Phosphatidylinositol 4,5-Bisphosphate Clusters the Cell Adhesion Molecule CD44 and Assembles a Specific CD44-Ezrin Heterocomplex, as Revealed by Small Angle Neutron Scattering*

Background: The mechanism by which the conserved CD44 cytoplasmic tail (CD44ct) functions is not well understood. Results: The disordered CD44ct interacts with FERM only in the presence of phosphatidylinositol 4,5-bisphosphate (PIP2). Conclusion: PIP2 clusters CD44ct and facilitates the assembly of a specific CD44-Ezrin heterotetramer complex. Significance: The study reveals the role of PIP2 in clustering CD44 and in assembling multimeric CD44-Ezrin signaling complexes. The cell adhesion molecule CD44 regulates diverse cellular functions, including cell-cell and cell-matrix interaction, cell motility, migration, differentiation, and growth. In cells, CD44 co-localizes with the membrane-cytoskeleton adapter protein Ezrin that links the CD44 assembled receptor signaling complexes to the cytoskeletal actin network, which organizes the spatial and temporal localization of signaling events. Here we report that the cytoplasmic tail of CD44 (CD44ct) is largely disordered. Upon binding to the signaling lipid phosphatidylinositol 4,5-bisphosphate (PIP2), CD44ct clusters into aggregates. Further, contrary to the generally accepted model, CD44ct does not bind directly to the FERM domain of Ezrin or to the full-length Ezrin but only forms a complex with FERM or with the full-length Ezrin in the presence of PIP2. Using contrast variation small angle neutron scattering, we show that PIP2 mediates the assembly of a specific heterotetramer complex of CD44ct with Ezrin. This study reveals the role of PIP2 in clustering CD44 and in assembling multimeric CD44-Ezrin complexes. We hypothesize that polyvalent electrostatic interactions are responsible for the assembly of CD44 clusters and the multimeric PIP2-CD44-Ezrin complexes.

for CD44 ligand binding, for the sensing and response of cells to their microenvironment, and for the localization and assembly of signaling complexes that regulate cell motility, migration, survival, growth, and proliferation.
The functions of Ezrin and the other ERM proteins are known to be regulated by a conformational autoinhibition mechanism, with the inactive protein being held in a closed conformation by head to tail intramolecular interactions (21,22). Binding to the phosphatidylinositol 4,5-bisphosphate (PIP2) lipid disrupts the autoinhibition and activates ERM proteins (23). In the activated ERM proteins, the N-terminal 4.1 Ezrin-Radixin-Moesin (FERM) domain binds to target proteins, which include the multi-PDZ scaffolding protein NHERF1 that in turn binds to a variety of membrane receptors and ion channels, whereas the C-terminal tail binds to cytoskeletal actin. ERM proteins organize a variety of signaling events at the interface between the cell membrane and cytoskeleton (22).
Previous biochemical studies have shown that PIP2 is involved in the interaction of CD44 with the ERM proteins (24). Biochemical and x-ray crystallography studies have determined that the FERM domain, which mimics an activated ERM protein, binds directly to positively charged amino acid residues in the juxtamembrane peptide of the cytoplasmic tail of CD44 (16,17). The role of PIP2 in CD44-Ezrin association has generally been attributed to the activation of ERM proteins. Here we report that the entire cytoplasmic tail of CD44 of 73 amino acid residues (CD44ct), which adopts a disordered "collapsed coil" conformation in solution, is unable to bind directly to the FERM domain of Ezrin (which mimics an activated Ezrin). Instead, PIP2 is necessary for CD44ct to form a complex with FERM. Using selective deuteration and contrast variation small angle neutron scattering (SANS), we show that PIP2 mediates the assembly of a discrete heterotetramer complex of CD44ct both with the FERM domain of Ezrin and with the full-length Ezrin. Additionally, PIP2 causes CD44ct to aggregate. These results reveal an important role of PIP2 in clustering CD44ct and in the assembly of multimeric CD44-Ezrin complexes by polyvalent electrostatic interactions.

