Structural insights into the tyrosine phosphorylation–mediated inhibition of SH3 domain–ligand interactions

Abstract

Src homology 3 (SH3) domains bind proline-rich linear motifs in eukaryotes. By mediating inter- and intramolecular interactions, they regulate the functions of many proteins involved in a wide variety of signal transduction pathways. Phosphorylation at different tyrosine residues in SH3 domains has been reported previously. In several cases, the functional consequences have also been investigated. However, a full understanding of the effects of tyrosine phosphorylation on the ligand interactions and cellular functions of SH3 domains requires detailed structural, atomic-resolution studies along with biochemical and biophysical analyses. Here, we present the first crystal structures of tyrosine-phosphorylated human SH3 domains derived from the Abelson-family kinases ABL1 and ABL2 at 1.6 and 1.4 Å resolutions, respectively. The structures revealed that simultaneous phosphorylation of Tyr⁸⁹ and Tyr¹³⁴ in ABL1 or the homologous residues Tyr¹¹⁶ and Tyr¹⁶¹ in ABL2 induces only minor structural perturbations. Instead, the phosphate groups sterically blocked the ligand-binding grooves, thereby strongly inhibiting the interaction with proline-rich peptide ligands. Although some crystal contact surfaces involving phosphotyrosines suggested the possibility of tyrosine phosphorylation–induced dimerization, we excluded this possibility by using small-angle X-ray scattering (SAXS), dynamic light scattering (DLS), and NMR relaxation analyses. Extensive analysis of relevant databases and literature revealed not only that the residues phosphorylated in our model systems are well-conserved in other human SH3 domains, but that the corresponding tyrosines are known phosphorylation sites in vivo in many cases. We conclude that tyrosine phosphorylation might be a mechanism involved in the regulation of the human SH3 interactome.

Introduction

Src homology 3 (SH3)⁴ domains are protein–protein interaction domains in eukaryotes involved in various intracellular signaling pathways (1, 2). More than 200 human proteins are known to possess one or more SH3 domains. These proteins play various roles in fundamental cellular processes, such as the regulation of cellular movement or proliferation, and are also involved in the development of cancer and several other disorders. SH3 domains are relatively small (∼60 residues) and have an antiparallel β-sandwich topology that is formed by five β-strands (βa, βb, βc, βd, and βe) connected by three loops (RT, N-Src, and distal loops) and a short 3₁₀ helix (Fig. 1A) (1, 2). SH3 domains usually bind short, proline-rich sequences within intrinsically unstructured regions of partner proteins. These linear motifs adopt a left-handed polyproline type II helix conformation. Some SH3 domains are also involved in intramolecular interactions. The most strictly defined consensus sequence of the SH3-binding linear motifs is PXXP, where X denotes any amino acid. There is usually a positively charged residue on either side of the consensus motif (+XXPXXP or XPXXPX+) (1, 2). The shallow, hydrophobic ligand-binding groove bounded by the RT and the N-Src loops is mostly composed of aromatic residues. This interface can be further divided into three smaller pockets (Pockets I-III). Pockets I and II are mainly hydrophobic and accommodate two residues of the binding motif (PX or XP, where X is a hydrophobic residue). One or more negatively charged residues of the RT loop can often be found in Pocket III orienting the ligands through electrostatic interaction with a positively charged residue on either side of the PXXP motif (Fig. 1A) (3).

Figure 1.

A, structure and ligand-binding groove of SH3 domains. The SH3 domain of the tyrosine kinase Src in complex with a bound ligand peptide is shown as an example (PDB entry 1QWE). The ligand-binding interface can be divided into three pockets that are mainly flanked by aromatic residues (blue and orange) and a conserved proline (yellow). Binding groove tyrosines starting from the N terminus are labeled as N, M1, M2, and C. These positions are referred to similarly throughout this work. One or more negatively charged residues of the RT loop (Asp in Src) orient the ligand (red) by interacting with a positively charged residue on either side of the PXXP motif (Arg in this example). B, secondary structural elements and sequence alignment of ABL1 SH3 and ABL2 SH3. Tyrosine phosphorylation sites (based on PhosphoSitePlus) are highlighted in red. Conserved residues of the binding groove are shown with backgrounds corresponding to the color-coding in A. Arrows, differences between ABL1 and ABL2. Sequence numbering is based on isoform 1B of both ABL1 (UniProt identifier P00519-2) and ABL2 (UniProt identifier P42684-1). Note that in ABL1 SH3 and ABL2 SH3, Phe and Trp residues occupy positions M1 and M2, respectively. C, protein sequences of Tyr-to-Phe mutant constructs used in this work.

A recent study showed that tyrosine phosphorylation occurs frequently in SH3 domains (4). Most of the involved tyrosines are part of the ligand-binding groove, and the result of this post-translational modification was the inhibition of partner binding in vivo in the majority of reported cases (4). However, to fully understand the role and function of tyrosine phosphorylation, detailed structural studies at atomic resolution complemented by biochemical and biophysical methods assessing SH3–ligand interactions would be crucial. Therefore, we have chosen ABL1 and ABL2, two SH3-containing nonreceptor tyrosine kinases belonging to the ABL family (5), as a model system to study SH3 domain tyrosine phosphorylation in vitro.

ABL kinases are key regulators of different signal transduction pathways in numerous cell types influencing proliferation, differentiation, survival, death, morphology, motility, and adhesions of cells (5). The SH3 domains of ABL1 and ABL2 or their constitutively active chimeric fusion protein forms (which are responsible for certain human leukemias, such as BCR-ABL (6)) were reported to be phosphorylated in vivo at three homologous tyrosine residues: Tyr⁸⁹ (7, 8) and Tyr¹³⁴ (8, 9) within and Tyr¹¹² outside of the ligand-binding groove in ABL1, whereas Tyr¹¹⁶ and Tyr¹⁶¹ (10) within and Tyr¹³⁹ (10) outside of the ligand-binding groove in ABL2 (Fig. 1B). Whereas tyrosine phosphorylation at multiple sites within the SH3-SH2 region was found to be necessary for full BCR-ABL transforming activity, substitution of Tyr⁸⁹ to phenylalanine was solely sufficient to strongly reduce BCR-ABL–mediated transformation (8). Inhibition of myeloid Src family kinases, which are thought to be the enzymes phosphorylating this region in vivo (7, 8), induced growth arrest and apoptosis in BCR-ABL–transformed cells (11).

