Direct Evidence for a Peroxide Intermediate and a Reactive Enzyme–Substrate–Dioxygen Configuration in a Cofactor-free Oxidase

Cofactor-free oxidases and oxygenases promote and control the reactivity of O2 with limited chemical tools at their disposal. Their mechanism of action is not completely understood and structural information is not available for any of the reaction intermediates. Near-atomic resolution crystallography supported by in crystallo Raman spectroscopy and QM/MM calculations showed unambiguously that the archetypical cofactor-free uricase catalyzes uric acid degradation via a C5(S)-(hydro)peroxide intermediate. Low X-ray doses break specifically the intermediate C5=OO(H) bond at 100 K, thus releasing O2 in situ, which is trapped above the substrate radical. The dose-dependent rate of bond rupture followed by combined crystallographic and Raman analysis indicates that ionizing radiation kick-starts both peroxide decomposition and its regeneration. Peroxidation can be explained by a mechanism in which the substrate radical recombines with superoxide transiently produced in the active site.


X-ray data collection
Data collections on UOX crystals belonging to space group I222 were carried out either using our in-house sealed-tube X-ray instrument (Agilent Nova) or synchrotron radiation (ESRF, Grenoble-France and Diamond Light Source, Didcot-U.K.). For synchrotron measurements X-ray dose calculations were performed with the program RADDOSE-3D [1] to take into account the larger crystal dimensions compared to the X-ray beam cross-section. The standard protocol for multiple data collections involved a series of identical acquisitions with a rotation range of 180° at low dose (data sets) interspersed by 'burn' acquisitions. Data collection parameters and statistics are given in Table S1. None of the general global features that typically accompany X-ray radiation damage effect such as increase of cell dimensions, increased overall ADP values are present here as the X-ray dose provided is about 40-100 lower than the Garman limit. [2] Typical specific damages in proteins are also rupture of disulfide bridges and decarboxylation of acidic residues. [3] Whilst there are no disulfide bridges in UOX, no sign of Asp/Glu decarboxylation are observed.

Crystallographic refinement
The programs COOT [4] and Refmac5 [5] were used throughout for model rebuilding and anisotropic refinement with riding hydrogen atoms, respectively.
Refinement statistics are given in Table S1. Refinement was additionally carried out with phenix.refine [6] and Shelxl [7] software packages for crossvalidation of occupancy (q) values. Isomorphous difference maps were calculated within Phenix [8] .
For occupancy refinement, the sum q PEROXIDE + q FLAT (FLAT refers to the planar radical moieties (Sub• + Sub*) arising from peroxide rupture) was constrained to unity whereas q DIOXYGEN was refined independently. The q PEROXIDE + q FLAT = 1.0 condition reflects the observation that geometrically invariant atoms in the PEROXIDE à FLAT reaction are present at full occupancy whereas electron density for the diatomic specie is lower. In test refinement runs we also removed the q PEROXIDE + q FLAT = 1.0 constrain. This resulted in occupancy values, whose sum was essentially unitary (0.97-0.99).
We therefore considered justified to keep this constrain throughout.

Kinetic fitting
Several reaction schemes were tested by performing kinetic fitting of dose-dependent occupancies with the package Dynafit. [9] The reaction scheme below provided the best results assessed by fitting quality (log(goodness of fit)) and lowest errors on the kinetic constants derived (errors lower than 35%) Changes in concentrations of the chemical species were computed by solving an initial value problem described by the following system of differential equations: We emphasize that although Scheme 1 allows a satisfactory kinetic fit of the data it contains speculative elements and at present should be considered as a working model.

Raman measurements
Non-resonant Raman spectra were recorded at 100 K at the ESRF (Grenoble) either off-line at ID29S-Cryobench laboratory or on-line at the ID29 beamline using an inVia Raman microspectrometer (Renishaw, Gloucestershire, UK) equipped with a near-IR 785 nm diode laser. [10] For online measurements both the spectrometer and the laser located at the Cryobench were connected via a 50 m-long optical fibre cable to the Raman head installed at the beamline and driven by a precise motorized device. The Raman laser was co-axial (antiparallel) with the X-ray beam. This experimental procedure allows a single crystal to be alternatively probed by Xray crystallography and Raman spectroscopy in an interleaved manner without any manual intervention on the sample. The same location of the crystal (within 10 µm) was probed by microspectrometry using a laser power of 50 mW at the sample position and 2000 s accumulated exposure time for each spectrum. As crystal orientation is known to strongly affect the relative band intensities in Raman spectra, e.g. due to polarization effects, the same spindle axis position was used for all Raman measurements. Non-resonant Raman data acquisition with laser irradiation at 785 nm does not result in detectable degradation of the structural and spectroscopic signatures of the samples. To obtain peak intensities, a curve-fitting procedure was applied using a group of Lorentzian peaks between 500 cm -1 and 700 cm -1 with a linear baseline.

UV-visible light absorption microspectrophotometry
UV-visible light absorption spectra of non-irradiated and irradiated crystals were obtained at the in crystallo spectroscopy facility ID29S Cryobench at the ESRF, Grenoble, France. The reference white light is provided by a balanced deuterium-halogen lamp (Mikropack DH2000-BAL, Ocean Optics). Spectra are recorded using a fixed-grating spectrophotometer with a CCD detector (QE65Pro, Ocean Optics).

