Tyrosine Phosphorylation as a Conformational Switch

Reversible protein phosphorylation is vital for many fundamental cellular processes. The actual impact of adding and removing phosphate group(s) is 3-fold: changes in the local/global geometry, alterations in the electrostatic potential and, as the result of both, modified protein-target interactions. Here we present a comprehensive structural investigation of the effects of phosphorylation on the conformational as well as functional states of a crucial cell surface receptor, αIIbβ3 integrin. We have analyzed phosphorylated (Tyr747 and Tyr759) β3 integrin cytoplasmic tail (CT) primarily by NMR, and our data demonstrate that under both aqueous and membrane-mimetic conditions, phosphorylation causes substantial conformational rearrangements. These changes originate from novel ionic interactions and revised phospholipid binding. Under aqueous conditions, the critical Tyr747 phosphorylation prevents β3CT from binding to its heterodimer partner αIIbCT, thus likely maintaining an activated state of the receptor. This conclusion was tested in vivo and confirmed by integrin-dependent endothelial cells adhesion assay. Under membrane-mimetic conditions, phosphorylation results in a modified membrane embedding characterized by significant changes in the secondary structure pattern and the overall fold of β3CT. Collectively these data provide unique molecular insights into multiple regulatory roles of phosphorylation.

Protein phosphorylation, initially discovered in the mid 1950s (1), today is considered as one of the most crucial cell signaling events. It is a reversible, ubiquitous switch which regulates nearly every aspect of prokaryotic and eukaryotic cell life and has been linked to many pathogenic processes. Phosphorylation involves a covalent attachment of the negatively charged phosphate group to the side chains of serine (86.4%), threonine (11.8%), and tyrosine (1.8%) residues in eukaryotes (2), which may result in local and/or global conformational rearrangement or induce transitions from order to disorder and vice versa (3). Moreover, it may alter protein function by modifying its interactions with the substrates or by varying the equi-librium between different conformational states. A total comprehension of these transitions is crucial for enhancing our knowledge of the signal transduction processes.
Integrins, a major class of non-covalent heterodimeric glycoprotein cell surface receptors, have been chosen for investigating the effects of phosphorylation in present work. Integrins are among the most studied and best characterized cell adhesion molecules. Each integrin subunit contains a large extracellular ligand-binding portion, a single membrane-spanning domain, and a short cytoplasmic tail devoid of any enzymatic activity (4). The unique bidirectional flow of information through integrins involves inside-out signals, which allow them to interact with extracellular soluble ligands, and ligand-dependent outside-in signals, which trigger the cellular response to cell adhesion. The integrins extracellular matrix (ECM) 2 interactions are controlled by integrins extracellular domains, whereas integrinscytoplasmic proteins interactions are controlled via their cytoplasmic tails (CTs). This integration of extracellular and intracellular compartments allows dynamic regulation of many cellular processes including cell migration, shape change, proliferation, and differentiation (5). Integrin regulated signaling pathways, which involve direct or indirect interaction of the integrin CTs with integrin-associated proteins, are often controlled through phosphorylation.
Although platelet integrin ␤ 3 CT includes several phosphorylation sites (two tyrosines, one serine and multiple threonines), only tyrosine phosphorylation is found to be specific for the outside-in signaling (6). However, despite the crucial role of tyrosine phosphorylation for ␤ 3 integrin function, the structural details describing the consequences of this process remain unknown. In this study we have investigated the effects of tyrosine phosphorylation on ␤ 3 CT under both, aqueous and membrane-mimic, environments. Our data demonstrate that, in comparison to the non-phosphorylated form (7,8), under aqueous conditions phosphorylation of Tyr 747 and/or Tyr 759 of ␤ 3 CT induces a novel fold which precludes ␣ IIb /␤ 3 complex formation, thereby preserving the activated state of the recep-tor. In presence of dodecyl-phosphocholine (DPC) tyrosine(s) phosphorylation results in significant conformational rearrangements of ␤ 3 CT coupled to a considerable perturbation of its interaction with the membrane. Together, these data define a critical role of tyrosine phosphorylation, in general, in the regulation of signal transduction as well as in controlling ␤ 3 integrin function.
Expression and Purification-Cloning, expression, and purification of ␣ IIb CT, ␤ 3 CT (non-phosphorylated form hereafter referred to as ␤ 3 NP), and MBP-␣ IIb have been described previously (7). To produce 15 N and/or 13 C isotopically labeled ␣ IIb CT and ␤ 3 CTs, cells were grown in M9 minimal medium containing 15 NH 4 Cl (1.1 g/liter) and 13 C glucose (2.5 g/liter) as the sole source of nitrogen and carbon. Tyrosine phosphorylation of ␤ 3 CT (mono-phosphorylated at Tyr-747 , hereafter referred to as ␤ 3 MP, and bi-phosphorylated at Tyr 747 and Tyr 759 , hereafter referred to as ␤ 3 BP) has been achieved in vivo by using TKB1 bacterial cell line from Stratagene following the manufacturer's protocol for the recombinant protein induction. Single amino acid mutations were made by using the QuikChange site-directed mutagenesis kit (Stratagene).
