Michaelis-like complex of mouse ketohexokinase isoform C

The structure of ketohexokinase C in complex with a natural substrate and product ADP mimicking the Michaelis complex is reported and a comparison with other structures with and without ligands shows that it undergoes a conformational change that places the enzyme in a catalytically competent form. These findings will be of interest to those in the areas of enzymology, enzyme structure–function relationships and drug discovery.

Abstract

Over the past forty years there has been a drastic increase in fructose-related diseases, including obesity, heart disease and diabetes. Ketohexokinase (KHK), the first enzyme in the liver fructolysis pathway, catalyzes the ATP-dependent phosphorylation of fructose to fructose 1-phosphate. Understanding the role of KHK in disease-related processes is crucial for the management and prevention of this growing epidemic. Molecular insight into the structure–function relationship in ligand binding and catalysis by KHK is needed for the design of therapeutic inhibitory ligands. Ketohexokinase has two isoforms: ketohexo­kinase A (KHK-A) is produced ubiquitously at low levels, whereas ketohexo­kinase C (KHK-C) is found at much higher levels, specifically in the liver, kidneys and intestines. Structures of the unliganded and liganded human isoforms KHK-A and KHK-C are known, as well as structures of unliganded and inhibitor-bound mouse KHK-C (mKHK-C), which shares 90% sequence identity with human KHK-C. Here, a high-resolution X-ray crystal structure of mKHK-C refined to 1.79 Å resolution is presented. The structure was determined in a complex with both the substrate fructose and the product of catalysis, ADP, providing a view of the Michaelis-like complex of the mouse ortholog. Comparison to unliganded structures suggests that KHK undergoes a conformational change upon binding of substrates that places the enzyme in a catalytically competent form in which the β-sheet domain from one subunit rotates by 16.2°, acting as a lid for the opposing active site. Similar kinetic parameters were calculated for the mouse and human enzymes and indicate that mice may be a suitable animal model for the study of fructose-related diseases. Knowledge of the similarity between the mouse and human enzymes is important for understanding preclinical efforts towards targeting this enzyme, and this ground-state, Michaelis-like complex suggests that a conformational change plays a role in the catalytic function of KHK-C.

1. Introduction

Fructose is a monosaccharide that is found in many plants, specifically in the fruits of flowering plants (Park & Yetley, 1993 ▸). It is commonly used as a food additive because it has a higher perceived sweetness than glucose or sucrose; therefore, lesser amounts can be used to achieve the same level of sweetness (Fontvieille et al., 1989 ▸). Upon ingestion, fructose is shuttled to organs including the liver, kidneys and brain (Mayes, 1993 ▸). In the liver, where the majority of fructose metabolism occurs, fructose is metabolized quickly as fructolysis lacks the positive- and negative-feedback mechanisms that tightly control glycolysis.

In the liver, the catabolism of fructose triggers an ATP-depletion pathway that results in the production of uric acid, which is thought to be linked to de novo lipogenesis and the production of excess triglycerides (Johnson et al., 2013 ▸). The accumulation of uric acid and triglycerides as a result of excess fructose consumption is associated with a number of disease states, including type-2 diabetes, obesity and non-alcoholic fatty liver disease (NAFLD), all of which have rapidly increased over the past forty years (Nakagawa et al., 2006 ▸).

The enzyme that initiates fructose metabolism in the liver is ketohexokinase C (KHK-C), which is also known as fructokinase, and catalyzes the ATP-dependent phosphorylation of fructose, forming fructose 1-phosphate. In humans, KHK is encoded by a gene that includes an intragenic duplication at exon 3 known as exon 3a and exon 3c. During mRNA processing the exons are alternatively spliced, resulting in two isoforms: ketohexokinase A (KHK-A) and KHK-C (Bonthron et al., 1994 ▸). KHK-A is expressed ubiquitously at low levels, whereas KHK-C is found mainly in the liver, kidney, islet cells and small intestine, which are the primary sites of fructose metabolism in the body (Hayward & Bonthron, 1998 ▸). The two isozymes differ in their kinetic behavior towards fructose, with KHK-C having a greater affinity for fructose, which is reflected by a tenfold lower K_m value for fructose (Asipu et al., 2003 ▸). Several of the residues encoded by exon 3 that differ between KHK-C and KHK-A are within 3–4 Å of the fructose-binding site.

Studies of KHK-knockout mice expressing neither the KHK-A nor the KHK-C isozyme have shown that the action of KHK is connected to the adverse metabolic effects of fructose. KHK-knockout mice are protected from chronic fructose-induced hypertriglyceridemia, insulin resistance, hepatic lipid accumulation and weight gain compared with wild-type KHK mice (Ishimoto et al., 2012 ▸). Based on these findings, one method to potentially treat fructose-associated diseases is through the inhibition of KHK.

