The CD2AP (CD2-associated protein) and CIN85 (Cbl-interacting protein of 85 kDa) adaptor proteins each employ three Src homology 3 (SH3) domains to cluster protein partners and ensure efficient signal transduction and down-regulation of tyrosine kinase receptors. Using NMR, isothermal titration calorimetry and small-angle X-ray scattering methods, we have characterized several binding modes of the N-terminal SH3 domain (SH3A) of CD2AP and CIN85 with two natural atypical proline-rich regions in CD2 (cluster of differentiation 2) and Cbl-b (Casitas B-lineage lymphoma), and compared these data with previous studies and published crystal structures. Our experiments show that the CD2AP-SH3A domain forms a type II dimer with CD2 and both type I and type II dimeric complexes with Cbl-b. Like CD2AP, the CIN85-SH3A domain forms a type II complex with CD2, but a trimeric complex with Cbl-b, whereby the type I and II interactions take place at the same time. Together, these results explain how multiple interactions among similar SH3 domains and ligands produce a high degree of diversity in tyrosine kinase, cell adhesion or T-cell signaling pathways.
Structured digital abstract
- CIN85 and Cbl-b bind by nuclear magnetic resonance (View Interaction: 1, 2)
- CD2 and CIN85 bind by nuclear magnetic resonance (View interaction)
- CD2 and CD2AP bind by nuclear magnetic resonance (View interaction)
- Cbl-b and CD2AP bind by nuclear magnetic resonance (View Interaction: 1, 2)
- Cbl-b and CIN85 bind by isothermal titration calorimetry (View Interaction: 1, 2, 3, 4)
- CIN85 and Cbl-b bind by X-ray scattering (View Interaction: 1, 2)
- CIN85 and CD2 bind by isothermal titration calorimetry (View interaction)
- CD2 and CD2AP bind by X-ray scattering (View interaction)
- CD2AP and CD2 bind by isothermal titration calorimetry (View interaction)
- CD2 and CIN85 bind by X-ray scattering (View interaction)
- CD2AP and Cbl-b bind by isothermal titration calorimetry (View Interaction: 1, 2, 3, 4, 5)
Casitas B-lineage lymphoma
cluster of differentiation 2
Cbl-interacting protein of 85 kDa
Cas ligand with multiple SH3 domains
- SH3 domain
Src homology 3 domain
Adaptor molecules are non-catalytic polypeptides that contain one or more domains that are able to bind to other protein or non-protein complexes that transmit intracellular signals involved in the regulation of cell growth, differentiation, migration and survival .
Formation of higher-order oligomers is required to mediate efficient transduction of signals in many cascades, such as internalization of epidermal growth factor receptors induced by receptor tyrosine kinases. The CIN85 (Cbl-interacting protein of 85 kDa)/CD2AP (CD2-associated protein) family of adaptor proteins [CD2AP is the mouse ortholog of Cas ligand with multiple SH3 domains (CMS)] is involved in orchestrating multiple steps in receptor tyrosine kinase signaling and endocytosis , and the family members share a similar domain organization with a high degree of sequence conservation.
CIN85 and CD2AP are implicated in down-regulation of diverse tyrosine kinases mediated by Casitas B-lineage lymphoma (Cbl) . Growth factor-induced tyrosine phosphorylation of Cbl, which enhances binding of the atypical proline-rich sequence present in Cbl/Cbl-b to CIN85 or CD2AP via its Src homology 3 (SH3) domains [3-5], is crucial for targeting receptor tyrosine kinases to clathrin-mediated endocytosis. In the cell, interaction of the SH3 domains of the CIN85/CD2AP family is involved in internalization of signal receptors by ligand stimulation . Association of CD2AP and CIN85 with the CD2 (cluster of differentiation 2) receptor is also induced upon T-cell activation [2, 6]. It has been proposed that central clustering of CD2 has an essential role in antigen receptor engagement and signaling . Direct molecular interaction of CD2AP with CD2 is crucial for CD2 clustering as well as T-cell polarization and recruitment of CD2 during formation of an immunological synapse [4, 6]. CIN85 has also been proposed to participate in this process as it is highly expressed in T cells and binds CD2 [2, 4].
The composition and stoichiometry of protein clusters involving such adaptor molecules is not fully established and may vary from dimeric to trimeric or larger oligomeric complexes. CIN85 and CD2AP facilitate the formation of large complexes required for efficient internalization of cell-surface receptors by means of three N-terminal SH3 domains (SH3A, SH3B and SH3C) [2, 3]. Most SH3 domains bind to proline-rich sequences containing a PxxP motif. They are recognized in either type I orientation (when the SH3 hydrophilic pocket accommodates basic residues N-terminal to the PxxP motif) or type II orientation (when the pocket is occupied by basic residues C-terminal to the motif). However, the N-terminal SH3 domain of CIN85 (SH3A) was found to recognize an atypical polyproline/arginine binding motif (PxxxPR) that is present in the ubiquitin ligases Cbl-b and c-Cbl [3, 4]. Crystallographic studies showed that a Cbl-b-derived peptide covering the proline-rich region from residue 902–912 interacts with two SH3 domains (Fig. 1) in a pseudo-symmetrical fashion forming a heterotrimeric complex . The peptide is simultaneously recognized by two SH3A domains; one SH3A engages the peptide in a manner resembling a type II interaction, whereas the second SH3A binds on the opposite side of the ligand, resembling a type I interaction. Moreover, isothermal titration calorimetry (ITC) data show a marked deviation from the 1 : 1 stoichiometry expected for a single-site interaction, which suggests formation of higher oligomers such as the trimers observed in the crystal structure . In support of this, heterotrimerization was also demonstrated in vivo .
