Hemeproteins

Albumin–heme complex was prepared by incubating 120 μM stock solution of human albumin (Sigma; A-8763) with heme at a 1:0.9 protein to heme molar ratio to ensure that no free, uncomplexed heme remained in the preparation [22 (link)]. Human hemopexin (Sigma; H-9291) and bovine methemoglobin (MP Biomedicals; 151234) were also used. Co-incubation of apo-HmuY or apo-Tfo with hemoproteins and HmuY in apo-form with Tfo-Fe(III)heme complex was carried out in PBS (pH 7.6 and 6) at 37°C and monitored by UV-visible spectroscopy using holo-Tfo and apo-HmuY each at 10 μM [14 (link),21 (link)].

Cyclic voltammetry and linear sweep voltammetry were performed using a potentiostat (Autolab PGSTAT128N, Eco Chemie, The Netherlands). With its control software (NOVA Eco Chemie), a three-electrode cell was used, which consisted of a carbon screen-printed electrode (area of 2 × 3 mm², Quasense. Co., Ltd., Bangkok, Thailand) performing as a working electrode, a platinum wire used as a counter electrode (length of 5.7 cm), and an Ag/AgCl electrode in saturated KCl (NF343 RF-1B, ALS Co., Ltd., Tokyo, Japan), used as a reference electrode. Cyclic voltammetry of MB/GP film was carried out using a phosphate-buffered solution containing no hemeproteins. Each of the electrodes was kept in the desiccator until being used for evaluation in the next experiment. The buffer solution was purged with high-purity nitrogen for 30 min before each experiment. X-ray photoelectron spectroscopy (XPS, PHI VersaProbe II) was carried out at the SUT-NANOTEC-SLRI Joint Facility, BL5.2: SUT-NANOTEC-SLRI, Synchrotron Light Research Institute. The excitation energy was 1486 eV using an Al Kα source. The morphology of MB/GP nanohybrids was characterized by using a scanning electron microscope (FE-SEM) (NovaTM NanoSEM 450), (JEM-1011, JEOL, Tokyo, Japan) operated at 10.0 kV, with a magnification of 2500× and 10,000×. Ultraviolet–visible (UV-Vis) absorption spectra of MB in aqueous solutions were detected with a UV-Vis spectrophotometer (CYT5FV, Agilent BioTek, San Diego, CA, USA). The characteristic functional groups of MB/GP nanohybrids were examined using Fourier transform infrared spectroscopy (INVENIO, Bruker, Germany). The zeta potentials of MB/GP nanohybrids were measured using electrophoretic light scattering (ELS) spectrophotometer (Dynamic Light scattering model Zetasizer Nano ZS, Malvern, England).

