Biochemical Methods to Map and Quantify Allosteric Motions in Human Glucokinase

Abstract

Allosteric regulation of protein function is ubiquitous in biology. Allostery originates from ligand-mediated alterations in polypeptide structure and/or dynamics, which produce a cooperative kinetic or thermodynamic response to changing ligand concentrations. Establishing a mechanistic description of individual allosteric events requires both mapping the relevant changes in protein structure and quantifying the rates of differential conformational dynamics in the absence and presence of effectors. In this chapter, we describe three biochemical approaches to understand the dynamic and structural signatures of protein allostery using the well-established cooperative enzyme glucokinase as a case study. The combined application of pulsed proteolysis, biomolecular nuclear magnetic resonance spectroscopy and hydrogen-deuterium exchange mass spectrometry offers complementary information that can used to establish molecular models for allosteric proteins, especially when differential protein dynamics are involved.

1. Introduction

Allostery is the process by which biomolecules transmit the impact of ligand binding at one site to another, often distal, functional site, allowing for regulation of activity. Allosteric regulation of protein function is a key mechanism for controlling biological processes (Guo and Zhou, 2016). In particular, allosteric regulation of enzyme activity plays an essential role in controlling metabolic flux and resource allocation. Textbook models typically describe allosteric regulation in terms of measurable changes in static, end-point protein structures. According to this view, effector binding toggles the enzyme between high-activity and low-activity states (Changeux, 2012). While these historic models have proven useful in developing a conceptual model to understand ligand-mediated control over enzyme activity, a growing number of mechanistic investigations demonstrate that allosteric regulation does not always require alterations of a protein’s three-dimensional structure (Motlagh et al., 2014; Tsai et al., 2014). It’s clear that a contemporary view of allostery must include effector-mediated alterations to protein intrinsic dynamics as well as structure (Popovych et al., 2006; Ferreon et al., 2013). As such, experimental methods to interrogate and map ligand-mediated changes in protein dynamics have become an increasingly important tool for understanding the mechanistic basis of allosteric regulation.

Mammalian glucokinase has emerged as a prototypic example of an allosteric enzyme in which protein dynamics play a key role in regulating catalytic activity. Understanding allosteric regulation of glucokinase is important to human health, as its activity mediates insulin secretion from pancreatic β-cells (Matschinsky, 1990). Glucokinase is a monomeric enzyme with a single binding site for its allosteric effector, glucose. The enzyme displays a mild cooperative kinetic response to increasing glucose concentrations that is manifested by a Hill coefficient of ~1.7. Allostery in glucokinase is fundamentally different from classical, equilibrium models. Unlike systems such as hemoglobin, cooperativity is solely kinetic in origin — ligand binding to the enzyme is not cooperative. Missense mutations in the glk gene that abolish cooperativity result in persistent hypoglycemic hyperinsulinemia (Gloyn, 2003). In addition, synthetic small-molecule glucokinase activators that reduce cooperativity have received attention as potential therapeutic agents to treat maturity onset diabetes of the young (Grimsby et al., 2003). Activators associate with a site near the hinge region of the protein, far from the glucose binding site, and function to reduce conformational heterogeneity and suppress millisecond dynamics. Due to its importance in glucose homeostatic maintenance, glucokinase has been the subject of numerous structural, biochemical, and biophysical investigations aimed at elucidating the molecular origins of its kinetic cooperativity (Larion and Miller, 2012).

Past studies in our lab and others have provided a molecular framework for understanding the mechanistic basis of glucokinase allostery. In this model, the unliganded enzyme transitions between at least two distinct conformations with a rate constant for exchange, k_ex, that is comparable to k_cat. Cooperativity in glucokinase is purely a kinetic phenomenon and mutations that alter the equivalency between k_ex and k_cat reduce cooperativity (Whittington et al., 2015; Sternisha et al., 2018). Two mechanistically distinct modes of activation have been identified in hyperactive, non-cooperative disease relevant enzyme variants. These mechanisms differ via their effects on either the protein conformational landscape sampled by the unliganded enzyme, or via their impact upon the dynamics of enzyme conformational changes, especially within a disordered mobile loop that governs product release. The development of an integrated model for glucokinase cooperativity, and disease-relevant glucokinase activation required the application of several complementary experimental methods. Together, these facilitated both mapping and quantifying the structural dynamics associated with allostery. In this chapter, we outline the three primary biochemical methods that have proven useful for understanding the dynamic signatures of glucokinase cooperativity. Described below are the general approaches for each of these method - pulsed proteolysis, biomolecular nuclear magnetic resonance spectroscopy and hydrogen-deuterium exchange mass spectrometry. Each method provides unique insight into protein dynamics that are complementary in nature. We expect that the application of these methods, both alone and in combination, can provide important advances to understanding the molecular basis of allostery for any protein system in which dynamics are postulated to be important.

2. Quantitative Mapping of Local Dynamics with Limited Proteolysis

Local dynamics and loop dynamics are important for enzyme function and regulation. Loops and other intrinsically disordered regions are often difficult to characterize experimentally. Limited proteolysis provides a quantitative, experimental method to probe local dynamics of transiently populated high-energy conformations, such as ligand binding loops, that undergo local unfolding (Chang and Park, 2009; Fontana et al., 2004; Park and Marqusee, 2004). In this approach, the enzyme of interest is exposed to a generic protease for fixed time intervals. By using a generic protease, the barrier to cleavage of the dynamic region is local unfolding (Hubbard, 1998) rather than the sequence specificity of the protease (Fig. 1A). Cleavage progress can be measured by SDS-PAGE analysis and/or enzyme assays if the function of the enzyme is disrupted by proteolysis.