KHK belongs to the phosphofructokinase B (PfkB) family of sugar kinases. Other members of this family, all of which use an alcohol moiety as the phosphoryl-group acceptor, include ribokinase, adenosine kinase, inosine kinase and 2-keto-3-deoxygluconate kinase (Cabrera et al., 2010 ▸). Members of the PfkB family are identified by the presence of two conserved sequence motifs. The first motif is a highly conserved GG motif located in the N-terminal region of these enzymes (residues 40 and 41 in mKHK-C, which has the same numbering as human KHK-C). These two glycine residues form part of the hinge between two protein domains (Fig. 1 ▸a). The second motif, which is found in the C-terminal region, is a DTXGAGD motif (residues 252–258 in mKHK-C) that is involved in the binding of ATP and forms the basis of an oxyanion hole that is used to stabilize the transition state during phosphoryl transfer (Mathews et al., 1998 ▸). This motif also includes a conserved aspartate residue, which serves as a catalytic base (Sigrell et al., 1999 ▸). Despite the similarity in tertiary structure which defines them as a family, the overall sequence identity among PfkB family members is less than 30% (Mathews et al., 1998 ▸; Sigrell et al., 1999 ▸). Several members, including KHK, have been shown to be biological dimers (Sigrell et al., 1997 ▸; Raushel & Cleland, 1977 ▸).

Figure 1.

(a) Subunit of the hKHK-C dimer (PDB entry 3nbv) highlighting the N-terminal GG motif (magenta, residues 40 and 41) that links the two domains and the C-terminal DTXGAGD motif (orange, residues 252–258) that forms part of the ATP-binding site. (b) The dimer of hKHK-C (PDB entry 3nbv) highlighting the residues from each subunit (purple and yellow) that form the β-clasp and Asp27 (green), which extends into the active site of the opposing subunit.

The structures of human KHK-C (hKHK-C) and mouse KHK-C (mKHK-C) have previously been determined (Trinh et al., 2009 ▸; Ebenhoch et al., 2023 ▸). Each subunit has two distinct domains: a central α/β-fold domain and a four-stranded β-sheet domain. hKHK-C is a dimer in solution, with an interface made through an extension of the four-stranded β-sheet domain that forms a β-clasp structure between the two subunits. The β-clasp comprises residues 29–31 from one subunit forming a β-sheet with a β-strand comprising residues 108–113 from the second subunit (Fig. 1 ▸b). Different conformations of this domain are found in both human and mouse KHK, as well as other PfkB family members, and seem to be correlated with different binding states (Trinh et al., 2009 ▸; Sigrell et al., 1999 ▸; Ebenhoch et al., 2023 ▸). A feature of KHK that emerges as a result of β-clasp rotation is that Asp27 from one subunit is placed into the active site of the other subunit (Maryanoff et al., 2012 ▸). We propose here that this rotation and placement of Asp27 is facilitated by substrate binding.

We aimed to determine the structure of mKHK-C to facilitate the use of mouse models to explore the preclinical efficacy of KHK-C inhibitors and to highlight differences between human and mouse KHK-C to aid in both the design and the development of KHK inhibitors. In addition, the structure of mKHK-C bound to substrates was sought in order to understand the structure–function relationship and the detailed catalytic mechanism of KHK, as the only structure of a liganded complex of mKHK is with an inhibitor that is selective for KHK-A over KHK-C (Ebenhoch et al., 2023 ▸). Here, we report an X-ray crystal structure of mouse KHK-C and note that residues identical to those in the human isozyme make binding interactions with ATP and fructose. The conformational change suggested by comparison of the structure of this ground-state, Michaelis-like complex with that of unliganded hKHK-C sheds light on the role that induced fit may have in catalytic function.

