Open Conformation of Ezrin Bound to Phosphatidylinositol 4,5-Bisphosphate and to F-actin Revealed by Neutron Scattering*

Background: The structure of activated ezrin is not known. Results: We have determined the conformation of activated ezrin upon binding to PIP2 and to F-actin. Conclusion: Activated ezrin forms more extensive contacts with F-actin than generally depicted. Significance: This study provides new insight into the mechanisms by which ezrin assembles signaling complexes at the membrane-cytoskeleton interface. Ezrin is a member of the ezrin-radixin-moesin family (ERM) of adapter proteins that are localized at the interface between the cell membrane and the cortical actin cytoskeleton, and they regulate a variety of cellular functions. The structure representing a dormant and closed conformation of an ERM protein has previously been determined by x-ray crystallography. Here, using contrast variation small angle neutron scattering, we reveal the structural changes of the full-length ezrin upon binding to the signaling lipid phosphatidylinositol 4,5-bisphosphate (PIP2) and to F-actin. Ezrin binding to F-actin requires the simultaneous binding of ezrin to PIP2. Once bound to F-actin, the opened ezrin forms more extensive contacts with F-actin than generally depicted, suggesting a possible role of ezrin in regulating the interfacial structure and dynamics between the cell membrane and the underlying actin cytoskeleton. In addition, using gel filtration, we find that the conformational opening of ezrin in response to PIP2 binding is cooperative, but the cooperativity is disrupted by a phospho-mimic mutation S249D in the 4.1-ezrin/radixin/moesin (FERM) domain of ezrin. Using surface plasmon resonance, we show that the S249D mutation weakens the binding affinity and changes the kinetics of 4.1-ERM to PIP2 binding. The study provides the first structural view of the activated ezrin bound to PIP2 and to F-actin.

adhesion molecules CD44, CD43, and ICAM-1/2/3 (19) or the G-protein couple receptor parathyroid hormone receptor (20,21). The FERM domain can also interact with transmembrane proteins via the scaffolding protein Na ϩ /H ϩ exchanger regulatory factor 1 or 2 (NHERF1 or NHERF2) (22,23); FERM binds to the C-terminal domain of NHERF proteins tightly (22,24), and the PDZ domains of the NHERF scaffolding proteins in turn bind to a number of transmembrane ion transport proteins and receptor complexes (25)(26)(27)(28). The last 34 residues of C-ERMAD bind to F-actin (29 -32). Because of the ability to interact with both the cell membrane components and the actin cytoskeleton, ERM proteins are membrane-cytoskeleton adapter proteins that form regulated signaling linkages between the assembled membrane signaling complexes and the actin cytoskeletal network.
The ERM proteins are regulated by an autoinhibitory mechanism, with the inactive protein being held in a closed and inactive conformation by head-to-tail-like intramolecular interactions (30,33). X-ray crystallography studies reveal that the FERM domains of all ERM proteins adopt a conserved cloverleaf-like structure with three subdomains, F1, F2, and F3 (34,35). In the closed ERMs, the central ␣-helical region folds back into an anti-parallel coiled coil (33). The C-ERMAD adopts an extended structure that binds extensively to the F2 and F3 subdomains, thus masking both the membrane-binding and the cytoskeleton-binding sites (36). Additionally, the crystal structure representing the full-length moesin from Spodoptera frugiperda (Sfmoesin) reveals that the N-terminal portion of the ␣-helical linker provides further protection to the FERM domain, further preventing FERM from binding to other proteins (33). Because of such tightly regulated intramolecular interactions, the inactive ERM proteins exhibit no binding to NHERF1 or to CD44 by the FERM domain or to F-actin by the C-ERMAD.
The ERM proteins are activated upon binding to the membrane signaling lipid phosphatidylinositol 4,5-bisphosphate (PIP 2 ) and/or by phosphorylation (37)(38)(39). Binding to PIP 2 is thought to release the head-to-tail intramolecular interaction in the ERMs. Phosphorylation at a conserved Thr in the C-ERMAD also contributes to ERM activation. This conserved residue is Thr-567 for ezrin, Thr-558 for moesin, and Thr-563 for radixin (40), which can be phosphorylated by a number of Ser/Thr kinases, including the Rho kinase (41), atypical protein kinase C (42), lymphocyte-oriented kinase (43), and MST4 (44). In cells, the phosphorylated ERMs are localized in the membrane extensions that are rich in actin (12,(45)(46)(47). Because this conserved Thr site is masked by the FERM domain in the dormant ERMs, PIP 2 binding is considered to cause conformational changes to make the C-terminal Thr site accessible to kinases for phosphorylation. It is proposed that PIP 2 binding and phosphorylation act sequentially in the activation of ezrin (48). A recent study shows that in the presence of PIP 2 , ezrin binding to F-actin is enhanced by the T567D phosphomimic mutation (44).
We have determined the molecular conformation of activated ezrin in the PIP 2 -bound and F-actin-bound states using contrast variation small angle neutron scattering (SANS). Similar to small angle x-ray scattering (SAXS), SANS determines the size, molecular mass, and shape of a protein in solution. Moreover, SANS has the capability of studying the structure of a multicomponent complex by contrast variation and deuterium labeling. By changing the D 2 O concentration (or deuterium content) of a buffer solution, one varies the neutron scattering-length density contrast between the buffer background and a particular component in a complex. Contrast variation SANS can retrieve not only the overall shape but also the internal structure of a protein⅐lipid membrane, protein⅐DNA, or a multiprotein complex.