Figure 1.

Quantitative measurement of local unfolding using limited proteolysis. (A) Under limited proteolysis conditions, the cleavage rate is governed by the rate of local unfolding of the cleavage site and the intrinsic rate constant for cleavage of the site. (B) The proteolytic cleavage pattern can be visualized via SDS-PAGE and the site of cleavage can be identified by N-terminal sequencing of the protein fragments. The intrinsic rate constant for cleavage, k_int, can be determined by measuring linear progress curves of cleavage of a fluorogenic peptide corresponding to the cleavage site (C) and fitting of the cleavage rates vs [peptide] to a linear equation (D). (E) Cleavage can be monitored by relative enzyme activities or gel quantitation producing exponential decay curves that give the observed cleavage rates. (F) Unliganded glucokinase displays a linear response to [thermolysin] allowing extraction of the equilibrium constant between the locally unfolded and folded states, K_op. (G) In the presence of saturating glucose, glucokinase displays a hyperbolic response allowing extraction of K_op and the rate constant of unfolding, k_op.

We have successfully used limited proteolysis to characterize the degree of local unfolding in an intrinsically disordered substrate binding loop (hereafter referred to as the mobile loop) in human glucokinase (Whittington et al., 2015). Cooperativity in glucokinase results from a matching of the rate of exchange between unliganded conformations and the rate of catalysis (Larion et al., 2015; Whittington et al., 2015). Previously, there was limited information on the structure and dynamics of this region despite its critical importance in glucokinase activity. We measured proteolysis kinetics for wild-type glucokinase and a series of disease associated mutations that lead to activation of glucokinase and loss of its characteristic cooperative response to glucose. Our use of proteolysis was key to the identification of two distinct activation mechanisms that disrupt cooperativity yet display identical glucose kinetic response curves, α-type activation leads to suppressed mobile loop dynamics and a more ordered structure, reducing or removing the exchange between unliganded conformations (Whittington et al., 2015). β-type activation leads to increased mobile loop dynamics increasing the rate of catalysis via a faster product release step (Whittington et al., 2015; Sternisha et al., 2018). These results underscore the potential for limited proteolysis in elucidating the functional basis of enzyme allosteric mechanisms.

2.1. Equipment

• Pipettes

• Microfuge tubes and glass culture tubes

• Vortex mixer

• Standard SDS-PAGE apparatus (BioRad MiniPROTEAN^® system)

• UV/Vis spectrophotometer (Cary Bio 100, Agilent)

• Fluorescence spectrophotometer (Cary Varian Eclipse, Agilent)

2.2. Buffers and Reagents

• Thermolysin from Bacillus proteolyticus rokko (Sigma T7902).

  • Diluted to 10 mg/mL in 2.5 M NaCl and 0.1 M CaCl₂ and stored at −20°C until use.

• Proteolysis assay solution: 50 mM HEPES, pH 7.6, 50 mM KCl, 10 mM CaCl₂, 10 mM DTT.

• 100 mM EDTA, pH 8.0 for quenching protease.

• Coomassie Brilliant Blue G Colloidal Stain (Sigma B2025).

• Purified recombinant protein of interest.

  • We used His⁶-tagged glucokinase that is purified by Ni-NTA (HisTrap, Cytiva) and size exclusion chromatography (Superdex 200, Cytiva) on an AKTA FPLC system (Cytiva).
• Enzyme assay system

  • For human glucokinase activity, we used an LDH/PK assay that links production of product to NADH oxidation (Larion and Miller, 2009).

2.3. Experimental Design and Procedure

2.3.1. Identification of cleavage site/s

1. Incubate 0.5 mg/mL glucokinase with 0.2 mg/mL thermolysin in the presence of 10 mM CaCl₂ for five minutes on ice.

2. Quench reaction by adding EDTA to 20 mM.

3. Separate proteins by running 10-50 μg of reaction per lane on an SDS-PAGE gel. We run 16.5% gels at 200 V for 1 h (Fig. 1B).

4. Transfer to a PVDF membrane at 20 V and 80 mA at 4°C for 1 h.

5. Stain with Coomassie and excise the bands with a razor blade.

6. Identify cleavage fragments. We used N-terminal sequencing via Edman degradation. For this, we used C-terminal His⁶ to prevent interference with fragment identification.

7. Glucokinase has one cleavage site (Fig. 1B). However, multiple cleavage sites could be identified using this method with carefully resolved fragments.

8. Alternative to Edman degradation, gel fragments could be identified by de novo peptide sequencing via mass spectrometry.

2.3.2. Determination of the intrinsic proteolytic rate constant (k_int)

1. To measure the intrinsic rate constant, k_int (equivalent to the second-order rate constant k_cat/K_M), for thermolysin proteolysis of the identified cleavage site, we ordered a four amino acid peptide corresponding to the site end-capped with a fluorescent molecule and a quencher (ABZ-K-G-F-K-pNA) from Peptide 2.0. Peptide was resuspended according to manufacturer’s instructions.

2. Measure linear progress curves of peptide cleavage at varying [peptide] and constant [thermolysin] (33 nM) by monitoring the increase in ABZ fluorescence (ext. λ = 323 nm; em. λ = 420 nm) over time (Fig. 1C) in proteolysis assay solution at room temperature. Initialize the reactions by addition of thermolysin. Linear fits of the progress curves yield cleavage rates in arbitrary units per min (a.u./min).

