ([I].[L])/K_d where I and L are integrin and ligand or Fab, respectively. Stabilizing integrins ≥ 99% in the desired conformations in this study required the use of Fabs at concentrations 80-fold higher than their EC50 values for each integrin preparation (Table S1 in Li and Springer, 2018; Li et al., 2017). For stabilizing intact α4β1 and α5β1 in extended and open states, some Fab EC50 values were in the hundreds of nM range, which required a combination of two compatible Fabs to conserve protein supplies. Compared to the integrin that was being stabilized here, Fabs were maintained throughout our experiments at molar excesses ranging from 9- to 48,000-fold over the integrin. This contrasts to the 1:1 ratio in integrin complexes isolated from free components for ns and cryoEM. Such complexes are out of equilibrium and must dissociate, the rate and extent of which are governed by the kinetics and equilibria such as measured here and previously, respectively (Li and Springer, 2018; Li et al., 2017).

To prevent confusion, we should discuss a recent elegant cryoEM structure of intact integrin α5β1 in nanodiscs bound both to TS2/16 Fv-clasp and Fn3_7-10 that showed for the first time that α5β1 recognizes both the RGD motif in Fn3 domain 10 and the synergy site in Fn3 domain 9 of fibronectin (Schumacher et al., 2021). Conformational equilibria of the α5β1 ectodomain determined with Fn3_9-10 and RGD peptides as ligands were similar (Figure 4 in Li et al., 2017), which makes it important to note that the claim that RGD, because it lacks the synergy site, cannot stabilize α5β1 headpiece opening, was not supported by cited articles (Schumacher et al., 2021). Addition of 1 mM RGD peptide to the α5β1 headpiece resulted in the presence of the headpiece in both the open and closed conformations as shown by nsEM (Figure 2B in Takagi et al., 2003). At saturating concentrations of ligands, integrin preparations vary in the percentage of ligand that is bound to each state (Figure 7B; see Figure 7—figure supplement 1 for a wide range of integrin preparations). The α5β1 headpiece used in Takagi et al., 2003 is ~50% open when saturably bound to a cyclic RGD peptide, in agreement with the nsEM results. The failure to open the closed α5β1 headpiece in crystals soaked with RGD was to be expected because the lattice contacts frustrated opening and in one publication a closed state-specific Fab was also bound (Nagae et al., 2012; Xia and Springer, 2014). In recent unpublished work, we found that while Mn²⁺ increases population of the EO state of cell surface α5β1, the large majority remains in the BC state (Anderson, Li, and Springer), in excellent agreement with the predominance of the BC state in Mn²⁺ in α5β1 in nanodiscs (Fig. 6 in Schumacher et al., 2021). Additionally, while α5β1 is less bent than some other integrins (Schumacher et al., 2021; Su et al., 2016), it still has distinct bent and extended conformations and it thus remains appropriate to describe α5β1 as having a BC state.

The advance in this study is the measurement of ligand binding and dissociation kinetics of integrin conformational states for the first time. Furthermore, we have made the surprising observation that the low-affinity integrin BC and EC states have faster on-rates than the high-affinity EO state. While we compare ${\displaystyle k_{\text{off}}\text{/}{k}_{\text{on}}}$ to previously determined equilibrium ${\displaystyle K_{\text{d}}}$ values, that is only to demonstrate that the values we have determined are not affected by slow equilibration among the conformational states, which would have invalidated our assumption that the apparent ligand-binding kinetics are the composites of the ligand-binding kinetics of each state at equilibrium.

