Cooperative dynamics across distinct structural elements regulate PTP1B activity

Protein Tyrosine Phosphatase 1B (PTP1B) is the canonical enzyme for investigating how distinct structural elements influence enzyme catalytic activity. While it is recognized that dynamics are essential for PTP1B function, the data collected thus far have not resolved whether distinct elements are dynamically coordinated, or, alternatively, if they fulfill their respective functions independently. To answer this question, we performed a comprehensive 13 C methyl relaxation study of Ile, Leu and Val (ILV) residues of PTP1B, which, due to its substantially increased sensitivity, provides a comprehensive understanding of the influence of protein motions on different timescales for enzyme function. We discovered that PTP1B exhibits dynamics at three distinct timescales. First, it undergoes a distinctive slow motion that allows for the dynamic binding and release of its two most N-terminal helices from the catalytic core. Second, we showed that PTP1B 13 C-methyl group side chain fast timescale dynamics and 15 N backbone fast timescale dynamics are fully consistent, demonstrating that fast fluctuations are essential for the allosteric control of PTP1B activity. Third, and most importantly, using 13 C ILV ct-CPMG relaxation measurements experiments, we demonstrated that all four catalytically important loops—the WPD, the Q, the E and the substrate binding loop (SBL)— work in dynamic unity throughout the catalytic cycle of PTP1B. Thus, these data show that PTP1B activity is not controlled by a single functional element, but instead all key elements are dynamically coordinated. Together, these data provide the first fully comprehensive picture on how the validated drug target PTP1B functions.


Abstract
Protein Tyrosine Phosphatase 1B (PTP1B) is the canonical enzyme for investigating how distinct structural elements influence enzyme catalytic activity. While it is recognized that dynamics are essential for PTP1B function, the data collected thus far have not resolved whether distinct elements are dynamically coordinated, or, alternatively, if they fulfill their respective functions independently. To answer this question, we performed a comprehensive 13 C methyl relaxation study of Ile, Leu and Val (ILV) residues of PTP1B, which, due to its substantially increased sensitivity, provides a comprehensive understanding of the influence of protein motions on different timescales for enzyme function. We discovered that PTP1B exhibits dynamics at three distinct timescales. First, it undergoes a distinctive slow motion that allows for the dynamic binding and release of its two most N-terminal helices from the catalytic core. Second, we showed that PTP1B 13 C-methyl group side chain fast timescale dynamics and 15 N backbone fast timescale dynamics are fully consistent, demonstrating that fast fluctuations are essential for the allosteric control of PTP1B activity. Third, and most importantly, using 13 C ILV ct-CPMG relaxation measurements experiments, we demonstrated that all four catalytically important loops-the WPD, the Q, the E and the substrate binding loop (SBL)work in dynamic unity throughout the catalytic cycle of PTP1B. Thus, these data show that PTP1B activity is not controlled by a single functional element, but instead all key elements are dynamically coordinated. Together, these data provide the first fully comprehensive picture on how the validated drug target PTP1B functions.

