Molecular mechanism of glycolytic flux control intrinsic to human phosphoglycerate kinase

Significance

Control of glycolytic flux plays an important role in energy production and metabolic homeostasis in cells. Phosphoglycerate kinase (PGK), a glycolytic enzyme, catalyzes reversible phosphotransfer to generate 3-phosphoglycerate (3PG) and adenosine triophosphate (ATP) in glycolysis. PGK controls glycolytic flux according to the intracellular [ATP]/[ADP] (adenosine diphosphate) ratio; however, its molecular mechanism remains unknown. Here, we report a protein-level regulation of the human PGK reaction by switching ligand-binding cooperativities between adenine nucleotides and 3PG, which is finely tuned to appropriately respond to changes in the intracellular [ATP]/[ADP] ratio. Our findings reveal a molecular mechanism intrinsic to human PGK that controls glycolytic flux by rapid adaptation to changes in the intracellular environment to avoid critical damage.

Abstract

Glycolysis plays a fundamental role in energy production and metabolic homeostasis. The intracellular [adenosine triphosphate]/[adenosine diphosphate] ([ATP]/[ADP]) ratio controls glycolytic flux; however, the regulatory mechanism underlying reactions catalyzed by individual glycolytic enzymes enabling flux adaptation remains incompletely understood. Phosphoglycerate kinase (PGK) catalyzes the reversible phosphotransfer reaction, which directly produces ATP in a near-equilibrium step of glycolysis. Despite extensive studies on the transcriptional regulation of PGK expression, the mechanism in response to changes in the [ATP]/[ADP] ratio remains obscure. Here, we report a protein-level regulation of human PGK (hPGK) by utilizing the switching ligand-binding cooperativities between adenine nucleotides and 3-phosphoglycerate (3PG). This was revealed by nuclear magnetic resonance (NMR) spectroscopy at physiological salt concentrations. MgADP and 3PG bind to hPGK with negative cooperativity, whereas MgAMPPNP (a nonhydrolyzable ATP analog) and 3PG bind to hPGK with positive cooperativity. These opposite cooperativities enable a shift between different ligand-bound states depending on the intracellular [ATP]/[ADP] ratio. Based on these findings, we present an atomic-scale description of the reaction scheme for hPGK under physiological conditions. Our results indicate that hPGK intrinsically modulates its function via ligand-binding cooperativities that are finely tuned to respond to changes in the [ATP]/[ADP] ratio. The alteration of ligand-binding cooperativities could be one of the self-regulatory mechanisms for enzymes in bidirectional pathways, which enables rapid adaptation to changes in the intracellular environment.

Glycolysis is a central glucose metabolic pathway that provides cellular energy, and its flux control plays an important role in adenosine triphosphate (ATP) production and glucose homeostasis in living cells. Excessively increased glucose metabolism is a hallmark of cancer, in which glycolysis is accelerated even in aerobic conditions (Warburg effect); thus, cancer cells can be distinguished from normal cells by their enhanced glucose metabolism (1). The intracellular [ATP]/[ADP] (adenosine diphosphate) ratio is a critical parameter that controls glycolytic rate (2). For instance, increases in [ATP] allosterically inhibit phosphofructokinase activity, which limits glycolytic flux under aerobic conditions (3). Glycolysis is thought to be mainly regulated by three committed steps catalyzed by hexokinase, phosphofructokinase, and pyruvate kinase, which have negative Gibbs free energies (ΔG) (−33.5 kJ/mol, −22.2 kJ/mol, and −16.7 kJ/mol, respectively), resulting in irreversible reactions (4). However, steps with small ΔG, which are near equilibrium (ΔG ∼= 0) also reportedly contribute to metabolic homeostasis and energy yield, demonstrating that glycolytic flux can be altered with only small changes in the concentration of glycolysis intermediates (5).

Phosphoglycerate kinase (PGK) is a glycolytic enzyme that catalyzes the reversible phosphotransfer from 1, 3-bisphosphoglycerate (1, 3-bPG) to ADP to generate 3-phosphoglycerate (3PG) and ATP. It is a near-equilibrium reaction and the first ATP generation step in glycolysis (forward reaction). It also participates in gluconeogenesis, catalyzing the opposite direction from glycolysis (reverse reaction). PGK is one of the key enzymes involved in reprogramming glucose metabolism to promote tumor growth in cancer cells (6, 7). Indeed, many types of cancer cells exhibit high PGK levels (8–13). Interestingly, mammalian cells treated with oligomycin, an inhibitor of ATP synthase that inhibits respiratory ATP production, exhibit an increased cumulative ΔG value of a coupling reaction between glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and PGK with decreasing [ATP]/[ADP] ratios, leading to enhanced glycolytic flux (5). The [ATP]/[ADP] ratio also controls the direction (forward or backward) of the PGK reaction flux (2). This evidence suggests that PGK is a key enzyme for glycolytic flux control by sensing the [ATP]/[ADP] ratio.

Human PGK (hPGK) is a 45-kDa monomeric protein consisting of two domains: an amino- (N-) terminal domain that binds to 1, 3-bPG or 3PG and a carboxyl- (C-) terminal domain containing a nucleotide-binding site. The relative orientation of these two domains stably adopts an open configuration regardless of the ligand-bound state, which is revealed by its crystal structures in complex with various ligands (14, 15). The phosphotransfer reaction proceeds by adopting a fully closed conformation through the hinge-bending motion demonstrated by small-angle X-ray scattering (SAXS) data (15, 16). Fully closed structures have only been solved in complex with transition-state analogs (TSAs) (17). These structural data provide a general picture of the PGK catalytic reaction (18); however, the detailed molecular mechanism of its regulation, how the turnover cycle is controlled, and how the reaction direction is switched by sensing the [ATP]/[ADP] ratio remains unclear.

Previous studies on pig muscle PGK (pmPGK) have shown that ligand binding of pmPGK is modulated by their counter ligands—the nucleotide binding is weakened by 3PG binding (19, 20), and, conversely, the affinity to 3PG is reduced in the presence of nucleotides (21). Thus, cooperative ligand binding is expected to be an essential component of the PGK catalytic reaction. However, these antagonistic effects were observed by indirect methods, using either fluorescent probes competing with nucleotide binding (19, 21) or by modifying reactive thiol groups of cysteine residues (20). Moreover, these experiments were performed under nonphysiological and low-salt conditions. The activity of yeast PGK is reportedly affected by salt concentration (22, 23); therefore, the ligand-binding cooperativity must be investigated under physiological salt conditions to understand its regulatory mechanism. Although many nuclear magnetic resonance (NMR) studies of PGK, including those containing ligand interactions (17, 24) or folding stability (25), have previously been reported, these experiments were performed under nonphysiological salt conditions.

We performed structural analyses of hPGK in various ligand-bound states under physiological salt conditions using NMR spectroscopy. This revealed that the negative and positive ligand-binding cooperativities between nucleotides and 3PG were the key drivers of the catalytic cycle in physiological environments. We proposed a reaction scheme for hPGK that illustrated a step-by-step sequence from substrate binding to product release during its turnover cycle.

Results

Ligand Bindings to hPGK at the Physiological Salt Concentration.