MATERIALS AND METHODS
Protein Expression and Purification-The proteins used in this study were expressed in bacterial cells. The entire cytoplasmic domain including amino acid residues 293-365 of mouse CD44 transcript variant 3 (CD44ct) was subcloned into a pET-32a vector that expresses a fusion protein of a His 6 plus thioredoxin tag at the N-terminal end of CD44ct. A TEV cleavage site was cloned between the thioredoxin tag and CD44ct. The FERM domain of Ezrin and full-length Ezrin was subcloned into a TOPO-pET151 vector as described previously (25)(26)(27)(28)(29). The vectors were transformed into Rossetta2 (DE3) competent cells. The protein purification procedures are the same as previously described (25)(26)(27). After protein purification, the fusion tag was cleaved by TEV and removed by gel filtration. The protocol for expressing and purifying deuterated proteins has been described previously (27,30,31). The nonexchangeable deuterium content of the purified deuterated proteins range from 0.65 to 0.75, as determined by matrix-assisted laser desorption time of flight mass spectrometry at the Columbia University Protein Core Facility.
Protein Concentration Determination-The concentrations of all proteins except CD44ct were determined by UV absorption at 280 nm, using the extinction coefficient calculated from ProtParam on the ExPASy Server (32). The extinction coefficient of the cytoplasmic tail of CD44 (CD44ct) is zero at 280 nm. The concentration of CD44ct was thus measured at 205 nm, using the Protein concentration (mg/ml) ϭ (Absorbance at 205 nm)/31 (33,34).
Static and Dynamic Light Scattering Experiments-Static light scattering experiments were performed using a DynaPro Nanostar (Wyatt Technology Corporation) with a laser of wavelength 824.7 nm at a fixed 90°scattering angle. Before light scattering experiments, the sample was centrifuged at 10,000 rpm for 5-10 min. Protein concentrations were varied from 0.5 to 2 mg/ml during light scattering measurements. Light scattering experiments were performed at 20°C.
CD Spectroscopy Experiments-CD experiments were performed with an Aviv 400 spectropolarimeter (Aviv Biomedical, Inc., Lakewood, NJ). The protein samples were dissolved at 0.1-0.2 mg/ml in buffer containing 20 mM phosphate (pH 7.5), 150 mM NaCl, and 1 mM DTT. Quartz cuvettes of 1-mm pathlength (Hellma USA, Plainville, NY) were used for CD experiments. The CD measurements performed at 20°C. The secondary structure content of each spectrum was calculated using a multilinear regression method that does not depend so much on accurate determination of protein concentration.
Gel Filtration Analysis of CD44ct Binding to Ezrin in Solution and in PIP2-A Superdex 200 10⅐300 GL gel filtration column was used to analyze the binding of CD44ct with Ezrin(T567D) or with FERM in solution and in PIP2, respectively. 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. To determine the binding in solution, 50 M Ezrin(T567D) or FERM is incubated with equal molar CD44ct for 1 h. The complexes were prepared by first mixing 50 M CD44ct with FERM or Ezrin(T567D) at 1:1 molar ratio, and then PIP2 at 10 molar ratio was added. The complex was incubated on ice for 1 h before the gel filtration to examine the shift in elution peak position.
Pulldown Experiments to Determine CD44ct Binding to Ezrin in Solution and in PIP2-The pulldown experiments were performed by incubating 15 l of 10 M of His 6 -Trx-CD44ct with 15 l of 10 M FERM, without PIP2 or in the presence of 10 molar ratio PIP2 in binding buffer, containing 20 mM imidazole, 20 mM phosphate (pH 7.0), 300 mM NaCl for 1 h on ice. Approximately 15 l of Ni-Magbeads (Genscript) were added to the incubation. After the beads were washed with binding buffer three times, the protein or protein complex was eluted in 100 l of buffer containing 500 mM imidazole, 20 mM phosphate buffer (pH 7.0), and 300 mM NaCl. The eluted protein or protein complex was analyzed by SDS-PAGE.
Isothermal Titration Calorimetry-Isothermal titration calorimetry (ITC) experiments were run in a MicroCal TM Auto-iTC 200 system at 25°C. All proteins were dialyzed for 12-18 h against the ITC buffer containing 10 mM HEPES (pH 7.5), 300 mM NaCl, 0.5 mM EDTA, and 0.5 mM ␤-mercaptoethanol. All samples were degassed for 10 min prior to experiments, and all measurements were carried out in duplicate. The heat of ligand dilution was obtained from the average heat of the last few injections after complete saturation of binding protein. This value was subtracted from all of the injection points during data analysis. All data were fit to a single site binding model using Bindworks to yield the stoichiometry (n), the association constant (K a ), and the enthalpy change (⌬H) for the binding reaction.