Introduction
Protein tyrosine phosphatase 1B (PTP1B, PTPN1) was the first non-receptor bound protein tyrosine phosphatase (PTP) isolated (1). Not surprisingly, it is also the best-studied member of the human PTP family (2). Since its discovery, PTP1B has been shown to have diverse roles in multiple cellular processes, especially glucose uptake, body mass regulation, motility and proliferation. As a consequence, PTP1B is a validated target for multiple diseases, especially diabetes and cancer (3).
PTP1B catalyzes the hydrolysis of phosphorylated tyrosine residues (4). The catalytic site is defined by: 1) the PTP loop ([I/V]HCxxGxxR[S/T]G), which includes cysteine C215, which functions as the catalytic nucleophile in the first step of hydrolysis; 2) the WPD loop, 179 WPD 181 , which contains aspartic acid D181, which functions as the proton donor and acceptor during phosphoryl transfer, 3) the E-loop, which plays a role in substrate recruitment; 4) the substrate binding loop (SBL), which restricts dephosphorylation to tyrosine residues; and 5) the Q-loop (residues 261-265), which includes the key glutamine residue Q262 that is responsible for coordinating a nucleophilic water, ensuring that PTPs do not function as phospho-transferases ( Fig. 1A) (5). Upon substrate binding, it is the WPD loop that undergoes that largest structural change, moving from an open (hydrolysis incompetent) to a closed (hydrolysis competent) position. Finally, PTP1B was also shown nearly 20 years go to contains an allosteric binding pocket that is ~20 Å away from the catalytic site, at the intersection of helices α3-α6-α7 (6).
Recently, a number of reports have shown that dynamics play critical roles in PTP1B enzyme activity and allostery (7)(8)(9). In addition to dynamics, these reports also highlighted the importance of structural rigidity in the extended WPD loop, particularly for proline residue Pro185 (7). This proline is essential for PTP1B activity, as it controls an indispensable CH/π switch that associates either with Trp179 within the WPD loop (closed state) or Phe269 from helix α6 (open state; Fig. S1). This CH/π switch controls WPD motion (7,10). Further, PTP1B helix α3 was identified as the mechanical/dynamics support that drives the transition between the open and closed states of the WPD loop and serves as the connector between the WPD loop and the allosteric pocket and helix α7 (7).
Together, these structural and dynamics data have revealed many key aspects of PTP1B activity and regulation. However, there remains a key unresolved question. Namely, are the observed dynamics for distinct structural elements independent of one another, or, alternatively, are their dynamics coordinated so that all catalytically critical residues work in dynamic unity throughout the 3 PTP1B catalytic cycle. The previously reported data were primarily based on 15 N-based protein backbone NMR measurements (7,11), which are excellent reporters on many aspects of protein function. However, when working with larger proteins such as PTP1B, some of these experiments have low sensitivity, decreasing measurement accuracy and making the data statistically unreliable.
Thus, to answer this question, we used 13 C methyl relaxation studies of Ile, Leu and Val (ILV) residues in PTP1B (12)(13)(14). The fast rotation of methyl groups in ILV residues ensures these NMR experiments have high sensitivity, allowing the dynamics of PTP1B to be determined at three distinct timescales -fast (ps/ns), intermediate (µs/ms) and slow (ms/s). Our data confirm our previous discovery (using only 15 N-based protein backbone NMR measurements) that fast motions are critical for the regulation of allostery in PTP1B. However, we also demonstrate that intermediate timescale motions regulate PTP1B activity. Critically, the motions observed in our 13 C ILV data extend far beyond the previously-reported movement of the WPD loop and have revealed that PTP1B substrate recruitment and dephosphorylation function in dynamic unison. Such behavior has also been seen for other enzymes, including the enzymes dihydrofolate reductase (DHFR) (15,16), cyclophilin A (17) and the kinases p38 (18) and ERK2 (19). Together, this work lays the foundation for a comprehensive understanding of conformation and dynamics for the entire PTP family specifically, and enzyme function and protein allostery, generally.

C ILV methyl assignment of PTP1B identifies reporters distributed throughout the protein
Fundamental insights into the dynamics and the function of a protein can be obtained by studying the 15 N backbone and/or the 13 C side chain dynamics, especially of Ile, Leu and Val residues. We recently reported the 15 N-based sequence specific backbone assignment of the folded catalytic domain of PTP1B (~35 kDa; residues 1-301; hereafter referred to as PTP1B) and leveraged these data to define how 15 N fast timescale dynamics contributes to PTP1B activity, especially allostery (20). However, the relatively large size of PTP1B, together with the inability to concentrate the protein to ≥250 µM, made 15 (21), which included nuclear Overhauser effect (nOe) experiments and measurements on 16 single amino acid variants, enabled us to complete the PTP1B 13 C ILV methyl group assignment both in its free ( Fig.  S2A; open; 98% assigned; only Leu88 is missing due to overlap) and active site inhibitor TCS-401 bound form ( Fig. S2B; closed; 96% assigned; Leu88 and Val49Cδ2 are overlapped; Ile219 is either broadened beyond detectability or overlapped). Our complete ILV assignment also correlates well with a partial assignment of the Ile-only region of PTP1B that was recently reported (10). NMR 15 N-based sequence-specific backbone assignments of large proteins, such as PTP1B, are often incomplete either due to slow H/D back exchange after protein expression in D2O-based medium or intermediate conformational exchange, which broadens peaks beyond detection (22). These are not issues for 13 C ILV methyl group assignment. Nevertheless, when counting the number of peaks in the Ile Cδ region of the spectrum, we observed 4 more peaks than expected (16 expected; 20 observed; Fig. S3A). The assignment showed that these peaks belong to a minor conformation of Ile10, Ile19, Ile23 and Ile246, which form the interface between the catalytic PTP1B domain and helices α1' and α2' (Fig. S3B). Thus, it is likely that helices α1' and α2' disengage from the core protein in a slow dynamics event (pB ~25%), as they are only loosely connected to the catalytic domain via L0, a 10 amino acid long linker (PTP1B residues 28-37).

