Cancer-associated somatic mutations in human phosphofructokinase-1 reveal a critical electrostatic interaction for allosteric regulation of enzyme activity

Glycolysis is an ancient pathway that breaks down glucose to generate energy, reducing agents, and biosynthetic precursors to support cellular functions. Adenosine triphosphate (ATP) produced by the glycolytic pathway is a major source of energy for cellular activities including muscle contraction, the transmission of neuronal signals, and active transport. In addition to ATP, glycolysis generates metabolic intermediates that serve as anabolic building blocks for cellular components such as DNA, lipids, and proteins. While glycolysis is essential for normal cell function, its dysregulation has been linked to the initiation and progression of many diseases, including cancer, diabetes, and neurodegenerative disorders. Aberrant activation of glycolysis can promote tumor growth and metastasis by providing cancer cells with energy and metabolic intermediates necessary for their survival [1]. Similarly, dysregulated glycolysis in diabetes can lead to the accumulation of toxic byproducts that damage tissues and impair insulin sensitivity [2]. In neurodegenerative diseases such as Alzheimer's, disrupted glucose metabolism has been shown to impair neuronal function and contribute to disease progression [3]. Therefore, determining the molecular mechanisms underlaying aberrant glycolysis in disease initiation and progression is essential for developing novel therapeutic interventions.

The rate of glycolytic flux is determined largely by the activity of rate-limiting enzymes, one of the most important of which is phosphofructokinase-1 (PFK1). PFK1 is known as the ‘gatekeeper’ of glycolysis catalyzing the reaction committing glucose to breakdown by irreversibly converting fructose 6-phosphate (F6P) to fructose 1,6-bisphosphate (F1,6bP). PFK1 is highly evolutionarily conserved, from bacteria to humans. Prokaryotic PFK1 protomers are approximately half the size of their mammalian counterparts and are less allosterically regulated [4,5]. Eukaryotic PFK1 evolved from a prokaryotic ancestral enzyme by gene duplication, tandem fusion, and evolution which created a novel tetramer interface, an amino-terminal ‘catalytic’ domain, and a carboxy-terminal ‘regulatory’ domain (Figure 1A,B) [6–8]. The increased size and complexity of eukaryotic PFK1 enabled enhanced allosteric regulation, with catalytic activity tightly controlled through multiple mechanisms including allosteric regulation by small molecule effectors, post-translational modification, and modulation of protein levels through transcriptional and post-translational mechanisms [9–13]. In general, signals of low cellular energy activate enzymatic activity while signals of high energy inhibit PFK1. PFK1 is inactive as individual or dimers of protomers but active as tetramers [14]. The shift between active tetramers and inactive protomers is regulated by small molecule effectors, with inhibitors promoting tetramer disassembly and activators promoting tetramer assembly [14–17].

Three PFK1 gene products exist in mammalian cells: muscle (PFKM), liver (PFKL), and platelet (PFKP). All three gene products are present at various levels in most tissues, except for skeletal muscle, which almost exclusively expresses PFKM, and the liver which predominantly expresses PFKL [18,19]. PFK1 can form homo- and hetero-tetramers, with some isoforms being able to assemble into larger catalytically active structures [20–22]. The genes encoding these isozymes are located on different chromosomes, with alternative splicing resulting in several transcript isoforms for each gene. The three human PFK1 isoforms are ∼70% identical at the amino acid level, with the differences in sequence accounting for isoform-specific regulation and activity. For instance, PFKM has the highest affinity for F6P and displays pH-dependent ATP inhibition while PFKP has greater activation by ADP/AMP and the highest level of inhibition by citrate [7,23,24]. The binding of biphosporylated hexoses, specifically fructose 1,6-bisphosphate (F1,6bP) and fructose 2,6-bisphosphate (F2,6bP), also has isoform-specific effects on PFK1. F2,6bP is the most potent natural agonist of PFK1; it both stabilizes and relieves ATP inhibition in all three human PFK1 isoforms. PFKP and PFKL show increased sensitivity to F2,6bP allosteric modulation compared with PFKM. Conversely, F1,6bP stabilizes all three human PFK1 isoforms but only relieves ATP inhibition in PFKL and PFKM [25–27].

