Characterization of the Site-Specific Incorporation of p-Benzoyl-l-Phenylalanine in c-Abl SH2.

Because protein tyrosine phosphorylation status is highly dynamic in living cells, we used the SH2 domain, which has a micromolar range of binding affinity toward phosphotyrosine peptides, to capture phosphotyrosine proteins in living cells combined with an in vivo phototrapping method to trap weak and transient phosphoproteins. The photo–cross-linking amino acid p-benzoyl-l-phenylalanine (pBpa) can genetically incorporate into a desired site of interest within protein domains in situ by means of orthogonal tRNA/aminoacyl-tRNA synthetase pairs (Fig. 1A). We have previously used this approach to investigate actin cytoskeletal signaling mediated by Rho family GTPase Activating Protein (GAP) SH3 domain protein interaction (PI) in living cells by trapping them as they occur (14). To test the utility of the approach to identify posttranslational modifications, we chose the SH2 domain from c-Abl, based on previous work suggesting it recognizes large numbers of phosphotyrosine proteins (15). Three amino acid positions near the phosphotyrosyl peptide binding pocket were selected based on the crystal structure of the SH2 domain of c-Abl (16) (Fig. 1B). Each c-Abl SH2 cDNA variant containing an amber codon [SH2amb1 (R152), SH2amb2 (R175), and SH2amb3 (N177)] was cloned into the pcpBpaV4 vector (14) and transfected into FreeStyle 293 cells. This pcpBpaV4 vector allows for the simultaneous expression of the modified SH2, together with the pBpa-specific variant of Escherichia coli tyrosyl-tRNA^Tyr (Eco-pBpaRS) and the Bacillus stearothermophilus suppressor tRNA^Tyr. Also, pcpBpaV4 adds affinity tags to the N terminus of the cloned insert, such that the SH2 domain can be efficiently purified by tandem affinity purification (TAP). This vector also allows for the detection of expressed SH2 domains by a C-terminal V5 epitope. pBpa was added to the growth media for 3 d posttransfection, after which the cells were treated with pervanadate (PV), followed by exposure to 365-nm light. Purified samples from these lysates were immunoblotted to determine SH2 expression by amber codon suppression (Fig. 1C). SH2amb1 expression was consistently low, yet SH2amb2 showed intense signals with high-molecular-weight (HMW) bands compared with SH2amb3, indicating covalent binding to cellular proteins. To test whether these HMW bands were photo–cross-linking–dependent, SH2amb2 was purified with and without UV light exposure (Fig. 1D). The results showed HMW bands were UV light-dependent, indicating most of the HMW bands were endogenous proteins phototrapped by SH2amb2. The HMW band intensity increased when cells were treated with PV, indicating the phototrapping was enhanced by increasing the levels of cellular phosphotyrosine. Moreover, under the same conditions, WT SH2 domain did not cross-link to cellular proteins, confirming that pBpa was required for cross linking in situ (Fig. S1). We also assessed the capacity of these three modified SH2 domains to bind a known c-Abl SH2 ligand using the peptide overlay method (Fig. 1E). Only SH2amb2 bound to the EGF receptor (EGFR)-derived peptide (17) in a phosphodependent manner. These results suggested that the unnatural amino acid incorporated into the SH2amb2 variant did not alter the recognition specificity toward a known phosphotyrosine peptide; however, further work would be needed to demonstrate if it retained the exact binding selectivity of the WT SH2 domain. The main goal of these experiments was to exploit the SH2 domain for its phosphobinding properties rather than to identify physiological partners of c-Abl. To this end, the overlay results confirmed that the binding was phosphodependent.

Targeting Modified SH2 Domains to Subcellular Compartments.