Analytical Ultracentrifugation Analysis of the PIP2⅐CD44ct⅐ FERM Complex-Both sedimentation velocity and sedimentation equilibrium experiments were performed on the PIP2⅐CD44ct⅐FERM complex. Before the ultracentrifugation experiment, the complex was prepared by first mixing 50 M CD44ct with 500 M PIP2, and 50 M FERM was added. The complex was purified by gel filtration using a Superdex 200 column. Samples were loaded into a six-chamber centerpiece at 2.1 M concentration and centrifuged at two rotor speeds (10,300 and 18,000 rpm) at 20°C and allowed to reach equilibrium in an Optima XL-I centrifuge (Beckman Coulter, Fullerton, CA). We used an absorbance optical data acquisition system and measured the sedimentation equilibrium data at 260-nm wavelength. Buffer viscosity, density, and specific volume were calculated using program SEDNTERP (35). The program SEDFIT (36) was used to sort and prepare equilibrium data for further analysis in program SEDPHAT (37). The sedimentation equilibrium data at two speeds were fit globally in SEDPHAT using a model that describes a single species of interacting systems.
Solution Small Angle X-ray Scattering Experiments-Solution x-ray scattering experiments were performed at Beamline X9 at the National Synchrotron Light Source, Brookhaven National Laboratory (38). Protein samples were loaded in a 1-mm diameter glass capillary. Triplicates of scattering data were acquired by flowing 20 l of protein sample or buffer during a 30-s exposure. Analysis of the simultaneous SAXS⅐WAXS data acquisition was performed using the program pyXS (39). Standard Guinier plots were calculated to identify aggregates (40).
SANS Experiments and Data Analyses-SANS experiments were performed on the extended Q range SANS instrument at the Spallation Neutron Source located at Oak Ridge National Laboratory (41). All measurements used a sample to detector distance of 4 m. Two wavelength settings were used: 60 Hz with a wavelength, , band of 2.5-6.1 Å and 30 Hz (frame-skipping mode) with two wavelength bands of 2.5-6.1 Å and 9.4 -13.4 Å. The former configuration provides a useful Q range (the wave vector transfer, Q ϭ 4 sin()/, where 2 is the scattering angle) of ϳ0.009 to ϳ0.22 Å Ϫ1 , whereas the latter provides additional low Q data (down to ϳ0.005 Å Ϫ1 ) with the same practical upper limit. The choice of configuration was determined by the expected size of the particles being studied.
Before SANS experiments, the proteins or protein⅐lipid complexes were dialyzed against buffers of different D 2 O concentrations. The samples were loaded into 1-mm-path length circular quartz cells (Hellma USA). SANS experiments were performed at 10°C. More details of the SANS experimental conditions and the SANS data reduction are described elsewhere (31,42). SANS data reduction followed standard procedures that are implemented in MantidPlot to correct for dark current (background radiation and electronic noise), the detector sensitivity, and the scattering contribution from the solvent and empty cells before being azimuthally averaged to produce I(Q) versus Q. The data were scaled into absolute units of cm Ϫ1 using a calibrated standard (43).
The length distribution function P(r), radius of gyration R g , the forward scattering intensity I(0) that is the neutron scattering intensity extrapolated to Q ϭ 0 Å Ϫ1 , and the maximum dimension D max were calculated from the scattering data using both the program GNOM (44). The molecular mass of a protein or protein complex can be determined by contrast variation SANS. Based a reference by Jacrot and Zaccai (45), an equation was derived to show that I(0) 0.5 versus the neutron scattering length density of the buffer o yields (27), where N is the number of the complexes in a volume of 1 cm 3 , The partial specific volume was assumed to be ϭ 0.73 cm 3 /g. Other details of contrast variation small angle scattering have been described previously (46,47). The three-dimensional shapes of "dummy bead" coordinates were generated using the program DAMMIN (48). Multiple calculations were performed using DAMMIN, and the generated 10 -20 structures were averaged and filtered using the program DAMAVER and DAMFILT (49). The three-dimensional density map was generated from the averaged coordinates using the program Situs (50). The fitting and docking of the high resolution structure to the density map were performed using Situs or UCS Chimera (51). For complexes of multiple components, the multiphase program MONSA restores the three-dimensional shape of the complex (48). The fitting and docking of the high resolution structure to the density map were performed using Situs (50) or UCSF Chimera (51).

