Allosteric activation of proto-oncogene kinase Src by GPCR–beta-arrestin complexes

G protein–coupled receptors (GPCRs) initiate signaling cascades via G-proteins and beta-arrestins (βarr). βarr-dependent actions begin with recruitment of βarr to the phosphorylated receptor tail and are followed by engagement with the receptor core. βarrs are known to act as adaptor proteins binding receptors and various effectors, but it is unclear whether in addition to the scaffolding role βarrs can allosterically activate their downstream targets. Here we demonstrate the direct allosteric activation of proto-oncogene kinase Src by GPCR–βarr complexes in vitro and establish the conformational basis of the activation. Whereas free βarr1 had no effect on Src activity, βarr1 in complex with M2 muscarinic or β2-adrenergic receptors reconstituted in lipid nanodiscs activate Src by reducing the lag phase in Src autophosphorylation. Interestingly, receptor–βarr1 complexes formed with a βarr1 mutant, in which the finger-loop, required to interact with the receptor core, has been deleted, fully retain the ability to activate Src. Similarly, βarr1 in complex with only a phosphorylated C-terminal tail of the vasopressin 2 receptor activates Src as efficiently as GPCR–βarr complexes. In contrast, βarr1 and chimeric M2 receptor with nonphosphorylated C-terminal tail failed to activate Src. Taken together, these data demonstrate that the phosphorylated GPCR tail interaction with βarr1 is necessary and sufficient to empower it to allosterically activate Src. Our findings may have implications for understanding more broadly the mechanisms of allosteric activation of downstream targets by βarrs.

G protein-coupled receptors (GPCRs), the largest group of membrane proteins, regulate virtually all physiological processes and represent the most common drug targets (1). GPCRs translate various extracellular stimuli into specific cellular responses via activation of signal transducers: heterotrimeric Gproteins and beta-arrestins (barr). barrs, initially discovered as proteins that desensitize G-protein signaling (2,3), are now recognized as signal transducers in their own right (reviewed in Ref. 4).
barr-dependent signaling begins with a two-step recruitment of barr to the activated receptor. The first step involves binding to the phosphorylated receptor tail (5) (Fig. 1A) that converts barr into an open active conformation characterized by, among other features, a 20-degree rotation between its N-and C-terminal domains (6). In addition, the C-edge loops of active barr interact with the lipid bilayer and function as a membrane anchor (7)(8)(9). The second step involves the engagement of the finger-loop region of barr with the receptor core (10) (Fig. 1A). The bimodal binding of barrs to GPCRs engenders the coexistence of two unique conformations of GPCR-barr complexes: the "tail" conformation and the "core" conformation, each with a distinct set of functions (11).
Upon activation by GPCRs, barrs interact with a diverse set of partners including several mitogen-activated protein kinases and Src family tyrosine kinases (12,13) among many others, and the physiological implications of these interactions are currently being explored in many laboratories. The nonreceptor tyrosine kinase Src is a pharmacologically important protooncogene, involved in the regulation of the cell cycle, adhesion, proliferation, and migration (reviewed in Ref. 14). It has been previously shown that barr mediates the recruitment of Src to the activated b2-adrenergic receptor and thus functions as an adaptor protein (12). In fact, Src was the very first signaling protein for which barr was shown to serve such an adaptor role. Moreover, recent data suggest the possibility of direct allosteric effects of barr on Src (15,16). However, the molecular mechanisms and the conformational basis of these effects have not been elucidated. Herein, we use purified proteins to evaluate whether barr can directly allosterically activate Src and to explore the conformational basis of such regulation.

