Distal Mutations Rewire Allosteric Networks to Control Substrate Specificity in PTP1B

Abstract

Protein tyrosine phosphatase 1B (PTP1B) is a key regulator of cellular signaling pathways, and its dysregulation is linked to diabetes, obesity, cancer, and immune dysfunction. While the catalytic mechanism of PTP1B is conserved across protein tyrosine phosphatases, its regulation by distal allosteric sites remains less understood. Here, we investigate how mutations at four allosteric sites (Y153, I275, M282, and E297) alter PTP1B substrate specificity and enzymatic dynamics. Kinetic analyses with phosphotyrosine peptides and p-nitrophenylphosphate reveal that allosteric mutants display distinct changes in catalytic efficiency (k_cat/K_m), in some cases reversing substrate preference relative to wild-type enzyme. Solution NMR spectroscopy and microsecond molecular dynamics simulations demonstrate that these mutations perturb long-range communication networks, disrupting coupling between α3 and α7 helices and altering acid-loop flexibility and active-site dynamics. Notably, the E297A mutation has the most pronounced effects, rigidifying the acid loop and weakening allosteric communication to the catalytic center. Community network analysis highlights the acid loop and α7 helix as central hubs linking distal sites to the active site. Together, these results establish that distal mutations can reshape PTP1B’s dynamic landscape, thereby modulating substrate specificity. This work expands understanding of allosteric regulation in PTP1B and provides a framework for targeting dynamic networks to control phosphatase activity.

Introduction

Protein tyrosine phosphatase 1B (PTP1B, UniProtKB: P18031) is a key signaling enzyme that regulates a myriad of biochemical pathways. Dysfunction of PTP1B results in several diseases including type-II diabetes, obesity, cancer, and immune system compromise.¹ Like other protein tyrosine phosphatases (PTPs), PTP1B employs a conserved catalytic mechanism (Fig. 1A).² The reaction proceeds in two steps. First, the enzyme cleaves phosphotyrosine containing substrates. An absolutely conserved cysteine thiolate, located in the so-called P-loop, containing the consensus sequence (I/V)-HCXAGXGR(S/T), nucleophilically attacks the substrate phosphorous. Simultaneously, the adjacent acid loop, which contains the conserved catalytic acid (D181), closes around the active site, positioning itself to protonate the departing phenyl group; prior works call this loop the WPD loop because the founding members of the PTP family have WPD at the N-terminus of this loop. It is now clear that most members do not have WPD in this loop, only the catalytic acid. Therefore we refer to this loop throughout this work as the acid loop. This cleavage releases the dephosphorylated peptide and generates a thiophosphate enzyme intermediate. In the second step, the intermediate is hydrolyzed through the coordinated actions of D181 and the conserved Q262 located on the Q-loop. Substrate specificity for phosphotyrosine over phosphoserine/phosphothreonine (pSer/pTy) arises partly from a deep active-site binding pocket formed by the pTyr loop, which contains Y46.³

Figure 1.

PTP1B mechanism and structure. (A) General mechanism of protein tyrosine phosphatases. After substrate binding (rate k₁), the nucleophilic cysteine on the P-loop attacks the phosphoester-containing substrate (k₂) to form a thiophosphate intermediate as the catalytic aspartic acid on the mobile acid loop donates a proton to the leaving group, R. In the second step, the same aspartic acid activates a water molecule coordinated by Q262 on the Q-loop to hydrolyze the intermediate (k₃). (B) The loop regions surrounding the active site are colored on a crystal structure of apo PTP1B (PDB entry 2CM2, left) and vanadate-bound PTP1B (PDB entry 3I80, right) with the catalytically important residues displayed in sticks. The coloring scheme is: pTyr-loop (residues 45 – 49) in green, E-loop (112 – 120) in light yellow, acid loop (177 – 188) in orange, P-loop (213 – 223) in red, and Q-loop (254 – 263) in violet. The left and right panels show the acid loop in the open conformation and closed conformations, respectively. (C) The location of the loops in (B) are shown (PDB entry 1G1G) using the same coloring scheme from (B). The allosteric sites in cyan consisting of β3, α2, α4, α6, α7, and a small part of α3 (193 – 197). Loops L4 (86 – 90), L11 (149 – 154), L16 (237 – 243), and L18 (280 – 285) are also shown in cyan. The same regions in panels (C) are shown in panel (D) with the same color scheme with the sites of mutations noted in black.

Overall, the catalytic site is shaped by several mobile loops, including the P-, acid-, Q-, pTyr, and E-loops (Fig. 1B,C) whose conformational dynamics are critical for both catalysis and substrate binding.^3,4,5–7 Notably, PTP1B as well as other PTPs are the only known enzyme family in which a catalytic residue is located on a mobile loop. This unique feature has led to the hypothesis that regulation of catalysis may be achieved by controlling the acid-loop motion.⁴ Support for this idea came from solution NMR studies showing that the acid-loop closure kinetics closely match the timescale for substrate cleavage in related PTPs.⁵

Consistent with its role in several disease states, PTP1B has been shown to act on multiple intracellular targets.^8–11 Biochemical and structural studies indicate that PTP1B preferentially dephosphorylates substrates with acidic or hydrophobic residues at the −1 position relative to the phosphotyrosine (pY) residue (with “−” indicating N-terminal to the pY residue, whereas “+” indicates amino acids just C-terminal to the pY group).^12,13 It has been suggested that PTP1B favors Met at the +1 position, with weaker preference for Gln and Ser, while disfavoring substrates with positively charged residues N-terminal to pY or with Gly/Pro flanking the pY site.¹³ In vivo, PTP1B has been shown to dephosphorylate the insulin receptor (IR),¹⁴ Janus kinase 2 (Jak2),¹⁵ p62dok,¹¹ and the epidermal growth factor receptor (EGFR)⁸, among others.¹⁶ Complementary, in vitro screening studies have identified a broad range of substrate diversity for pY peptides cleaved by PTP1B.^17,18 This broad substrate specificity implies that PTP1B activity is subject to multiple layers of regulation within cells. Reported mechanisms include reversible oxidation,^19,20 post-translational modification^21–24, and subcellular localization.^25,26 In addition, PTP1B harbors allosteric sites, though their physiological roles remain poorly understood. While there is limited detail on in vivo allostery of PTP1B, we wondered if perhaps allosteric sites on PTP1B could alter its substrate specificity. One such allosteric site, located between α3 and α6, was identified in an inhibitor screen.²⁷ Ligand binding at this site restricted closure of the acid-loop and limited catalytic activation. Subsequent studies have revealed and characterized additional allosteric sites in PTP1B.^28,29–34

