Enabling microbial syringol conversion through structure-guided protein engineering

Significance Lignin is an abundant but underutilized heterogeneous polymer found in terrestrial plants. In current lignocellulosic biorefinery paradigms, lignin is primarily slated for incineration, but for a nonfood plant-based bioeconomy to be successful, lignin valorization is critical. An emerging concept to valorize lignin uses aromatic–catabolic pathways and microbes to funnel heterogeneous lignin-derived aromatic compounds to single high-value products. For this approach to be viable, the discovery and engineering of enzymes to conduct key reactions is critical. In this work, we have engineered a two-component cytochrome P450 enzyme system to conduct one of the most important reactions in biological lignin conversion: aromatic O-demethylation of syringol, the base aromatic unit of S-lignin, which is highly abundant in hardwoods and grasses.

Determination of [FAD] and non-heme [Fe] in GcoB. FAD was released from GcoB by denaturing 200 μL of a protein (0.024 µM) solution with 5 μL saturated ammonium sulfate, pH 1.4 (7% v/v H2SO4). Precipitated protein was pelleted by centrifugation and the UV/vis spectrum of the FAD-containing supernatant was measured. The absorbance at 454 nm, εFAD = 11.3 mM -1 cm -1 , and total protein concentration determined by the BCA assay (Pierce) were used to determine [FAD] bound to GcoB. An extinction coefficient for GcoB-bound FAD was estimated via the slope of a line relating absorbance at 454 nm to [GcoB-FAD].
To determine the Fe-S content of GcoB, the protein was first denatured as described above. 50 μL of supernatant was added to 25 μL of 5% w/v sodium ascorbate to reduce the iron. 100 μL of bathophenanthroline disulfonate (0.1% w/v in ddH2O) was added and the sample was incubated for 1h. The resulting Fe(II) complex was quantified via its absorbance at 535 nm (ε535 = 22.14 mM -1 cm -1 , determined using FeSO4 standards). An extinction coefficient for GcoB-bound 2Fe-2S cluster was estimated via the slope of a line relating absorbance at 423 nm to [GcoB-2Fe-2S].
Steady state kinetics and substrate dissociation constants. Steady state kinetics of F169A. 0.2 µM each of F169A GcoA and GcoB were dissolved in air-saturated buffer (25 mM HEPES, 50 mM NaCl) in a cuvette at pH 7.5, 25 °C. 100 µg/mL catalase was added to each reaction to capture any H2O2 formed during the uncoupled reaction. A saturating amount of NADH (≥ 5KM, 300 µM) was added and a background rate of NADH oxidation in air (~210 µM O2) recorded via continuous scanning of the UV/vis spectrum (Varian Cary 50). 20-300 µM guaiacol or syringol 2-20 mM stock dissolved in DMSO was added and the reaction was monitored via measurement of UV/vis spectra for several minutes.
The initial velocity was determined by disappearance of the characteristic NADH absorbance at 340 nm (ε344 = 6.22 mM -1 cm -1 ). A plot of vi versus [guaiacol] was fit to equation 4 to obtain kcat, KM, and kcat/KM. For specific activity determination, the above method was used but with saturating (300 µM) guaiacol, syringol, or 3

vi = Vmax[S]/(KM + [S]) (4)
Determination of substrate dissociation constants (KD) with F169A. 0-60 µM of guaiacol, syringol, or 3MC in 0.5 or 1 µM aliquots, were titrated into a cuvette containing 3 µM F169A GcoA in 25 mM HEPES, 50 mM NaCl, pH 7.5. The spectrum after each substrate addition was recorded, beginning with no substrate bound. The solution reached equilibrium before the next addition. A difference spectrum was made to illustrate the shift from a low-spin aquo-heme complex to the high-spin substrate-bound complex (spectral shift from 417 nm to 388 nm). The resulting difference spectra showed a peak at 388 nm, and a trough at 417 nm. The absorbance at 388 nm (Abs388-417 nm) was plotted as a function of [substrate], yielding a quadratic curve that was fit to equation (5) to determine the KD.
Where L0, Et, KD, and ΔAbsmax are the ligand concentrations, total protein (subunit) concentration, the equilibrium dissociation constant, and the maximum Abs388-417 nm, respectively. Reported values are the average of 2 or more measurements.