GPCR-barr1 complexes allosterically activate Src in vitro by promoting Src autophosphorylation
To study whether barr1 directly activates Src in vitro, we expressed and purified Src, barr1, and chimeric M2 muscarinic and b2-adrenergic receptors (M2V2 and b2V2, respectively). To ensure homogeneous phosphorylation, a synthetic phosphopeptide (V2Rpp) derived from the C-terminal tail of the vasopressin 2 (V2) receptor was ligated to the C termini of both receptors using sortase (17). V2Rpp has eight phosphorylated residues and confers tight binding to barrs (6). M2V2 and b2V2 were reconstituted in high-density lipoprotein particles (HDL, lipid nanodiscs) as this environment closely mimics a native membrane and enables testing functional outcomes in response to different ligands (Fig. 1B). We chose ;12-nm diameter MSP1D1E3 nanodiscs, previously found optimal for high-resolution structural studies of M2V2-barr1 complex (9). The extended lipid surface of MSP1D1E3 enables anchoring of C-domain of barr1 to the lipid bilayer, which stabilizes the complex (9).
Ligand binding and allosteric coupling of barr1 to HDLreconstituted receptors were verified by radioligand-binding assays ( Fig. 1, C-H). [ 3 H]NMS saturation ligand binding at HDL-M2V2 and 125 I-CYP saturation ligand binding at HDL-b2V2 show apparent affinities of 1.14 6 0.43 nM and 29.2 6 4.4 pM for the respective radioligand ( Fig. 1, C and F). Consistent with previous studies (17, 18), allosteric coupling of barr1 results in a significant increase in agonist, but not antagonist, affinity at respective GPCRs ( Fig. 1, D, E, G, and H).
We then measured the rate of phosphorylation of a synthetic peptide substrate (AEEEIYGEFEAKKKK) by Src using a continuous kinase colorimetric assay (19). In this assay the rate of peptide phosphorylation is coupled via the pyruvate kinase/lac-tate dehydrogenase enzymes to the oxidation of NADH measured through the decrease in absorbance at 340 nm. A progress curve of Src begins with a short lag phase with no or little changes in NADH absorbance ( Fig. 2A). This lag phase is associated with the slow activation step caused by the disruption of an autoinhibited conformation of Src and intermolecular autophosphorylation of catalytic Tyr-416 (19). Once Src is fully activated, it quickly phosphorylates the substrate, which causes a rapid decrease in absorbance ( Fig. 2A). The presence of lag phase in Src activity is clearly illustrated by comparing the progress curves of WT Src and a purified kinase domain of Src (SH1). SH1 represents a constitutively active form of the enzyme (20) and demonstrates no lag phase (Fig. 2B). Due to Figure 1. Pharmacological characterization of M2V2 and b2V2 reconstituted in lipid nanodiscs: ligand binding and allosteric coupling of barr1. A, schematic representation of GPCR-b-arrestin1 binding: barr1 recruitment begins with binding to the phosphorylated receptor tail (1) and is followed by the engagement with the receptor core (2). C-edge loops of active barr1 anchor it to the membrane (L, ligand; FL, finger loop). B, schematic representation of chimeric M2 muscarinic (M2V2) and b2-adrenergic (M2V2) receptors reconstituted in HDL particles (lipid nanodiscs). A synthetic phosphopeptide mimicking a phosphorylated C-terminal tail of V2 receptor was ligated to the receptors' C termini using sortase. The receptor is colored in red, the phosphorylated C-tail of V2 receptor is shown in yellow, MSP1D1E3 is shown in green. C, [ 3 H]NMS saturation ligand binding at HDL-M2V2. D and E, competition ligand binding assays using [ 3 H]NMS (1 nM) at HDL-M2V2 and a dose of agonist iperoxo (D) or antagonist atropine (E) in the absence (control) or presence of 1 mM barr1. F, 125 I-CYP saturation ligand binding at HDL-b2V2. G and H, competition ligand binding assays using 125 I-CYP (60 pM) at HDL-b2V2 and a dose of agonist isoproterenol (G) or antagonist ICI-118551 (H) in the absence (control) or presence of 1 mM barr1. C-H, points in respective curves represent mean 6 S.D. from three independent experiments. Asterisks (*) in B and E indicate significant difference in IC50 values between control versus barr1 competition curves (p , 0.05, one-way ANOVA with Bonferroni's post test). the presence of the slow activation step, the Src kinetics before and during the activation process is divergent from Michaelis-Menten kinetics. Because the lag phase is indicative of the degree of autoinhibition of the enzyme, the initial velocity of the reaction (V0) represents the most accurate parameter to measure Src activity (21,22).