These results suggested the expression of soluble SH2amb2 (hereafter referred to as Solu-SH2) could be used to phototrap phosphotyrosine-containing proteins within cells. Next, we targeted Solu-SH2 to specific subcellular sites within the cell to bias the trapping of phosphotyrosine proteins at these locations. Three types of subcellular localization signals were fused to the N terminus of the Solu-SH2 domain to target it to F-actin, mitochondria, or cellular membranes (Fig. 2A). To confirm whether the localization signals effectively targeted the pBpa-containing SH2 to subcellular compartments, we coexpressed each SH2 domain variant with control vectors (Fig. 2 B–J). Each targeted SH2 protein colocalized with control fluorescent proteins, demonstrating the pBpa SH2 variants were targeted to the desired subcellular compartment effectively, although some soluble signal was also detected. Also, we verified the presence of tyrosine-phosphorylated proteins in each subcellular structure (Fig. 2 K–S). At the sites where actin-targeted EGFP or membrane-targeted enhanced yellow fluorescent protein (EYFP) was expressed, phosphotyrosine staining was observed (Fig. 2 K–P). Additionally, the mitochondria-targeted EYFP demonstrated partial overlap with phosphotyrosine staining (Fig. 2 Q–S). Each SH2 variant was expressed in FreeStyle 293 cells and exposed to UV light, followed by TAP, and immunoblot analysis was performed to determine the photo–cross-linked proteins. Each SH2 variant displayed a unique banding pattern (Fig. 2T), suggesting each variant captured and enriched a unique subset of phosphoproteins.

Targeted SH2 Domains Captured Unique Proteins from Subcellular Regions.

Purified fractions from each of the three distinct subcellular compartments [actin-targeted SH2 (Act-SH2), mitochondria-targeted SH2 (Mito-SH2), and membrane-targeted Sh2 (Mem-SH2)] and without a targeting sequence (Solu-SH2) were subjected to solution trypsin digest, and tandem MS was performed to identify captured proteins. To compare identified proteins between each SH2pBpa variant fraction, hierarchical clustering was performed (Fig. 3A). Proteins uniquely enriched in each fraction (termed PI clusters) were identified, thus removing common proteins that copurified with the baits independent of their location. These could result from a low level of nontargeted SH2 due to overexpression of the tagged domain or from nonspecific interactions. These PI clusters included 97 proteins from Solu-SH2, 46 proteins from Act-SH2, 24 proteins from Mito-SH2, and 22 proteins from Mem-SH2 (Dataset S1). The proteins within each PI cluster were heavily enriched for the subcellular region targeted by the SH2pBpa variants. For example, Act-SH2 phototrapping identified proteins known to be regulators of or localized to actin, including RhoA, myosin VI, gelsolin, LIMA1, and IQGAP1. By comparative analysis with the PhosphoSite and Phospho.ELM databases, potentially unique phosphoproteins were identified in each enrichment fraction (Fig. S2), indicating the targeting and phototrapping approach could capture unique potential phosphotyrosine proteins from subcellular compartments. In addition, BCAR1, which is known to interact with c-Abl in an SH2-dependent manner, was captured in the soluble, actin, and mitochondria fractions (18). Thus, consistent with our EGFR peptide overlay, these data indicate the pBpa-containing SH2 domain retains its natural binding selectivity. Importantly, in each subcellular-targeted SH2 PI cluster, the portion of the target subcellular localization predicted by gene ontology for the identified proteins was increased compared with the Solu-SH2 PI cluster (Fig. 3B). These results further demonstrate that targeting SH2 domain to each subcellular compartment biased the phototrapping of phosphoproteins that naturally occur at that subcellular region in situ.

Identification of Potential SH2 Binding Phosphorylation Sites by Peptide Array Overlay.

