Phosphoinositide Specificity of and Mechanism of Lipid Domain Formation by Annexin A2-p11 Heterotetramer*

Annexin A2 is a phospholipid-binding protein that forms a heterotetramer (annexin II-p11 heterotetramer; A2t) with p11 (S100A10). It has been reported that annexin A2 is involved in binding to phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P2) and in inducing membrane microdomain formation. To understand the mechanisms underlying these findings, we determined the membrane binding properties of annexin A2 wild type and mutants both as monomer and as A2t. Our results from surface plasmon resonance analysis showed that A2t and annexin A2 has modest selectivity for PtdIns(4,5)P2 over other phosphoinositides, which is conferred by conserved basic residues, including Lys279 and Lys281, on the convex surface of annexin A2. Fluorescence microscopy measurements using giant unilamellar vesicles showed that A2t of wild type, but not (K279A)2-(p11)2 or (K281A)2-(p11)2, specifically induced the formation of 1-μm-sized PtdIns(4,5)P2 clusters, which were stabilized by cholesterol. Collectively, these studies elucidate the structural determinant of the PtdIns(4,5)P2 selectivity of A2t and suggest that A2t may be involved in the regulation of PtdIns(4,5)P2 clustering in the cell.

Annexins are a family of peripheral proteins that bind anionic phospholipids in a Ca 2ϩ -dependent manner (1)(2)(3)(4)(5). Structures and in vitro functions of annexins have been well characterized. Annexins have a variable N-terminal region and a conserved C-terminal core that is composed of four (eight in case of annexin A6) ␣-helical annexin folds (2). The C-terminal core is the Ca 2ϩ -dependent membrane-binding module that contains multiple Ca 2ϩ -binding sites on its convex membrane-binding surface (2). The N-terminal region of annexins is attached to the concave side of the C-terminal core and thought to be involved in interactions with other proteins and post-translational modifications (3)(4)(5). In addition to their membrane-binding activities, annexins have been reported to have other in vitro activities, including membrane aggregation and lateral aggregation on the membrane surface (6). Despite the wealth of structural and functional information on annexins, their physiological functions are only beginning to emerge with recent genetic and cell studies (3)(4)(5).
Annexin A2 is an abundant cellular protein that has been implicated in numerous physiological processes (3)(4)(5)7). Annexin A2 interacts with an EF-hand protein p11 (also known as S100A11) with high affinity via its N-terminal region, forming a symmetric heterotetramer, (annexin A2) 2 -(p11) 2 (A2t) 2 (8,9). Annexin A2 has been reported to exist either as a monomer or A2t in mammalian cells (3, 10 -12). Annexin A2 and A2t have been shown to have high vesicle aggregating activity (13) and form a monolayer of protein clusters when bound to the lipid bilayer with anionic phospholipids accumulating underneath the protein clusters (14). Mounting evidence indicates that annexin A2 and A2t are involved in organizing cholesterol-rich lipid rafts (15,16) and linking them to cytoskeletal proteins (17)(18)(19)(20). It has been also reported that annexin A2 (21,22) and A2t (22) bind to phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P 2 ) with high specificity and affinity and that this activity is linked to the organization of actin at membrane sites that are enriched in PtdIns(4,5)P 2 . Together with previous reports showing that annexin A2 binds cholesterol-containing membranes (23,24) and PtdIns(4,5)P 2 is localized in cholesterol-rich lipid rafts in the plasma membrane (25)(26)(27)(28), these results suggest that annexin A2 plays a role in regulating the formation of PtdIns(4,5)P 2 -rich lipid rafts or lipid raft-like structures. However, it is not known whether the annexin A2 dynamically controls the organization of these structures or it passively binds to the PtdIns(4,5)P 2 -rich regions. Furthermore, the affinity and specificity of annexin A2 and A2t for PtdIns(4,5)P 2 have not been quantitatively determined, which makes it difficult to assess their capability to compete with other PtdIns(4,5)P 2binding proteins under physiological conditions.
In this study, we systematically and quantitatively determined the phosphoinositide binding specificity and affinity of annexin A2 and A2t by surface plasmon resonance (SPR) analysis and identified the structural determinant of its phosphoinositide specificity. We also investigated the formation of PtdIns(4,5)P 2 -rich membrane domains on giant unilamellar vesicles (GUV) induced by A2t and mutants under various conditions. Our study provides new insight into the mechanism by which A2t mediates the organization of PtdIns(4,5)P 2 -rich membrane domains.
Vector Construction and Mutagenesis-The cDNA of full-length human annexin A2, which was a generous gift from Dr. Volker Gerke, was subcloned into the vector pET-21a(ϩ) (Novagen, Madison, WI), between the restriction sites BamHI and XhoI. A stop codon was introduced just before the restriction site XhoI in order to exclude the C-terminal hexahistidine tag from the sequence during protein expression. The cDNA of p11 (a generous gift from Dr. James Seilhamer) was also subcloned into pET-21a(ϩ) in a similar fashion. K279A and K281A mutants of annexin A2 were generated by the overlap extension polymerase chain reaction mutagenesis. The vector pGEX-4T-1 (Novagen, Madison, WI), which has an N-terminal glutathione S-transferase tag and a thrombin cleavage site was used for subcloning the cDNA of the phospholipase C␦1 (PLC␦1) pleckstrin homology (PH) domain between the restriction sites BamHI and XhoI. All of the above constructs were transformed into DH5␣ cells for plasmid isolation, and their DNA sequences were verified. Subsequently, these plasmids were transformed into BL21(DE3) cells for protein expression.