Dysregulation and mutations of PFK1 have been implicated in the initiation and progression of various human diseases. For example, somatic missense and frameshift mutations in PFKM can cause glycogen storage disease type VII (Tarui disease), an autosomal recessive disease that impairs glycogen breakdown [28,29]. It is unclear whether somatic mutations in PFK1 contribute to other diseases such as cancer. Many cancer cells exhibit elevated glucose uptake and lactic acid excretion regardless of oxygen levels, a phenomenon known as the Warburg effect [30,31]. This altered metabolic program supplies energy and building blocks required to sustain rapid proliferation and adaption to the tumor microenvironment. Consistent with increased glycolytic flux, PFK1 expression and activity are elevated in cancer cell lines and primary tumors [32,33]. However, recent findings have demonstrated that metabolic reprogramming of cancer cells is more complex, suggesting that metabolic flexibility is essential for cancer initiation and progression [34–36]. In line with this idea, post-translational modifications which inhibit or degrade PFK1, such as glycosylation and ubiquitination, are also increased in cancers [10,11,37]. Hundreds of cancer-associated somatic mutations have been identified between the three human PFK1 genes. However, the impact of these mutations and their contribution to disease initiation and progression is unknown. While most disease-associated mutations in PFK1 are thought to inactivate the enzyme in a recessive manner, such as those identified in Tarui disease, we previously identified a cancer-associated mutation in PFK1, PFKP-D564N, that acted in a dominant manner when expressed in cells [15]. In this study, we characterized how cancer-associated somatic mutations of this critical aspartic acid-arginine pair alter the biochemical and structural regulation of PFK1, providing insights into the molecular mechanisms underlying allosteric regulation of human PFK1.

Aspartic acid 564 is located near the catalytic interface of human PFKP on helix 6′ (E. coli numbering) [38]. The carboxylic acid group of aspartic acid 564 faces the guanidine group of arginine 319 on helix 11 and the pair was predicted to form an electrostatic interaction. Somatic mutations to either the aspartic acid or arginine were found in multiple PFK1 isoforms (Figure 1C). In PFKL, mutations to both the homologous aspartic acid and arginine residues were identified, where aspartic acid 553 was changed to asparagine in kidney cancer and the arginine was replaced with a glutamine in lung cancer. While PFKM did not have mutation to the orthologous residues, mutations in adjacent residues, aspartic acid 309 and arginine 555, have been reported in Tarui disease and cancers, respectively [39].

We asked if the Asp–Arg pair was evolutionarily conserved by performing sequence and structural alignments. We found that residues homologous to Arg319 and Asp564 were conserved in single-cell eukaryotes such as Saccharomyces cerevisiae and in prokaryotes such as Geobacillus stearothermophilus PFK1 (Figure 1D,E and Supplemental Table S1). S. cerevisiae has two PFK1 subunits, PFK1 and PFK2, which form a hetero octamer. The octamer is a dimer of tetramers, with each tetramer equivalent in size and shape to the mammalian PFK1 tetramer. Both PFK1 and PFK2 subunits have residues orthologous to arginine 319 and aspartic acid 564 and are predicted to form an electrostatic interaction. PFK1 from prokaryotes is half the size of their eukaryotic orthologs, with the prokaryotic dimer equivalent in size to the eukaryotic protomer. Using PFK1 from G. stearothermophilus as a representative structure of bacterial PFK1, we found that the arginine and aspartic acid residues are conserved and located at the dimer interface, with the arginine from one protomer interacting with the aspartic acid from a second protomer. In summary, the location of aspartic acid 564 and an electrostatic interaction with arginine 319 are evolutionarily conserved from bacteria to humans, suggesting an important role in PFK1 activity.

We generated recombinant human protein to determine the impact of cancer-associated mutations on PFK1 structure, activity, and regulation. Wild-type and mutant PFK1 proteins were expressed in a baculovirus expression system and purified to homogeneity, as determined by Coomassie-stained SDS–PAGE gels (Supplemental Figure S1). While we successfully generated PFKP-D564N and PFKL-D553N, we were unable to express and purify PFKL-R310Q, and therefore PFKL-R310Q was excluded from further analysis. We hypothesize that the less conservative nature of the arginine to glutamine substitution decreased the stability of the mutant protein. Native-PAGE analysis of purified protein was performed to determine if the mutations dramatically altered the quaternary structure of the enzyme. PFK1 protomers are ∼90 kDa in size, but assemble into tetramers of ∼480 kDa, as shown by Native-PAGE (Supplemental Figure S1). Bands larger than tetramers were also observed, predominantly in lanes containing PFKL, consistent with the ability of PFKL to form filaments of stacked tetramers [20]. No significant differences were observed between the wild-type and mutant PFK1 proteins, suggesting that the mutations do not alter the assembly of PFK1 into tetramers.