In the present study, we performed a database analysis to identify phosphorylated SH3 domains in the human proteome. Four major tyrosine phosphorylation sites were found, all affecting conserved tyrosine residues in positions N, M1, M2, and C of the ligand-binding groove (see Fig. 1A). The SH3 domains of ABL1 and ABL2 kinases were used to investigate the functional and structural effects of tyrosine phosphorylation of the two most prominent sites, N (Tyr⁸⁹ in ABL1 and Tyr¹¹⁶ in ABL2) and C (Tyr¹³⁴ in ABL1 and Tyr¹⁶¹ in ABL2), using classical biochemical binding assays, ¹H-¹⁵N NMR spectroscopy, CD spectroscopy, dynamic light scattering (DLS), small-angle X-ray scattering (SAXS), and X-ray crystallography. To the best of our knowledge, these are the first phosphorylated SH3 domain structures published so far. Comparison of phosphorylated, nonphosphorylated, and ligand-bound ABL SH3 structures revealed no large-scale structural rearrangements; however, ligand binding was strongly inhibited by either of the introduced phosphate groups. These were sterically blocking the binding groove, making it physically inaccessible for the ligands. Phosphorylation of homologous tyrosine residues in other SH3 domains might block ligand binding through a similar molecular level mechanism in general.

Results

Identification of conserved tyrosine phosphorylation sites within human SH3 domains

To assess the importance of tyrosine phosphorylation among SH3 domains and identify novel potential regulatory sites, all human SH3 sequences were aligned (Table S1) and subsequently filtered by the occurrence of tyrosine phosphorylation (using the PhosphoSitePlus database; Table S2). We found 168 individual tyrosine phosphorylation sites in 94 different human SH3 domains. Phosphorylation sites were mapped onto the aligned sequences (Table S2). The four most prominent sites are highlighted on the sequence logo representation of the alignment (Fig. 2). Note that all of these modifications affect positions N (10), M1 (12), M2 (50), and C (55) of the ligand-binding groove (see Fig. 1A). The unfiltered multiple alignment demonstrates that tyrosine residues in these positions are well-conserved (Table S1). The specific occurrence of Tyr at positions N, M1, M2, and C among human SH3 domains is 43.2, 52.0, 15.0, and 53.5%, respectively.

Figure 2.

Sequence logo representation of the tyrosine-phosphorylated human SH3 domains. Key conserved residues of the ligand-binding groove are highlighted in gray background. The four most prominent tyrosine phosphorylation sites (positions 10, 12, 50, and 55, with 32, 13, 20, and 35 occurrences in PhosphoSitePlus, respectively) are indicated in red. (See also the full alignment in Table S2. Note that only positions homologous to ABL1 SH3 are shown here for simplicity.)

The two most common tyrosine phosphorylation sites are positions N and C, with 32 and 35 occurrences in the database, respectively. These residues are located in close proximity to each other in the binding grooves of SH3 domains (see Fig. 1A), and their aromatic rings form the hydrophobic “Pocket I.” We found two other major phosphorylation sites at positions M1 and M2, with 13 and 20 occurrences, respectively. These residues are also important in the formation of the ligand-binding groove. The overlap of major phosphorylation sites with key structural elements of the binding groove strongly suggests that phosphorylation of either one or more tyrosines at position N, M1, M2, and/or C, hereafter referred to as TyrN, TyrM1, TyrM2, and TyrC, respectively, might inhibit ligand binding.

In vitro phosphorylation of the SH3 domain of ABL1 and ABL2

First, we were screening for tyrosine kinases that can be expressed in an active form in Escherichia coli and can be used to catalyze the selective phosphorylation of our SH3 domains. Briefly, kinase domains from distant tyrosine kinase families (SRC, ABL, ephrin, and Tec) were tested both as His-tagged and GST-tagged proteins. We observed acceptable expression levels of autophosphorylated, active His-EphB1, His-HCK, and GST-ABL1 (data not shown). Both ABL1 SH3 and ABL2 SH3 were selectively and effectively phosphorylated by His-EphB1 at TyrN and TyrC (Fig. 1B), as was shown by the combination of anion exchange chromatography and MS (Fig. S1). Even after extensive incubation in the presence of excess ATP, tyrosine residues outside of the binding groove (Tyr¹³⁹ in ABL2 or Tyr¹¹² in ABL1) were not phosphorylated at detectable levels in our experiments. Kinase reactions were scaled up to obtain doubly phosphorylated ABL1 SH3 and ABL2 SH3 (referred to as ABL1 SH3^pYn/pYc and ABL2 SH3^pYn/pYc, respectively, throughout this work) on a milligram scale (see “Experimental procedures”). As we could not achieve acceptable separation of the monophosphorylated SH3 domains (Fig. S1), the following tyrosine-to-phenylalanine mutants were generated: ABL1 SH3^Y89F, ABL1 SH3^Y134F, ABL2 SH3^Y116F, and ABL2 SH3^Y161F, referred to as ABL1 SH3^YnF, ABL1 SH3^YcF, ABL2 SH3^YnF, and ABL2 SH3^YcF, respectively (Fig. 1C). With this approach, the effects of the two individual phosphorylation events could be investigated separately. Importantly, each phosphorylated Tyr-to-Phe mutant was eluted as a single peak from our anion-exchange column (data not shown) corresponding to a singly phosphorylated protein as shown by MS. The lack of chromatographic peaks corresponding to doubly phosphorylated variants showed that EphB1 is indeed specific for binding groove tyrosines.