To investigate the conformational consequences of hPGK on ligand binding, we used methyl-transverse relaxation-optimized spectroscopy (TROSY) (26, 27) combined with selective ¹³C-isoleucine δ1-methyl (Ileδ1-[¹³CH₃]) labeling (28), which is suitable for analyzing high-molecular-weight proteins. All 19 Ile resonances on the ¹H-¹³C heteronuclear multiple quantum coherence (HMQC) spectrum of Ileδ1-[¹³CH₃]–labeled hPGK were completely assigned by site-directed mutagenesis (Fig. 1A and SI Appendix, Figs. S1 and S2).

Fig. 1.

Conformational consequence of hPGK ligand binding. (A) Mapping of isoleucine residues on the structure of hPGK (Protein Data Bank [PDB] code: 2XE7). The isoleucine residues are represented as magenta sticks. The N- and C-terminal domains are colored in cyan and green, respectively. The CHH (residues 186 through 202) is colored in blue. 3PG and ADP are shown as stick models. (B) Overlay of a portion of the ¹H-¹³C HMQC spectra of 0.05 mM Ileδ1-[¹³CH₃]–labeled hPGK, as recorded with increasing [3PG] (i), increasing [MgADP] (ii), and increasing [MgAMPPNP] (iii). The spectra are processed identically and plotted using the same contour levels. Resonance colors correspond to different [ligand]/[hPGK], as shown in the same color on the right side of the spectra. The directions of the CS changes are indicated by black arrows. The resonance positions of Ile-367 at different ratios of [3PG]/[hPGK] are connected by a red dashed line. Spectra are selectively extracted from all titration points and overlaid for clear viewing. (C) The ligand-binding equilibrium model included four states: ligand-free (E; E_nzyme), nucleotide-bound (EN; E_nzyme-N_ucleotide), 3PG-bound (EP; E_nzyme-₃P_G), and nucleotide-3PG–bound (ENP; E_nzyme-N_ucleotide-₃P_G). The K_D values of the nucleotides in the absence and presence of 3PG [K_D(N) and K_D(N/P), respectively], and those of 3PG in the absence and presence of nucleotides [K_D(P) and K_D(P/N), respectively], are defined by (i). According to the energy conversion principle (35), the K_D values correlate with each other, as shown in (ii). K_D(N) and K_D(P) values are fixed to those determined by individual ligand titration experiments. K_D(N/P) and K_D(P/N) can be simplified as K_D(N)/α and K_D(P)/α, respectively, as shown in (iii). (D) CS changes in the Ile-367 resonance induced by the two-ligand binding experiments with MgADP and 3PG. (i) The titration spectra of hPGK with MgADP in the presence of a 20-fold excess of [3PG] at the ratio of [MgADP]/[hPGK] = 0 (shown in black) and 40 (shown in magenta). (ii) The titration spectra of hPGK with 3PG in the presence of a 30-fold excess of [MgADP] at the ratio of [3PG]/[hPGK] = 0 (shown in black) and 60 (shown in magenta). The directions of the CS changes are indicated by black arrows. The theoretical ¹³C CSs of the Ile-367 resonances in the ligand-free, MgADP-bound, 3PG-bound, and MgADP-3PG–bound states calculated by the global CS-fitting analysis are shown by gray dashed lines as E, ED, EP, and EDP, respectively. The linear trajectory between the E and EP states is indicated by the red dashed line. (E) Changes in the ¹³C CS of the Ile-367 resonances induced by ligand binding. ¹³C CSs of Ile-367 resonances are plotted against [3PG]/[hPGK] in the absence of MgADP (i), [MgADP]/[hPGK] in the presence of a 20-fold excess of [3PG] (ii), [MgADP]/[hPGK] in the absence of 3PG (iii), and [3PG]/[hPGK] in the presence of a 30-fold excess of [MgADP] (iv). Error scales are estimated using NMR digital resolutions. The theoretical ¹³C CSs of the ligand-free, MgADP-bound, 3PG-bound, and MgADP-3PG–bound states are shown by the green lines as E, ED, EP, and EDP, respectively. (F) Same as D, except for the two-ligand binding experiments with MgAMPPNP and 3PG. (i) The titration spectra of hPGK with MgAMPPNP in the presence of a 20-fold excess of [3PG] at the ratio of [MgAMPPNP]/[hPGK] = 0 (shown in black) and 60 (shown in magenta). The spectrum in the TSA-complex state is superimposed, as shown in blue. (ii) The titration spectra of hPGK with 3PG in the presence of a 60-fold excess of [MgAMPPNP] at the ratio of [3PG]/[hPGK] = 0 (shown in black) and 20 (shown in magenta). The theoretical ¹³C CSs of the Ile-367 resonances in the ligand-free, MgAMPPNP-bound, 3PG-bound, and 3PG-MgAMPPNP–bound states are shown by the gray dashed lines as E, EX, EP, and EXP, respectively. (G) Similar to E, except that the ¹³C CSs of the Ile-367 resonances are plotted against [3PG]/[hPGK] in the absence of MgAMPNP (i), [MgAMPPNP]/[hPGK] in the presence of a 20-fold excess of [3PG] (ii), [MgAMPPNP]/[hPGK] in the absence of 3PG (iii), and [3PG]/[hPGK] in the presence of a 60-fold excess of [MgAMPPNP] (iv). The theoretical ¹³C CSs of the ligand-free, MgAMPPNP-bound, 3PG-bound, and 3PG-MgAMPPNP–bound states are shown by the green lines as E, EX, EP, and EXP, respectively.

We first titrated each ligand separately to hPGK in the presence of 120 mM KCl with 5 mM KPi (hereafter referred to as the physiological salt concentration). Although 3PG binds to the N-terminal domain, the titration of 3PG to hPGK showed chemical shift perturbations (CSPs) widely distributed throughout the protein in the fast exchange on the NMR timescale (SI Appendix, Fig. S3). Assuming that these CSPs linearly correlate with the shift in the population between the free and 3PG-bound states, we performed a global chemical shift (CS) fitting analysis and determined the dissociation constant (K_D) value of 3PG for hPGK to be 112 μM (Table 1 and SI Appendix, Fig. S4). The largest CSP was observed for the resonance of Ile-367 at the hinge region (Fig. 1 B, i) and SI Appendix, Fig. S3). As the relative orientation of Ile-367 with the central hinge helix (CHH; residues 186 through 202) changes upon 3PG binding (SI Appendix, Fig. S5A), the observed CSP reflects an allosteric conformational change around the hinge region that is induced by 3PG binding. It can be utilized as a representative probe for the conformational equilibrium between the 3PG-bound and -unbound conformations of hPGK (29).

Table 1.