Product analysis. Formaldehyde determination. A colorimetric assay using tryptophan can be used to quantify the amount of formaldehyde produced in F169 GcoA/B reactions with guaiacol, syringol, or 3MC. 0.2 µM each of F169 GcoA mutants and GcoB were dissolved in airsaturated buffer (25 mM HEPES, 50 mM NaCl) in a cuvette at pH 7.5, 25 °C. 100 µg/mL catalase was added to each reaction to capture any H2O2 formed during the uncoupled reaction. 200 μM NADH was added and the background rate recorded. 100 (guaiacol, syringol, or 3MC) or 200 (syringol) μM of substrate was then added and the reaction monitored until there was no more change, due to either substrate, NADH or O2 depletion, whichever occurred first. 200 µL of the reaction was then quenched by adding 200 µL of a 0.1% tryptophan solution in 50% ethanol and 200 µL of 90% sulfuric acid. Upon thorough mixing, 40 µL of 1% FeCl3 was added. The solution was then incubated in a heating block for 90 min at 70 °C. After cooling, the absorbance was read at 575 nm and the [formaldehyde] calculated by using ε575 nm = 4.2 mM -1 cm -1 , obtained with formaldehyde as a standard (1). A negative control included everything but the substrate and was used as a baseline.
HPLC for product identification and specific activity measurement. Analyte analysis of the above end-point reactions (100 μM guaiacol, syringol, or 3MC, or 200 μM syringol) was performed on an Agilent 1100 LC system (Agilent Technologies) equipped with a G1315B diode array detector (DAD). Each sample and standard was injected at a volume of 10 μL onto a Symmetry C18 column 5 μm, 4.6 x 150 mm column (Waters). The column temperature was maintained at 30 °C and the buffers used to separate the analytes of interest was 0.01% TFA in water (A)/ acetonitrile (B). The separation was carried out using a gradient program of: For specific activity determination, 300 μM guaiacol, syringol, or 3MC were added from 0.1 M DMSO stocks to air saturated buffer (25 mM HEPES, 50 mM NaCl, pH 7.5), with a final volume of 1 mL. The [analyte] was measured via the above HPLC method. Then, 0.2 μM F169A/ GcoB and 100 μg/mL catalase were added. Upon addition of 300 μM NADH, the timer was started and 50 μL removed every 10 (guaiacol and syringol) or 30 (3MC) seconds. The reaction of each aliquot was immediately quenched with 12.5 μL saturated ammonium sulfate, 7% v/v H2SO4 (pH 2.0) prior to loading onto the HPLC column.
[Substrate] disappearance was referenced to GcoA (0.2 µM) and fit to a linear line to determine specific activity.
Data analysis. The consumption of NADH and subsequent production of formaldehyde and aromatic product were measured in triplicate and the error reported as ±1 standard deviation. To determine the statistical significance between NADH consumption and the products produced, the p-value was determined for all runs containing guaiacol, syringol, and 3MC. These values were calculated using two degrees of freedom, a tail value of two, and assuming a t-statistical value for unequal variances.
Uncoupling reactions. Detection of H2O2 via horseradish peroxidase (HRP) and Amplex Red assay. The reaction between 100 μM guaiacol or 3MC with 0.2 μM F169A/GcoB and 100 μM NADH in air-saturated buffer (25 mM HEPES, 50 mM NaCl, pH 7.5) was monitored continuously in a quartz cuvette, using the NADH consumption assay described above. The same thing was done for syringol, but with either 100 or 200 μM NADH. When there was no longer any change in the spectra, e.g., the reaction was completed, 100 μL was removed from the cuvette and pipetted into a 96-well microplate. A 5 mL solution containing 50 μL of 10 mM Amplex Red (prepared in DMSO and stored at -20 °C) and 100 μL of 10 U/mL HRP was made up in the above buffer. 100 μL of this was added to each of the wells with each reaction sample. The plate was incubated in the dark at room temperature for 30 min, at which point the absorbance at 572 nm was recorded by a Varioskan Lux microplate reader (Thermo Scientific). The absorbance was compared to a standard curve with 0-100 μM H2O2 to quantify the amount of peroxide produced in the reactions.

