Modeling Alcohol Dehydrogenase Catalysis in Deep Eutectic Solvent/Water Mixtures

Abstract The use of oxidoreductases (EC1) in non‐conventional reaction media has been increasingly explored. In particular, deep eutectic solvents (DESs) have emerged as a novel class of solvents. Herein, an in‐depth study of bioreduction with an alcohol dehydrogenase (ADH) in the DES glyceline is presented. The activity and stability of ADH in mixtures of glyceline/water with varying water contents were measured. Furthermore, the thermodynamic water activity and viscosity of mixtures of glyceline/water have been determined. For a better understanding of the observations, molecular dynamics simulations were performed to quantify the molecular flexibility, hydration layer, and intraprotein hydrogen bonds of ADH. The behavior of the enzyme in DESs follows the classic dependence of water activity (a W) in non‐conventional media. At low a W values (<0.2), ADH does not show any activity; at higher a W values, the activity was still lower than that in pure water due to the high viscosities of the DES. These findings could be further explained by increased enzyme flexibility with increasing water content.


S3
finally concentrated to 5.0 mL. The obtained purified HLADH was firstly frozen at -80°C for overnight and subsequently freeze-dried at 0.03 mbar and -60°C for 40 h. 5.0 mL of purified HLADH with the protein concentration of 23 mg/mL was obtained from 21.4 g wet cells. After the lyophilization, 112 g lyophilized HLADH was obtained, thus the productivity is 5.23 mg protein/g cells ( Figure S1).

Synthesis of glyceline
Choline chloride (ChCl) and glycerol (Gly) were directly weighed in a flask in a molar ratio of 1:2. Then, the mixture was heated and stirred (IKA Magnetic stirrer, 250 rpm) at 80 o C until a colorless liquid was formed (normally less than one hour).

Determining the water activity (aw) of glyceline-water binary mixtures
Binary mixtures (5 mL) of glyceline with various water contents of 0-20% (v/v) and 100% (v/v) were freshly prepared and incubated at 60 o C for 1 hour in 25 mL sealed glass bottles. The thermodynamic water activity (aw) of these mixtures were then determined at room temperature (24-25 o C) using HMT337 Humidity and Temperature Transmitter (vaisala, Vantaa, Finland). Here, the aW was measured based on the ratio of the water vapor pressure in the glyceline-water mixtures (p) to the vapor pressure of pure water (p0) (aw = p/p0).

Determining the viscosity of glyceline-water binary mixtures
The dynamic viscosity (η) of glyceline-water mixtures was measured on 2 mL sample with a Brookfield Digital Rheometer (Model DV-III Ultra, Brookfield Engineering Laboratories Inc., MA, USA) equipped with a spindle CPE41, at different shear rates between 0.2 s -1 and 500 s -1 at room temperature (24-25 o C). S5 Figure S4. The viscosity (η) of glyceline-water mixtures as a function of water content. The data points are connected by a solid line and a dash line to guide the eye.

Determining the Ti of HLADH in glyceline-water binary mixtures
Tycho NT.6 instrument (NanoTemper Technologies) was used to determine the inflection temperature (Ti) of HLADH with the concentration of 1.0 mg/mL in glyceline with various water contents. The Tycho NT.6 capillaries (high precision glass capillaries specifically developed for use with Tycho NT.6) containing 10 μL sample were placed on the sample holder. A temperature gradient of 30°C/min from 35°C to 95°C was applied and the fluorescence of intrinsic tryptophan and tyrosine of HLADH was recorded at 330 nm and 350 nm. Tycho utilizes a fast and defined thermal ramp to unfold HLADH and identifies the inflection temperature (Ti) that represent unfolding transition(s) or discrete changes in the structural integrity of HLADH and can be used for comparing thermal stabilities of HLADH in glyceline with various water content.

Determining the t1/2 of HLADH in glyceline-water binary mixtures
The half-life times of HLADH in various glyceline-water mixtures were determined by incubating 500 μL of 1.0 mg/mL purified HLADH at 60°C. The samples were taken at aliquot time points ranging from 0 to 20 h and the residual activities were detected at 25°C. The half-life times (t1/2) were calculated based on Equation 1.

Gas chromatography analysis
Aliquots samples (50 µL) from each reaction system were taken at definite time intervals and mixed with 250 µL of ethyl acetate (2 mM methyl benzoate as the internal standard). After centrifuging (13,000 rpm; 1 min) and separating the two phases, the EtOAc layer was dried with anhydrous MgSO4. All reaction components were then analyzed by gas chromatography (GC) and the methods were developed with ß-DEX 120 column (30 m x 0.25 mm x 0.25 µm, Supelco ® Analytical, USA; catalogue reference: 24304). Peaks were identified by standards. The details could be seen below in Table S2. The concentrations of target components were calculated based on the corresponding GC calibration in each reaction system.