Next, we asked if a disruption of the Asp–Arg pair altered enzyme activity or allosteric regulation of PFKP. Consistent with our previous findings [15], recombinant PFKP-D564N had decreased maximum velocity of reaction, decreased affinity for the substrate fructose 6-phosphate (F6P), and decreased cooperativity when compared with wild-type PFKP (Figure 2A and Table 1). We tested if the mutation altered the allosteric regulation of three effectors. First, we examined the response to ATP, which is both a substrate and allosteric inhibitor of PFK1. The amino-terminal catalytic domain of eukaryotic PFK1 contains an inhibitory ATP binding site with mM affinity [8,40]. Under the conditions tested, both wild-type and mutant PFKP showed similar inhibition by ATP, with ∼50% inhibition observed at 4 mM (Figure 2B). This suggests that disruption of the Asp–Arg electrostatic interaction at the domain interface does not significantly affect allosteric regulation by ATP. Next, we asked if activation by adenosine monophosphate (AMP), a signal of low cellular energy, was altered. AMP/ADP binds with high µM affinity to its effector-binding site located in the center of the PFK1 protomer at the interface between the catalytic and regulatory domains [8,41]. As AMP relieves ATP inhibition, we performed the assays at the inhibitory ATP concentration of 3 mM. Wild-type PFKP had a 10-fold increase in catalytic activity in response to increasing concentrations of AMP (Figure 2C), while PFKP-D564N was only activated 3-fold. We also determined how the addition of AMP effected the affinity for F6P. AMP concentration of 0.5 mM increased the affinity of wild-type PFKP 2-fold, from 2.16 ± 0.08 mM in the absence to 0.88 ± 0.03 mM in the presence of AMP (Figure 2D and Table 2). PFKP-D564N also showed an ∼2-fold increase in affinity for F6P, from 4.64 ± 0.68 mM to 2.25 ± 0.34 mM. These data indicate that Asp–Arg pair disruption attenuates, but does not prevent, AMP from relieving ATP inhibition. Finally, we asked if the mutation altered allosteric regulation by biphosphate sugars. F1,6bP and F2,6bP bind to an effector-binding site in the regulatory domain that is composed of residues from two protomers across the catalytic interface [15,42–45]. F1,6bP stabilizes all three human PFK1 isoforms while it only relieves ATP inhibition of PFKL and PFKM but not PFKP due to isoform-dependent coordination of F1,6bP in the allosteric sugar-binding site [25,26]. We asked if the mutation altered the ability of F1,6bP to stabilize wild-type and mutant PFKP. We focused our attention on allosteric regulation by F1,6bP as mammalian PFKP has only been co-crystallized with F1,6bP (39) and not F2,6bP. PFKP undergoes irreversible, time-dependent thermal inactivation at 55°C in the absence of ligands, with ∼10% activity remaining after 60 min of incubation (Figure 2E). PFKP-D564N lost activity at a decreased rate compared with the wild-type protein, with ∼50% activity remaining after 60 min of incubation, suggesting that PFKP-D564N has increased stability compared with the wild-type enzyme. F1,6bP protected against the loss of activity in both wild-type and mutant PFKP, both of which retained activity in the presence of 200 µM of the sugar. The mutation had no significant impact on affinity for F1,6bP, with an apparent dissociation constant of 159 nM ± 39 and 509 nM ± 310 observed for PFKP and PFKP-D564N, respectively (Figure 2F). In summary, PFKP-D564N was less active and was more resistant to thermal inactivation when compared with the wild-type protein.

We next asked if the aspartic acid to asparagine mutation had a similar effect on the activity and allosteric regulation of PFKL. The disruption of the Asp–Arg pair had a more profound effect on PFKL than on PFKP, with PFKL-D553N having almost no measurable activity under conditions that the wild-type enzyme was active (Figure 3A). We screened assay conditions to include activators and found that partial activation of the enzyme could be achieved in the presence of AMP and ammonium sulfate (Figure 3B and Table 2). Interestingly, the addition of either AMP or ammonium sulfate individually did not cause a measurable increase in PFKL-D553N activity. Therefore, the response of PFKL-D553N to allosteric effectors was measured in the presence of 5 mM ammonium sulfate and 1 mM AMP. However, as the wild-type enzyme showed no allosteric inhibition in these conditions (Figure 3B), we were only able to compare the relative regulation of the wild-type and mutant enzymes. We tested the regulation of wild-type and mutant PFKL by ATP, AMP, and F1,6bP. We found that ATP inhibited both wild-type and mutant PFKL in a similar manner, with >80% inhibition achieved by 4 mM ATP (Figure 3C). To test the ability of AMP to activate PFKL, PFKL-D553N was assayed under conditions in which no ATP inhibition was observed (0.125 mM; Figure 3D) while wild-type PFKL was assayed at inhibitory ATP concentrations (3 mM). Increasing concentrations of AMP activated both wild-type and mutant PFKL (Supplemental Figure S3A). However, the relative increase in activation was higher for wild-type PFKL than for PFKL-D553N, 22-fold compared with 3.5 times, respectively. Finally, we asked if activation of PFKL by F1,6bP was altered. Wild-type PFKL showed robust activation by F1,6bP (Figure 3E). In contrast, PFKL-D553N showed no activation by F1,6bP under any of the conditions evaluated (Figure 3E and Supplemental Figure S3). Taken together, these results indicate that the aspartic acid to asparagine mutation had a more severe impact on PFKL than PFKP, decreasing enzymatic activity and attenuated the ability of F1,6bP activate the enzyme.