Tyrosine phosphorylation strongly interferes with binding of ABL1 SH3 and ABL2 SH3 to peptide ligands derived from 3BP-2 and ABI2

The intrinsic tryptophan fluorescence intensity signal was used to determine binding affinities of in vitro phosphorylated and nonphosphorylated ABL1 and ABL2 SH3 domains for peptide ligands corresponding to proline-rich motifs of 3BP-2 and ABI2 (HPPAYPPPPVPT and TPPTQKPPSPPMS, respectively), known binding partners of ABL1 (12, 13). A remarkable blue shift was observed in the spectra of both ABL1 SH3 and ABL2 SH3 upon the addition of the ligands (Fig. 3A). As the highest difference in the fluorescence intensity between the ligand-bound and free SH3 domains was observed at 318 nm, we followed all titrations at this wavelength. Phosphorylation of WT ABL1 and ABL2 SH3 domains at both tyrosines resulted in total loss of ligand binding (Fig. 3, B and C). Dissociation constants obtained for Tyr-to-Phe mutants (3BP-2, ∼4–11 μm; ABI2, ∼100–200 μm) were generally close to those obtained for WT proteins (3BP-2, ∼2 μm; ABI2, ∼100–150 μm), showing that the introduced mutations had only minor effects on ligand binding (Table 1). ABL1 SH3^YnF/pYc and ABL2 SH3^YnF/pYc showed no interaction with either of the ligands. A strong inhibition of ligand binding was also observed in the case of ABL1 SH3^pYn/YcF and ABL2 SH3^pYn/YcF; however, some “residual” binding could be detected (3BP-2, K_d > 100 μm; ABI2, very weak).

Figure 3.

Tryptophan fluorescence–based titrations of ABL1 SH3 and ABL2 SH3. A, tryptophan fluorescence emission spectra of free and 3BP-2 peptide–bound ABL1 SH3 and ABL2 SH3. Ligand binding was associated with a remarkable blue shift. The highest intensity difference was observed at 318 nm. B, complex formation between ABL1 and ABL2 SH3 variants and 3BP-2 followed by fluorescence intensity changes at 318 nm. Both the doubly phosphorylated (ABL1 SH3^pYn/pYc and ABL2 SH3^pYn/pYc) and Tyr-C–phosphorylated (ABL1 SH3^YnF/pYc and ABL2 SH3^YnF/pYc) proteins showed total loss of ligand binding. The inhibitory effect of TyrN phosphorylation (ABL1 SH3^pYn/YcF and ABL2 SH3^pYn/YcF) was weaker but still substantial. C, complex formation between ABL1 and ABL2 SH3 variants and ABI2 followed by fluorescence intensity changes at 318 nm. Although the ABI2-derived ligand showed much weaker binding compared with 3BP-2 (dissociation constants around 100 μm), the effects of phosphorylation were similar. Error bars, S.D. of three independent measurements.

Table 1.

Dissociation constants determined by intrinsic tryptophan fluorescence–based titrations using 3BP-2–derived (HPPAYPPPPVPT) and ABI2-derived (TPPTQKPPSPPMS) ligand peptides

Errors represent the S.E. of fitting.

Crystal structures of tyrosine-phosphorylated ABL1 and ABL2 SH3 domains reveal atomic level details of the inhibition of ligand binding

To obtain atomic-level insights into the mechanism of inhibition and to understand the somewhat different impact of the two phosphorylation events on ligand binding, we crystallized the doubly phosphorylated SH3 domains ABL1 SH3^pYn/pYc and ABL2 SH3^pYn/pYc and the nonphosphorylated ABL2 SH3 as no atomic resolution structure of this domain had been reported previously. Crystal structures were solved at 1.6, 1.4, and 2.0 Å resolutions, respectively (Fig. 4). Crystallographic statistics are reported in Table S3.

Figure 4.

Comparison of the nonphosphorylated and phosphorylated SH3 domains. Structural alignment of ABL1 SH3 (light blue; PDB code 4JJC) with ABL1 SH3^pYn/pYc (dark blue, PDB code 5NP2) (A) and ABL2 SH3 (yellow; PDB code 5NP3) with ABL2 SH3^pYn/pYc (orange; PDB code 5NP5) (B). Key residues of the binding grooves are labeled and shown as sticks.

The sequences of the SH3 domains of ABL1 and ABL2 are highly similar. There are only four different residues in the βc strand and one in the N-Src loop; Cys¹¹⁹, Ala¹²¹, Gln¹²², Thr¹²³, and His¹¹⁴ in ABL1 correspond to Ser¹⁴⁶, Val¹⁴⁸, Arg¹⁴⁹, Ser¹⁵⁰, and Gln¹⁴¹ in ABL2, respectively (see Fig. 1B). By comparing the ABL1 and ABL2 SH3 domain structures (Fig. S2), it seems unlikely that these differences could significantly influence the structure or function of the ligand-binding groove. The most significant differences based on Cα–Cα distances can be found in the RT (SGD region) and distal loops (TKNG and SKNG regions in ABL1 and ABL2, respectively) (Fig. S2); however, it seems very likely that different crystal contacts are the determining factor here.

With the comparison of the phosphorylated SH3 domains with their nonphosphorylated counterparts, no large-scale structural rearrangements could be observed (Fig. 4). We could only see slight differences in certain regions (e.g. VASGDN region in the RT loop and KN region in the distal loop; Fig. S2); however, these again seem to be more related to differences in crystal contacts rather than being the result of tyrosine phosphorylation. Moreover, there are no differences in side-chain conformations that could be obviously interpreted as the result of tyrosine phosphorylation.

To understand the regulatory effect of phosphorylation on ligand binding, we aligned the phosphorylated SH3 structures to a ligand-bound ABL1 SH3 structure (PDB entry 1ABO (14); Fig. 5A). This comparison clearly showed that the phosphate group of pTyrC occupies the center of the binding groove, changing the hydrophobic character of two neighboring pockets (Pockets I and II) to hydrophilic, thus making it impossible for the protein to accommodate the ligand in the right position (Fig. 5B). As ligand binding would result in serious steric clashes between the peptide backbone and the phosphate group, it is not surprising that this modification has a very strong inhibitory effect (see Fig. 3, B and C). The phosphate group of pTyrN is located at the edge of the binding groove (Fig. 5A). Although it is pointing outward, it still occupies some space that would be necessary for the peptide backbone to bind (in Pocket I). The somewhat weaker inhibition in this case (Fig. 3, B and C) might be explained by the higher flexibility of this side chain (see the different conformations in the two crystallographic monomers) and the greater distance of the phosphate oxygens from the peptide backbone (Fig. 5C). Moreover, the presence of this phosphate group affects mainly the C-terminal end of the peptide, potentially allowing some “residual” interaction with other parts of the binding groove.