3. Linear fit of cleavage rates vs [peptide] yields k_int (Fig. 1D).

4. To convert a.u. to moles of peptide, incubate varying [peptide] with a μM amount of thermolysin in proteolysis assay solution to completely convert the peptide to the cleaved form. Linear fit of a.u. of the completed reaction vs [peptide] gives the conversion factor.

5. The presence of 200 mM glucose (a saturating concentration for human glucokinase) did not affect the k_int value.

2.3.3. Proteolysis and Determination of Proteolytic Rates

1. We empirically determined the amount of enzyme to use in our proteolysis assays to give a maximum rate of 0.5 absorbance units per min of enzyme activity in the enzyme assays. Additionally, this amount allowed visualization of the same proteolysis assays on SDS-PAGE gels.

2. Final reaction volumes were 500 μL.

3. We equilibrated N-terminal His⁶ tagged human glucokinase (5 – 10 μM) in proteolysis assay solution for 10 min at room temperature.

4. Proteolysis was initiated by addition of thermolysin. To measure proteolysis kinetics, [thermolysin] is varied and [glucokinase] is constant.

5. Time points were taken by removing 40 μL aliquots and adding them to microfuge tubes with 10 μL of 100 mM EDTA to quench the reaction.

6. Since glucokinase has one cleavage site (the mobile loop) and cleavage disrupts activity, we monitored proteolysis via enzyme activity. We assayed 5 μL of the quenched reaction aliquots using an NADH-linked LDH/PK assay that is standard in our lab. Enzyme activity provides an extremely sensitive proxy of the cleavage state of glucokinase.

7. Alternatively, SDS-PAGE can be used to monitor proteolysis. We mixed 10 μL aliquots of the quenched reaction with 5 μL of SDS sample buffer with 5% BME and heated at 80°C for 5 min.

8. After staining with Coomassie, the relative amounts of full-length glucokinase can be quantitated via gel densitometry using Image-J (NIH) (Fig. 1B).

2.3.4. Data Fitting and Analysis

1. Plots of relative remaining activity (or gel density) vs time give single-exponential decay curves (Fig. 1E). Fitting gives the observed proteolysis rate, k_obs.

2. Plots of proteolysis rates vs [thermolysin] allow extraction of proteolysis kinetic constants (Fig. 1A).

3. In the absence of glucose, glucokinase displays a linear response to increasing thermolysin (Fig. 1F). Fitting of a linear function allows determination of the equilibrium constant of local unfolding of the cleavable site, K_op.

4. In the presence of saturating amounts of glucose, k_obs values are an order of magnitude lower and activity displays a hyperbolic response to [thermolysin] (Fig. 1G). Fitting to a hyperbolic function allows determination of K_op as well as the rate constant for closing of the cleavable site, k_op. The rate constant for local unfolding of the cleavable site, k_cl, is easily calculated (Fig 1A).

2.4. Data Interpretation, Notes, and Helpful Advice

Caution should be taken in extrapolating locally unfolded states to global dynamics and structure, as measured by other techniques. For example, a locally unfolded state — unobservable with other techniques — may only be present in one conformation of a protein sampling multiple conformations at equilibrium, and it may be unclear which of the two conformations contain the locally unfolded state. In addition, the dynamics of local unfolding may occur on very different timescales than global dynamics. If the protein of interest has multiple cleavage sites and/or cleavage of a site does not disrupt enzymatic activity, then the cleavage rate must be determined by gel quantitation. Many general proteases can be used to measure local unfolding. For example, we also performed a proteolysis reaction on human glucokinase using trypsin and proteinase K. These two proteases have orthogonal sequence specificities to thermolysin yet gave the same proteolysis pattern on SDS-PAGE. This suggested that the cleavage was indeed dependent on local unfolding of the cleavage site and not on sequence specificity of the protease. The optimal range of [protease] will need to be determined empirically for the protein of interest. Human glucokinase is more susceptible to proteolysis (K_op = 0.01) compared to some model proteins probed with this technique, such as ribonuclease H (K_op = 0.000087) (Park and Marqusee, 2004) and maltose binding protein (K_op = 0.00000084) (Chang and Park, 2009). A higher concentration of thermolysin may be necessary to observe cleavage in less dynamic proteins. Alternatively, a protease with higher activity, such as proteinase K, could be used. If other proteases are used, then an appropriate quench agent must be used. Trypsin can be quenched with EDTA, and Proteinase K must be quenched with PMSF. Elucidation of the thermodynamic mechanism of local unfolding can be accomplished by measuring proteolysis kinetics at varying temperatures. Van ‘t Hoff plots of the equilibrium constant of unfolding, K_op, can indicate whether local unfolding is enthalpically and/or entropically driven. One can also combine global unfolding studies with local unfolding quantitation via limited proteolysis of one or more cleavage sites and, using temperature variation assays, one can determine the thermodynamic mechanisms underlying each event.

3. ¹³C Methyl-Detected NMR Methods Probing Protein Dynamics and Allostery

NMR spectroscopy provides an atomic resolution, residue-specific measurement of local dynamics over a very large range of timescales (Fig. 2A) (Grutsch et al., 2016, Farber & Mittermaier, 2015). Exploiting the spectroscopic properties of the methyl group widens applicability of these techniques to larger systems, not initially amenable to NMR studies, when paired with deuteration, site-specific isotopic labeling, and the methyl-TROSY effect (Schutz & Sprangers, 2020). These methyl-specific properties provide better resolution in spectra due to faster local rotational motion compared to the rest of the molecule. In addition, the strong initial magnetization originating on the three neighboring hydrogen nuclei of methyl groups and the prevalence of these substituents in polypeptides makes this approach broadly applicable (Boswell & Latham, 2019; Schutz & Sprangers, 2020). Indeed, due to the relatively high distribution of methyl groups throughout the hydrophobic core and binding sites of many proteins, rigorous descriptions of local and global conformational exchanges associated with allosteric communication and ligand binding can be determined from their NMR-observables. In the case of human glucokinase, these approaches allowed the identification of domain-specific structural changes linked to allostery and provided quantification of the timescale of the conformational rearrangements responsible for kinetic cooperativity.

