Conformational equilibria and intrinsic affinities define integrin activation

Abstract We show that the three conformational states of integrin α5β1 have discrete free energies and define activation by measuring intrinsic affinities for ligand of each state and the equilibria linking them. The 5,000‐fold higher affinity of the extended‐open state than the bent‐closed and extended‐closed states demonstrates profound regulation of affinity. Free energy requirements for activation are defined with protein fragments and intact α5β1. On the surface of K562 cells, α5β1 is 99.8% bent‐closed. Stabilization of the bent conformation by integrin transmembrane and cytoplasmic domains must be overcome by cellular energy input to stabilize extension. Following extension, headpiece opening is energetically favored. N‐glycans and leg domains in each subunit that connect the ligand‐binding head to the membrane repel or crowd one another and regulate conformational equilibria in favor of headpiece opening. The results suggest new principles for regulating signaling in the large class of receptors built from extracellular domains in tandem with single‐span transmembrane domains.


Materials and Methods
Materials hTERT-BJ cells were a gift from Dr. Robert A. Weinberg (Whitehead Institute for Biomedical Research). Mouse anti-human antibodies J143 (4th International Workshop on Leukocytes), SAM-1 (5th International Workshop on Leukocytes), #481709 (R&D Systems), LM142 (EMD Millipore) and TS2/4 1 were from the indicated sources. FITC-conjugated goat anti-mouse IgG was purchased from Sigma. α 5 specific blocking antibody mAb16 was purified from hybridoma provided to us by Dr. Kenneth M. Yamada. α 4 specific antibody Natalizumab was from commercial source.
Negative stain electron microscopy (EM) EM specimen preparation, data collection and processing were as described 2 .
Protein and carbohydrate composition Unclasped α 5 β 1 ectodomain with shaved (30 µg), high-mannose (60 µg), and complex (30 µg) N-glycans were separately loaded on an Agilent liquid chromatography system equipped with a TSKgel BioAssist G4SWXL analytical size exclusion column (Tosoh Bioscience), a DAWN HELEOS II multi-angle light scattering detector, an Optilab T-rEX refractive index detector and a variable wavelength UV detector (Wyatt Technology Corporation). Data were processed in ASTRA 6 using the protein conjugate model (d /d = 0.185 and 0.145 for protein and carbohydrate components, respectively) 4 .
Quantitative comparison of α 5 -, α 4 -and β 1 subunit expression levels on K562 and Jurkat cells was the same as described about except that cells were only incubated with 3.75 µg/mL Alexa647-conjugated primary antibodies before subject to washing and flow cytometry.
For determining EC 50 values for extension-stabilizing and open-stabilizing Fabs, we made the assumption that the increase in FP was directly proportional to the increase in concentration of Fab-bound open α 5 β 1 . This assumption is reasonable because the affinity of the EO conformation is so much higher than that of the BC and EC conformations for cRGD. Therefore, data were fit to a dose response curve: In the case of closure-stabilizing Fabs where α 5 β 1 was used at a high concentration in the assay, EC 50 significantly deviates from d due to depletion of Fab and FITC-cRGD. Therefore, we wrote equations S3-S10 as described below, and fit data to Eq. S11 below. In the assay, FITC-cRGD and its complex with α 5 β 1 free of Fab were the major sources of FP obs ; the α 5 β 1 ·Fab complex essentially does not bind FITC-cRGD due to its extremely low affinity, which was evident in Fig. 2 and Fig. S1 where at high concentrations of closure-stabilizing Fab, FP obs dropped to the value of free FITC-cRGD (0.09). Experimentally, α 5 β 1 was at 100 nM and cRGD was at 5 nM; in experiments to determine Fab d (Fig. S1) most of the FP signal was due to cRGD bound to the open α 5 β 1 conformation. Because the observed decrease in FP was due to Fabbinding to α 5 β 1 and stabilizing it in the closed conformation, we first wrote the equations for Fab-binding to α 5 β 1 and then considered the effect on α 5 β 1 binding to cRGD: are total concentrations of α 5 β 1 (100 nM) and closure-stabilizing Fab in the assay, respectively; [α 5 β 1 ] ′ is the concentration of Fab-free α 5 β 1 at equilibrium. In the following equations, we make the reasonable assumption that only Fab-free α 5 β 1 contributes to the FP signal: where FP L and FP α 5 β 1 ·L are FP of free FITC-cRGD and α 5 β 1 ·FITC-cRGD complex, respectively; [α 5 β 1 ] ′ is defined in Eq. S9. Fitting the FP obs and [Fab] tot data to Eq. S11 yielded ens(Basal):Fab d , FP L and FP α 5 β 1 ·L .