We asked if the structure of PFKP-D564N could provide insight into how disrupting the Asp–Arg pair might result in the biochemical changes observed experimentally. We determined the crystal structure of PFKP-D564N at 3.6 Å resolution (Figure 4). The atomic model had a good agreement with the crystallographic data and the expected geometric parameters (Table 3). The asymmetric unit contained one tetramer, with four protomers having essentially the same conformation. Tetrahedral anions, namely phosphate and sulfate, were bound to PFKP-D564N. Each protomer contained two phosphates; two protomers contained a phosphate bound to the allosteric sugar binding site and a prokaryotic effector site while two protomers contained both phosphates bound to prokaryotic effector sites. In contrast with previous PFKP structures, PFKP-D564N crystalized in the presence of 200 µM ammonium sulfate. We found 13 sulfate ions bound asymmetrically to the tetramer. Five sulfate residues were bound to allosteric sugar binding sites while the remaining ions were located on the surface of the tetramer. Despite the presence of ATP in the crystallization buffer, we did not observe the nucleotide bound to either the catalytic or the inhibitory binding sites. Overall, PFKP-D564N had similar structure to the wild-type enzyme (Figure 4A) with asparagine 564 found in close proximity to arginine 319. The change in amino acid identity from aspartic acid to asparagine was predicted to weaken the electrostatic interaction with guanidinium group of arginine (Figure 4B); however, it remained unclear how disrupting the Asp–Arg pair results in the changes in activity and regulation of PFKP-D564N. Thus, it is possible that the PFKP-D564N crystal lattice packing preferentially selected a certain protein conformation. To overcome this limitation, we used molecular dynamic simulations to ask how disruption of the Asp–Arg pair altered the structure of the enzyme free from crystallization constrains. We observed the backbone movements of wild-type PFKP (PDB: 4WL0) and PFKP-D564N over time and determined the root square mean deviation for wild-type (green) and mutant (blue) protein (Figure 4C). We found that wild-type PFKP had decreased movement compared with PFKP-D564N. This indicates that PFKP-D564N is further from its preferred state when locked in an active conformation in the crystal lattice and suggests that PFKP-D564N may preferentially occupy the less active t-state.

Finally, we asked if the expression of PFKP-D564N could alter glucose metabolism in a clinically relevant model since previous effects of the mutation were evaluated in a rat adenocarcinoma cell line expressing EGFP-labeled PFKP [15]. To address this question, FLAG-tagged wild-type PFKP and PFKP-D564N were stably expressed in the human breast cancer cell line MDA-MB-231 using a lentiviral expression system. EGFP was used as a control for viral transduction. Wild-type and mutant FLAG-PFKP were expressed at similar levels, as determined by western blot analysis (Figure 5A). The cellular rate of glycolysis, as determined by the extracellular acidification rate (ECAR) was monitored in real time using a Seahorse XF (Figure 5B). Expression of FLAG-PFKP had a minor impact on cellular glucose metabolism compared with EGFP control cells. Cells expressing FLAG-PFKP had a similar rate of basal glycolysis (Figure 5C), consistent with the high allosteric regulation of PFK1. However, a small decrease in the glycolytic capacity and glycolytic reserve (Figure 5D) was observed, suggesting that exogenous expression of FLAG-PFKP decreases the cell's maximal rate of glycolytic flux. The ratio of glycolytic activity (ECAR) to mitochondrial activity (oxygen consumption rate; OCR) was slightly decreased in FLAG-PFKP expressing cells, suggesting that expression of FLAG-PFKP rendered the cells slightly more glycolytic but did not dramatically alter the basal cellular phenotype (Figure 5E). In contrast, the expression of FLAG-PFKP-D564N displayed a dominant negative effect on cellular glycolysis in the presence of endogenous PFKP as FLAG-PFKP-D564N-expressing cells had a significant decrease in the basal rate of glycolysis when compared with either EGFP or FLAG-PFKP expressing cells. Consistent with this finding, cells expressing FLAG-PFKP-D564N were more reliant on mitochondrial activity than either EGFP or FLAG-PFKP expressing cells. Interestingly, the maximal glycolytic capacity of wild-type and mutant FLAG-PFKP expressing cells was the same, suggesting that low cellular energy can lead to activation of FLAG-PFKP-D564N. When oxidative phosphorylation (OXPHOS) was inhibited by oligomycin, glycolytic flux increases to generate ATP. Therefore, an increased glycolytic reserve is consistent with activation of glycolysis in FLAG-PFKP-D564N-expressing cells by signals of low cellular energy such as AMP, which was able to activate the mutant enzyme (Figure 2C).