Previous studies on α4β1 and α5β1 integrins showed that their affinities were intrinsic to the type of integrin, the ligand, and the state; that is, independent of the type of integrin preparation: cell surface, soluble ectodomain or headpiece, or N-glycosylation status (Li and Springer, 2018; Li et al., 2017). Similar to intrinsic affinities, the results here on ligand-binding kinetics were consistent with on-rates and off-rates that are intrinsic to integrin conformational states. On-rates for Fn3_9-10 binding to EO states of intact α5β1 on cell surfaces and the α5β1 ectodomain were identical and off-rates differed by only 1.8-fold. Similarly, on- and off-rates for the EC state of the intact cell-surface and ectodomain forms of α5β1 differed only by 1.1-fold and 1.9-fold, respectively. Moreover, BC and EC states for both types of integrins had similar kinetics, showing that ligand-binding on- and off-rates were determined by whether the headpiece was closed or open, and not by extension. These findings are in agreement with the essentially identical intrinsic affinities of the two integrin α5β1 closed states (Li and Springer, 2018; Li et al., 2017). In further agreement, crystal structures of the integrin αIIbβ3 ectodomain in the BC state and of the αIIbβ3 closed headpiece fragment, which has no interactions with the lower legs and thus serves as a model for the EC conformation (Zhu et al., 2008; Zhu et al., 2013), show essentially identical conformations of the ligand-binding site.

We checked whether kinetics might be influenced by bound Fabs, in contrast to equilibria, which showed no dependence (Li et al., 2017). Among two Fabs used to stabilize the EO state, ligand association and dissociation was slower with 12G10, which binds near the ligand-binding site in the βI domain than with HUTS4, which binds distally in the hybrid domain (Figure 4F). As Fabs generally decrease dynamic protein motions in their epitopes (Wei et al., 2014) and may also sterically slow binding, the kinetics measured using HUTS4 Fab more likely approximate integrin kinetics in the absence of Fab and are reported in Figure 7C.

The kinetics of the EC and BC states were calculated from measurements on extended or basal ensembles after correction for the kinetics in these ensembles contributed by the EO state. As a check on these measurements, we also measured k_off in the presence of mAb13 Fab, which after conformational conversion of EO•L to EC•L + BC•L trapped the closed states so that their dissociation could be measured. The lower limit of ${\displaystyle k_{\text{off}}^{\text{C}}}$ and ${\displaystyle k_{\text{off}}^{\text{EC}}}$ determined from these experiments (Figures 5C and 6C) is in good agreement with off-rates of the closed states calculated from ensembles and also from K_d•k_on (Figure 7C).

Typical protein-protein on-rates as found for antibody-antigen interactions are in the range of 10⁵–10⁶ M^–1 s^–1 (Alsallaq and Zhou, 2008). The on-rates for the BC and EC states were in this range. In contrast, the on-rates for EO states for the corresponding integrin-ligand pairs were slower and suggest a hindrance to ligand binding. The open conformation of integrins has a tighter ligand-binding pocket around the metal ion-dependent adhesion site (MIDAS) than the closed state (Dong et al., 2017; Nagae et al., 2012; Schumacher et al., 2021; Xia and Springer, 2014; Xiao et al., 2004; Zhu et al., 2013; Figure 7D). Movement of the β1-α1 loop toward the ligand and the MIDAS Mg²⁺ ion upon βI domain opening partially buries the Mg²⁺ ion and is expected to slow binding of the Asp sidechain, which must fit into a tight pocket with a specific geometry dictated by partially covalent and highly directional Asp sidechain metal coordination and hydrogen bonds to the β1-α1 loop backbone amide nitrogens.

The ~1000-fold higher affinities of the EO than the closed conformations for both α4β1 and α5β1 integrins are achieved by the ~25,000-fold slower off-rate of the EO conformation (Figure 7C). Similar to the differences in on-rates, the differences in off-rates can be understood in terms of the structural details in the ligand-binding pocket and the much higher affinity of the EO state. The tighter Asp sidechain binding pocket and greater burial of the Asp provide a barrier to dissociation (Figure 7D). The number of hydrogen bonds of the Asp sidechain to the β1-α1 loop backbone increases in the open state (Zhu et al., 2013). Furthermore, the greater burial of these polar bonds and increased network of hydrogen bonds around them increase their strength.