2. Methods

2.1. Expression and purification of mouse KHK-C

The sequence encoding mKHK (EST clone I5201 made in pME18S-FL3) was amplified using mutagenic primers to introduce a His₆ tag, a linker sequence (ENLYFQGPRG) at the N-terminus of the open reading frame (ORF), an NcoI restriction-enzyme site that overlaps with the ATG start codon, and a HindIII site after the TGA stop codon. The primer sequences were 5′-GGC GCC ATG GGC CAC CAC CAC CAC CAC CAC GAA AAC CTG TAC TTC CAA GGC CCG CGT GGC GAA GAG AAG CAG ATC CTG TG-3′ and 5′-GGC GAA GCT TCA CAC AAT GCC ATC AAA-3′. The resulting PCR product was subcloned into pPB1, a high-copy plasmid with ampicillin resistance (Beernink & Tolan, 1992 ▸). The construct was used to express the mKHK-C protein in BL21 (DE3) competent Escherichia coli cells. The expression of mKHK-C was performed in a 1 l culture as described previously (Asipu et al., 2003 ▸), with the exception that purification used a pre-packed 5 ml GE HisTrap FF column. Briefly, the harvested cells were resuspended in 20 ml lysis buffer consisting of 250 mM MOPS/KOH, 10 mM EDTA, 2 mM DTT, 1% glycerol, pepstatin A, 2.5 mM PMSF and DNAse and RNAse (50 µg ml⁻¹ each). Cell lysis was performed using a French press and the lysate was centrifuged at 34 500g for 30 min at 4°C. The supernatant fraction was passed over a GE HisTrap FF column (5 ml) using a peristaltic pump at a flow rate of 0.5 ml min⁻¹. After loading the entire fraction, the flow was stopped and the column was incubated for 4 h at 4°C. Nonspecific proteins were eluted with 200–400 ml 10 mM HEPES/KOH pH 7.5 containing 20 mM imidazole. The His₆-tagged protein was eluted with 20 ml buffer consisting of 10 mM HEPES/KOH pH 7.5 plus 0.5 M imidazole and the eluent was dialyzed twice against a >100-fold volume of 10 mM HEPES/KOH pH 7.5, 0.5 mM DTT. The dialysate was concentrated to 15 mg ml⁻¹ protein using an Amicon Ultrafiltration unit with a 10 kDa molecular-weight cutoff membrane. A dye-binding assay was used to determine the protein concentration (Bradford, 1976 ▸). The protein was >95% pure as assessed by SDS–PAGE and the polydispersity was <40% as assessed by dynamic light scattering (DLS). The final yield was 50 mg of mKHK per litre of cell culture. The protein was stored as 15 mg ml⁻¹ aliquots at 4°C in 10 mM HEPES/KOH, 0.5 mM DTT pH 7.5.

2.2. Activity and kinetic assays

Buffer optimization for purified mKHK-C was performed against 96 different conditions, varying the pH, salts, glycerol content and ionic strength, using a protein thermal shift assay with detection by differential scanning fluorimetry (DSF) analysis (Huynh & Partch, 2015 ▸).

Protein-activity and substrate-kinetic assays were carried out using an enzyme assay in which the activity of KHK was coupled to those of pyruvate kinase and lactate dehydro­genase (Adelman et al., 1967 ▸). Conversion of NADH to NAD⁺ by lactate dehydrogenase was monitored using the decrease in the absorbance at 340 nm over a 5 min period at 22°C. The assay cocktail consisted of 33 mM TEA–HCl pH 7.4, 100 mM KCl, 20 mM MgCl₂, 1.33 mM phosphoenolpyruvate, pyruvate kinase/lactate dehydrogenase enzymes mix (4–10 U ml⁻¹) and NADH (300 µM). For specific activity, the concentrations of substrates used were 3 mM ATP and 20 mM fructose. Steady-state kinetic assays contained 0.03–20 mM fructose and 0.02–3.0 mM ATP, with the highest concentration of one substrate being used when varying the other. All absorbances were measured in triplicate, allowing the removal of outliers, and assays were repeated a minimum of five times.

2.3. Crystallization, X-ray data collection and processing

Initial crystallization conditions were screened using mKHK-C in the presence of ADP, fructose and nitrate (as a metaphosphate mimic) with Index HT (Hampton Research) by vapor diffusion in hanging-drop geometry. The final optimized condition used a 1:1:2 ratio of well solution (1.2 M ammonium citrate tribasic pH 7, 20% glycerol), ligand mixture (1.3 M KNO₃, 100 mM MgCl₂, 220 mM fructose, 52 mM ADP) and purified protein (15 mg ml⁻¹ in 10 mM HEPES/KOH pH 7, 0.5 mM DTT) in a drop volume of 2 µl. Drops were equilibrated against 500 µl well solution at 17°C. Within seven days, these conditions yielded cubic and rectangular crystals with the longest dimensions varying from 20 to 100 µm (Fig. 2 ▸). The crystals were protected for cryocooling by the addition of glycerol to a final concentration of 20% and were cryocooled in a stream of gaseous nitrogen at 100 K. X-ray diffraction data were collected at the Boston University Chemical Instrumentation Center (BU-CIC) using a Bruker AXS X8 Proteum-R X-ray source with a PLATINUM135 charge-coupled area detector. Data collection was performed at 100 K using 0.5° oscillations and a total of 269.50° of data were collected. The Proteum program (Bruker) was used to scale and integrate the data. Data-collection statistics are presented in Table 1 ▸.