Figure 2.

Methyl dynamics application and workflow. (A) Timescales of protein motion and NMR experiments capable of measuring them. Protein dynamics span femtosecond to hour long timeframes (top); a variety of NMR experiments have been designed for routine measurement of these dynamics (bottom). (B) Recommended workflow for methyl assignment and dynamics measurements detailed in this section. Dashed lines indicate steps that can aide, but are not explicitly mandatory for, unambiguous assignment. (C) Chemical exchange saturation transfer (CEST) profile showing a 1.5 ppm difference between two states with population A = 94.5% (solid line) and B = 6.5% (dashed line), an exchange rate of 50/s. (D) The effective R₂ profile as a function of Carr-Purcell-Meiboom-Gill (CPMG) pulse train frequency for an exchange event K_ex=1600/s between population A = 98.85% and B = 1.15%. (E) Triple quantum relaxation violated coherence transfer (RVCT) profile for a methyl side chain rotation with n = 100/s showing rigidification upon perturbation, n = 75/s

Dynamics measurements from NMR extract site-specific information from relaxation dispersion profiles or determining order parameters (Boswell & Latham, 2019). Routine analysis (Figure 2B) can provide the degree of flexibility of methyl groups, distribution of states, P_A, P_B, etc., a rate constant for inter-state exchange, k_ex, differences in chemical shift between states, Δω_AB, and the rates of longitudinal and transverse relaxation of each nucleus in each discreet state, R₁ and R₂ (Weisner & Sprangers 2015). Probing the dynamic features of a protein in NMR first requires assigning the signal from each methyl to individual residues in a 2D correlation spectrum. Single (HSQC) and multiple (HMQC) coherence transfers as well as transverse relaxation-optimized spectroscopy (TROSY) can report the chemical shift of each isotopically labeled ¹³C¹H₃. Each “base” experiment can be modified to contain a pulse program element which serves to generate the information required to determine structural dynamics. Detailed here are routine chemical exchange saturation transfer (CEST; Yuwen et al., 2017), Carr-Purcell-Meiboom-Gill (CPMG; Yuwen et al., 2019), and relaxation-violated coherence transfer (RVCT; Boswell et al., 2018) experiments used to investigate methyl group dynamics in proteins.

CPMG-based dynamics measurements have detected and quantified the presence of major conformational exchange in the small domain of unliganded human glucokinase, which provided a molecular basis for the kinetic cooperativity model of catalysis (Larion et al., 2015). Using isoleucine methyl groups as the reporter, relaxation dispersion profiles from residues both near and far from the glucose binding site were fitted with a general two-state exchange model and revealed a minor-state population of 16.5% with k_ex = 509 s⁻¹ at 37°C (Larion et al., 2015). This minor state represents an ensemble of adoptable conformations sampled by the unliganded enzyme, which is diminished and/or eliminated in the presence of glucose or activating mutations (Larion et al., 2015). While intermediate exchange limited further dissection of the minor state ensemble, this NMR study provided a site-specific description of allosteric effects governing the rate of glucokinase catalysis in a substrate dependent manner. From these studies, ¹³C methyl dynamics measurements served to expand structural descriptions of glucokinase function by probing important, yet dynamic processes.

3.1. Equipment and Reagents

• High quality NMR tubes

• Two NMR field strengths

• Pulse-field gradient triple resonance probe

• Isotopic Labeling Materials. (See Table 1)

• >99% D₂O

• M9 minimal medium salts

Table 1.

Keto- and amino acid labeling materials to incorporate isolated ¹³C methyl groups.

Residue	Position	Labeled Material	Dueteration	Scrambling Suppression	Reference
I	δ1	L-Ileδ1-[13CH3]	[2H,12C]-D-glucose		Tugarinov et al., 2006
	δ1	2-keto-3-d2-4-13C-butyrate	[2H,12C]-D-glucose		Tugarinov et al., 2006
	γ2	2-hydroxy-2-ethyl-D5-3-oxobutanoate-4-13C	[2H,12C]-D-glucose	d7,α-ketoisovalerate	Ayala et al., 2012
L	δ1 or δ1/2	L-Leuδ[13CH3,12CD3] or [13CH3]2*	[2H,12C]-D-glucose		Tugarinov et al., 2006
	δ1	2-keto-3-methyl-d3-3-d1-4-13C-butyrate	[2H,12C]-D-glucose		Tugarinov et al., 2006
	δ1/2	2-keto-3-methyl-13C-3-d1-4-13C-butyrate*	[2H,12C]-D-glucose		Tugarinov et al., 2006
V	γ1 or γ1/2	L-Valγ[13CH3,12CD3] or [13CH3]2*	[2H,12C]-D-glucose		Tugarinov et al.,2006
	γ1	2-keto-3-methyl-d3-3-d1-4-13C-butyrate	[2H,12C]-D-glucose		Tugarinov et al., 2006
	γ1/2	2-keto-3-methyl-13C-3-d1-4-13C-butyrate*	[2H,12C]-D-glucose		Tugarinov et al., 2006
A	β	L-Ala β-[13CH3]	[2H,12C]-D-glucose	L-isoleucine-d10, succinate-d 4, R-keto-isovalerate-d7	Ayala et al., 2009
M	ε	L-Met ε-[13CH3]	[2H,12C]-D-glucose		Gelis et al., 2007
		4-13C-Methyl-thio-2-ketobutyrate	[2H,12C]-D-glucose		Fischer et al., 2007
T	Yγ2	U-2H Thr-γ2[13CH3]	[2H,12C]-D-glucose	2-keto-3-d2-4-13C-butyrate, L-glycine-d5	Velyvis et al., 2012

Note:

^*indicates materials that are incompatible with relaxation-violated coherence transfer methodologies.