Here, we report a comprehensive structural and functional characterization of cancer-relevant mutations in human PFK1. Aspartic acid 564 in PFKP is located at the interface of the regulatory and catalytic domains and forms an electrostatic bond with asparagine 319. In addition to PFKP-D564N mutation in colon cancer identified previously [15], mutations to both members of the Arg–Asp pair were also found in PFKL (Figure 1C). Asp–Arg pair located at the interface of the regulatory and catalytic domains is conserved from bacteria to humans (Figure 1D,E and Supplemental Table S1), suggesting that these residues are important for the enzyme function. Using purified recombinant proteins and enzyme kinetics we show that mutations in Asp–Arg pair decrease enzyme activity and alter allosteric regulation. Allosteric signals from effectors binding to the catalytic domain, such as ATP inhibitory signal, and from effectors binding at the catalytic and regulatory domain interface (AMP), are transmitted to the enzyme's active site. However, the activation by AMP and F1,6bP was attenuated in the Asp-to-Arg mutants compared with the wild-type enzymes. Disruption of the Asp–Arg pair had a stronger impact on PFKL compared with PFKP, possibly due to the greater allosteric regulation and decreased stability of PFKL. We determined the crystal structure of PFKP-D564N at 3.6 Å resolution to examine molecular mechanism of reduced allosteric regulation. Notably, this is the first reported structure of a human disease-associated mutation in PFK1. Molecular dynamic simulations revealed that PFKP-D564N was further from its preferred state when locked in an active conformation in the crystal lattice.

Aerobic glycolysis entails increased uptake of glucose from the environment and excretion of lactic acid regardless of oxygen availability. This metabolic program confers advantage to cancer cells by promoting fast ATP production and providing biosynthetic precursors necessary to support rapid cell proliferation [46]. Consistent with this idea, several studies demonstrated that high PFK1 expression correlates with aggressive cancer phenotype and poor patient survival [10,13]. On the other hand, context-dependent metabolic plasticity has been described throughout cancer progression and metastatic cascade [47,48]. For example, high proliferation rate of tumor cells and hypoxia inside a tumor core result in elevated oxidative stress. Cancer cells adapt to reactive oxygen species (ROS) stress by redirecting glucose catabolism from glycolysis to other pathways, such as the pentose phosphate pathway (PPP). The PPP supplies nucleotide precursors for cell proliferation and NADPH used for both ROS detoxification and catabolic metabolism [49]. Likewise, several studies demonstrated that decreased PFK1 activity in cancer switches glucose flux to the PPP, resulting in better cell viability under ROS stress [11,34,50]. Consistent with those reports, we show here that cancer-associated somatic mutations in PFK1 lower activity and alter allosteric regulation of the enzyme. We observed that breast cancer cells expressing FLAG-PFKP-D564N displayed a decreased basal rate of glycolysis compared with control cells or cells expressing wild-type FLAG-PFKP (Figure 5). These data suggest that cancer mutations reducing PFK1 activity may provide a selective advantage for cancer cell survival in tumor microenvironments where ROS stress occurs. Oxidative stress can also be caused by matrix detachment and high oxygen content in blood circulation during metastatic cascade; therefore, low PFK1 activity may promote cancer progression at multiple stages of the disease. These observations are in good agreement with those obtained for another rate-limiting glycolytic enzyme, pyruvate kinase. Cancer-associated mutations in pyruvate kinase decrease the enzyme activity and impair allosteric activation by F1,6bP, conditions that support a highly proliferative metabolic program [51]. In summary, inactivating mutations in PFK1 may contribute to malignant metabolic reprogramming to benefic adaptation to tumor microenvironment and cancer cell survival during metastatic cascade.

Understanding the molecular mechanisms of how somatic mutations alter PFK1 activity and regulation may provide insights into drug design for treating cancer, glycogen storage disease VII, and other conditions affected by altered glycolysis. The finding that a PFK1 mutation can have a dominant impact on the metabolic program of the cell may impact the way we treat Tarui disease. Individuals with Tarui disease have impaired glycolytic function, which causes the accumulation of glycogen in tissues including muscle, liver, and red blood cells. Patients with Tarui disease often display exercise intolerance, muscle weakness, cramps, and myoglobinuria [52]. Gene replacement therapy has been proposed as a potential treatment for the disease. However, this strategy requires the mutant protein does not interfere with the exogenous replacement enzyme. Therefore, knowledge of the individual mutations will be essential to predict the efficacy of these treatments. Alternatively, mutations in PFK1 which act in a dominant manner could be to targeted by small molecule pharmaceuticals to specifically activate the enzyme.