The intrinsic ligand-binding kinetics of integrin conformational states described here are consistent with previous kinetic observations. These studies showed that activating integrin ensembles with Mn²⁺ or activating IgG or Fab, using conditions that in retrospect would partially but not completely shift integrin ensembles to the EO state, decreased the ligand off-rates of integrins α4β1 and α5β1 (Chigaev et al., 2001; Takagi et al., 2003). The extremely long lifetime of the α5β1 complex with fibronectin in the EO state, around several hours, also explains why the α5β1 complex with fibronectin in Mn²⁺ was much more rapidly reversed by mAb 13 IgG specific for the closed conformations than by competitive inhibitor (Mould et al., 2016; Mould et al., 2014). One important difference from most previous work, including our own (Takagi et al., 2003), is our ability to fit all of our data to a 1:1 binding model. All our fits were global; that is, data at all ligand concentrations were used to create the fits in each panel shown in figures. This is the most rigorous way to fit data. When more variables are introduced such as a rate for conformational change to a different state or additional binding and dissociation rates, better fits can always be achieved. Notably, our data fit the 1:1 binding model whether one, two, or three states were present in the ensemble. Therefore, we believe that the rates of interconversion are relatively fast and attribute the difference from some previous results to homogenous protein preparations free from aggregates and homogenous incorporation of binding partners either as native proteins on intact cells or as purified proteins immobilized uniformly through single, site-specific biotins to streptavidin biosensors.

A major impetus for these studies was to determine the pathway for activation of integrins in cells, that is, the activation trajectory. Of key importance is how integrins on the cell surface first engage ligands. Integrin signaling is governed by cytoskeletal force and the force-stabilized, high-affinity, EO conformation is the only state competent to mediate cell adhesion (Alon and Dustin, 2007; Astrof et al., 2006; Li and Springer, 2018; Li et al., 2017; Nordenfelt et al., 2016; Nordenfelt et al., 2017; Sun et al., 2019; Zhu et al., 2008). Mechanotransduction occurs when an integrin binds to ligand anchored in the extracellular environment and at the same time its cytoplasmic domain binds to a cytoskeletal adaptor and links to actin retrograde flow, resulting in a tensile force transmitted through the integrin that stabilizes the EO conformation over the BC conformation. Thus, the on- and off-rates of ligand binding to integrins are among the key parameters that determine the efficiency of cytoskeletal force regulation. We found that the closed states, with loose ligand-binding pockets, have higher on-rates for ligand binding, making them the most efficient state for encountering ligand.

What is the cell biological context in which we should think about our finding that the closed states are kinetically poised to bind ligand as the first extracellular event in integrin activation? And what is integrin inside-out signaling? This term was first used in a review on adhesion molecules of the immune system (Springer, 1990) after the discovery that integrins on lymphocytes (Dustin and Springer, 1989) and platelets (reviewed in Hynes, 1992) only became active in adhesion after these cells were stimulated through other receptors. The commonality is that lymphocytes and platelets in the bloodstream do not exhibit polarization or actin retrograde flow, but do so after stimulation, consistent with the model that integrins are activated by the tensile force transmitted through them between actin retrograde flow and extracellularly embedded ligands (Li and Springer, 2017) and findings that integrins are aligned with actin retrograde flow (Nordenfelt et al., 2017; Swaminathan et al., 2017) and transmit tensile force (Nordenfelt et al., 2016; Sun et al., 2016). Our results on the kinetics of ligand binding of different integrin conformational states are relevant to both integrins on cells in suspension as studied here and to unbound integrins on adherent cells since the kinetics are intrinsic to the EO, EC, and BC states.

Our results on ligand-binding kinetics provide a stepping stone to a remaining important question, the kinetics of integrin conformational change. Equilibria show that the ligand-bound open state is ~1000-fold more favored than the ligand-bound closed state. Thus, following step 1 of binding of ligand to the closed state, step 2 of conformational change to the more stable ligand-bound EO state should be rapid, but remains to be measured.

It should be emphasized that the populations of integrin conformational states we have measured on cells in suspension are for unliganded integrins. On adherent cells, integrins bound to ligand on the substrate belong to a separate pool of liganded integrins, which are removed from and not included in the populations of unliganded integrins we have measured. Ligand-bound integrins may dominate on the ventral surface of adherent cells, as emphasized by the predominance of extended integrin LFA-1 on the ventral surface of lymphoid cells bound to the LFA-1 ligand ICAM1, and the greater abundance of bent LFA-1 on the ventral surface of the same cells bound to the α4β1 ligand fibronectin (Moore et al., 2018).