Binding affinities and kinetic parameters of the ligands for hPGK in the absence and presence of counter ligands^*

Ligand [Ex-model]†	KD/mM		α	ΔΔG/kJ mol−1	koff/s−1
	−3PG	+3PG			−3PG	+3PG
MgADP [F to I]	0.224 (0.216 to 0.232)	0.501 (0.477 to 0.527)	0.448 (0.426 to 0.470)	2.07 (1.94 to 2.20)	811 (749 to 880)	1,348 (1,227 to 1,493)
	0.237 (0.203 to 0.266)	0.320 (0.297 to 0.354)	0.740 (0.671 to 0.797)	0.775 (0.584 to 1.03)	—	—
MgAMPPNP [F]	0.894 (0.852 to 0.942)	0.656 (0.624 to 0.687)	1.36 (1.30 to 1.43)	−0.791 (−0.920 to −0.675)	N/A‡	N/A‡
	1.08 (1.04 to 1.17)	0.737 (0.663 to 0.806)	1.47 (1.34 to 1.63)	−0.990 (−1.26 to −0.759)	—	—
	−nucleotide	+MgADP	α	ΔΔG/kJ mol−1
3PG [F]	0.112 (0.106 to 0.119) 0.158 (0.140 to 0.175)	0.251 (0.239 to 0.263)	0.448 (0.426 to 0.470)	2.07 (1.94 to 2.20)	N/A‡	N/A‡
		0.213 (0.198 to 0.235)	0.740 (0.671 to 0.797)	0.775 (0.584 to 1.03)	—	—
		+MgAMPPNP	α	ΔΔG/kJ mol−1
		0.0823 (0.0782 to 0.0862)	1.36 (1.30 to 1.43)	−0.791 (−0.920 to −0.675)	N/A‡	N/A‡
		0.108 (0.0967 to 0.118)	1.47 (1.34 to 1.63)	−0.990 (−1.26 to −0.759)	—	—

*The values obtained from the NMR data are presented in the upper rows, whereas those obtained from the ITC data are displayed in the lower rows with italics for each ligand. The K_D values in the presence of counter ligands are calculated from these α-values (Fig. 1C). Values in brackets represent the 95% bootstrap CIs.

†Exchange model (Ex-model) used for the global CS-fitting analyses are shown in square brackets. [F to I] and [F] indicate “fast-to-intermediate” and “fast” exchanges, respectively.

‡Not applicable owing to fast exchange.

Titration of hPGK with MgADP resulted in large CSPs for residues close to the nucleotide-binding site (SI Appendix, Figs. S1 and S3). The largest CSP was observed for the Ile-370 resonance with the appearance of signal broadening in the middle of titration (Fig. 1 B, ii), indicating the fast-to-intermediate exchange between the free and MgADP-bound states on the NMR timescale. Therefore, we performed a global CS-fitting analysis by assuming fast-to-intermediate exchange between the free and MgADP-bound states and determined the dissociation rate (k_off) and the K_D values to be 811 s⁻¹ and 224 μM, respectively (Table 1 and SI Appendix, Fig. S6). Considering the k_off value, the exchange-induced broadening was reasonably observed only for the Ile-370 resonance because the ¹³C CS difference between the free and MgADP-bound states (Δδ¹³C = 218 Hz) is more than twofold larger than other resonances (SI Appendix, Figs. S3 and S6), which coincided with the prediction from the fast exchange limit (30). Notably, both K_D and k_off values reasonably agreed with those obtained by line-shape analysis applied to the Ile-370 resonance using the program TITAN (31) (SI Appendix, Fig. S7A). As Ile-370 is in close proximity to the phosphate-binding region of a nucleotide (PBR; residues 371 through 375) (32) (SI Appendix, Fig. S5C), the observed CSP was directly caused by binding of the phosphate groups of ADP.

The titration of hPGK with MgAMPPNP, a nonhydrolysable ATP analog, generally exhibited CSPs in the fast exchange time regime. In contrast to the MgADP titration, very little CSP was observed for the Ile-370 resonance (Fig. 1 B, iii and SI Appendix, Fig. S3), indicating that phosphate groups of AMPPNP interacted less with the PBR, which was in agreement with the crystal structure of pmPGK bound with MgATP in which phosphate groups of ATP are distal from the PBR (33) (SI Appendix, Fig. S5D). The K_D value of MgAMPPNP for hPGK was likewise determined by the global CS-fitting analysis assuming fast exchange between the free and MgAMPPNP-bound states and found to be 894 μM (Table 1 and SI Appendix, Fig. S8).

We employed global CS fitting rather than individual CS fitting because CSPs are widely distributed throughout the protein. K_D values of each ligand determined by the individual CS fittings for each resonance were in close agreement with those determined by the global CS fitting; however, fitting errors (within 95% bootstrap CIs) of the global CS fittings were remarkably smaller than those of the individual CS fittings for each resonance (SI Appendix, Fig. S7B). The reliability of the global CS-fitting analysis was confirmed by isothermal titration calorimetry (ITC), which provided K_D values similar to those obtained from the global CS-fitting analysis (Table 1 and SI Appendix, Fig. S9 A–C and F). Notably, ligand affinities were highly affected by salt concentrations, especially for 3PG, as the K_D value of 3PG for hPGK was ∼1,000-fold lower at nonphysiological salt concentrations than at physiological concentrations (SI Appendix, Figs. S10–S13 and Table S1). Stronger ligand affinities for hPGK at nonphysiological salt concentrations were also reported in a previous study (34) (SI Appendix, Table S1).

Negative Cooperative Binding of MgADP and 3PG to hPGK.

To examine the binding cooperativity of the nucleotides and 3PG to hPGK, MgADP or MgAMPPNP was titrated to hPGK in the presence of 3PG and vice versa via two-ligand binding experiments. As CSs of all resonances were linearly perturbed in all titration experiments (SI Appendix, Fig. S14), global CS-fitting analysis was performed similarly as described in Ligand Bindings to hPGK at the Physiological Salt Concentration by assuming fast exchange for 3PG and MgAMPPNP bindings and fast-to-intermediate exchange for MgADP binding. In this context, we built a four-state ligand-binding equilibrium model as shown in Fig. 1C and introduced a cooperativity factor, α, determined by so-called “exact analysis,” in which changes in the populations of all four states during the ligand titrations were considered (35). Nucleotides and 3PG bind to hPGK with negative cooperativity when α < 1 but with positive cooperativity when α > 1 (see more details in the Fig. 1C legend).

We first analyzed the binding cooperativity between MgADP and 3PG (SI Appendix, Fig. S15). The fitting quality was considerably good; however, a small but systematic fitting error beyond the 95% bootstrap CI was observed for several titration curves (Fig. 1 E, iv or SI Appendix, Fig. S15, ¹³C-I220). However, these error ranges were within the NMR digital resolution (SI Appendix, Fig. S16) and thus can be regarded as negligible. The α-value was determined to be 0.45, corresponding to the affinity reduction of 3PG and MgADP in the presence of their counter ligands by ∼2-fold, which compensated the free-energy reduction (ΔG) of binding by ∼2.1 kJ/mol (ΔΔG ∼= +2.1 kJ/mol) compared to their absence (Table 1). Therefore, MgADP and 3PG bound to hPGK with negative cooperativity at physiological salt concentrations. In addition, the k_off value of MgADP increased from 811 s⁻¹ to 1,348 s⁻¹ (Table 1). The ITC analysis also confirmed the negative binding cooperativity, but its α-value was larger than that determined from the NMR data (Table 1 and SI Appendix, Fig. S9 D and F). A large unfavorable enthalpic contribution to the cooperativity between 3PG and MgADP (ΔΔH = +16.1 kJ/mol as seen in SI Appendix, Fig. S9F) masked the apparent exothermic heat by ligand binding, especially at high ligand concentrations, indicating the ineffectiveness of ITC analysis for precise α (35). Indeed, a very small apparent exothermic heat (less than 1 kJ/mol) was observed by 3PG injection under nearly saturated conditions with MgADP (SI Appendix, Fig. S9D), resulting in increased uncertainty in the thermodynamic parameters obtained from the curve fittings owing to the low signal-to-noise ratio, leading us to employ the α-value from the NMR data for further analyses. Titration of MgADP to hPGK in the presence of 3PG demonstrated that the Ile-367 resonance shifted downfield from the 3PG-bound (EP) state toward the MgADP-bound (ED) state, crossing over the MgADP-3PG-bound (EDP) state (Fig. 1 D, i and E, ii and SI Appendix, Fig. S15). This indicated that the initially bound 3PG gradually dissociated from hPGK with increasing ED population. Conversely, in the presence of MgADP, the Ile-367 resonance shifted slightly downfield from the ligand-free (E) state along the linear trajectory between the E and EP states (Fig. 1 D, ii and E, iii), suggesting that MgADP binding shifts the conformational equilibrium toward the 3PG-unbound conformation. Upon the addition of 3PG, the Ile-367 resonance shifted upfield toward the EP state beyond the EDP state (Fig. 1 D, ii and E, iv and SI Appendix, Fig. S15), representing the dissociation of bound ADP with increasing EP population. Similar line crossings were also observed for several residues (SI Appendix, Fig. S15).