Building on our previous work, which identified multiple distal sites of PTP1B where mutation perturbs catalytic activity (Fig. 1D),^34,35 we sought to determine how these mutations influence substrate specificity, protein structure, and enzyme dynamics. Here, we combine experimental and computational approaches to explore how allosteric mutations trigger cascades of altered long-range interactions that rewire conformational dynamics on nanosecond to millisecond timescales. These dynamical changes remodel the native active site environment, ultimately reshaping the conformational ensemble of catalytic residues and substrate-binding determinants.

Material and Methods

Site-directed Mutagenesis.

Site-directed mutagenesis was performed using the QuickChange Site-Directed Mutagenesis Kit using primers listed in SI Table 1. The correct DNA plasmid sequence was confirmed by DNA sequencing at the Keck DNA Sequencing Core (Yale University).

Protein expression and Purification.

The expression plasmids containing the gene for WT and mutant PTP1B were transformed into BL21(DE3) cells. For kinetics experiments, E. coli were grown in 50 mL of LB broth with 50 μg/mL kanamycin at 37 °C overnight and transferred to 1 L of LB with 50 μg/mL kanamycin and grew at 37 °C. Protein induction was achieved with 0.5 mM IPTG when the OD₆₀₀ of the cell culture reached a value of 1.0 and expression was continued at 25 °C for 18 hours at which time the cells were harvested by centrifugation at 4 °C and 7400 x g for 45 minutes.

For NMR studies, E. coli were grown in 5 ml of H₂O LB at 37 °C overnight. After this initial growth culture, 2 mL of the LB starting culture was centrifuged at 3000 x g for five minutes at room temperature and the pellet was resuspended in 20 mL of LB prepared using recycled D₂O (typically ~ 90% ²H content). After eight hours of growth at 37 °C, the D₂O LB culture was centrifuged at 3000 x g for 10 minutes at room temperature and the pellet was resuspended in 100 mL of 2x M9 minimal medium prepared with recycled D₂O and grown at 37 °C overnight and subsequently was centrifuged for ten minutes at 3000 x g and room temperature. The resulting cell pellet was resuspended in 1 L of 2x M9 minimal medium prepared with 99.9% D₂O, 1 g of ¹⁵NH₄Cl and 2 g of glucose/dextrose and grown until OD₆₀₀ reached ca. 0.8. Half an hour before induction, 60 mg/ml α-ketobutyric acid and 100mg/ml α-ketoisovaleric acid (Cambridge Isotope Labs) was added for the ¹³C labeling of methyl groups of isoleucine, leucine, and valine (ILV). The cells were then induced with 0.5mM IPTG and induction was continued for 18 hours before the cells were harvested at 4 °C and 7400 x g for 45 minutes.

PTP1B purification.

The cell pellet was resuspended in 5 mL of lysis buffer (50 mM Tris, 500 mM NaCl, 5% glycerol, pH 8.0 for each 1 gram of pellet). One tablet of cOmplete Mini EDTA-free protease inhibitor was also added and the cells were then lysed by sonication on ice. The cell lysate was centrifuged for 45 minutes at 4 °C at 27000 × g and the supernatant was filtered through a 0.45 μm filter and loaded to nickel column. The column was washed with 12 CV of wash buffer (50 mM Tris, 500 mM NaCl, 20 mM imidazole, 5% glycerol) and the protein was eluted with 12 mL of elution buffer (50 mM Tris, 500 mM NaCl, 300 mM imidazole, 5% glycerol). Fractions containing the His-tagged PTP1B were pooled and the affinity tag was removed with TEV protease under dialysis in 50 mM Tris, 500 mM NaCl, 20% glycerol, 0.5 mM 2-mecaptoethanol, pH 8.0 at 4 °C with a PTP1B:TEV concentration ratio of ca. 3:1. The resulting mixture was loaded to the nickel column again and the cleaved PTP1B was then collected from the flow-through while the his-tageed TEV was retained on the column. PTP1B containing fractions were pooled and concentrated using and Amicon concentrator at 4 °C. Typically, ¹³C-ILV labeled samples were then immediately used in NMR experiments.

Pre-steady-state (Stopped-flow) Kinetics.

Purified PTP1B (1–301) was dialyzed in 50 mM Tris, 50 mM Bis-tris, 100 mM sodium acetate, pH 6.0 and concentrated to 120 – 260 μM. Pre-steady-state kinetics experiments were performed using 20 mM p-nitrophenylphosphate (pNPP) as the substrate at 3.5 °C to resolve the initial burst phase on an Applied Photophysics SX40 instrument. The change of absorbance was measured at 410 nm for 200 ms and quantified using a pNP extinction coefficient of 1846 cm⁻¹M⁻¹. A total of 4000 points were collected for a single run. Specifically, two separate syringes with 120 – 260 mM of enzyme and 40 mM of pNPP were installed onto the instrument. The final concentration of enzyme of pNPP in the reaction chamber ranged from 60 – 130 μM and 20 mM, respectively.