We find that the wild-type ezrin and phosphomimic mutants adopt a closed conformation in solution. PIP 2 binding is sufficient to induce the opening of ezrin. Additionally, using gel filtration, we find that the conformational opening of ezrin in response to PIP 2 binding is cooperative, but the cooperativity of conformational opening is abolished by a phospho-mimic mutation S249D in the FERM domain. Furthermore, ezrin binding to F-actin requires the simultaneous binding of ezrin to PIP 2 . Once bound to F-actin, the opened ezrin forms more extensive contacts with F-actin than previously thought. This study provides the first view of how activated ezrin interacts with the membrane component PIP 2 and with F-actin.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The human cDNA encoding the full-length ezrin was subcloned into the pET151/ D-TOPO vector (Invitrogen). The T567D, S249D, and S249D/ T567D mutants were generated with the QuikChange II sitedirected mutagenesis kit (Agilent Technologies). The plasmids were transformed into Rosetta 2 (DE3) competent cells (EMD Biosciences). The bacterial cells were grown at 37°C to an absorbance of 0.8 at 600 nm and were induced with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside for 2 h. The proteins were purified by a Ni 2ϩ -chelating column and by gel filtration using a Superdex 200 10/300 GL column (GE Healthcare). The tag of the purified protein was cleaved by acetyl tobacco etch virus protease (Invitrogen). The cleaved tag and residual uncleaved proteins were removed by a Ni 2ϩ -chelating column. The purity of the proteins is above 95% as estimated from SDS-PAGE (Fig. 1B).
For producing deuterated proteins, bacteria cells were grown at 37°C in sterile D 2 O M9 medium (D 2 O 99.9%, Cambridge Isotope Laboratories) until the absorbance at 600 nm reached 0.7-0.8. The cells were induced with 0.25 mM isopropyl ␤-D-1thiogalactopyranoside for 11-12 h. Purification of the deuterium-labeled protein was similar to that for the unlabeled protein. The nonexchangeable deuterium content of the purified deuterated proteins ranged from 0.63 to 0.67, as determined by matrix-assisted laser desorption time-of-flight mass spectrometry at the Columbia University Protein Core Facility. At such deuteration levels, the scattering length density of the deuterated protein approximately matched that of 100% D 2 O buffer. To determine whether deuteration caused any conformational changes in ezrin, we compared the conformation of the hydrogenated ezrin and deuterated ezrin ( d ezrin) in buffer by SAXS and SANS (supplemental Fig. S1). The hydrogenated ezrin and indicating that deuterium labeling does not cause conformational changes in the protein.
Lipid Micelle Preparation-The short chain lipid 1,2-diheptanoyl-sn-glycero-3-phosphocholine (DHPC) and PIP 2 ammonium salt from porcine brain, dissolved in 20:9:1 CHCl 3 / MeOH/H 2 O, were purchased from Avanti Polar Lipids, Inc. Before the experiments, the solvent from the PIP 2 solution was removed in a speed-vac for 1 h, and the dry PIP 2 film was dissolved in the buffer of 25 mM Tris (pH 7.5), 300 mM NaCl, 1 mM DTT.
F-actin Filament Preparation-Nonmuscle actin from a human platelet of higher than 99% purity was purchased from Cytoskeleton Inc. One milligram of lyophilized actin was resuspended by adding 100 l of 20 mM Tris-HCl (pH 7.5), 0.2 mM CaCl 2 , 0.2 mM ATP, and 0.2 mM DTT. The protein concentration was determined by measuring the absorbance at 280 nm, using the molar extinction coefficient ⑀ ϭ 42,680 M Ϫ1 cm Ϫ1 and a molecular mass ϭ 41.7378 kDa. A 10ϫ F-actin polymerization buffer of 500 mM KCl, 20 mM MgCl 2 , and 10 mM ATP was added to achieve a final concentration of 1ϫ to initiate actin polymerization. Actin was allowed to polymerize at room temperature for 1 h.
Surface Plasmon Resonance Experiments-SPR experiments were performed on a Biacore X100 (GE Healthcare). For studying the binding of FERM or FERM(S249D) to PIP 2 , purified FERM and FERM(S249D) were dialyzed overnight in SPR binding buffer, containing 10 mM Hepes (pH 7.5), 300 mM NaCl, 3 mM EDTA. An L1 chip was coated with 0.4 mM PIP 2 ϩ 16 mM DHPC to a response unit of 424.9. The analyte, FERM, or FERM(S249D) was injected onto the chip at a flow rate of 30 l/min at 10°C. The sensor chip was regenerated by passing 20 mM NaOH after each analyte injection. The sensorgrams were fit with a 1:1 kinetic binding model with the manufacturer's supplied program to obtain the rate constant k on and k off , as well as the dissociation constant K d .