3.2. Procedures

3.2.1. D₂O Acclamation (Tugarinov et al., 2006)

1. From a starter culture of natural abundance H₂O-M9 medium, centrifuge cells and resuspend 20 mL of H₂O-M9 medium to an optical density (OD₆₀₀) = 0.1. Grow to OD₆₀₀ ≈ 0.6.

2. Spin down cells and resuspend cells in 100 mL D₂O-M9 medium + 2 g/L deuterated D-glucose to OD₆₀₀ = 0.1, grow to OD₆₀₀ ≈ 0.5, then dilute to 200 mL with D₂O-M9 and grow again to OD₆₀₀ ≈ 0.5

3. Dilute to a volume upon which addition of labeling materials results in a final volume of 1 L.

4. Grow to an OD₆₀₀ = 0.25 or an OD that corresponds to 1 h prior to induction, then add solutions or solids of labeling materials to appropriate concentrations based on desired labeling scheme(s).

5. Continue to express as normal. Limiting expression to 6-8 h after induction is recommended to prevent scrambling of label to other carbons.

3.2.2. Methyl Group Resonance Assignment

1. Many robust methyl-assignment algorithms have been developed for streamlining methyl-detected NMR measurements in large proteins. Methyl Assignments Using Satisfiability (MAUS; (Nerli et al., 2021) software demonstrates the state-of-the-art and was designed for automated assignment of all ¹³C methyl-containing sidechains with assignment accuracy and completeness near 100% and 50-80%, respectively. Inputs include a user-annotated 2D HMQC spectrum, two 3D NOE spectra (50ms and 300ms mixing times) or one 4D NOE spectrum (300ms mixing time) and a PDB structure or model.

2. In combination with automated assignment, unambiguous determination of methyl group identity can be achieved by single-residue mutation and recollection of 2D ¹³C¹H correlation spectra. When not drastically altering the structure or function of the protein, mutations offer the potential to make crowded regions more interpretable.

3.2.3. Chemical Exchange Saturation Transfer

1. Measure a 2D spectrum recording the intensity of each peak (I₀) with an empty exchange period of 150-300 ms.

2. Recollect this spectrum, now with a weak (10-100 Hz) B₁ field during the exchange period. Optimize this parameter with coarse sampling of the indirect dimension.

3. Vary the offset of the B₁ field, sweeping through the carbon spectral width in a step size equivalent to the desired resolution (1.0 – 0.5 ppm or lower).

4. For each offset position, record the intensity of each peak (I). Typically, 2 or more replicates of each offset frequency are summed prior to interpretation.

5. Plot the intensity ratio (I/I₀) of each peak versus the offset, producing the CEST profile for each peak. (Fig. 2C)

6. CEST profiles can be fit using Bloch-McConnell equations to extract values of the P_B, k_ex, Δω_AB as well as R1 and R2 relaxation rates for major and minor states.

3.2.4. Carr-Purcell-Meiboom-Gill

1. Record 2D spectrum, note the intensity of each peak (I₀).

2. Recollect this data with a CPMG pulse train element of constant time (T) ~20-80ms. Increase the frequency (V_CPMG) of the refocusing pulses for each data set, ~50-1000Hz, in 10-20 (or more) steps.

3. The effective transverse relaxation rate (R_2,eff) can be extracted from 2D correlation spectra by:

   $$R_2,eff\left(v_{CPMG}\right)=-\frac{1}{T}\ln \left(\frac{I{v}_{CPMG}}{Io}\right)$$

4. Data sets should be recorded at two different field strengths and profiles fit simultaneously to extract a robust solution for exchange parameters.

5. Plotting R_2,eff versus V_CPMG gives the residue-specific dispersion profile. (Fig. 2D)

6. Fitting profiles numerically will provide P_B, k_ex, and Δω_AB as well as R1 and R2 relaxation rates for major and minor states.

3.2.5. Relaxation-Violated Coherence Transfers

1. 2D Spectra are recorded using standard literature procedures (Sun et al. 2011). Interleave collection of 2D spectra with increasing relaxation delays track the buildup of methyl-1H triple-quantum coherences occurring at increasing lengths of relaxation delays (T).

2. To plot a well-defined build up curve, 10-20 sets of experiments with relaxation delays of 2-50ms are recommended.

3. Intra-methyl dipolar cross-relaxation rates (η), which is proportional to the amplitude of motion within the methyl groups, can be calculated by fitting the absolute ratio of peak intensities from forbidden and allowed coherence transfer pathways as a function of relaxation delay, T. (Fig. 2E)

   $$\left|I_f/I_a\right|=\frac{C(\eta \tanh {({\eta }^2+{\delta }^2)}^{0.5}T)}{{({\eta }^2+{\delta }^2)}^{0.5}-\delta \mathit{tanh}({({\eta }^2+{\delta }^2)}^{0.5}T)}$$

   Where C = ¾ and δ, which accounts for relaxation effects caused by external protons, is 0 as the protein should be highly deuterated.