The kinetic rate constants when [S] >> [E] were extracted by fitting the average of five runs to the exponential equation as described previously.³⁶ The absorbance was transformed and normalized by dividing the absorbance by the concentration of the enzyme and the extinction coefficient of pNP. Briefly, for the PTP reaction scheme in Figure 1A, R is the p-nitrophenyl group. For burst kinetics observed for PTP1B, the rapid increase in absorbance due to the time-dependent increase in [p-Nitrophenolate] upon cleavage and hydrolysis of the phosphate group is modeled as

$$[p-\text{nitrophenolate}]=\text{A}t+\text{B}\left(1-e^{-bt}\right)+\text{C}$$ (1)

In equation 1, t = time in seconds, b = (k₂ + k₃) and the linear portion, A = k₂k₃ / (k₂ + k₃), and B is the amplitude of the burst. Thus, fitting of the stopped flow data yields both forward chemical rate constants for cleavage (k₂) and hydrolysis (k₃). Stopped flow data was fit with the Kintec Explorer software package.

Steady-state kinetics with pNPP.

All steady-state pNPP kinetic experiments were performed at 30 °C in 50 mM Tris, 50 mM Bis-tris, 100 mM sodium acetate, pH 6.0 in triplicate. Initially, 100 nM of PTP1B was added to buffered solutions containing 0.4, 1.0, 2.0, 4.0, 8.0, 12.0, and 20.0 mM of pNPP. After 10, 20, 30, and 40 seconds, 200 mL aliquot of this mixture was added to 800 uL of 1.0 M NaOH to quench the reaction. The absorbance at each time point was measured at 405 nm and converted to concentrations using an extinction coefficient of 18000 cm⁻¹M⁻¹. The linear portion of this fixed time point assay was used to construct the initial rates, which were combined with [S] and the data fit with the Michaelis-Menten expression in Prism v10 (Graphad).

Steady-state kinetics with peptides.

Phosphorylated peptide substrates were purchased from Lifetein and Genescript in HPLC purified form at 95% purity and were used without further purification. Peptide cleavage/hydrolysis reactions were monitored continuously at 282nm, which shows and absorbance for tyrosine when pY is dephosphorylated.³⁷ For the peptide stock solution, Insulin Receptor (IR) peptide (sequence Ac-TDpYYRKG-NH₂), modified IR peptide (sequence Ac-TDpYYRAG-NH₂), Janus kinase2 (Jak2) peptide (sequence Ac-KEpYYKVK-NH₂), and p62dok peptide (sequence Ac-TALpYSQVQ-NH₂) were dissolved in 50 mM Tris, 50 mM Bis-tris, 100 mM sodium acetate at a pH of 6.0, 7.0, 6.0, and 5.0, respectively, to prevent aggregation. For each peptide pY is the location of the phosphotyrosine group and Ac (NH₂) are acetyl (amino) groups. The reaction solution was prepared by adding appropriate amount of the stock solution to the kinetics buffer (50 mM Tris, 50 mM Bis-tris, 100 mM sodium acetate, pH 6.0) so that the final peptide concentration was 30 – 130 μM for IR, 30 – 100 μM for modified IR, 50 – 350 μM for Jak2, and 40 – 150 μM for p62^dok. The pH in the final reaction was 6.0. To initiate the reaction, 25 nM of PTP1B was added to the buffered peptide solution at 30 °C and the absorbance was monitored continuously at every 0.2 s at 282 nm. The reaction was performed in triplicate at four difference substrate concentrations. The entire progress curves for all [S] were globally fit with Prism version10 using the Lambert function^38,39 approximation of the integrated rate expression eq 2.

$$P=S-K_m\cdot \text{ln}1.2\frac{\frac{s}{K_m}{e}^x}{\text{ln}2.4\left(\frac{S}{K_m}{e}^x\right)\text{ln}\left(1+2.4\cdot \frac{S}{K_m}{e}^x\right)}+\mathit{\text{offset}}$$ (2)

to obtain k_cat and K_m values.

The four optimal concentrations of peptides used above were determined by first estimating the kinetic constants in trial experiments with 50 and 100 μM of peptides and 25 nM of WT enzyme followed by fitting the progress curves to obtain an initial k_cat and K_m. A Monte-Carlo simulation was then performed with the estimated k_cat and K_m including 5% random uncertainty to determine the optimal values of [S] to use.

NMR Experiments.

¹H-¹⁵N TROSY and ¹H-¹³C methyl TROSY two-dimensional spectra for WT and mutant PTP1B were acquired at 14.1 T. ¹⁵N spectra were acquired with 3412 (1706) t₁ (t₂) points with spectral widths of 10000 (2500) Hz respectively. ¹³C spectra were acquired with 3778 (1889) t₁ (t₂) points with spectral widths of 8503 (3200) Hz respectively. All NMR experiments were conducted at 19° C as calibrated with a methanol standard and at pH = 6.8. ¹³C methyl multiple quantum relaxation dispersion experiments⁴⁰ were conducted at 16.5 T with 3778 (1889) t₁ (t₂) points, spectral widths of 8503 (4400) Hz, carbon carrier frequency at 19.5 ppm, using 20 ms constant time relaxation period, with τ_cp values of 0.41667, 0.5, 0.45455, 0.625, 0.55556, 0.7143, 1, 1.25, 1.66667, 2.5, and 5 ms. Relaxation rates were determined using Chemex. (Reference Bouvignies, G. Gbouvignies/ChemEx, 2025. https://github.com/gbouvignies/ChemEx). Amide (¹H/¹⁵N) chemical shifts were taken from the BioMagResBank (accession #5474)⁴¹ and methyl (¹H/¹³C) ILV chemical shift assignments were obtained from Torgeson⁴² and from Cui³⁵.

MD Simulations.

All molecular dynamics (MD) systems were constructed based on the 1.95 Å crystal structure of human PTP1B (PDB ID: 1SUG), comprising residues 1–301. Three systems were prepared: wild-type (WT), E297A, and I275A. Each system was modeled using psfgen and solvated in a cubic box of TIP3P water (90 × 90 × 90 ų),⁴³ with 0.15 M NaCl added to mimic physiological ionic strength and ensure system neutrality.