The Cytoplasmic Tail of CD44 Is Disordered and Adopts a "Collapsed Globule" Conformation
The entire cytoplasmic tail of CD44 (CD44ct) has 72 amino acid residues (Fig. 1A). Because the extinction coefficient of CD44ct at 280 nm is zero, it is difficult to detect CD44ct alone by gel filtration. We thus have first analyzed the fusion protein of CD44ct with an N-terminal thioredoxin tag (Trx-CD44ct). Gel filtration shows that Trx-CD44ct migrates as a single peak (Fig. 1B). Static light scattering shows that the measured molec-ular mass of Trx-CD44ct is close to the theoretical molecular mass of a monomer ( Fig. 1C and Table 1). The gel filtration and light scattering results thus indicate that Trx-CD44ct is a monomer in solution at protein concentration lower than 2 mg/ml. Above protein concentration of 2 mg/ml, static light scattering indicates that Trx-CD44ct has weak self-association (Fig. 1C).
A comparison of the SANS data from the deuterated CD44ct ( d CD44ct) and deuterated fusion protein d Trx-CD44ct provides an estimation of the shape and the oligomer state of CD44ct in solution ( Fig. 1, E-H). SANS shows that the R g and D max of d Trx-CD44ct are larger than those d CD44ct (Table 1 and Fig. 1G). The three-dimensional shape of d CD44ct reconstructed from SANS can be docked into the three-dimensional envelope of monomeric d Trx-dcD44 (Fig. 1H). These comparisons suggest that d CD44ct is a monomer in solution.
CD spectroscopy shows that CD44ct is largely disordered without apparent secondary structure features (Fig. 1D). The disorder in CD44ct is also exhibited in the SANS data. The relatively large R g and D max of the 72-residue d CD44ct as compared with a well folded theoredoxin of 109 residues indicate that d CD44ct is not a compact globular structure ( Fig. 1G and Table 1). The docking experiment also suggests that CD44ct is not as compact as the thioredoxin globular domain amino acid residues (Fig. 1H).
The Kratky plot from SANS can be used to show whether a protein is a compact globular structure, a random coil-like chain, or a partially folded protein (52,53). The Kratky plot of a folded globular protein is a bell-like curve because of the presence of a clear interface between the measured particle and surrounding solvent, whereas that of a random coil chain tends to increase linearly with Q (52). The plateau behavior of the Kratky plot suggests that, in solution, d CD44ct is neither a well folded globular protein nor a true random coil chain-like structure (Fig. 2C). Thus, altogether, the static light scattering, CD, and SANS results indicate that CD44ct is a monomer that is somewhat collapsed but without significant secondary structure.

Binding to PIP2 Unfolds CD44ct and Causes CD44ct to Aggregate
Upon adding PIP2 to CD44ct, the CD spectrum shows that CD44ct remains largely disordered with no apparent changes in secondary structure (Fig. 1D). SANS, performed at the contrast-matching point of PIP2 in 20% D 2 O, indicates that d CD44ct forms aggregates in PIP2 solution (Fig. 2). The protein concentration normalized forward scattering I(0)/c of d CD44ct in PIP2 solution is ϳ14 times higher than that of d CD44ct in solution, suggesting that ϳ14 d CD44ct molecules reside in the aggregates ( Fig. 2A and Table 1). R g and D max of d CD44ct in PIP2 are significantly larger than that of d CD44ct in solution (Table 1 and Fig. 2B), also indicating that d CD44ct aggregates upon binding to PIP2.