Figure 2.

Typical crystals of mKHK-C complexed with ADP and fructose. The crystals measure approximately 20–100 µm on each edge.

Table 1. X-ray crystallographic data-collection and refinement statistics for mKHK-C bound to ADP, fructose and nitrate.

Values in parentheses are for the highest resolution shell.

Source	BU-CIC
Detector	PLATINUM135 CCD
Wavelength (Å)	1.54
Data-collection temperature (K)	100
Resolution range (Å)	37.96–1.79 (1.86–1.79)
Space group	C2221
a, b, c (Å)	42.25, 78.45, 150.77
Total reflections	306134 (14710)
Unique reflections	23192 (1886)
Multiplicity	13.2 (7.8)
Completeness (%)	96.8 (79.3)
Mean I/σ(I)	14.3 (6.4)
Wilson B factor (Å2)	9.33
R merge	0.14 (0.30)
R meas	0.15 (0.32)
CC1/2	0.99 (0.95)
Reflections used in refinement	23191 (1886)
Reflections used for Rfree	1954 (160)
R work	0.17 (0.26)
R free	0.22 (0.32)
No. of atoms
Non-H	2726
Protein	2294
Fructose	12
ADP	27
Nitrate	4
Water	389
No. of protein residues	296
R.m.s. deviations
Bond lengths (Å)	0.003
Bond angles (°)	0.60
Ramachandran plot
Favored (%)	97.9
Allowed (%)	2.1
Outliers (%)	0
Rotamer outliers (%)	0.4
Clashscore	3.7
B factors (Å2)
Average	12.6
Protein	10.9
Fructose	5.8
ADP	24.6
Nitrate	25.1
Water	22.04

2.4. X-ray crystal structure determination and model refinement

Phases were determined by automated molecular replacement with Phaser in Phenix (Liebschner et al., 2019 ▸) using one protomer from the structure of wild-type human KHK-C (PDB entry 3q92; Maryanoff et al., 2011 ▸) as the search model with water and ligands removed. The model and phases were improved by rigid-body refinement and debiased with AutoBuild in Phenix. The electron-density maps (σ_A-weighted with coefficients 2F_o − F_c) calculated from the model after rigid-body refinement displayed strong electron density for all areas of the protein as well as for the ligands ADP, fructose and nitrate. Iterative rounds of model building using Coot (Emsley et al., 2010 ▸) were performed alternately with rounds of positional, group temperature-factor and simulated-annealing refinement in Phenix. When R_work reached 0.29, the models for ADP, fructose and nitrate were placed in density using LigandFit in Phenix. A summary of the refinement statistics can be found in Table 1 ▸.

2.5. Analysis of ligand-induced conformational changes

Ligand-induced conformational changes in mKHK-C were assessed using two structures of KHK: chain A of the 2.3 Å resolution hKHK-C structure (PDB entry 3nbv; Gibbs et al., 2010 ▸) and chain A of the mKHK-C–ADP–fructose structure reported here. DynDom was used to analyze possible domain motions by comparing the two structures (Hayward & Lee, 2002 ▸).

2.6. Structure deposition

Atomic coordinates and structure factors have been deposited in the Protein Data Bank (PDB entry 6p2d).

3. Results

3.1. Assessment of the quality, purity and activity of mKHK-C

The homogeneity and purity of the mKHK-C preparation were assessed using SDS–PAGE and DLS (Supplementary Fig. S1). The protein was shown to be >95% pure, with 99.9% of the mass in a single population with 39.5% polydispersity. Although a polydispersity of <20% is generally desirable for crystallization, we proceeded with crystallization trials. Enzyme-activity assays were performed to measure the values of the steady-state kinetic parameters k_cat and K_m, using a nonlinear regression analysis fitting of the saturation data to the equation for a hyperbola with GraphPad Prism version 6.01 (Table 2 ▸).

Table 2. Steady-state kinetic parameters for mouse and human KHK-C.

Data are means ± standard errors; n = 9.

	mKHK-C	hKHK-C†
Km, ATP (µM)	270 ± 20	150 ± 10
Km, fructose (µM)	1700 ± 180	800 ± 180
kcat (s−1)	7.1 ± 0.2	7.6 ± 0.7
kcat/Km, ATP (s−1 µM−1)	0.026 ± 0.002	0.05 ± 0.006
kcat/Km, fructose (s−1 µM−1)	0.0042 ± 0.0005	0.010 ± 0.002

^†Kinetics parameters from Asipu et al. (2003 ▸).