Protein Expression and Purification-One liter of sterile Luria broth medium containing 100 g/ml ampicillin was inoculated with BL21(DE3) cells harboring each construct and grown at 37°C until the optical density at 600 nm reached 0.6. Protein expression was then induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside (Research Products, Mount Prospect, IL). After 8 h of incubation at 37°C, cells were harvested by centrifugation (5,000 ϫ g for 10 min at 4°C). The resulting pellet was resuspended in 20 ml of precooled lysis buffer, containing 100 mM Tris-HCl (pH 7.5), 200 mM NaCl, 2 mM dithiothreitol, 1 mM EGTA, 0.5 mM phenylmethylsulfonyl fluoride, 5 g/ml leupeptin, 5 g/ml pepstatin, 5 g/ml aprotinin, a protease inhibitor tablet, and 0.1% Triton X-100. This suspension was sonicated for 6 min (30 s of sonication followed by 30 s of incubation on ice) and then centrifuged for 1 h (40,000 ϫ g at 4°C). The supernatant was treated with 50% (NH 4 ) 2 SO 4 (final concentration) for 45 min and centrifuged at 40,000 ϫ g for 20 min to remove insoluble proteins. This supernatant was then applied to an 80-ml butyl-Sepharose column (Amersham Biosciences) pre-equilibrated with 50% (NH 4 ) 2 SO 4 in the lysis buffer. Annexin A2 (or a mutant) was eluted with a linear gradient of (NH 4 ) 2 SO 4 from 50 to 0% in the lysis buffer. The eluted samples were dialyzed against 10 mM HEPES buffer, pH 7.5, containing 50 mM NaCl, 1 mM EGTA, and 1 mM dithiothreitol and applied to a DEAE-Sepharose column (Amersham Biosciences) equilibrated with the same buffer. Elution was carried out with 200 ml of 10 mM HEPES buffer, pH 7.5, containing 1 M NaCl, 1 mM EGTA, and 1 mM dithiothreitol. The recombinant p11 was purified by the same protocol. The purified recombinant annexin A2 was dialyzed against 10 mM HEPES buffer, pH 7.5, containing 50 mM NaCl, 1 mM EGTA, 1 mM dithiothreitol, 6 mM CaCl 2 , and a 2-5-fold molar excess of p11. These samples were then loaded onto a 5-ml Heparin HP HiTrap TM column (Amersham Biosciences) equilibrated with the same buffer without p11. The column was washed with 100 ml of the same buffer to get rid of excess p11, and the heterotetramer was eluted with 50 ml of elution buffer containing 30 mM HEPES, pH 7.5, 160 mM NaCl, 10 mM EGTA, and 1 mM dithiothreitol. The heterotetrameric nature of A2t was confirmed by the polyacrylamide gel electrophoresis performed under nondenaturing conditions (i.e. in the absence of SDS and dithiothreitol) using a 16% polyacrylamide gel (see Fig. 1).
For the expression of the PLC␦1 PH domain, 1 liter of Luria broth containing 100 g/ml ampicillin was inoculated with BL21(DE3) cells harboring the PH domain construct and grown at 37°C until the optical density at 600 nm reached 0.4. Protein expression was induced by the addition of 50 mg of isopropyl-1-thio-␤-D-galactopyranoside, and cells were harvested by centrifugation (5,000 ϫ g for 10 min at 4°C) after a 12-h incubation at 25°C. The resulting pellet was resuspended in 10 ml of 30 mM HEPES buffer, pH 7.5, containing 160 mM NaCl, 50 M phenylmethylsulfonyl fluoride, and 0.1% Triton X-100. This solution was sonicated for 6 min (30 s of sonication followed by 30 s of cooling on ice) and then centrifuged for 1 h (40,000 ϫ g at 4°C). After filtering the supernatant into a 50-ml Falcon tube, 500 l of the glutathione S-transferase-Tag TM resin (Novagen, Madison, WI) were added. After incubating this mixture on ice for 45 min with mild shaking at 80 rpm, it was poured onto a column prerinsed with 50 ml of 30 mM HEPES buffer, pH 7.5, containing 160 mM NaCl. After washing the nonspecifically bound protein with 50 ml of 30 mM HEPES buffer, pH 7.5, containing 160 mM NaCl, 1 ml of the same buffer containing 4 units of thrombin was added in order to cleave the glutathione S-transferase tag, and the column was then sealed for a 6-h incubation at 25°C. The protein was then eluted in five fractions using 500 l of 30 mM HEPES buffer, pH 7.5, containing 160 mM NaCl. The protein purity was checked on a 16% polyacrylamide gel, and all of the protein samples were subsequently concentrated. The protein concentrations were determined by the bicinchoninic acid method (Pierce).