It is likely that in addition to phosphorylated proteins, nonphosphorylated proteins associated with the SH2 binding proteins would be present in the purified samples. To identify the potential tyrosine-phosphorylated proteins and the sites on these proteins that could serve as the SH2 docking site, we performed peptide array analysis. Sequences surrounding tyrosine sites from a sample of the most abundant proteins (based on peptide spectral counts) that were found in each SH2 PI cluster were arrayed onto membranes as phosphotyrosine and nonphosphotyrosine peptides. These arrays were then incubated with GST c-Abl SH2 domain, and bound GST-SH2 was detected by immunoblotting. A total of 182 peptides from 25 proteins were analyzed (Dataset S2). The resulting intensity for binding at each peptide, nonphosphopeptide vs. phosphopeptide, was measured and normalized to a positive control peptide derived from EGFR. For example H-Ras harbors nine potential phosphotyrosine sites, one of which (Y157) was reported in a phosphoproteomic study to be phosphorylated (19). Overlay analysis confirmed the Y157 could function as an SH2 binding site in addition to other potential phosphotyrosine sites, such as Y32 (Fig. 4 A and B). Representative results from parallel antiphosphotryosine blotting and GST-SH2 overlays verified that negative binding results were not due to variability in the phosphotyrosine peptide synthesis (Fig. S3), consistent with other studies demonstrating the highly reproducible nature of phosphotyrosine peptide array synthesis and SH2 domain overlay (20). Of a total 182 peptides, 28 were previously identified as phosphorylated (Fig. 4B). The SH2 overlay screening identified 10 of these peptides. Additionally, 49 previously undescribed potential SH2-binding phosphorylation sites were identified. Binding sequences identified from the peptide array were aligned to deduce the preferred c-Abl SH2 binding sequence flanking the phosphotyrosine site (Fig. 4C). Previously the c-Abl SH2 domain was reported to show selectivity for a common pY[−][−]ϕ motif, where [−] and ϕ designate negatively charged and hydrophobic residues, respectively (21). The binding data obtained here are consistent with this specificity, although the c-Abl SH2 domain has shown a broader binding spectrum, which, in turn, makes it useful for trapping phosphotyrosine peptides in vivo.

RhoA Can Be Tyrosine-Phosphorylated on Y34 and Y66.

To characterize further one of the phosphoproteins identified by the spatial phototrapping strategy, RhoA, which was identified in the Act-SH2 PI cluster, we carried out a peptide array overlay analysis with the c-Abl SH2 domain and found that RhoA had two potential SH2 binding phosphorylation sites, Y34 and Y66 (Fig. 5A). Recently one site, Y34, was identified by a large-scale phosphoproteomics study, further validating endogenous RhoA as a tyrosine kinase substrate (19). Rho-family GTPases share a common domain structure, consisting of a GDP/GTP-binding motif (G box) and switch I and II regions, which mediate binding to regulators [GAP and guanine nucleotide exchange factor (GEF)] and downstream effectors (22, 23). Interestingly, Y34 and Y66 were located in these switch regions, suggesting they could be important regulatory sites of RhoA (24) (Fig. 5B). To test whether RhoA could be tyrosine-phosphorylated in cells, RhoA was expressed in HEK 293T cells, followed by immunoprecipitation, and was immunoblotted with an antiphosphotyrosine antibody (Fig. 5C). RhoA was tyrosine-phosphorylated, and the phosphorylation was stimulated by PV treatment. To screen for candidate tyrosine kinases involved in RhoA phosphorylation, we coexpressed RhoA with several active tyrosine kinases (Fig. 5D and Fig. S4). RhoA was phosphorylated by Bcr-Abl and Src. We also confirmed earlier work suggesting endogenous RhoA could be phosphorylated (19). Interestingly, endogenous RhoA phosphorylation was detected only when coexpressed with Bcr-Abl in the presence of PV, indicating the phosphorylation of RhoA may be tightly regulated by phosphatase activity (Fig. 5E). To map the responsible phosphorylation site in situ, mutagenesis analyses were performed (Fig. 5F). Y34F and Y66F RhoA mutants reduced tyrosine phosphorylation levels to 50% and 20%, respectively, compared with controls when they were coexpressed with Src (Fig. 5G). When these RhoA mutants were coexpressed with Bcr-Abl, phosphorylation levels of Y34F and Y66F RhoA mutants dramatically decreased to 32% and 17%, respectively. As expected, when Y34 and Y66 were both mutated to phenylalanine, phosphorylation was completely abolished. Together, these observations indicate that Y34 and Y66 are the two predominant phosphorylation sites, and that the Src kinase and Bcr-Abl are the two candidate kinases that may phosphorylate these sites.

Tyrosine 34 and 66 of RhoA Appear to Be Critical for Effector Binding.