Mass Spectroscopy-Mass spectral analyses were performed on a quadrupole time-of-flight (Q-TOF) mass spectrometer (QSTAR Elite) equipped with an ESI source. The data acquisition was under the control of the Analyst QS software (Foster City, CA). All samples were dissolved in methanol:water mixture to achieve final concentration of 40 M. Samples were infused into the ESI source at a flow rate of 10 l/min by using the built-in syringe pump. Typical source conditions for Q-STAR were as follows: capillary voltage (5500V), declustering potential (215V), resolution (15000, full width-half maximum).
Tryptophan Fluorescence Quenching-Steady-state fluorescence was measured with a SPEX Fluorolog FL3-22 spectrometer (Jobin Yvon, Edison, NJ) equipped with double-grating excitation and emission monochromators. Emission scans for Trp fluorescence ( ex ϭ 295 nm; em ϭ 310 to 400 nm at 1-nm intervals with 2 nm and 4 nm excitation and emission bandpass, respectively) were performed using samples (20 -37 M protein, pH 5.9) in 4 ϫ 4 mm quartz microcells. Iodide Trp quenching was measured by titrating two equivalent samples in parallel with aliquots from either 2.5 M KI or 2.5 M KCl stocks, each containing 5 mM Na 2 S 2 O 3 . Maximum intensities from each KI titration point were corrected by the KCl-containing samples to account for dilution and ionic strength and analyzed according to the Stern-Volmer law: (F 0 /F) Ϫ 1 ϭ K SV [I Ϫ ] where F 0 and F are the net intensities in the absence and presence of I Ϫ , respectively, and K SV is the Stern-Volmer quenching constant. All experiments were repeated three times.
EC Transfection and Adhesion Assay-Lung EC at passage 2 were suspended at 5 ϫ 10 6 /ml in Optimem media with 10 g/ml pCDNA3.1 (empty or expressing ␤ 3 WT or mutants) and 2 g/ml pMAX-GFP. 100-l portions were transferred to nucleofection cuvettes and pulsed using amaxa Biosystems nucleofector, program M-003. The cells were plated in growth media on VN-coated dishes, and sodium butyrate (at 5 mM final concentration) was added. 96-well plates were coated 1 h at 37°C with 0.1 g/ml VN, blocked 1 h at 37°C with 1% heatdenatured BSA in PBS and washed in 3ϫ PBS. Transfected cells were harvested 72 h post-transfection by brief trypsinization, washed, and suspended in DMEM:F12 containing 0.2% BSA. 2 ϫ 10 4 cells in 60 l volume were plated per well and allowed to adhere for 25 min. The wells were washed 5ϫ in DMEM:F12 and the cells fixed 10Ј in 2% formaldehyde in PBS. GFP images at 5x magnification were acquired and composite images encompassing the entire wells were constructed. Amounts of GFP-positive cells per well were quantified using ImageJ software. Portion of the transfected cells were analyzed by FACS using ␣ v ␤ 3 antibodies to compare expression levels of the ␤ 3 constructs.
NMR Spectroscopy-Chemical shift assignments of ␤ 3 CT have been determined previously (7) and have been modified to address the effects of phosphorylation and changes in pH values. All the NMR experiments were performed on Varian 600MHz and 800 MHz equipped with inverse-triple resonance cold-probes and were processed with NMRPipe (9) and analyzed by CCPN software suite (10). Earlier we have used aqueous conditions (pH 6.1) to understandthe ␣ IIb ␤ 3 heterodimer interface (7). However, because of the apparent solubility issues of phosphorylated constructs at pH 6.1, pH of the ␤ 3 MP samples was reduced to 5.9 to achieve sufficient concentrations for structure determination. Both ␤ 3 NP and ␤ 3 MP did not demonstrate substantial pH-dependent conformational differences judging by their chemical shift perturbation data (supplemental Fig. S4A). 1 H-15 N HSQC titration experiments (Fig. 3, C and D) were performed in water at 25°C at pH 6.1. Transferred NOESY experiments (Fig. 3, E and F) for different peptides were performed at 25°C at pH 6.1. Different ratios of the peptides to the binding partner were investigated to find the optimal range for NOE transfer for each particular analysis. The resonance assignments of unlabeled peptides were made using conventional two-dimensional 1 H-1 H TOCSY and NOESY spectra (11) by CCPN software suite (10).
To characterize the structures and membrane-binding properties of phosphorylated ␤ 3 CTs, 0.07-0.9 mM 15 N-and/or 13 Clabeled ␤ 3 CTs (␤ 3 NP, ␤ 3 MP and ␤ 3 BP) were dissolved in 60 -300 mM deuterated DPC solution (Sigma-Aldrich) prepared in 20 mM sodium phosphate buffer, 5 mM Ca 2ϩ at pH 5.9. The pH was monitored with pH strips (EMD Chemicals). All NMR experiments involving membrane-mimetic conditions were performed at 40°C. To determine the location of different ␤ 3 constructs relative to the micelle surface, steric acid compounds (16-DSA and 5-DSA) were used. Both were dissolved in 50 mM deuterated DPC solution to make 50 mM stock solution. These solutions were then added to the proteinϩ DPC solution to achieve the following final ratios of protein: 5/16 DSA: DPC: 1:10:750 (␤ 3 NP); 1:12:1000 (␤ 3 MP); 1:14:1000 (␤ 3 BP). The effects of the spin labels were observed by comparing the peak intensities (supplemental Fig. S6) in 1 H-15 N HSQC spectra. For calculating the intensity ratios, the spectra were processed with 10 Hz exponential broadening in direct dimension and zerofilled to 2048 ϫ 1024 data points in t2 and t1, respectively. For the NMR dynamics study of ␤ 3 NP, 1 H-15 N NOE, 15 N T 1 , and T 2 data under aqueous (25°C) and membrane mimetic conditions (40°C) were collected on a Varian Inova 600 MHz spectrometer. 15 N T 1 values were measured from the spectra recorded with 8 different durations of the delay: T ϭ 30, 90, 150, 250, 400, 600, 800, 1200 ms. 15 N T 2 values were determined from spectra recorded with 8 different durations of the delay: T ϭ 10, 20, 30, 50, 70, 90, 110, 150 ms. Steady-state hetero-nuclear 1 H-15 N NOE values were determined from spectra recorded with 5 s relaxation delay and the presence and absence of a proton presaturation period of 5 s. T 1 , T 2 , and NOE values were extracted by a curve-fitting subroutine included in the CCPN software suite (10) (supplemental Fig. S7). The rotational correlation time ( c ) values were estimated to be ϳ5 ns in aqueous solution and 9 nanoseconds in DPC using TENSOR-2 (32) indicating the differences in overall tumbling associated with micelles binding.