Are the populations of unliganded integrin conformational states the same on suspended and adherent cells? We do not know the kinetics of binding of adaptors such as talin or kindlin to integrins or what fraction of unliganded integrins on suspended and adherent cells are bound to adaptors. The adaptor talin appears to interfere with association between integrin α and β subunit transmembrane/cytoplasmic domains in the BC state (Lau et al., 2009; Zhu et al., 2009), and thus if talin is bound to a higher proportion of unliganded integrins in adherent cells than suspended cells, the populations of unliganded EC and EO states might be increased relative to the BC state. Adaptor-bound unliganded integrins might also show preferential localizations in cells. Extended and unliganded β1 integrins have been shown to localize along the leading edge of fibroblast lamellae and growth cone filopodia to probe for adhesion sites (Galbraith et al., 2007). High-affinity, unliganded integrin αVβ3 was also shown to be recruited to lamellipodia in endothelial cell migration (Kiosses et al., 2001). Rapid ligand binding, together with rapid cytoskeletal adaptor binding, would enable their coincidence to regulate integrin activation, thus providing a seamless method for activating integrins at cellular locations where actin, talin, and kindlin are activated and extracellular ligand is available.

IgGs of 8E3 (Mould et al., 2005), 9EG7 (Bazzoni et al., 1995), 12G10 (Mould et al., 1995), HUTS4 (Luque et al., 1996), mAb13 (Akiyama et al., 1989), and SNAKA51 (Clark et al., 2005) were produced from hybridomas and purified by protein G affinity; Fabs were prepared with papain digestion in phosphate-buffered saline (PBS with 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, and 1.8 mM KH₂PO₄, pH7.4) with 10 mM EDTA and 10 mM cysteine and papain: IgG mass ratio of 1:500 for 8 hr at 37°C, followed by Hi-Trap Q chromatography in Tris-HCl pH 9 with a gradient in the same buffer to 0.5 M NaCl. These conformation-specific Fabs are used to stabilize integrin α4β1 and α5β1 into ensembles of defined conformational states. To ensure saturable population of target conformations, conformation-specific Fabs were used at concentrations (shown in the figure legends) well above the concentration giving half-maximum responses, that is, their EC50 values on each integrin preparation (Table S1 in Li and Springer, 2018; Li et al., 2017).

Integrin α5β1 ectodomain (α5 F1 to Y954 and β1 Q1 to D708) with secretion peptide, purification tags, and C-terminal clasp (Takagi et al., 2001) were produced by co-transfecting the pcDNA3.1/Hygro(-) vector coding the α-subunit and pIRES vector coding the β-subunit into HEK 293S GnTI^−/− (N-acetylglucosaminyl transferase I-deficient) cells. Stable transfectants were selected with hygromycin (100 μg/ml) and G418 (1 mg/ml), and proteins were purified from culture supernatants by His tag affinity chromatography and Superdex S200 gel filtration after cleavage of C-terminal clasp and purification tags with TEV protease (Li et al., 2017).

FITC-labeled α4β1-specific probe, 4-((N′-2-methylphenyl)ureido)-phenylacetyl-L-leucyl-L-aspartyl-L-valyl-L-prolyl-L-alanyl-L-alanyl-L-lysine (FITC-LDVP) and its unlabeled version (LDVP), were from Tocris Bioscience (Avonmouth, Bristol, UK). Human VCAM D1D2 (mature residues F1 to T202) were expressed and purified from HEK 293S GnTI^−/− cell line supernatants by affinity chromatography and gel filtration (Yu et al., 2013). VCAM D1D2 was fluorescently labeled with Alexa Fluor 488 NHS Ester (Thermo Fisher Scientific). Human Fn3_9-10 S1417C mutant (mature residues G1326 to T1509) and its synergy and RGD site (R1374A&P1376A&R1379A&S1417C&Δ1493-1496) mutated inactive version were expressed in Escherichia coli and purified as described (Li et al., 2017; Takagi et al., 2001). Fn3_9-10 S1417C mutant was fluorescently labeled with Alexa Fluor 488 C5 maleimide (Thermo Fisher Scientific) at residue Cys-1417. Both Fn3_9-10 S1417C mutant and its inactive version were biotinylated with Maleimide-PEG11-Biotin at residue 1417 (Thermo Fisher Scientific) in PBS.

