Visualizing protein breathing motions associated with aromatic ring flipping

Aromatic residues cluster in the core of folded proteins, where they stabilize the structure through multiple interactions. Nuclear magnetic resonance (NMR) studies in the 1970s showed that aromatic side chains can undergo ring flips—that is, 180° rotations—despite their role in maintaining the protein fold1–3. It was suggested that large-scale ‘breathing’ motions of the surrounding protein environment would be necessary to accommodate these ring flipping events1. However, the structural details of these motions have remained unclear. Here we uncover the structural rearrangements that accompany ring flipping of a buried tyrosine residue in an SH3 domain. Using NMR, we show that the tyrosine side chain flips to a low-populated, minor state and, through a proteome-wide sequence analysis, we design mutants that stabilize this state, which allows us to capture its high-resolution structure by X-ray crystallography. A void volume is generated around the tyrosine ring during the structural transition between the major and minor state, and this allows fast flipping to take place. Our results provide structural insights into the protein breathing motions that are associated with ring flipping. More generally, our study has implications for protein design and structure prediction by showing how the local protein environment influences amino acid side chain conformations and vice versa. Nuclear magnetic resonance spectroscopy and X-ray crystallography capture how aromatic side chains can flip within the core of a protein via the generation of a local void volume.

Aromatic residues cluster in the core of folded proteins, where they stabilize the structure through multiple interactions. Nuclear magnetic resonance (NMR) studies in the 1970s showed that aromatic side chains can undergo ring flips-that is, 180° rotations-despite their role in maintaining the protein fold [1][2][3] . It was suggested that large-scale 'breathing' motions of the surrounding protein environment would be necessary to accommodate these ring flipping events 1 . However, the structural details of these motions have remained unclear. Here we uncover the structural rearrangements that accompany ring flipping of a buried tyrosine residue in an SH3 domain. Using NMR, we show that the tyrosine side chain flips to a low-populated, minor state and, through a proteome-wide sequence analysis, we design mutants that stabilize this state, which allows us to capture its high-resolution structure by X-ray crystallography. A void volume is generated around the tyrosine ring during the structural transition between the major and minor state, and this allows fast flipping to take place. Our results provide structural insights into the protein breathing motions that are associated with ring flipping. More generally, our study has implications for protein design and structure prediction by showing how the local protein environment influences amino acid side chain conformations and vice versa.
Aromatic residues make up a considerable fraction of the hydrophobic core of folded proteins, where they stabilize the structure through CH-π (refs. [4][5][6], π-π (refs. 7,8 ) and cation-π (refs. 9,10 ) interactions as well as hydrogen bonds. NMR studies in the 1970s demonstrated that aromatic side chains can undergo ring flips-that is, 180° rotations of the χ 2 dihedral angle (Cβ-Cγ axis)-even when engaged in stabilizing interactions in the hydrophobic core [1][2][3] . These ring flips require concerted movements of the surrounding residues (large-amplitude protein breathing motions), and ring flipping rates as a function of temperature and pressure have been used to report on these motions by deriving activation energies and volumes [11][12][13][14][15][16][17][18][19][20][21] . However, the structural details of ring flipping and the associated breathing motions have remained unknown, probably owing to difficulties in stabilizing ring flipping transition states or intermediates that are amenable to structure elucidation.
Here we capture ring flipping events of a buried tyrosine residue in the SH3 domain of the JNK-interacting protein 1 ( JIP1). We show using NMR relaxation dispersion that the aromatic ring of this tyrosine residue populates a minor-state conformation (3%), and we design single point mutations to stabilize this conformation and capture its high-resolution structure using X-ray crystallography. The structure reveals how the intricate network of hydrogen bonds and CH-π interactions is rearranged in the minor state. We show how a substantial void volume is generated around the tyrosine ring during the structural transition from the major to the minor state, which can be associated with the breathing motions that allow fast-timescale ring flipping events to take place. Our results provide structural insights into aromatic ring flipping and its associated protein breathing motions.