Similarly to Cbl-b and other target proteins, CD2 is also recognized via its atypical polyproline/arginine sequence PxxxPR [3, 7]. The crystal structure of the CMS N-terminal SH3A domain (which shares 100% sequence identity with the SH3A domain of CD2AP) in complex with a CD2-derived peptide also suggested formation of heterotrimers, providing a conserved feature in the molecular mechanism of cluster formation involving the CIN85/CMS family of adaptor proteins . However, these findings were later disputed because no evidence was observed for the existence of a heterotrimer between CIN85 and Cbl-b in solution [3, 9], although the experimental conditions were not directly comparable .
All three SH3 domains in CIN85 and CD2AP recognize the same polyproline-rich consensus motifs, and each of them may therefore bind a molecular partner. Moreover, SH3-mediated multimerization implies that ligands such as Cbl also contribute to molecular clustering by recruiting the adaptor molecules . SH3-mediated oligomerization has only been reported for the N-terminal SH3A domain, but not for the SH3B or SH3C domains [5, 8].
To clarify the composition and stoichiometry of the clusters that the SH3A domains of adaptor proteins CD2AP and CIN85 form in solution with their natural targets CD2 and Cbl-b, we have performed a thermodynamic and structural characterization combining three complementary solution techniques: ITC, NMR and small-angle X-ray scattering (SAXS).
The results clearly show that the interaction mechanism of the SH3A domain is different for the two adaptor proteins. The CD2AP-SH3A domain forms a dimer with both CD2 and Cbl-b ligands, with CD2 only in the type II orientation, but with Cbl-b in both orientations (type I and II). In the case of the SH3A domain of CIN85, the novel interaction with CD2 leads to a type II dimer, as for CD2AP-SH3A, but the interaction with Cbl-b, in agreement with the X-ray studies and in vivo experiments , leads to formation of a heterotrimer consisting of two SH3 domains sandwiching a Cbl-b peptide. The experiments show that both SH3A domains are able to interact in type I and type II orientations depending on whether the atypical proline targets contain an arginine at the N-terminus (for type I) and/or at the C-terminus (for type II). The atypical PxxxPR recognition sequence thus needs to be extended to RxPxxxPR to include the type I interaction. The formation of a dimer or a trimer possibly correlates with steric differences associated with the n-Src loop of the SH3A domains of CIN85 and CD2AP. The differential mode of recognition observed between CIN85 and CD2AP when interacting with their cellular partners provides the molecular basis for the distinct biological behavior observed between the two adaptors.
Characterization of the interaction between Cbl-b and the SH3A domains of CIN85 and CD2AP using ITC and NMR
ITC titration of CIN85-SH3A with the Cbl-b peptide shows an exothermic binding thermogram. Stoichiometry of the complex (n), dissociation constant Kd, and thermodynamic parameters binding entalpy (ΔH), entropy (ΔS) and Gibbs energy (ΔG) were obtained by analyzing the data using a model of n identical and independent sites (Table S1). The value of n obtained in the titration of CIN85-SH3A with Cbl-b peptide was close to 0.5. In order to check the stoichiometry and the suitability of the binding model, we also performed the reverse experiment, obtaining a value of n close to 2.0, in agreement with the direct experiment. These results suggest a 2 : 1 stoichiometry, whereby two molecules of CIN85-SH3A interact simultaneously with one molecule of Cbl-b, in accordance with previous ITC data, in vivo experiments and the crystal structure of the trimeric complex  (Fig. 1). To further investigate and confirm our results, we followed the binding of the CIN85-SH3A domain to Cbl-b by NMR. Titrations of 15N-labeled CIN85-SH3A with increasing amounts of Cbl-b peptide at 25 °C caused a selective shift of amide proton and nitrogen resonances of several SH3A residues (Fig. 2A), indicating a specific union between the protein and the peptide. Most of the residues in the binding site are in a fast-to-intermediate exchange regime. The residues of CIN85-SH3A that are more affected by binding Cbl-b are in the RT loop, the n-Src loop, the β-III, β-IV and the 310 helix (blue bars in Fig. 2B,E). Interestingly, during the NMR titration, some residues (e.g. W36 and F52) show an unusual curvature. Such curvature is most pronounced for the 1H–15N shifts of the W36 indole Nε–Hε moiety (Fig. 3E, red contours). This tryptophan is conserved in all SH3 domains as it plays a crucial role in binding polyproline stretches. The indole moiety of this tryptophan is thus an excellent probe for monitoring changes that take place in the binding interface (Fig. 1).
The presence of curvature in the HSQC titration is indicative of two or more binding events . Such curvature was also observed previously for the SH3A domain of CIN85, and was explained by the simultaneous formation of two dimers (type I and type II) rather than a trimer . However, no experimental evidence was reported regarding the relationship between the directionality of the curvature and the type of interaction that take place in solution. Moreover, in our titrations, the W36 indole signal broadens out during the intermediate steps of the titration, indicative of exchange between different conformational, possibly oligomeric, states.
The crystal structure of the trimeric complex  indicates that arginines 904 and 911 in Cbl-b are crucial for the type I and II interactions, respectively. If the curvature is indeed due to simultaneous binding in the type I and II orientations, elimination of one or the other arginine should result in a linear shift change indicative of simple, one-site, type I or type II binding . We designed two mutants to validate this hypothesis, Cbl-bR904G to eliminate the type I interaction (as there is a glycine at the equivalent position in CD2), and Cbl-bR911A to eliminate the type II interaction. Titrations of CIN85-SH3A with these mutant peptides were monitored by both ITC (Fig. 3B,D, respectively) and NMR. Figure 3E shows that, in both cases, the curvature of the W36 indole Nε–Hε peak disappears during the NMR titrations; the isolated type II interaction (green contours) shows a linear downfield shift change, while the isolated type I interaction (blue contours) shows a linear upfield shift change of the same W36 indole. The linearity of the shift change is related to a single binding event , and our results indicate that the directionality of the shift change is related to the formation (downfield shift change for type II) or not (upfield change for type I) of a hydrogen bond involving the tryptophan indole proton. Indeed, inspection of the trimeric CIN85-SH3A:Cbl-b structure (PDB entry 2BZ8), shows that the W36 Nε–Hε moiety is involved in a hydrogen bond with the carboxyl group of R909 in Cbl-b (type II orientation), while the distance between W36 Nε-Hε moiety and the carboxyl group of R905 is too long to form a hydrogen bond in the type I orientation.