Positive Cooperative Binding of MgAMPPNP and 3PG to hPGK.

The binding cooperativity of MgAMPPNP and 3PG was investigated by assuming fast exchange (SI Appendix, Fig. S17), and the α-value was determined to be 1.36, indicating a slight increase in the affinity of MgAMPPNP and 3PG. It enhanced the free-energy reduction (ΔG) of binding by ∼0.8 kJ/mol (ΔΔG ∼= −0.8 kJ/mol) in the presence of their counter ligands compared to those in their absence (Table 1), demonstrating the positive binding cooperativity between MgAMPPNP and 3PG at the physiological salt concentration. A similar α-value was also obtained from the ITC experiments (Table 1 and SI Appendix, Fig. S9 E and F). Titration of MgAMPPNP to hPGK in the presence of 3PG induced the Ile-367 resonance to shift further upfield from the EP state (Fig. 1 F, i and G, ii and SI Appendix, Fig. S17), suggesting additional conformational changes at the hinge region that are anticipated to be open to a closed conformational change through the hinge-bending motion. Therefore, we recorded the HMQC spectra of hPGK in the TSA-complex state, which adopts a fully closed conformation (17). The addition of ammonium fluoride (NH₄F) to hPGK in the EDP state resulted in the disappearance of the original peaks and the appearance of new cross-peaks at different CSs, which is represented as the 3PG-MgF₃⁻-ADP TSA-complex state (SI Appendix, Fig. S18A). Notably, the Ile-367 resonance in the TSA-complex state appeared upfield from the EP state on the linear trajectory between the E and EP states (Fig. 1 F, i). As the relative orientation of the CHH in the TSA complex moved further away from Ile-367 (SI Appendix, Fig. S5B), the simultaneous binding of MgAMPPNP and 3PG shifts the conformational equilibrium toward the closed conformation with increasing MgAMPPNP-3PG–bound (EXP) population (Fig. 1 F, i). In contrast to MgADP binding, the Ile-367 resonance shifted slightly upfield from the E state in the presence of MgAMPPNP (Fig. 1 F, ii and G, iii), indicating that MgAMPPNP binding induces a shift in the conformational equilibrium toward the 3PG-bound conformation. Upon the addition of 3PG, the Ile-367 resonance shifted toward the EXP state beyond the EP state (Fig. 1 F, ii and G, iv and SI Appendix, Fig. S17). As the affinity of MgATP for PGK is nearly the same as that of MgAMPPNP (33), MgATP and 3PG also bind to hPGK, most likely with positive cooperativity under physiological salt conditions.

Dynamic Conformational Change of hPGK.

To investigate the molecular mechanism of hPGK domain closure through the hinge-bending motion, we performed mutation studies in which Lys-215 was mutated to Ala (K215A) and Asp-374 was substituted with Lys (D374K) or Asn (D374N), both of which function as critical hydrogen bonds to maintain the closed conformation (17) (Fig. 2A). As with the wild type, ligand titration experiments were performed (SI Appendix, Figs. S19–S28). The affinities of the K215A mutant to both MgADP and 3PG were similar to those observed in the wild type (Table 2 and SI Appendix, Table S2). However, no cross-peaks corresponding to the TSA complex emerged upon the addition of a 1,000-fold excess of NH₄F (SI Appendix, Fig. S18B), indicating that the K215A mutant hardly formed the TSA complex. This agrees with previous evidence that Lys-215 stabilizes the transfer phosphate of the nucleotide (36). The D374K mutant was also unable to form the TSA complex (SI Appendix, Fig. S18B), although it has a similar affinity for 3PG (Table 2) and a stronger affinity for MgADP (SI Appendix, Table S2). Notably, both mutants lost PGK activity in the forward reaction (Table 2). In contrast, two-ligand titrations of the D374N mutant with MgADP and 3PG exhibited larger CSPs (SI Appendix, Figs. S19 and S28), and some resonances overlapped well with those of the wild type in the TSA-complex state (Fig. 2B). The D374N mutant showed strong positive binding cooperativity between MgADP and 3PG (Table 2 and SI Appendix, Table S2); however, its k_cat value decreased by ∼60% compared with the wild type (Table 2 and SI Appendix, Fig. S29). We also collected paramagnetic relaxation enhancement (PRE) data to confirm domain closure. We designed the K130C/E259I mutant (hereafter hPGK^PRE), in which an MTSL tag was attached to a newly introduced cysteine in the N-terminal domain and an isoleucine was introduced in the C-terminal domain (SI Appendix, Fig. S30). PRE effects should be observed only in the closed conformation based on the calculated distances between the paramagnetic center of the MTSL and the δ1-methyl of Ile-259 (Fig. 2C and SI Appendix, Fig. S31). A large PRE effect was observed for the Ile-259 resonance of the hPGK^PRE wild type in the TSA-complex state, which corresponds to the fully closed conformation (17) (Fig. 2C and SI Appendix, Fig. S32). A similar large PRE effect was observed for the hPGK^PRE D374N mutant in the EDP state (Fig. 2C), showing that the strong positive binding cooperativity between MgADP and 3PG stabilizes the closed conformation.

Fig. 2.

Dynamic change of hPGK from the open to closed conformations. (A) Important hydrogen-bond interactions in the active site of the hPGK-3PG-MgF₃⁻-ADP TSA complex (PDB code: 2WZB). Lys-215 and Asp-374 are represented as cyan and blue sticks, respectively. ADP and 3PG are shown as stick models. The magnesium ion and an oxygen atom of a hydration water molecule are indicated in magenta and gray spheres, respectively. MgF₃⁻ moieties are represented as small orange spheres. Black dashed lines indicate hydrogen bonds. (B) Overlay of ¹H-¹³C HMQC spectra of 0.05 mM Ileδ1-[¹³CH₃]–labeled hPGK wild type (WT) and D374N mutant. Black, blue, and red spectra are shown for the hPGK wild type in presence of a 30-fold excess of [MgADP] and a 60-fold excess of [3PG] (WT; EDP), for the hPGK wild type in the presence of a 400-fold excess of [NH₄F] in addition to the black spectrum (WT; TSA complex), and for the hPGK D374N mutant in the presence of a 10-fold excess of [MgADP] and a 10-fold excess of [3PG] (D374N; EDP), respectively. Typical overlapped resonances of the red with the blue, but not with the black, are labeled. (C) PRE effects of the MTSL-labeled hPGK K130C/E259I mutant (hPGK^PRE WT) and the K130C/E259I/D374N mutant (hPGK^PRE D374N). (Top) Model structures of MTSL-labeled hPGK, which are built based on crystal structures of the 3PG-ADP–bound state (PDB code: 2XE7), and the TSA-complex state (PDB code: 2WZB). The MTSL attached at Cys-130 and the residue Ile-259 are represented as brown sticks. (Middle) PRE profiles of the MTSL-labeled hPGK^PRE WT and hPGK^PRE D374N. PRE profiles of the Ile-259 resonance in the free, the EDP, and the TSA-complex states of the hPGK^PRE WT are shown in black bars, whereas that in the EDP state of hPGK^PRE D374N is represented as a red bar. (Bottom) Schematic representations of the fully closed conformation of the hPGK WT and D374N mutant (D374N). Fully closed conformations of hPGK are depicted in green. ADP and 3PG are represented as blown triangles and red ellipses, respectively. The MgF₃⁻ and magnesium ion are shown as a blue square and magenta circles, respectively.