Figure 5.

Alignment of the crystal structure of ABL1 SH3^pYn/pYc and ABL1 SH3 in complex with a peptide ligand derived from 3BP-1 (PDB code 1ABO (14)). A, for simplicity, only the ligand (salmon) and the phosphorylated domain are shown. Surfaces of the three neighboring hydrophobic pockets within the binding groove are shown in different colors. B, the phosphate group of pTyrC (pTyr134) occupies the binding groove, making it totally inaccessible for the ligand. (Both Pocket I and Pocket II are affected.) C, although the phosphate group of pTyrN (pTyr89) is located at the edge of the binding groove and this side chain seems to be more flexible, steric clashes with the ligand can be expected. (Mainly Pocket I is affected.)

Tyrosine phosphorylation does not significantly affect the structure of ABL1 SH3 and ABL2 SH3 in solution

To show that the introduced phosphate groups or Tyr-to-Phe mutations did not induce significant structural changes in solution, we acquired far-UV CD spectra of all SH3 derivatives (Fig. S3) and determined the secondary structure contents by BeStSel (http://bestsel.elte.hu/index.php)⁵ (44, 61) (Table S4). All studied SH3 variants were nearly exclusively composed of relaxed and right-twisted antiparallel β-sheets (∼10 and ∼18 residues, respectively), turns (∼5 residues), and other structures (∼27 residues), whereas no helices or parallel or left-twisted antiparallel β-sheets were observed. The fact that only slight differences were detected in the secondary structure contents strongly suggests that the three-dimensional structure of ABL1 SH3 and ABL2 SH3 is significantly affected neither by the mutations nor by phosphorylation.

Similarly, ¹H-¹⁵N HSQC spectra of nonphosphorylated and phosphorylated ABL1 SH3 derivatives (Fig. S4) indicate only minor structural perturbations upon tyrosine phosphorylation (Fig. S5), showing that the overall structural integrity of the protein was not affected by these molecular level alterations. The behavior of the N- and C-terminal regions (Phe⁸⁵–Thr¹⁰² and Ser¹³²–Val¹³⁸, respectively) seems to be correlated, as phosphorylation of either TyrN or TyrC mutually affects the environment of the other. This is manifested in a similar overall pattern of phosphorylation-induced chemical shift changes over the SH3 sequence for each investigated derivative with the most pronounced changes occurring if both residues are phosphorylated. This correlation and synergism most likely arise from the spatial proximity of TyrN and TyrC, as they are located close to each other within the ligand-binding groove forming the same “functional unit” (Pocket I) of the domain.

SH3 domain tyrosine phosphorylation does not regulate the dimerization of ABL kinases

ABL1 has been shown to oligomerize at high levels of expression in COS cells (15). Moreover, dimerization of the full-length ABL kinases might promote their activation (16–18). As an introduced phosphate group may promote (or interfere with) the formation of dimers/oligomers, and several SH3 domains have already been shown to form homodimers by different mechanisms (19–22), we assessed the effect of phosphorylation on the oligomeric state of ABL1 SH3 and ABL2 SH3. We found that pTyrN was indeed involved in crystal contacts in both the crystal structures of ABL1 SH3^pYn/pYc and ABL2 SH3^pYn/pYc (Fig. S6). This residue forms a salt bridge with a Lys residue (Lys¹⁰³ in ABL1, Lys¹³⁰ in ABL2) of a symmetry-related molecule, indicating a possible monomer–dimer equilibrium that may exist in solution and might be regulated by phosphorylation of TyrN. Moreover, all four crystal structures (ABL1 SH3, ABL2 SH3, ABL1 SH3^pYn/pYc, ABL2 and SH3^pYn/pYc; PDB entries 4JJC, 5NP3, 5NP2, and 5NP5, respectively) shared a very similar hydrophobic crystal contact surface built by the same cluster of residues that seemed ideal for forming homodimers (Fig. S6).

However, both SAXS and DLS analysis showed that the radius of gyration (R_g) of ABL1 SH3 increased by only ∼0.1 nm upon phosphorylation. This is consistent with the incorporation of the phosphate groups, but not with dimerization (Fig. S7 and Table S5). Direct comparison of our crystal structure with the SAXS profile assuming a monomer–dimer equilibrium model indicated ∼10 and ∼20% dimeric fraction for ABL1 SH3 and ABL1 SH3^pYn/pYc, respectively (χ² ∼0.8). The fact that neither the protein concentration (data not shown) nor phosphorylation shifted the monomer/dimer ratio effectively in these experiments suggests that dimerization of this SH3 domain is not regulated by tyrosine phosphorylation, and the observed dimeric fraction was most likely the result of nonspecific protein aggregation.

To complement these studies, oligomerization was further investigated by NMR relaxation experiments (Table S6 and Fig. S8). Phosphorylation resulted in a slight perturbation of atomic relaxation rates of ABL1 SH3 with no measurable effect on macroscopic dynamic properties. Singly phosphorylated variants (ABL1 SH3^YnF/pYc and ABL1 SH3^pYn/YcF) showed a similar behavior. The highly similar relaxation properties are manifested in nearly identical values of rotational correlation times (τ_c) for the four ABL1 SH3 variants (Table S6) corresponding to the expected value for a ∼7-kDa globular protein in aqueous buffer. Based on τ_c, R_g was calculated for all SH3 variants. A good correlation between the DLS/SAXS- and NMR-derived parameters was found, although DLS and SAXS slightly overestimated R_g. This is most likely explained by the fact that whereas the DLS and SAXS derived parameters are affected by the presence of oligomers, the NMR-derived parameters reflect solely the properties of the monomeric fraction. The oligomeric fraction might be heterogeneous and thus not be detectable in NMR. We conclude that neither the phosphorylation of pTyrN nor the phosphorylation of pTyrC changes the monomeric state of ABL1 SH3 (or ABL2 SH3; data not shown) in solution.