We then performed the assay in the presence of barr1 and GPCR-barr1 complexes. Whereas free barr1 did not affect Src activity (Fig. 2C), barr1 in complex with either phosphorylated M2V2 or b2V2 activated by iperoxo or BI-167107, respectively, caused a significant increase in the rate of peptide phosphorylation by Src (Fig. 2D, Table S1). Addition of GPCR-barr1 complexes reduces this lag phase and leads to a more rapid decrease in absorbance (Fig. 2E). M2V2 and b2V2 alone did not lead to Src activation (Fig. 2D) indicating that this process is mediated by barr1. The initial velocity of peptide phosphorylation by SH1 is 5.6-fold higher than that of WT Src (3.16 6 0.45 versus 0.56 6 0.16 nmol min 21 , respectively), whereas in the presence of M2V2-barr1 and b2V2-barr1 we observed a 3.4-and a 2.6fold increase in the initial rate, respectively (Table S1). Taken together, our data demonstrate that receptor stimulated barr1 directly activates Src.
Kinase assay data suggest that GPCR-barr1 complexes reduce the lag phase in Src activation (Fig. 2E). We thus hypothesized that interactions of GPCR-barr1 complexes with Src promotes Src autophosphorylation. To test this hypothesis, we analyzed the time course of Tyr-416 autophosphorylation by Western blotting (Fig. 2F). In the presence of the agonist-activated M2V2-barr1 complex, a phosphorylation of Tyr-416 was observed within 15-30 s after adding ATP, whereas for the Src alone reaction autophosphorylation was detected only at later time points. These data suggest that GPCR-stimulated barr1 activates Src by reducing the lag phase in Src autophosphorylation.
Phosphorylated GPCR tail interaction with b-arrestin 1 is sufficient to confer the activation of Src We next sought to elucidate which conformations of GPCR-barr complexes contribute to Src activation. Previously shown that b2V2-barr1 complex represents a dynamic mixture of tail (partially engaged) and core (fully engaged) conformations (10). Recent structural and biophysical data demonstrate that in addition to tail and core interactions with the receptor, the C-edge of active barr also engages the lipid nanodisc and functions as a membrane anchor (Figs. 1A and 3A) (8,9).
First, we tested whether an interaction of barr1 with the receptor core is required for Src activation. Importantly, because C termini of both M2V2 and b2V2 are phosphorylated, barr1 will bind to the receptor tail regardless of the presence of an agonist. Interestingly, we observed full activation of Src even without agonist stimulation of b2V2 (Fig. S1). This result suggests that allosteric activation of Src is primarily driven by interaction of barr1 with the phosphorylated receptor tail. However, existing evidence of high (basal) constitutive activity of GPCR reconstituted in lipid nanodiscs (23) required additional experiments to verify this hypothesis.
To prevent barr coupling to the receptor core, we expressed and purified a barr1 mutant with a deleted finger loop (barr1DFL). barr1DFL is unable to interact with the receptor core (11), thus all GPCR-barr1DFL complexes will remain in the tail conformation (Fig. 3A). Interestingly, both M2V2-b arr1DFL and b2V2-barr1DFL complexes fully retain the ability to activate Src suggesting that core interaction is dispensable for allosteric activation of the kinase (Figs. 3B).
We next tested whether the engagement of barr1 with the lipid bilayer plays a role in barr1 ability to activate Src. Molecular dynamic simulations have shown that interactions between C-edge loops of barr1 and the membrane stabilize the active conformation of barr1 (9) that might be crucial for allosteric activation of Src. We thus formed GPCR-barr complexes with a barr1 mutant deficient in lipid interaction (barr1DDD) (9) (Fig. 3A) and measured Src activity. M2V2-barr1DDD and b2V2-barr1DDD complexes activate Src as efficiently as complexes formed with WT barr1 and barr1DFL indicating that anchoring of barr1 to the membrane is inessential for allosteric activation of the enzyme (Figs. 3B).