3.2. Structure of mKHK-C

The structure of mouse KHK-C bound to ADP and fructose was determined with phases calculated by molecular replacement using the coordinates of human KHK-C (PDB entry 3q92) with waters and ligands removed. Crystallographic data-collection and refinement statistics are summarized in Table 1 ▸. The structure had a single subunit in the asymmetric unit; the resulting electron density was clearly defined, with residues 3–298 (out of 298 residues) modeled. The structure was refined to 1.79 Å resolution, with a final R_work of 0.17 and R_free of 0.22. In the final rounds of refinement, ligands and water molecules were included in the model. The Ramachandran plot output by Phenix showed that 98% of the residues were in favored regions, with no outliers. Although nitrate ion was added to the crystallization conditions, along with ADP and fructose, to mimic the transition state, as previously observed for creatine kinase (Lahiri et al., 2002 ▸), a nitrate ion was not observed in the phosphoryl-binding subsite but instead was liganded to Arg141 just outside the active site (3.6 Å from the fructose C6 hydroxyl). Density for the ligands ADP and the β-d-fructofuranose form of the substrate were well resolved in the active site (Fig. 3 ▸), with ligand B factors similar to those of nearby side chains. This configuration of the fructose was consistent with biochemical data demonstrating specificity for this anomer (Raushel & Cleland, 1973 ▸). Although the divalent metal cation Mg²⁺ often associates with nucleotides, its affinity is higher for ATP than for ADP (Gout et al., 2014 ▸), and Mg²⁺ ion was not observed in the structure. The overall secondary structure of the protomer is as described previously (Trinh et al., 2009 ▸; Ebenhoch et al., 2023 ▸), with two distinct domains: a central α/β-fold and a four-stranded β-sheet. The α/β-fold domain is approximately defined as residues 5–12, 42–89 and 122–297. The β-sheet domain that forms a β-clasp structure in the biological dimer is composed of residues 13–40 and 96–111 (Fig. 4 ▸a). An overlay of the main-chain C^α atoms of the functional mKHK-C dimer, which was visualized by a crystallographic symmetry operator in PyMOL, with those of the hKHK-C dimer (PDB entry 3nvb) liganded by fructose and the ATP analog AMP-PNP gave an overall root-mean square deviation (r.m.s.d) of 0.45 Å, showing that they are in a similar conformation. A B-factor putty view in PyMOL shows that the residues in the β-clasp domain have higher B factors than the residues in the α/β-fold domain (Fig. 4 ▸b).

Figure 3.

Composite OMIT electron-density maps (blue cages) contoured at 1.0 r.m.s. for bound ADP (a) and fructose (b) generated with Phenix and displayed in PyMOL.

Figure 4.

(a) Ribbon structure of the mKHK-C dimer colored by domain. The α/β-fold domains are colored blue in each subunit and the lid/β-sheet domains are colored beige. The two β-sheet domains come together to form a β-clasp structure as commonly found in PfkB family members. Fructose bound in the active site is shown as yellow sticks and ADP is shown as green sticks. (b) B-factor putty view of the mKHK-C dimer with low B factors colored with cool tones and high B factors colored with warm tones. The lid/β-sheet domain colored green, yellow and orange has higher B factors than the central α/β-fold domain colored purple and blue.

3.3. Ligand-binding sites

Numerous published structures of both human KHK-A and KHK-C bound to substrates, products or inhibitors have provided insight into substrate recognition and catalysis by KHK. As in the human structures, the substrate-binding sites of mKHK-C are located between the α/β-fold domain and the β-sheet domain, with the β-sheet domain acting as a lid for the opposing subunit that allows the entrance and exit of substrates and products (Fig. 4 ▸).

Tunnel and solvent-accessibility analysis of mKHK-C using CAVER 3.0 (Chovancova et al., 2012 ▸) shows that when the ligands are removed a tunnel is accessible to solvent (Fig. 5 ▸). However, when the ligands are present CAVER 3.0 is unable to locate any tunnels as the ligands occupy the binding sites.

Figure 5.

Tunnel and solvent-accessibility analysis of binding sites in mKHK-C using CAVER 3.0 (Chovancova et al., 2012 ▸). Using Gly286 (orange) as the starting point, CAVER 3.0 was run in both the presence and absence of ADP and fructose bound to mKHK-C. When the ligands are removed, a tunnel (magenta) is vacant and accessible to solvent. However, when the ligands are present CAVER 3.0 is unable to locate any tunnels as the ligands are occupying the binding sites. The figure shows an overlay of ligands and CAVER 3.0 tunnels in a composite view.