The expected one-site binding was also confirmed by ITC (Fig. 3B,D), with one-to-one stoichiometry in both cases (Table S1). The Kd of 2.0 μm for the type II interaction compared to 46.9 μm for the type I interaction indicates preferential type II binding of the Cbl-b peptide by CIN85-SH3A.
Cbl-b contains another basic residue, K907, close to the R904 that is involved in the type I interaction with the acidic pocket of the SH3A domain . To validate its importance in type I binding and its contribution to the curvature, we mutated K907 to a leucine (as in CD2). Titrations of CIN85-SH3A with the Cbl-bK907L mutant followed by ITC and NMR show a similar pattern to wild-type Cbl-b. Curvature of the Nε–Hε indole moiety of W36 was also observed for this mutant (Fig. S1A), supporting a major role for R904 in the type I interaction. The chemical shift changes of the W36 side chain during the titration are displaced slightly more downfield than in the case of wild-type Cbl-b, indicating a probably higher contribution of type II interactions. These small differences are probably due to loss of the hydrogen bond between K907 of Cbl-b and Y10 of the CIN85-SH3A in the trimeric complex, thus destabilizing the type I interaction.
To understand the molecular details of the various members of the CIN85/CD2AP family of adaptor proteins, we also performed the study of the interaction between the peptide Cbl-b and the SH3A domain of protein CD2AP. The ITC titration of CD2AP-SH3A with the peptide Cbl-b (Fig. 4C) gave a stoichiometry of 1 : 1, with a Kd of 0.19 μm (Table S2), which indicates formation of a heterodimeric complex. NMR titrations of 15N-labeled CD2AP-SH3A with increasing amounts of Cbl-b peptide at 25 °C caused a selective shift of 1HN and 15N resonances for several SH3A residues (yellow bars in Fig. 2C,E). Most of the residues in the binding site are in a fast-to-intermediate exchange regime. However, the beginning of the RT loop in CD2AP-SH3A in particular is less affected (Fig. 2F,G) compared to CIN85-SH3A. Interestingly, the NMR titration shows again the unusual curvature of the Nε–Hε moiety of W37 (equivalent to W36 in CIN85-SH3A; Fig. 4E, red contours). The presence of such curvature suggests the existence of two binding sites as for CIN85-SH3A. However, in the case of CD2AP-SH3A, we did not observe extensive line broadening during the titration, indicating the absence of extensive conformational exchange phenomena, in agreement with the ITC titrations that show a 1 : 1 stoichiometry. All these features indicate that, in solution, the predominant complexes are dimers formed by interaction of one molecule of CD2AP-SH3A and one molecule of Cbl-b arranged either in the type I or type II orientation instead of a trimeric state.
In order to understand the origin of the curvature in the 15N-HSQC spectra, we also characterized the interaction between CD2AP-SH3A with the mutants Cbl-bR904G and Cbl-bR911A. As shown in Fig. 4E, the curvature disappears for both mutants. The isolated type II interaction (green contours) shows a linear downfield change for the Nε–Hε shift of W37, and the isolated type I interaction (blue contours) shows a linear upfield change for the same 15N–1H shift. Again the directionality of the shift change (downfield or upfield) is related to the ability of W37 Nε–Hε moiety of CD2AP-SH3A to form a hydrogen bond with Cbl-b. In the type II interaction, W37 Nε–Hε moiety forms a hydrogen bond with R908-CO that is absent in the type I interaction (PDB entry 2J6F).
As in the case of CIN85-SH3A, ITC experiments show one-to-one stoichiometry for both mutants (Fig. 4B,D and Table S2), with a Kd of 0.71 μm for the type II interaction and a Kd of 46.5 μm for the type I interaction. These novel results highlight the crucial role of R904 in the type I interaction of Cbl-b with the SH3A domain of CD2AP, and are in complete agreement with the type I recognition pattern of the SH3A domain of CIN85. As observed for CIN85, NMR titration of CD2AP-SH3A with Cbl-bK907L again showed a slightly downfield-displaced curvature in the 15N-HSQC spectra (Fig. S1B). This indicates that K907 in Cbl-b stabilizes the type I interaction for both SH3A domains.
Characterization of the interaction between CD2 and the SH3A domains of CD2AP and CIN85 using ITC and NMR
The aforementioned results suggest that other targets of CIN85 and CD2AP, such as CD2, that lack an N-terminal positively charged residue, should bind solely as type II with a 1 : 1 stoichiometry. ITC experiments using CD2AP-SH3A with CD2 (Fig. 4A) confirm the expected 1 : 1 stoichiometry, with a Kd of 3 μm (Table S2). To obtain information about the nature of the interactions driving the binding , this experiment was repeated at various temperatures and with various buffers. The enthalpy of binding is constant over a temperature range between 15 and 30 °C, which, under the assumption of a linear relationship with the heat capacity change (ΔCp), leads to a ΔCp equal to zero. This suggests that the peptide undergoes very little if any structuring upon binding, and that there is a negligible contribution from the de-solvation to the binding process [14, 15]. Moreover, no protonation/deprotonation processes are coupled to the binding, as the binding enthalpy is independent of the ionization enthalpy of the buffer in which the reaction takes place, as may be concluded from ITC titrations performed in buffers with different protonation enthalpies (data not shown).