Crystallography
Purified protein was buffer exchanged into 10 mM HEPES pH 7.5 and concentrated to an A280 value of 12, as measured on a NanoDrop 2000 spectrophotometer (Thermo Fisher). Crystals of GcoA were grown with 2.4 M sodium malonate and 200 mM substrate, dissolved in 40% DMSO where necessary. Crystals were cryocooled directly in liquid N2 without further addition of cryoprotectants. All data were collected at Diamond Light Source (Harwell, UK). For each crystal, 1800 images were taken at 0.1 o increments using the default wavelength of 0.9795Å on beamline i04. Data was captured on a Pilatus 6M-F detector. All phases were solved by molecular replacement from the original WT GcoA structure in complex with guaiacol with all non-polypeptide components removed. Data were processed, phased, and models were built and refined using Xia2 (2-6) and the Phenix suite (7)(8)(9)(10).

Strain construction
For genomic manipulations in P. putida KT2440, an antibiotic-sacB counter and counterselection method was utilized as described in Blomfield at al. (11) and modified for P. putida KT2440 (12). Diagnostic PCR was utilized to confirm correct integrations or deletions using the 2X myTaq TM system (Bioline). Specific strain construction details are included in Table S8, plasmid construction details are included in Table S9, and oligo sequences are included in S10.
To transform cells for episomal expression in P. putida, cells were initially grown shaking at 225 rpm, 30 °C, overnight in Luria-Bertani (LB) medium containing 10 g/L tryptone, 5 g/L yeast extract, and 5 g/L NaCl, and used to inoculate LB the following day to an OD600 of ~0.1 and grown until they reached an OD600 of 0.6. Cells were then washed three times using 300 mM sucrose and resuspended in a volume 1/50 of the original culture (13). Plasmid DNA was added, and this mixture was then transferred to 1 mm electroporation cuvettes and electroporated at 1.6 kV, 25 F, 200 Ω. Following the addition of 950 uL of SOC media (NEB), cells were shaken for an additional 45 minutes at 30°C, 220 rpm, and plated on solid media containing appropriate antibiotics.

Shake flask experiments and in vivo product analyses
Strains evaluated in shake flask experiments were initially grown overnight in LB media with appropriate antibiotics from glycerol stocks and resuspended the following day in M9 minimal media (6.78 g/L Na2HPO4, 3 g/L KH2PO4, 0.5 g/L NaCl, 1 g/L NH4Cl, 2 mM MgSO4, 100 µM CaCl2, and 40 µM FeSO4 . 7H2O) supplemented with 20 mM glucose (Sigma-Aldrich) and 50 mg/mL kanamycin. Cells were grown until they reached an OD600 of 1, at which point syringol (Sigma-Aldrich) was added to a final concentration of 1 mM. Syringol concentrations were quantified using any 1 H NMR, and spectra were collected using a Bruker Avance III HD 400 MHz Spectrometer and analyzed using Bruker TopSpin3.7 software. For identification of unknown compounds, samples were analyzed via LC-MS-MS using the Acquity UPLC system (Waters Inc.). Compounds were separated using a C18 (evo) column (Kinetex) with the mobile phase comprised of a water, acetonitrile, and .1% formate. Flow was directly analyzed by SYNAPT HD-MS using electron spray ionization (ESI) in negative ion mode.
Bioinformatics Analysis CYP255A sequences were retrieved by a blastp search against the non-redundant protein sequence database (14) using GcoA as the query sequence. From the blast results, only sequences with a query cover of at least 66% and sharing a sequence identity of greater than 28% but less than 97% with the query sequence were retained. In total, 482 homologs of GcoA were retrieved. A multiple sequence alignment (MSA) was carried out using MAFFT (15) with Biopython (16). Conservation analysis was implemented by computing the relative entropy for each site in the MSA (17). The relative entropy is given by equation (6): where pi is the frequency of amino acid i in the given site and p i MSA is the frequency of amino acid i in the MSA.