Chemical Labeling of Proteins-The purified p11 (2 mg/ml) was treated with a 10-fold molar excess of Texas Red TM C 2 -maleimide for 2 h at room temperature in 30 mM HEPES buffer, pH 7.5, which was purged with a stream of nitrogen gas before use to remove oxygen. The labeling reaction was subsequently quenched by adding an excess amount of 2-mercaptoethanol, and the labeled protein was separated from the reagents using a Sephadex G25 column (Sigma) eluted with 30 mM HEPES buffer, pH 7.5, containing 50 mM NaCl. The fractions corresponding to the protein peak were pooled and dialyzed against 30 mM HEPES buffer, pH 7.5, containing 50 mM NaCl and 6 mM CaCl 2 for 24 h at 4°C. The labeling efficiency of p11 was estimated using the equation, mol of dye/mol of protein ϭ (absorbance of the labeled protein at 582 nm)/((molar absorptivity of Texas Red at 582 nm (⑀ ϭ 112,000 M Ϫ1 cm Ϫ1 ) ϫ (protein concentration)). Under our labeling conditions, ϳ0.7 mol of Texas Red was incorporated per mol of p11. The labeled p11 was then incubated on ice with purified annexin A2 (wild type or mutants) for 2 h, and the labeled heterotetramer was purified using a HiTrap TM Heparin HP column as described above. The PLC␦1 PH domain was labeled with a 10-fold molar excess of fluorescein-5-isothiocyanate for 2 h at room temperature, and the labeled protein was separated using a Sephadex G25 column eluted with 30 mM HEPES buffer, pH 7.5, containing 160 mM NaCl.
SPR Measurements-All SPR equilibrium binding measurements were carried out at 23°C as described (30,31). The sensor chip Pioneer L1 (Biacore) was coated with vesicles according to a protocol described previously (32). Typically, after washing the sensor chip surface, 90 l of phospholipid vesicles of different lipid composition were injected at a flow rate of 5 l/min to give a response of 5000 resonance units. The control surface was then coated with POPC/POPE (80:20) vesicles to give the same resonance unit response as that of the active surface. All of the equilibrium binding measurements were carried out at a steady flow rate of 5 l/min in order to give sufficient time for the R values of the association phase to attain saturating response units (R eq ). R eq values were then plotted against protein concentrations (C), and the K d value was determined by a nonlinear least squares analysis of the binding isotherm using the equation Spectrofluorometric Measurements on Large Unilamellar Vesicles (LUV)-LUV were prepared by an extrusion method using a Liposofast microextruder and a 100-nm polycarbonate filter (Avestin, Ottawa, Canada). Spectrofluorometric measurements using LAURDAN-containing LUV were carried out with a Hitachi F4500 fluorescence spectrophotometer at an excitation wavelength of 364 nm.
Microscopy Measurements on GUV-GUV were prepared by the electroformation method using a home-built device as described previously (33). Briefly, GUV were grown in deionized water at 60°C for 30 min by spreading ϳ3 l of the lipid stock with various compositions on platinum wires. During GUV growth, the platinum wires were connected to a function generator (Hewlett-Packard, Santa Clara, CA) for 30 min, and a low frequency AC field (sinusoidal wave function with a frequency of 10 Hz and an amplitude of 3 V) was applied. After 45 min, the temperature was lowered to 40°C, and the frequency generator was switched off after the system attained this temperature. All subsequent measurements were carried out at 40°C in 10 mM HEPES buffer, pH 7.5, with 0.16 M NaCl and different concentrations of Ca 2ϩ .
All microscopy measurements were carried out using a custom-built combination laser-scanning and multiphoton microscope that was described previously (34). Briefly, a 920-nm ultrafast pulsed beam from a tunable Tsunami laser, set up for femtosecond operation (Spectra Physics, Mountain View, CA), was spatially filtered and launched into the scan head. The beam was directed toward the primary dichroic mirror (Chroma Technology, Brattleboro, VT) and then toward the XY scan mirrors (model 6350, Cambridge Technologies, Cambridge, MA). A Prairie Technologies scan lens (Middleton, WI) was used to focus the laser light, collimated by the 1ϫ Zeiss tube lens, and directed toward a 40ϫ water-corrected 1.2 numerical aperture Zeiss objective, mounted on a Zeiss 200 M platform (Carl Zeiss Inc.). Light excited by a 920-nm ultrafast pulse was collected on a nondescanned pathway by the Peltiercooled 1477P style Hamamatsu photomultiplier tubes. The light was reflected and filtered using appropriate optics. Instrument control was accomplished with the help of ISS amplifiers, an ISS three-axis scanning card (Champaign, IL), and two ISS 200-KHz analog lifetime cards. All the microscopic experiments were controlled by a data acquisition program, SimFCS (Laboratory for Fluorescence Dynamics, University of Illinois, Urbana-Champaign, IL).