Structure Calculation- Table 1 lists detailed structural statistics of the final fifteen lowest energy conformers of ␤ 3 NP, ␤ 3 MP, and ␤ 3 BP under aqueous and membrane-mimetic conditions along with the two-dimensional and three-dimensional NMR experiments utilized for individual structure determination. For ␤ 3 MP and ␤ 3 BP in presence of DPC micelles, the backbone, and , dihedral angle restraints were obtained by using Talosϩ (12). All the initial structure calculations were performed using CYANA 2.1 (13). Hydrogen bond restraints (in the case of ␤ 3 MP under aqueous conditions) were introduced during the final stages of calculations. Sixty lowest energy structures from CYANA were subjected to molecular dynamics simulations in explicit water (14) using CNS (15). For ␤ 3 NP under membrane-mimetic conditions, sixty structures were calculated by utilizing previously (8) determined NOE and dihedral restraints with the help of CYANA and later were refined in explicit water to maintain consistency and for a more accurate comparison. None of the structures have NOE and dihedral angle violations more than 0.5 Å and 5°, respectively. The Protein Structure Software suite (PSVS; courtesy of CABM Structural Bioinformatics Laboratory, Rutgers State University of New Jersey) was used for structure quality assessment and validation.

RESULTS
Preparation of Tyrosine-phosphorylated Integrin ␤ 3 Cytoplasmic Tail for Structural Analysis-Phosphorylation has long been considered as a critical regulatory apparatus in signal transduction and nuclear magnetic resonance (NMR) spectroscopy is a pertinent technique for deciphering the emanating conformational changes imparted by phosphorylation. However, the first step in investigation of phosphorylation by NMR, production of the phosphorylated, isotopically labeled proteins in adequate amounts, is usually an uphill task. In the case of ␤ 3 CT (see supplemental Fig. S1 for sequence details), the TKB1 bacterial cell line (Stratagene), carrying Elk tyrosine kinase gene controlled by the trp promoter, was found to yield sufficient quantities of (tyrosine)-phosphorylated protein. Although it has been suggested that this in vivo approach is not as efficient as the in vitro technique due to the deleterious effects of Elk tyrosine kinase on the bacteria (16), we could produce reasonable amounts of phosphorylated ␤ 3 by using unusually short induction times (IPTG induction of ␤ 3 at A 600 ϳ0.3 followed in 2 h by tryptophan induction of Elk). Supplemental Fig. S2 (SM) depicts the reversed phase, RP-HPLC chromatogram of ␤ 3 CT expressed in TKB1 cells and the deconvoluted mass spectra of the three HPLC peaks. The MS analysis reveals that the peaks eluting at 23, 24, and 26% of acetonitrile gradient correspond to the bi-phosphorylated, mono-phosphorylated, and non-phosphorylated ␤ 3 CT, respectively. Our initial assumption was that we will achieve almost equal populations of the two monophosphorylated ␤ 3 CT constructs (pY 747 and pY 759 respectively) which would be very difficult to separate. However, a closer inspection of superimposed 1 H-15 N HSQC spectra of these three HPLC peaks, Fig. 1A and supplemental Fig. S3A, demonstrates that the middle peak in the chromatogram is pY 747 -␤ 3 CT and does not contain any pY 759 -␤ 3 CT. Though Elk is known to be a promiscuous tyrosine kinase, the efficiency of phosphorylation in ␤ 3 CT appeared to be very different for the two tyrosine residues (Tyr 747 , Tyr 759 ). While we could produce NMR quantities of pY 747 -␤ 3 CT (hereafter referred to as ␤ 3 MP) and extremely limited quantities of pY 747 , pY 759 -␤ 3 CT (hereafter referred to as ␤ 3 BP); we could not generate any pY 759 -␤ 3 CT. One possible explanation for the lack of pY 759 product may be the flexibility and dynamic nature of the ␤ 3 CT's C terminus in the absence of pY 747 . To substantiate this hypothesis, we introduced conservative point Y747F or Y759F mutations in ␤ 3 CT construct by using site-directed mutagenesis (QuikChange). Supplemental Fig. S3B depicts the superimposition of 1 H-15 N HSQC spectra of these mutants with the wild type ␤ 3 CT. Expression of these mutants in TKB1 cells confirmed our observation. As in the case of their wild type counterparts, we could produce pY 747 for Y759F mutant (supplemental Fig.  S3C), but were unable to phosphorylate the Y747F mutant.