Molecular dynamics (MD) simulations
Unrestrained MD simulations. Unrestrained MD simulations were performed on both WT GcoA and F169A mutant with either guaiacol or syringol bound at the active site. The previous constructed model of WT GcoA with guaiacol bound at the active site was used as the initial conformation to generate these structures (1). The heme group was kept in a hexacoordinate state (Compound I). Similar to previous work, five histidine residues, including His131, His221, His224, His255, and His343 were doubly protonated (+1 charge), while the remaining histidine residues were kept neutral (singly protonated). Na + cations were added to each system to achieve charge neutrality, resulting in ~74,000 atoms for each system.
The ff14SB Amber force field (18) parameters were used for the enzyme and Generalized Amber Force Field (GAFF) (19,20) parameters were used for the various substrates, as reported previously (1). Force field parameters for the heme group were taken from ref (21). Particle mesh Ewald (PME) (22) was used for long-range interactions and the cutoff distance for nonbonded interactions was 9 Å. All the simulations were conducted using NAMD program(23) with a time-step of 2.0 fs. SHAKE algorithm (24) was used to keep bonds to hydrogen atoms fixed. Langevin thermostat with a collision frequency of 1.0 ps -1 was used to keep the temperature at 300 K. Each system was relaxed in NPT ensemble for 500 ps first, followed by 80 ns of production was performed in the NVT ensemble. For the initial NPT dynamics, the pressure was held at 1 atm using a Nose− Hoover barostat coupled to a Langevin piston (25,26) with a damping time of 100 fs and a period of 200 fs.
As in our previous work (1) we employed the following reaction coordinate to describe the opening/closing motion of GcoA: ξ = RMSDopen -RMSDclosed, where RMSDopen is the RMSD relative to the most open structure of WT GcoA/apo obtained via microsecond MD trajectory, and RMSDclosed is relative to the closed crystal structure. The Cα atoms of residues 5:35 and 154:210 were chosen to calculate the RMSD.
Replica exchange thermodynamic integration (RETI) simulations. In order to assess relative free energies of binding in the active site of GcoA, RETI simulations were performed (27, 28) using NAMD 2.12 (29). Simulations were performed in the NPT ensemble with simulation parameters identical to those of the MD simulations described above, unless noted otherwise. Initial configurations for RETI were produced from the crystal structure appropriate for each system along with modified AMBER prmtop files to accommodate the dual-topology paradigm employed by NAMD for alchemical simulations. Each TI replica was simulated for between 150 -250 ns until convergence was achieved. The temperature was maintained at 308 K through Langevin dynamics using a collision frequency of 1.0 ps −1 . The cutoff distance for nonbonded interactions was 12.0 Å.
Two separate types of alchemical transitions were performed, with transformations targeting either the substrate or the enzyme. In each case, the transition parameter  describes the progress of alchemical transformation. In the first kind, a syringol molecule at  = 0 undergoes an alchemical transition to a guaiacol molecule at  = 1. This simulation was performed with the substrate molecule in three different contexts: in a periodic water box, in the active site of solvated WT GcoA, and in the active site of solvated GcoA F169A mutant. In the second group of transformations, residue 169 of the GcoA is transformed from Phe at  = 0 to Ala at  = 1. This transformation was also performed in three separate contexts: with guaiacol at the active site, with syringol at the active site, and with empty active site (apo). Although these alchemical transitions were made between the entire substrate molecule (in the first set) or protein residue (in the second set), in both cases many atoms are shared in common, so zero-length 10 kcal/mol/Å 2 "bonds" were applied between equivalent heavy atoms, "pinning" them together. These additional restraints eliminate sampling of unphysical conformations that can slow the convergence of the calculations (30)(31)(32) (33). The van der Waals contributions of both disappearing and appearing residues were simultaneously varied with the reaction coordinate . Electrostatic contributions of disappearing residues were turned off over the first half of the reaction coordinate, while those of the appearing residues were turned on over the second half of the reaction coordinate, such that both appearing and disappearing atoms were uncharged at  = 0.5. Exchanges between adjacent replicas were attempted every 2 ps between alternating replica pairs, yielding an overall exchange attempt rate of once per 4 ps, consistent with literature guidelines (34). The combination of these parameters yielded an acceptance ratio for exchanges of approximately 50%. In each case, the final 100 ns were used to compute each ΔG reported in Figure S11. Third order spline interpolation was used to integrate the average dU /dλ obtained from simulations at each window over the final 100 ns. Error estimates were obtained using the methodology detailed in Steinbrecher et al (35). To avoid underestimating noise error due to autocorrelations of the dU /dλ timeseries, means and standard errors were computed by sampling all of the dU /dλ data at a rate higher than the output rate. For each simulation, this rate was set by finding the longest triple e-folding time of the dU /dλ autocorrelation, as determined by an exponential fit. Typical values of this time were 3-4 ps.