Simulations were performed in NAMD⁴⁴ using the CHARMM36m⁴⁵ force field for proteins. Each system underwent a three-stage equilibration protocol: (1) 1 ns with water and ions relaxed while restraining protein atoms, (2) 5 ns with side chains and solvent relaxed and the backbone restrained, and (3) 5 ns with all atoms unrestrained. Production trajectories were run for 200 ns × 8 replicas per system.

The temperature was maintained at 310 K using Langevin dynamics (damping coefficient γ = 1.0 ps⁻¹), and pressure was controlled at 1 atm using an anisotropic Langevin piston barostat⁴⁶. Simulations used a 2-fs integration time step. Bonded and nonbonded interactions were computed every step. Lennard-Jones interactions used a 12 Å cutoff with a switching function applied from 10–12 Å. Electrostatics were evaluated using the particle mesh Ewald method,⁴⁷ in which pairwise interactions were computed directly within 12 Å, and interactions beyond this distance were calculated on the Ewald mesh, updated every other step.

MD trajectories were analyzed using VMD,⁴⁸ MDiGest,⁴⁹ and MDAnalysis^50,51. All RMSF, correlation, and community calculations were performed on Cα atoms for each residue. Per-residue RMSF values for each simulation were computed for Cα atoms using VMD and were averaged over eight replicas for analysis. RMSF percent differences from WT and CSP data were mapped onto the protein structure using MDAnalysis and were visualized using Pymol⁵².

Pairwise correlations and eigenvector centralities⁵³ were computed using the generalized correlation coefficients⁵⁴ derived from mutual information⁵⁵ using default parameters as implemented in MDiGest. The resulting correlation matrices and eigenvector centralities were each averaged over eight replicas for each protein for analysis.

Communities were generated from the averaged correlation and distance matrices using the Louvain heuristic scheme⁵⁶ with 20 iterations as implemented in MDiGest. Communities smaller than five residues were redistributed to maximize modularity as implemented in MDiGest. Community structure visualizations and the corresponding eigenvector centrality plots were generated using MDiGest, Pymol, and Matplotlib.⁵⁷

Distance and dihedral measurements were performed using MDAnalysis. For residues with multiple equivalent atoms, the center of mass of those atoms were used to calculate distances. The corresponding histograms for each protein were generated using pandas⁵⁸, NumPy⁵⁹, and Matplotlib using all measurements from the eight replicas. Bin sizes of 0.1 Angstroms and 3.6 degrees were used for distance and dihedral histograms respectively.

Results and Discussion

Allosteric mutants alter PTP1B-substrate interactions.

We generated single alanine point mutants of PTP1B at positions I275, M282, E297, and Y153 previously identified as residues that influence catalytic turnover.³⁴ The locations of these mutations are shown in Figure 1D. All are distal from the nucleophilic C215 in the P-loop, with distances of 30 Å (M282), 23 Å (I275), 21 Å (E297), and 15 Å (Y153). Each mutant was expressed successfully and retained catalytic activity.

Kinetic parameters for phosphotyrosine-containing peptides were determined by global fitting of full progress curves at multiple substrate concentrations, monitoring absorbance changes at 282 nm.³⁷ For the small-molecule substrate p-nitrophenylphosphate (pNPP), parameters were obtained from linear-rate profiles using standard Michaelis-Menten analysis. The results are summarized in Figures 2, 3 and SI Figures 1 and 2.

Figure 2.

Steady-state and pre-steady kinetics of PTP1B. (A) Steady-state (left) and stopped-flow (right) experiments were performed for WT and allosteric mutants with pNPP as the substrate. (B) Progress curves representing the formation of dephosphorylated peptides as a function of time with WT PTP1B. Black lines are the non-linear least squares fit to the gray data points. The peptide sequences are shown at the top of each panel. Identical experiments for the mutants are shown in the SI.

Figure 3.

Bar graph representation of the catalytic efficiency of k_cat/K_m with different PTP1B peptide and pNPP substrates. The substrate is indicated at the top of each panel and highest (lowest) k_cat/K_m values are indicated in red (blue). The mutant PTP1B enzyme is indicated at the bottom of each panel.

Specificity constants (k_cat/K_m) for peptides substrates ranged from 5 × 10⁵ M⁻¹s⁻¹ to 8 × 10⁶ M⁻¹s⁻¹, consistent with previously reported values for PTP1B.³⁷ However, the k_cat/K_m values vary substantially among mutants and substrates, in some cases switching the relative order of substrate specificity. For example, for the IR peptide, I275A displayed the highest k_cat/K_m while E297A was lowest. In contrast, with the p62dok peptide, E297A had the highest k_cat/K_m and I275A the lowest. Overall, WT PTP1B generally exhibited the highest k_cat values for peptide substrates, with the exception of the IR variant peptide (SI Fig 1 & 2). For pNPP, WT showed a lower k_cat than I275A but higher than the other mutants. Notably, in most mutants, reduced K_m values compensated for lower turnover numbers, resulting in elevated k_cat/K_m values relative to WT (Figure 3, SI Fig 2). These steady-state data demonstrate that allosteric mutations differentially affect kinetic constants across substrates, indicating that distal sites can rewire PTP1B substrate specificity.

The E297A mutant shows the largest deviations in k_cat/K_m values relative to WT, with the exception of the Jak2 substrate (SI Figure 2). Specifically, k_cat/K_m values for E297A differ from WT by −18%, +61%, +24%, and −14% for the IR, p62^dok, mutant IR, and Jak2 substrates, respectively. Y153A and I275A also exhibit deviations from WT, though to a lesser extent than E297A, with both increases and decreases observed depending on the substrate. In contrast, M282A consistently yields higher k_cat/K_m values than WT, ranging from only 0.2% above WT for the IR substrate to 31% above for p62dok. Collectively, these data indicate that substrate-dependent differences in k_cat/K_m between WT and mutants are substantial. Notably, the kinetic trend observed with the small-molecule substrate pNPP diverges from those of the peptide substrates. Given this distinction, we examined pre-steady-state kinetics of WT and mutant enzymes using pNPP as the substrate (Fig. 2A).