Gel Filtration Analysis of Conformational Opening upon Binding to Lipid-A super Superdex 200 10/300 GL gel filtration column was used to analyze the conformational opening of ezrin and mutants in DHPC and PIP 2 . The buffer used for these gel filtration analyses is 25 mM Tris (pH 7.5), 300 mM NaCl, 0.5 mM DTT, and 0.1 mM EDTA. Before the experiment, 9.9 M ezrin or a mutant and 50.5 mM DHPC were incubated with PIP 2 at different molar ratios for 1 h. The experiments were repeated with a protein concentration of 6.  The F o versus PIP 2 concentration data were either fit with a sigmoidal function F o ‫؍‬ B max (L n )/(K n ϩ L n ), where L is the ligand concentration, K is the mid-point of transition, and B max the top asymptote.
Solution Small Angle X-ray Scattering-SAXS experiments were performed with an in-house apparatus, utilizing a Micro-MaxTM-007 HF Microfocus rotating anode generator as the x-ray source (Rigaku/MSC). In this study, a 0.014 Ͻ Q Ͻ0.32 Å Ϫ1 range was covered, where Q ϭ 4sin/ is the magnitude of the scattering vector; is half the scattering angle, and is the wavelength of the x-ray. Details about SAXS data reduction and analysis have been described previously (49 -51).
Solution Small Angle Neutron Scattering-SANS experiments were performed at the Bio-SANS (CG3) at the High Flux Isotope Reactor and at the EQ_SANS at the Spallation Neutron Source (52), Oak Ridge National Laboratory. At the Bio-SANS, the neutron wavelength, , was 6 Å, with a wavelength spread, ⌬/, of 0.14 obtained with a velocity selector. Scattered neutrons were detected with a 1 ϫ 1 m 2 helium-filled two-dimensional position sensitive detector with 192 ϫ 192 pixels. Two sample-to-detector distances, 8 and 1.7 m, were used to cover a Q range between Q min ϭ 0.008 Å Ϫ1 and Q max ϭ 0.4 Å Ϫ1 . The data acquisition time from the samples varied from ϳ20 min to 4 h at each detector position to ensure sufficient data statistics. The two-dimensional raw counts were corrected for nonuniform detector response and electronic dark current, which represents the ambient radiation background and electronic noise and azimuthally averaged to produce a one-dimensional profile I(Q). The data processing procedure for EQ-SANS has been described previously (53). At EQ-SANS, the Q range covered is between Q min ϭ 0.008 Å Ϫ1 to Q max ϭ 0.5 Å Ϫ1 . Data were placed on an absolute scale in units of cm Ϫ1 through the use of pre-calibrated secondary standards (54).
Before SANS experiments, the protein, protein⅐lipid, and protein⅐lipid⅐F-actin complexes were dialyzed against buffer containing the desired D 2 O volume fraction for two times, each time for about 8 h. The buffer used for SANS experiments contains 25 mM Tris (pH 7.5), 300 mM NaCl, 1 mM DTT. Protein concentrations in buffer and lipid were measured by UV absorption spectroscopy at 280 nm, using the calculated extinction coefficients based on the amino acid sequence of the recombinant proteins. For ezrin in complex with both the lipid and F-actin, the concentrations were estimated based on the concentrations of the stock solution. The protein concentrations used in the SANS experiments range between 1.0 and 1.85 mg/ml. At these protein concentrations, the inter-molecular interactions are negligible (see supplemental Figs. S2 and S3). The SANS buffer background at each D 2 O volume fraction was taken from the dialysis buffer. The sample cells used for SANS experiments are 1-mm quartz cuvettes.
SANS Data Analysis and Three-dimenional Shape Reconstruction-The length distribution function P(r), radius of gyration R g , the forward scattering intensity I(0), and the maximum dimension D max were calculated from the scattering data using the program GNOM (55). R g and I(0) can also be obtained from Guinier fitting (see supplemental Table S1). The threedimenional "dummy bead" coordinates were generated using the program DAMMIN (56). Multiple calculations were performed using DAMMIN, and the generated 10 structures were averaged and filtered using the program DAMAVER and DAMFILT (57). The normalized spatial discrepancy (NSD) value, which is a measure of reproducibility of the generated three-dimenional shape, is given in the figure legends. The three-dimenional density map was generated from the averaged coordinates using the program Situs (58). The fitting and docking of the high resolution structure to the density map were performed using Situs or UCS Chimera (59).
The scattering from an F-actin or a protein⅐F-actin complex can be considered as the scattering from a long rod (60, 61) as shown in Equation 2, with P c (r) the cross-section length distribution function, and J o (QR) the zero-order Bessel function. The cross-section forward scattering intensity I c (0) is related to the mass per unit length of the complex M L (see supplemental material) (62). The cross-section length distribution function of the filament, P c (r), was obtained using the program GNOM (56), which also gives the cross-section radius of gyration R c , and the cross-section maximum dimension D max,c. The scattering Q range of 0.02 Ͻ Q Ͻ0.20 Å Ϫ1 was used to calculate P c (r).

Closed and Autoinhibited Conformation of Ezrin and Phospho-mimic Mutants in Solution-Previous biochemical studies
have shown that phosphorylation at Thr-567 in the C-ERMAD of ezrin contributes to ezrin activation (12, 45-47). A Thr 3 Asp or Ser 3 Asp mutation is often employed to mimic the negative electrostatic charges of a phosphorylated Thr or Ser (47,63,64). Recently, we have identified a new conserved phosphorylation site Ser-249 in the FERM domain of ezrin, and we have found that cells expressing the phospho-mimic ezrin(S249D) and ezrin(S249D/T567D) show significantly weakened cell-cell adhesion, as well as altered subcellular localizations of ezrin. We have characterized the oligomer states of the full-length wild-type ezrin and the phospho-mimic mutants, ezrin(T567D) and ezrin(S249D/T567D), using gel filtration and static light scattering.