Phosphorylation of Tyr 747 Results in Structural Rearrangement of ␤ 3 CT under Aqueous Conditions-The superimposition of 1 H-15 N-HSQC spectra for ␤ 3 NP, ␤ 3 MP and ␤ 3 BP under aqueous conditions is shown in Fig. 1A and the subsequent chemical shift perturbations are presented in Fig. 1C. From these data, it is clear that phosphorylation not only affects the nearby residues ( 743 NN-LpYKEA 750 and 757 ITpYRGT 762 ) but also influences the membrane-proximal region (K 716 -D 723 ). These latter chemical shift perturbations are common to both, ␤ 3 MP and ␤ 3 BP constructs, suggesting an intramolecular interaction between the membrane-proximal and 744 NPLpY 747 regions. This interaction is a direct result of Tyr 747 phosphorylation, most probably due to a formation of the salt bridge between the negatively charged phosphate group and the positively charged/polar side-chain(s) of the N-terminal amino acid(s). To better understand this change, we acquired 1 H-1 H two dimensional (2D) Nuclear Overhauser Enhancement Spectroscopy (NOESY) and three-dimensional 15 N-edited NOESY spectra of ␤ 3 MP, which allowed us to structurally characterize ␤ 3 MP under aqueous conditions. The overall fold of ␤ 3 MP is shown in Fig. 2, A and B (PDB ID: 2ljf, see statistics in Table 1). Interestingly, the negatively charged phosphate group of pY 747 creates a salt-bridge with the positively charged side-chain of Lys 738 , which, in turn, affects the orientation of Trp 739 indole side chain. This bulky, hydrophobic side chain forges contacts with the methyl groups of membrane-proximal residues (Leu 717 is shown as an example, Fig. 2B) and the resulting compact conformation is then further stabilized by hydrogen bond between the side-chains of D 740 -N 743 . The first turn of the membrane-proximal helix, found in the ␣ IIb ␤ 3 heterodimer (7), is not formed in ␤ 3 MP and the helical region spans only from Lys 725 to Lys 729 . The C-terminal region (F 754 -T 762 ) is dynamically unstructured. Interaction of Trp 739 with the membraneproximal residues supports the notion that it could be situated near, but not necessarily within, the membrane. This is consistent with our prior findings that both Trp 739 and Tyr 747 can interact with phospholipids (8). Compared with the compact conformation of ␤ 3 MP, ␤ 3 NP is much more dynamic and largely unstructured, as has been reported previously (7,17), except for a reverse turn formed around the 744 NPLY 747 motif and helical tendencies in the membrane proximal region (Fig. 2, E and F, statistics of the ensemble are presented in Table 1). Moreover, the dynamic nature of ␤ 3 NP under aqueous conditions is supported by NMR relaxation measurements as most of the HetNOE values are below 0.3 (supplemental Fig. S7, A-D). Sequential connectivity maps for both ensembles are provided (supplemental Fig. S5).
It is imperative to mention that in the case of ␤ 3 MP, although the chemical shift changes due to pY 747 are very modest and absolute chemical shift values are close to the random coil values, we were able to observe long-range NOEs. In dynamic systems where the folded, transient, and unstructured conformers are in fast exchange, the observed NOEs can be averaged over multiple conformers, and not all the contacts are satisfied by a single conformer. Any attempt to fulfill all such NOE contacts simultaneously might lead to an over-constrained system. Hence, to independently confirm this novel compact conformation of ␤ 3 MP is not artificial, we performed iodide quenching of tryptophan fluorescence to determine the relative exposure of W 739 indole side chain of ␤ 3 MP and ␤ 3 NP to the aqueous environment (Fig. 2C). The measured K SV values were lower for ␤ 3 MP (K SV ϭ 3.09 Ϯ 0.43 M Ϫ1 ) than for ␤ 3 NP (K SV ϭ 4.35 Ϯ 0.56 M Ϫ1 ). Our steady-state emission scans strongly suggest that the higher K SV values of the ␤ 3 NP are not due to the higher unquenched fluorescence lifetime because the specific fluorescence intensities and max values are equivalent for both proteins. This indicates that Trp 739 in ␤ 3 MP is indeed better shielded from solvent exposure than its non-phosphorylated counterpart by virtue of its compact fold and due to the hydrophobic interactions with the N-terminal residues. In addition, we investigated the importance of pY 747 -K 738 salt bridge by using NMR salt titrations. The rationale behind these experiments was that the electrostatic interactions responsible for the ␤ 3 MP fold should be affected by changes in salt concentrations. Supplemental Fig. S4B displays the distribution of chemical shifts for ␤ 3 MP under the different ionic strength buffers. High salt concentration, indeed, has a major impact on the changes in chemical shifts resulting in significant decline of chemical shift perturbations throughout the membrane proximal region of ␤ 3 CT.