Discussion

Identification of major regulatory tyrosine phosphorylation sites in SH3 domains

In this work, we performed a database analysis for tyrosine phosphorylation in SH3 domains. A previous study identified TyrN within the conserved ALpYDY motif as the most preferred phosphorylation site (4). Here we found that TyrC within the YFPSNpYV motif (Fig. 2) is another prominent site, with even more occurrences in the human phosphoproteome (Table 2 and Table S2). Our experimental strategy allowed us to obtain phosphorylated SH3 domains in large quantities that meet the requirements of in vitro biochemical and even structural studies. We observed that phosphorylation of either TyrN or TyrC in the SH3 domain of ABL1 or ABL2 interferes with partner binding. Simultaneous phosphorylation of both residues, which is catalyzed by several Src-family kinases in vivo, has a similar effect (8). In other human SH3 domains, phosphorylation not only occurs most frequently on these tyrosine residues, but in many cases, both modifications were observed in the same protein (see Table 2 and Table S2). All of these findings strongly suggest that tyrosine kinases may generally accept both TyrN and TyrC as substrates, and phosphorylation of one of these residues does not interfere with the phosphorylation of the other.

Table 2.

N- and C-type tyrosine phosphorylation in SH3 domains

ABI1, ABL interactor 1; ABI2, ABL interactor 2; ABL1, Abelson tyrosine-protein kinase 1; ABL2, Abelson tyrosine-protein kinase 2; ARHGAP12, Rho GTPase-activating protein 12; ARHGAP42, Rho GTPase-activating protein 42; BAIAP2L1, brain-specific angiogenesis inhibitor 1–associated protein 2–like protein 1; BCAR1, breast cancer anti-estrogen resistance protein 1; BTK, tyrosine-protein kinase BTK; CRKL, Crk-like protein; CTTN, Src substrate cortactin; DNMBP, dynamin-binding protein; EPS8, epidermal growth factor receptor kinase substrate 8; EPS8L2, epidermal growth factor receptor kinase substrate 8–like protein 2; EPS8L3, epidermal growth factor receptor kinase substrate 8–like protein 3; FGR, tyrosine-protein kinase Fgr; FRK, tyrosine-protein kinase FRK; FYN, tyrosine-protein kinase Fyn; GRAP2, GRB2-related adapter protein 2; GRB2, growth factor receptor–bound protein 2; ITK, tyrosine-protein kinase ITK; ITSN1, intersectin-1; ITSN2, intersectin-2; LASP1, LIM and SH3 domain protein 1; LYN, tyrosine-protein kinase Lyn; MAP3K21, mitogen-activated protein kinase kinase kinase 21; MAP3K9, mitogen-activated protein kinase kinase kinase 9; NEDD9, neural precursor cell–expressed developmentally down-regulated protein 9; NOSTRIN, nitric-oxide synthase traffic inducer; PACSIN3, protein kinase C and casein kinase substrate in neurons protein 3; PIK3R1, phosphatidylinositol 3-kinase regulatory subunit α; PLCG2, phospholipase C-γ-2; SASH1, SAM and SH3 domain–containing protein 1; SASH3, SAM and SH3 domain–containing protein 3; SH3GL1, endophilin-A2; SH3PXD2A, SH3 and PX domain–containing protein 2A; SH3PXD2B, SH3 and PX domain–containing protein 2B; SH3RF1, E3 ubiquitin-protein ligase SH3RF1/SH3 domain–containing RING finger protein 1; SH3RF3, SH3 domain–containing RING finger protein 3; SORBS2, sorbin and SH3 domain–containing protein 2; SPTA1, spectrin α chain, erythrocytic 1; SPTAN1, spectrin α chain, non-erythrocytic 1; SRC, proto-oncogene tyrosine-protein kinase Src; TEC, tyrosine-protein kinase Tec; TXK, tyrosine-protein kinase TXK; VAV1, proto-oncogene vav; VAV3, guanine nucleotide exchange factor VAV3; YES1, tyrosine-protein kinase Yes; GEF, guanine nucleotide exchange factor; HTP, high-throughput; LTP, low-throughput.

Effects of TyrC phosphorylation

Phosphorylation of TyrC resulted in total inhibition of ligand binding in our experiments. As TyrC lies between two hydrophobic binding pockets in the middle of the binding groove, the introduced phosphate group fully occupies the space that would be necessary for ligands with polyproline-type II helix conformation. In agreement with this, all TyrC-type phosphorylation events are known to inhibit partner binding in other proteins. For example, the EGF-induced phosphorylation of cortactin SH3 disrupts the interaction with TIP150 (23). Phosphorylation of TyrC in the C-terminal SH3 domain of GRB2 by BCR-ABL or EGFR results in the inhibition of Sos binding (24). This mechanism represents a negative regulatory loop limiting EGF-mediated signaling in time (24) and also allows other signaling pathways to antagonize EGF-induced MAP kinase activation and cell proliferation (25, 26). Phosphorylation of TyrC in VAV1 SH3 interferes with autoinhibitory intramolecular interactions, leading to the activation of the protein (27).

Effects of TyrN phosphorylation

Phosphorylation of TyrN was found to inhibit ligand binding significantly in this work (see Fig. 3, B and C). Interestingly, whereas this modification in endophilin A2 is known to disrupt the interaction with dynamin (28, 29), it does not inhibit the interaction with focal adhesion kinase (28). Moreover, TyrN-type phosphorylation of the adaptor protein ABI1 seems to enhance the interaction with ABL1 (30). Phosphorylation of TyrN in the Tec family tyrosine kinase BTK was found to inhibit binding to WASP, while not influencing the interaction with c-Cbl, and was even required for the interaction with kinase-active SYK (31). These observations indicate that TyrN (and maybe TyrC) phosphorylation can be part of more complex molecular level mechanisms in vivo. For example, other regulated domain-level interactions may exist in parallel in multidomain proteins and their complexes that depend on the inhibition of SH3 domain function. TyrN phosphorylation might also be involved in the activation of autoregulated multidomain proteins by disrupting autoinhibitory intramolecular interactions. One example is the Tec-family tyrosine kinase ITK, where TyrN phosphorylation extends the duration of kinase activation (32, 33). This modification and also the phosphorylation of the homologous site in BTK may serve as “checkpoints” altering the interaction profile of these kinases with other proteins, thereby controlling the transition between early and late activation events (32, 34, 35).