Full details of the raw data and set-up and analysis procedures are given in the Supporting Information Tables S1–S15 and Figures S1–S25. We will focus here on the main trends and results, however.
The molecular dynamics (MD) set-up was started from the published CYP1A2 and melatonin structures as taken from the Protein Databank [18 (link)] under PDB entries 2HI4 [49 (link)] and ML1. The substrate and crystal water molecules were removed from the 2HI4 pdb file in Chimera UCSF [100 (link)] and chain A was selected. The heme was manually modified into a Cpd I structure with an Fe−O bond length set to 1.686 Å. Hydrogen atoms were added to the structure in Ambertools using pH 7 conditions [101 (link)]. Protonation states of key residues were manually corrected through visual inspection of their local environment. The melatonin structure was geometry optimized in Gaussian-09 [102 ] using a density functional theory method at the B3LYP/6-311 + G* level of theory [103 (link),104 (link)] and converted into PDB format. The PDBs of our CYP1A2 and melatonin structures were assigned as receptor and ligand for substrate docking by AutoDock Vina [78 (link)] with a simulation box with a size of 20.0626 × 23.4102 × 21.855 ų. The ten lowest-energy structures after docking were saved separately into PDB format as the orientation of the substrate with respect to the heme was different in each of them. The maximum energy difference between the best and worst binding mode was set to 2 kcal mol⁻¹. The maximum number of binding modes was set to ten.
MD parameters for the heme complexes were calculated from QM methods by taking the heme complex with its first-coordination sphere ligands, namely the four ligands of the heme, one ligand of cysteine on the L-helix and the distal oxo ligand. The MCPB.py routine implemented in AmberTools 2018 [105 (link)] was used to generate the additional parameters for the MD simulations. The enzyme model was solvated in a rectangular box with a 10 Å distance between the box edges and the enzyme and filled with TIP3P defined water molecules [106 (link)], while the standard amino acids were described by the ff14SB12 forcefield [107 (link)]. The system was neutralized by adding Na⁺ and Cl⁻ ions to the surface of the model. After that, the prepared structure was minimized, heating to 310 K and finally a production run was performed. The minimization was performed in a single step without any constraints with steepest descent of 2000 cycles. Next, the enzyme was heated up from 0 to 310 K in 10 ns. Lastly, the production was run for 100 ns under the following conditions: constant temperate and pressure at 310 K and 1 bar, respectively. The 100 ns MD simulation was run sequentially for 20 times of 5 ns at a time. For model VI of CYP1A2 reactants and model III of CYP1A1 [65 (link)], we expanded the MD simulation to 1 μs.
The results from the MD simulations were collected into a database and analyzed in detail. The stability of each model was evaluated by checking the Root Mean Square Deviation (RMSD) of the atom positions of the various groups in the model, including the water shell. Most MD runs stabilized their RMSD within 60 ns. Therefore, the results of each model after 60 ns were further analyzed. In particular, the ten structures with lowest total energy were taken and their RMSD compared to the starting structure. This process ensured the similarity of those ten structures. Next, all residues within a radius of 5 Å from melatonin substrate were listed. Due to the mobility of the substrate in the binding pocket, its interactions play an important role in its possible reactivity. As such, all ten models are different and show different protein–substrate interactions. The occurrence of the interaction between the substrate and the residues nearby were collected and compared. The result from this process highlights all residues that potentially can interact with the substrate.
For a number of MD snapshots, QM cluster models were created of up to 350 atoms in size that contain the heme, substrate and heme–substrate protein interactions. To this end, a PDB file from the MD simulation was taken and trimmed to the appropriate size and shape. To this end, amino acid side chains pointing out of the active site were truncated to Gly residues by replacement of the side chain with a hydrogen atom. These large QM cluster models are known to reproduce experimental rates and selectivities well [108 (link),109 (link),110 (link)]; however, the larger the model, the more calculation time is required. Initial geometry optimizations were run at the UB3LYP level of theory [103 (link),104 (link)] in Gaussian-09 [102 ] and utilized a basis set designated BS1 with LANL2DZ with core potential on iron [111 (link)] and 6–31G* on C, H, O, N and S. Single-point calculations with basis set BS2 were performed to correct the energies, whereby a 6–311 + G* basis set was used on C, H, O, N and S. The effect of solvent was tested through single-point calculations at the same level of theory but with a dielectric constant mimicking chlorobenzene included. To test the effect of the basis set on iron, a subsequent set of single point calculations at basis set BS3 improved the iron basis set to cc-pVTZ [112 (link)], and solvent through the conductor-type polarized continuum model with dielectric constant mimicking chlorobenzene. The UB3LYP/BS1 approach was used for geometry optimizations, analytical frequency calculation and constrained geometry scans, while BS3 was used in order to obtain more accurate energy results. Each model was calculated in the doublet and quartet spin states with overall charge of −2. Free energies were calculated at a temperature of 298.15 K and 1 bar pressure, whereby vibrational frequencies and entropies are corrected using the quasi-harmonic approximation [113 (link)]. In general, quasi-harmonic corrections give the same trends as without the corrections, and do not change the order of the transition states. The quantum chemical calculations were validated with a range of methods and little effect on the structures and energetics was found when the computational method or basis set were changed [114 (link),115 (link),116 (link)]. Transition states were verified by analyzing the imaginary mode, and all had a single imaginary frequency for the correct transition.

Fresh pork longissimus lumborum (LL) was purchased from a local market, immediately shipped to the laboratory under cold storage. Washed pork muscle (WPM) was prepared as described formerly (16 (link)). Before the treatments, fresh pork LL was kept at −40°C for 1 d to remove heme proteins better. The pork LL was minced. Then the minced muscle was mixed with sodium phosphate buffer (m/v,1:4), homogenized, and centrifuged to collect precipitation. This process was repeated five times, and the collected precipitate was called WPM. The whole washing process was carried out at a low temperature (4°C). After pre-washing, about 0.2 g WPM was used to determine the moisture content of samples. The remaining WPM was vacuum-packaged in polyvinylidene chloride material and stored at −40°C for further tests.