To further validate and test the biological significance of this fold, we performed site-directed mutagenesis, point mutation K738E (Quikchange) (see supplemental Fig. S1 for sequence details) and investigated both in vitro and in vivo properties of this mutant. Residue Lys 738 was selected for this analysis as it plays a critical role in the conformation change of ␤ 3 MP and as per our knowledge has not been implicated in any interactions with known integrin modulators, such as talins or kindlins. Hence the effect of this mutation on integrin activation state should be attributed to the structural integrity rather than external factors. Theoretically, this K738E charge reversal should abolish the salt-bridge in ␤ 3 MP, making the conformational change due to Tyr 747 phosphorylation challenging. As expected, the chemical shift perturbations in the membrane proximal region due to Tyr 747 phosphorylation are smaller than its wild type (WT) counterpart and are randomly distributed  NOVEMBER 25, 2011 • VOLUME 286 • NUMBER 47 throughout the sequence with the major local effect shifted in C-terminal direction (supplemental Fig. S4C). Thus it can be argued that under in vitro conditions pY 747 -K738E mutant probably exhibits behavior similar to that of ␤ 3 NP. The effect of this charge reversal was next tested in ␤ 3 integrin-dependent endothelial cell (EC) adhesion to vitronectin (VN). EC isolated from ␤ 3 knock-out (KO) mice were transfected with expression vectors for either WT␤ 3 or substitution mutants Y747F, K738E, Y747F-K738E (see supplemental Fig. S1 for sequence details). As shown in Fig. 2D, expression of both K738E and Y747F mutants showed decreased adhesive response as compared with WT ␤ 3 integrin expressing cells. Importantly, adhesion of K738E-expressing cells was similar to that of Y747F mutant and no further inhibition was observed in double K738E/Y747F mutant cells. As an additional control, K738E mutation was tested to find out whether it can reduce integrin ability to undergo activation, for example by preventing its interactions with talins or kidlins. As expected, this mutation did not result in diminished soluble fibrinogen binding mediated by ␣ IIb ␤ 3 integrin (supplemental Fig. S4D). Thus, the resulting differences in the activation states between this mutant and the wild type are related to the internal structural integrity. Combined, these data confirm the critical role of K 738 /Y 747 for the regula-tion of integrin-mediated cell adhesion and biological significance of novel ␤ 3 MP conformation.

Tyrosine Phosphorylation in Integrins
Because of the challenges with sample preparation, we could not investigate structural details of ␤ 3 BP under aqueous conditions, but considering the similarity of chemical shifts of N-terminal residues between ␤ 3 MP and BP (Fig. 1, A and C) and the dynamic nature of the C terminus, it is safe to suggest that ␤ 3 BP accommodates a conformation similar to ␤ 3 MP and is distinct from ␤ 3 NP. Together these data indicate that upon phosphorylation ␤ 3 CT undergoes a substantial structural change, leading to a more compact conformation, which, in turn, might affect interactions between the receptor and intracellular adaptors.
Tyrosine Phosphorylation Preserves the Activated State of Integrin by Preventing the Interaction between ␤ 3 and ␣ IIb Cytoplasmic Tails-In a previous study (7), we have structurally characterized the cytoplasmic domain of ␣ IIb ␤ 3 heterodimer. Our data revealed the underlying mechanism by which the inter-subunit clasp, R 995 (␣ IIb CT)-D 723 (␤ 3 CT) along with several other electrostatic and hydrophobic contacts (Fig. 3A), maintains the integrin in a resting state. Termination of these interactions eventually results in integrin activation. As tyrosine phosphorylation leads to a conformational rearrangement of ␤ 3 CT, we were curious to find out whether these changes  (12). b Hydrogen bonds were introduced in the last stage of structure calculations. c The following residues are considered for the rmsd calculations: I) ␤ 3 NP(water): residues 720 -735 II) ␤ 3 MP(water): residues 720 -745 III) ␤ 3 NP(DPC): residues 722-745 IV) ␤ 3 MP(DPC): residues 720 -745 V) ␤ 3 BP(DPC): residues 720 -745. d After refinement in explicit water by using CNS (15). e All residues, calculated using the Protein Structure Software suite. f 400 ms mixing time. g 300 ms mixing time. h 150 ms mixing time.