Theoretical calculations
MD simulations were initiated based on the available crystal structures. The starting structures were prepared using the CHARMM [11] software and the MD simulations were performed with NAMD [12] (version 2.9). The coordinates of the hydrogen atoms were generated with CHARMM using the standard protonation states for the titratable residues. As we lack experimental information about the protonation and the total charge of the substrate, we identified and tested several valence models. The positions of the hydrogen atoms of the ligand were optimized in the absence of solvent at the B3LYP/6-31+G* level of theory using QCHEM [13] and the atomic charges in the substrate were calculated using the CHELPG [14] algorithm. The protein was placed in the center of a cubic box that extends at least 10 Å in all directions from the system. This cube was filled with water molecules and the total charge of the system was neutralized using either K + or Clions. The Particle Mesh Ewald method was used to treat the electrostatics of the periodic boundary conditions with a non-bonded cutoff of 13.5 Å and with an electrostatic force shifting function and a van der Waals switching function between 10 and 12 Å. The CHARMM27 force field was used throughout the simulations. The protein and the solvent were energy minimized following a two-step procedure, while the substrate was kept fixed at its crystallographic coordinates. Firstly, the non-backbone atoms and non-crystallographic water molecules were minimized for 9000 steps, subsequently all the atoms were After the minimizations with an added electron, the HOMO and HOMO-1 orbitals of the system and the CHELPG atomic charges were calculated using Gaussian 09. [15]     A total of seven data collections, DC(i), interspersed by 'burn datasets', B(i), and online Raman microspectrophotometry measurements, R(i), were acquired from a single UOX:5-PMUA crystal. Dose calculations were performed using the program RADDOSE-3D (www.raddo.se) (20) to take into account the bigger crystal size compared to the X-ray beam dimensions The diffraction weighted dose (DWD) represents the effective average dose that is observed in the diffraction pattern [16] . Raman microspectrophotometry probes the surface of the irradiated crystal volume. The dose at the surface considering the experimental geometry was calculated from the RADDOSE-3D output: Average_dose_surface = average_dose_irradiated_volume/1.54. Figure S5. Protocol employed for the crystallographic measurement performed on UOX:5-PIU at the I02 beamline (DLS, Dicot). A total of eleven data collections, DC(i) interspersed by 'burn datasets', B(i), were acquired from a single UOX:5-PIU crystal. Dose calculations were performed using the program RADDOSE-3D (www.raddo.se) (20) to take into account the bigger crystal size compared to the X-ray beam dimensions The diffraction weighted dose (DWD) represents the effective average dose that is observed in the diffraction pattern [16] . Raman microspectrophotometry was not performed on this sample due to the low intensity of 5-PIU specific bands ( Figure S3). Figure S6. Dose-dependent rupture of the C5-Op1 bond of 5-PIU. (A) Data sets for UOX:5-PIU were collected on the same crystal at increasing X-ray dose according to the protocol in Figure S5. Upon X-ray exposure the C5-Op1 bond breaks resulting in the accumulation of O 2 and the planar UA species (UA• + UA*, see Scheme 1 in the main text). The (W2) water molecule moves close to the O8 atom at H-bond distance. Organic moieties and O 2 and are shown as ball-and-stick models. Waters in close proximity are shown as spheres. Carbon, oxygen, and nitrogen atoms are shown in brown, red and blue, respectively. 2mF o -DF c electron density maps at the 1.0σ level are shown in brown. Hydrogen bonds are shown as dashed black lines; (B) (F o(x kGy) -F o(2.2 kGy) )exp(iϕ (2.2 kGy) ) isomorphous difference maps contoured at the +5.5σ (green) and -5.5σ (brown) levels; They highlight the bond breaking process with the concomitant flattening of the residual organic specie and trapping of O 2 as well as the reorganization of the solvent network. The solvent molecule W2 is mostly formed as a result of the partial relocation of W5 and W6. Oxygen and nitrogen atoms are in red and blue, respectively. Broken black lines indicate H-bonds between dioxygen and UOX residues. Distances are in Å. Figure S8. UV/Vis changes upon exposure of UOX:5-PIU crystals to X-rays. The spectrum before X-ray exposure is shown in brown as a dotted line. The spectrum after X-ray exposure (1.5 MGy) is represented by the continuous line. The inset shows the difference spectrum. Following X-ray exposure there is an increase in absorbance in the broad 310-550 nm region exhibiting a maximum at around 360-380 nm. This is consistent with the formation of a resonance-stabilized urate radical species. [17]   were fit to the reaction scheme in panel B using the package Dynafit. [9] Lines represent the dose-dependent occupancies of the different species calculated from the five kinetic constants derived from the fitting procedure; (B) Reaction scheme. The scheme accounts for the recombination reaction between UA• (one-electron oxidation product of UA dianion) and superoxide promoted by the one-electron reduction of the UA•-O2 complex by radiolytically-produced solvated electrons. O* refers to any O 2 decay species (for example OH•) whilst UA* indicates a non-reactive form of the substrate (for example the one-electron oxidation product of Sub•). UA* and UA• are crystallographically indistinguishable. (C) Dose-dependent kinetic constants obtained from the least-squares fitting procedure.  Table S2. Experimental and theoretical peroxide geometries. Selected 5-PMUA and 5-PIU distances and angles are compared to those of 5-PMUA theoretical models calculated at the MP2/6-31+G* level of theory. Four different valence states within the protein environment were considered. 5-PMUA peroxide with a protonated hydroxyl group (charge = -1) is not reported as it is 62 kcal/mol less stable than the isoelectronic 5-PMUA hydoperoxide (charge = -1). Overall, the best geometry agreement between experiment and theory is for 5-PMUA hydoperoxide (charge = -1).