Our findings thus suggest that the interaction of barr1 with a phosphorylated receptor tail is sufficient to allosterically activate Src. We then tested if active barr1 with only a phosphorylated C-terminal tail of the V2 receptor (barr1-V2Rpp) induces activation of Src similarly to GPCR-barr1 complexes. Indeed, we achieved the same level of Src activation in the presence of barr1-V2Rpp and GPCR-barr1 (Fig. 3C, Table S1).
We next sought to ascertain whether phosphorylated GPCR tail interaction with barr 1 is absolutely required for allosteric activation of Src. We thus tested Src activity in the presence of barr1 and nonphosphorylated C-terminal tail of the V2 receptor (V2Rnp) (Fig. 3D). As expected, no significant increase in Src activity was observed (Fig. 3D). Furthermore, the presence of barr1 and nonphosphorylated M2V2 receptor (M2V2np) (Fig. 3A) also did not activate Src (Fig. 3E), probably due to the reduced binding of barr1 to the nonphosphorylated receptor. To test the binding of barr1 to both M2V2 and M2V2np, we performed a M1-FLAG pulldown assay (Fig. S2). Even though M2V2np binds a small amount of barr1, it is not sufficient to trigger the activation of Src, as the large portion of barr1 predominantly remains in an inactive conformation. Taken together, these data indicate that the phosphorylated GPCR tail interaction with barr1 is necessary and sufficient to drive the allosteric activation of Src.

b-Arrestin 1 mediates allosteric activation of Src by interacting with SH3 domain
Our findings demonstrate that active barr1 mediates allosteric activation of Src by promoting autophosphorylation of the enzyme. We next wanted to delineate the molecular mechanism of the activation. First, we tested the binding of active and inactive conformations of barr1 to different regions of Src in vitro using a GSH S-transferase (GST)-pulldown assay (Fig.  4A). barr1 weakly interacts with both SH3 and SH1 domains of Src, which is consistent with previously published data on barr1-Src interactions in cells and in vitro (12,13). Interest-ingly, the SH3 domain of Src binds tighter to active barr1 (barr1-V2Rpp), whereas the SH1 domain interacts more strongly with the inactive form of barr1 (Fig. 4A). To understand which of these interactions contribute to the allosteric activation of Src, we performed a competitive colorimetric kinase assay. In this assay we tested the ability of barr1-V2Rpp to activate Src in the presence of an excess of either purified SH3 domain or a kinase dead mutant of the SH1 domain D386N (SH1 KD). The presence of SH1 KD does not impact the ability of barr1-V2Rpp to activate Src (Fig. 4B). In contrast, an excess of SH3 domain completely blocks the barr1-mediated activation of Src (Fig. 4B) suggesting that the SH3 domain binds to barr1 and thus interferes with the mechanism of activation through direct competition with Src. Thus, the activation of Src by b-arrestins requires its interaction with the SH3 domain of the enzyme.

Discussion
Arrestins play a plethora of roles in GPCR signaling. In addition to receptor desensitization, internalization, and intracellular trafficking, arrestins also function as independent signal transducers (reviewed in Ref. 4). However, the precise mechanisms of signal transduction via arrestins remain elusive. Here, we demonstrate that GPCR-activated barr1 exerts direct allosteric activation of the proto-oncogene kinase Src in vitro.