At 3.5° C, pre-steady-state analysis yielded k₂ (cleavage) and k₃ (hydrolysis) values of 120 ± 6 s⁻¹ and 15.0 ± 0.1 s⁻¹ for WT, 124 ± 4 s⁻¹ and 12.1 ± 0.1 s⁻¹ for I275A, and 111 ± 5 s⁻¹ and 11.8 ± 0.1 s⁻¹ for M282A. The close agreement among these values indicates that the I275A and M282A have minimal effects on the individual cleavage and hydrolysis steps for pNPP. This conclusion is consistent with steady-state kinetic measurements at 30°C, where pNPP k_cat values were 68 ± 2 s⁻¹ (WT), 73 ± 1 s⁻¹ (I275A), and 65 ± 1 s⁻¹ (M282A). Similarly, the K_m values for I275A (1.0 ± 0.1 mM) and M282A (1.2 ± 0.1 mM) were comparable to WT (1.0 ± 0.1 mM).

In contrast, mutations Y153A and E297A in the allosteric region (Fig. 1C, D)³⁴ produced pronounced reductions in catalytic efficiency. Both mutants exhibited approximately two-fold decreases in k₂ (62 ± 3 s⁻¹ for Y153A; 71 ± 7 s⁻¹ for E297A), and in k₃ (7.0 ± 0.04 s⁻¹ for Y153A; 8.6 ± 0.1 s⁻¹ for E297A), along with a two-fold increase in K_m (2.4 ± 0.1 mM for Y153A; 2.0 ± 0.1 mM for E297A).

Taken together, these experiments demonstrate that the allosteric mutants not only alter the individual rate constants for pNPP relative to WT but also modulate substrate specificity (k_cat/K_m) across different peptide substrates. This suggests that the allosteric sites play a direct role in regulating PTP1B-peptide interactions. To further elucidate the connection between the allosteric and active sites, we employed solution NMR spectroscopy in combination with μs-timescale molecular dynamics (MD) simulations. For the remainder of this work, we focus on E297A, which exhibits the most pronounced deviations from WT, and I275A, which behaves similarly to WT with the IR peptide but diverges with the other substrates.

Allosteric mutations have distinct effects across the PTP1B structure.

We mapped the effects of allosteric mutations on the active site and substrate-binding pocket using solution NMR spectroscopy and μs-MD simulations. Specifically, we combined NMR chemical-shift perturbation (CSP) analysis, NMR relaxation experiments, and differential root mean square fluctuations (ΔRMSF) analysis of the protein backbone from MD simulations (Figure 4).

Figure 4.

NMR and MD comparison between PTP1B enzymes. ¹⁵N(amide), ¹³C(ILV) NMR chemical shift perturbations (CSP) between apo WT and apo I275A/E297A and the root-mean-square-fluctuation differences obtained from MD simulation in the bottom panel. The differences are shown as a function of amino acid residue number. For the CSP plots, prolines, unobserved and overlapped residues are shown as negative black tics. Amino acid residues whose chemical shifts change significantly making their assignments ambiguous are indicated as blue bars. That these residues’ chemical shifts move significantly indicates they are impacted by the mutation, therefore we color them differently from the other residues. Residues with a CSP above the 1.5 standard deviation over the 10% trimmed mean (horizontal black dotted line) are colored in red. The RMSF difference was obtained by subtracting the average WT RMSF from the mutant’s. Errors were computed for each residue using the standard error of the mean over eight replicas and are indicated with shaded regions surrounding the data points. Data representing I275A and E297A are colored in red and blue, respectively. The PTP1B secondary structure is indicated at the top of the figure and loop regions are indicated with colored bars as in Figure 1B & C. The location of the I275A and E297A mutants are shown as red and blue stars, respectively.

The NMR experiments probed both the amide backbone using ¹⁵N-based experiments and the hydrophobic sidechains using ¹³C-based experiments targeting the methyl groups of isoleucine, leucine, and valine (ILV). CSP values, calculated as the differences between WT PTP1B and the I275A and E297A mutants, are shown in Figure 4. Statistically significant CSP are highlighted as red bars and mapped onto the three-dimensional structure of PTP1B (Fig. 5). While I275A and E297A share some similarities in their CSP profiles, they also display distinct and mutation-specific differences.

Figure 5.

Mutation induced changes in PTP1B. (Left) Amide CSP from Figure 4 are plotted onto the crystal structure of PTP1B. The magnitude of the CSP difference is depicted with a gray-to-red color change and an increase in radius of the backbone sausage. Residues exhibiting significant amide CSP above the 1.5 standard deviation over the trimmed mean are colored in dark red. (Right) PTP1B Root-mean-square-fluctuation percent differences (ΔRMSF) from WT are plotted onto the crystal structure, with the blue-to-red color gradient indicated more rigid than WT to more flexible than WT respectively. The increase in radius also represents the size of the ΔRMSF values by percentage. Residues that are overlapped, not observable, or unassigned are indicated in black. The P-loop, acid loop, and site of mutation are noted with arrows.

For the I275A mutant, significant chemical shift changes are observed at the N-terminal helices α1’ and α2’ (residues 1–30), within the acid-loop, and extensively throughout the C-terminal region beyond α4, including the Q-loop, α6, and α7. Additional perturbations are detected at K41, A77, L144, T178 and F182. Importantly, differences in chemical shifts in acid-loop residues T178 and F182 indicate that this mutation alters the chemical environment of this remote, catalytically essential element. The strong perturbations at the N-terminus are consistent with structural packing interactions, as the I275 Cδ atom lies within 3.0 Å and 3.3 Å from the γ and β carbons of M3 and E4, respectively.