affect the formation of ␣ IIb ␤ 3 heterodimer. Superimposition of ␤ 3 MP structure with ␣ IIb ␤ 3 heterodimer revealed a steric clash between ␤ 3 MP and the ␣ IIb subunit indicating the possible difficulties in clasp formation (Figs. 3B and 5A). To test this prediction, we performed chemical shift mapping experiments similar to those done before to define ␣ IIb ␤ 3 cytoplasmic clasp (7). Non-labeled ␤ 3 NP, ␤ 3 MP, and ␤ 3 BP were titrated with 15 Nlabeled ␣ IIb CT and the associated chemical shift perturbations were monitored. Chemical shift changes plotted as a function of residue numbers in ␣ IIb CT (Fig. 3C) indicate a gradual decrease upon ␤ 3 CT phosphorylation. These observations were also supported by the opposite experiments, where non-labeled ␣ IIb CT is titrated into solutions of 15 N-labeled ␤ 3 NP, ␤ 3 MP, and ␤ 3 BP. Fig. 3D shows the subsequent chemical shift changes plotted as a function of the residue number in ␤ 3 CT. It is important to mention that residues Lys 748 and Glu 749 are more perturbed in ␤ 3 MP than in ␤ 3 NP. This could indicate either the appearance of a new binding site, or an internal conformational rearrangement in ␤ 3 tail. To examine the possibility of interac-tion between ␣ IIb CT and residues Lys 748 and Glu 749 of ␤ 3 CT, we utilized the transferred NOE (trNOE) method (18). This approach is well suited for characterization of such weak interactions and has been used in the elucidation of ␣ IIb ␤ 3 structure (7). The method detects appearance of additional peaks in the ligand's NOESY spectra upon its interaction with the target protein (19). ␣ IIb CT was fused to maltose-binding protein (MBP) tag to increase the molecular weight as higher molecular weight allows more favorable NOE transfer, the effect proven experimentally even despite of some independent local ␣ IIb CT motion (7). Because the full-length ␤ 3 tail has limited solubility (20), we have used shorter ␤ 3 peptides, NMP␤ 3 containing the N-terminal residues, including pY 747 and BP␤ 3 Pep containing the C-terminal residues, including both pY 747 and pY 759 (see supplemental Fig. S1, SM, for sequence details). No additional peaks were detected in NOESY spectra of these peptides upon addition of MBP-␣ IIb under any of the conditions tested (Fig. 3,  E and F). Thus based on all these data, it can be concluded that the chemical shift perturbations observed for residues Lys 748 Transferred NOESY experiments indicating the lack of ␣ IIb -␤ 3 MP/BP interactions; E, superimposition of NOESY spectra for NMP␤ 3 in absence(black) and presence (lime) of MBP-␣ IIb at the ratio 20:1; F, superimposition of NOESY spectra for BP␤ 3 Pepin absence (black) and presence (lime) of MBP-␣ IIb at the ratio 20:1. Peaks marked with * represent peaks from MBP-tag. All the transferred NOESY experiments were performed at 25°C (400 ms mixing time) in 0.25ϫ PBS buffer in presence of 6 mM CaCl 2 to stabilize ␣ IIb as described earlier (Vinogradova et al.,25). and Glu 749 are due to an internal structural rearrangement. And tyrosine phosphorylation of ␤ 3 CT, indeed, prevents ␤ 3 from making and/or maintaining contacts with ␣ IIb CT, thereby preserving the activated state of the receptor.
Phosphorylation Affects ␤ 3 CT Interaction with the Membrane-To address the effects of tyrosine(s) phosphorylation on ␤ 3 CT's interaction with lipid bilayer, we next investigated ␤ 3 MP and ␤ 3 BP in DPC detergent micelles. DPC has been used extensively as a membrane mimetic for NMR studies and was previously utilized to structurally characterize ␤ 3 NP (8) and other similar constructs (21,22). In the presence of DPC micelles, ␤ 3 NP exhibits much more structured conformation even without its binding partner ␣ IIb CT (8). Moreover, we have demonstrated that several residues of ␤ 3 NP (Trp 739 , Thr 741 , Ala 742 , Pro 745 , and Tyr 747 ) could interact with DPC micelles and these interactions initiate the formation of a second short ␣-helical region (Leu 746 -Asn 756 ), which is not generally observed in either aqueous ␤ 3 or ␣ IIb ␤ 3 heterodimer. These conclusions are also corroborated by the NMR relaxation measurements (supplemental Fig. S7, E-H). The whole ␤ 3 NP is tumbling along with the micelles except the last three C-terminal residues (RGT 763 ) which are undergoing a significant local motion. However, both ␤ 3 MP and ␤ 3 BP behaved very differently as compared with ␤ 3 NP in DPC micelles. The superimposition of 1 H-15 N-HSQC spectra of ␤ 3 NP/MP/BP in DPC is depicted in Fig. 1B and the resultant chemical shift perturbations are presented in Fig. 1D. The shift differences between non-phosphorylated and phosphorylated constructs are quite significant (Fig. 1D). This may be due to the alterations in secondary structural elements because of the distinctive ␤ 3 -membrane interaction. Interestingly, the chemical shifts are almost identical for the affected mid-region (residues E 731 -T 753 ) in ␤ 3 MP and ␤ 3 BP cases, indicating the likeness of the conformations in membrane environment. For ␤ 3 BP, as expected, the additional shift changes were associated with the second phosphorylation site (the C terminus residues, F 754 -T 762 ) highlighting the possible differences from ␤ 3 MP. To confirm these hypotheses, we have performed a full scale NMR structural investigation of ␤ 3 MP (and partially for ␤ 3 BP) in DPC micelles. We have also tested the interactions of all three constructs (␤ 3 NP/MP/BP) with the DPC micelles using the Paramagnetic Relaxation Enhancement (PRE) approach (23).
The overall folds of ␤ 3 NP, ␤ 3 MP and ␤ 3 BP are shown in Fig.  4, A, C, and E, respectively (PDB IDs: 2ljd and 2lje for ␤ 3 MP and ␤ 3 BP, respectively. The statistics are shown in Table 1; see under "Experimental Procedures" for additional details. Sequential connectivity maps are provided in supplemental Fig.  S5). The prominent structural features of ␤ 3 NP in presence of DPC micelles, a membrane-proximal ␣-helix (K 716 -R 734 ) followed by a flexible loop and another short helix (Y 747 -T 755 ) (8), are extensively modified due to tyrosine phosphorylation. The membrane-proximal ␣-helix is slightly longer (K 716 -K 738 ), however the C-terminal region directly following pY 747 is no longer helical except for some helical tendencies in A 750 -T 753 region. Another surprising finding is that the kink at residues D 723 /R 724 in ␤ 3 NP, which allows the helix to bend, bringing the flexible loop (K 738 -A 742 ) into possible contact with the membrane surface, has shifted toward the residues K 725 /E 726 in ␤ 3 MP and ␤ 3 BP. The angles between the two portions of the membrane-proximal helices are not very well defined in all three structures. The recent structural study (21) of the nonphosphorylated ␤ 3 construct, where several additional N-terminal transmembrane residues of ␤ 3 were cross-linked with ␣ IIb subunit, has reported that the residue Arg 724 of ␤ 3 CT formed a single-residue hinge and the angle between the two parts of the membrane proximal helix, defined based upon intramolecular NOEs between residues Phe 727 and Ile 721 , is about 100°. We however, could not find evidence of the above mentioned NOEs in any of our NOESY experiments. The biological significance of mutual orientation of these two portions of the membraneproximal helices requires further investigation as it could easily reflect the consequences of higher surface curvature of the micelles, in comparison with mostly flat lipid bilayer.