The NMR titration followed by 15N-HSQC experiments shows a linear downfield shift change for the W37 indole Nε–Hε moiety (Fig. 4E, black contours), indicative of simple two-state binding . The direction of the shift changes is the same as for the Cbl-bR904G mutant that was used to eliminate type I binding, and is in full agreement with formation of a strong hydrogen bond between the W37 Nε–Hε moiety of CD2AP-SH3A and the R330-CO of CD2 (PDB entry 2J7I). These results thus confirm our hypothesis that peptides without a positively charged residue at the N-terminal position with respect to the PxxxPR motif show a severely compromised interaction in type I mode with an SH3 domain, and that the directionality of the downfield change is associated with type II interaction. To confirm the type II binding mode, a 2D NOESY spectrum of the CD2AP-SH3A:CD2 peptide complex was recorded and compared to the free CD2AP-SH3A spectrum (Fig. 5). This shows that, in the complex, additional NOEs involving the indole He1 proton of W37 may be assigned to protons in the CD2 peptide belonging to R330, P331 and R332. These inter-molecular NOEs may only be explained in the crystal structure by binding of the SH3A domain to the peptide CD2 in type II mode, where all the protons are within 4.5 Å distance of the W37 Nε–Hε proton. In the case of a type I binding, most of these protons, except those of R330, are as far as 6.5–9.5 Å away, and are thus not observable in a NOESY spectrum.
NMR titration of CD2AP-SH3A with CD2 shows that the residues that are more affected (green bars in Fig. 2C) are exactly the same as those during titration with Cbl-b (yellow bars in Fig. 2C). This suggests that, at the end of the titration in both cases, a dimeric type II complex is the predominant form. These results are in accordance with the same NOESY patterns found at the end of every titration (data not shown). Moreover, eliminating the arginine responsible for the type II interaction by mutating it into an alanine (mutant CD2-R332A) shows a very small linear downfield shift change of the W37 Nε–Hε moiety upon addition of a large excess of peptide, indicating very weak residual type II binding (data not shown). A Kd of more than 2 mm was estimated from these chemical shift changes.
We also studied the interaction of CIN85-SH3A with CD2. As for CD2AP, ITC titration (Fig. 3A) showed a 1 : 1 stoichiometry, with a Kd of 9.9 μm (Table S1). NMR titration of CIN85-SH3A with CD2 shows that the residues that are more affected (red bars in Fig. 2B,D) are the same as those during titration with the Cbl-b peptide (blue bars in Fig. 2B). This suggests that, at the end of the titration, in both cases, the predominant form is a dimeric type II complex. These results are in accordance with the closely similar NOE patterns in the 2D NOESY spectra recorded at the end of the titrations of CIN85-SH3A with CD2 and Cbl-b peptides. Moreover, the NMR titration with CD2 showed a linear downfield shift change of the W36 Nε–Hε moiety (Fig. 3E, black contours), indicative of formation of a dimeric complex in a type II orientation. This type II interaction was also confirmed using the mutant CD2-R332A in NMR titrations. Again a small linear downfield shift change of the W36 Nε–Hε moiety was observed with a high excess of CD2-R332A, indicating residual type II binding (with a Kd of more than 4 mm as estimated from the chemical shift changes). All these results show that both the CIN85 and CD2AP SH3A domains interact with CD2, predominantly in the type II orientation, in contrast with the Cbl-b peptide, which is able to interact with both SH3A domains in type I and type II orientations (Figs 3 and 4).
Structural characterization of the interaction between the SH3A domains of CD2AP and CIN85 with CD2 and Cbl-b using SAXS
The results obtained from the combination of ITC and NMR data support strongly a scenario where CD2AP-SH3A forms only heterodimeric complexes with CD2 and Cbl-b, whereas CIN85-SH3A forms both heterodimers (with CD2) and heterotrimers (with Cbl-b). We recorded scattering curves (Fig. 6A) of various SH3:peptide complexes at various ratios to correlate the features observed in the NMR titrations with direct structural changes in the complexes. We assumed the presence of three major species present in detectable amounts at equilibrium throughout the titration. These species consist of monomers, heterodimers (formed in type I or type II orientations) and heterotrimeric complexes where two SH3 domains sandwich a peptide. We then used the program OLIGOMER , which fits any experimental SAXS curve by linear combination of a number of input scattering curves. This allows monitoring of the evolution of the volume fractions of individual species present at various stages of the titration.
We selected the equivalent of three or four stages of the NMR titration (starting point with zero concentration of peptide, a middle point with a 1 : 0.5 SH3:peptide ratio, the equimolar point with a 1 : 1 SH3:peptide ratio, and a point above saturation with a 1 : 1.6 SH3:peptide ratio) to assess the relative composition of the populations in solution. Figure 6, Figures S2 and S3, and Table S3 show the analysis of the experimental SAXS curves. In the case of the interaction of CD2AP-SH3A with CD2 and Cbl-b, the SAXS data confirm the overwhelming presence of dimeric complexes as the main species in solution, with only traces of trimers. CIN85-SH3A, on the other hand, shows a different behavior. It interacts with CD2 to form mainly dimers in solution; however, at half the equimolar point and above, trimers account for approximately 10% of the population (Fig. 6 and Table S3). With Cbl-b, the situation is even more unusual. The experimental scattering curves may only be explained if we consider the presence of a trimeric complex in significantly large amounts in equilibrium with dimeric complexes. Figure 6 shows that the trimeric species dominates in solution at a 1 : 0.5 (SH3:peptide) molar ratio, and decreases to approximately half of the dimeric population above the equimolar point. Similar behavior was found for interaction of CIN85-SH3A with the Cbl-bK907L peptide (Fig. 6 and Fig. S2), where, despite the mutation of K907, the scattering curves still indicate formation of a trimeric complex, in agreement with the ITC and NMR results, confirming the key role of R904 in the type I interaction.
