Catalytic surface radical in dye-decolorizing peroxidase: a computational, spectroscopic and site-directed mutagenesis study

Dye-decolorizing peroxidase (DyP) of Auricularia auricula-judae has been expressed in Escherichia coli as a representative of a new DyP family, and subjected to mutagenic, spectroscopic, crystallographic and computational studies. The crystal structure of DyP shows a buried haem cofactor, and surface tryptophan and tyrosine residues potentially involved in long-range electron transfer from bulky dyes. Simulations using PELE (Protein Energy Landscape Exploration) software provided several binding-energy optima for the anthraquinone-type RB19 (Reactive Blue 19) near the above aromatic residues and the haem access-channel. Subsequent QM/MM (quantum mechanics/molecular mechanics) calculations showed a higher tendency of Trp-377 than other exposed haem-neighbouring residues to harbour a catalytic protein radical, and identified the electron-transfer pathway. The existence of such a radical in H2O2-activated DyP was shown by low-temperature EPR, being identified as a mixed tryptophanyl/tyrosyl radical in multifrequency experiments. The signal was dominated by the Trp-377 neutral radical contribution, which disappeared in the W377S variant, and included a tyrosyl contribution assigned to Tyr-337 after analysing the W377S spectra. Kinetics of substrate oxidation by DyP suggests the existence of high- and low-turnover sites. The high-turnover site for oxidation of RB19 (kcat> 200 s−1) and other DyP substrates was assigned to Trp-377 since it was absent from the W377S variant. The low-turnover site/s (RB19 kcat ~20 s−1) could correspond to the haem access-channel, since activity was decreased when the haem channel was occluded by the G169L mutation. If a tyrosine residue is also involved, it will be different from Tyr-337 since all activities are largely unaffected in the Y337S variant.

The structure of DyP was solved by molecular replacement using the crystal structure of B. adusta DyP (PDB entry 4AFV) as the search model and the program AUTOMR of the PHENIX package [2]. The final model was obtained by successive refinement rounds (PHENIX package) followed by manual building with Coot [3] using σ A weighted 2Fo-Fc and Fo-Fc electron density maps. Solvent molecules were introduced in the refinement, as implemented in the PHENIX package, and visually inspected. A total of 5% of reflections was used to calculate the R free value throughout the refinement process. The structures of all the variants were solved using that of WT DyP as a model, and the final models were obtained in the same way as explained above. The structures were validated with MolProbity [4]. PyMOL [5] and Deep view/Swiss Pdb-Viewer (www.expasy.org/spdbv) were used for structure visualization and analysis, and image generation.

MALDI-TOF after steady-state turnover
Eventual changes in the molecular mass of WT DyP, similar to those produced during activation of the T. cervina LiP by formation of a VA adduct with the catalytic tyrosine residue, were analyzed by MALDI-TOF using an Autoflex III instruments, and 2,5-dihydroxyacetophenone matrix, after enzyme reaction with H 2 O 2 and VA, as previously described [6]. For these analyses, 10 M enzyme was incubated for 1 h at 25 °C, in sodium tartrate (pH 3) containing 0.5 mM H 2 O 2 and 10 mM VA, and compared with a control without H 2 O 2 and VA. The reaction mixture was centrifuged into a 30 kDa Centricon Millipore, diafiltrated with 3 vol of 20 mM tartrate (pH 5).

Computational analyses: PELE, MD and QM/MM calculations
The model used is based on the recombinant A. auricula-judae DyP crystal structure solved here. As the optimum pH for oxidation of RB19 by DyP is 3.5 we prepared the starting structure accordingly. All ionizable residues were inspected with Schrodinger's protein preparation wizard [7] and with the H++ web server [8]. His-304 was assessed to be δ-protonated, His-115 εprotonated, and the remaining are double protonated (positively charged). At this low pH several aspartic (residues 8, 12, 84, 129, 189, 246 and 270) and glutamic (residues 158, 220, 225 and 432) acids are in their acidic form, while all other ones are found in their anionic states. Electrostatic potential atomic charges on the RB19 substrate (Fig. S1A), to be used in PELE and molecular dynamics (MD), were obtained from an optimization with Jaguar [9] at the DFT/M06-L level with the 6-31G** basis set and a PB implicit solvent.
Once the initial protein structure was prepared, RB19 was placed manually in 20 initial random positions on the protein's surface and the protein-ligand conformational space was explored with PELE [10]. This is a Monte Carlo based algorithm that produces new configurations through a sequential ligand and protein perturbation, side chain prediction and minimization steps. New configurations are then filtered with a Metropolis acceptance test, where the energy is described with an all-atom OPLS force field and a surface generalized Born solvent. In this way it is possible to locate and characterize local and global minima structures for the most favorable protein-ligand interactions. Results shown are based on 160 independent 48-h PELE simulations. Enhanced local sampling on Trp-377 surface site was obtained with a 5 ns MD simulation with DESMOND [11]. Such analysis allowed us to investigate the effect of solvent and charge fluctuations on the oxidative tendency of Trp-377 and RB19.
QM/MM calculations were performed with Qsite [12]. This method allows for the incorporation of the complete protein structure (as well as solvent and ions) at the atomic level while electronic structure based methods are employed in a sub-section of the system. Most calculations were performed at the M06-L(lacvp*)/OPLS level. Spin densities for the highly reactive compound I were computed at the quartet spin state (easier to converge than the isoenergetic doublet one). Spin density on residues was computed by adding all the atomic spin density contributions. Those residues included in the QM region with/without the haem cofactor (compound I) and a substrate (RB19) molecule are indicated in each case. Additionally, electron transfer pathway calculations were performed with the QM/MM e-pathway approach (using the Hartee-Fock method) [13]. This approach maps those residues with higher probability to participate in the electron transfer pathway by finding, iteratively, those regions of the transfer region with lower ionization energy. In particular, we computed the pathway from Trp377 by considering in the QM region His-304, Ile-305, Arg-306, Lys-307, Thr-308, Arg-309, Pro-310, Arg-311, Leu-323, Ser-324, Ala-325, Leu-373, Gln-374, Gln-375 and Asp-395.