Complementarity between NMR and MD techniques is evident in this mutant. Both approaches reveal altered dynamics at the N-terminal region, α6 and α7, as well as the acid loop (Figs. 4 & 5). However, MD simulations additionally capture deviations between I275A and WT not apparent by NMR, particularly in the E-loop, pTyr loop, Q-loop, and P-loop. Notably, the P-loop is typically rigid in both NMR and computational studies of the WT enzyme,⁶⁰ suggesting that the I275 mutation perturbs multiple aspects of the active site structure and dynamics beyond local packing interactions.

The E297A mutant exhibits more extensive chemical shift perturbations in α7 than I275A, consistent with the location of E297 within this helix. In contrast, α5 shows fewer perturbed residues in E297A than in I275A. Both mutants display CSP in the acid-loop (T177, F182, G183, and V184) and Q-loop (L260, I261, and T263). Simulations suggest that each mutation induces rigidification of the acid loop on the fast timescale (Fig 4, 5). However, the pTyr recognition loop residues in E297A exhibits a smaller increase in RMSF relative to WT.

At the structural level, E297 disrupts the hydrogen bond linking α7 and α3 by eliminating the interaction between the E297 sidechain O^ε and the N193 side-chain with N^δ2 (Fig. 6A, 6E). By comparison, I275A perturbs packing between α6 and the N-terminal helices α1’ and α2’ (Fig. 6D); distance distributions for molecular interactions in this region are significantly broadened and shifted towards shorter distances for I275A (SI Fig. 3). Consistent with these computational results, NMR relaxation dispersion experiments reveal changes near α3 at L195 in both mutants (Fig. 6B). E297A further produces unique dynamical effects in α7. Relaxation dispersion and R₂ analyses show marked perturbations for L294 and V287, with residues involved in hydrophobic packing within the α7 C-terminus (Fig. 6C & SI Fig. 5). Specifically, the V287 ¹³C^γ1(¹³C^γ2) R₂ values for WT PTP1B are 52.2 ± 0.6 s⁻¹ (65.5 ± 1.2 s⁻¹), whereas in E297A and I275A these rates are, respectively, 25.9 ± 0.1 s⁻¹ (34.9 ± 0.5 s⁻¹) and 43.4 ± 0.5 s⁻¹ (unassigned in I275A). Likewise, for L294 the ¹³C^δ1(¹³C^δ2) R₂ values for WT, E297A, and I275A are 12.9 ± 0.04 s⁻¹ (24.9 ± 0.1 s⁻¹), 15.1 ± 0.02 s⁻¹ (15.6 ± 0.03 s⁻¹), and 14.0 ± 0.05 s⁻¹ (23.2 ± 0.2 s⁻¹), respectively. These data confirm that both mutations perturb the dynamics of α7, with E297A exerting the stronger effect, as shown by the R₂ values (SI Table 3).

Figure 6.

α3 and α7 helices are connected by a hydrogen bond network. (A) The distance profile from MD simulations between O^ε of the E297 sidechain with N^δ2 sidechain of N193. Carr-Purcell-Meiboom-Gill relaxation dispersion of apo PTP1B collected at 800 MHz with a total time of 20 ms showing (B) L195^δ1, (C)V287^γ1, and (D) L294^δ2. (E) Snapshot from the MD simulation showing that E297 interacts with (1) S151, (2) Y153, (3) Y152 in L11 and (4) N193 in α3, indicated by dotted lines.

Notably, residues L294 and V287 as well as helix α7 more broadly, have been shown to rigidify upon binding of the active-site inhibitor TCS-401.⁴² Previous studies also established the role of α7 in PTP1B allosteric communication, whereby active-site ligand binding stabilizes the C-terminus and dampens conformational exchange throughout PTP1B.⁴² Together, these findings reinforce the central role of α7 as a conduit linking distal mutations and allosteric regulation to the catalytic core.

Previous computational studies of WT PTP1B identified W291 in α7 as a structural anchor within the hydrophobic pocket formed by F280 and A189.⁶¹ Disruption of this region in the E297A mutant strengthens the interaction between W291 and F280, as observed in MD simulations, while simultaneously increasing the distance between W291 and A189, which lies at the C-terminal region of the acid loop (SI Figure 4A). This altered packing is accompanied by decreased ¹³C R₂ values for nearby I281 (Fig 7, SI Table 3) and by substantial changes in the χ1 and χ2 dihedral angle distributions of W291, indicating conformational rearrangements of this aromatic side chain (SI Fig. 4B).

Figure 7.

(A) Bar graph of the transverse relaxation rates of I281 in apo PTP1B at a τ_cpmg of 0.41667 ms at 800 MHz and 20 ms total relaxation time. (B) Carr-Purcell-Meiboom-Gill relaxation dispersion of apo PTP1B collected at 800 MHz with a total time of 20 ms showing V184^γ1 and I281.

Other residues within this hydrophobic cluster, including L192, L195, and L272, also show significant reductions in R₂ values in both I275A and E297A compared to WT (SI Table 3). Notably, these residues are located near the allosterically critical “197” site identified by Keedy et al,²⁹ and L192 independently identified in computational work by Kamerlin and coworkers as part of a core allosteric communication network that governs PTP1B catalytic activity.⁶⁰ Collectively, our data reinforces these findings, showing that perturbations in α7 propagate through this conserved residue network, ultimately relaying allosteric effects to the active site.

The altered interaction between A189 and W291 is particularly notable, as A189 resides at the C-terminus of the acid-loop. Additional perturbations are observed in this region, including changes to contacts between S187 and residues 190 and 191, which cap α3 and serve as a hinge point for acid-loop motion (SI Fig. 4C).³¹

Both I275A and E297A mutants exhibit reduced ¹³C R₂ values for residues near the acid loop compared to WT, with the strongest effect in E297A. For example, the ¹³C^δ1(δ2) R₂ values of L110 decrease by 28 s⁻¹ (36 s⁻¹) in E297A relative to WT values of 63.3 ± 1.5 s⁻¹ (75.5 ± 3.6 s⁻¹) (SI Table 3). In contrast, the reductions in I275A are more modest (9 s⁻¹ and 8 s⁻¹). L110 resides in the E-loop, which has been shown by MD simulations to exhibit coordinated motions with the P-loop in the active site and aids in acid loop closure.⁶⁰ Residue L110 lies in close proximity to the active site – 5 Å from the catalytic C215 and 6 Å from the N-terminal hinge of the acid loop – positioning it as a key structural mediator. Thus, while both mutations perturb PTP1B dynamics, E297A exerts the more pronounced effect, consistent with its stronger impact on catalysis.