Unlike d CD44ct alone in solution (Fig. 2C), the Kratky plot of d CD44ct in PIP2 has a peak at QR g Ϸ 1.8 Å Ϫ1 , suggesting that the d CD44ct aggregates are globular-like clusters (Fig. 2D). However, at QR g Ͼ 3.5, in which SANS probes the internal structure of the aggregates, the Kratky plot increases with Q, a behavior that is typical of a random coil. The SANS results thus indicate that upon binding to PIP2, d CD44ct unfolds into an open and random coil-like conformation and that the opened d CD44ct aggregates into large globular clusters. Fig. 3D shows the gel filtration chromatogram of CD44ct alone in solution, CD44ct dissolved in the 1,2-diheptanoyl-snglycero-3-phosphocholine (DHPC) at 1:100 molar ratio, and CD44ct in PIP2 at 1:10 molar ratio. In solution or in DHPC micelles, CD44ct migrates as a monomer, suggesting that DHPC does not cause CD44ct to aggregate. However, in the presence of PIP2, CD44ct migrates as large aggregates, which is  consistent with the contrast-matching SANS results. Comparing the gel filtration results of CD44ct in PIP2 and in DHPC2 thus indicates that the aggregation of CD44ct is specific to PIP2.

PIP2 Is Necessary for CD44ct Binding to Ezrin
Using gel filtration and pulldown experiments, we have analyzed the binding of CD44ct to the full-length Ezrin (Fig. 3A). A phosphomimetic Ezrin(T567D) mutant was selected in this experiment, based on previous findings that a phosphomimetic Ezrin(T567D) mutant is more susceptible to adopt an open conformation than the wild-type full-length Ezrin upon binding to PIP2 and is thus more active in binding to target proteins than the wild-type Ezrin (23,31).
The gel filtration results show that CD44ct does not interact with Ezrin(T567D) in solution but forms a complex with Ezrin(T567D) only in the presence of PIP2 (Fig. 3A). This result is consistent with previous findings that PIP2 is involved in the interaction of Ezrin with CD44 (24). Previous studies have attributed the role of PIP2 in the CD44-Ezrin interaction as disrupting the head to tail autoinhibition in ERM proteins and opening the molecular conformation of ERM. The exposed FERM domain of an active ERM protein is generally considered to be capable of binding directly to CD44.
Nevertheless, to our surprise, we find that in solution CD44ct does not bind to the FERM domain of Ezrin, which is a mimetic of active Ezrin (54). Gel filtration, pulldown, and ITC experiments all show that CD44ct does not interact with the FERM domain in solution (Fig. 3, B, D, and E). Instead, only in the presence of PIP2, do CD44ct and FERM form a complex, as shown by gel filtration and pulldown experiments (Fig. 3, C and  D). These results suggest that PIP2 not only serves as a conformational opener for the full-length Ezrin but is also necessary to mediate the interaction of CD44ct with FERM or with the fulllength Ezrin.

Contrast Variation SANS Reveals the Assembly of a 2CD44ct⅐2FERM or a 2CD44ct⅐2Ezrin Heterotetramer Complex in PIP2
The Stoichiometry and the Structure of the PIP2⅐ d -CD44ct⅐ d FERM Complex-To determine how PIP2 mediates CD44-FERM interaction, we have first performed contrast variation SANS on d CD44ct in complex with deuterated FERM ( d FERM) and with PIP2 (Fig. 4A). With selective deuteration of a subunit in a multicomponent complex, contrast variation SANS can resolve the conformation of each component, as well as the stoichiometry and architecture of the whole complex (27,31,46). The molar ratio of incubation is PIP2⅐ d CD44ct⅐ d -FERM ϭ 10:1:1 based on our gel filtration and pulldown experiments, which show that the interaction of CD44ct with FERM is more complete at this molar ratio of incubation than at lower molar ratios of PIP2.
SANS shows that the PIP2⅐ d CD44ct⅐ d FERM complex has well defined P(r) functions at each contrast (Fig. 4B). The normalized forward scattering intensity I(0) versus the neutron scattering length density of the buffer o shows a straight line (Fig. 4C). These results suggest that d CD44ct and d FERM form a discrete complex in the presence of PIP2. According to Equation 1, the molecular mass of a protein complex can be determined by contrast variation SANS from the slope of Fig. 4C to be Mw ϭ 107,826 Ϯ 2,770 g/mol, which corresponds to the molecular mass of a 2 d CD44ct⅐2 d FERM heterotetramer plus 8 -9 PIP2 lipid molecules ( Table 1).