Regulation of ABL kinases by SH3 domain tyrosine phosphorylation

Here, we presented the crystal structures of two closely related tyrosine-phosphorylated SH3 domains: ABL1 SH3^pYn/pYc and ABL2 SH3^pYn/pYc. Only minor changes were observed in the structures upon phosphorylation; however, the ability of the SH3 domains to bind peptide ligands was compromised by the phosphate groups, mainly by steric blocking of the binding groove. Activity of the kinase domain in ABL1 and ABL2 are regulated by autoinhibition (36, 37). In the “closed,” inactive conformation, the SH3 domain binds to the proline-rich linker between the SH2 and kinase domains (36, 37). Tyrosine phosphorylation within the ligand-binding groove of the SH3 domain disrupts this interaction in vivo (7, 8). Although we could not observe any interaction between the recombinant ABL1 and ABL2 SH3 domains and a synthetic peptide corresponding to this linker in our experiments (data not shown), our results demonstrate that either the phosphorylation of TyrN or TyrC strongly interferes with ligand binding. Phosphorylation of any of these residues or both may likely lead to the activation of ABL kinases and inhibit their SH3-mediated interactions with external ligands. Interestingly, whereas the nonreceptor tyrosine kinase SRC is regulated by a similar autoinhibitory mechanism, phosphorylation of TyrC only inhibits partner binding and is not required for the activation of the kinase (38).

Although our crystal structures raised the possibility of a monomer–dimer equilibrium regulated by the phosphorylation of TyrN, we demonstrated by DLS, SAXS, and NMR that neither the phosphorylation of TyrN nor the phosphorylation of TyrC affects the monomeric state of these domains in solution.

No phosphorylation of Tyr¹¹² in ABL1 or Tyr¹³⁹ in ABL2 was observed in this study. These residues are not part of the ligand-binding groove, and they do not contribute to any intramolecular interaction necessary for the formation of the autoinhibited conformation of ABL kinases (37). Therefore, it seems to be unlikely that there is any effect associated with their phosphorylation on the known functions of ABL1 SH3 or ABL2 SH3, including the binding of external ligands. To the best of our knowledge, there has been no information reported on the regulatory functions of these phosphorylation events so far.

Summary

In this work, we recognized novel major tyrosine phosphorylation sites in the binding groove of SH3 domains and demonstrated that SH3 domain tyrosine phosphorylation strongly inhibits the interaction with proline-rich ligands due to the steric blocking of the ligand-binding groove by the phosphate groups. Phosphorylation is not followed by significant structural rearrangements or oligomerization of the domains. Based on the occurrence of phosphotyrosine residues in homologous positions of many human SH3 domains, this seems to be a widespread mechanism regulating a significant part of the SH3 interactome in cells.

Experimental procedures

Sequence analysis of SH3 domains and identification of phosphorylation sites

Sequences of human SH3 domains were obtained from UniProt (40) and subsequently aligned by Clustal Omega (41). The alignment was reviewed, and the boundaries of SH3 domains were corrected manually. Tyrosine phosphorylation sites were mapped based on both high- and low-throughput data from PhosphoSitePlus (42). All nonphosphorylated SH3 domain sequences were removed, and the resulting multiple alignment was used as a template to generate a sequence logo by using WebLogo (43).

DNA constructs

The TEV protease gene cloned into a bacterial expression vector was a kind gift of Dr. László Nyitray. DNA constructs harboring the coding region of human ABL1, ABL2, and EphB1 kinases were obtained from Addgene. Sequences corresponding to the SH3 domain of ABL1 (Uniprot accession number P00519, residues 64–120) and ABL2 (Uniprot accession number P42684, residues 110–166) and the kinase domain of EphB1 receptor (Uniprot accession number P54762, residues 612–892) were amplified by PCR and subcloned into a modified pET vector encoding a His tag followed by the TEV protease recognition site by using NdeI and BamHI restriction enzymes. Mutants of the ABL1 and ABL2 SH3 domains were generated by PCR by using a third oligonucleotide carrying the base changes necessary for the mutations. All constructs were verified by DNA sequencing (Eurofins Genomics).

Protein expression and purification

WT and mutant SH3 domains of ABL kinases and the kinase domain of EphB1 receptor were expressed in E. coli Rosetta pLysS (Novagen) and Arctic Express (Agilent Technologies) cells, respectively. Cells were grown in 2YT medium up to A₆₀₀ = 0.7 at 37 °C and induced overnight with 0.5 mm isopropyl 1-thio-β-d-galactopyranoside at 16 °C. The ¹⁵N-enriched SH3 domains were expressed in E. coli Rosetta pLysS cells. Cultures were grown in 4 × 1000 ml of 2YT medium up to A₆₀₀ = 0.7 at 37 °C. Cells were harvested by centrifugation and resuspended in 1 liter of minimal medium containing 1 g/liter [¹⁵N]NH₄Cl and induced by 0.5 mm isopropyl 1-thio-β-d-galactopyranoside at 37 °C for 4 h. Purification of all proteins followed the same protocol. After centrifugation, cells were resuspended in lysis buffer (50 mm Na₂HPO₄, 500 mm NaCl, 0.25 mm TCEP, pH 8.0) and subsequently lysed by lysozyme treatment, one freeze-thaw cycle, and sonication. Proteins were purified by immobilized metal affinity chromatography by using Profinity^TM IMAC nickel-charged resin (Bio-Rad). Fractions containing the protein of interest were pooled and subsequently dialyzed against lysis buffer. His tags were cleaved by TEV protease digestion. To remove the protease, residual undigested protein molecules, and digested His tags, samples were loaded again onto a nickel-charged IMAC column. The flow-through was collected, dialyzed against buffer A (20 mm Tris, pH 8.0) and further purified by anion-exchange chromatography (HiTrapQ column) applying a linear gradient of buffer B (20 mm Tris, 1 m NaCl, pH 8.0). Finally, protein samples were dialyzed against Tris-buffered saline (TBS) supplemented with TCEP (20 mm Tris, 150 mm NaCl, 0.5 mm TCEP, pH 7.6), concentrated by ultrafiltration, aliquoted, and stored at −80 °C.