Table 2.

Binding affinities of 3PG for the hPGK wild type and various mutants in the absence and presence of MgADP and their kinetic parameters^*

	KD/mM		α	KM/mM	kcat/s−1
	−MgADP	+MgADP
WT	0.112 (0.106 to 0.119)	0.251 (0.239 to 0.263)	0.448 (0.426 to 0.470)	0.20 ± 0.06	795 ± 50
K215A	0.108 (0.102 to 0.113)	0.150 (0.146 to 0.155)	0.716 (0.695 to 0.739)	nd†	nd†
D374K	0.110 (0.104 to 0.115)	0.140 (0.132 to 0.147)	0.785 (0.745 to 0.829)	nd†	nd†
D374N	0.111 (0.106 to 0.115)	0.0656 (0.0656 to 0.0675)	1.69 (1.64 to 1.69)	0.16 ± 0.03	315 ± 15

*The K_D values in the presence of MgADP are calculated from the α-values (see Fig. 1C). Values in brackets represent the 95% bootstrap CIs. K_M and k_cat values, with their SDs, are calculated from three independent experiments.

†Not detectable.

Real-Time Monitoring of the hPGK Catalytic Reaction.

Because 1, 3-bPG is extremely unstable in aqueous solutions, its chemical synthesis is highly challenging. Therefore, we utilized a coupled enzymatic reaction of PGK and GAPDH together with a lactate dehydrogenase (LDH)-coupled nicotinamide adenine dinucleotide (NAD⁺)-regenerating system (37) to observe the hPGK reaction at different nucleotide concentrations using real-time NMR under physiological salt conditions (Fig. 3A). The complete conversion of glyceraldehyde-3-phosphate (GAP) to 1, 3-bPG by GAPDH catalysis was confirmed by the ¹H-NMR signals of their methylene protons in the absence of hPGK (Fig. 3 B, i and ii and SI Appendix, Fig. S33). A coupled reaction was initiated by adding GAPDH and LDH to the reaction mixture containing 0.05 mM hPGK and 0.3 mM GAP. Before adding GAPDH and LDH, GAP signals were still observed regardless of the nucleotide concentrations, confirming that GAP did not bind to hPGK (Fig. 3 C, i and E, i).

Fig. 3.

Real-time monitoring of hPGK catalytic reaction. (A) The sequential enzyme reaction scheme from the GAPDH to hPGK in the presence of an LDH-coupled NAD-regenerating system. (B) 1D ¹H-NMR spectra of the phosphoglycerate species. ¹H-NMR signals of methylene protons of GAP (i), 1, 3-bPG (ii), and 3PG (iii) are indicated by black, orange, and red arrows, respectively. The complete conversion from GAP to 1, 3-bPG by GAPDH catalysis are confirmed in the absence of hPGK (i and ii). 3PG signals are separately assigned in the presence of GAPDH and LDH and in the absence of hPGK (iii). The asterisks mark the signal from an impurity. (C–E) The coupling enzymatic reaction between GAPDH and hPGK in the absence of nucleotides (C), in the presence of 0.15 mM [MgADP] (D), and in the presence of 0.15 mM [MgADP] and 2 mM [MgAMPPNP] (E). (i) Ligand-detected 1D ¹H-NMR spectra of phosphoglycerate species. The spectrum before and after adding GAPDH and LDH are shown at the bottom and top, respectively. GAP signals are indicated by black arrows. The free and exchange signals of 3PG between the free and PGK-bound state are indicated by red and magenta arrows, respectively. The asterisks mark the signal from an impurity contained in the reaction mixtures. (ii) Overlay of a portion of ¹H-¹³C HMQC spectra of 0.05 mM Ileδ1-[¹³CH₃]–labeled hPGK. The spectra before adding GAPDH and LDH are shown in black. After adding the nucleotides in D and E, the spectra are changed to blue, as indicated by blue arrows. Upon addition of GAPDH and LDH in C–E, the spectra are further changed to red, as indicated by red arrows. The theoretical ¹³C CSs of the Ile-367 resonances in the ligand-free, the 3PG-bound, the MgADP-bound, the MgAMPPNP-bound, the MgADP-3PG–bound, and the MgAMPPNP-3PG–bound states are shown by gray dashed lines as E, EP, ED, EX, EDP, and EXP, respectively. Note that each CS is calibrated with respect to the CS of samples in the reaction mixture (−0.08 ppm). The linear trajectory between the E and EP states are shown by red dashed lines. (iii) Schematic representations of the ligand bindings to hPGK. hPGK are depicted in blue; 1, 3-bPG, 3PG, ADP, ATP, and AMPPNP are represented as orange stars, red ellipses, brown triangles, dark green squares, and light green squares, respectively. Magnesium ions are shown as magenta circles.

In the absence of nucleotides, the addition of GAPDH and LDH resulted in the disappearance of the GAP signals, as expected; however, no appearance of the 1, 3-bPG signal was observed (Fig. 3 C, i). The complete conversion of GAP to 1, 3-bPG by GAPDH catalysis yielded the equilibrium of 1, 3-bPG between the free and hPGK-bound states (Fig. 3 C, iii), which likely induced the exchange line broadening of the 1, 3-bPG signals. This suggests that the affinity of 1, 3-bPG for hPGK is in the nanomolar to micromolar range at physiological salt concentrations. Because all hPGKs were occupied by 1, 3-bPG (Fig. 3 C, iii), the Ile-367 resonance was upfield shifted to that corresponding to the 1, 3-bPG–bound state (Fig. 3 C, ii, red), exhibiting a nearly equal CS to the EP state. In the presence of 0.15 mM [MgADP], the free 3PG signals (see Fig. 3 B, iii) were observed upon the addition of GAPDH and LDH (Fig. 3 D, i). At most, half of the amount of 1, 3-bPG (0.15 mM) generated by GAPDH catalysis should be further converted to 3PG by PGK catalysis, resulting in the copresence of 3PG and 1, 3-bPG (Fig. 3 D, iii). Because of the higher affinity of 1, 3-bPG (nanomolar to micromolar) than 3PG (submillimolar, Table 1), hPGK preferentially binds 1, 3-bPG (Fig. 3 D, iii), resulting in the detection of free 3PG signals (Fig. 3 D, i). The Ile-367 resonance also shifted to the same position as that observed in the absence of nucleotides (Fig. 3 D, ii, red), which corresponds to the 1, 3-bPG–bound state. In the presence of 0.15 mM [MgADP] and 2 mM [MgAMPPNP], no clear free 3PG signals were observed; instead, the broad signal, confirmed as the exchange signal of 3PG between the free and PGK-bound states (SI Appendix, Fig. S34), was detected upon the addition of GAPDH and LDH (Fig. 3 E, i). This demonstrates that hPGK can bind to 3PG despite the existence of 1, 3-bPG. However, the large excess of MgAMPPNP over that of MgADP enabled 3PG binding to hPGK by positive binding cooperativity with MgAMPPNP, leading 3PG to the equilibrium between the free and PGK-bound states (Fig. 3 E, iii). In this case, hPGK adopts the equilibrium among the 1, 3-bPG–bound, EP, and EXP states that reflect the signal broadening of Ile-367 resonance along the ¹³C-axis (Fig. 3 E, ii).