DFT calculations
Density Functional Theory (DFT) calculations were performed using Gaussian 09 (36). A truncated model containing the porphyrin pyrrole core, Fe center and a methanethiol to mimic cysteine as Fe-axial ligand was used. Geometry optimizations and frequency calculations were performed using unrestricted B3LYP (UB3LYP) (37)(38)(39) with the LANL2DZ basis set for iron and 6-31G(d) on all other atoms. Transition states had one negative force constant corresponding to the desired transformation. Enthalpies and entropies were calculated for 1 atm and 298.15 K. A correction to the harmonic oscillator approximation, as discussed by Truhlar and co-workers, was also applied to the entropy calculations by raising all frequencies below 100 cm -1 to 100 cm -1 (40,41). Single point energy calculations were performed using the dispersion-corrected functional (U)B3LYP-D3(BJ) (42,43) with the LANL2DZ basis set on iron and 6-311+G(d,p) on all other atoms, within the CPCM polarizable conductor model (diethyl ether, ε = 4) (44,45) to have an estimation of the dielectric permittivity in the enzyme active site. The use of a dielectric constant ε=4 has been proved to be a good and general model to account for electronic polarization and small backbone fluctuations in enzyme active sites (46,47). All stationary points were verified as minima or first-order saddle points by a vibrational frequency analysis.      GcoA-F169A bound to syringol additionally contains another water molecule, but it has weak electron density, indicating low occupancy and/or a degree of disorder (not shown). We hypothesize that these ordered water molecules occupy space in the active site that helps to maintain the substrate in a productive binding pose.  Figure S7. MD simulations indicate that WT GcoA bound with syringol lengthens the key distance for the rate-limiting step for demethylation, namely the distance between the methyl hydrogen and the oxygen atom bound by Compound I heme. The GcoA-F169A mutation brings this distance back to that seen with guaiacol, promoting demethylation. It should be noted that this shift as displayed in MD simulations is significantly subtler than the substrate shift seen in crystal structures ( Figure 1A of main text). Error bars quoted in the table represent the standard error of the mean.   Figure S10. Microsecond MD simulations indicate the increased flexibility and displacement of F169 in WT GcoA when syringol is bound rather than guaiacol. These analyses are performed on three simulations (each) of WT GcoA with guaiacol and with syringol. These simulations were performed in our original study of GcoA (1); however our previous publication focused on the utilization of guaiacol and presented the results with syringol only very briefly (the fifth entry of Figure S21 in the Supplementary Information). Time traces of the reaction coordinate (RC) for opening and closing as well as the RMSD of the three active site Phe residues were presented in the SI of the prior publication. A) Time traces of the RMSD of residue F169 and B) probability distributions of the same indicate that F169 deviates substantially from its crystal structure position when syringol is bound. The movement of F169 is somewhat correlated with the motion of the substrate access loop. The probability distribution shows that each guaiacol simulation spends significant time with F169 very near the crystal structure position (first peak) whereas none of the syringol simulations display this peak. C) RMSF averaged over three simulations each for WT GcoA with either guaiacol or syringol. The significant area of deviation is in the substrate access loop (which includes F169), indicating greater dynamic flexibility when syringol is bound.  Figure 2 in that publication). In our 80 ns MD simulations, the RMSD of the combined F and G helices, as well as the adjoining loop fluctuate around an RMSD value of 3 Å, whereas the adjoining loop alone fluctuates around 5 Å, as shown in the graphs here. Thus, the degree of opening by this metric is on the same order, but lower by about 50% from that observed in P450cam crystal structures.