For both ezrin and ezrin(T567D), the gel filtration chromatograms show two peak fractions, one at elution volume 12.3 ml and the other at 14.7 ml (Fig. 1B). Static light scattering indicates that the fraction at elution volume 12.3 ml is a dimer, although the fraction at 14.7 ml is a monomer (supplemental Fig. S3). The dimer fraction of ezrin(S249D/T567D) is significantly reduced compared with that of ezrin or ezrin(T567D) (Fig. 1C). Additionally, the monomer fraction of ezrin or ezrin(T567D) does not convert to a dimer fraction after gel filtration separation (Fig. 1D), and the dimer fraction decreases after storing the proteins on ice for several days.
We then performed solution SAXS experiments on the monomer and the dimer fractions of ezrin (supplemental Fig.  S1). The radius of gyration (R g ) and the maximum dimension (D max ) of the monomeric ezrin from SAXS are listed in Table 1. For the dimer fraction of ezrin, SAXS yields similar R g and D max values as the monomer fraction, suggesting that ezrin has converted to the folded monomer conformation during the exper- iment. We thus focus on analyzing the monomer fraction of ezrin in the SANS experiments. Fig. 2 shows the SANS results from deuterated wild-type ezrin ( d ezrin) and deuterated ezrin(T567D) ( d ezrin(T567D)) in solution. Overall, R g , D max , and P(r) values of d ezrin are identical to those of d ezrin(T56D) ( Table 1 and Fig. 2, A and B). Fig.  2C shows the three-dimenional molecular envelopes of d ezrin and d ezrin(T567D), ab initio reconstructed from SANS. For comparison, the reconstructed three-dimenional maps are docked to the crystal structure of Sfmoesin (Protein Data Bank code 2I1K) that represents the closed and auto-inhibited conformation of an intact monomeric ERM protein (Fig. 2C) (33). The crystal structure of the autoinhibited form of Sfmoesin has a central helical linker composed of two helices folded into antiparallel coiled-coil conformation. The comparison indicates that both d ezrin and d ezrin(T567D) adopt a closed conforma-tion in solution, and the phospho-mimetic d ezrin(T567D) does not have apparent conformational changes when compared with the wild-type protein.  (Fig. 3B). However, the P(r) functions of d ezrin(S249D) and d ezrin(S249D/ T567D) show a more pronounced shoulder at r 32 Ϸ90 Å, and R g and D max values of d ezrin(S249D/T567D) are slightly larger than the wild-type protein ( Fig. 3B and Table 1).
Using SPR, we have also estimated the fraction of ezrin, ezrin(T567D), ezrin(S249D), and ezrin(S249D/T567D) that is capable of binding to target protein NHERF1. Previous biochemical and structural studies have shown that the conserved NHERF1-binding site is located in the F3 subdomain of FERM,

, D max , and I(0)/c of d ezrin* and mutants in buffer, PIP 2 -bound, and F-actin-bound states
Static light scattering experiments (see supplemental Fig. S3), which measure the absolute molecular mass, show that the fraction of ezrin eluted at 14.6 ml in the Superdex 200 10/300 GL column is monomeric. The parameters listed here are from P(r) function calculations. See supplemental Table S1 for R g and I(0) obtained from Guinier fit. which is masked by C-ERMAD in the closed form of full-length ezrin (33,65,66). Using the monomer fraction from gel filtration, we find that an insignificant 0.4% fraction of the wild-type ezrin is capable of binding to NHERF1, whereas 16.2% ezrin(T567D), 18.3% ezrin(S249D), and about 27% ezrin(S249D/ T567D) are capable of binding NHERF1 (supplemental Table S2). The SANS and SPR results thus confirm that in the wild-type ezrin, FERM is tightly auto-regulated or masked by C-ERMAD. The results also show that although the phospho-mimic mutants are largely folded, they are more dynamic than the wild-type protein because a considerable fraction of the mutants is capable of binding to NHERF1 (supplemental Table  S2). It is likely that inter-domain motions between the FERM and the C-ERMAD domain are activated in the phospho-mimic mutants, so that a fraction of the mutants can sample the conformational states that are capable of binding to NHERF1.

R g D max
Phospho-mimetic Ezrin(S249D) Mutation Abolishes the Cooperativity of Conformational Opening of Ezrin in Response to PIP 2 Binding-We have performed a gel filtration analysis of the binding of ezrin, ezrin(T567D), or ezrin(S249D/T567D) to the short chain phospholipid DHPC and to the signaling lipid PIP 2 . After 9.9 M ezrin or ezrin(T567D) is incubated with 50.5 mM DHPC alone for an hour, the dimer fraction eluting at 12.3 ml disappears, and only the monomer fraction elutes at 14.7 ml in the gel filtration chromatogram (Fig. 4A and supplemental  Fig. S4). DHPC thus can disrupt the dimer fraction of ezrin or   Fig. S1), DHPC does not bind to ezrin or to ezrin(T567D) significantly to affect the size of the eluted protein.