RESULTS
Phosphoinositide Specificity of Annexin A2 and A2t-It has been recently reported that annexin A2 and A2t have high specificity for PtdIns(4,5)P 2 (21,22); however, this putative PtdIns(4,5)P 2 specificity has not been quantitatively measured. We therefore prepared mixed vesicles of POPC/POPE containing 3 mol % of each of seven phosphoinositides and quantitatively determined by SPR analysis the affinity of annexin A2 and A2t for these vesicles coated onto the sensor chip. Annexin A2 and p11 have been shown to form a heterotetramer with high affinity (8,9). Fig. 1 indicates that A2t exists as a heterotetramer under our experimental conditions. Representative sensorgrams for A2t and vesicles and a binding isotherm generated from the sensorgrams are shown in Fig. 2. The SPR method not only allows sensitive and quantitative determination of K d values (32,35) but also circumvents the vesicle aggregation during binding measurements, because it uses vesicles immobilized on the sensor chip. This is important, because in the vesicle pelleting assay, the charge and size of vesicles can significantly affect the pelleting efficiency, which in turn complicates the interpretation of binding data (35). For the first set of measurements, a relatively high Ca 2ϩ concentration (0.1 mM) was employed to ensure that the proteins show detectable affinities for all phosphoinositide-containing vesicles  under the same conditions. The control surface was coated with POPC/ POPE (80:20) vesicles, because neither annexin A2 nor A2t showed detectable affinity for these zwitterionic vesicles.
This in turn suggests that A2t has a defined, albeit not optimized, PtdIns(4,5)P 2 binding site that is located not in p11 but in annexin A2.
The Effect of Other Lipids on Membrane Affinity of A2t-It has been reported that annexin A2 is involved in the organization of lipid rafts in the plasma membrane of mammalian cells (15,16). To see if A2t has the physical properties that are consistent with this reported activity, we further investigated the membrane binding properties of A2t. First, we measured the affinities of A2t for the vesicles whose lipid headgroup compositions recapitulate those of mammalian cell membranes (see the footnotes to TABLE TWO) to assess the differential affinities of A2t for various cell membranes. This approach has been successfully applied to determine and account for the cellular targeting specificity of various membrane targeting domains and peripheral proteins (37,38). For these measurements, we employed 5 M Ca 2ϩ to simulate the cellular environment as much as possible. Lower Ca 2ϩ concentrations could not be used, because SPR signals were too small to analyze under such conditions.
As shown in TABLE TWO, A2t showed considerable affinity for the mimetic of inner plasma membrane at 5 M Ca 2ϩ while showing no detectable affinity for the mimetics of other cellular membranes. It should be noted that A2t exhibited no detectable binding to POPC/ POPE/PtdIns(4,5)P 2 (77:20:3) vesicles at 5 M Ca 2ϩ . Therefore, A2t has a preference for the lipid composition of the inner plasma membrane, suggesting that the plasma membrane is a main site for A2t actions. This preference became more pronounced when 3 mol % PtdIns(4,5)P 2 was incorporated into the plasma membrane mimetic. The K d value for this membrane is 27 nM even at 5 M Ca 2ϩ . Among various cellular membranes, the inner plasma membrane is known to contain the highest content of anionic lipids and cholesterol (see the footnotes to TABLE TWO), both of which have been shown to enhance the membrane affinity of annexin A2 and A2t (23,24). To further investigate the effect of cholesterol on the membrane affinity of A2t, we measured the binding of A2 to the plasma membrane mimetic without cholesterol. A2t showed no detectable affinity for this membrane even in the presence of 3 mol % PtdIns(4,5)P 2 , demonstrating the importance of cholesterol in membrane binding of A2t. Interestingly, A2t did not show any binding to POPC/POPE/cholesterol (55:20:25) vesicles even at 1 mM Ca 2ϩ (data not shown), indicating that A2t does not have affinity for cholesterol

PtdIns(4,5)P 2 Clustering by Annexin A2
itself. Thus, it appears that the cholesterol enhances the affinity of A2t for the plasma membrane by modulating the membrane structure. Structural Determinant of Phosphoinositide Specificity of Annexin A2 and A2t-PtdIns(4,5)P 2 -binding proteins can be subdivided into two groups (38,39). The first group contains a well defined structural module, such as PH domain, that specifically recognizes PtdIns(4,5)P 2 in a binding pocket, whereas the second group lacks a well defined binding pocket but utilizes a cluster of surface cationic residues to bind PtdIns(4,5)P 2 . Annexin 2 should belong to the second group, since it does not have an established PtdIns(4,5)P 2 -binding domain. To identify the annexin A2 residues involved in PtdIns(4,5)P 2 binding, we surveyed the PtdIns(4,5)P 2 -binding motifs found in several actin-binding proteins belonging to the second group, including gelsolin, villin, gCAP39, MARCKS, and cortexillin I. These proteins have a consensus nonapeptide motif, (R/K)LXXX(R/K)X(R/K)(R/K) (see Fig. 3), which is directly involved in PtdIns(4,5)P 2 binding (40,41). A multiple sequence alignment by ClustalW (42) (Fig. 3) revealed that annexin 2 also possesses near the C terminus a putative PtdIns(4,5)P 2 -binding motif. Among several conserved basic residues in this region, Lys 279 , Lys 281 , and Arg 284 are located on the convex surface of the annexin core (see Fig. 4).
