The catalytic inactivation of the N-half of human hexokinase 2 and structural and biochemical characterization of its mitochondrial conformation

The high proliferation rate of tumor cells demands high energy and metabolites that are sustained by a high glycolytic flux known as the ‘Warburg effect’. The activation and further metabolism of glucose is initiated by hexokinase, a focal point of metabolic regulation. The human hexokinase 2 (HK2) is overexpressed in all aggressive tumors and predominantly found on the outer mitochondrial membrane, where interactions through its N-terminus initiates and maintains tumorigenesis. Here, we report the structure of HK2 in complex with glucose and glucose-6-phosphate (G6P). Structural and biochemical characterization of the mitochondrial conformation reveals higher conformational stability and slow protein unfolding rate (ku) compared with the cytosolic conformation. Despite the active site similarity of all human hexokinases, the N-domain of HK2 is catalytically active but not in hexokinase 1 and 3. Helix-α13 that protrudes out of the N-domain to link it to the C-domain of HK2 is found to be important in maintaining the catalytic activity of the N-half. In addition, the N-domain of HK2 regulates the stability of the whole enzyme in contrast with the C-domain. Glucose binding enhanced the stability of the wild-type (WT) enzyme and the single mutant D657A of the C-domain, but it did not increase the stability of the D209A mutant of the N-domain. The interaction of HK2 with the mitochondria through its N-half is proposed to facilitate higher stability on the mitochondria. The identification of structural and biochemical differences between HK2 and other human hexokinase isozymes could potentially be used in the development of new anticancer therapies.


Minimum Energy Pathway of conformational transition from open to closed state.
Atomically detailed simulations were used to study the mechanism of the open to closed conformational transition for the FL-HK2 enzyme. By fixing the two end states at open and closed states, the minimum energy path determined by connecting the two states using steepest descent path (SDP) methodology [1]. The enzyme undergoes a collective motion during the conformational transition. The opening and closing of the four active sites are triggered by the movement of the small subdomains of the homodimeric HK2 that point to the outside of the enzyme.
To study the role of helix-α 13 Figure 1D). The change in the groove width was estimated from the distance between T88 and T336.
A salt bridge was detected between R468 on the linker helix-α 13 and D202 on the β-sheet of the small subdomain of the N-domain ( Figure 1C). To decipher the effect of salt bridge on helix-α 13 angle and overall to conformation of HK2 enzyme, MD simulations were conducted on the mutant R468A to determined its effect on the enzyme conformation and the binding stabilities of the substrates. Abolishing of the salt bridge between R468A and D202 leads to a wide distribution of θ angle in the R468A mutant in comparison to the WT enzyme ( Figure 1C and 1D). Lower values of the θ angle lead to widening of the active site and hence weakening of the H-bonding network with the substrates, resulting in a rapid dissociation of the glucose from the active site of R468A mutant ( Figure 1E). The salt bridge is proposed to remotely modulate glucose binding by widening the grove width of the active site.