PTP1B fast timescale 13 C ILV relaxation data correlate with PTP1B 15 N backbone relaxation data
To define the PTP1B motions at different timescales, we recorded T1 and T1ρ/T2 fast timescale 13 C ILV side chain relaxation data ( Fig. 2A, S4). Values of these parameters typically correlate with the distance of the 13 Cmethyl group from the backbone, i.e. the further the distance the higher the flexibility. The fast timescale 13 C ILV relaxation data show this is also true for PTP1B, where on average the methyl groups of Valine residues rotate the slowest, followed by Leucine and Isoleucine (Fig. S5A,B). Plotting the PTP1B T2 13 C ILV side chain relaxation data against the protein sequence and comparing it directly with our previously reported 15 N backbone relaxation data (T1, T2) shows an overall similar behavior. This includes a sharp increase in overall dynamics of helix α7, which we previously established as the central modulator of allostery in PTP1B ( Fig. 2A,B). The only region where the T2 13 C ILV side chain and 15 N backbone relaxation data differ is for PTP1B helix α3, which modulates WPD motion and subsequently PTP1B activity; helix α3 did not show the increased dynamics identified in the 15 N backbone relaxation data (7).

C ILV assignment and side chain dynamics of closed PTP1B
Next, we determined the consequences of active site inhibitor binding on PTP1B chemical shifts and dynamics. TCS401 is a small (306 Da) active site inhibitor that binds PTP1B with a KD of ~26 ± 2 µM (23). We and others have shown that the binding of small molecule inhibitors and/or substrate peptides leads to the closure of the WPD loop and a rearrangement of PTP1B helix α3, which rigidifies helix α7. As expected, direct comparison of the 15 N-backbone chemical shift perturbation (CSP) data with that of 13 C ILV shows excellent overlap (Fig. S6). Notably, PTP1B helices α1' and α2' still disengage from the core enzyme (pB ~29%) in TCS401saturated PTP1B, showing that binding of TCS-401 does not influence this slow dynamics event (Fig. S3C).
To determine if altered dynamics accompanies active site inhibitor binding, we repeated the 13 C ILV side chain relaxation dynamics measurements with TCS401saturated PTP1B. The most significant changes were observed in helix α7, where the dynamics detected in the TCS401-free state were quenched in the TCS401-saturated state. This is identical to what was observed in the 15 N relaxation data (R1, R2) that probe fast timescale backbone dynamics ( Fig. 2A) (7). Val287 and Leu294 show the largest changes, which define the core of the PTP1B allosteric site, at the intersection of helices α3, α6 and α7 (Fig. 2B). Thus, these results further highlight the critical role of helix α7 and its intrinsic dynamics for the PTP1B allosteric network. Finally, consistent with the PTP1B 15 N backbone relaxation data, and the 13 C ILV CSP data Val155 also has a statistically significant reduction in 13 C ILV side chain fast timescale dynamics ( Fig. 2A). Val155 abuts the L11 loop, which includes Tyr152 and Tyr153, two residues that are important for the allosteric pathway in PTP1B. Together, these data show that the fast-timescale dynamics results between the 15 N backbone relaxation and the 13 C ILV side chain relaxation data are consistent, a critical step before evaluating the 13 C ILV side chain ct-CPMG data.