In addition to P450cam, we also note P450 BM3, which also displays crystallographic evidence of open (e.g. PDB codes 1BVY (50)) and closed (e.g. 1JPZ (51) and 1FAG (52) For GcoA, a similar metric is the distance between alpha carbons of Gly44 and Gly178, which is 6.35 Å in the closed state (5NCB). In our simulations of WT GcoA, the difference between the simulation with syringol bound and that with guaiacol bound is slightly more than 3 Å. As shown in the graph below, the average distance over the final 40 ns of the simulation is 8.0 Å (guaiacol) and 11.1 Å (syringol).
We note that, to date, efforts to crystallize GcoA in the apo state have proven unsuccessful. As a result, all existing crystal structures demonstrate GcoA in the closed state. Thus, the range of motion of the substrate access lid for GcoA is currently unverified experimentally. The range of motion that we observe in our 80 ns simulations is not necessarily the full range of motion in the catalytic cycle of GcoA (and very likely is not). Additionally, it is possible that the primary effect of the increased flexibility and propensity for the substrate access lid to open when syringol is bound could be upon substrate binding/egress or on the catalytic demethylation step.
Figure S11. DFT calculations for possible reaction pathways for syringol degradation catalyzed by Compound I (CpI). A) Target pathway for syringol demethylation to 3MC catalyzed by P450 CpI. Our calculations indicate syringol demethylation follows a similar pathway to that previously described for guaiacol (1). Free energy barriers for the rate-limiting initial hydrogen atom transfer (HAT) from the methoxy group to the heme-bound oxygen atom are very similar for guaiacol and syringol (ΔG ‡ = 18.6 and 18.9 kcal·mol -1 , respectively), and also are the optimized transition state (TS) geometries. Consequently, the demethylation reaction likely proceeds similarly for both substrates when they can bind productively with the CpI reactive species in the enzyme active site. Unproductive pathways b (B) and c (C) lead to undesired products (and possibly uncoupling) and are intrinsically lower in energy than pathway a. This is indicating that enzymatic environment that forces an specific substrate binding pose, as demonstrated by X-ray structures and MD simulations, is preventing pathways b and c to take place and promotes desired pathway a. Gibbs energies, obtained at B3LYP-D3BJ/6-311+G(d,p)+Fe(LanL2DZ)(PCM=Diethylether)//B3LYP/6-31G(d)+Fe(LanL2DZ), are given in kcal/mol. Distances and angles in DFT optimized key TSs are given in Å and degrees, respectively.
Figure S12. Replica exchange thermodynamic integration (RETI) simulations indicate the change in free energy for different alchemical transformations. Each solid arrow indicates a RETI calculation that was performed and has the associated free energy change for the process in the direction of the arrow. Dashed arrows represent transformations that were not performed. A full four-step thermodynamic cycle was completed, represented by the four boxes in the lower right. The closure in performing these four steps theoretically is equal to zero. We calculate a closure of -1.6 kcal/mol, which is on the order of the expected cumulative error one would expect given the errors in the individual steps (approximately 0.2 kcal/mol).