Protein dynamics induced by a tyrosine residue
The SH3 domain of JIP1 undergoes exchange between two distinct conformations, as evidenced by 15 [22][23][24][25]. These experiments quantify the kinetics of exchange processes and provide the difference in chemical shift between a major and a minor state, together with their relative populations [26][27][28][29][30] . The data confirm that exchange contributions to the transverse relaxation are present for residues within three regions of the protein (Extended Data Fig. 1d, e, Supplementary Table 1). These residues are located spatially close to tyrosine 526 (Y526) (Extended Data Fig. 1f). A mutation of Y526 to alanine (Y526A) shows no conformational exchange (Extended Data Fig. 1d, e), and conserves the protein backbone conformation, as evidenced from its crystal structure that we obtained at 1.5-Å resolution (Extended Article Data Fig. 3, Extended Data Tables 1, 2). These results show that the relaxation dispersion that affects around 40% of the residues in the SH3 domain arises from a single exchange process, with Y526 being the origin of the observed exchange.
The 15 N and 1 H N relaxation dispersion data were analysed simultaneously according to a two-site exchange model in which a highly populated major state interconverts with a low-populated minor state (Extended Data Fig. 4a, b, Supplementary Discussion). The analysis of the data gives a population of the minor state of p minor = 2.8 ± 0.1% and an exchange rate constant of k EX = 2,600 ± 70 s −1 . The derived chemical shift differences, Δδ CPMG , span a range of 4.7 ppm for 15 N and 1.1 ppm for 1 H N suggesting that there are substantial structural changes between the major and the minor state (Extended Data Fig. 4c, d).
The side chain of Y526 is found in an unusual conformation in the crystal structure (Protein Data Bank (PDB) code 2FPE (ref. 31 ), Extended Data Table 1), characterized by a χ 2 dihedral angle of 2° (Fig. 1a). Normally, this eclipsed conformation is energetically unfavourable because of steric interactions with the backbone (Fig. 1b), and it is rarely found in proteins as χ 2 angles are preferred where Cδ 1 and Cδ 2 are staggered with respect to Cα (ref. 32 ). However, the eclipsed conformation of the aromatic ring of Y526 is stabilized by CH-π interactions from V517, Q520 and A541 and by π-π interactions with H493 (Fig. 1a).

Proteome-wide SH3 sequence analysis
To investigate the contribution from the surrounding residues in stabilizing the eclipsed conformation of Y526, we analysed the sequences of all identified human SH3 domains 33 . We categorized the sequences according to the identity of the amino acid at the position of Y526 in the JIP1 SH3 domain ( JIP1-SH3) (Extended Data Fig. 5a), and we retained the sequences carrying a phenyl-based amino acid (Tyr or Phe) at this position, amounting to 33 SH3 domains. Sequence alignments reveal a large variation in the size of the amino acids at positions 493, 517 and 541, whereas at position 520 most sequences contain Gln, Arg or Lys (72%) (Extended Data Fig. 5b, c). To study the size correlation between the amino acids at positions 493, 517 and 541 and their influence on the conformation of the aromatic residue at position 526, we carried out a principal component analysis (PCA) by assigning a size score (n) to each amino acid according to the number of heavy atoms in their side chains. This analysis reveals two well-separated groups, with the SH3 domain of JIP1 belonging to group 1 (Fig. 1c). In group 2, five SH3 domains are found for which high-resolution crystal structures are available; these include three SH3 domains of the sorbin and SH3 domain-containing proteins 1 and 2 (SORBS1 and SORBS2) 34 , and the SH3 domains of the dedicator of cytokinesis protein 2 (DOCK2) 35 and of the tyrosine protein kinase CSK 36 (Extended Data Fig. 5b). Notably, all group 2 structures show a favourable, staggered side chain conformation (of Cδ 1 /Cδ 2 with respect to Cα) of the corresponding tyrosine, with the χ 2 dihedral angle ranging from −40° to −64° (Fig. 1c). We therefore hypothesized that SH3 domains of group 1 have eclipsed conformations, whereas group 2 have staggered conformations. To test our hypothesis, we determined two crystal structures of SH3 domains of the scaffold protein POSH ('plenty of SH3 domains'), for which the first SH3 domain belongs to group 2 and the fourth SH3 domain belongs to group 1 (Extended Data Table 2, Fig. 1c). Consistent with the PCA, the crystal structure of the first SH3 domain of POSH (POSH-SH3-1) shows a staggered conformation of the corresponding tyrosine residue (Y172), which is stabilized by CH-π interactions from L163 (position 517) and T185 (position 541) (Fig. 1d, e). POSH-SH3-4 shows a similar structure to JIP1-SH3, with an eclipsed conformation of the corresponding phenylalanine residue (F867) stabilized by CH-π interactions from V858 (position 517), K861 (position 520) and G882 (position 541), and by π-π interactions with H834 (position 493) (Fig. 1f, g). Our data therefore suggest that the conformation of the aromatic ring at position 526 is determined by steric interactions dictated by the size of the surrounding amino acids.