SH3 domain phosphorylation

In vitro kinase reactions were performed by the recombinant kinase domain of EphB1 receptor in a buffer containing (20 mm Tris, 150 mm NaCl, 2 mm ATP, 10 mm MgCl₂, 0.5 mm TCEP, 0.01% Triton X-100, pH 7.6) at 30 °C overnight. Typical concentrations of substrates and the kinase were 500 and 10 μm in reaction mixtures, respectively. Reactions were stopped by the addition of 0.5 m EDTA, pH 7.6 (20 μm final concentration). Samples were diluted with buffer A to reach a final NaCl concentration of 15 mm. Phosphorylated SH3 domains were separated by HPLC by using a Sepax Proteomix SAX-NP3 (strong anion exchanger) column and a linear gradient of buffer A and buffer B. Fractions were analyzed by MS to identify the number of modified tyrosines and their positions within the sequence. (See details in the supporting information.)

Peptide synthesis

The synthetic peptides corresponding to the SH3-binding motif of 3BP-2 (HPPAYPPPPVPT; Uniprot accession number P78314, residues 200–211) and ABI2 (TPPTQKPPSPPMS; Uniprot accession number Q9NYB9, residues 175–187) were obtained from GeneScript. The N and C termini of the peptides were acetylated and amidated, respectively.

Ligand-binding assays (fluorescence spectroscopy)

Titrations of the ABL1 and ABL2 SH3 domains by peptide ligands corresponding to the ABL1 SH3-binding linear motif of 3BP-2 and ABI2 (HPPAYPPPPVPT and TPPTQKPPSPPMS, respectively) were followed by the intrinsic tryptophan fluorescence signal. Prior to titrations, SH3 samples were dialyzed against PBS supplemented with 0.5 mm TCEP, and the peptide ligands were also dissolved in the same buffer. Titrations were performed by pipetting the titrant solution (3BP-2 or ABI2 peptide, 500 μm and 1.7 mm, respectively) in 2-μl steps into the cuvette. The starting volume and concentration of the SH3 solutions in the cuvette were 800 μl and 5 μm, respectively. Experiments were carried out at 25 °C by using a FluoroMax-3 spectrofluorometer (HORIBA Scientific). Excitation of the fluorophores was performed at 295 nm with a spectral bandwidth of 2 nm. Emission scans were acquired in the wavelength range from 310 to 400 nm with an emission bandwidth of 2 nm. Titrations were followed at 318 nm (the difference between the fluorescence intensities of the apo- and ligand-bound states was the highest at this wavelength). The average signal intensity and S.D. of three independent titrations were plotted against the volume of the titrant. Data were fitted by the following equation to determine the dissociation constants,

$$I=a\left[A\right]{\prime }_t+\left(b-a\right)\frac{K_d+\left[A\right]{\prime }_t+\left[B\right]{\prime }_t-\sqrt{{\left(K_d+\left[A\right]{\prime }_t+\left[B\right]{\prime }_t\right)}^2-4\left[A\right]{\prime }_t\left[B\right]{\prime }_t}}{2}$$ (Eq. 1)

where [A]′_t = C_AV_A/(V_A + V_B) is the total concentration of the SH3 domain in the cuvette at each point of titration, [B]′_t = C_BV_B/(V_A + V_B) is the total concentration of the ligand peptide in the cuvette at each point of titration, K_d is the dissociation constant, C_A and C_B are the concentrations of the stock solutions, V_A and V_B are the total volumes of the stock solutions mixed at the given point of titration, and a and b are the theoretical fluorescence intensities of the ABL1 SH3 solution and the ABL1 SH3–ligand peptide complex solution at 1 m concentration, respectively.

CD spectroscopy

CD spectra were recorded at 25 °C in the wavelength range from 190 to 250 nm (1-nm increments) on a JASCO J-715 circular dichroism spectropolarimeter (JASCO Inc.). The scanning speed was set to 20 nm/min, and the spectral bandwidth was 2 nm. Protein samples were dialyzed against CD buffer (50 mm phosphate, pH 7.4) overnight and diluted to a concentration of 20 μm prior to the experiments. The spectrum of the buffer was subtracted from each protein spectrum. Corrected far-UV CD spectra were analyzed by the BeStSel (http://bestsel.elte.hu/index.php)⁵ secondary structure prediction server (44, 61).

Protein crystallization

All proteins (ABL2 SH3, ABL2 SH3^pY116/pY161, and ABL1 SH3^pY89/pY134) were dialyzed twice against TBS supplemented with TCEP and sodium azide (20 mm Tris-HCl, 150 mm NaCl, 3 mm NaN₃, 0.5 mm TCEP, pH 7.6) and concentrated to ∼750 μm using Amicon Ultra-15 3K centrifugal filter devices (Millipore). Crystals were grown using the hanging-drop vapor diffusion method at 293 K with a reservoir solution of 0.8 m sodium citrate, 0.1 m sodium cacodylate, pH 6.5, in the case of ABL1 SH3^pY89/pY134; 1.8 m disodium dl-malate, pH 7.0, in the case of ABL2 SH3; and 1.8 m (NH₄)₂SO₄, 0.1 m sodium acetate, pH 4.6, in the case of ABL2 SH3^pY116/pY161. Drops were composed of 1 μl of reservoir solution and 1 μl of protein solution. Crystals were soaked in reservoir solution supplemented with 15% glycerol for ∼30 s and subsequently cooled in liquid nitrogen.