Based on their location within the switch regions, one potential functional role of phosphorylation of RhoA at Y34 and Y66 could be to regulate effector binding in vivo. To investigate this possibility, the activity of RhoA vs. RhoA mutants was tested using a two-photon excitation laser scanning microscopy with fluorescence lifetime imaging (2pFLIM) assay for RhoA effector [Rho binding domain (RBD)] binding (Fig. 6A). Fluorescence lifetime decay curves of mEGFP-RhoA were measured to calculate the binding fraction of RhoA (25, 26). Constitutively active RhoA (RhoAQ63L) had an approximately twofold higher fraction of RBD binding compared with WT RhoA (Fig. 6B). To test the potential role of phosphorylation, tyrosine 34 and 66 were mutated to glutamic acid. Although acidic amino acids cannot perfectly mimic phosphotyrosine, they are the only substitutions that have successfully been used to approximate tyrosine phosphorylation in regulating protein activity (27, 28). In contrast to active RhoA, RhoAQ63L(Y34,66E) had a dramatic decrease in RBD binding. This binding fraction was even lower than that of WT RhoA, suggesting phosphorylation at these sites could have a negative effect on RhoA activity (Fig. 6B). To test this notion further, RhoAQ63L was coexpressed with constitutively active Src (SrcCA) or dominant negative Src (SrcDN) (Fig. S5A). SrcCA reduced the binding of RhoAQ63L (17.4%) compared with SrcDN (Fig. 6C). Because the glutamine 63 to leucine mutation in the switch 2 region of GTPases both increases basal GDP/GTP exchange and inhibits intrinsic GTPase activity, the RhoAQ63L is relatively insensitive to the activity of cellular GAPs and GEFs (29–31). Thus, the effect of Src was a likely a direct effect on the sensor. This effect was likely via Src phosphorylation of RhoA vs. the RBD domain of the sensor because control experiments demonstrated that although Src could phosphorylate RhoA, it could not phosphorylate the RBD domain (Fig. S5B). Together, these data suggest Y34 and Y66 are critical for the ability of RhoA to interact with downstream effectors on activation and that phosphorylation at these sites would likely inhibit activation-dependent effector binding. This possibility was tested using WT RhoA vs. RhoA(Y34,66E) in a RhoA activation assay in response to EGF treatment (Fig. 6D). RhoA(Y34,66E), WT RhoA, and dominant negative RhoA (RhoAT19N) were expressed in the cells and stimulated with EGF to induce activation (32). As expected, WT RhoA was potently activated by EGF treatment. In contrast, RhoA(Y34,66E) was not responsive and behaved similar to RhoAT19N (Fig. 6 D and E).

Role of Y34 and Y66 of RhoA in Effector Binding and RhoA-Induced Actin Stress Fiber Formation.

These results suggested phosphorylation of RhoA at Y34 and Y66 may inhibit binding to cellular effectors. To test this possibility, GFP-tagged RhoAQ63L or its mutant, RhoAQ63L(Y34,66E), expressed in cells was immunoprecipitated to compare interacting partners (Fig. 7A). MS-based comparative analysis of the immunoprecipitates showed that RhoA effectors involved in the regulation of the actin cytoskeleton, such as DIAPH1, DAAM1, ROCK1/2, and PKN, were detected only in the RhoAQ63L immunoprecipitations (Fig. 7B). Also, Rhotekin, which was used as an effector binding for 2pFLIM assay, was associated with RhoAQ63L but not with RhoAQ63L(34,66E). Importantly, RhoGAP1, a GTPase activating protein for RhoA and Cdc42 (33, 34), was associated with both RhoA mutants, suggesting that altering tyrosine 34 and 66 does not disrupt this binding. These observations indicate that Y34 and Y66 are critical for effector binding but not for GAP binding. Importantly, it also shows the mutations did not globally disrupt RhoA folding or function but did selectively impair its ability to bind effectors. To determine the role of each tyrosine with regard to effector binding, an RBD pull-down assay was performed (Fig. 7 C and D). Both Y34 and Y66 were critical for Rhotekin binding, suggesting phosphorylation of either site could inhibit RhoA effector functions.