Inhibiting aerobic glycolysis by therapeutically targeting glycolytic enzymes is a current proposed strategy to limit cancer progression. Proliferative cancer cells are dependent on glucose to provide energy and biosynthetic precursors. However, this strategy is limited by the number of existing drugs, specific to a particular enzyme. Non-specific drugs, such as 2-deoxyglucose, are available but their use is limited by poor clinical outcomes due to adverse side-effects in patients. This is likely due to off-target effects due to multiple metabolic fates of glucose in the cell, such as processing through the PPP and protein glycosylation. Targeting of specific isoforms of glycolytic enzymes is an alternate approach that may circumvent the negative impacts of 2-deoxyglucose. However, knowledge of the identity and function of these enzymes in the cell as well as generation of specific targeting molecules will be essential for the success of this approach. Specific small molecule modulators of several glycolytic enzymes, including PFKFB3 and glucose transporters (GLUTs) inhibitors as well as pyruvate kinase isoform M2 (PKM2) activators and inhibitors, are currently being evaluated for clinical efficacy [53–59]. Until recently, there were no specific and high-affinity compounds targeting PFK1. Our studies on disruption of the Asp–Arg pair in PFK1 highlight the importance of helices 11 and 6′ in allosteric regulation and reveal that mutated amino acids on these helices contribute to dysregulated glucose metabolism in cancers and glycogen storage disease (Figure 1A) [15,39]. Residues in helices 11 and 6′ are involved in binding to allosteric effectors. For example, the allosteric activator AMP/ADP [8,41] binds to a central pocket and interacts with residues on these helices. Recently, the first non-natural agonist for PFK1, NA-11, has been reported. NA-11 interacts with residues on helices 11 and 6′, binds to the AMP/ADP effector site activating PFK1 with nM affinity [42]. Furthermore, a critical residue imparting isoform-selectivity of NA-11, lysine 315 (PFKL numbering), is located on helix 11. While there are no specific and high-affinity compounds inhibiting PFK1, the presumptive citrate-binding site includes lysine 567, located adjacent to aspartic acid 564 on helix 6′ [60]. Our findings raise the possibility of generating small molecule inhibitors of PFK1 acting to disrupt the Asp–Arg electrostatic interaction or to alter the relative orientation of helices 11 and 6′. Such compounds would be a valuable resource to determine the role of PFK1 in cancer cell metabolic reprogramming and have high potential for therapeutic use.

In summary, we used somatic mutations of PFK1 to inform our understanding of allosteric regulation by identifying key amino acid residues that are involved in the regulation of enzyme activity. Mutations to a critical Asp–Arg pair located at the catalytic-regulatory domain interface of PFK1 inhibits enzymatic activity in vitro and decreases glycolytic flux when expressed in cells in a dominant manner. Targeted compounds that disrupt the Asp–Arg pair may allow for highly selective therapeutic intervention to inhibit PFK1 activity.

Recombinant human PFKL (NM_002626), PFKP (NM_002627.4), and the mutants were generated as described [15,20] using the Bac-to-Bac Expression system (Invitrogen) in sf9 cells. Cells were lysed in a lysis buffer (20 mM HEPES, pH 7.5; 80 mM potassium phosphate; 1 mM 2-mercaptoethanol; 10% glycerol; 10 mM imidazole, and EDTA-free Protease Inhibitor Cocktail (Abcam)) with a Dounce homogenizer. The lysates were cleared by centrifugation and the supernatants were incubated with HisPur cobalt resin (ThermoFisher Scientific). The resin was washed with lysis buffer, with high salt lysis buffer containing 1 M NaCl, and with lysis buffer again. PFK1 was eluted in lysis buffer supplemented with 100 mM imidazole. Fractions containing protein were pooled, and the buffer was exchanged for freezing buffer (20 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) pH 7.5; 1 mM DTT; 500 µM ammonium sulfate; 5% glycerol; 1 mM ATP, and 100 µM EDTA) using a HiPrep 26/10 desalting column (GE Healthcare Life Sciences) and concentrated using an Amicon Ultracel-30K Centrifugal Filter Unit (MilliporeSigma). Protein concentration was determined using a Bradford Protein Assay kit (ThermoFisher Scientific), and aliquots snap frozen in liquid nitrogen and stored at −80°C. The quaternary structure of proteins was determined by Native-PAGE using NativePAGE Novex Bis-Tris gel system (ThermoFisher Scientific) according to the manufacturer's instructions and using Light Blue Cathode Buffer. Proteins were diluted in Native PAGE buffer (ThermoFisher Scientific) with the addition of 15 mM HEPES pH 7.5; 37.5 mM KCl; 0.75 mM DTT; and 0.9 mM ammonium sulfate. Proteins were resolved in 3–12% NativePAGE Bis-Tris gel (ThermoFisher Scientific). Proteins were identified using the Pierce Silver Stain kit (ThermoFisher Scientific) according to manufacturer's instructions.