Ligand-Bound States of hPGK in Different Cellular Environments.

To examine the ligand-bound state of hPGK in native cellular environments, we performed in-cell NMR measurements. We prepared perdeuterated hPGK to overcome the intrinsic poor sensitivity of in-cell NMR and delivered it into HeLa cells by electroporation (38, 39). The exogenously delivered hPGK was mostly distributed in the cytosol (SI Appendix, Fig. S35A). For in-cell NMR experiments, we managed the cellular environment using a bioreactor, which continuously supplies fresh media to maintain appropriate cell-culture conditions during NMR measurements (38, 40). In the absence of the bioreactor, the in-cell NMR spectrum of hPGK recorded for the first 30 min overlapped well with its in vitro spectrum in the ED state based on the Ile-367 and Ile-370 resonances (Fig. 4 A, i and SI Appendix, Fig. S35 B and C). This indicates that hPGK predominantly binds to ADP, in which the forward reaction is fully activated to generate ATP, which is in agreement with the fact that intracellular ATP is entirely depleted within 30 min under these conditions (40). In contrast, in the bioreactor, the Ile-367 resonance appeared between the E and EP states in the in-cell NMR spectrum recorded for the first 90 min (Fig. 4 A, ii and SI Appendix, Fig. S35 B and D), suggesting that the PGK reaction in the forward direction was inhibited to some extent by the presence of the 3PG-bound population. The cell state under this condition is considered normal, as the intracellular ATP concentration is essentially constant during NMR measurements (40). Note that the Ile-367 resonance is also shifted upfield by 1,3-bPG binding (Fig. 3 C, ii). However, the exogenous hPGK concentration in HeLa cells was estimated to be ∼20 μM based on its NMR signal intensities in the cell lysate, and the intracellular cytoplasmic concentration of 1, 3-bPG is ∼0.4 μM, which is ∼100-fold lower than that of 3PG (∼45 μM) (41), implying that the 1, 3-bPG–bound state is unlikely to be detectable.

Fig. 4.

Catalytic mechanism of hPGK. (A) Comparison of the in-cell NMR spectra of Ileδ1-[¹³CH₃]– and U-²H–labeled hPGK delivered into HeLa cells in the absence (i) and presence (ii) of the bioreactor. A portion of in-cell ¹H-¹³C HMQC spectra (shown in black) is overlaid with the in vitro spectra of the EP (show in magenta) and ED (shown in cyan) states. The Ile-367 and Ile-370 resonances are labeled with the same colors with respect to each spectrum. The theoretical ¹³C CSs of the Ile-367 resonances in the E, ED, EP, and EDP states are shown by gray dashed lines. Note that each CS was calibrated with respect to the CS of the undeuterated samples (−0.07 ppm). The linear trajectory between the E and EP states are shown by red dashed lines. (B) The simulated population shifts of the ligand-bound states in response to the changes in the [ATP]/[ADP] ratio. The population shifts of the E, EP, ED, EDP, EX, and EXP states against the [ATP]/[ADP] ratio are plotted in solid black, dashed black, solid red, dashed red, solid blue, and dashed blue lines, respectively, in the experimental condition (i), assuming no ligand-binding cooperativities [α(ADP) = 1 and α(ATP) = 1] (ii), assuming further stronger ligand-binding cooperativities [α(ADP) = 0.2 and α(ATP) = 5.0] (iii), and at the nonphysiological salt concentration in the Hepes-KOH buffer (iv). S and N represent the starved and normal conditions, respectively. The populations of the ED and EDP states in the starved conditions are shown in filled red circles and triangles, whereas those in the normal condition are indicated in open red circles and triangles. The populations of the EX and EXP states in normal conditions are represented in open blue circles and triangles. (C) Schematic representation of the catalytic reaction scheme of hPGK established by this study. A catalytic turnover cycle is illustrated with a step-by-step sequence from substrate binding to product release. The open, partially closed, and fully closed conformations of hPGK are depicted in blue, cyan, and green, respectively; 1, 3-bPG, 3PG, ADP, and ATP are represented as orange stars, red ellipses, brown triangles, and dark green squares, respectively.

Simulation of the Population Shift among Different Ligand-Bound States.

Based on the K_D and the α-values determined by the two-ligand binding experiments, we simulated the population shift in response to changes in the [ATP]/[ADP] ratio. We fixed the total PGK concentration to 130 μM, as reported in the cytosolic fluid of rabbit psoas muscle fibers (42), the free 3PG concentration to 45 μM (41), and free [ATP] + [ADP] to 5 mM in all simulations. Since mammalian [ATP]/[ADP] ratios are expected to range from 1 to >100 (2, 43, 44), we defined the [ATP]/[ADP] ratio as 1 and 100 in the starved and normal conditions, respectively.

In the starved condition, the major population is the ED state, which has a large capacity for 1, 3-bPG binding (Fig. 4 B, i). Assuming no binding cooperativity between MgADP and 3PG (Fig. 4 B, ii), the ED population decreases, whereas the EDP population increases, indicating that the negative binding cooperativity between MgADP and 3PG contributes to accelerating the forward reaction. In cells, ATP depletion caused by hypoxia results in a switch to anaerobic glycolysis for energy production (45). Thus, the hPGK forward reaction is activated, which requires rapid 3PG release. This is in accordance with our ligand-detected NMR results in the presence of MgADP (Fig. 3D). Moreover, this activation indeed occurs in the native cellular environment, as confirmed by in-cell NMR experiments under ATP-depleting conditions (Fig. 4 A, i). Under normal conditions, the EX and EXP populations were dominant (Fig. 4 B, i), indicating that most of the hPGK binds to ATP and some also bind to 3PG. In cells under this condition, it is not necessary to accelerate glycolysis, as mitochondrial respiration is a major energy source. 3PG was observed in the PGK-bound state in the presence of excess MgAMPPNP over MgADP (Fig. 3E), indicating the partial inhibition of the PGK forward reaction to prevent the overproduction of ATP. It specifically occurred in the normal cellular environment as well, as observed in the in-cell NMR experiment (Fig. 4 A, ii). It was also revealed that the reverse reaction during gluconeogenesis only occurred when the EXP population was abundant (Fig. 4 B, i). If there is no binding cooperativity between MgATP and 3PG, the EXP population decreases (Fig. 4 B, ii), indicating that the positive binding cooperativity between MgATP and 3PG enhances the reverse reaction. If the ligand-binding cooperativities are stronger, the EXP population increases beyond the EX population under normal conditions (Fig. 4 B, iii), resulting in the overregulation of the forward reaction. At nonphysiological salt concentrations, the ED population was almost absent under the starved condition, and under normal conditions, ∼100% of the EXP population was detected (Fig. 4 B, iv) because the strong affinity of 3PG for hPGK prevents both the forward and reverse reactions. In contrast, at physiological salt concentrations, the affinity of 3PG for hPGK is reduced to a range of submillimolar concentrations of K_D. The K_D values could be influenced by the accumulation of chloride ions at the 3PG-binding site, as demonstrated by a molecular dynamics simulation of bacterial PGK in the presence of KCl (46). The same K_D order of 3PG and nucleotides enables quick alternation between positive and negative ligand-binding cooperativities. These findings clearly show that hPGK intrinsically modulates its function through the ligand-binding cooperativities between adenine nucleotides and 3PG, which is finely tuned to respond to changes in the [ATP]/[ADP] ratio at physiological salt concentrations.