4. From measurement of η, the amplitude of motion of the methyl 3-fold axis is directly reported. Monitoring the effect of structural perturbations on this parameter can define local and regional changes in flexibility concomitant of allostery.

3.3. Data Interpretation and Model Assumptions

While protein dynamics can be described in detail with NMR-based techniques, the limitations and assumptions of the data collection and interpretation must be noted. As the number of methyl signals increases with the size of the molecule, the ability of NMR to potentially detect every carbon nucleus becomes the source of spectral overlap and assignment ambiguity. Analysis of mutants might not always yield the full breadth of information required to deconvolute such situations. Even in the case of an ideal protein, local dynamics can occur at rates comparable to the NMR timescale and result in severe line-broadening, a condition termed intermediate exchange. The experiments themselves also suffer from the low sensitivity of NMR. If a minor state is sampled by less than 1-2% of the population, the sensitivity of the experiments described above is quite low. As for data fitting, the conformational ensemble of a protein is a continuum of states that interconvert over some timespan, thus using a two- or even multi-state model to describe exchange rates generalizes the occurrence of physical events. Nonetheless, NMR has proven indispensable for addressing elements of protein dynamics which have not been ascertainable by other means.

3.4. Notes and Helpful Advice

To reduce the effects of external relaxation mechanisms, RVCT experiments specifically require incorporating ¹³C¹H₃ at only one methyl-group position of Leucine and Valine. CEST and CPMG datasets can be fitted using ChemEx chemical exchange analysis following the user guide developed by creator Guillaume Bouvignies (https://gbouvignies.github.io/ChemEx/docs/user_guide). Care should be taken to optimize the temperature used in assignment and dynamics measurements. To this end, collecting relatively quick test-spectra at various temperatures before the assignment step can help ensure optimal results. Prior determination of the temperature range tolerated by the protein under investigation, e.g., by using activity assays, is helpful to establish the window under which NMR dynamic investigations can be conducted. If the rotational correlation time of the molecule is known or can be extracted from other measurements, the methyl 3-fold symmetry axis order parameter (S²_axis) can be calculated from RVCT experiments. Values for S²_axis range from 0 to 1 for the most flexible to the most rigid methyl groups, respectively. These solutions can provide information on the type of chemical exchange the methyl group undergoes (i.e. jumps between rotameric states, or large and small scale motions within rotameric states). Growth in D₂O can be much slower than in H₂O, thus proper acclimation is crucial to efficient expression and recovery of the target protein.

4. Hydrogen-Deuterium Exchange Mass Spectrometry to Interrogate Allostery

Hydrogen-deuterium exchange mass spectrometry (HDX-MS) is a powerful technique that is widely applied to study the conformations of individual proteins and protein complexes (Busenlehner and Armstrong, 2005). This technique can also be used to locate protein binding sites, probe allosteric effects and monitor protein folding/unfolding dynamics. HDX-MS is complementary to other high resolution biophysical techniques, such as nuclear magnetic resonance (NMR) and X-ray crystallography, and it can accommodate large proteins, low sample concentrations and complex sample matrices. Most HDX MS experiments employ a continuous labelling strategy in which the protein is exposed to deuterium for various amount of time from seconds to hours (Fig. 3). The reactions are then quenched, and the labeled proteins undergo proteolytic digestion, followed by LC-MS analysis. This strategy allows the monitoring of deuterium incorporation into the protein structure as a function of time, providing information on conformations and dynamics.

Figure 3.

General workflow of a hydrogen/deuterium exchange mass spectrometry experiment.

Biochemical and biophysical studies of human glucokinase suggest that activated disease variants can be segregated into two mechanistically distinct classes, termed α-type and β-type (Whittington et al., 2015; Sternisha et al., 2018). Prior NMR experiments were limited in describing the structural differences between wild-type glucokinase and each activated variant, since they relied on only 17 sparsely distributed isotopic labels located on isoleucine side chain (Larion et al., 2012; Larion et al., 2015). To provide a more complete picture of the structural consequences of enzyme activation, the application of HDX-MS proved valuable. In the studies described below, we utilized solution phase HDX monitored by a 21 Tesla Fourier Transform-Ion Cyclotron Resonance (FT-ICR) mass spectrometer for wild-type glucokinase, as well as α-type and β-type variants. In conjunction with the pulsed proteolysis and NMR methods described above, HDX-MS results facilitated the development of a detailed model of the structural and dynamical alterations associated with enzyme activation.

Comparative HDX-MS provided specific localized insights into the nature of differential structure and dynamics related to allostery in human glucokinase. Experiments conducted on unliganded enzyme variants demonstrates that a disordered active-site loop, which folds upon glucose binding, is protected from exchange in the α-type variant. The α-type variant also displays increased exchange within a β-strand located near the enzyme’s hinge region, which becomes more solvent-exposed upon glucose binding. In contrast, β-type activation causes no substantial difference in global or local exchange relative to unliganded, wild-type glucokinase. Together with data from quenched-flow experiments, the HDX-MS results demonstrate that α-type activation results from a shift in the conformational ensemble of unliganded glucokinase toward a state resembling the glucose-bound conformation, whereas β-type activation is attributable to an accelerated rate of product release (Fig. 4) (Sternisha et al., 2018). This work elucidates the molecular basis of naturally occurring, activated disease variants and provides insight into the structural and dynamic origins of glucokinase’s unique kinetic cooperativity.

Figure 4.