Structure of the minor state
Next, we investigated whether the minor state detected by NMR corresponds to a staggered conformation of the side chain of Y526. We sought to stabilize the minor state relative to the major state by   Dashed lines indicate CH-π (black) and π-π (red) interactions. b, Side chain conformation of Y526 in JIP1-SH3, illustrating the steric interactions between the δ 1 nuclei of the aromatic ring and the backbone. c, PCA of a dataset comprising the size of the amino acid side chains at positions 493, 517 and 541 within SH3 domains that contain Y or F at position 526 (Extended Data Fig. 5b). Two groups are observed, which correspond to eclipsed (group 1) or staggered (group 2) conformations of the aromatic ring. JIP1-SH3 is indicated in blue; POSH-SH3-1 and POSH-SH3-4 are indicated in green; and SH3 domains for which crystal structures have been determined previously are shown in red. d, Crystal structure of POSH-SH3-1, showing a staggered conformation of Y172. e, Unbiased electron density maps (Fo-Fc) of Y172 and the surrounding residues in POSH-SH3-1. f, Crystal structure of POSH-SH3-4, showing an eclipsed conformation of F867. g, Unbiased electron density maps (Fo-Fc) of F867 and the surrounding residues in POSH-SH3-4. h, i, Results of the PCA, illustrating the size and nature of the residues that surround the aromatic ring in position 526 in group 1 (h) and group 2 (i) SH3 domains. The size of the spheres in each position is proportional to the average size (n av ) of the amino acid side chain across group members.
introducing single point mutations. The PCA suggests that a staggered conformation in group 2 is favoured over an eclipsed conformation in group 1 when residues with larger side chains are found in position 541, when isoleucine or leucine are occupying position 517 and when smaller residues are found in position 493 (Fig. 1h, i). Accordingly, we designed four different mutants of JIP1-SH3 (H493A, V517A, V517L and A541L) and obtained their high-resolution structures by X-ray crystallography (Extended Data Tables 1, 2). Of note, three mutants (H493A, V517A and A541L) induce a staggered conformation of Y526, with χ 2 dihedral angles ranging from −41° to −75°, whereas V517L shows an eclipsed conformation of Y526 and an almost identical structure to the wild-type protein (Fig. 2, Extended Data Fig. 6a). The high resolution of these structures, ranging from 1.4 to 1.9 Å, allows unambiguous determination of the conformation of Y526 and the surrounding residues, as demonstrated by their unbiased electron density maps (Fig. 2).