This trend extends to V184 within the acid loop, where ¹³C relaxation dispersion profiles are similar for WT and I275A but flat for E297A, indicating a loss of millisecond-timescale motions for E297A (Fig. 7). Complementary MD simulations also reveal rigidification of the acid-loop on the faster nanosecond timescale, as reflected in negative ΔRMSF values (Fig. 4). These dynamical changes propagate beyond the acid-loop, influencing other elements of the active site. Further support comes from ¹³C multiple-quantum methyl relaxation dispersion experiments, where E297A shows no dispersion in contrast to WT, underscoring a loss of conformational exchange in the acid loop of E297A in the millisecond timescale.

To further probe the impact of I275A and E297A on the PTP1B active site, we analyzed chemical shift changes upon binding of the transition-state analog vanadate (VO₄²⁻).^62,63 A correlation plot of VO₄²⁻-binding induced methyl ¹³C chemical shifts for WT, E297A, and I275A is shown in Figure 8 and SI Figure 6, with notable deviations highlighted. For I275A, affected residues include I261, L250, and L299. Importantly, I261 lies adjacent to Q262, the catalytic glutamine residue that orients a water molecule for the hydrolysis step during catalysis. The sidechain of L250 is involved in direct hydrophobic packing with the sidechain of I261, while L299 is located in α7, further highlighting the aforementioned coupled interactions between the active site and α7. In E297A, deviations also involve I261 along with I57, V113 (E-loop), V213 (P-loop), I275, I281, V287 and L294.

Figure 8.

Correlation plots showing the vanadate-induced chemical shift changes comparing WT and mutant enzymes (I275A red, E297A blue). (A) ¹⁵N^H (left) ¹H^N (right) chemical shift changes (apo – bound). In each panel residues with significant deviations from the diagonal, indicating mutant-specific differences in response to vanadate binding are circles. For ¹⁵N^H (left), these residues are S80 (Group1), L232, I261, T263, A264, W291, D298 (Group 2), G277 and G283 (Group 3). For ¹H^N (right) these residues are A264 (Group 1), G183 (Group 2), W16 (Group 3), and G283 (Group 4). (B) ¹³C^me (left) ¹H^me (right) chemical shift changes (apo – bound). In each panel residues with significant deviations from the diagonal, indicating mutant-specific differences in response to vanadate binding are circles. For ¹³C^me (left) these residues are I261 (Group 1), V287 (Group 2), L250 and L299 (Group 3 and 4). For ¹H^me (right) these residues are L110 and V155 (Group 1), L59 and V212 (Group 2), L192, I219, L250, I281, V287. The perpendicular distance between the data points and the diagonal were measured. Those above 1.5 standard deviations from the 10% trimmed mean are mapped onto the crystal structure (PDB entry 3I80) in spheres with a color gradient in (C) and (D), where the VO4²⁻ in the center is marked in blue. The sites of mutation are indicated with arrows.

Analysis of the VO₄²⁻-binding induced methyl proton chemical shift changes (Fig. 8B) revealed a similar set of affected residues for I275A, including L250, L192, L195, I219, and I281. Notably, I219 is located in the P-loop, while L192 is near the acid loop and packs against α6, suggesting that mutation at I275 perturbs ligand interactions at the active site. I281, positioned at the C-terminal end of α6, is also affected. For E297A, deviations were detected at L110 in the N-terminus of the E-loop, V155 in “197” site,²⁹ and I281 in loop L18.

The correlation plot of ¹⁵N amide chemical shifts (Figure 8A) from VO₄²⁻- binding reveals additional perturbations in the protein backbone of the I275A and E297A variants. Affected residues include L232, A264, and K292. The connection between L232 in helix α4 and the acid loop has been reported previously, as this residue consistently exhibits perturbations upon mutations within the acid loop.³⁴ Similarly, residues near Q262, including A264 show clear chemical shift perturbations. K292, located in the allosteric helix α7, also displays significant deviations relative to the wild type.

In the E297A variant, chemical shit deviations extend to multiple regions: S80 in β3, T177 in the acid loop, T263 in Q-loop, G283 in L18, as well as W291, K292, and D298 in α7. Notably, S80 lies within the 78–95 peptide segment, which undergoes enhanced amide exchange with deuterated buffer upon binding of the allosteric inhibitor BB3.²⁸

Amide proton shift deviations (Figure 8A) further highlight structural disruptions in I275A, particularly in the N-terminal helices α1’ and α2’ (D11, W16). Additional effects are observed at A77 (immediately preceding β3), A264 (following the Q-loop), and A278 (within α6). For E297A, significant perturbations are detected in I10 (in α1’), A27 (after α2’), G183 and V184 (acid loop), A228, Q262, V274 (in α6), and G283 (in L18).

Alteration of Allosteric Communities.

We have investigated allosteric interactions in PTP1B using community network analysis derived from molecular dynamics simulations.^64,65 In WT PTP1B (Fig. 9 and SI Fig. 7), the acid loop and α7 helix cluster belong to the same community (light brown) and are dynamically coupled, consistent with prior reports.^{27,29,31,35,42,61} In both E297A and I275A, however, this community splits: α7 forms an independent community, indicating destabilization of α7 and loss of coupling with the acid loop. Previous studies have shown that destabilization of α7 promotes the open (inactive) conformation of the acid loop.^{27,29,31,42,61} Although the mutation sites are about 20 Å apart, E297A and I275A may inhibit PTP1B activity via a similar mechanism.

Figure 9.