Example of data interpretation based on HDX-MS results for wild-type human glucokinase compared to α-type and β-type activated variants. Left: Deuterium uptake profiles and maximum-entropy fits vs. HDX reaction periods for selected amino acid (AA) segments of wild-type enzyme (blue), an α-type variant (red) and a β-type variant (green). Lower right: ARDDs for α-type variant minus wild-type glucokinase mapped onto the crystal structures of glucose-bound (PDB 1V4S) and unliganded (PDB 1V4T) human glucokinase. Regions with a decreased level of deuterium exchange in the α-type variant are shown in green and regions with an increased level of deuterium exchange are shown in orange.

4.1. Equipment and Software

• HTC Pal auto-sampler (Eksigent Technologies, Dublin, CA)

• Jasco HPLC (Jasco, Easton, MD)

• 21 Tesla Fourier transform ion cyclotron resonance mass spectrometer (FT-ICR MS)

• Pro-Zap MS C₁₈ column (1.5 μm, 500 Å pore size, 2.1 × 10 mm, Dr. Maisch Gmbh, Ammerbuch, Germany)

• Pipettes

• 96-well plates and Agilent HPLC vials

• Xcalibur Software (ThermoFisher, San Jose, CA)

• In-house software package ((Kazazic et al.; Zhang, 2013; Zhang et al.))

4.2. Reagents and Buffers

• D₂O deuterium oxide, 99.9 atom% D

• H₂O, LC-MS grade

• Tris (2-carboxyethyl)phosphine (TCEP)

• Urea

• Formic Acid, LC-MS grade

• Acetonitrile, LC-MS grade

• D₂O buffer: 50 mM KPO₄, pH 7.6 ± 0.1, 50 mM KCl, 10 mM DTT, in deuterium solution

• H₂O buffer: 50 mM KPO₄, pH 7.6 ± 0.1, 50 mM KCl, 10 mM DTT, in water solution

• Quench solution: 200 mM TCEP and 8 M urea in 1.0% formic acid

• Proteolysis solution: 40% saturated protease type XIII (Sigma-Aldrich) in 1.0% formic acid (final pH ~ 2.3)

• Mobile phase A - acetonitrile/H₂O/formic acid, 5%/94.5%/0.5% (v/v)

• Mobile phase B - acetonitrile/H₂O/formic acid, 95%/4.5%/0.5% (v/v)

4.3. Experiment Design and Procedure

4.3.1. Continuous deuterium labeling and quenching

1. To initiate each hydrogen deuterium exchange reaction, 5 μL of purified glucokinase was mixed with 45 μL of corresponding D₂O buffer.

2. For the blank control, the dilution was performed in H₂O buffer instead of D₂O.

3. Reactions were conducted at 1-2°C to reduce the back-exchange.

4. The HDX reaction periods were 0.5, 1, 4, 15, 30, 60, and 480 min.

5. Add 25 μL of pre-cooled quench solution to the 50 μL of diluted protein.

4.3.2. In-solution digestion

1. After quenching, add 25 μL of proteolysis solution into each sample. The protease digestion was performed for 3 min at 1 ~ 2 °C.

2. Each HDX reaction and assay was performed in triplicate.

4.3.3. Online liquid chromatography and Ultra-high-resolution mass spectrometry

1. After proteolysis, 45 μL of the resulting peptides were loaded onto a Pro-Zap MS C₁₈ column for separation and desalting.

2. A rapid gradient from 2% to 95% B over 2 min was performed to elute peptides at a flow rate of 0.3 mL/min.

3. To increase electrospray ionization efficiency, a post-column four-way union was used to split the flow by 1:100.

4. The ionized LC eluent was directed to a positive electrospray ionization 21 Telsa FT-ICR mass spectrometer for measurement. Mass spectra were collected from m/z 400 ~ 1300 at a high resolving power (m/Δm_50% = 300,000 at m/z 400). The electrospray ionization source voltage was 3.5 kV, and the heated capillary temperature was 325°C. The automatic gain control (AGC) target was set at three million charges for broadband spectra.

5. All raw data were collected by Xcalibur software.

4.4. Data Analysis

4.4.1. Sequence coverage

Prior to HDX MS experiments, a digestion experiment without deuterium incubation was performed to test the applicability of HDX in probing structural differences. Treatment with protease XIII yielded 106 common overlapping peptides that cover 93% of the primary structure of wild-type glucokinase. Only those proteolytic peptides common to both wild-type and activated variants are compared for the following HDX analysis. With many overlapping peptides, a peptide-level resolution of conformational comparison can be achieved.

4.4.2. Determination of deuterium incorporation

HDX experiments were monitored by mass spectrometry with mass spectra collected across an LC gradient. A custom software package was used to identify eluting peptides and determine the centroid mass of deuterated peptide. For control experiments, the software searches through all LC-MS scans to find proteolytic peptides whose monoisotopic masses fall within the 2 ppm mass error tolerance window. For each peptide, the software searches for matched deuterium-incorporated peaks based on a mass ‘sub-window’. Later, the experimental m/z centroid can be calculated, and further compared to that of nondeuterated peptide to obtain deuterium incorporation levels (i.e., number of deuterium divided by number of exchangeable amide hydrogens) as shown in Figure 3. Percentages of deuterium incorporation at each time point can be calculated by dividing the level of deuterium incorporation by maximum update, D_max (theoretical exchangeable hydrogen minus one).

4.4.3. Calculation of averaged relative deuterium-uptake difference (ARDD)

The averaged relative deuterium uptake difference (ARDD) between each exchange period is calculated by the following equation:

$$\text{ARDD}={\sum }_i\frac{A\left(t_i\right)-B\left(t_i\right)}{A\left(t_i\right)}$$

In which A is the deuterium uptake for variant glucokinase at a specific time (t_i) and B is the deuterium uptake for wild-type glucokinase at the same time point (t_i). Time course HDX data in a kinetic plot with number of deuteration against labeling time were fitting by an MEM fitting method (Zhang et al.).