Discussion

In this study, we demonstrated that the ligand-binding cooperativities of hPGK induced a shift in the population among different ligand-bound states under physiological salt conditions. Based on these ligand-binding cooperativities, we propose a reaction scheme for the hPGK forward reaction, as shown in Fig. 4C. The reaction starts in the ADP-bound state (step 1). Our simulation revealed that hPGK consistently binds to either ADP or ATP under cellular conditions (Fig. 4 B, i). In addition, 1, 3-bPG binding (step 2) triggers phosphate transfer by adopting a fully closed conformation through the hinge-bending motion (step 3). This is an essential step for PGK catalysis, as demonstrated by the defect in the activity of the K215A and D374K mutants (Table 2), both of which cannot stabilize the fully closed conformation. The completion of the phosphate transfer results in the ATP- and 3PG-bound state (steps 4 and 4′) with a partially closed conformation, as the positive binding cooperativity between 3PG and MgAMPPNP shifts the equilibrium toward the closed conformation (Fig. 1 F and G). Under normal conditions, an equilibrium for 3PG exists between the free and PGK-bound states (Fig. 4 B, i), allowing spontaneous 3PG dissociation (step 4), which contributes to the physiological glycolytic flux. The dissociation of 3PG shifts equilibrium toward an open conformation. The replacement of ATP by ADP (step 5), which depends on the [ATP]/[ADP] ratio (Fig. 4 B, i), results in the ADP-bound state and consequently proceeds into the next cycle (steps 6 through 1). Under starved conditions, the forward reaction is accelerated by the quick replacement of ATP by ADP because of the decreasing [ATP]/[ADP] ratio, resulting in the ADP- and 3PG-bound state (steps 4′ through 5′), shifting the equilibrium to the open conformation (Fig. 1 D and E). The negative binding cooperativity between MgADP and 3PG then induces 3PG dissociation (step 5′) in preparation for the subsequent cycle (steps 6 through 1). This negative binding cooperativity potentially causes the equilibrium between ADP and 3PG dissociation; however, 3PG dissociation is preferred because of the higher [ADP] than [3PG] in the starved condition, resulting in only a small EP population (Fig. 4 B, i). As expected from the increase in the k_off value of MgADP in the presence of 3PG (Table 1), this negative binding cooperativity could also increase the k_off value of 3PG, which accelerated the 3PG dissociation. In contrast, ADP binding stabilized the fully closed conformation of the D374 mutant (Fig. 2 B and C), which hindered rapid 3PG dissociation and resulted in a slowdown in its turnover rate (steps 4′ through 6). Indeed, the k_off value of MgADP in the presence of 3PG was significantly lower than that of the wild type (SI Appendix, Table S2). As expected from the strong positive binding cooperativity between MgADP and 3PG for this mutant, the k_off value of 3PG could also be reduced in the presence of MgADP. These findings clearly demonstrated that an intrinsic fully closed conformation was enzymatically unfavorable, in agreement with the results of the SAXS measurements, in which the PGK was in an open conformation majorly with short periods of domain closure during the catalytic reaction (15). Thus, the reaction scheme proposed here indicates that the intracellular [ATP]/[ADP] ratio is a critical factor in PGK regulation. The importance of MgATP levels for enzyme regulation has also been shown for protein kinase A (PKA), which controls the ability of allosteric ligands to stabilize open versus closed topologies of the regulatory subunit of PKA (PKA-R), which alters the affinity to its catalytic subunit (PKA-C) (47). The PGK reaction is near equilibrium (ΔG ∼= 0) on glycolysis (5); thus, even minute changes of free energy of the ligand-binding cooperativities (ΔΔG ∼= −0.8 to 2.1 kJ/mol, as presented in Negative Cooperative Binding of MgADP and 3PG to hPGK and Positive Cooperative Binding of MgAMPPNP and 3PG to hPGK) can greatly contribute to glycolytic flux regulation. This was fully consistent with previous evidence that glycolytic flux is increased with only small changes in metabolite concentration (5). In contrast, stronger binding cooperativities could be unfavorable for the flexible control of the flux, as demonstrated in our simulation (Fig. 4B). The contribution of a shallow free-energy landscape to enzyme function has previously been shown for PKA-R, in which the population of two conformers within a nearly degenerated free-energy landscape can be reversibly controlled to function as an allosteric conformational switch (48).

The regulation of enzyme activity by sensing a product or substrate is typically known as allosteric regulation. In glycolysis, the enzymes involve three committed irreversible steps by hexokinase (49), phosphofructokinase (50), and pyruvate kinase (51), which are allosterically regulated. In contrast, PGK catalyzes a bidirectional reaction to flexibly control the glycolytic flux in near-equilibrium steps; therefore, a mechanism different from allosteric regulation is required. Previously, PGK regulation by ADP binding at the secondary site has been proposed (52). However, our global CS-fitting analyses were satisfactorily fitted to a single nucleotide site model at physiological salt concentrations. The alternation of positive and negative ligand-binding cooperativities thus acts as a simple but efficient switch for PGK regulation. Switches between positive and negative cooperativities have been reported for PKA-C (53) and Src kinase (54). These kinases bind ATP and unphosphorylated substrates via positive cooperativity for PKA-C and negative cooperativity for Src kinase, while ADP and the phosphorylated product reveal a negative binding cooperativity for PKA-C and a positive binding cooperativity for Src kinase. In contrast to these kinases that catalyze unidirectional pathways, PGK catalyzes a bidirectional pathway. It binds 3PG (the product of the glycolytic pathway) via negative cooperativity with ADP, while positive cooperativity with ATP, which is a distinct switch mechanism from enzymes catalyzing unidirectional reactions that can be an intrinsically programmed regulation mechanism for an enzyme in a bidirectional pathway.

PGK possesses a self-regulatory mechanism at the protein level as well as various transcriptional regulations. Quantity control by transcriptional regulation generally takes minutes to hours during the cell cycle (55, 56). In contrast, self-regulation at the protein level occurs on a faster timescale, enabling a quick response to changes in the intracellular environment, which can be an intrinsic emergency response for cells to avoid critical damage.

Materials and Methods

Sample Preparation.