The Stuhrmann plot, defined as Rg 2 plotted versus 1/⌬ , where ⌬ is the neutron scattering length density contrast of the whole complex against solvent, can be used to estimate the distribution of scattering length density of the complex (55). The dependence of the R g 2 on the contrast is approximated by the following expression, where R s is the R g of the complex at infinite contrast, A relates to the distribution of density as a function of distance from the center of mass, and B relates to the distance separating the center of mass of the particle shape from the center of mass of neutron density. A straight line with a positive slope in the Stuhrmann plot (Equation 2), as shown in Fig. 4D (Fig. 4B). The three-dimensional shape of the d CD44ct⅐ d FERM complex was reconstructed from SANS using the program DAMMIN (48) (see Fig. 8A). A 2-fold symmetry was imposed when reconstructing the three-dimensional shape of the complex based on the information obtained from molecular mass analysis of the complex.
The Stoichiometry and the Structure of the PIP2⅐ d -CD44ct⅐ h FERM Complex-We have then performed contrast variation SANS on the complex of d CD44ct in complex with the hydrogenated FERM ( h FERM) in PIP2 at the incubation molar ratio of PIP2⅐ d CD44ct⅐ h FERM ϭ 10:1:1 (Fig. 5). Again, the well behaved P(r) functions and the straight line of the normalized I(0) 0.5 versus o plot suggest that the PIP2⅐ d CD44ct⅐ h FERM complex is a discrete complex with defined molecular mass (Fig. 5, B and C). The slope of Fig. 5C yields the molecular mass of the complex to be Mw ϭ 97,772 g/mol, corresponding to the molecular mass of two h FERM and two d CD44ct plus 7-8 PIP2 lipid molecules in the complex (Fig. 5C and Table 1). The number of PIP2 lipids in the PIP2⅐ d CD44ct⅐ h FERM complex is consistent with that in the PIP2⅐ d CD44ct⅐ d FERM complex within experimental error.  Table 1  With selective deuteration, the SANS data at different contrasts reveal the structures of PIP2⅐ d CD44ct, PIP2⅐ h FERM, and the whole PIP2⅐ d CD44ct⅐ h FERM complex, respectively. In 40% D 2 O buffer that is the contrast-matching point of hydrogenated FERM, the Kratky plot shows that PIP2⅐ d CD44ct adopts an open, random coil-like conformation (Fig. 5D). In 100% D 2 O at the contrast point of d CD44ct, the three-dimensional shape of the PIP2⅐ h FERM component indicates that h FERM is a dimer, and the Kratky plot indicates that PIP2⅐ h FERM is a multidomain entity (Fig. 5D). In 0% D 2 O in which all components in the complex contribute to scattering, the bell-like Kratky plot suggests a compact globular structure of the whole PIP2⅐ d CD44ct⅐ h FERM complex (Fig. 5D).
In addition to SANS, we have performed light scattering on the PIP2⅐CD44ct⅐FERM complexes (Fig. 6). The dynamic light scattering data suggest that both the PIP2⅐CD44ct and PIP2⅐CD44ct⅐FERM complex are monodispersed, suggesting that both types of complexes are discrete entities (Fig. 6, A and  B, inset, and the % polydispersity values). The molecular mass of PIP2-CD44ct complex, as measured from static light scattering is 200 kDa. The hydrodynamic radius R h of the PIP2⅐CD44ct complex, as measured from dynamic light scattering, is 72 Ϯ 1 Å. For the PIP2⅐CD44ct⅐FERM, the molecular mass is 133 kDa, and R h is 65 Ϯ 2 Å. Additionally, analytical ultracentrifugation experiments indicate that complex migrates as a discrete spe-cies. The best fit of the sedimentation velocity data of the major peak (Fig. 7A) gives a frictional ratio of f/f o ϭ 1.359. The higher value of f/f o suggests that the sedimented species likely assumes an elongated shape. Sedimentation equilibrium experiments were performed at two speeds, 10,300 and 18,000 rpm. The sedimentation equilibrium data could fit into a "single species of interacting system" model and gave a molecular mass of ϳ134.9 Ϯ 6.7 kDa (Fig. 7B). The size and the molecular mass of the complex measured from light scattering and analytical ultracentrifugation experiments are larger than the SANS result. This is because light scattering and analytical centrifugation experiments measure the protein-lipid complex, whereas contrast-matching SANS measures protein assembly in the protein-lipid complex.