The product of catalysis, ADP, interacts directly with mKHK-C via two residues: Gly257 and Arg108. The backbone amide NH of Gly257 forms a hydrogen bond to a terminal O atom of the β-phosphoryl group of ADP, which is the same interaction as observed with AMP-PNP (Trinh et al., 2009 ▸). The guanidino group of Arg108 forms two additional hydrogen bonds to another O atom of the β-phosphoryl group (Fig. 6 ▸a). The ADP is also positioned in the binding site by two additional residues that make hydrogen bonds bridged by waters. The β-phosphate of ADP is also stabilized by the helix dipole (residues 255–269), as is typical for many nucleotide-binding proteins. Hydrophobic interactions with the adenine ring are made by Pro246, Ala285 and Cys289, and Ala224, Ala226, Gly229, Ala230, Ala256 and Phe260 surround the ribosyl moiety; again these interactions have also been observed with AMP-PNP (Trinh et al., 2009 ▸). The proline at position 246, which is 3.8 Å from the C3 atom of adenine, may be responsible for the specificity of KHK towards ATP versus GTP (Adelman et al., 1967 ▸).

Figure 6.

(a) Detailed view of the ADP-binding site and direct interactions with chain A of mKHK-C. The ligand is anchored by hydrogen bonds from two residues: Gly257 and Arg108. (b) Detailed view of the fructose-binding site and direct interactions with chain A of mKHK-C. The ligand is anchored by hydrogen bonds from four residues: Ser97, Gly41, Asn45 and Asp15.

In the fructose-binding site of mKHK-C, the fructose C2 hydroxyl makes a hydrogen bond to the Asn45 amide. The C3 hydroxyl makes hydrogen bonds to Gly41 NH and Asp15. The C4 hydroxyl makes hydrogen bonds to Ser97 and Asp15. The negatively charged side chain of Asp15 forms hydrogen bonds to both the C3 and C4 hydroxyls of fructose (Fig. 6 ▸b). These fructose-binding interactions have previously been described in hKHK structures (Trinh et al., 2009 ▸) and are consistent with biochemical data with various substrate analogs that report specificity for the β-d-fructofuranose ring and the configuration at C3, but less specificity for the C4 and C5 configurations (Raushel & Cleland, 1973 ▸). The weaker (3.3–3.7 Å) hydrophobic interactions with fructose made by Gly40, Leu254, Val13, Asn42 and Leu112 have not previously been described.

3.4. Analysis of the hinge movement of the β-sheet domain

DynDom (Hayward & Lee, 2002 ▸) was used to detect and quantify differences in conformation between the mKHK-C structure reported here and several hKHK structures in the PDB. Firstly, our liganded mKHK-C structure was compared with an unliganded hKHK-C structure (PDB entry 3b3l; Trinh et al., 2009 ▸) to determine whether ligand binding results in movement of the β-clasp domain. Notably, the unliganded mKHK-C structure (PDB entry 8omg; Ebenhoch et al., 2023 ▸) is in the same conformation as unliganded hKHK-C, with an r.m.s.d. of 0.49 Å. The analysis was run on both chain A and chain B of PDB entry 3b3l. The β-clasp of PDB entry 3b3l (chain B) is rotated 11.6° closer to the α/β-fold domain than that in mKHK-C, whereas the β-clasp position of PDB entry 3b3l (chain A) was rotated more than 22° towards the ATP-binding site of the opposing subunit compared with that in mKHK-C (Fig. 7 ▸a). We assign β-clasp conformations similar to PDB entry 3b3l (chain B) as the open conformer as the clasp is rotated away from the ATP-binding site of the opposing subunit, and those similar to PDB entry 3b3l (chain A) as the closed conformer because the clasp is rotated towards the ATP-binding site of the opposing subunit, serving as a lid for the active site. Residues 113–117 which link the α/β-fold domain and the β-clasp domain were identified by DynDom as bending residues responsible for the different conformations of the β-clasp; the motions were classified as a mix of screw and hinge movements.

Figure 7.