All the constructs for the expression of wild-type and mutant hPGKs were cloned into the pGEX-6P-3 plasmid (Cytiva) with an N-terminal glutathione S-transferase (GST) tag and a PreScission protease recognition site. The resulting expression constructs were analyzed using an ABI3730xl DNA analyzer (Thermo Fisher Scientific) at the Support Unit for Bio-Material Analysis in the Research Resources Division of the Institute of Physical and Chemical Research (RIKEN) Center for Brain Science. The GST-tagged hPGK proteins were expressed in the Escherichia coli BL21 Star (DE3) strain (Thermo Fisher Scientific). For the selective Ileδ1-[¹³CH₃]–labeled hPGK, the cells were grown overnight in 5 mL Luria-Bertani (LB) medium and subsequently inoculated into 1 L M9 minimal medium at 37 °C. A total of 1 h before the induction of protein expression, 50 mg/L [methyl-¹³C, 3-²H₂]-α-ketobutyric acid (Cambridge Isotope Laboratories, Inc.) was added to the cell culture and protein expression was induced by adding 1 mM isopropyl β-d-1-thiogalactopyranoside when the cells reached an optical density at 600 nm of 0.6, and the culture was incubated at 24 °C overnight. For the expression of uniformly ²H-labeled hPGK, 4 g/L [²H₇]-d-glucose and D₂O were used. To purify hPGK, the cells were lysed by sonication in 20 mM Tris-Cl buffer (pH 8.0) containing 200 mM NaCl, 2 mM dithiothreitol (DTT), and a tablet of cOmplete protease inhibitor mixture (Merck). The lysate was centrifuged, the supernatant was applied to a GSTrap column (Cytiva), and the bound proteins were eluted with reduced glutathione. The GST tag was cleaved by PreScission protease (Cytiva) at 4 °C overnight and separated from hPGK using a GSTrap column (Cytiva). The hPGK protein was further purified by chromatography using a Superdex 200 size-exclusion column (Cytiva) followed by a Mono Q anion-exchange column (Cytiva).

NMR Analyses.

Two-dimensional ¹H-¹³C HMQC experiments were conducted using 0.05 protein solutions of Ileδ1-[¹³CH₃]–labeled hPGK in 25 mM Hepes-KOH buffer (pH 7.2) containing 120 mM KCl and 5 mM KPi with 10% D₂O on Avance III 600-MHz or 700-MHz spectrometers with a QCI CryoProbe (Bruker BioSpin) at 37 °C unless stated otherwise. The ligand-detected one-dimensional (1D) ¹H NMR experiments were performed under the same conditions as the ¹H-¹³C HMQC experiments, except that 25 mM Pipes buffer (pH 7.2) containing 120 mM KCl and 5 mM KPi with 10% D₂O were used. The ¹H-NMR spectra were recorded with a Carr–Purcell–Meiboom–Gill (CPMG) T₂ filter (57, 58) to suppress the ¹H-NMR signals derived from hPGK with an excitation sculpting sequence for water suppression. In-cell NMR spectra were recorded using band-selective excitation along the ¹³C-dimension to suppress the t₁ noise derived from the intense signals of HeLa cells with an ¹H CS of 0.8 to 1.2 ppm. All in-cell NMR experiments were initiated for less than 30 min after sample preparation. The durations of in-cell NMR experiments were ∼30 min and ∼90 min in the absence and presence of the bioreactor (38) (InsightCell; Bruker BioSpin), respectively. All NMR spectra were processed using TopSpin 3.6 software (Bruker Biospin), and the data were analyzed using the NMRFAM-SPARKY software (59).

Global CS-Fitting Analysis.

To analyze the one- or two-ligand titration NMR experiments, we constructed a two-part observation model. The first part computed concentrations of ligand-bound protein species, which, for the two-ligand system, were computed numerically using the same formula as previously described (35). The second part computed the CSs, which were performed by assuming either fast or intermediate chemical exchange. To simulate intermediate exchange, we adopted an approximation assuming no intrinsic transverse relaxation (30). The best-fit parameters, K_D, k_off, and α, which minimized the sum of the squared residual errors, were searched using the simplex method using MATLAB R2020a (MathWorks). Bootstrap CIs were calculated with the percentile method using 1,000 bootstrap samples, which were constructed by residual resampling (60). For more details, see SI Appendix, SI Theory.

Coupled Enzymatic Reaction of PGK with GAPDH.

1, 3-bPG was enzymatically synthesized by a GAPDH reaction in the presence of the LDH-coupled NAD⁺-regenerating system as described previously (37), with some modifications. Briefly, the reaction was performed in a 4-mm NMR sample tube to detect NMR signals immediately after the enzyme reaction. Before the GAPDH reaction, 250 μL reaction mixture containing 25 mM Pipes buffer (pH 7.2) containing 120 mM KCl, 2 mM NAD⁺, 2 mM sodium pyruvate, 2 mM KH₂PO₄, and 0.3 mM GAP with 10% D₂O was prepared. Subsequently, the GAPDH reaction was performed by adding five units of GAPDH (rabbit muscle; Merck) and 20 units of LDH (rabbit muscle; Merck) to the reaction mixture. To monitor the sequential reaction from GAPDH to hPGK, hPGK (0.05 mM) and various concentrations of nucleotides were added to the reaction mixture.

Simulation of Ligand-Bound States of PGK.

The PGK concentration of each ligand-bound state was calculated for a given cellular ligand concentration, dissociation constants, and cooperativity factors as follows:

$$\left[\text{E}\right]=\frac{{[\text{E}]}_{\text{total}}}{1+\frac{[3\text{PG}]}{K_{\text{D}(3\text{PG})}}+\frac{[\text{ADP}]}{K_{\text{D}(\text{ADP})}}+\frac{[3\text{PG}]\left[\text{ADP}\right]{\mathit{\alpha }}_{(\text{ADP})}}{K_{\text{D}(3\text{PG})}{K}_{\text{D}(\text{ADP})}}+\frac{[\text{ATP}]}{K_{\text{D}(\text{ATP})}}+\frac{[3\text{PG}]\left[\text{ATP}\right]{\mathit{\alpha }}_{(\text{ATP})}}{K_{\text{D}(3\text{PG})}{K}_{\text{D}(\text{ATP})}}}$$

$$\left[\text{EP}\right]=\frac{\left[\text{E}\right][3\text{PG}]}{K_{\text{D}(3\text{PG})}}$$

$$\left[\text{ED}\right]=\frac{\left[\text{E}\right][\text{ADP}]}{K_{\text{D}(\text{ADP})}}$$

$$\left[\text{EDP}\right]=\frac{\left[\text{E}\right]\left[3\text{PG}\right]\left[\text{ADP}\right]{\mathit{\alpha }}_{\left(\text{ADP}\right)}}{K_{\text{D}\left(3\text{PG}\right)}{K}_{\text{D}\left(\text{ADP}\right)}}$$

$$\left[\text{EX}\right]=\frac{\left[\text{E}\right][\text{ATP}]}{K_{\text{D}(\text{ATP})}}$$

$$\left[\text{EXP}\right]=\frac{\left[\text{E}\right]\left[3\text{PG}\right]\left[\text{ATP}\right]{\mathit{\alpha }}_{\left(\text{ATP}\right)}}{K_{\text{D}\left(3\text{PG}\right)}{K}_{\text{D}\left(\text{ATP}\right)}},$$

where square brackets denote the cellular concentration of the chemical species; E, EP, ED, EDP, EX, and EXP denote the ligand-bound states of PGK defined in Negative Cooperative Binding of MgADP and 3PG to hPGK and Positive Cooperative Binding of MgAMPPNP and 3PG to hPGK; ${\left[\mathrm{E}\right]}_{\mathrm{total}}$ denotes the total PGK concentration; K_D is the dissociation constant of the specified ligand; and α is the cooperativity factor between 3PG and the specified nucleotide.

Data Availability

All study data are included in the article and/or supporting information.