Force field for MD simulations of biomolecules
For MD simulations a force field model is needed, which describes the interatomic and intra-atomic interactions between the simulated molecules. Prior to the application of MD simulations to large protein systems the validation of the used force field is of great importance. In particular, the application of the novel solvent group of DESs requires a careful validation of the force field parameters.
The recently published OPLS-DES force field by Doherty and Acevedo [1] was tested to reproduce the density of mixtures of choline chloride with varying water content.
Therefore, MD simulations of pure glyceline as well as in mixtures with a water content S10 up to 47% (mol/mol) using the software package GROMACS version 2018 [2] were compared to an experimental density correlation. [3] Compared to the publication of Doherty and Acevedo, [1] the simulation parameters have been tuned to be better suitable for simulations containing a protein considering the parameters usually used for protein simulations and considering that the system is much larger. As all bonds between hydrogen and other atoms are fixed using the LINCS algorithm, a time step of 1 fs is not necessary. Instead a larger time step of 2 fs would be more efficient in particular for large system sizes. In addition to the time step the cut-off radius for the van-der-Waals and electrostatic interactions has been changed compared to Doherty and Acevedo. [1] The force field parameters of OPLS-DES have been fitted using a cut- OPLS-DES force field [1] (black squares and dash line) and experimental correlation [3] (black solid line). S11

MD simulations procedure for HLADH in glyceline-water mixtures
Owing to large viscosities, MD simulations of DESs have proven to be challenging due to their slow dynamic behavior. [4] The challenges begin with the equilibration of the systems, which is largely influenced by the slow dynamics of most DES. In order to overcome this issue, Perkins et al. [4a, 4b] suggested to use compression and decompression scheme to ensure an efficient equilibration of highly viscous systems, which was priory applied for polymeric systems. [5] This scheme also later adopted by Mainberger et al., [4c] was tested in this work for the MD simulations of HLADH in glyceline-water mixtures. However, the resulting simulations show large energy drifts, which occur even at long simulations times larger than 80 ns. Therefore, a different equilibration scheme based on temperature annealing is proposed to equilibrate the simulation boxes. A similar scheme has been applied by Liu et al. for ionic liquids. [6]  All MD simulations of this work were performed with the software package GROMACS version 2018.6. [2] The OPLS-DES force field [1] and the TIP3P force field [7] have been used for the DES glyceline and water, respectively. The protein interactions have been modeled with the OPLS-AA/M force field. [8] The temperature curve applied for the equilibrations is displayed in Figure S10. Whereby the used equilibration scheme for the protein systems consists of five consecutive steps. Starting with the S12 crystallographic structure of HLADH (PDB entry 1HEU) embedded in a solution using packmol [9] , an energy minimization with the steepest decent algorithm was performed for 5000 steps. Afterwards, initial velocities for all molecules according to a Maxwelldistribution at 278 K are assigned and a 1 ns simulation at 298 K using a velocity rescale thermostat and by constraining the positions of all enzyme atoms has been performed. In order to ensure a proper mixing of the liquid around the constrained enzyme, the temperature is subsequently increased to 500 K during a period of 1 ns and then kept constant for 20 ns. A temperature of 500 K has been chosen as this is the maximal temperature the OPLS-DES force field was validated for. After cooling down the simulated system to 298 K during a period of 1 ns the enzyme structure is

Conversion of volume fractions and mole fractions for glyceline water mixtures
In terms of molecular dynamics simulations, the experimentally used volume fractions need to be transferred to mole fractions in order to relate it to the MD simulations. A corresponding conversion plot can be found below (Figure S11), the concentrations investigated in this work are listed in Table S3. Whereby, the mole fractions in the MD simulations have been calculated by the following equation:

Crystallographic structure of HLADH
As starting point for the MD simulation, the experimentally measured crystallographic structure of HLADH (PDB entry 1HEU) was used. A schematic illustration of the HLADH structure is displayed in Figure 3. In general, the enzyme can be divided into S14 two parts: the substrate binding domain (residues 1-175, 319-549, 693-748) and the coenzyme binding domain (residues 176-318, 550-692). [10] The coenzyme binding domain is located in the middle of the protein structure, whereas the substrate binding domain, which includes the active centers are located at the left and right site of the protein structure illustrated in Figure 3.
HLADH has a dimeric structure of two identical subunits connected via hydrogen bonds. [10][11] Hence, the enzyme owes two active centers in the middle of each monomer (Figure 3, highlighted in blue). Though, the three amino acids of each active center (Cys46, His67, Cys174) belong to one chain (for instance chain A), the substrate binding pockets also consists of amino acids of the other chain (in this example chain B). The amino acids along the substrate pocket of chain A, according to Eklund et al. [10][11] are summarized in Table S4. This means that HLADH is only functional in form of the dimeric structure, and not in terms of the single monomers. In case of pure water the loops consisting of the residues 120-128 and 616-622 (highlighted in purple) are also excluded from the calculations. Table S4. Amino acids of the substrate binding pocket of chain A according to Eklund et al.. [10][11] Chain A Chain B

Root mean square deviations of the -atoms of HLADH in different glyceline-water mixtures
The RMSD of the C α -atoms of HLADH with respect to the crystallographic structure  In addition, the structural changes for both active centers including the binding site pockets are illustrated in Figure S14. Besides the simulations in pure water and 10% (v/v) water the RMSD of the pocket and active center are similar for both chains.
However, at 10% (v/v) the binding site of chain A is closer to the crystalline state compared to chain B. A similar trend could be found for the pure aqueous environment. This is counterintuitive, as the enzyme structure is consisting of two identical chains and is perfectly symmetric. An analysis of the minimal distance of the binding site and pocket residues to the solvent molecules revealed a closer contact to choline and chloride ions for chain A compared to chain B ( Figure S15). The minimal distance to chloride was much lower than in all other cases, particularly compared to the pure glyceline case (data not shown). The strong interactions between chloride and the enzyme may lead to different structures, which could explain the differences observed in the simulations. This is particularly interesting as it coincides with the beginning of a catalytic activity. Nevertheless, this may not explain the huge difference for the pure water case, as both catalytic centers are equally hydrated in those simulations.

Intra-protein hydrogen bonds
The intra-protein hydrogen bonds have been calculated with the gmx hbond tool implemented in the GROMACS package for HLADH in all investigated glyceline-water mixtures. Whereby, a cut-off distance of 0.25 nm and a cut-off angle of 30° have been used. The resulting values are shown in Figure S17.