Structural details of breathing motions
The two variants H493A and V517A show almost identical crystal structures (Extended Data Fig. 6b) and an equivalent stabilization mechanism of the aromatic ring of Y526. Whereas the wild-type structure exhibits a classic β-bulge at residue 518 (ref. 37 ), the transition from the eclipsed to the staggered conformation induces a local inversion (in-out) at residues 518-520, which leads to the formation of a canonical β-strand, as observed in the structures of the H493A and V517A variants ( Fig. 3a-c). This transition allows the side chain of L519 to rearrange and form CH-π interactions with the ring of Y526 (Fig. 2c, d); and, at the same time, large-scale movements of E518, Q520, E522 and Y524 are observed (Fig. 3a). We note that SH3 domains in both group 1 and group 2 of the PCA show classic β-bulges at position 518 (Extended Data Fig. 5d), which suggests that the presence of this structural motif is not determinant of the side chain conformation of the phenyl ring in position 526.
The A541L mutation also triggers a staggered conformation of Y526 and a rearrangement of the 517-522 region; however, the stabilization mechanism of the aromatic ring is different. A looping out of the β-strand between residues 517 and 522 is observed (Extended Data Fig. 6c-e), which allows the side chain of A521 to reorient and to stabilize the staggered conformation of the ring through CH-π interactions together with L541 and V517 (Fig. 2b). At the same time, the side chains of L519, Q520 and E522 and D523 undergo large-scale movements to accommodate the flipped ring within the pocket (Extended Data Fig. 6c).
Altogether, the different mutants show that the dynamics of the region encompassing residues 517-522 are key for the formation of the minor state. The experimental 13 C chemical shifts for residues in this region are characteristic of random coil conformations (Extended Data Fig. 1b) and, compared to other regions of secondary structure, the relaxation-derived order parameters (S 2 ) are lower (Extended Data Fig. 2g) and the crystallographic B-factors are higher. This supports the idea of the 517-522 region being intrinsically dynamic, with a fluctuating hydrogen-bonding network that is prone to structural transitions.
Next, we sought to determine which of the two crystal structures (H493A/V517A-like or A541L-like) best capture the conformation of the wild-type minor state detected by NMR relaxation dispersion. The H493A and V517A variants show almost identical crystal structures and for a subset of residues, the chemical shifts of which are unaffected by the mutations, the resonances of the two variants fall on a straight line together with the resonances of the wild-type protein (Fig. 3d, e). This suggests that they are in fast-intermediate exchange between two conformations represented by the crystal structures of the wild-type protein and of the H493A/V517A variants. In agreement with this, both variants show line broadening and chemical exchange contributions as detected by 15 N and 1 H N relaxation dispersion (Extended Data Figs. 7a-c, 8a-c). Analysis of these data (Extended Data Figs. 7d, e, 8d, e) shows that the structural features of the minor state of the wild-type protein are captured by the H493A/V517A crystal structures, as shown by the excellent agreement between the Δδ CPMG values for the wild-type protein and the two variants (Extended Data Figs. 7f, g, 8f, g). In addition, the analysis yields exchange rates between the staggered (canonical β-strand) and the eclipsed (classic β-bulge) conformation of k EX = 2,830 ± 70 s −1 (H493A) and k EX = 6,800 ± 300 s −1 (V517A), compared to k EX = 2,600 ± 70 s −1 determined for the wild-type protein. Finally, the observable chemical shifts of the two variants (Fig. 3d, e), in conjunction with analysis of the relaxation dispersion data (Supplementary Discussion), show that the H493A mutation slightly stabilizes the minor state relatively to the major state, whereas the V517A mutation almost inverts the relative populations of the major and minor states (Fig. 3f). For completeness, we note that the A541L crystal structure is not representative of the minor state conformation (Extended Data Fig. 9), although it shares structural features that are necessary for fast ring flipping of Y526 (see below).