Affinity of α 5 β 1 ectodomain for Fn3 9-10 from competitive binding and Eqs. S27-S28 α 5 β 1 ectodomain affinities for Fn3 9-10 was measured by using Fn3 9-10 to compete binding of FITC-cRGD peptide ligand. α 5 β 1 ectodomain (270 nM in the absence of Fabs, 20 nM in the presence of HUTS4 Fab, 90 nM in the presence of 8E3 Fab, or 70-10,000 nM in the presence of mAb13 Fab or mAb13 plus 9EG7 Fabs) was equilibrated with 0-10,000 nM Fn3 9-10 (competitor) for 2 hr. The mixture was incubated with 5 nM FITC-cRGD (ligand) for 2 hr, and FP was measured. Since only α 5 β 1 free of Fn3 9-10 could bind FITC-cRGD, the equations are identical to those for binding affinities for closure-stabilizing Fabs (Eq. S3-S11). Substituting Fab with C (for competitor) in Eq. S9 and S11: where FP obs is the measured FP; FP L and FP α 5 β 1 ·L are FP of free FITC-cRGD and α 5 β 1 ·FITC-cRGD complex, respectively; [α 5 β 1 ] tot , [L] tot and [C] tot are total concentrations of α 5 β 1 , FITC-cRGD ligand (5 nM) and Fn3 9-10 competitor in the assay, respectively; [α 5 β 1 ] ′ is the concentration of Fn3 9-10 -free α 5 β 1 , either free of FITC-cRGD or with FITC-cRGD bound; Fab-binding affinities of α 5 β 1 conformational ensembles and Eqs. S29-S40 The basal ensemble (i.e., in the absence of Fabs) of intact α 5 β 1 on the cell surface or of its ectodomain fragment in solution comprises three overall conformational states-bent-closed (BC), extended-closed (EC), and extended-open (EO). Suppose a Fab binds (and thus stabilizes) one or more of the BC, EC and EO states in the basal ensemble to form BC·Fab, EC·Fab and/or EO·Fab complexes with intrinsic Fabbinding affinities BC a , EC a and/or EO a , respectively: If the Fab does not bind a state ( = BC, EC or EO), then a = 0 and [ ·Fab] = 0. For Fabs that bind two states and with equal affinities, i.e., a = a ( = EC and = EO for extension-stabilizing Fabs, = BC and = EC for closure-stabilizing Fabs), Fab-binding does not change the relative distribution of the two states: The experimentally measured affinity of the basal ensemble for Fab, Binding affinity is often expressed as dissociation constant ( d = 1/ a ) in the biological sciences to facilitate comparison to concentrations of the reactants. Following this convention, Eq. S33 can be rewritten as: The total probability (population) of all Fab-bound species in the ensemble is For the α 5 β 1 headpiece, a similar analysis shows that at a sufficiently high concentration of Fab, the Fab-stabilized state dominates the ensemble.
In Eq. S39, [Fab] is the concentration of free Fab. In most experiments, Fabs are used at concentrations much higher than that of α 5 β 1 ; therefore, [Fab] can be approximated as the total Fab concentration minus the highest α 5 β 1 concentration used ([Fab] tot − [α 5 β 1 ] tot ) in calculating α 5 β 1 ·Fab : True ligand-binding affinities ( ens d ) of α 5 β 1 conformational ensemble members and Eqs. S41-S72 We now consider using Fabs to stabilize conformational ensembles in specific states. A complication is that in the absence of 100% binding of the Fab, unbound α 5 β 1 can exist in other states. We derive the equations for determining the contributions of both Fab-bound α 5 β 1 ( α 5 β 1 ·Fab ) and unbound α 5 (Fig. S2), and we use our derived equations to calculate true affinity, ens d . Intact α 5 β 1 or its ectodomain fragment contain three states in their conformational ensemble, BC, EC and EO. Suppose they bind a ligand, L, to form the BC·L, EC·L, and EO·L complexes with intrinsic ligandbinding affinity (association constants) BC a , EC a and EO a , respectively: If a Fab is also present in the ensemble, then the Fab-bound state(s) BC·Fab, EC·Fab and/or EO·Fab can bind the ligand L to form BC·Fab·L, EC·Fab·L, and/or EO·Fab·L complexes with intrinsic ligandbinding affinity (association constants) BC·Fab a , EC·Fab a and EO·Fab a , respectively. Suppose the Fab-bound α 5 β 1 states bind ligand with the same affinity as their corresponding native α 5 β 1 states (see main text for justification), then: Calculation of free energy of each conformational state and Eqs. S78-S84 Using EO as the reference state (Δ EO = 0), the relative free energies of the BC and EC states, Δ BC and Δ EC , are related to the probabilities (populations) BC , EC and EO through the Boltzmann distribution (as also shown in Fig. 1C): where is known as the partition function. Solve Eq. S78-S81 for Δ BC and Δ EC , and substituting BC , EC and EO with Eq. S73-S75: Likewise, for α 5 β 1 headpiece, using O as the reference state (Δ O = 0) Calculation of free energies associated with conformational changes and Eqs. S85-S94 α 5 β 1 activation is associated with ectodomain extension and headpiece opening. These two types of conformational changes are not necessarily separate, independent steps; nor must they occur in a predefined order, allowing the conformational change among the three integrin states to be defined as from one state to another, or as interchange between one state and two other states. Indeed, we have previously described scenarios for different orders of steps 7 , and movies show how headpiece opening may either follow or precede extension (Supplemental Movies EV1-EV3). States such as bent-open may also be possible (Movie EV3); however, since these states have never been visualized by electron microscopy or small-angle X-ray scattering, their populations must be small, and thus the presence of their populations in the numerator or denominator of conf (defined below) has little effect on conf values. Extension from BC to EC defines E conf and its associated free energy Δ E conf :  Binding of closure-stabilizing Fab SG/19 to α 5 β 1 (100 nM) influenced binding of FITC-cRGD (5 nM) to α 5 β 1 as monitored by FP. d values were obtained from fitting the FP data to Eq. S11. Errors are fitting errors from triplicates.      Cells were stained with 2.5 µg/mL primary antibody, followed by 2 µg/mL FITC-conjugated secondary antibody. Anti-α L serves as negative control.