For ␤ 3 BP in the presence of DPC micelles, the lack of distance restraints (single three-dimensional 15 N-edited NOESY-HSQC experiment) resulted in inadequate structural convergence. Because of the challenges in preparation of 13 C, 15 N-labeled  (29). A, C, E, backbone superimposition on fifteen lowest energy conformers of ␤ 3 NP, ␤ 3 MP, ␤ 3 BP, respectively; B, D, F, ribbon representation of ␤ 3 NP, ␤ 3 MP, ␤ 3 NBP conformers closest to the mean structures. The intensity ratios from the PRE experiments (␤ 3 NP/MP/BPϩ16-DSAϩDPC) are mapped onto the surfaces of ␤ 3 NP, ␤ 3 MP, and ␤ 3 BP respectively. Green to orange to white color gradient is used to map the PRE intensities (respective color keys are shown adjacent to the figure). Overlapping residues, residues with missing information, and prolines are marked in gray. ␤ 3 BP we could not perform 3D 13 C -edited NOESY-HSQC experiment. To circumvent this issue, since the chemical shifts for the N-terminal residues (K 716 -A 750 ) between ␤ 3 MP and ␤ 3 BP were virtually identical (Fig. 1, B and D), we introduced the additional distance restraints corresponding to these N-terminal amino acids from ␤ 3 MP in ␤ 3 BP structure calculations. These additional restraints have resolved the issue of convergence. As expected, the structure is very similar to ␤ 3 MP except a slightly sharper kink in membrane proximal helix and the orientation of C terminus residues ( 752 S-G 762 ). The crucial differences in C terminus arise due to the phosphorylation of Tyr 759 , which probably affects the orientation of the 756 NITYR 760 motif. In contrast to ␤ 3 MP, where Tyr 759 interacts with the membrane (for more details, see below), in the case of ␤ 3 BP, pY 759 is pointing in opposite direction away from the membrane due to the repulsion between the negatively charged phosphate groups of pY 759 and DPC.
To determine how tyrosine(s) phosphorylation alters the membrane binding, we utilized the PRE approach. Two paramagnetic relaxation agents which selectively partition in hydrophobic environment, 5-doxyl stearic acid (5-DSA) and 16-DSA, (24) were introduced into the DPC micelles and the consequent drop in the intensities of the amide peaks of all three ␤ 3 constructs was monitored. The doxyl moiety in 16-DSA is attached to the very end of the aliphatic chain and thus gets localized at the center of DPC micelles. In 5-DSA, on the other hand, the doxyl moiety is situated close to the polar head group and the membrane-water interface. Both these tags were utilized to determine the membrane-embedded residues. Supplemental Fig. S6 represents the intensity ratios of the backbone amide groups of ␤ 3 NP/MP/BP upon titration with 5-DSA and 16-DSA and Fig. 4, B, D, and F depict these ratios (selected for 16-DSA) mapped on the surfaces of ␤ 3 NP/MP/BP structures representing the direct contacts with the micelles. In the case of ␤ 3 NP, we confirmed some of our earlier intermolecular NOEs findings (8). The membrane-proximal residues (L 717 -I 721 ) and Tyr 747 of ␤ 3 NP are, indeed, inserted into the membrane. Moreover, the region 738 KWD 740 is associated with the membrane surface judging by the drop in intensity ratios upon titration with 5-DSA, but not with 16-DSA (supplemental Fig.  S6, A and B). Surprisingly, the C-terminal 756 NITYR 760 motif of ␤ 3 NP also shows significant drop in peaks intensities in both cases (5-DSA and 16-DSA) and is probably associated with the membrane. Previously we could not detect any intermolecular NOEs to support this finding, which is due to the highly dynamic nature of this C-terminal region as confirmed by 15 N relaxation data (supplemental Fig. S7). Although the intensity ratios are rather similar for all ␤ 3 constructs, we do see specific differences in the patterns of membrane association upon tyrosine phosphorylation. In ␤ 3 MP, as in the case of ␤ 3 NP, the membrane-proximal residues (L 717 -I 721 ) are membrane embedded and the C-terminal 756 NITYR 760 motif is either membrane embedded or associated. However, unlike ␤ 3 NP, neither 738 KWD 740 nor pY 747 are inserted into or associated with the membrane (supplemental Fig. S6, C, D), which is, most probably, a direct result of tyrosine phosphorylation. The charge repulsion between the negatively charged phosphate groups of ␤ 3 MP and DPC may not allow the pY 747 to come within close proximity to the membrane even in the presence of counterbalancing positively charged choline group. This, in turn, might affect the orientation of 738 KWD 740 motif, resulting in Trp 739 pointing in opposite direction from the membrane (Fig. 4, C and F). In ␤ 3 BP, similar to ␤ 3 NP/MP, the membraneproximal residues (L 717 -I 721 ) are inserted into the membrane. However, due to the phosphorylation of both tyrosines, neither the 738 KWD 740 , pY 747 nor 756 NITYR 760 motifs are inserted or associated with the membrane. In summary, we have not only confirmed our earlier findings about ␤ 3 interaction with phospholipids but also have gained some novel insights into how tyrosine phosphorylation affects these interactions and causes major conformation changes.