When PIP 2 is incubated with ezrin in the presence of DHPC, the gel filtration chromatogram starts to show a fraction that elutes at 12.9 ml (Fig. 4A). Our contrast-matching SANS shows that PIP 2 binding induces large conformational changes but does not alter the oligomeric states of ezrin (see Fig. 5A and Table 1). The peak fraction at a 12.9-ml elution volume thus indicates significant conformational changes in the PIP 2 -bound ezrin (Fig. 4A). For ezrin or ezrin(T567D), the peak height of the opened conformation fraction increases with increasing PIP 2 concentrations until the PIP 2 /protein molar ratio reaches 230 ( Fig. 4A and supplemental Fig. S4). In addition, with increasing PIP 2 concentrations, the "closed" form of ezrin also becomes more expanded in the gel filtration chromatogram (Fig. 4A), suggesting the existence of intermediate states. These intermediate states may be either PIP 2 -bound closed monomer or partially open monomer. At the PIP 2 /protein molar ratio, the ezrin⅐PIP 2 complex only elutes as an open conformation, suggesting that ezrin is fully opened. The lipid⅐protein complexes preclude light scattering from determining the size and molecular mass of the protein. These results indicate that PIP 2 binds to ezrin or ezrin(T567D) and causes conformational changes in ezrin or ezrin(T567D).
The fraction of conformational opening (F o ) of ezrin or ezrin(T567D) in response to PIP 2 binding can be calculated from the peak areas of the opened and closed forms of the protein (see "Experimental Procedures"). In Fig. 4B, F o of E, three-dimenional shapes of opened d ezrin in PIP 2 reconstructed from SANS. The largest NSD value is 0.673 from the 10 models used to generated the averaged structure. F, threedimenional shape of opened d ezrin(T567D) in PIP 2 reconstructed from SANS. The model is generated by averaging 10 models generated from DAMMIN. The largest NSD value is 0.610 from the 10 models used to generated the averaged model. OCTOBER 26, 2012 • VOLUME 287 • NUMBER 44 JOURNAL OF BIOLOGICAL CHEMISTRY 37125 ezrin(T567D) shows a sigmoidal response to PIP 2 concentration with a mid-point of transition of 80.0 Ϯ 3.5 and a Hill coefficient of 3.1 Ϯ 0.4 (Fig. 4B). For the wild-type ezrin, F o also shows a sigmoidal response to PIP 2 concentration with a midpoint of transition 136 Ϯ 40 mM and a Hill coefficient of 2.1 Ϯ 0.5 (Fig. 4B). These analyses illustrate the cooperative opening of the conformation of ezrin or ezrin(T567D) in response to PIP 2 binding. The Hill coefficients of PIP 2 binding to the wildtype of ezrin or to ezrin(T567D) suggest that more than one PIP 2 molecule is required in the process of opening and activating ezrin.

Open Conformation of Ezrin
Ezrin(S249D/T567D) shows quite a different PIP 2 binding and conformational opening behavior as compared with that of ezrin or ezrin(T567D) (Fig. 4B and supplemental Fig. S3). The F o curve is noncooperative, with a Hill coefficient of 1.05 Ϯ 0.07 (Fig. 4B), and a mid-point of transition of 227.6 Ϯ 39. The ezrin(249D/T567D) mutant has abolished the cooperativity of conformational opening in response to PIP 2 binding, and the opening of the ezrin(S249D/T567D) also requires a higher PIP 2 concentration. Comparing the F o value of the wild-type ezrin, ezrin(T567D), and ezrin(S249D/ T567D) indicates that the S249D mutant affects the PIP 2 binding behavior.
Although F o is an indication of the conformational transition, the mid-point of F o is not necessarily the same as the dissociation constant K d of ezrin to PIP 2 binding. We have thus performed SPR analysis of the affinity and kinetics of PIP 2 binding to FERM and FERM(S249D) (Fig. 4, C and D, and Table 2). SPR shows that PIP 2 has a considerably higher affinity for FERM with K d ϭ 77.1 nM (protein concentration) than for FERM(S249D) with K d ϭ 1207 nM. Moreover, the kinetic processes of PIP 2 binding to FERM and to FERM(S249D) are also different (Fig. 4, C and D, and Table 3). The k on values of FERM to PIP 2 binding is about two times slower than FERM(S249D), and k off indicates that FERM(S249D) dissociates from PIP 2 about 10 times faster than FERM. The FERM(S249D) mutation thus has significantly altered both the PIP 2 binding affinity and kinetics, suggesting that FERM(S249D) is less competent to interact with PIP 2 than the wild-type FERM.