Void volume enables ring flipping
Aromatic 1 H-13 C heteronuclear single quantum coherence (HSQC) spectra show averaging of the NMR signals of the tyrosine ε 1 /ε 2 nuclei of Y526 (Fig. 4a, Extended Data Fig. 10a, Supplementary  Fig. 1), which shows that full 180° ring flipping occurs. This poses the question of the timescale of the full ring flipping and its relation to the observed minor state. To answer this question, we acquired L-optimized TROSY-selected aromatic side chain 13 Cε CPMG (ref. 38 ) and on-resonance R 1ρ (ref. 39 ) relaxation dispersion data of Y526. These data are entirely explained by the exchange process between the major and minor state, with a negligible contribution from the V517L Q520 V517  Article full ring flipping event (Fig. 4b, c), demonstrating that ring flipping of Y526 is fast (k EX > 50,000 s −1 ) (Supplementary Discussion). This observation agrees with a 1-μs molecular dynamics (MD) simulation that shows several 180° ring flipping events of Y526 (Fig. 4d, Extended Data Fig. 10b, c). During the structural transition between the major and the minor state (Supplementary Video 1), a void volume is created around the ring of Y526 that corresponds to a pocket expansion of 65 Å 3 ; this is mainly due to the structural reorganization of the side chain of Q520 (Fig. 4e, f). This cavity expansion is in agreement with previous studies that have reported activation volumes between 40 and 85 Å 3 for ring flipping events of aromatic residues in other proteins 12,14,21,40,41 . The expansion is followed by a compaction of the surrounding protein environment as the ring becomes stabilized by CH-π interactions from L519 (Fig. 4e, f).
Collectively, our results are consistent with a model in which fast protein breathing motions along the structural trajectory between the major and the minor state generate the necessary void volume for ring flipping to take place by lowering the energy of the transition state ( Fig. 4e-g, Supplementary Video 2). Occasionally, the β-bulge to β-sheet transition is completed and the aromatic ring becomes trapped in a staggered conformation that is stabilized by CH-π interactions with L519-a process that gives rise to the observed relaxation dispersion. These events are rare and occur on a slow timescale (Figs. 3f, 4g), but they constitute an important tool for observing the trajectory of the protein breathing motions coupled to aromatic ring flipping. The initial generation of void volume around the ring is almost identical along the structural trajectory between the major state (wild type) and the A541L crystal structure (Extended Data Fig. 6f). Thus, all mutants-including A541L, which stabilizes Y526 in a staggered conformation by a different mechanism-share the same initial structural trajectory and report on identical breathing motions.

Conclusions
Our results provide structural insights into the protein breathing motions that are associated with aromatic ring flipping. We reveal how the dynamics of the region encompassing residues 517 to 522 are key for accommodating the ring flipping process of Y526. Notably, the transition from the eclipsed, major conformation to the staggered, minor conformation is associated with a structural change from a rare, classic β-bulge to a common, canonical β-strand conformation (Supplementary Video 1, Fig. 3a-c). Breathing motions along the structural trajectory between the major and the minor state generate the necessary void volume for fast ring flipping of Y526 to take place (Supplementary Video 2). Although a recent NMR study suggested extensive local unfolding as the source of cavity creation 41   generated through distinct structural rearrangements, while maintaining the overall protein fold.
More generally, our study shows how the local environment in the protein core can lead to a priori energetically unfavourable conformations of amino acid side chains, and how subtle changes in this environment can lead to major structural rearrangements to revert to preferred amino acid conformations. Our results therefore have implications for protein design and structure prediction, and for how novel biological functions can be acquired during the course of evolution; for example, by altering the delicate balance between hydrogen bonds and CH-π interactions. Finally, the combination of sensitive NMR methods to detect low-populated states, protein design using proteome-wide sequence analyses and high-resolution crystallography could be a strategy to further discover the structural details of sparsely populated protein states and their link to function.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-04417-6.
To obtain unlabelled proteins, Escherichia coli BL21(DE3) cells transformed with one of the constructs were grown in lysogeny broth (LB) medium at 37 °C until the optical density at 600 nm reached 0.7. Protein expression was induced by the addition of isopropyl β-d-1-thiogalactopyranoside (IPTG) to a final concentration of 1 mM. The cultures were grown for an additional 15 h at 20 °C (POSH-SH3s) or 4 h at 37 °C ( JIP1-SH3). The cells were collected by centrifugation and frozen at −20 °C or −80 °C. Isotopically 15 N/ 13 C-and 15 N-labelled proteins were produced by growing transformed E. coli BL21(DE3) cells in M9 minimal medium containing 1 g l −1 of 15 N-NH 4 Cl and 2 g l −1 of 13 C 6 -d-glucose or 12 C 6 -d-glucose. To obtain 15 N-labelled protein with tyrosine residues site-selectively labelled at the ε position with 13 C, transformed E. coli BL21(DE3) cells were grown in M9 minimal medium containing 1 g l −1 of 15 N-NH 4 Cl, 2 g l −1 of NaH 13 CO 3 and 2 g l −1 of [2-13 C]-glycerol 43 .
All SH3 domains were purified by Ni affinity chromatography followed by size-exclusion chromatography. Cell lysis was carried out by sonication using purification buffer (POSH: 50 mM Tris pH 7.0/8.0, 500 mM NaCl, 1 mM β-mercaptoethanol; JIP1: 50 mM HEPES pH 7.0, 150 mM NaCl) supplemented with protease inhibitor tablets (Roche). The washing buffer used for Ni affinity chromatography was the same as the purification buffer with the addition of 20 mM imidazole. The elution buffer was the same as the purification buffer with the addition of 500 mM (POSH) or 300 mM ( JIP1) imidazole. Nickel affinity chromatography was followed by cleavage by the TEV protease, a second Ni affinity column and size-exclusion chromatography on a Superdex 75 (GE Healthcare). This column was equilibrated with 50 mM HEPES pH 8.0, 500 mM NaCl, 2 mM DTT for POSH-SH3-1, 50 mM HEPES pH 7.0, 500 mM NaCl, 2 mM DTT for POSH-SH3-4 and 50 mM HEPES pH 7.0, 150 mM NaCl for JIP1-SH3.