(a) Overlay of the mKHK-C dimer subunit (blue) with two subunits of hKHK-C (PDB entry 3b3l): chain B (teal) and chain A (purple). Relative to mKHK-C (blue), the β-sheet domain of chain B is rotated 11.6° away from the active site of the opposing subunit (open conformation) and the β-sheet domain of chain A is rotated 22.2° towards the active site of the opposing subunit (closed conformation). Our mKHK-C structure is between the two conformations. (b) The relationship between the number of ligands bound to KHK and the β-clasp conformation. The histogram shows the percentage of KHK subunits in the closed conformation grouped by the number of subsites occupied in the opposing subunit. Subunits with no ligands are rarely in the closed position (yellow). Subunits with only one ligand bound are about equally distributed between the open and closed conformations (purple). When both the ATP- and fructose-binding sites are occupied in the opposing subunit the closed conformation is invariably observed (green). The binding of two ligands in the adjacent subunit of the mKHK-C dimer allows the interactions that hold the β-clasp in the closed conformation. The number of subunits measured is given in parentheses. A list of the structures used in this analysis is given in Supplementary Table S1.

This analysis was repeated, comparing our liganded mKHK-C structures with hKHK structures in the PDB (44 dimer subunits from 21 different deposited structures) that were bound to no ligands, a single ligand (in either the fructose or ATP site) or two ligands (in both the fructose and ATP sites). An open or closed conformation was assigned to each subunit based on whether the β-clasp domain rotated towards the αβ domain, as in chain B of PDB entry 3b3l, or rotated towards the ATP-binding site of the adjacent subunit, as in chain A of PDB entry 3b3l, both relative to the conformation of the β-clasp domain of mKHK-C, which defined the limit for a closed conformation (Supplementary Table S1). Human KHK structures with no ligands were mostly in the open conformation (ten open and four closed). Those with only one ligand bound in the opposing subunit were almost equally distributed between the open and closed conformations (12 open and 14 closed), whereas those with two ligands bound in the opposing subunit were solely determined in the closed conformation (five closed and zero open), as found for mKHK-C (Fig. 7 ▸b).

4. Discussion

The use of the pPB1 expression vector (Beernink & Tolan, 1992 ▸) enabled the purification of large amounts of mKHK-C with sufficient purity and homogeneity for biochemical assays and crystallographic studies. The steady-state kinetic constants for mKHK-C were shown to be similar to the previously reported values for KHK (Table 2 ▸). The K_m value of 270 µM of mKHK-C for ATP is similar to the value of 150 µM reported for hKHK-C. The K_m value of 1700 µM of mKHK-C for fructose is also similar to the previously reported K_m value of 800 µM of hKHK-C for fructose. The similar kinetic parameters indicate that the interpretation of any preclinical data using mice as a model for the human target will be clearer.

The mKHK-C structure was determined to 1.79 Å resolution, allowing the placement of ligands and the mapping of protein–ligand interactions. The His tag was not removed after purification; both the tag and the first few N-terminal residues were not visible and are most likely to be disordered in the crystalline form. The tertiary structure of mKHK-C is composed of a central α/β-fold and a β-sheet domain that forms a β-clasp dimer interface as previously reported for hKHK-C.

Both subunits in the biological dimer of our mKHK-C structure are bound to the fructose substrate as well as to ADP, the product of catalysis. As in other members of the PfkB family, magnesium is required for activity; therefore, a high concentration of MgCl₂ was used in the crystallization buffer. However, there was no electron density corresponding to Mg²⁺ ions in or around the active site in the crystal structure. This was expected, as Mg²⁺ is often not found in structures with nucleotide diphosphates bound. In addition, we introduced nitrate into the crystals (at 1.3 M) to mimic the γ-phosphoryl group of ATP, but we were not able to resolve this ion in the active site.

The tertiary structure of mKHK-C has high structural similarity to that of hKHK-C as well as other representative members of the PfkB family, including ribokinase, adenosine kinase and 2-keto-3-deoxygluconate kinase. An overlay of each of these structures with mKHK-C resulted in r.m.s.d. values of 0.5–2.6 Å, with hKHK-C being the most similar, followed by ribokinase, adenosine kinase and 2-keto-3-deoxygluconate kinase in order of decreasing similarity (data not shown). Despite their high level of structural similarity, mKHK-C and 2-keto-3-deoxygluconate kinase share <20% sequence identity. Structural comparison of the ATP-binding sites of the family members shows significant differences (data not shown). The ATP-binding site of KHK is composed of hydrophobic residues, creating a flatter and narrower binding pocket compared with those in the other family members. For members of the PfkB family with a dimeric oligomerization state, such as ribokinase, the β-sheet domain from the adjacent subunit extends into the active site, limiting accessibility, and acts as a lid controlling the entrance and exit of substrates and products (Sigrell et al., 1999 ▸; Fig. 4 ▸). Functional completion of an active site by another protomer is not uncommon and occurs in other enzyme families. In addition, conformational changes similar to those in mKHK-C have also been described in other PfkB family members. In adenosine kinase a semi-open conformation of the active site has been described that results from a rotation of the hinge domain acting as a lid for the active site in the opposing subunit (Zhang et al., 2007 ▸). In ribokinase, a conformational change initiated by the binding of a monovalent ion shifts a loop in the enzyme towards the ATP-binding site, causing an increase in the substrate-binding affinity (Andersson & Mowbray, 2002 ▸). The structure of mKHK liganded to an inhibitor that is selective for KHK-A over KHK-C (PDB entry 8omd) is in the same closed conformation as that reported here for mKHK-C (PDB entry 6p2d), with an r.m.s.d. of 0.44 Å. The findings from this analysis, in agreement with those reported here, show that an inhibitor occupying the ATP-binding site is able to select for a conformation that is also used in catalysis (Ebenhoch et al., 2023 ▸; Fig. 7 ▸b).