A2t-mediated Lipid Ordering and PtdIns(4,5)P 2 Clustering-Annexin A2 has been reported to induce lipid raft formation in mammalian cells (15,16). However, these results are based mainly on detergent extraction and antibody-mediated visualization methods, in combination with cholesterol depletion, which provide only indirect evidence and may also cause artificial lipid clustering (43). To directly measure the effects of annexin A2 and A2t on the membrane organization, we employed a well defined in vitro system in which interactions of A2t and annexin A2 with the vesicles of different lipid composition and sizes could be directly monitored by various fluorescence techniques. We first measured spectrofluorometrically the annexin-mediated changes in lipid ordering in the plasma membrane mimetic vesicles (LUV of 100-nm diameter) containing 3 mol % PtdIns(4,5)P 2 and 0.1 mol % LAURDAN. It has been reported (33,44) that an increase in the membrane order causes a blue shift in the fluorescence emission of the membrane-incorporated LAURDAN. As shown in Fig. 5A, A2t caused a significant blue shift of the LAURDAN emission spectra, suggesting that it induces membrane ordering. In contrast, (K279A) 2 -(p11) 2 (Fig.  5B) and (K281A) 2 -(p11) 2 (Fig. 5C) only slightly enhanced the emission intensity without a detectable spectral shift. Neither did annexin A2 cause a significant spectral shift (data not shown). Also, we measured the effect of cholesterol on the A2t-mediated membrane ordering. As shown in Fig. 5D, A2t did not cause the blue shift of LAURDAN fluorescence when cholesterol was removed from the above vesicles, suggesting that the presence of cholesterol is essential for A2t-mediated membrane ordering.
We then directly monitored the A2t-mediated clustering of PtdIns(4,5)P 2 using GUV of the plasma membrane mimetic containing 1 mol % BODIPY-labeled PtdIns(4,5)P 2 and 2 mol % unlabeled PtdIns(4,5)P 2 . GUV (diameter Ͼ 10 m) are an excellent model system for cell membranes that allow direct visualization of various membrane processes, including structural changes of membranes (45). For these   DECEMBER 30, 2005 • VOLUME 280 • NUMBER 52  measurements, p11 was chemically labeled with Texas Red, and this fluorescently labeled p11 was incorporated into A2t, (K279A) 2 -(p11) 2 , and (K281A) 2 -(p11) 2 , respectively. A green fluorescence protein tag on either annexin A2 or p11 could not be used here, because it interfered with heterotetramer formation and membrane binding. p11 has a single surface-exposed Cys (Cys 82 ) and another internal Cys (9). Treatment of p11 with Texas Red C 2 -maleimide and partial purification yielded the labeled protein, which incorporated ϳ0.7 mol of Texas Red/mol of p11, implying that only Cys 82 is labeled. Most importantly, the labeled protein formed a heterotetramer with annexin A2 (see Fig. 1), and this labeled heterotetramer was indistinguishable from the unlabeled A2t with respect to the affinity for POPC/POPE/PtdIns(4,5)P 2 (77:20:3) vesicles (K d ϳ35 nM at 0.1 mM Ca 2ϩ ).

PtdIns(4,5)P 2 Clustering by Annexin A2
In the absence of proteins, the BODIPY-labeled PtdIns(4,5)P 2 showed homogeneous distribution on the GUV surfaces (Fig. 6E, left). Ca 2ϩ up to 1 mM had no effect on the distribution of PtdIns(4,5)P 2 (data not shown). The addition of 80 nM A2t to the GUV in the presence of 5 M Ca 2ϩ , however, caused the formation of nearly 1-m PtdIns(4,5)P 2 clusters (see Fig. 6, A and B, for clustering in two separate GUV). Most important, PtdIns(4,5)P 2 spots were invariably colocalized with A2t spots (Fig. 6, A and B), showing that PtdIns(4,5)P 2 clustering is directly linked to the PtdIns(4,5)P 2 binding of A2t. More than 80% of GUV used in this study showed the same trend. This notion was further supported by the finding that 300 -800 nM (K279A) 2 -(p11) 2 and (K281A) 2 -(p11) 2 bound to GUV but did not form either PtdIns(4,5)P 2 clusters or protein aggregates under the same conditions (Fig. 6, C and D). Under the same conditions, the annexin A2 monomer (unlabeled) up to 1 M did not cause PtdIns(4,5)P 2 clustering (Fig. 6E). Given that A2t has a greater tendency to laterally aggregate than annexin A2 at low Ca 2ϩ concentrations (e.g. 5 M), this indicates that A2t-mediated PtdIns(4,5)P 2 clustering derives not only from the capability of A2t to specifically bind PtdIns(4,5)P 2 but also from its ability to laterally aggregate on the vesicle surfaces.