Expression and purification of JIP1-SH3 variants
Expression and purification of the JIP1-SH3 variants (Y526A, V517A, V517L, A541L and H493A) were performed following the same protocol as for JIP1-SH3, except that the cultures were grown for 15 h after induction at 20 °C (instead of 4 h at 37 °C).

Thermal stability measurements of JIP1-SH3
The stability of JIP1-SH3 was measured by differential scanning fluorimetry using a Prometheus NT.48 (Nanotemper) instrument with the emission wavelengths set to 330 and 350 nm and an excitation power of 10%. The melting curve for wild-type JIP1-SH3 was measured at a protein concentration of 4 mg ml −1 in 50 mM HEPES, 150 mM NaCl at pH 7.0 by using Prometheus Standard Capillaries (PR-C002). The temperature scan rate was fixed at 1 °C per min from 20 °C to 95 °C. The melting temperature (T m ) was calculated from the peak of the first derivative of the intrinsic protein fluorescence intensity ratio at 350 nm and 330 nm throughout the duration of the temperature ramp.

N relaxation measurements of JIP1-SH3
Measurements of 15 N relaxation rates (R 1 , R 2 and heteronuclear NOEs) of JIP1-SH3 were obtained using standard HSQC-type pulse sequences 48 at a 1 H frequency of 600 MHz (Agilent, operated with VnmrJ v.3.1). The relaxation rates were measured at four different temperatures: 15, 25, 35 and 45 °C. The magnetization decay was sampled at (0, 100, 200, 400, 600, 800, 1,100, 1,500 and 1,900) milliseconds (ms) for longitudinal and at (10,30,50,70,90,130,170,210 and 250) ms for transverse relaxation. Technical replicates of one or two of these delays were acquired to estimate the uncertainty on the relaxation rates using a Monte Carlo approach. Details of the Lipari-Szabo model free analysis can be found in the Supplementary Discussion.

N and 1 H N CPMG relaxation dispersion of JIP1-SH3 and its variants
All 15 N CPMG relaxation dispersion experiments 24 were carried out at 15 °C using a constant-time relaxation delay of 32 ms with CPMG frequencies (ν CPMG ) ranging from 31.25 to 1,000 Hz and a 1 H decoupling field of 11 kHz. The 1 H N relaxation dispersion experiments were carried out at 15 °C using the published pulse sequence 25 with a constant-time relaxation delay of 20 ms and CPMG frequencies ranging from 50 to 2,000 Hz. Uncertainties on peak intensities extracted from the relaxation dispersion experiments were estimated using the pooled s.d. calculated from repeat measurements (technical replicates of one to three ν CPMG values), each pool being the set of repeat points per ν CPMG and per peak. Uncertainties on R 2eff values were propagated from the peak intensity uncertainty using a Monte Carlo approach. The following relaxation dispersion experiments were acquired: JIP1-SH3: 15