Ab initio shape reconstruction (Fig. 6B) of the various species provides additional support for our description of the populations in solution as obtained using OLIGOMER . For modeling, we used the program DAMMIF  without constraints imposed on the simulated annealing protocol. All the models were very reproducible in independent runs, with mean normalized spatial discrepancy (NSD) values below 1.0. These models are highly consistent with the pseudo-atomic models of free SH3A, dimers and trimers used in our calculations, and validate the presence of trimers in solution (Fig. 6B,C).
Receptor tyrosine kinases control many important biological processes and therefore need to be tightly regulated. Down-regulation of the receptor tyrosine kinases is essential to normal cell function, and the consequences of any failure are frequently associated with severe diseases such as cancer . Key to this down-regulation is the formation of large protein clusters, in which CIN85 and CD2AP play an important role by recruiting several ligand molecules. Both adaptors contain three SH3 domains that recognize atypical polyproline/arginine binding motifs that are present in Cbl-b/c-Cbl and CD2. Cbl-b/c-Cbl was shown to simultaneously contact two N-terminal CIN85/CMS SH3 domains, providing additional molecular functions to the ubiquitin ligase . The stoichiometry of the complexes formed has been controversial for Cbl and CIN85-SH3A [9, 10, 12]. Using several complementary biophysical techniques, we aimed to fully characterize the interactions established by the family of adaptors, including both the CIN85 and CD2AP N-terminal SH3 domains (SH3A), with two cellular ligands, Cbl-b and CD2.
Atypical polyproline recognition: the role of arginine residues for type I and type II binding
The CIN85/CD2AP family of adaptor proteins does not bind to canonical polyproline sequences, but they have been shown to recognize an atypical motif PxxxPR [3-5, 7]. The C-terminal arginine drives the type II interaction in the SH3–peptide complex. Based on our results, the polyproline sequence should be extended to include the arginine responsible for the type I interaction (as present in Cbl-b) to R–2x–1P0x1x2x3P4R5 (Fig. 1). We use this numbering to discuss the contributions of the various residues in binding to the SH3A domains of CIN85 and CD2AP.
ITC, NMR and SAXS experiments (Fig. 3A,E, black contours, and Fig. 6C) show that the SH3A domain of CIN85 is able to interact with the atypical proline sequence of the peptide CD2, predominantly forming a type II dimer as it lacks an arginine at position R–2. The strong downfield shift of the W36 Nε–Hε moiety in the NMR titration (black arrow in Fig. 3E) is indicative of formation of a hydrogen bond with R330-CO in the type II orientation, similarly to the interaction of CD2AP-SH3A with CD2 (PDB entry 2J7I).
In the case of peptides that present the extended atypical proline-rich sequence, including R–2, such as Cbl-b, the interaction with the SH3A domain involves both type II and type I modes of binding. Using ITC, we determined a 2 : 1 stoichiometry for CIN85-SH3A with the Cbl-b peptide, indicating formation of a trimeric structure in solution, consisting of a Cbl-b peptide sandwiched by two CIN85-SH3A domains, as shown in the crystal structure. Previous studies reported that formation of the heterotrimer was due to a crystallization artifact and disputed the formation of such complex in solution [9, 10, 12]. In our work, we have resolved the supposed controversy as the existence of a trimeric complex is in agreement with our results from ITC, NMR and the ab initio reconstructed shape obtained from analysis of the SAXS curves. In support of this, heterotrimerization was also shown in vivo . This formation is not an exception as ligand-induced SH3 multimerization has now been observed in CIN85, β-PIX (PAK-interacting exchange factor) , CTTN (Cortactin)  and Fyn-SH3 (a Src family tyrosine kinase) .
Mutants Cbl-bR911A and Cbl-bR904G, designed to eliminate type II and type I interactions respectively, highlight the importance of the N-terminal (R–2) and C-terminal (R5) arginines for formation of the trimeric complex (Fig. 1). As expected, the ITC titrations indicate that both Cbl-bR904G and Cbl-bR911A preclude formation of trimers, forming type II and type I dimers, respectively, with CIN85-SH3A. In accordance, when analyzing the NMR signals of CIN85-SH3A W36 indole 15N-1H cross peaks during the Cbl-b titrations, linear shift changes were observed during titration with both Cbl-b mutants instead of the curvature observed with wild-type Cbl-b (Fig. 3). The directionality of the chemical shift change of the indole proton of the tryptophan may be correlated with the orientation of the peptide, indicating the interaction type taking place: the downfield change (black arrow in Fig. 3E) is related to formation of hydrogen bonds and is indicative of a type II interaction in the SH3 domains. However, the upfield shift change (blue arrow in Fig. 3E) is related to unfavorable distances for hydrogen bond formation, and is representative of the type I interaction of SH3 domains. These chemical shift perturbation profiles may thus be used to designate the type I or type II orientation of other studied targets for the N-terminal and most likely the other SH3 domains of both CIN85 and CD2AP. Figure 3E shows an unusual curvature and a strong decrease in intensity during titration of wild-type Cbl-b. The curvature points to the occurrence of both type I and II interactions, while the strong broadening of NMR signals indicates extensive exchange between multiple conformers, supporting the formation of possibly type I and type II dimers but also a heterotrimeric complex between CIN85-SH3A and Cbl-b.
To completely characterize the versatile interactions of the CIN85/CD2AP family of adaptor proteins, we also studied the interaction of the SH3A domain of CD2AP with the peptides CD2 and Cbl-b, as we did with CIN85-SH3A.