4.5. Data Interpretation – An Example Case Study with Human Glucokinase Variants

Using the described HDX-MS method, we measured the averaged relative deuterium uptake differences between wild-type glucokinase and two mechanistically distinct activated enzyme variants. We found that segments comprising residues 53-73, 159-186, and 275-302 of the α-type variant display apparent alterations in deuterium exchange relative to that of the wild-type enzyme (Fig. 4). Conversely, β-type glucokinase did not show significant (>10%) deuterium-uptake change compared to wild-type enzyme in any region of the protein. These results indicate that α-type activation, but not β-type activation, is associated with changes in the structure and/or dynamics of specific regions of the unliganded enzyme.

To gain mechanistic insight into the differences in hydrogen-deuterium exchange observed in the α-activated variant, we mapped ARDD values for each region onto the corresponding crystal structures of both unliganded and glucose-bound glucokinase. As shown in Figure 4, regions that exhibited a reduced rate of deuterium uptake in the α-variant include the mobile active site loop comprised of residues 159-180 and a peptide segment spanning residues 275-302. The observation of a significantly decreased deuteration level in segment 159-180 is attributed to a disorder-order transition that occurs when glucose binds, in accord with previously published NMR data (Larion et al., 2012; Whittington et al., 2015). Segment 275-302 comprises part of the glucose-binding pocket; therefore, its decrease in deuteration level likely results from the formation of a more closed conformation, which limits the proton exchange with the surrounding deuterated solvent (Fig. 4). Conversely, we observed that a segment of the α-variant spanning residues 62-73 exhibited elevated levels of deuterium incorporation compared to that of its wild-type counterpart. This region is located within a loop/β-strand motif that connects the large and small domains. Notably, previous studies have identified several activating substitutions within this region, including S64P, a variant that displays NMR features characteristic of the α-type activation mechanism (Larion et al.; Pal & Miller). HDX-MS revealed an increased rate of deuterium uptake in this region, in agreement with a comparison of the unliganded and glucose-bound crystal structures, demonstrating that this region is more solvent exposed in the glucose-bound state (Fig. 4). HDX characterized data for activated variants, in combination with prior NMR results (Larion et al., 2012; Whittington et al., 2015) provide compelling evidence that the unliganded α-activated glucokinase variant resembles the glucose-bound conformation in several regions.

4.5. Notes and Helpful Advice

Prior to initiating an HDX-MS experiment, intact protein analysis and gel electrophoresis analysis is recommended to confirm primary structure and sample purity. The deuterium labeling reaction is highly sensitive to experimental conditions such as temperature, ionic strength and pH. The hydrogen-deuterium exchange reaction, following quenching and protease digestion should be performed at the well-controlled temperature of 1~ 2°C to minimize the back-exchange. One also should take caution in choice of the buffer system used in the labeling reaction to make sure it has sufficient buffering capacity to ensure constant pH. Use of an HDX automation platform is recommended to pursue to ensure experiment repeatability and precision. The back-exchange level should be well-controlled by maintaining consistent low pH and low temperature during the experiment. Level of back-exchange can be determined by analysis of ‘maximally labelled’ peptide standards or protein digest. In a well-designed and controlled HDX-MS experiment, most of peptides have a back-exchange rate of roughly 30%.

For protease treatment, most fully automated HDX-MS experiments have used an immobilized pepsin column, because pepsin lacks cleavage specify and is active at low pH. There are varieties of other proteases active under acidic conditions that have been used including protease XIII, XVI, plasmephsin, and nepenthesin. In addition, a co-immobilized, a dual protease column has been developed to further improve the digestion efficiency and sequence coverage. Quench solution will often contain a denaturant such as guanidine and a reducing agent that helps initiate unfolding of the protein before exposure to the acid functional protease. TCEP is the best reductant option as it provides efficient disulfide breakage at the low pH and low temperature required. During assignment of unique peptides, some ambiguously assigned isobaric peptides are likely to appear and should be carefully analyzed. In general, the pair of isobaric peptides that span very different sequences are disregarded, whereas those that only differ in one amino acid position are kept for HDX analysis. HDX-MS data from multiple charge statues and overlapping peptides should be used to confirm the reported structure related deuterium-uptake differences.

5. Summary and Conclusions

New and mechanistically diverse examples of allosteric proteins continue to be discovered (Gunasekaran et al. 2004). Moreover, there is a growing interest within the synthetic biology community to engineer cooperative kinetic and/or thermodynamic responses into molecular systems (Gorman et al., 2019). Understanding the molecular origins of allostery requires more than simply elucidating the structures of individual conformations within the ensemble. Quantifying the rates of conformational rearrangements occurring in response to ligand concentrations is essential for developing a quantitative and predictive allosteric model. The biochemical approaches outlined in this chapter were invaluable in establishing a detailed kinetic and structural model for allostery in human glucokinase. Together, these methods demonstrated that glucokinase cooperativity stems from millisecond timescale dynamic interconversions between multiple conformations within the enzyme’s small domain, which are facilitated by more rapid flexibility within a 30-residue intrinsically disordered loop (Sternisha and Miller, 2019). We envision that the combination of the three experimental approaches described herein will be generally useful for interrogating the mechanistic basis of allostery in other proteins, especially when a diverse range of conformation fluctuations, occurring across a broad range of timescales, are associated with cooperativity.