To investigate the effect of cholesterol on the A2t-mediated PtdIns(4,5)P 2 clustering, we repeated the measurements using the same The arrows indicate the locations of PtdIns(4,5)P 2 /A2t clusters. The GUV of the plasma membrane mimetic containing 1 mol % BODIPY-PtdIns(4,5)P 2 and 2 mol % unlabeled PtdIns(4,5)P 2 in 10 mM HEPES buffer, pH 7.5, with 0.16 M NaCl and 5 M Ca 2ϩ were used unless specified otherwise. Images were taken every 5 s. DECEMBER 30, 2005 • VOLUME 280 • NUMBER 52 GUV minus cholesterol. Since A2t has lower affinity for this membrane, Ca 2ϩ concentration was raised to 20 M (K d ϳ150 nM for these vesicles at 20 M Ca 2ϩ ) to facilitate the membrane binding of A2t. Under these conditions, A2t-bound GUV formed protein and PtdIns(4,5)P 2 clusters on the membrane surfaces; however, the number of clusters was significantly reduced, and the clusters were only transient, as shown by the time lapse images in Fig. 6G. The clusters formed in the presence of cholesterol (see Fig. 6, A and B) lasted much longer than these transient clusters. This underscores the importance of cholesterol in the formation and stabilization of the PtdIns(4,5)P 2 clusters.

PtdIns(4,5)P 2 Clustering by Annexin A2
Last, we measured the effect of another PtdIns(4,5)P 2 -binding protein, the PH domain of PLC␦1, to see whether the PtdIns(4,5)P 2 clustering is a generic property of any PtdIns(4,5)P 2 -binding proteins or specific to A2t. We incrementally added the PLC␦1 PH domain to GUV of the plasma membrane mimetic containing 1 mol % BODIPY-labeled PtdIns(4,5)P 2 and 2 mol % unlabeled PtdIns(4,5)P 2 . With the PH domain concentration up to 1 M, under which condition the vesicle surfaces should be fully covered by the PH domain (note that K d ϭ 52 nM for the PH domain-PtdIns(4,5)P 2 vesicle binding; see TABLE ONE), no PtdIns(4,5)P 2 clustering was detected (Fig. 6F). We then chemically labeled the PLC␦1 PH domain with fluorescein to simultaneously monitor vesicle binding of A2t and the PH domain. For these measurements, BODIPY-labeled PtdIns(4,5)P 2 was not included in GUV to circumvent the spectral overlap with the fluorescein-labeled PH domain. When the fluorescein-labeled PH domain was added to the mixture after the Texas Red-labeled A2t was allowed to interact with the GUV of plasma membrane mimetic containing 3 mol % unlabeled PtdIns(4,5)P 2 in the presence of 5 M Ca 2ϩ , the PH domain was able to rapidly bind to the GUV (Fig. 6H, right) without disrupting A2t clusters (Fig. 6H, left and middle). Notice that the locations of A2t clusters in the left and middle panels are different because they were monitored at different times, and PtdIns(4,5)P 2 /A2t clusters diffused laterally on the membrane. This suggests that A2t-induced PtdIns(4,5)P 2 clusters are readily accessible to other PtdIns(4,5)P 2 -binding proteins and may thus serve as sites for various PtdIns(4,5)P 2 -mediated processes.

DISCUSSION
Although PtdIns(4,5)P 2 is a minor component of membrane lipids that is mainly found in the inner leaflet of plasma membrane, it plays important regulatory roles in diverse cellular processes, including actin polymerization (46, 47) and vesicle trafficking (48,49), and also serves as the precursor for the generation of second messengers, including diacylglycerol, inositol (1,4,5)-trisphosphate (50), and PtdIns(3,4,5)P 3 (51). To account for these diverse roles of PtdIns(4,5)P 2 , the presence of spatially confined PtdIns(4,5)P 2 pools in the plasma membrane has been hypothesized (48,52,53). Biochemical studies have indicated that PtdIns(4,5)P 2 is concentrated in membrane microdomains, such as cholesterol-rich lipid rafts (25)(26)(27)(28). Also, direct visualization of PtdIns(4,5)P 2 by the green fluorescent protein-tagged PLC␦1 PH domain has shown the presence of PtdIns(4,5)P 2 clusters in the plasma membrane of various cells (54 -57), some of which are much larger than the putative size of lipid rafts (i.e. Ͻ250 Å). These results indicate that PtdIns(4,5)P 2 in the plasma membrane may exist in clusters, but it is still controversial whether it is confined within the cholesterol-rich lipid rafts.