barr-mediated effects on Src and extracellular signal regulated kinases (ERK1/2) have been explored in two recent studies (15,16). In particular, it was shown that Src activation downstream of dopamine D1 receptor in HEK 293 cells solely depends on barr2, whereas ERK1/2 activation involves both Gprotein and barr2 (16). Yang et al. (15) showed that barr1 and GRK6-phosphorylated b2-adrenergic receptor-barr1 complexes promoted Src activity in vitro. These studies, however, do not address the structural and conformational basis for the barr-mediated allosteric activation of these enzymes. Here, we tested the contribution of five different conformational arrangements of barr1 and GPCR-barr1 complexes to the activation of Src (Figs. 3A and 4C). With respect to receptor engagement these are: 1) fully engaged (tail-and-core of receptor are bound and barr1 is membrane-anchored (M2V2-b arr1); 2) fully receptor engaged barr1-deficient in membrane interaction (M2V2-barr1DDD); 3) partially receptor engaged (tail-bound) membrane anchored (M2V2-barr1DFL); 4) only receptor tail-bound (barr1-V2Rpp); and 5) inactive barr1: barr1, barr1-V2Rnp, and M2V2np-barr1. Apart from inactive barr1, all conformations, including barr1-V2Rpp, activate Src to a similar level, suggesting that the barr1-mediated allosteric activation of Src depends only on the receptor tail-bound conformation of barr1 and does not require its interaction with either the receptor core, or the membrane. Importantly, these findings suggest that initiation of signaling via barrs may precede the termination of G-protein-mediated signaling, which requires barr interaction with the receptor core. These results are also consistent with previously published data showing that the tail conformation of the b2V2-barr1 complex retains the ability to mediate receptor internalization, a process known to be Src-dependent (11,13,24). Moreover, the interactions of barr1 with the core of b2V2 and V2 receptors are also dispensable for ERK2 binding and activation (24,25). Interestingly, in contrast to the previously published work (15), we did not observe a statistically significant activation of Src in the presence of inactive barr1. This result can be explained by a different barr1:Src ratio (5:1 in our study versus 6:1 in Ref. 15) and slightly different conditions of the experiment. In addition, in the previous study (15) Src activity was determined by fitting the initial rate to the Michaelis-Menten equation to obtain K m and k cat , whereas we used the initial velocity to monitor the lag phase, similarly to previously published papers on activation of Src-family kinases (21,22).
Intriguingly, we observed identical allosteric effects on Src by barr1 stimulated by two different receptors, M2V2 and b2V2, sharing the same phosphorylated V2 receptor tail and thus mimicking class B GPCRs. As shown previously (12,15), WT b2-adrenergic receptor, a class A GPCR, can also activate Src through barr1 in an agonist and phosphorylation-dependent manner. These findings further buttress the conclusion that it is the phosphorylated receptor C terminus that orchestrates barr1-mediated Src activation. Moreover, the phosphorylation pattern in the receptor tail is known to affect barr-mediated recruitment of Src (15). In our study, nonphosphorylated M2V2np did not activate Src through barr1 (Fig. 3E). These results support the barcode hypothesis, the notion that the receptor phosphorylation pattern induces a range of specific conformations of barr1 that direct barr1-mediated signaling (26,27). It is therefore tempting to speculate that a receptor tail with a different phosphorylation pattern might elicit a different outcome on Src recruitment and activation even for the same receptor.
We found that GPCR-activated barr1 reduces the lag phase in Src activation by promoting its trans-phosphorylation (Fig.  2, E and F). Furthermore, we showed that allosteric activation of Src requires the interaction of barr1 with the SH3 domain of the enzyme (Fig. 4B). The most plausible mechanism of the activation is therefore a disruption by barr1-SH3 interactions of Src intramolecular contacts that normally constrain the activity of the kinase (Fig. 4C). SH3 domains recognize lefthanded type II polyproline sequences PXXP (where X is any amino acid) (28). In an autoinhibited conformation of Src, the SH3 domain interacts with the type II helix of the linker between the SH2 and kinase domains (29) (Fig. 4C). This linker does not have the classical PXXP signature, therefore, a partner with an optimal polyproline sequence will easily displace the SH2 linker unlocking the autoinhibited conformation (30). This event represents a common mechanism of Src-family kinase activation and has been documented by several structural and biochemical studies (31,32).