PFKP-D564N was crystallized in a buffer containing 125 µM ATP by a micro batch method at 22°C, nucleation of rod-shaped crystals occurred within four days. A solution of PFKP-D564N (5 mg/ml) was mixed in a 1 : 1 ratio with crystal solution (0.2 mM ammonium sulfate; 0.1 M Bis-Tris pH 5.5; 25% w/v Polyethylene glycol monomethyl ether 3350). Crystals were harvested and cryoprotected in 0.2 mM ammonium sulfate; 0.02 M Bis-Tris pH 5.5; 28% w/v Polyethylene glycol monomethyl ether 3350, and 30% ethylene glycol and were flash frozen in liquid nitrogen. Data collection was performed on The Northeastern Collaborative Access Team (NE-CAT) 24-IDE beamline at The Advanced Photon Source (APS). The diffraction images were processed using XDS [61], pointless, and aimless [62–64]. The structure of ATP-bound PFKP (PDB accession number 4xyk) [15] was used to determine the structure using the molecular replacement method on Phenix Phaser-MR software [65] and structural refinement was performed with phenix.refine [66]. The data processing and refinement statistics are summarized in Table 3.

Images highlighting the position of Arg319 and Asp564 of human PFKP (PDB file 4XYK) were generated in the PyMOL Molecular Graphics System, Version 2.0 (Schrödinger, LLC). Hydrogen bonds were predicted by the PyMOL ‘Polar Contacts' command. However, the hydrogen bonds were not rigorously determined due to the low resolution of available PFK1 crystal structures. Structural alignment of G. stearothermophilus (PDB file 6PFK), S. cerevisiae (PDB file 3O8O), human PFKL (PDB file 7LW1), and human PFKP (PDB file 4XYK) was performed in PyMOL. The molecular modeling and docking studies were done using the software suit of MOE 2020.0901 (Chemical Computing Group). The protein crystal structure 4XZ2 was obtained from the Protein Data Bank (39), and prepared for downstream applications by adding hydrogens, fixing missing side chains, and adding charges at pH 7.4. Molecular dynamics simulations were done using YASARA 20.7.4 modeling suite (YASARA Biosciences GmbH) using the NOVA force field. The protein structures were mutated from D564N with the protein tools in MOE 2020.09 and minimized with the AMBER10 force field [67].

In vitro, PFK1 activity was determined as previously described [15,20]. Briefly, the auxiliary enzymes aldolase, triosephosphate isomerase, and glycerol phosphate dehydrogenase were purchased from MilliporeSigma, desalted using an Amicon Ultracel-10K Centrifugal Filter Unit, and resuspended in 2× Buffer containing 100 mM HEPES pH 7.5; 200 mM KCl; 2 mM DTT, and 0.3 mM NADH. The temperature was equilibrated to 25°C for 5 min before initiating the reaction with 10 mM MgCl₂. Absorbance at 340 nm was measured using a Varioskan LUX Multimode microplate reader (ThermoFisher Scientific). For assays containing F1,6bP, the rate of ADP production was determined using the auxiliary enzymes lactate dehydrogenase (LDH) and pyruvate kinase (MilliporeSigma) as described below. One unit of activity was defined as the amount of enzyme that catalyzed the formation of 1 µmol of fructose 1,6-bisphosphate per min. All experiments were performed with at least two separate protein preparations.

The assays were performed as previously described [27] with some modifications. PFKP was diluted to 100 ng/µl in a buffer containing 50 mM HEPES pH 7.5, 100 mM KCl, 1 mM DTT and incubated at 55°C for indicated times. After incubation, the residual PFK1 activity was measured using a pyruvate kinase/LDH auxiliary enzyme plate reader assay in a buffer containing 50 mM HEPES pH 7.5, 100 mM KCl, 1 mM DTT, 0.25 mM NADH, and 0.2 mM phosphoenolpyruvic acid, 0.25 mM ATP, 5 mM F6P, and auxiliary enzymes. To determine F16bP altered PFKP stability, 200 µM F1,6bP were added to the samples before the incubation at 55°C. In experiments determining F1,6bP affinity, recombinant wild-type and mutant PFKP was diluted to 100 ng/µl in a buffer and proteins were incubated with 0–20 µM F1,6bP for 60 min at 55°C and the residual enzymatic activity measured.

MDA-MB-231 (MDA-MB-231) cells were purchased from ATCC and maintained in high-glucose DMEM (Gibco) supplemented with 10% heat-inactivated FBS (Gibco) and penicillin-streptomycin (Gibco) in 5% CO₂ at 37°C. Cells were confirmed to be mycoplasma-free by using LookOut mycoplasma detection kit (MilliporeSigma) and authenticated by STR profiling. EGFP, FLAG-PFKP, and FLAG-PFKP D564N were cloned into pLenti6.3 vector (ThermoFisher Scientific). Lentivirus particles were generated using HEK293T cells (ATCC) and ViralPower lentiviral particles packaging mix (ThermoFisher Scientific) according to the manufacturer instructions. MDA-MB-231 cells were infected with the virus particles in presence of 5 µg/ml of Polybrene (MilliporeSigma), selected with 15 µg/ml blasticidin (Gibco), and protein expression was verified by western-blotting.