Open Conformation of Ezrin Bound to PIP 2 -We have performed contrast-matching SANS experiment to determine the conformational changes in d ezrin and d ezrin(T567D) upon binding to PIP 2 . The scattering of neutrons from PIP 2 becomes "invisible" in 20% D 2 O because the neutron scattering length density of the buffer matches that of the lipid. At the contrastmatching point of the PIP 2 lipid, SANS determines the conformational changes of the deuterated proteins that have sufficient coherent neutron scattering without the interference scattering from the lipid. The concept and applications of contrastmatching small angle scattering have been described elsewhere (27,50,51,60,62,67,68).  Table 1). Similarly, I(0)/c of d ezrin(T567D) in 20% D 2 O buffer is also the same as that in PIP 2 ( Fig. 5B and Table 1). Using both static light scattering and SANS, we have confirmed the molecular mass of the monomer fraction of ezrin and ezrin(T567D) (Fig. 2 and supplemental  Fig. S4). Comparing I(0)/c thus indicates that d ezrin and d ezrin(T567D) remain as monomers in PIP 2 solution, and PIP 2 does not cause oligomer state changes in ezrin. Nevertheless, in PIP 2 , the size of d ezrin or d ezrin(T567D) increases significantly when comparing with their respective closed forms in solution. The size of d ezrin expands to R g ϭ 67.4 Ϯ 1.2 Å and D max ϭ 240 Ϯ 5 Å and that of d ezrin(T567D) has also increased, with R g ϭ 68.0 Ϯ 1.0 Å and D max ϭ 240 Ϯ 5 Å (Fig. 5, C and D, and Table 1). Thus, contrast-matching SANS reveals the monomeric and open structures of d ezrin and d ezrin(T567D). Fig. 5, E and F, gives the three-dimenional shapes of open d ezrin and d ezrin(T567D) in PIP 2 ab initio reconstructed from SANS using the program DAMMIN (56). The docked atomic model is taken from the crystal structure of Sfmoesin (33), but the two antiparallel central helices are unwind and one of the helices rotates about 120 -180°. In the three-dimenional map, the center-of-mass distance between FERM and the C-ERMAD is about 180 Å, which agrees with a previous biophysical finding that the moesin ␣-helical coiled coil becomes an unfolded rod-like structure (69). In addition, the three-dimenional map shows extra density in the central hinge region between the two central helical halves. This is likely due to swivel-like motions between the FERM and   C-ERMAD about the hinge connecting the two halves of the central helices. Previously, we have shown the highly fluctuating region of a protein tends to be overestimated by the ab initio reconstruction method (70).  Fig. S5, A and B). The closed ezrin alone thus does not bind to F-actin. However, when d ezrin(T567D) is incubated with both PIP 2 and F-actin in 40% D 2 O buffer, SANS detects significant conformational changes. R g increased to 93.2 Ϯ 1.7 Å and D max to 300 Å (Fig. 6, A and B, and Table 1). Considering that 40% D 2 O is close to the contrast-matching point of PIP 2 , the deuterated d ezrin(T567D) dominates the scattering, and the detected size changes in 40% D 2 O mainly reflect the conformational changes of d ezrin(T567D). Fig. 6C presents the ab initio reconstructed three-dimenional image of the open d ezrin(T567D) in the PIP 2 ⅐ d ezrin(T567D)⅐F-actin complex, which adopts an elongated spiral shape. These results also show that PIP 2 is required for ezrin to bind to F-actin.

Open Conformation of Ezrin Bound to F-actin-We
At the contrast points other than 40% D 2 O, the hydrogenated F-actin filament contributes to scattering (Fig. 7A). The scattering from the F-actin filament or from the PIP2⅐ d ezrin(T567D)⅐F-actin complex can be considered as that from long rod-like structures with random orientations. Analyzing the small angle scattering data QI(Q) of such rodlike structures typically yields structural information about the cross-section of the filament complex (60,71). The crosssection length distribution function P c (r) of the PIP 2 ⅐ d ezrin(T567D)⅐F-actin complex are shown in Fig. 7B at three contrasts of 0, 20, and 100% D 2 O.
In 100% D 2 O, P c (r) of the PIP 2 ⅐ d ezrin(T567D)⅐F-actin complex is similar to that of F-actin alone and to the computed P c (r) of an F-actin model (see Fig. 7D). The PIP 2 ⅐ d ezrin(T567D)⅐Factin complex has similar cross-section maximum dimension (D c, max ) and cross-section radius of gyration (R c ) as the F-actin filament ( Table 2). This result suggests that the scattering comes mainly from the F-actin filament in 100% D 2 O.
In 0 and 20% D 2 O, P c (r) ratios of the PIP 2 ⅐ d ezrin(T567D)⅐Factin complex have D c, max ϭ 130 Å, and the cross-section center-of-mass distance between F-actin and d ezrin(T567D) is about 100 Å (Fig. 7B), which is less than half the length of the fully opened ezrin. This comparison suggests that the opened ezrin does not bind to the F-actin filament vertically with only the C-ERMAD domain in contact with actin. Our SPR experi-  OCTOBER 26, 2012 • VOLUME 287 • NUMBER 44 ments find that FERM binds to F-actin with high affinity (supplemental Fig. S6A), in agreement with previous reports (32,72). In addition, we find that the helical linker region of ezrin also binds to F-actin (supplemental Fig. S6B). Based on the open structure of d ezrin(T567D), the cross-section structure of the PIP 2 ⅐ d ezrin(T567D)⅐F-actin complex, and the SPR binding results, we propose a model that the opened ezrin binds longitudinally along the F-actin filament (see Fig. 8A). In this model, ezrin interacts with F-actin more extensively beside the putative C-ERMAD.

Open Conformation of Ezrin
Using the SANS data in 0 and 20% D 2 O as a constraint, we have performed rigid-body modeling. The rigid-body modeling docks the open form ezrin (from Fig. 7C) to a 20-mer F-actin with different orientations. The results indicate that the model shown in Fig. 8A fits best the SANS data in 0 and 20% D2O. Alternative models that do not fit the SANS data as well are shown in Fig. 9. Fig. 7C shows the normalized cross-section forward scattering as a function of scattering length density of the buffer background. The square root of the slope gives the mass per unit length (M L ) of the complex (see supplemental material). The M L of complex is listed in Table 2, together with that of F-actin filament. A Stuhrmann plot of the cross-section radius of gyration R gc 2 against the contrast 1/⌬ does not give a straight line, indicating that the distribution of d ezrin(T567D) and PIP 2 around the hydrogenated F-actin filament is not symmetric.