barr1 has three PXXP sequences ( 88 PPAP 91 , 121 PNLP 124 , and 175 PERP 178 ), and previous studies have demonstrated that mutations of Pro-91 and Pro-121 drastically reduced barr1-Src interactions and kinase activation (12,15). It is currently unknown whether one particular site is involved in the interaction or they are inter-changeable. In a computationally generated docking model of GPCR-b-arrestin-Src complex SH3 domain is positioned in the pocket between both Pro-91 and Pro-121 polyproline motifs of barr1 (33). Intriguingly, none of the proline sites in barr1 represent a canonical sequence for the SH3 domain of Src that requires consensus sequences RXXPXXP or PXXPXR, in which positively charged arginine is important for high affinity binding (28,34,35). Perhaps, due to the dynamic nature of the GPCR-barr1-Src signaling module low affinity interactions between SH3 and barr1 are preferred. Further structural studies will shed light on the detailed molecular mechanism of Src recruitment and allosteric activation by barrs.
In conclusion, we demonstrate that binding of barr1 to the phosphorylated receptor tail instigates a distinct barr1-mediated signaling pathway via allosteric activation of Src. A combination of biochemical approaches used in this study can easily be applied to explore allosteric activation of other downstream targets by barr in vitro. Taken together, our findings represent an important step forward toward understanding more broadly the mechanisms of signal transduction via barrs.

Radioligand-binding assays
Saturation and competition radioligand-binding assays were performed at HDL-reconstituted M2V2 and b2V2 (17, 18). All binding assays were carried out until equilibrium at room temperature in a buffer composed of 20 mM HEPES, pH 7.4, 100 mM NaCl, 0.2 mg/ml of BSA and 0.18 mg/ml of ascorbic acid. To prepare barr1-V2Rpp complex, barr1 was incubated with 2-fold molar excess of V2Rpp and Fab30 for 1 h at room temperature and then the complex was purified by SEC in 20 mM HEPES, pH 7.5, 150 mM NaCl, and 1 mM tris(2-carboxyethyl)phosphine hydrochloride (TCEP). To prepare M2V2-b arr1 and b2V2-barr1 complexes, M2V2 (20 mM) and b2V2 (20 mM) receptors were preincubated with 5-fold molar excess of iperoxo or BI-167107, respectively, for 20 min on ice. After incubation, 20 mM barr1 and 20 mM Fab30 were added and the mixture was incubated for 1 h on ice.

Continuous colorimetric kinase assay
Continuous colorimetric kinase assay was performed as previously described (19). All reactions (200 ml) contained Src, 100 mM HEPES, pH 7.5, 150 mM NaCl, 5 mM MgCl 2, 1 mM phosphoenolpyruvate, 0.3 mM NADH, 0.25 mM optimal Src peptide (AEEEIYGEFEAKKKK), 2 mM sodium orthovanadate, 1 mM TCEP, 0.005% Triton X-100, 4 units of pyruvate kinase, and 6 units of lactic dehydrogenase. The concentration of Src in all experiments was 25 nM and the concentration of barr1 or GPCR-barr1 complex was 125 nM unless stated otherwise. In reactions containing barr1, barr1-V2Rpp, or GPCR-barr1 complexes, the reaction mixture was incubated for 1 h on ice. Reactions were started by the addition of ATP to a final concentration of 0.1 mM, and the decrease in NADH absorbance was monitored over 40 min at 25°C using a CLARIOstar microplate reader (BMG Labtech). The initial velocity of the reaction (V 0 ) was determined using a nonlinear regression curve fit in GraphPad Prism software. The change in absorbance was then converted to the product concentration using the Beer-Lambert law and to the amount of product formed in the reaction volume per minute. Statistical comparisons were determined by one-way ANOVA followed by a Dunnett's multiple comparison test.