Based on these results, the functional consequences to RhoA-induced cellular actin reorganization were assessed (Fig. 7 E and F). For this assay, RhoA (wild), RhoAQ63L, and mutants of RhoAQ63L containing either phenylalanine or glutamic acid at positions Y34 and Y66 were overexpressed in serum-starved HeLa cells and the percentage of transfected cells with elevated actin stress fiber formation was quantified. As expected, RhoAQ63L strongly induced actin stress fibers in HeLa cells, almost sixfold that of WT RhoA, which weakly stimulated stress fiber formation. In contrast, the RhoAQ63L(Y34,66E) mutant completely lost the ability to induce F-actin formation, whereas RhoAQ63L(Y34,66F) retained significant activity. These data indicate the difference in ability to stimulate cellular stress fiber formation is due to the negative charge of glutamic acid compared with the structurally similar phenylalanine. Interestingly, these results almost perfectly mirror the effector binding results of Fig. 6B, which also showed the Y34,66E form of RhoA is less active than WT RhoA.

Plasmids.

A mammalian expression vector that coexpresses Eco-pBpaRS and B. stearothermophilus suppressor tRNA^Tyr (a gift from Shigeyuki Yokoyama, University of Tokyo, Tokyo, Japan) was modified for TAP (14). A strep-tag, followed by a calmodulin binding peptide (CBP), a V5 epitope, two tobacco etch virus protease cleavage sites, an HRV3C cleavage site, and a tandem protein A sequence, was inserted after a mammalian expression promoter. Unique RsrII and SbfI restriction sites were placed between the CBP and V5 epitope for subcloning SH2 domain coding sequences. The SH2 domain (aa 120–220, Uniprot P00519) derived from c-Abl was inserted, and the targeted sequences listed below were placed on the N terminus of SH2. Lifeact, MGVADLIKKFESISKEE (49); membrane (Mem), MLCCMRRTKQVEKNDEDQKI (50) (derived from neuromodulin); and mitochondria (Mito), MSVLTPLLLRGLTGSARRLPVPRAKIHSL (51) (derived from cytochrome c oxidase subunit 8A) were used. The Lifeact was a gift from Michel Bagnat (Duke University). Other plasmids used in this study are detailed in SI Materials and Methods.

Photo-Cross Linking.

Photo-cross linking was performed as described (14). Briefly, on the day of transfection, a 1-L culture of FreeStyle 293 cells was grown in FreeStyle 293 medium (GIBCO) to a density of 1 × 10⁶ cells/mL. Transfection was done following a modified polyethylenimine (PEI) protocol (52). One milligram of the vector encoding TAP-tagged SH2 domain plus Eco-pBpaRS and suppressor tRNA^Tyr was mixed with 2 mg of PEI in 50 mL of FreeStyle 293 medium and added to the cell culture after 10 min of incubation. pBpa (270 mg) was dissolved in 1.1 mL of 1 M NaOH and added to the cell culture following addition of 7.5 mL of 1 M Hepes buffer to the culture. After 3 d of incubation at 37 °C, the cells were treated with 0.5 mM PV for 5 min, followed by exposure to UV light (365 nm) in a custom-built chamber. After 30 min of photo-cross linking in the chamber at 4 °C, cells were collected by centrifugation and subjected to TAP purification. Detailed methods for TAP purification can be found in SI Materials and Methods.

MS.

MS was carried out as previously described (14). Briefly, all MS data were acquired on an LTQ-Orbitrap XL mass spectrometer (Thermo Scientific) operating in positive-ion mode with an electrospray voltage of 2.0 kV. The instrument was set to acquire a precursor MS scan in the Orbitrap XL mass spectrometer from m/z 40–2,000, with r = 60,000 at m/z 400 and a target automatic gain control setting of 1e6 ions. In a data-dependent mode of acquisition, tandem MS spectra of the five most abundant precursor ions were acquired in the Orbitrap XL mass spectrometer at r = 7,500 at m/z 400 with a target AGC setting of 2e5 ions. Maximum fill times were set to 1,000 ms for full MS scans and to 500 ms for tandem MS scans with minimum tandem MS triggering thresholds of 5,000 counts. For all experiments, fragmentation occurred in the LTQ linear ion trap, Finnigen LTQ (Thermo Electron Corporation), with a collision-induced dissociation energy setting of 35%, and a dynamic exclusion of 60 s was used for previously fragmented precursor ions. Details of qualitative identifications from raw liquid chromatography/tandem MS data can be found in SI Materials and Methods.