EPR equipments
CW X-band (9.8 GHz) measurements were performed with a Bruker E500 Elexsys Series using the Bruker ER 4122SHQE cavity, equipped with an Oxford helium continuous flow cryostat (ESR900). W-band (94.17 GHz) experiments were recorded on a Bruker Elexsys E600 spectrometer, operating in continuous wave, equipped with a 6T split-coils superconducting magnet (Oxford Instrument), using a continuous helium flow cryostat (Oxford Instrument).

Haem pocket site-directed variants: Electronic absorption and EPR spectra
Asp-168 in DyP occupies the position of conserved distal histidine in classical peroxidases. This residue and the neighbor Arg-332 are expected to play a central role in catalysis since both the D168N and the R332L variants obtained were fully inactive on the different substrates.
The electronic absorption spectra of DyP resting and H 2 O 2 -activated (2 eq) states at pH 3 (the optimal pH for DyP activity, as shown in Fig. S6) are included in Fig. S4A. The latter exhibit the main peak at 403 nm, and small maxima at 529, 558, 597 and 620 nm, all of them being characteristic for compound I. However, the D168N and R332L variants were unable to react with H 2 O 2 (pH 3) as shown by their unchanged absorption spectra ( Fig. S5A and B). When 2 eq of H 2 O 2 were added to the resting enzyme at pH 7 (which showed reduced absorbance at 406 and 630 nm and the 506 nm maximum displaced to 502 nm with respect to the pH 3 resting state) a compound II-like spectrum, with main peak at 418 nm and small maxima at 530, 554 and 634 nm, was obtained (Fig. S4B). Unexpectedly, no compound II-like spectrum could be obtained at pH 3 by self-reduction or using other H 2 O 2 concentrations.
The involvement of Asp-168 and Arg-332 in DyP reactivity with H 2 O 2 , suggested by the electronic absorption spectra, was investigated by EPR. The resting state spectrum of the D168N variant showed a ferric iron in a rhombic environment (g xx = 6.06, g yy =5.6 and g zz = 2.0) (Fig. S5C, black trace). After H 2 O 2 activation at pH 3 (Fig. S5C, red trace) only a small decrease in the ferric iron signal is recorded and no protein radicals are formed. An almost undetectable amount of compound I in the form of the porphyrin radical can be identified superimposed to the ferric g= 2 region of the EPR spectrum (Fig. S5C, inset). The results were similar for the R332L variant (Fig.  S5C), although a small radical signal was observed (4% radical yield for R332L at pH 3 compared with 58% for WT DyP, under the same experimental conditions) (Fig. S5D, inset).
The similarities in the haem pocket architecture of unrelated DyP and classical (plant-fungalprokaryotic) peroxidases would result from adaptative convergence. This is the case of: i) the conserved proximal histidine (A. auricula-judae DyP His-304) acting as the fifth haem iron ligand in both enzyme types; and ii) the opposite haem side residues (DyP Asp-168 and Arg-332) most probably contributing to reaction with H 2 O 2 , as distal histidine and arginine do in classical peroxidases. The latter was demonstrated by the loss of DyP ability to form compound I after the D168N and R332L mutations (as shown by electronic absorption and EPR spectra). A central role of the conserved aspartic acid or arginine in H 2 O 2 reaction have been claimed in the B. adusta [14] and R. jostii DyP [15], respectively, but the present results show that both residues are required for A. auricula-judae DyP activity, although some protein radical signal was still observed in the EPR spectrum of the R332L variant. As reported for other DyPs [14], no compound II spectrum was observed during self-reduction of the A. auricula-judae DyP (at physiological pH 3), a fact that suggests fastest self-reduction of compound II than of compound I, in contrast with other fungal peroxidases [16]. At the other side of the haem, Asp-395, also conserved in other DyPs [14], would help proximal His-304 to modulate the haem electron-deficiency, as the proximal aspartic acid in classical peroxidases [17].       Fig. S9).