Src autophosphorylation assay
Src autophosphorylation reactions were performed as previously described (21). Src was diluted to 12.5 nM in a buffer containing 100 mM HEPES, pH 7.5, 150 mM NaCl, 5 mM MgCl 2 , 20 mg/ml of BSA, 2 mM sodium orthovanadate, and 1 mM TCEP. In reactions containing 125 nM M2V2-barr1 complex, the reaction mixture was incubated for 1 h on ice before adding ATP. The reactions were initiated by addition of ATP to a final concentration of 0.1 mM and carried out on ice. At various time points, 50-ml aliquotes of the reaction were quenched with 15 ml of 43 SDS loading buffer and subjected to SDS-PAGE and Western blotting. The active form of Src was detected with anti-Src (phospho-Y418) antibody (Abcam, ab4816, 1:5000 dilution). The total Src was detected by anti-Src antibody (EMD Millipore 05-184, 1:2000 dilution) on a separate SDS-PAGE gel. The optical density of the bands was quantified in ImageJ and statistical differences were determined by Mann-Whitney test in GraphPad Prism software.

M1-FLAG pulldown assay
To test the binding of barr1 to M2V2 and M2V2np, 10 mM of the receptor was preincubated with a 5-fold molar excess of iperoxo and the positive allosteric modulator LY211,960 for 30 min on ice and then with a 2-fold molar excess of barr1 and Fab30 for 2 h on ice. 50 ml of M1-FLAG resin equilibrated in 20 mM HEPES, pH 7.5, 100 mM NaCl buffer was added thereafter, and the mixture was rotated for 1 h at 4°C and then additional 30 min after adding 2 mM CaCl 2 . After incubation the M1-FLAG beads were collected by centrifugation and washed with 20 mM HEPES, pH 7.5, 100 mM NaCl, 2 mM CaCl 2 buffer three times. The proteins were eluted with 0.2 mg/ml of FLAG-peptide and 5 mM EDTA in 20 mM HEPES, pH 7.5, 100 mM NaCl buffer. Receptor glycosylation was removed by incubation with a 1:10 protein ratio of peptide:N-glycosidase F to the receptor for 60 min at room temperature in the presence of 1% Nonidet P-40, 0.5% SDS, 40 mM DTT. The samples were subjected to SDS-PAGE, visualized by Instant Blue Coomassie stain (Expedeon), and quantified by ImageJ.

GST-pulldown assay
For detection of barr1 binding to GST-SH3, 20 mM barr1 was preincubated with 3-fold molar excess of V2Rpp and Fab30 for 1 h at room temperature, then 10 mM GST-SH3 was added and incubation continued for another 1 h. For detection of SH1 binding to GST-barr1, 10 mM GST-barr1 was preincubated with 3-fold molar excess of V2Rpp for 1 h at room temperature, then 20 mM SH1 was added and incubation continued for another 1 h. 50 ml of GST beads (GoldBio) equilibrated in 20 mM HEPES, pH 8.0, 150 mM NaCl buffer were added thereafter, and the mixture incubated for 1 h at room temperature with rotation. After incubation the GST beads were collected by centrifugation and washed with 20 mM HEPES, pH 8.0, 150 mM NaCl buffer three times. The proteins were eluted from GST beads with 40 ml of 20 mM reduced GSH in 20 mM HEPES, pH 8.0, 150 mM NaCl buffer, then mixed with 43 SDS loading buffer, subjected to SDS-PAGE, and Western blotting and detected by EMD Millipore 05-184 antibody (SH3, 1:5000 dilution), Abcam ab1187 anti-His 6 tag antibody (SH1, 1:5000 dilution), and Cell Signaling Technology 30036S antibody (barr1, 1:2000 dilution).

Data availability
All data presented are available upon request from Robert J. Lefkowitz (lefko001@receptor-biol.duke.edu).
Acknowledgments-We thank Prof. John Kuriyan for helpful discussions throughout this work. We are grateful to Darrell Capel, Xingdong Zhang, and Xinrong Jiang for technical assistance and Yangyang Li, Quivetta Lennon, and Victoria Brennand for administrative assistance. Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.