ITC (Fig. 4A) and SAXS (Fig. 6) experiments show that the SH3A domain of CD2AP interacts with the atypical proline-sequence of peptide CD2, forming dimers. Using NMR, the observation of specific inter-molecular NOEs between the tryptophan indole proton and several side-chain protons in CD2 may only be rationalized by type II binding (Fig. 5). Eliminating the electrostatic interactions of R332 (R5) with the negatively charged glutamate residues in the RT loops of the SH3 domains by mutating it to an alanine almost completely abolishes the binding (the Kd is three orders of magnitude greater for the R332A mutant). This indicates that these electrostatic interactions are the main driving force for SH3 binding to CD2. It also confirms that the lack of a positively charged residue in the N-terminal part of the CD2 peptide at position R–2 results in prevalent type II binding. This interaction is enthalpically driven as shown by ITC, with a favorable negative enthalpy change that compensates for the positive entropy change (Fig. S4, and Tables S1 and S2). The affinity is higher for the CD2AP-SH3A domain mainly due to greater enthalpy, although the interaction in the case of CIN85-SH3A domain has a lower entropic penalty, indicative of a more flexible interaction (Fig. S4, and Tables S1 and S2).
Interestingly, the interaction of the SH3A domain of CD2AP with Cbl-b, which presents the extended atypical proline-rich sequence, includes formation of type I and type II dimers, as confirmed by the 1 : 1 stoichiometry of ITC experiments and the ab initio shape of SAXS. Crystallographic studies on CMS-SH3 with Cbl-b  showed a Cbl-b peptide sandwiched between two CMS-SH3 domains simultaneously interacting in type II and type I modes (PDB entry 2J6F). However, the type I interaction involved K907 instead of R904.
During NMR titration of the SH3A domain of CD2AP with Cbl-b, the signals of the W37 indole 15N-1H cross peaks (Fig. 4E) show a curvature, like during titration of CIN85-SH3A. However, despite the fact that the affinity is one order of magnitude higher, in this case there is no broadening of the signals, indicating the absence of extensive conformational exchange phenomena.
To further characterize the differences between the curvatures that we observed in the NMR titrations of CD2AP and CIN85 with Cbl-b, we tried to simulate the curvature observed during the titration with wild-type Cbl-b by using the Kd values resulting from fitting of chemical shift changes of the tryptophan indole Nε–Hε moiety from NMR titration of the isolated type I interaction (with Cbl-bR911A) and type II interaction (with Cbl-bR904G). The Kd values were used to calculate the fraction of free SH3 and bound type I and type II dimers at every point of the titration (see 'Results'). We then fitted the chemical shift changes of the tryptophan indole Hε1 of the SH3A domain of CD2AP and CIN85 during titration with wild-type and mutant Cbl-b. As a proof of principle, the experimental data for the isolated type II and type I interactions may be perfectly simulated (Fig. 7: green, type II; blue, type I) for both CIN85 and CD2AP-SH3A domains. Using the experimental Kd values from isolated type I and type II interactions, and considering the existence of two independent type I and type II dimers in solution, the titration of CD2AP-SH3A with wild-type Cbl-b may be simulated perfectly (Fig. 7B, red line). However, the titration of CIN85-SH3A with wild-type Cbl-b cannot be simulated using this approach, especially at a low concentration of Cbl-b (Fig. 7A, red line). This indicates that, in addition to the formation of type I and type II dimers, other multimeric states (e.g. trimers) exist in solution. These simulations are thus in agreement with the stoichiometries determined by ITC and the SAXS analysis. In the case of CIN85, at a low concentration of Cbl-b, the trimeric state is highly populated, and when the concentration of Cbl-b is higher than that of the protein, the principle species in solution are type II dimers, as reflected by the W36 indole chemical shift approaching that of purely type II interactions at the end of the titration (Fig. 7). Thus, under saturating ligand conditions, the main species are type II heterodimers. This is in full agreement with the similar inter-molecular NOE patterns involving the tryptophan indole He1 proton, the higher affinity for the type II interaction (Tables S1 and S2), and the ratios obtained from fitting the SAXS curves (Fig. 6C and Table S3). The fact that the mechanisms of clustering are different at different concentrations of peptide may have important biological implications.
SH3A domains of CD2AP and CIN85 have different decoding mechanisms: biological implications
Even though the SH3A domain of CD2AP is able to interact with Cbl-b, forming type I and type II dimers, it is unable to form a trimeric complex involving Cbl-b R904 and R911 simultaneously, as for the SH3A domain of CIN85. This may be rationalized by superimposing its structure on CIN85-SH3A in the X-ray structure of the trimeric complex (Fig. 8). The length and structural differences of the n-Src loop of the SH3A domains of CIN85 and CD2AP (Fig. S5) are responsible for the ability to form a trimer (Fig. 8); residue E35 of one CD2AP-SH3A molecule would clash with Y8 of the other SH3 molecule, thus preventing formation of a trimer with Cbl-b. Moreover, R909 (x3) of the Cbl-b peptide is unable to take up the same orientation as in the CIN85-SH3A:Cbl-b trimeric complex. Therefore, only a trimeric arrangement like the one reported in the CMS/Cbl-b crystal structure , involving K907 (x1), is possible. Thus, it may be that CD2AP is able to form type II dimers with Cbl-b R911 (R5), type I dimers with R904 (R-2) like CIN85-SH3A, and heterotrimers involving R911 (R5) in the type II orientation and K907 (x1) in the type I orientation. This hypothesis is supported by the observation that the affinity of isolated type I binding (Cbl-bR911A) is exactly the same for the CD2AP and CIN85-SH3A domains (see Tables S1 and S2), indicating a very similar interaction. According to our results, it appears that, for the SH3A domain of CD2AP, the individual dimers are more stable than a trimeric arrangement, although there are clear differences in experimental conditions with respect to the crystal structures and the in vivo assays .