X-ray diffraction experiments were performed at beamline P13 of the PETRA III synchrotron source (EMBL Hamburg). Data were integrated with XDS (45), and the phase problem was solved by PHASER (46). A high-resolution X-ray structure of ABL1 SH3 was used as a search model (PDB entry 4JJC). The MR search identified four, two, and two molecules in the asymmetric unit of structure ABL2 SH3, ABL2 SH3^pY116/pY161, and ABL1 SH3^pY89/pY134, respectively. Structure refinement was carried out in Phenix (47), and modeling was done in Coot (48). The structures were deposited in the PDB under reference codes 5NP3, 5NP5, and 5NP2. Crystallographic data and refinement statistics are shown in Table S3.

SAXS experiments

SAXS experiments were carried out at the BM29 beamline at ESRF, Grenoble. Prior to the measurements, samples were dialyzed against TBS supplemented with 0.5 mm TCEP. A dilution series was measured from 2.8 to 0.4 mg/ml to reduce the interparticle effect from the scattering profile, but no concentration-dependent changes were observed during data collection. Data were analyzed using the ATSAS program package (49). Primary data were analyzed with PRIMUS (50). Fitting to the monomeric crystal structure was performed with CRYSOL (51), and to fit with the monomer–dimer ensemble, we used OLIGOMER (50). The structure of the dimer was assumed to be identical to the crystallographic dimer stabilized by the interactions of pTyr⁸⁹ in ABL1 SH3 (Fig. S6).

Dynamic light scattering

The hydrodynamic size distribution profile was determined by an Avid Nano W130i instrument at 25 °C. Prior to the measurements, samples were dialyzed against TBS supplemented with 0.5 mm TCEP. Protein concentration of ABL1 SH3 and ABL1 SH3^pTyr89/134 was 5.05 and 3.82 mg/ml, respectively. R_g values were calculated from the experimental hydrodynamic radii (R_h) using the scaling factor 0.775.

NMR spectroscopy

Phosphorylation-induced structural changes were monitored by two-dimensional ¹H-¹⁵N HSQC NMR measurements (52) performed on uniformly U-¹⁵N–enriched WT and mutated (Y89F and Y134F) ABL1 SH3. Experiments were carried out on a 600-MHz Varian NMR system spectrometer equipped with a 5-mm indirect detection triple resonance (¹H¹³C¹⁵N) z axis gradient probe in PBS buffer, pH 7.2, containing 3 mm NaN₃ and 0.5 mm TCEP at 25 °C.

Resonance assignment was obtained on uniformly U-¹⁵N–enriched ABL1 SH3 at 25 °C using a combination of 2D ¹H-¹⁵N HSQC (52), 3D ¹H-¹⁵N TOCSY-HSQC (τ_m = 120 ms), and ¹H-¹⁵N NOESY-HSQC (τ_m = 250 ms) (53), 2D ¹H-¹H TOCSY (τ_m = 120 ms) (54), ¹H-¹H NOESY (τ_m = 50, 100, and 250 ms) (55), and natural abundance ¹H-¹³C HSQC (56) experiments. Spectral processing, computer-assisted spin-system analysis, and resonance assignment were carried out using Felix 2004 (Accelrys, Inc.). To obtain the ¹H,¹⁵N assignment of phosphorylated derivatives, the assignment of nonphosphorylated ABL1 SH3 was transferred. Combined (¹HN, ¹⁵N) chemical shift perturbations were calculated using the equation, Δδ_1HN,15N = √((Δδ²_HN+Δδ²_N/25)/2) (57).

The ¹⁵N relaxation measurements (52, 58, 59) were collected on nonphosphorylated and phosphorylated derivatives (pY89, pY134, and pY89pY134) of U-¹⁵N–enriched ABL1 SH3 at 25 °C at a magnetic field strength of 14.1 teslas (corresponding to ¹⁵N Larmor frequency of 60 MHz). Backbone amide ¹⁵N R₁ values were measured from a series of eight spectra with the following relaxation delay times: 50, 100, 170, 240, 340, 480, 630, and 800 ms and 50, 100, 170, 240, 420, 610, 720, and 800 ms. Amide ¹⁵N R₂ values were obtained similarly: 10, 30, 50, 70, 110, 170, 210, and 250 ms and 10, 30, 50, 90, 130, 150, 170, and 230 ms. Fourier transformation of free induction decays was performed using Felix 2004 (Accelrys, Inc.). For relaxation analysis, the transformed and phased spectra were imported into CCPNMR. Rotational correlation times (τ_c) were calculated from ¹⁵N R₁, R₂ relaxation rates using the corresponding equation in Kay et al. (60) considering the J(0) and J(ωN) spectral density terms. R_h values were derived based on the equation, τ_c = 4πη_H2OR_h³/3k_BT, where η_H2O is the viscosity of water, T is the temperature, and k_B is the Boltzmann constant.

Data and material availability

Source data for biochemical experiments that support the findings of this study are available from the corresponding authors upon reasonable request.

Author contributions

B. M. carried out the experiments, analyzed data, and wrote the paper. L. R., V. V., and L. B. conceived the project, supervised the research, and prepared the manuscript. G. G. participated in the X-ray and SAXS measurements. O. T. performed the NMR measurements. K. K. and B. S. helped in data analysis and drawing figures. A. C. and G. K. assisted in the tryptophan fluorescence-based titrations. M. D. helped in the generation of recombinant proteins. S. S. contributed to the in vitro phosphorylation experiments. L. N. helped in writing the manuscript. I. L. and B. G. V. run the Biostruct Laboratory and assisted in the in-house X-ray scattering experiments. All authors analyzed the results and approved the final version of the manuscript.

Addendum

While this manuscript was being prepared for publication, an article by Dionne et al. (39) appeared reporting the phosphorylation of TyrC residues in the SH3 domains of NCK by the receptor tyrosine kinase Eph4a. Their findings confirm our conclusions showing that phosphorylation of TyrC inhibits ligand binding and functions as a negative regulatory loop through which receptor tyrosine kinases terminate signaling.