In addition, we have also performed SANS on d ezrin(S249D/ T567D) that is incubated simultaneously with hydrogenated scaffolding protein NHERF1 and with F-actin. SPR binding experiments indicate that, compared with the wild-type ezrin, a considerable fraction of this double mutant ezrin(S249D/ T567D) is capable of binding to NHERF1 (supplemental Table  S2). We thus posit that incubating d ezrin(S249D/T567D) with both NHERF1 and F-actin may trap and stabilize the open structure. SANS was performed on the incubation that contains 28.9 M hydrogenated NHERF1, 28.9 M deuterated d ezrin(S249D/T567D), and 124.4 M hydrogenated F-actin in 0, 40, and 100% D 2 O buffer.
In 40% D 2 O, in which both NHERF1 and F-actin are invisible, d ezrin(S249D/T567D) undergoes large conformational changes when compared with the closed form in solution, with R g increases to 64.2 Ϯ 1.5 Å, and D max expands to 240 Ϯ 5 Å (supplemental Fig. S7 and Table 1). The reconstructed threedimenional shape of d ezrin(S249D/T567D) adopts an L-shaped open structure (supplemental Fig. S8C). When incubating d ezrin(S249D/T567D) with NHERF1 alone at a 1:1 molar ratio, we find that d ezrin(S249D/T567D) does not have apparent conformational changes using contrast-matching SANS. Thus, d ezrin(S249D/T567D) requires the FERM domain to be occupied by NHERF1 and F-actin simultaneously to adopt an open structure.
In 40% D 2 O, the overall size of open d ezrin(S249D/T567D) in the NHERF1⅐ d ezrin(S249D/T567D)⅐F-actin complex is smaller than d ezrin(T567D) in the PIP 2 ⅐ d ezrin(T567D)⅐F-actin complex (see Table 1). The larger size of d ezrin(T567D) in complex with PIP 2 and F-actin is probably due to the scattering from the PIP 2 molecules bound to d ezrin(T567D). Alternatively, PIP 2 is more capable of extending ezrin(T567D) and stabilizing the open conformation on F-actin than NHERF1. In 0% D 2 O, all three components of the NHERF1⅐ d ezrin(S249D/T567D)⅐F-actin complex are visible to neutrons; P c (r) of the complex has a D c, max of 110 Å and an extra peak at about 100 Å as compared with F-actin filament alone (supplemental Fig. S9A). In 100% D 2 O in which d ezrin(S249D/T567D) is invisible and only the hydrogenated NHERF1 and F-actin scatter neutrons, P c (r) of the complex also has a peak at about 100 Å (supplemental Fig.  S9A). This peak is due to the contribution from NHERF1, and the maximum of the second peak in P c (r) indicates that the cross-section center-of-mass distance between the F-actin filament and NHERF1 is about 95 Å. Because the full-length of an open ezrin is about 240 -300 Å in the NHERF1⅐ d ezrin(S249D/ T567D)⅐F-actin, the short cross-section center-of-mass distance between NHERF1 and F-actin can only implicate that the extended d ezrin(S249D/T567D) binds intimately in the F-actin filament, forming extensive contacts with F-actin besides the canonical C-ERMAD (supplemental Fig. S9B).
To summarize, binding of ezrin to F-actin requires either PIP 2 or NHERF1 to be bound to the FERM domain of ezrin. Once bound to F-actin, ezrin does not stand perpendicular on the F-actin filament with only the C-ERMAD domain in contact with F-actin, as often depicted in the published cartoon pictures. Instead, our model shows that once bound to F-actin, ezrin forms extensive contacts with F-actin.

DISCUSSION
We have determined the structural changes of the full-length ezrin upon binding to PIP 2 and to F-actin using contrast variation SANS. Using gel filtration, we show that the conformational opening of ezrin in response to PIP 2 binding is cooperative, but the cooperativity is disrupted by the phospho-mimetic mutation S249D in the FERM domain. Using SPR, we find that the S249D mutation weakens the binding affinity of FERM domain for PIP 2 and changes the kinetics of FERM to PIP 2 binding. Furthermore, our study indicates that ezrin binding to F-actin requires the simultaneous binding of ezrin to either PIP 2 or in the case of the double mutant ezrin(S249D/T567D) to the scaffolding protein NHERF1. According to cross-section analysis of the SANS data, the cross-section center-of-mass distance between F-actin and the bound ezrin is significantly shorter than the full length of the activated ezrin, suggesting that the opened ezrin is collapsed on F-actin and forms extensive contact with the fila-

Open Conformation of Ezrin
ment. This model of the ezrin/F-actin interaction is thus different from that generally depicted in the published literature, in which the activated ezrin binds vertically with only the C-ERMAD in contact with F-actin.