In the mKHK-C structure, two defining features of the PfkB family are present: the DTXGAGD motif and the anion hole used in catalysis. The GAGD motif in mKHK-C is positioned at the interface of the γ-phosphoryl group of ATP and fructose (Fig. 1 ▸). As in other PfkB family members, Asp258 in the GAGD motif is in a position to serve as a catalytic base to deprotonate the C1 hydroxyl of fructose, and the backbone amides of Ala256 and Gly257 would be proximal to the γ-phosphate and C1 hydroxyl of fructose in the active site, forming the oxyanion hole used to stabilize the transition state during phosphoryl transfer (Mathews et al., 1998 ▸). Another possibility is proton abstraction through a water-mediated tetrad of waters 478 and 557–His137–Ser169–Trp135. All three residues are conserved in vertebrate KHK enzymes, but not across the family as for Asp258 of the DTXGAGD motif.

The electron density generated by our data is well defined around the ligands for mKHK-C and the B factors indicate that the ligands are well ordered. In mKHK, Arg108 plays a role in electrostatic stabilization of the developing negative charge at the β-phosphate in the transition state as the bond to the γ-phosphate breaks, which has been described for many phosphoryl-transfer enzymes, as well as in replacing solvent interactions in binding to the water-protected active site, similar to the case of creatine kinase (Lahiri et al., 2002 ▸). In the absence of the γ-phosphoryl group of ATP, the guanidino group interacts with the β-phosphoryl group of ADP. This role of Arg108 was not described in the original structures of KHK (Trinh et al., 2009 ▸). Gly257 has been shown to form a hydrogen bond to the γ-phosphoryl group in the ATP-binding site of hKHK-C (Trinh et al., 2009 ▸) and to the β-phosphate of ADP as reported here, which is consistent with the role described above. Gly257, which is at the N-terminus of the helix dipole, is positioned to additionally help to stabilize the developing negative charge on the β-phosphate in the transition state.

Residues located on the β-clasp in the opposing subunit are involved in stabilizing ligands in the active site when mKHK-C is in the closed conformation. In the fructose-binding site, the β-d-furanose form of the sugar substrate is selected by several appropriately placed residues that hydrogen-bond to the hydroxyl groups of residues in chain A; additionally, Asp29 from chain B interacts with O5 of fructose via two bridging water molecules. These interactions of residues in chain B are allowed by the hinge rotation promoted by the binding of ligands in chain A. Based on our data and the analysis of liganded structures in the PDB, we propose that the β-clasp is only placed into a catalytically competent conformation when both the ATP-binding and fructose-binding sites are occupied. Conversely, when there are fewer than two ligands bound, the open and closed conformations of the β-clasp are fairly iso­energetic. Of course, it is difficult to judge whether induced fit is part of the mechanism from X-ray structural data alone, and additional data from FRET or hydrogen–deuterium exchange studies would be informative.

5. Conclusion

Based on the high sequence identity and structural and kinetic similarities between mKHK-C and hKHK-C, any preclinical studies using a mouse animal model for the inhibition of KHK will be more easily interpreted. Here, we report a structure of mouse KHK-C that suggests a hinge rotation of the β-sheet domain towards the active site of the opposing subunit. This rotation change is most likely initiated by the binding of ligands to the other chain of the biological dimer. Upon rotation, the β-sheet domain closes the active site of the opposing subunit and acts as a lid controlling the entrance and exit of ligands. Based on the Michaelis-like complex of mouse KHK-C reported here, we hypothesize that the closed conformation is the catalytically active form of KHK. Although a similar conformational change has been linked to catalysis in other PfkB family members, further exploration of the hinge rotation and the placement of catalytic residues is necessary to determine their roles in catalysis.

Supplementary Material

PDB reference: mouse ketohexokinase C in complex with fructose and ADP, 6p2d