We conclude that whether the SH3 domains of this family of adaptor proteins interact in type I or type II mode is determined by the sequence of the ligand, in particular the pseudosymmetry of positive amino acids capable of interacting with the hydrophilic pocket in the SH3 domain. However, the formation of a trimeric complex is related to the structural differences between the two adaptor SH3 domains, especially in the n-Src loop, which determine whether the trimeric complex is sterically feasible.
In the case of CIN85-SH3:Cbl-b interactions, combined use of ITC, NMR and SAXS has been crucial to clarify the previous discrepancy regarding the recognition mode of the SH3A domains, and to identify novel aspects of the interaction mechanisms of these domains with atypical proline-rich sequences, providing the structural basis for the signal transduction pathways involving these two adaptor proteins. Although a contribution to protein multimerization has only been reported for SH3A domains [5, 8], we cannot exclude the possibility that the other two SH3 domains (SH3B and SH3C) may also participate in a similar manner, as they have been shown to recognize similar partners [2, 3, 12]. Further studies are required to understand how the three SH3 domains of CD2AP and CIN85 adaptor proteins cooperate to bind a single ligand.
Despite the high degree of sequence conservation and domain organization between the two adaptor molecules and many overlapping molecular functions in cultured cells, investigations with mice suggest that CIN85 and CD2AP may have distinct functions in vivo [21, 22]. This distinct biological behavior may be explained, at least in part, by the differential mode of recognition observed between the two adaptors when interacting with their cellular partners. These properties of CIN85 and CD2AP contribute to the existence of another level of organization within the molecular clustering required for down-regulation and signaling.
The protein construct of the SH3A domain of CD2AP was cloned into the pET21a vector (Novagen, Billerica, MA, USA), and the N-terminal domain from human CIN85 was cloned into plasmid pET28a (Novagen, Billerica, MA, USA, and modified with a PreScission protease cleavage site provided by Dr. R. Campos, CNIO, Madrid, Spain) containing the SH3A gene covalently linked to an N-terminal 6xHis tag and a PreScission protease cleavage site. Both constructs were verified by DNA sequencing.
SH3A of CD2AP was over-expressed without the His tag in the Escherichia coli Rosetta (DE3) strain at 37 °C. SH3A of CIN85 was over-expressed with the His tag in E. coli strain BL21 (DE3) at 37 °C. In both cases, protein expression was induced using 1 mm isopropyl thio-β-d-galactoside at 37 °C over 18 h. The purification protocols for the SH3A domains of labeled and unlabeled CD2AP without the His tag and unlabeled CIN85 with the His tag have been described previously [5, 23]. Isotope-enriched samples of the SH3A domain of CIN85 with the His tag were obtained by growing the bacteria in M9 minimal medium supplemented with 15N-labeled ammonium chloride and 13C-enriched glucose (Cambridge Isotope Laboratories, USA & Cortecnet, France). After overexpression, cells were centrifuged at 5180 g for 15 min. Protein was expressed in inclusion bodies. To break the inclusion bodies, the pellet was dissolved and homogenized in 50 mm Tris buffer (pH 8.0), 400 mm NaCl, 20 mm imidazole with 8 m urea. After sonication, the lysate was ultracentrifuged at 48 300 g, and the supernatant was passed through a Ni–Sepharose resin (GE Healthcare, Little Chalfont, Buckinghamshire, UK) equilibrated with the same buffer. The protein was eluted using an imidazole gradient (from 20 to 500 mm) in the presence of 8 m urea. Afterwards, the protein with the His tag was refolded by successive dialysis in buffers containing 3, 1.5 urea and 0 m urea. The His tag was cleaved by incubation with PreScission protease (GE Healthcare) (5 units·mg−1) at room temperature over 16 h. Additional purification was performed by gel filtration chromatography using a Hiload 16/60 Superdex-75 column (GE Healthcare) equilibrated with 50 mm sodium phosphate buffer, pH 6.0. Protein solutions were concentrated using an Amicon ultracentrifugal unit (Millipore, Billerica, MA, USA), and stored at −20 °C after shock freezing in liquid nitrogen.
CD2- and Cbl-b-derived peptides
Peptides containing stretches of 13 residues including the atypical PxxxPR recognition sequence for CD2 (residues 322–334) and Cbl-b (residues 901–913) and mutants (Table 1) with more than 98% purity were purchased from ChinaPeptides Co. Ltd (Shanghai, China) and SynBioSci Co. (San Francisco, CA, USA). A tyrosine residue was added at the N-terminus to ensure accurate determination of the concentration. All peptides are N-terminally acetylated and C-terminally amidated. The purity of the peptides was confirmed by mass spectrometry.
Isothermal titration calorimetry (ITC)
ITC experiments were performed on a VP-ITC instrument (MicroCal Inc., Northampton, MA, USA). The SH3A domains of CD2AP and CIN85 (40–90 μm) were titrated with the various peptides (0.6–1.9 mm) in 50 mm cacodylate buffer, pH 6.0, at 25 °C. Additionally, SH3A of CD2AP (50 μm) was titrated with CD2 peptide (1 mm) in various buffers (50 mm MES and 50 mm cacodylate) at pH 6.0, and at various temperatures (15, 20, 25 and 30 °C). Most titration schemes consisted of a preliminary 3 μL injection (neglected in data analysis), followed by 30 subsequent 6 μL injections. In other cases, we used a variable volume profile titration of 30 injections based on simulations. In general, 6 or 8 min were left between each injection in order to reach equilibrium. Heats of dilution, determined by titrating the peptides (and the SH3A domains, in the case of reverse titrations) into buffer alone, were subtracted from the raw titration data before analysis. Data were fitted by least-squares procedures assuming a model of n identical and independent binding sites [,] using MicroCal Origin version 7.0 (MicroCal Inc., Northampton, MA, USA). All the ITC experiments were repeated three times, and the thermodynamic parameters and the error calculated as the standard deviation from the mean value are reported in Tables S1 and S2 and shown in Fig. S4.