Since the rate of lateral diffusion of PtdIns(4,5)P 2 within the lipid bilayer should be much faster than the rate of its local biosynthesis (39), PtdIns(4,5)P 2 clustering would require diffusion barriers that can be provided by proteins that are capable of binding to PtdIns(4,5)P 2 and cytoskeletal proteins. Although MARCKS (39) and other proteins (58) have been implicated in reversible clustering and sequestration of PtdIns(4,5)P 2 in the plasma membrane, direct evidence for this interesting idea has not been reported. Among a large number of PtdIns(4,5)P 2binding proteins, A2t is suited for this putative role due to its unique properties; i.e. A2t binds the membrane in a Ca 2ϩ -and/or PtdIns(4,5)P 2dependent manner (21,22), laterally aggregates on the membrane surface (14), interacts with F-actin (59), and is abundant in mammalian cells. Furthermore, recent reports have indicated the potential involvement of annexin A2 and A2t in regulation of membrane microdomain formation (15,16). The present study quantitatively shows that A2t possesses all the biophysical properties that are necessary for its putative role in organizing PtdIns(4,5)P 2 clusters in the plasma membrane and also provides direct evidence for A2t-induced PtdIns(4,5)P 2 clusters in a well defined model membrane system.
On the basis of qualitative lipid overlay assay and vesicle pelleting assay, Hayes et al. (21) recently reported that annexin A2 has high specificity and affinity for PtdIns(4,5)P 2 . Rescher et al. (22) also reported that A2t as well as annexin A2 have high affinity and selectivity for PtdIns(4,5)P 2 in vesicle pelleting and lipid plate binding assays. In both studies, PtdIns(4,5)P 2 binding activity of annexin A2 and A2t was linked to the binding of these proteins to the sites of membrane-associated actin assembly in the cell. We reexamined the specificity and affinity of annexin A2 and A2t for PtdIns(4,5)P 2 by means of SPR analysis. Our SPR measurements confirm that annexin A2 and A2t bind PtdIns(4,5)P 2 and that A2t binds PtdIns(4,5)P 2 -containing membranes with high affinity. The affinity of A2t for PtdIns(4,5)P 2 -containing vesicles is comparable with that of the PLC␦1 PH domain and epsin ENTH domain (36). However, our measurements reveal that neither annexin A2 nor A2t has high specificity for PtdIns(4,5)P 2 . The largest difference between PtdIns(4,5)P 2 and any phosphoinositide was less than 4-fold when they were incorporated in POPC/POPE/phosphoinositide (77: 20:3) vesicles. This discrepancy in PtdIns(4,5)P 2 specificity between the published data and ours should arise mainly from different methods used for binding measurements (see "Results"). A typical PtdIns(4,5)P 2specific protein, such as PLC␦1 PH domain or epsin ENTH domain (36), shows much more pronounced selectivity for PtdIns(4,5)P 2 over other phosphoinositides. Thus, cellular specificity of A2t (or annexin A2) for PtdIns(4,5)P 2 would derive from the combination of modest PtdIns(4,5)P 2 selectivity of A2t and the relative abundance of PtdIns(4,5)P 2 over other phosphoinositides. In addition, the strong preference of A2t for the inner plasma membrane over other cellular membranes should help its specific interaction with PtdIns(4,5)P 2 enriched in the plasma membrane.
There is some confusion in the literature as to whether annexin A2 monomer or A2t is involved in vesicle trafficking and membrane microdomain organization (5). As reported previously (22), the modest PtdIns(4,5)P 2 selectivity of A2t derives not from p11 but from the annexin A2 molecule. However, the affinity of annexin A2 monomer for the PtdIns(4,5)P 2 -containing membrane seems to be too low to be physiologically significant. It has been long known that A2t has lower Ca 2ϩ requirement for phospholipid binding and granule aggregation than annexin A2 (60). However, the difference in membrane affinity between A2t and annexin A2 has not been quantified due in part to the complication associated with a conventional vesicle pelleting assay (see "Results"). Our SPR data show that A2t has at least 10 times higher affinity for PtdIns(4,5)P 2 -containing vesicles than annexin A2 at 0.1 mM Ca 2ϩ and that the difference is much bigger at lower Ca 2ϩ concentrations. Presumably, A2t has much higher membrane affinity than annexin A2, because two membrane binding modules of A2t function either additively or synergistically, as proposed for the membrane bind-ing of low affinity PH domains (61). Due to its low affinity, the annexin A2 monomer would not be able to compete with other PtdIns(4,5)P 2binding proteins for the same pools of PtdIns(4,5)P 2 unless its membrane binding is supplemented by interaction with other membranebound proteins. These results thus indicate that A2t is the functional form of annexin A2 in the cell as far as the PtdIns(4,5)P 2 -dependent membrane binding is concerned.