Hierarchical Clustering.

Relative protein abundance in the sample prepared by TAP was quantified using spectral counting (53). To normalize the relative protein abundance, the spectral counts were expressed as a percentage of the total spectra observed in the sample. Mean normalized spectral counts were obtained from multiple independent experiments (n ≥ 2 for each targeted SH2 domain). Hierarchical clustering was performed based on the uncentered Pearson correlation of the mean normalized spectral counts (54, 55).

Peptide Array Synthesis and GST or GST-SH2amb Protein Overlay.

Peptides (19mer) were synthesized as previously described (56) using Auto-Spot Robot ASP 222 (INTAVIS AG). The GST-SH2 domain was purified as previously described (57). The peptide arrays were incubated with the GST-SH2 domain at a concentration of 100 nM in 5% (wt/vol) BSA for 2 h. After a wash with Tris-buffered saline/Tween-20, the GST fusion proteins were detected using HRP-conjugated anti-GST antibody (BETHYL). Positive peptide spots were densitometrically quantified using ImageJ (National Institutes of Health). Phosphodependent binding strength was estimated by subtracting the intensity of phosphopeptide from nonphosphopeptide and then normalized to the control peptide. A peptide showing a binding strength over 20 was assessed as a phosphotyrosine-dependent SH2 binding peptide. Amber codon-modified SH2 proteins were purified by TAP purification and incubated with the peptide arrays derived from EGFR, followed by UV exposure, and bounded SH2 was detected with anti-V5 antibody.

Immunoprecipitation Experiments.

Immunoprecipitation experiments using HEK 293T cells were performed as previously described (58). Briefly, transfected cells were lysed and centrifuged, and the supernatant was incubated with V5 antibody-agarose (Sigma) or GFP trap-agarose (ChromoTek). After beads were washed with lysis buffer, sample buffer was added and subjected to immunoblotting. For endogenous RhoA precipitation, cells were lysed with radioimmunoprecipitation assay buffer, followed by sonication, and centrifuged, and the supernatant was incubated with agarose-conjugated anti-RhoA (26C4; Santa Cruz Biotechnology). Further details can be found in SI Materials and Methods.

Cell Preparation and Image Analysis.

Transfected FreeStyle 293 cells were plated onto poly-d-lysine–coated coverslips, fixed, and stained with anti-V5 antibody (1:500; Invitrogen) or antiphosphotyrosine antibody (1:100, 4G10; Millipore). Images were taken on a Zeiss LSM 710 laser scanning confocal microscope with a 63×/1.4 N.A. oil-immersion objective or 20×/0.8 N.A. objective. Further details for quantification of phalloidin staining are provided in SI Materials and Methods.

2pFLIM Assay.

Details of the 2pFLIM assay have been described previously (26). mEGFP was excited with a Ti/sapphire laser (Mai Tai; Spectraphysics) tuned at a wavelength of 920 nm. Fluorescent images were collected using a 60×/0.9 N.A. objective (Olympus). For fluorescence lifetime imaging, a photoelectron multiplier tube with low transfer time spread (H7422-40p; Hamamatsu) was used. Fluorescence lifetime images were obtained using a time-correlated single-photon counting board (SPC-140; Becker and Hickl) controlled with custom software based on Scanimage software (open source software, Karel Svoboda, Janelia Farm, Ashburn, VA).

Statistical Tests.

Statistical analysis was done using Excel 2007 (Microsoft Corporation). Unless otherwise noted, significance was evaluated using a two-tailed Student t test.