We find that the opening of ezrin or ezrin(T567D) in response to PIP 2 binding is cooperative. Cooperative binding warrants a robust regulation of a biochemical process in response to a ligand or an effector molecule. A cooperative regulation of ezrin activation by PIP 2 may suggest the need for acute spatial-temporal regulation of ezrin functions in the cellular context. Although PIP 2 includes only about 0.5-1% of the cell membrane phospholipids, PIP 2 is highly localized in a variety of subcellular compartments and microdomains due to local synthesis and sequestering of PIP 2 (73)(74)(75). PIP 2 is particularly localized in the apical membrane of epithelial cells, in lamellipodia, in microvilli, and at the cell junctions, in which ezrin plays important roles in assembling and maintaining these specialized subcellular structures (14, 76 -79). An effective spatial-temporal regulation of the assembly and disassembly of the protein complexes is required for the dynamic turnover of these subcellular structures. As a result of cooperative activation of ezrin by PIP 2 , PIP 2 and ezrin may contribute significantly to the dynamics of these specialized subcellular structures. Also, there is increasing evidence that the activated ezrin binds to target proteins that trigger the subsequent propagation of downstream allosteric binding signals in the membrane cytoskeleton (15,80,81). For instance, ezrin activates NHERF1 and induces long range allostery in NHERF1 so as to strengthen the interactions of NHERF1 with transmembrane proteins and other signaling proteins (24,27,70). In turn, NHERF1 also allosterically activates other proteins, such as the scaffolding protein PDZK1 that binds to downstream targets for the assembly microvillus structures on the cell surface (15,81). Ezrin is a crucial player in protein dynamics, long range allostery, and signal transduction (70,82).
Altering the subtle cooperativity of ezrin activation in response to PIP 2 binding could have substantial impact on cellular functions. Indeed, we find that ezrin phosphorylated at Ser-249 or the ezrin(S249D) mutant is no longer localized in the apical membrane or at the cell-cell junctions in polarized epithelial cells as the wild-type ezrin, but it is largely degraded or forms clustered aggregates in the cytoplasm. Cells expressing the ezrin(S249D) mutant show altered morphology and weakened cell-cell adhesion as compared with cells expressing the wild-type ezrin. Furthermore, the localization of adherens junction marker proteins, E-cadherin and ␣-catenin, are diffuse and reduced at the cell-cell junctions in cells expressing ezrin(S249D) mutants. It remains to be determined if other phosphorylation sites, such as Tyr-  145 (83), in the FERM domain can influence the PIP 2 binding behavior.

Open Conformation of Ezrin
The S249D mutation affects both the affinity and kinetic rate constants of FERM to PIP 2 binding, even though this Ser-249 is outside the two patches of positively charged residues that are necessary for FERM binding to PIP 2 (35,38,84). Thus, mutating the Ser-249 to the negatively charged Asp may affect the long range electrostatic field on the surface of the FERM domain. Electrostatic interactions are likely to be the basis of FERM to PIP 2 binding, which are found in many other protein domains that bind to phosphoinositide lipids with limited structural specificity (74,85).
Without PIP 2 , the phospho-mimic mutants adopt essentially closed conformation in solution as the autoinhibited wild-type ezrin. However, a considerable fraction of the phospho-mimetics is active and capable of binding to NHERF1 when compared with the wild-type protein. It is likely that the phospho-mimetic mutants are more dynamic so that a small fraction of the mutants is open for a period of time and becomes competent to bind to NHERF. The SANS results on the T567D mutant provide an alternative view from a previous electron microscopy study that finds this mutant to be completely open (86). This is because SANS samples an ensemble of molecules, whereas EM selectively looks at a particular population of molecules. A small fraction of activated and open ezrin(T567D) may not contribute significantly to the ensemble averaged R g and D max values measured by SANS. It would be interesting to determine the kinetics and dynamics of ezrin opening in future studies (87).
Our results show that ezrin binding to F-actin requires the simultaneous binding of ezrin to either PIP 2 or to the scaffolding protein NHERF1. Furthermore, the neutron scattering cross-section analyses of both the PIP 2 ⅐ d ezrin(T567D)⅐F-actin and the NHERF1⅐ d ezrin(S249D/T567D)⅐F-actin complexes suggest that the opened ezrin does not stand vertically on the F-actin filament. Instead, we propose a model that ezrin collapses on F-actin forming extensive contacts. Previous studies (32) and our own binding experiment show that the FERM domain has the capability to interact with F-actin strongly. The FERM domain is likely to also bind F-actin and is anchored in the actin filament.
Our structural model of a collapsed ezrin spans about 10 actin monomers on the F-actin filament. This model corroborates the findings from previous biochemical studies that ezrin binding to actin is saturable with a 1:8 -10 molar ratio (32,88). The model we presented here indicates that ezrin acts a spatial ruler on the F-actin filament.
In cells, through extensive contacts with F-actin, ezrin can bring the cell membrane close to the underlying F-actin cytoskeletal network. Indeed, an electron tomography study of the membrane skeleton reveals that the actin filaments are closely associated with the cytoplasmic surface of the plasma membrane within 10.2 nm (89). Ezrin and NHERF1 are distributed along almost the entire microvillus structure (14). The growing F-actin filaments can have intimate interactions with the lipid membrane and support the expanding cell membrane. In this scenario, ezrin may play active roles in regulating the adhesion and tension between the membrane and the cytoskeleton, as required for forming many subcellular structures and for regu-lating many transmembrane proteins at the cell surface (16, 90 -92).