This study also identifies the specific PtdIns(4,5)P 2 -binding residues in the annexin A2 molecule. Annexin A2 contains the basic "nonapeptide" motif that is found in many actin-and PtdIns(4,5)P 2 -binding proteins (40,41). This motif is located in one of five Ca 2ϩ binding loops. Since this loop protrudes in the middle of the convex surface (Fig. 4A) and has a highly positive electrostatic potential (Fig. 4B), it is expected to make direct contact with the anionic membrane. Among several cationic residues in this region, Lys 279 , Lys 281 , and Arg 284 are surface-exposed, and at least two of them, Lys 279 and Lys 281 , are shown to be directly involved in PtdIns(4,5)P 2 binding. The K279A and K281A mutations reduce the affinity of A2t for the plasma membrane mimetic by less than 3-fold but decrease the affinity for the PtdIns(4,5)P 2 -containing plasma membrane mimetic by about 20-fold. Thus, these residues should be directly involved in PtdIns(4,5)P 2 binding rather than interacting nonspecifically with anionic membrane surfaces. A recent crystal structure of the ANTH domain of Ap180/CALM showed that a cluster of cationic residues bind the phosphate groups of PtdIns(4,5)P 2 (62). Likewise, the surface cationic residues of annexin A2, particularly Lys 279 and Lys 281 , would form a surface binding site for the inositol headgroup of PtdIns(4,5)P 2 . The modest phosphoinositide selectivity of annexin A2 suggests that the arrangement of these cationic residues is not fully optimized for PtdIns(4,5)P 2 binding.
The identification of PtdIns(4,5)P 2 -binding residues also helped to elucidate the mechanism by which A2t induces PtdIns(4,5)P 2 clustering. A2t, but not (K279A) 2 -(p11) 2 and (K281A) 2 -(p11) 2 , causes lipid ordering and PtdIns(4,5)P 2 clustering in LUV and GUV of the plasma membrane mimetic containing 3 mol % PtdIns(4,5)P 2 . Thus, the PtdIns(4,5)P 2 binding activity of A2t appears to be essential for its effect on membrane organization. The ϳ1-m size of PtdIns(4,5)P 2 clusters formed on GUV is much larger than the putative size of cholesterol-rich lipid rafts (Ͻ250 nm), indicating that the observed PtdIns(4,5)P 2 clusters do not represent the lipid raft-like structure. Since the resolution of light microscopy does not allow direct monitoring of lipid rafts, our study cannot distinguish whether A2t induces the formation of PtdIns(4,5)P 2 clusters from preexisting cholesterol-and PtdIns(4,5)P 2rich lipid rafts or from randomly distributed lipids. In either case, the PtdIns(4,5)P 2 -clustering activity of A2t would seem to derive from its ability to laterally aggregate on the membrane surface. This is because the annexin A2 monomer and the PLC␦1 PH domain, both of which bind PtdIns(4,5)P 2 -containing vesicles but show little to no tendency to laterally aggregate on the membrane surface, do not show such activity.
It has been reported that cholesterol increases the affinity of annexin A2 for anionic vesicles (63) and regulates the subcellular distribution of annexin A2 (24). Our SPR measurements show, however, that annexin A2 does not bind cholesterol itself, which is consistent with a recent report on the binding of A2t to solid-supported lipid membranes (64). Thus, cholesterol should have an indirect effect on the membrane binding of A2t. It has been reported that cholesterol induces the formation of lipid microdomains (65). Therefore, cholesterol may enhance the membrane affinities of annexin A2 and A2t by mediating the formation of anionic lipid-rich microdomains and/or clusters and thereby increasing the local concentrations of anionic lipids. Our measurements show that A2t cannot effectively induce the formation of stable PtdIns(4,5)P 2 patches in the absence of cholesterol. Thus, cholesterol may play a dual role of promoting the initial membrane adsorption of A2t and facilitating the A2t-mediated formation of stable PtdIns(4,5)P 2 patches.
On the basis of our results and known properties of other PtdIns(4,5)P 2 -binding proteins (38), we propose the mechanism by which A2t induces PtdIns(4,5)P 2 clustering in model membranes. A2t binds to anionic membranes via Ca 2ϩ -dependent nonspecific electrostatic interactions. This initial binding is followed by lateral diffusion of A2t on the membrane surface to specifically bind PtdIns(4,5)P 2 . As A2t molecules laterally aggregate on the membrane surface, protein-bound PtdIns(4,5)P 2 molecules also form patches, which is facilitated by the presence of cholesterol in the membrane. These A2t-induced PtdIns(4,5)P 2 clusters are still accessible to other PtdIns(4,5)P 2 -binding proteins, as evidenced by undeterred binding of the PLC␦1 PH domain to the PtdIns(4,5)P 2 clusters. In the cell, A2t is expected to bind the plasma membrane (or early endosomes under certain conditions) containing PtdIns(4,5)P 2 and cause similar PtdIns(4,5)P 2 clustering in a Ca 2ϩ -dependent manner, thereby generating spatially separated PtdIns(4,5)P 2 pools. In this case, dual interactions of A2t with both the membrane and the cytoskeleton will severely limit the mobility of PtdIns(4,5)P 2 clusters. Whether these PtdIns(4,5)P 2 clusters correspond to lipid rafts or not, they may function as the sites for cell signaling, vesicle trafficking, or actin assembly. Obviously, further studies are necessary to elucidate if and how A2t mediates PtdIns(4,5)P 2 clustering in the cell. This in vitro study provides important new structural and mechanistic information that will form the basis of such cell studies.