Our data show that the phosphorylation of ␤ 3 precludes this weak interaction between ␣ and ␤ cytoplasmic tails and, therefore, may play a critical role in maintaining the active state of the receptor during outside-in signaling. Closer examination of ␤ 3 MP structure reveals several features supporting this conclusion. The superimposition of ␤ 3 MP over ␤ 3 NP in complex with ␣ IIb (PDB ID 1M8O) shows severe steric clashes between ␤ 3 MP and the ␣ IIb subunit (Fig. 5A). In addition, the heterodimeric complex of integrin tails involves both hydrophobic and electrostatic interactions with Asp 723 and Glu 726 of ␤ 3 CT forming a salt bridge with Arg 995 of ␣ IIb . However, in the case of ␤ 3 MP, Asp 723 and Glu 726 are pointing in opposite direction. Plus, in ␣ IIb ␤ 3 complex, Leu 718 of ␤ 3 is involved in a hydrophobic interaction with Val 990 of ␣ IIb . In ␤ 3 MP, however, the membrane proximal hydrophobic residues interact with Trp 739 aromatic side chain. Based on our titration experiments, we can speculate that these intramolecular interactions in ␤ 3 MP, although weak, are comparatively stronger than the weaker intermolecular interactions necessary to form ␣ IIb ␤ 3 complex. Because of several technical difficulties, we do not have structural data for the ␤ 3 BP under aqueous conditions. However, based on the similarity of phosphorylation-dependent chemical shift perturbations in the membrane-proximal region (K 716 -D 723 ), we can propose that the ␤ 3 BP maintains a conformation similar to ␤ 3 MP. Moreover, the NMR titration experiments for ␤ 3 BP show a significant reduction in the associated chemical shift perturbations, suggesting that the bi-phosphorylation of ␤ 3 CT also favors disruption of the inter-subunit clasp. Our structural analysis emphasizes the role of electrostatic interaction between Lys 738 and pY 747 in the maintenance of the compact fold of phosphorylated ␤ 3 CT. When expressed in ␤ 3 -null endo-thelial cells, K738E, Y747F, and K738E/Y747F mutants diminished cell adhesion response as compared with WT integrin, thus supporting the notion that the phosphorylation-induced fold supports ␤ 3 integrin functional activity.
In this study, we also have accumulated the first direct evidence that tyrosine phosphorylation affects the structure and the association of ␤ 3 CT with membrane. As visualized in Fig.  5B, charge repulsion pushes the phosphorylated tyrosine(s) away from the membrane, probably exposing alternative motifs to interact with different potential integrin-associated proteins, thus providing an additional level of complexity to the regulatory mechanisms employed in integrin signaling. A model for diversity of such interactions is presented in Fig. 5C based on our results. The resting state of the receptor, with the clasp between ␣ and ␤ subunits located within the cytosol is depicted in Fig. 5C, I. When integrin is activated, for example by talin head domain (Fig. 5C, II), ␣ and ␤ subunits are separated. After talin dissociation the re-clasping can be prevented and the activated state of the receptor maintained by tyrosine phosphorylation of the ␤ subunit within cytosol (Fig. 5C, IV) as well as by membrane association of the membrane-proximal helices of a single or both subunits (Fig. 5C, III). The other functionally significant outcome of tyrosine(s) phosphorylation is the redirection of the ␤ subunit mid-region and/or C terminus away from the phospholipid bilayer, thus allowing different adaptor proteins to bind, and, as exemplified in the case of Shc (Fig. 5C,  V), propagating outside-in signaling events within cytoplasm.
To conclude, in this study, we have performed detailed NMR analysis of the effects of tyrosine(s) phosphorylation on integrin ␤ 3 CT under both aqueous and membrane-mimetic conditions. We have shown that the phosphorylation causes significant conformational rearrangement in ␤ 3 CT under solution conditions where the pY 747 containing segment folds back and interacts with the membrane-proximal region. This arrangement prevents the ␤ 3 CT from binding to ␣ IIb CT, thus likely dictating an unclasped state of the receptor necessary to mediate integrin outside-in signaling. Moreover, tyrosine(s) phosphorylation under membrane-mimetic conditions modifies ␤ 3 CT's interaction with the membrane and perturbs its overall fold. By preventing the phosphorylated tyrosines containing regions from being inserted into or associated with the lipid bilayer phosphorylation might shift the equilibrium of integrin interactions with the different cytoplasmic adaptor proteins, adjacent receptors, and/or cytoskeleton. Our data provide a structural basis for the critical role of Tyr 747 phosphorylation in controlling ␤ 3 CT function and shed light upon molecular details of how phosphorylation may play multiple roles in regulating different states of cell surface receptors, suggesting a more complex paradigm than a simple two state (active/inactive) model.