Nuclear magnetic resonance (NMR) spectroscopy
15N-labeled and 15N/13C-labeled samples of the SH3A domains of CD2AP and CIN85 were prepared in a range of concentrations between 0.3 and 1 mm in 93% H2O/7% D2O, 50 mm sodium phosphate at pH 6.0. All NMR spectra were recorded at 25 °C on a Varian (Agilent Technologies, Santa Clara, CA, USA) NMR Direct-Drive Systems 600 or 800 MHz spectrometer, the latter equipped with a salt tolerance triple-resonance PFG-Z cold probe. The backbone resonances of CIN85-SH3A were assigned from a combination of 2D 15N-HSQC and 3D HNCO, HNCACB and CBCACONH spectra . The backbone resonances of CD2AP-SH3A were previously assigned .
15N-labeled SH3A of CD2AP and CIN85 was titrated with increasing amounts of the peptides (stock solution of 6.3 mm) at pH 6.0 and 25 °C. The progress of the titrations was monitored by recording 2D 1H–15N HSQC spectra. The magnitude of the chemical shift perturbation (Δδ) was calculated as described previously .
To evaluate the titration experiments, the 1H–15N chemical shifts of the indole moiety from W36 of CIN85-SH3A and W37 of CD2AP-SH3A were plotted against the ligand concentration and fitted to an equation that takes into account the Hill coefficient for possible cooperative effects:
where [Pt], [Lt], Kd and Bmax are the total protein concentration, ligand concentration, dissociation constant and chemical shift deviation at saturation, respectively, and α is the Hill coefficient. When the Hill coefficient = 1, this equation becomes the standard equation for a single binding model . Origin 7.0 (OriginLab Corporation, Northampton, MA, USA) was used to fit the experimental data with the aforementioned equations in order to obtain the experimental Kd values. The Kd values were used to calculate the fraction of free and bound SH3A domain at each titration point considering that the protein can recognize the two binding sites of the ligand forming heterodimers in an independent way, and taking into account that the most favorable interaction (smaller Kd) occurs first. This analysis was used to simulate the chemical shift perturbations of the tryptophan indole Hε1 of the SH3A domain of CD2AP and CIN85 during the NMR titration with Cbl-b peptide and the mutants Cbl-bR904G and Cbl-bR911A.
2D-NOESY spectra (mixing times of 150 ms) were recorded at the end of every NMR titration to check the inter-molecular NOE pattern of the indole proton of W36 of CIN85-SH3A and W37 of CD2AP-SH3A. For this purpose, the CD2 proton resonances were assigned from a combination of homonuclear 2D 1H–1H TOCSY (mixing time of 20 and 70 ms), 2D 1H–1H ROESY (mixing time of 200 ms) and 2D 1H–1H NOESY (mixing time of 200 ms) experiments. A 0.80 mm CD2 solution was titrated with increasing amounts of unlabeled CD2AP-SH3A domain (stock solution of 5.8 mm). The progress of the titrations was monitored by recording 2D 1H–1H TOCSY (mixing time of 70 ms) and 2D 1H–1H NOESY (mixing time of 200 ms) spectra. Proton chemical shifts for CD2 were assigned at each titration point. All NMR data were processed using NMRPipe  and analyzed by CcpNmr  or NMRViewJ .
Small-angle X-ray scattering (SAXS)
SAXS data for characterizing the various complexes formed between the SH3A domains of CD2AP and CIN85 with the peptides CD2 and Cbl-b were collected at beamline X33 of the Deutsches Elektronen Synchrotron (Hamburg, Germany). The camera length was 2.7 m, and the wavelength was 0.15 nm, with 2 min of exposure time for data collection. The data were averaged, background-subtracted, and merged to generate the scattering curve using PRIMUS . The radius of gyration (Rg) was calculated from Guinier analysis as implemented in PRIMUS (see Fig. S2 for further details), and also from the entire scattering curve using the indirect Fourier transform package GNOM [16, 33]. CRYSOL  was used to compare experimental and theoretical scattering curves (see Fig. S3 for further details). We used OLIGOMER  for ab initio shape reconstruction of the various species. For modeling, we used the program DAMMIN  without constraints imposed on the simulated annealing protocol.
Samples of CD2AP-SH3A were prepared in absence and presence of the peptides CD2 and Cbl-b (1:1 and 1:1.6 ratios for both peptides, one at each time). Samples of CIN85-SH3A were also prepared in absence and presence of CD2 and Cbl-b (1:0.5 and 1:1.6 ratios for both peptides, one at each time), and in presence of Cbl-bK907L in the ratio 1 : 0.5. Table S3 provides more details of the sample preparation. All the samples were measured at four concentrations (7, 5, 3 and 1 mg·mL−1) in 50 mm sodium phosphate buffer at pH 6.0.
We thank Ari Ora (University of Helsinki, Department of Biosciences) and Nayra Cardenes (Simons Center for Interstitial Lung Disease, University of Pittsburgh, PA, USA) for initial help with the production of CD2AP-SH3A and CIN85-SH3A. This research was funded by grant number FQM02838 from the Junta de Andalucía. M.A.C. is supported by a Formación de Profesorado Universitario (FPU) grant from the Ministerio de Educación de España. A.G.-P. is a Fonds Wetenschappelijk Onderzoek post-doc fellow. J.L.O.-R. is a FEBS post-doc fellow. Work in N.A.J.vN.'s laboratory is supported by the Vlaams Instituut voor Biotechnologie and the Flanders Hercules Foundation. Work in J.B.'s laboratory was supported by the Ministerio de Ciencia e Innovación (SAF2009-10667 and SAF2012-31405) and the Generalitat Valenciana (ACOMP/2012/024).