Tyrosine assignments and 13 C CPMG relaxation dispersion of Y526
The 13 Cε-1 Hε tyrosine resonances were assigned at 45 °C by acquiring a two-dimensional (2D) plane of a BEST-TROSY intra-residue HNCACB experiment 44 , an aromatic BEST constant-time 1 H-13 C HSQC experiment 50 and a (Hβ)Cβ(CγCδCε)Hε experiment 51 linking the Cβ chemical shifts directly to the Hε chemical shifts. The spectra were manually peak-picked using NMRFAM-Sparky 45 and 13 Cε-1 Hε tyrosine resonances were assigned manually ( Supplementary Fig. 1). The acquisition of aromatic BEST constant-time 1 H-13 C HSQC experiments at different temperatures (between 5 and 45 °C) enabled the final assignment at 15 °C.
Aromatic L-optimized TROSY-selected 13 C CPMG 38 and R 1ρ 39 relaxation dispersion experiments were carried out at 15 °C on a 1 mM uniformly 15 N and site-selective 13 C-labelled JIP1-SH3 sample in 50 mM HEPES, 150 mM NaCl at pH 7.0. CPMG relaxation dispersion experiments were carried out at magnetic field strengths of 600 MHz, 700 MHz and 850 MHz (Bruker) using a constant-time relaxation delay of 20 ms with CPMG frequencies ranging from 100 to 1,000 Hz. The R 1ρ relaxation dispersion experiment was recorded on-resonance with Y526 at 700 MHz (Bruker) using B 1 field strengths ranging from 700 to 10,000 Hz with a 20 ms relaxation delay. Error bars were derived from repeat measurements as described above for the 15 N and 1 H N relaxation dispersion experiments. Analysis of the 13 C data was carried out using ChemEx or using available analytical expressions for R 1ρ relaxation in the presence of two-site exchange 52 (Supplementary Discussion).

Comparison of JIP1-SH3 to other human SH3 domains
The sequences of 320 human SH3 domains were obtained 33 , aligned using Clustal Omega 53 and categorized according to the identity of the amino acid at the position of Y526 in JIP1-SH3. A new alignment was performed using only the sequences that carry Y or F at this position, amounting to a total of 33 human SH3 domains. The sequences of the SH3 domains of the RIMS-binding proteins 1, 2 and 3 and the metastasis-associated in colon cancer protein 1 (MACC1) were not included in this alignment, as they contain longer insertions compared to the JIP1-SH3 sequence. For each SH3 domain, the amino acids corresponding to residues 493, 517 and 541 of JIP1-SH3 were assigned a size score according to the number of heavy atoms in their side chains Crystallization of JIP1-SH3 and the variants Y526A, V517A, V517L, A541L and H493A JIP1-SH3 and its variants were concentrated to a final concentration of 20 mg ml −1 ( JIP1-SH3, JIP1-SH3(V517A), JIP1-SH3(A541L) and JIP1-SH3(V517L)), 10 mg ml −1 ( JIP1-SH3(Y526A)) and 4 mg ml −1 ( JIP1-SH3(H493A)) after size-exclusion chromatography by using Amicon Ultra-4 3.0-kDa centrifugal filters (Merck). All crystals were obtained in 0.1 M HEPES pH 7.5, 1-5% PEG 400 and 2-2.5 M ammonium sulfate at 20 °C by the hanging-drop vapour diffusion method in 24-well plates (Hampton research) 55 . Drops of 2-3 μl consisting of 1:1 or 2:1 parts of protein solution and reservoir solution were vapour-equilibrated against 500 μl of reservoir solution. All crystals appeared after two days and were collected by transferring them to a mother liquor solution containing 20-30% trehalose, frozen and kept in liquid nitrogen.
The initial condition identified for POSH-SH3-1 was 0.2 M NaF, 20% PEG 3350 from the PEGs-I screen (Qiagen) at 4 °C. Needles appeared after 3 to 7 days. Further optimization was done using the hanging-drop vapour diffusion method at 4 °C in 24-well plates. Drops of 2 μl consisting of equal parts protein solution at 2.5 mg ml −1 and reservoir solution (0.2 M NaF, 22% PEG 3350) were vapour-equilibrated against 500 μl of reservoir solution. Hexagonal crystals appeared after three days and were collected after five days by transferring them to a mother liquor solution containing 5% ethylene glycol as cryoprotectant, frozen and kept in liquid nitrogen.