Cells were lysed in a SDS sample buffer (63 mM Tris-hydrochloride, pH 6.8; 2% sodium dodecyl sulfate (SDS), and 10% glycerol) and protein concentration was determined using a BCA assay (ThermoFisher Scientific). An amount of 30 µg of proteins were separated by 10% polyacrylamide gel and transferred onto a polyvinylidene difluoride (PVDF) membrane (Millipore). Membranes were blocked in 5% non-fat milk in Tris Buffered Saline with 0.1% Tween-20. After the membranes were incubated with primary and secondary antibodies, immune complexes were detected using a Pierce ECL substrate (ThermoFisher Scientific) in an Amersham Imager 600 imager (GE Healthcare Life Sciences). The antibodies used in this study were against actin (Millipore, MAB1501; 1 : 5000), PFKL (Abcam, ab181064; 1 : 1000), FLAG-tag (Sigma, F1804; 1 : 1000).

OCR and ECAR of live cells was analyzed on a Seahorse XFe96 Analyzer (Agilent Technologies). Cells were plated onto Seahorse XF96 microplate (Agilent Technologies) and allowed to attach overnight. Before assay, cells were equilibrated for 1 h in a non-CO₂ incubator at 37°C with XF DMEM base medium (Agilent Technologies) supplemented with 2 mM Glutamine (Agilent Technologies). Metabolic functions were examined through sequential injection of 10 mM glucose, 1 µM oligomycin, and 100 mM 2-deoxy-glucose (2-DG). The data were normalized by cell number measured by a CyQUANT Cell Proliferation Assay (ThermoFisher Scientific) according to the manufacturer's recommendations. Initial experiments were performed to ensure linearity of the cell number vs fluorescence intensity for each cell line. EGFP-expressing cells showed an increased baseline compared with other cell lines, due to the expression of the fluorescent protein, and cell number was quantified independently. All other cell lines showed similar results and the cell number was estimated from a single standard curve. Analysis was performed with the Agilent Seahorse Wave Desktop software.

The data represents means ± SEM. Statistical analysis was performed with GraphPad Prism software and a P-value <0.05 was considered statistically significant. At least three experiments with at least two independent protein preparations were performed to evaluate enzymatic activity.

All individual experiments described have been repeated at least three times. Initial observations of the impact of PFK1 mutations were made by Jacob Miller and Bradley Webb. These were reproduced independently and expanded upon by Maria Voronkova and Heather Hansen. Key results have been repeated independently by three experimentalists.

X-ray coordinates for PFKP-D564N have been deposited with the RCSB PDB databank and are accessible under the ID 7TFF.

The authors declare no competing financial interests.

This work was supported by West Virginia Start-up funding to B.A.W. and A.R., Visual Sciences CoBRE project leader funding to B.A.W. (P20GM144230), and NIGMS grants to A.R. (R01GM133857) and W.J.G. (U54GM104942 and P20GM109098). This work is based upon research conducted at the Northeastern Collaborative Access Team beamlines, which are funded by the National Institute of General Medical Sciences from the National Institutes of Health (P41GM103403). The WVU Mitochondria, Metabolism & Bioenergetics Working Group, supported by the Community Foundation for the Ohio Valley Whipkey Trust.

Bradley Webb: Conceptualization, Validation, Investigation, Visualization, Methodology, Writing — original draft, Project administration, Writing — review and editing. Maria A. Voronkova: Conceptualization, Investigation, Methodology, Writing — original draft, Writing — review and editing. Heather L. Hansen: Validation, Investigation, Writing — original draft, Writing — review and editing. Madison P. Cooper: Validation, Investigation, Methodology. Jacob Miller: Conceptualization, Investigation, Methodology. Narayanasami Sukumar: Methodology. Werner J. Geldenhuys: Visualization, Methodology, Writing — review and editing. Aaron Robart: Visualization, Methodology, Writing — review and editing.

The authors would like to thank Mark Pinti and Morgan Glass for their excellent technical assistance. Seahorse analysis was performed by the WVU Mitochondria, Metabolism & Bioenergetics Working Group.

2-DG

2-deoxy-glucose

AMP

adenosine monophosphate

ATP

adenosine triphosphate

ECAR

extracellular acidification rate

HEPES

4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid

LDH

lactate dehydrogenase

OCR

oxygen consumption rate

PFK1

phosphofructokinase-1

PPP

pentose phosphate pathway

ROS

reactive oxygen species