C ILV intermediate timescale dynamics of PTP1B
Next, we measured 13 C ILV ct-CPMG side chain relaxation data (all 13 C ILV ct-CPMG for the studies reported here are recorded at two magnetic fields; 14.1 and 18.8 T), which reports on µs-ms motions, for free PTP1B. Enzymatic reactions for signaling enzymes such as PTP1B often have catalytic turnover rates (PTP1B 15-60 s -1 ) that can be correlated with protein dynamics in the μs-ms timescale. In ct-CPMG measurements, residues undergoing μs-ms conformational exchange dynamics show changes in the effective relaxation rate R2 (R2,eff), which are measured as a function of the repetition frequency (νCPMG). Plots of R2,eff vs νCPMG are curved for those residues experiencing μs-ms timescale dynamics. Fitting these curves to a two-state model (Carver-Richards) allows the populations of (pA and pB) and the exchange rates between (kex) these populations to be extracted. A significant number of residues in free (38 residues) and TCS401-saturated (23 residues) PTP1B exhibited μs-ms conformational exchange dynamics. Using an approach that we used previously to analyze p38 MAP kinase dynamics (18), we next identified residues that experience similar exchange dynamics. Briefly, this approach identifies groups of residues with similar fluctuations and thus provides increased statistical significance when evaluating ct-CPMG data.
In free PTP1B, we identified two groups of residues with uniform μs-ms conformational exchange dynamics ( Table 1). The two clusters showed fast μs exchange dynamics with kex of 3550-6160 s -1 . We identified a small cluster of five residues (group 1) with very fast exchange dynamics (kex = 6160 ± 360 s -1 and pB = 1.8 ± 1.0%). This cluster exclusively includes residues on the surface of PTP1B and thus these dynamics likely reflect interactions with bulk solvent (Fig. 3A). In contrast, a large cluster (33 residues; group 2) shows exchange dynamics of kex = 3550 ± 70 s -1 and pB = 3.2 ± 0.4% (pA open PTP1B form; pB closed or similar to the PTP1B closed form) and includes residues from most secondary structure elements of PTP1B, including the WPD loop and all other regions of PTP1B that have identified functions in the catalytic activity and allostery of PTP1B. This demonstrates that, in its free form, PTP1B exhibits largely uniform exchange dynamics (Fig. 3A).
Comparing our 13 C ILV ct-CPMG side chain relaxation data (38 residues) with previously reported 15 N backbone ct-CPMG data (3 residues; fit with a kex = 900 s -1 and pB = 2.3%; Fig. 1A) shows similarity (11). Our pB's are similar to the ones resulting from the 15 N backbone ct-CPMG data evaluation ( 13 C ILV ct-CPMG pBs, 3.2%; 15 N backbone ct-CPMG pB, 2.3%). The populations extracted from our 13 C ILV ct-CPMG side chain relaxation data have larger errors, due to the 4fold increase in kex, which makes fitting accurate populations more difficult. This 4-fold faster 13 C kex, most likely represents a fast (probably side-chain associated) motion that drives an underlying slower (backbone associated) motion that was detected in the 15 N data. Together, these data show that free PTP1B exhibits mostly fast exchange dynamics and that these can be partitioned into two statistically significant groups. However, the largest, most significant group encompasses most of the protein and highlights the largely uniform exchange behavior throughout free PTP1B.

Active site binding alters 13 C ILV intermediate timescale dynamics of PTP1B
Next, we evaluated the 13 C ILV ct-CPMG side chain relaxation data for TCS401-saturated (inhibitor bound) PTP1B (WPD loop is closed; helix α7 becomes rigidified). The data show a large reduction in the overall exchange dynamics of PTP1B when the active site is occupied. We were able to fit three statistically relevant residue groups ( Table 2). Group 1 (3 residues), which exhibits the fastest exchange dynamics (kex = 2100 ± 240 s -1 ; pB = 0.4 ± 0.03%), is a small group and includes only residues in β-strands 8 and 9, which are distant from the PTP1B active site. The exchange dynamics for Group 2, which is the largest group (13 residues), are significantly slower (kex = 780 ± 50 s -1 ; pB = 2.1 ± 0.2%) and composed of residues surrounding the PTP1B active site and residues present within helices α3/4/6 and β-strands 4/9/10/11. Finally, group 3 (7 residues) exhibits the slowest exchange dynamics (kex = 550 ± 40 s -1 ; pB = 7.8 ± 0.6%). All of these residues belong to structural elements that are critical for PTP1B substrate recruitment and activity, including Val184 in the WPD loop (activity), Val113 and Leu119 of the E-loop (substrate recruitment), Val49 of the substrate recruitment loop (substrate recruitment) and Ile261 of the Q-loop (activity; Figs. 3B,C). The linked dynamics of these disparate sites in the TCS401-bound state strongly suggests that motions within the active site become synchronized upon binding of inhibitors/substrates. Only a limited comparison can be made between our 13 C ILV ct-CPMG data and the previous 15 N backbone ct-CPMG data of closed PTP1B, as the latter analysis is based on a single residue (A198 in helix α3) and was recorded using a peptide with a nonhydrolysable tyrosine analog (11). Our data shows that the population of the PTP1B open state in TCS401-saturated PTP1B (pB) is 7.8%, compared to 14% for peptide-saturated PTP1B. This difference in population either reflects the less optimal fit of a single residue or simply reflects the higher affinity of TCS401 for PTP1B versus a substrate-like peptide (low µM vs. high µM to low mM), leading to a higher population of the closed state of PTP1B in our data.