Structure determination
Crystal diffraction was performed at the ESRF beamlines ID30A, ID23-1, ID23-2 using the MXCube software 56,57 , at the automated beamline MASSIF-1 58 or at the Diamond beamlines I04 and I04-1, all equipped with Pilatus detectors (Dectris). Indexing and integration was performed using the XDS 59 , the autoProc 60 or GrenADeS 61 program suites. Data reduction for JIP1-SH3(H493A) was carried out with Pointless and Aimless 62,63 . Molecular replacement of the wild-type JIP1-SH3 structure was carried out in Phaser 64 using the PDB code 2FPE (chains A-B) as a search model. The structures of JIP1-SH3 mutants were obtained by using our wild-type JIP1-SH3 structure as a search model. The initial solutions were improved through cycles of manual adjusting in Coot 65 and refined by using Refmac5 66 . Aimless, Phaser and Refmac were all used as programs of the CCP4 suite 67 .
The structure of POSH-SH3-4 was determined by molecular replacement using a homology model that was built on the basis of the SH3 domain structure of SORBS1 (PDB code: 2LJ1, chain A), which has 45% sequence identity. The structure of POSH-SH3-1 was determined by molecular replacement using as a search model the SH3 domain of human tyrosine protein kinase C-Src (PDB code: 2SRC). Crystallography applications were compiled and configured by SBGrid 68 .

Structural trajectory and void volume calculations
The structural trajectory between the major and minor conformation was generated with Chimera 69 by morphing between the wild-type JIP1-SH3 structure (PDB 7NYK) and the structures of the two variants JIP1-SH3(H493A) (7NYL) and JIP1-SH3(V517A) (7NYM). To calculate changes in the volume of the Y526 pocket, protons were added to all structures of the trajectory and Y526 was replaced by glycine to allow calculation of the complete pocket volume by POVME 3.0 using a distance cut-off of 1.09 Å, corresponding to the van der Waals radius of a hydrogen atom 70 . A similar strategy was used to generate the structural trajectory between the wild-type JIP1-SH3 structure (PDB 7NYK) and the structure of the JIP1-SH3(A541L) variant (7NYO).

MD simulations of JIP1-SH3
MD simulations were carried out using ACEMD v.3.3.0 71 and the Charmm36m force field parameters 72 . Using VMD 73 , coordinates of the dimer from PDB 2FPE were inserted in the box of dimensions with a minimum distance of 2 Å in each direction between each atom and any box side. The box was then filled with water molecules and an amount of Na + and Cl − corresponding to [NaCl] = 0.1 M. Electrostatic interactions were evaluated using Particle-Mesh Ewald (PME) electrostatics with a cut-off distance of 9 Å. Van der Waals forces were calculated with a cut-off of 9 Å and a switching function active from 7.5 Å to smoothly reduce the potential to zero. An integration step of 2 fs and holonomic constraints on all hydrogen-heavy atom bond terms were used. The energy of the system was minimized using conjugate-gradient minimization for 500 steps. Random velocities from a Maxwell distribution with T = 298.15 K were assigned to atoms. Then, the system was equilibrated first for 100 ps in the NVE ensemble and then for 1 ns in the NPT ensemble. In the latter case, temperature and pressure were controlled using the Langevin thermostat with a damping constant of 1 ps −1 and Berendsen barostat with a relaxation time of 400 fs, respectively. Finally, a 1 μs trajectory was calculated in the NVT ensemble using the Langevin thermostat with a damping constant of 0.1 ps −1 . Trajectories were processed and analysed using the MDAnalysis Python package 74 .