The most surprising of these thermodynamic results is the relatively low free energy difference between syringol and guaiacol bound at the active site of WT GcoA. This can explained by the fact that these simulations indicate that binding syringol in the WT can produce binding complexes in which the substrate is flipped "upside down" from the configuration seen in crystal structures (see Fig. S12). This may help explain the experimental result that while WT demethylation activity is greatly reduced when exchanging syringol for guaiacol, the binding preference is less dramatic. Substrate flipping could reduce the crowding penalty in WT GcoA. The analogous set of simulations in the GcoA-F169A mutant (as well as mutation transformations with either syringol or guaiacol bound) showed comparatively few flips (for substrate transformation in WT, the substrate was flipped ~90% of the time, compared to ~10% in the other three cases). Experimentally, syringol binding affinity is not much changed from the WT to GcoA-F169A (KD values indicate a difference in binding free energy of 0.3 kcal·mol -1 ), even as the activity is improved markedly. The RETI results indicate the possibility that this may be because KD measurements detect all binding at the active site capable of displacing a Fe-coordinating water molecule whereas catalytic turnover will only result from productive binding with catalytically relevant positioning of substrate relative to CpI heme. RETI results present the possibility that this is because the GcoA-F169A mutant is able to keep both substrates oriented correctly whereas WT GcoA is not. Figure S13. Alternate substrate binding configurations seen in RETI simulations. These results indicate a flip of substrate that is unique to the ΔGsub,WT calculation, which involves an alchemical transformation between guaiacol and syringol in the active site of WT GcoA. Given the differences in KD, activity, and crystal structures, the expected result for this transformation would be that guaiacol would be significantly favored at the WT active site. It is therefore surprising that this transformation results in a free energy change that slightly favors syringol (by 0.83 ± 0.069 kcal/mol); this is explained by the fact that the substrate adopts a unique binding pose not seen in crystal structures or any other MD simulations. A) Exemplified by syringol, in all 20 windows (in which syringol is gradually "turned off" and guaiacol is gradually "turned on") of the ΔGsub,WT transformation, this "upside down" configuration dominates, spending only about 10-15% of the simulation time in a configuration similar to that seen in crystal structures. By contrast, in the other transformations (ΔGmutate,gua, ΔGsub,F169A, and ΔGmutate,syr) the "upright" substrate configuration is displayed in upwards of 90% of the simulation time. This is represented in panel B by a representative frame from the ΔGsub,       ), which was then referenced to NADH consumed. b The same reaction done above was repeated to quantify H2O2 produced using the Amplex Red/HRP assay (see SI Appendix, Methods). The amount of hydrogen peroxide was then referenced to NADH consumed.  (55)  pCJ004 Deletion of catA2 in P. putida KT2440 Construction of pCJ004 was described previously in Johnson and Beckham (12). pCJ005 Replacing catBCA with Ptac:xylE in P. putida KT2440 The 5' targeting region was amplified from P. putida KT2440 genomic DNA with primer pair oCJ042/oCJ043 (1,104 bp, which incorporated the tac promoter), xylE (969 bp) was amplified from P. putida mt-2 (ATCC 23973) genomic DNA with primer pair oCJ044/oCJ045, and the 3' targeting region was amplified using primer pair oCJ046/oCJ047 (1033 bp). These fragments were then assembled into pCM433 digested with AatII and SacI (7991 bp).

pCJ011
Deletion of pcaHG in P. putida KT2440 Construction of pCJ011 was described previously in Johnson and Beckham (12). pCJ020 Insertion of tac promoter upstream of pcaHG in P. putida KT2440 Construction of pCJ020 was described previously in Johnson and Beckham (12). pCJ021 Episomal expression of gcoAB on pBTL-2 Construction of pCJ021 was described previously in Tumen-Velasquez, et al. (56). pCJ105 Deletion of catBCA in P. putida KT2440 The 5' (1054 bp) and 3' (1044 bp) targeting regions were amplifed from pCJ005 with primers oCJ542/oCJ543 and oCJ544/oCJ545, respectively, and assembled into pK18sB (Genbank MH166772 (57) Movie S1. Motion of active site Phe residues in WT GcoA with guaiacol bound. 80 ns MD simulation demonstrates the relative stability of F169 (center), F75 (left), and F395 (right). Compound I heme is shown in sticks along with Fe molecule (orange) and bound oxygen atom (red) in sphere representation.
Movie S2. Motion of active site F residues in WT GcoA with syringol bound. 80 ns MD simulation reveals the increased flexibility of F75 (left), F395 (right), and especially F169 (center). Compound I heme is shown in sticks along with Fe molecule (orange) and bound oxygen atom (red) in sphere representation.
Movie S3. Substrate access loop is relatively stable in WT GcoA with guaiacol bound. 80 ns MD simulation reveals a relatively stable substrate access loop (residues 154-210, shown in raspberry color) when guaiacol (green "sticks" representation) is bound at the active site. Compound I heme is shown in sticks along with Fe molecule (orange) and bound oxygen atom (red) in sphere representation.
Movie S4. Substrate access loop is flexible in WT GcoA with syringol bound. 80 ns MD simulation reveals a more active substrate access loop (residues 154-210, shown in raspberry color) when syringol (pink "sticks" representation) is bound at the active site.
Compound I heme is shown in sticks along with Fe molecule (orange) and bound oxygen atom (red) in sphere representation.