It is of interest to point that when preparing the protein/lipid complex for the light scattering and analytical centrifugation experiments, we first mixed the PIP2 and CD44ct at a 10:1 molar ratio, and FERM was added at 1 molar ratio. The complex was incubated overnight and separated by gel filtration before the centrifugation experiments. The sequence of mixing the different components for centrifugation experiments is different from the sample preparation for SANS experiments when we first mixed the CD44ct and FERM at equal molar ratio, and PIP2 of 10 molar ratio was added to form the complex. With either method of sample preparation, similar molecular mass  Table 1  and size of the complex are obtained. These results suggest that the PIP2-CD44ct aggregate is reversible and that the PIP2-CD44ct aggregates dissociate on the time scale of incubation to form a stable tetramer with FERM. The exact kinetics of PIP2-CD44ct association and dissociation is a subject for future investigation.
The three-dimensional shape of the PIP2⅐ d CD44ct⅐ h FERM complex was reconstructed from SANS using the program MONSA (48) at different contrasts, assuming that the complex has three phases in terms of neutron density distribution (Fig. 8). A 2-fold symmetry was imposed when reconstructing the three-dimensional shape of the complex, based on the stoichiometry and symmetry information obtained from the Stuhrmann plot and the molecular mass analyses. The recon-structed three-dimensional shape of the PIP2⅐ d CD44ct⅐ d FERM complex is shown in Fig. 8B.
Comparing the three-dimensional shapes of the PIP2⅐ d -CD44ct⅐ d FERM and PIP2⅐ d CD44ct⅐ h FERM shows that the two differently deuterated complexes have similar overall architecture (Fig. 8). Further, the three-dimensional shape of PIP2⅐ d CD44ct⅐ h FERM reveals the internal structure and the relative position of each component in the heterotetramer complex. In this heterotetrameric complex, two h FERM domains, shown in green, assemble into a dimer that are flanked by two d CD44ct molecules shown in gold. The dimeric FERM is confirmed by the SANS data from the PIP2⅐ d CD44ct⅐ h FERM complex at 100% D 2 O, which is the contrast-matching point of the deuterated d CD44ct component (Fig. 5E).    and in PIP2 and dCD44 (Fig. 9). All the SANS experiments shown in Fig. 9 were performed in 20% D 2 O, which is the contrast-matching point of PIP2. The SANS data of d Ezrin(T567D) in solution and in PIP2 are taken from our previous publication (31). In that study, we showed that Ezrin or Ezrin(T567D) adopts a closed conformation, whereas the PIP2-bound Ezrin or Ezrin(T567D) adopts an open and monomer conformation. We also showed that upon binding to PIP2, Ezrin(T567D) is more likely to adopt an open conformation than the wild-type Ezrin. We thus chose d Ezrin(T567D) in this study for determining the effects of CD44ct and PIP2 on the assembly of Ezrin.
Comparing the protein concentration normalized forward scattering I(0)/c indicates that the PIP2⅐ d CD44ct⅐ d Ezrin-(T567D) complex has a molecular mass 2.15 times that of d Ezrin(T567D) in solution, whereas I(0)/c of Ezrin(T567D) in PIP2 is approximately the same as that of d Ezrin(T567D) in solution ( Fig. 9A and Table 1

DISCUSSION
Our results first show that PIP2 binding to CD44ct unfolds CD44ct, and the unfolded CD44ct molecules either cluster into large aggregates among themselves or assemble with Ezrin into a specific heterotetramer complex. The study thus reveals an autoregulation mechanism in the cytoplasmic tail of CD44, as well as the important roles of PIP2 in CD44 clustering.