Supplementary Methods
Modeling the open state of HK2. The crystal structure of the HK2 enzyme was determined in the closed state. The apo-enzyme could not be crystalized from all the different variants constructed here and could not acquire the open conformation of HK2. Computational homology modeling was used to generate the open state of HK2. The structure of yeast hexokinase yeast [2] was used as a template in MODELLER [3] then targeted molecular dynamics (TMD) simulation protocol [4] implemented in Gromacs 4.05 [5] was used with the structure-based model for biomolecules (SMOG) potential [6]. SMOG potential allowed rapid sampling of the conformational space while keeping the secondary structure stoichiometry intact during the closed to open transition. The approach has been utilized heavily to study biomolecular processes including the mechanism of conformational transitions of large biomolecular complexes [7]. A TMD simulation of 250000 steps in reduced temperature was used to obtain the open conformation of HK2. The temperature was kept constant using velocity verlet scheme [8]. Nonbonded interactions were treated with 15Å cutoff. Stochastic dynamics integrator with a time step of 0.001 and a friction coefficient of 0.1 was used to integrate the equations of motion. The trajectory created with TMD later used to study the minimum energy pathway of the transition.
Modeling the transition pathway. The 96 configurations equally spaced from initial TMD trajectory and minimized each using a classical Molecular Mechanics force field. The pathway from closed to open was refined using SDP methodology [1] implemented in MOIL [9]. In this approach given the two ends configurations R C and R O , a discrete set of coordinates between the two end states R i where index i = 2, 3.., N -1 with the optimized the function Here R(l) is the coordinate vector as a function of arc-length l . ÑU(R(l)) is the gradient of the potential energy. GBSA implicit solvent model [10] and OPLSAA force field was used to account for the interactions of glucose with the protein [11]. ATP parameters were adopted from previous work [12] with minimization protocol similar to work done previously [12]. Minimum energy path of closed to open state is shown in Movie 1. The trajectory was smoothed by averaging two consecutive frames for visualization purposes.

Molecular Dynamics Simulation Protocol.
Explicit water All Atom Molecular Dynamics Simulations were carried out in Gromacs 4.05 suit of programs [5]. OPLSAA [13] forcefield is used for protein and glucose. ATP parameters were adopted from [12]. Water is modeled using SPCE [14]. Smith and Dang parameters [15] were used for K + and Cl -, and Mg 2+ coordinates were acquired from [16]. The protein was solvated with a simulation box of 7.7 x 11.8 x 8.2 nm 3 . To neutralize the system and to mimic experimental conditions, 21 Cland 22 K + ions were added by randomly replacing water molecules. To equilibrate water and ions, the positions of the heavy atoms on the enzyme were restrained then run molecular dynamics simulation for 10 ns using constant pressure (NPT) simulation. The pressure was kept at 1 bar using Parrinello-Rahman scheme [17]. NPT simulation was followed by constant volume (NVT) simulation for another 10 ns with the position restrains. Unrestrained NVT simulations were later carried out for sampling the conformational states. In all simulations, Leap Frog integration scheme was used with a time step of 2 fs. The temperature was set to 300K using the Velocity-Scaling method implemented in Gromacs [8]. Van der Walls interactions were calculated with 7-10 Å switching scheme. Electrostatic interactions were treated by Particle Mesh Ewald summation method [8] with cubic interpolation order of 4 and grid spacing of 1.6 Å with a real-space cutoff of 12 Å. All bonds in the enzyme and water were constrained by LINCS algorithm.

SUPPLEMENTARY FIGURES AND TABLES:
Supplementary Figure HK2. (A−B) The N-and C-halves of HK2 with large (white) and small (gold) subdomains. The MBP of the N-half is dark blue. Helix α 13 (red) protrudes out of the active site at the end of the N-and C-halves. Two 5 stranded βsheets (blue) encloses the active site in addition to helices  5 (green) and  13 . Inset: Helix  5 that carries the catalytic residue D209 or D657 of the N-and C-halves, respectively, is perpendicular to helix  13 . (C) The glucose (white) and ATP (yellow) binding pockets in HK2 with the later modeled based on the crystal structure of HK4−ATP complex. (D) Overlay of the monomers of HK1 (aquamarine) and HK2 (yellow) in complex with glucose and G6P. Alignment was only possible on N-half but not the FL enzyme. Even though the structural folds of HK1 and HK2 domains are identical, a 16° bent on the linker helix of HK1 prevented its alignment to HK2. This figure was prepared using PyMol (Schrodinger LLC).  * The molecular weight (MW) was calculated using ExPasy Proteomics tools [19].
Supplementary Table 2: Crystallographic data and refinement statistics, related to Figure 1, and Supplementary Figure 2. Numbers in parentheses represent the highest resolution bin.

Data collection
Space Group: P2 1