Allosteric communities derived from the generalized correlation coefficient from 1.6 μs of MD simulations. Communities are grouped by color. Note the change in the α7/α3 community in WT in favor of the lone α7 community in I275A and E297A. E297A also shows a switch in the community to which the P-loop belongs whereas the P-loop in I275A remains as it is in WT PTP1B.

Differences emerge in the L11 community (dark brown), another key feature in the allosteric regulation of acid loop dynamics. In WT PTP1B, the L11 community (dark brown) spans the central β-sheet adjacent to the acid loop, E-loop, and P-loop. β-sheets have been shown to facilitate long-range correlations in proteins, which could link L11 to these important loops.⁶⁶ In I275A, L11 is isolated from the β-sheet, while in E297A, L11 and the central β-sheet merge with the S-loop community (red), which is directly C-terminal to the α3 helix. Both NMR and computational studies have implicated this region in modulating acid loop dynamics and PTP1B activity.^67–70 Thus, differences in the allosteric organization of L11 likely contributes to the distinct catalytic behavior of the two mutants.

Beyond these structural rearrangements, mutations also reshape the allosteric network governing substrate recognition. In particular, the pTyr loop community (orange in WT and I275A) splits into two communities in E297A. Since the pTyr loop harbors key residues for peptide binding, its dynamic decoupling from adjacent loops in E297A is likely to alter binding affinity. In WT PTP1B, the Q-loop and the N-terminal region of α6 belong to the same community (gold), which also includes G220 and I219 of the catalytic P-loop. By contrast, in both E297A and I275A, the Q-loop and these P-loop residues disengage from α6 and instead couple with the c-terminal end of α4 (light pink). Both α4 and α6 have been implicated in regulating PTP1B activity;^31,34,35,69 suggesting that altered connectivity between these elements (i.e., the Q-loop and P-loop) contributes to changes in activity and substrate recognition.

In WT PTP1B, the P-loop and E-loop belong to the same community (light green). However, in both E297A and I275A, the two loops separate into distinct communities. Prior work has shown that the E-loop is coordinated with the P-loop, and its motions are further correlated with the acid loop in WT PTP1B.^60,71 Furthermore, increased flexibility in the E-loop (Figure 4) has been suggested to influence both substrate binding and catalytic activity.⁶⁰ Altogether, the differential alterations in catalytically important loops — the acid loop, pTyr loop, Q-loop, P-loop, and E-loop — are likely to contribute to altered substrate-specificity. To identify the most influential structural elements within the correlation network, we computed eigenvector centrality from the correlation coefficient matrix.⁵³ Eigenvector centrality can measure the strength of each residue’s connectivity and communication in the correlation network, where a large eigenvector centrality represents a large influence on the network.⁵³ Previous computational studies using dynamic cross-correlation identified the acid loop and nearby regions, such as α3, as displaying high eigenvector centrality.⁷¹ Consistent with these findings, our analysis highlights high centrality for residues in the acid loop, as well as in α7 (SI Figure 8). Although α3 residues exhibit low centrality in WT PTP1B, both E297A and I275A display the largest increases in centrality within α3 and α7 compared to WT. Prior studies have reported dynamic coupling between these helices, suggesting they jointly regulate acid loop dynamics.^31,35,42,61 Despite evidence of α7 destabilization in both mutants, its motions remain central to PTP1B dynamics. Importantly, different communities exhibit distinct centrality changes across mutants: I275A shows large increases in the α1’ helix (dark blue), whereas E297A shows minimal changes in α1’ but strong centrality increases in the Q-loop/P-loop community (light pink). These mutation-specific alterations in network centrality ultimately lead to differential substrate interactions and altered catalytic constants.

Conclusions

Enzymes exist as ensembles of conformations, each with distinct ligand-binding affinities and catalytic efficiencies. To rationalize the observed changes in kinetic parameters, we propose a kinetic model analogous to those described previously.^72,73 Our kinetic, computational, and NMR data indicate that K_i,1 is biased in WT to favor the E* state relative to I275A and E297A. This shift accounts for the primary effect of the mutants—reduced peptide K_m values compared to WT—summarized in Scheme 1.

Scheme 1.

Kinetic model for PTP1B considering alternate conformations in both apo and substrate (ligand)—bound states.

From Scheme 1, a Michaelis-Menten-like expression can be derived:

$$v=\frac{V_{m,eff}[S]}{K_{m,eff}[S]}$$ 3

Here, the effective kinetic constants are modified by additional equilibria (K_i,1 and K_i,2). We propose that the mutations alter these equilibria, shifting the balance between more versus less active and binding-competent versus incompetent ensembles. Moreover, the distinct effects on k_cat suggest that mutations differentially perturb K_i,2 in a substrate-dependent manner.

Accordingly, the effective parameters are:

$$K_{m,eff}=K_m\frac{1+{\left(K_{i,1}\right)}^{-1}}{1+{\left(K_{i,2}\right)}^{-1}}$$ 4

$$V_{m,eff}=\frac{V_m}{1+{\left(K_{i,2}\right)}^{-1}}$$ 5

with inhibitory equilibrium constants defined as K_i,1 = [E]/[E*] and K_i,2 = [ES]/[(E**S)].

Overall, these findings demonstrate that allosteric mutations reshape the PTP1B energy landscape to regulate both substrate binding and catalysis. They further suggest that studies of enzyme allostery based on small-molecule substrates may not fully capture the cellular function of an enzyme compared to the more natural peptide substrates.

Supplementary Material

SI Table 1 containing oligonucleotide sequences; SI Table 2 containing enzyme kinetic data; SI Table 3 containing NMR relaxation rates; SI Figure 1 showing enzyme reaction progress curves; SI Figure 2 showing fits to kinetic data from SI Figure 1; SI Figure 3–4 showing analysis of Molecular Dynamics simulations; SI Figure 5 showing NMR relaxation dispersion curves; SI Figure 6 showing NMR chemical shift perturbations; SI Figure 7 showing Eigenvector analysis of WT and mutants;