Second, the results further reveal that PIP2 assembles CD44ct and Ezrin into a heterotetramer complex. Previous studies have shown Ezrin and other ERM proteins are regulated by a self-inhibition mechanism, with the inactive ERM protein being held in a closed conformation by head to tail intramolecular interactions (21, 22, 56 -58). Binding to PIP2 and phosphorylation disrupt the autoinhibition and activate ERM proteins (23). It is generally accepted that in the activated ERM proteins, the exposed FERM domain binds directly to CD44. In addition to CD44, the FERM domain of activated Ezrin also binds to multi-PDZ scaffolding protein NHERF1 that in turn binds to a variety of transmembrane receptors and ion transport proteins, whereas the exposed actin-binding site interacts with cytoskeletal actin (25)(26)(27)(28)(29)(30). Previous studies also show that the basic residues in the juxtamembrane region of CD44ct are involved in direct binding to the FERM domain of ERM proteins (16,17). Indeed, we find that a truncated CD44 juxtamembrane peptide can bind directly to the FERM domain. Nevertheless, our experiments show that the entire CD44ct does not bind to FERM. Instead, PIP2 is required for the entire CD44ct tail to form a complex with FERM. Moreover, SANS shows that in the PIP2-CD44ct-FERM complex, PIP2 clusters are sandwiched between CD44 and FERM. Thus, PIP2 molecules are involved in mediating the interaction of CD44 with FERM in the heterotetramer assembly shown in Fig. 10.
The above results of PIP2 mediating the assembly of multimeric CD44ct-FERM complexes may be explained by earlier findings that the polybasic residues in a peptide sequester PIP2

Assembly of CD44 and Ezrin Multimeric Complexes by PIP2
MARCH 6, 2015 • VOLUME 290 • NUMBER 10 into clusters (59,60) and that PIP2 can also sequester into clusters of transmembrane proteins that possess positively charged polybasic residues in the juxtamembrane (61). Polyvalent electrostatic interactions are responsible for such clustering of PIP2 and cytoplasmic tails of transmembrane proteins (61). The juxtamembrane region of CD44ct contains a stretch of multiple basic residues (Fig. 1A), similar to those basic residues in the juxtamembrane region of syntaxin-1A (61). The FERM domain of ERM proteins possesses two PIP2-binding motifs that are also composed of multiple basic residues (23,31,62,63). We hypothesize that the negatively charged PIP2 molecules function as molecular fasteners that bridge specific heterocomplex formation of CD44ct with Ezrin by polyvalent ionic interactions. Future experiments could use mutagenesis and high resolution structural studies to test this hypothesis.
In cells, PIP2 is localized in the microdomains of the plasma membranes, in lamellipodia of migrating cells, as well as in the apical membrane and at the cell-cell junctions of polarized epithelial cells (60, 64 -70). CD44 is also localized in these microdomains of cells. PIP2 is involved in the spatial-temporal regulation of the assembly and disassembly of the transmembrane protein complexes (71), which are required for effective cellular sensing and response to extracellular ligand binding and for regulating the cellular dynamics of cell adhesion, migration, and turnover of subcellular structures (58). Our findings point to the important roles of the PIP2 lipid in dynamically regulating CD44 conformation and clustering and in mediating the assembly of discrete multimeric complex of CD44 with Ezrin. Such membrane multimeric complexes of CD44 and Ezrin may be involved in the assembly of actin cytoskeleton-dependent receptor clusters at the cell surface (18) or in virus-host cell interactions (72).
The regulation of CD44 activity is likely to be multifaceted. CD44 is enriched in the cholesterol-rich lipid raft microdomains (19). Studies have shown that the lipid raft-associated CD44 is inactive to bind to Ezrin. Lipid raft thus negatively regulates cell migration (73,74). Our study shows that PIP2 activates CD44 and mediates clustering of CD44 and the interaction of CD44 with Ezrin. Thus, it is likely that PIP2 and the cholesterol-rich lipid raft play opposite roles in regulating the function of CD44. At present, the structural mechanism of CD44ct autoinhibition is not known.
In summary, Fig. 8 shows the schematic presentation of the findings of this study that PIP2 induces aggregation of CD44ct and that PIP2 assembles multimeric CD44-Ezrin complexes. Future studies should determine the structure of CD44ct at a resolution higher than the SANS data presented in this study and address how full-length CD44 binding to ligand triggers the changes in PIP2 localization and the dynamic exchange of CD44 with the lipid raft and with PIP2-associated microdomains.