Discussion
Protein dynamics at multiple timescales are essential for protein function and regulation (24)(25)(26). NMR spectroscopy is distinctively qualified to report on protein dynamics on multiple timescales, from ps to hours. 15 Nbackbone dynamics, which report on the dynamics of the amide 15 N-H N vector, is the most widely used technique, as it has one reporter for each amino acid (except for proline). However, this approach is of limited applicability for larger proteins (≥ 35 kDa) because their increased overall correlation time τc limits sensitivity and resolution. Thus, for these systems the use of 13 C-methyl groups for dynamics measurements, especially from Isoleucine, Leucine and Valine (ILV) residues, has become a routine approach. 13 C-H vectors in ILV methyl groups rotate rapidly and thus provide high sensitivity and sharp lines even for very large proteins. Here, we have used 13 C ILV relaxation measurements to define the dynamics of PTP1B at different timescales in order to understand how PTP1B dynamics correlates with enzymatic function. In particular, while it is well-accepted that multiple structural elements of PTP1B are essential for catalysis (WPD-, SBL-, Q-loops), we set out to determine if and how these elements are dynamically coordinated, or, alternatively, if they fulfill their respective functions independently of one another.
Using 13 C-methyl group dynamics relaxation experiments, we discovered that PTP1B exhibits dynamics at three distinct timescales. First, and somewhat unexpectedly, we discovered that PTP1B undergoes a distinctive slow motion that allows for the dynamic binding and release of helices α1' and α2' from the core catalytic domain. Among the non-receptor PTP family, these helices are unique to PTP1B and its most closely related homolog, TCPTP/PTPN2 (2), suggesting they might have a specific function. If and how these helices contribute to PTP1B (and TCPTP) activity is unknown; however, an intriguing possibility is that the dynamic binding and release of these helices may facilitate substrate recruitment. Second, we confirmed that, in PTP1B, both PTP1B 13 C-methyl group side chain fast timescale dynamics (this work) and 15 N backbone fast timescale dynamics in elements distal from the active site, particularly helix α7 (7), are essential for the allosteric control of PTP1B activity. Thus, these data confirm that, in PTP1B, allostery is governed by local fast timescale dynamics, which modulates intermediate timescale dynamics that controls the catalytic activity of PTP1B.
Third, we performed 13 C ILV ct-CPMG relaxation measurements to measure µs-ms exchange dynamics, as this timescale is most often correlated with enzymatic function. In agreement to a previous report, we see a significant 6.5-fold reduction of the WPD-loop motion when the active site is occupied (open, kex = 3550 ± 70 s -1 ; closed; kex = 550 ± 40 s -1 ). However, our comprehensive relaxation data, which reports on dynamics throughout PTP1B (up to 56 distinct residues), significantly expands this observation. Namely, the data show that the functionally critical elements of PTP1B structure fluctuate coherently and distinctly from the rest of the protein when a substrate analog/inhibitor binds the PTP1B active site. This includes residues from the WPD loop, the substrate recruitment E-loop, the SBL (substrate recruitment loop) and the Qloop. This shows that all catalytically critical residues work in dynamic unity throughout the catalytic cycle of PTP1B.
It is well recognized that conformational dynamics/plasticity can have an essential role in the catalytic cycle of multiple enzymes. For instance, in cyclophilin A, the observed rate-limiting dynamics reflect coordinated motions across its active site, even in the absence of obvious backbone changes in the corresponding crystal structures (17). However, in the case of PTP1B and other PTPN family members, crystal structures show clearly defined changes in the position of the WPD loop (open/closed) that, in principle, might depict the full range of conformational heterogeneity that allow for the catalytic activity (27). Indeed, similar limited motions that are rate limiting have been reported for the enzyme triosephosphate isomerase (28). Furthermore, previous results indicated that motions of the PTP1B WPD loop were apparently correlated to the catalytic rate (11), suggesting this limited model of conformational heterogeneity (WPD loop open/closed) was likely appropriate.
However, further experiments demonstrated that the rate of catalysis of PTP1B is curiously disconnected from the WPD loop's dynamic equilibrium and suggested the involvement of adjacent structural elements in rate-limiting dynamics and thus catalytic activity (7,10). Indeed, crystallographic evidence from the related protein tyrosine phosphatase PTPN7 (HePTP) showed coordinated motions of the E-loop and WPD-loop in response to depletion of a phosphate analogue from the binding site (29). Thus, it is likely that coherent/coordinated fluctuations within the active site, either in response to loop opening or in a completely closed state, control catalysis. Our new data adds significant support to this latter view. Critically, the greater density of probes provided by uniform ILV-labeling allows for the comprehensive observation of dynamic effects throughout the PTP1B active site and shows that all catalytically significant structural elements/loop are engaged in a coherent fluctuation in the presence of an inhibitor, which functioned as a proxy-substrate in our analysis. Thus, while the WPD loop mobility is important, it is clearly not the sole contributor to PTP1B dynamics and function at the PTP1B active site.
Taken together, these data suggest that in PTP1B, intermediate dynamics are important for substrate binding and product release while fast dynamics, in elements distal from the chemical reaction, are important for the allosteric control of the overall function of PTP1B. As these structural features are highly conserved within the PTP family of proteins, it is likely that this pattern of dynamic influence is conserved throughout the PTP family (30). Furthermore, it is likely that sites within these networks can be targeted by new allosteric approaches to modulate PTP function.
Multiple labeling schemes were used for different measurements. For the assignment of PTP1B ILV, PTP1B was expressed with 1 g/L 15 NH4Cl, 4 g/L [ 2 H, 13