Jurkat and K562 cell lines were purchased from ATCC, which have been authenticated by STR profiling. Monthly mycoplasma contamination tests carried out in our lab confirmed the cells were free of mycoplasma.

Jurkat and K562 cells (10⁶ cells/mL in RPMI-1640 medium, 10% FBS) were washed twice with assay medium (Leibovitz’s L-15 medium, 10 mg/mL BSA) containing 5 mM EDTA, twice with assay medium alone, and suspended in assay medium. Cells at 2 × 10⁶ cells/mL were incubated with indicated concentration of Fabs for 30 min at 22°C. Addition of FITC-LDVP, Alexa488-VCAM D1D2 (1.6 labeling ratio), or Alexa488-Fn3_9-10 (1.0 labeling ratio) at indicated concentrations initiated association. Association was measured as MFI at successive time points after addition of the fluorescent ligands. Addition of 500-fold higher concentration of the unlabeled ligand at the end of the association phase initiated the dissociation phase. Background MFI for FITC-LDVP, Alexa488-VCAM D1D2, and Alexa488-Fn3_9-10 in the presence of 10 mM EDTA was subtracted (Figure 2—figure supplement 1).

Kinetic traces including both the association phase and the dissociation phase at different analyte concentrations were globally fitted to the following function:

where t is time, R_t is response at time t, t_D is the time that dissociation starts, [A] is the analyte concentration, and R_max is the maximum response. The first term fits the data in the association phase, and the second term fits the data in the dissociation phase. The prefactor of the first term is 1 prior to t_D and becomes 0 after t_D, whereas the prefactor for the second term is 0 prior to t_D and becomes 1 after t_D. Nonlinear least-square fit of R_t, [A], and t to the above equation yields the on-rate, k_on, off-rate, k_off, and R_max.

Binding kinetics of unclasped high-mannose α5β1 ectodomain and Fn3_9-10 was measured by BLI (Wallner et al., 2013) with streptavidin biosensors on an Octet RED384 System. The reaction was measured on 96-well plate (200 μL/well) in buffer with 20 mM Tris HCl (pH 7.4), 150 mM NaCl, 1 mM Ca²⁺, 1 mM Mg²⁺, and 0.02% Tween20. Streptavidin biosensors were hydrated in reaction buffer for 10 min before starting the measurements. Each biosensor was sequentially moved through five wells with different components: (1) buffer for 3 min in baseline equilibration step; (2) 35 nM biotin-Fn3_9-10 for 1 min for immobilization of ligand onto the biosensor; (3) indicated concentrations of Fabs for 5 min for another baseline equilibration; (4) indicated concentrations of α5β1 ectodomain and Fabs for the association phase measurement; and (5) indicated concentrations of Fabs for the dissociation phase measurement. Each biosensor has a corresponding reference sensor that went through the same five steps, except in step 2 the ligand was replaced with 35 nM inactive version of Fn3_9-10 with both the RGD-binding site and the synergy site (PHSRN) mutated. Background-subtracted response in both the association and dissociation phases, and at different α5β1 ectodomain concentrations, was globally fit to the 1 vs. 1 Langmuir binding model, with ${\displaystyle k_{\mathrm{o}\mathrm{n}}^{\mathrm{a}\mathrm{p}\mathrm{p}}}$ and ${\displaystyle k_{\mathrm{o}\mathrm{f}\mathrm{f}}^{\mathrm{a}\mathrm{p}\mathrm{p}}}$ as shared fitting parameters and maximum response (R_max) for each biosensor as individual fitting parameter. The equilibrium binding response (R_eq) was calculated from ${\displaystyle k_{\mathrm{o}\mathrm{n}}^{\mathrm{a}\mathrm{p}\mathrm{p}}}$ and ${\displaystyle k_{\mathrm{o}\mathrm{f}\mathrm{f}}^{\mathrm{a}\mathrm{p}\mathrm{p}}}$ values at each α5β1 ectodomain concentration and fit to a dose–response curve to calculate K_d values as a check on ${\displaystyle k_{\text{off}}^{\text{app}}/k_{\text{on}}^{\text{app}}}$ values. To calculate R_eq, fitted ${\displaystyle k_{\text{on}}^{}}$, ${\displaystyle k_{\text{off}}^{}}$, and R_max values at each α5β1 ectodomain concentration [A] were used to calculate R_eq at a time 1000-folder longer than the ‘binding time’, that is,${\displaystyle t=1000\ast \frac{1}{k_{\text{on}}[A]}}$ with the following equation:

The measured on- and off-rates (${\displaystyle k_{\text{on}}^{\text{app}}}$ and ${\displaystyle k_{\text{off}}^{\text{app}}}$) for each defined ensemble containing two or three states shown in Figures 2—4 was approximated by the on- and off-rates of each state weighted by its population in the ensemble (Figure 7A, Eqs. 1–4). At steady state, the population of the free integrin states and the ligand-bound integrin states was calculated based on the previously determined population and intrinsic ligand-binding affinity of each state (Figure 7—figure supplement 1, Eqs. S5–S10) in the respective integrin α4β1 and α5β1 preparations (Li and Springer, 2018; Li et al., 2017; Figure 7B). Specifically, ${\displaystyle \frac{K_{\text{a}}^{\text{EO}}}{K_{\text{a}}^{\text{EC}}}\text{ and }\frac{K_{\text{a}}^{\text{EO}}}{K_{\text{a}}^{\text{BC}}}}$ (in Figure 7—figure supplement 1B, Eqs. S8–S10) are the intrinsic ligand-binding affinity ratios of the EO state and the closed states. For integrin α4β1, the ratios were averaged to 729 ± 211 from six α4β1 preparations, including α4β1 headpiece with high-mannose N-glycans, α4β1 ectodomain with high-mannose N-glycans, α4β1 ectodomain with complex N-glycans, and intact α4β1 on three different cell lines (Li and Springer, 2018); for integrin α5β1, the intrinsic ligand-binding affinity ratio of the EO state and the closed states were averaged to 3106 ± 1689 from eight soluble α5β1 preparations that varied in the presence or absence of the lower legs, of a loose clasp in place of the TM domain, and in whether the N-linked glycan was complex, high mannose, or shaved (Li et al., 2017). Using ${\displaystyle k_{\text{on}}^{\text{EO}}}$ and ${\displaystyle k_{\text{off}}^{\text{EO}}}$ rates experimentally measured in Figures 2—4, ${\displaystyle k_{\text{on}}^{\text{EC}}}$ and ${\displaystyle k_{\text{off}}^{\text{EC}}}$ were derived from the ${\displaystyle k_{\text{on}}^{\text{app(EC+EO)}}}$ and ${\displaystyle k_{\text{off}}^{\text{app(EC+EO)}}}$ measured in extended ensembles, respectively (Figure 7A, Eqs. 1 and 2). By including the values for ${\displaystyle k_{\text{on}}^{\text{EC}}}$ and ${\displaystyle k_{\text{off}}^{\text{EC}}}$ in addition to ${\displaystyle k_{\text{on}}^{\text{EO}}}$ and ${\displaystyle k_{\text{off}}^{\text{EO}}}$, ${\displaystyle k_{\text{on}}^{\text{BC}}\text{ and }{k}_{\text{off}}^{\text{BC}}}$ were then derived from ${\displaystyle k_{\text{on}}^{\text{app(BC+EC+EO)}}}$ and ${\displaystyle k_{\text{off}}^{\text{app(BC+EC+EO)}}}$ measured in basal ensembles, respectively (Figure 7A, Eqs. 3 and 4).

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank Kelly L Arnett at the Center for Macromolecular Interactions of Harvard Medical school for training and consultation on BLI measurement. We thank Taekjip Ha for suggestions on our manuscript. We thank Margaret Nielsen for making cartoons for each figure and for figure preparation. This work was funded by NIH R01 HL131729 ('Activation trajectories of integrin a5b1').

© 2021, Li et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.