Protein Purification
PTP1B was purified as previously described into NMR Buffer (10 mM HEPES pH 7.4, 150 mM NaCl, 5 mM DTT) (7,20). Purified protein was either used immediately or flash frozen in liquid nitrogen for storage at −80 °C. Samples for NMR dynamics experiments were aliquoted into 550 ul aliquots, flash frozen in liquid nitrogen and lyophilized for 24 hours. The protein was then resuspended in 550 µl D2O. Typical PTP1B protein yields are ∼55 mg/L in D2O-based M9 minimal medium and variant yields are ~21-48 mg/L in M9 minimal media at 98% purity.

NMR measurements
All PTP1B 13 C-methyl ILV assignment NMR measurements were performed using a Bruker Advance Neo 800 MHz NMR spectrometer with a TCI HCN-active z-gradient cryoprobe at 298 K. A 3D ( 13 C, 13 C, 1 H) HMCM(CG)CBCA experiment functioned as the initial assignment step using PTP1B produced following scheme 1 (32). A 3D ( 13 C, 13 C, 1 H) HSQC-NOESY-HSQC (τm= 200 ms) and a 3D ( 13 C, 15 N, 1 H) HSQC-NOESY-HMQC (τm = 400 ms) spectra were recorded using PTP1B produced following scheme 3. The final concentration of PTP1B for these measurements was 0.35 mM in NMR Buffer containing 10% D2O. All data were processed using NMRPipe (33) or Topspin 4.0.5 and analyzed using NMRFAM-SPARKY (34). 13 C-methyl relaxation NMR measurements ( 13 CHD2) were performed on Bruker Advance Neo 600 and 800 MHz spectrometers equipped with TCI HCN-active z-gradient cryoprobes at 298 K. Data was recorded on 2 H, 12 C, 15 N-labeled PTP1B with 13 CHD2 labeled ILV methyl groups (scheme 4), either free or inhibitor/TCS401-saturated at a final protein concentration of 0.25 mM in NMR buffer and 100% D2O. Sample concentration was tightly monitored to ensure no effect on τc and thus T1 measurements. TCS401 inhibitor was carefully titrated to achieve full saturation. Upon saturation (chemical shifts of interacting residues stopped changing; usually at 1:3 ratio), additional TCS401 (to 1:6 ratio) was added to ensure that all experiments were performed under fully inhibitor-saturated conditions and thus that the observed CPMG dispersions are independent of ligand on/off exchange events. All relaxation data was recorded as a pseudo-3D in a fully interleaved manner.

Relaxation analysis
T1 and T1 values were calculated using NMRviewJ using the peak intensities (jitter function) and exponential decay fitting function. Errors were also determined via relaxation curve fitting. T2 was extracted from T1 by: / , where β is the effective rotation angle for each 15  .

Data availability
All data are contained within the article.

Funding
This work was supported by the American Diabetes Association Pathway to Stop Diabetes Grant 1-14-