Structural basis of the radical pair state in photolyases and cryptochromes

We present the structure of a photoactivated animal (6-4) photolyase in its radical pair state, captured by serial crystallography. We observe how a conserved asparigine moves towards the semiquinone FAD chromophore and stabilizes it by hydrogen bonding. Several amino acids around the final tryptophan radical rearrange, opening it up to the solvent. The structure explains how the protein environment stabilizes the radical pair state, which is crucial for function of (6-4) photolyases and cryptochromes.

The crystals were first grown at 4 • C in hanging drop plates with a reservoir of 100 mM bis-tris pH=6.5, 200 mM lithium sulphate monohydrate, 22 % PEG 3350, 0.5 % Ethyl acetate. After two days, the macrocrystals were crushed and used as seeds for batch crystallization 3 . Prior to collecting data, the crystals were spun down, 90 % of the mother liquor was removed and the content of the vial was mixed in 1:3 ratio with hydroxy ethyl cellulose (HEC) matrix. HEC matrix was produced by dissolving 22 % of HEC in water (w/w).

Data collection
The data were collected at beamline X06SA at the Swiss Light Source Synchrotron (SLS). The sample was extruded through a high viscosity injector with a 75 µm diameter nozzle 4 . The sample flow rate was set to 0.08 µl/min (300 µm/s) for both dark and light steady state data sets. For the light data set, a CW 473 nm laser(Crystalaser CL-473-150) with a fluence of 612 mW/mm2 was employed.
From the flow speed and distance between nozzle tip and x-ray interaction point (90 µm) we estimate that the sample was illuminated for approximately 0.3s before being probed by the x-ray. The sample was prepared and data were collected at room temperature.

Data analysis
The diffraction images were indexed, integrated and merged with Crystfel 0.9.1 5 . Indexamajig was used for indexing patterns with xgandalf 6 . These patterns had indexing ambiguity issues related to the tetragonal Bravais lattice of the crystals. This issue was addressed and corrected with ambigator by using the operator h,k,-l. The resulting stream file was merged, scaled and post-refined with partialator. The partial reflection were treated using the unity model in partialator. The dark state structure is phased using MR (PDB ID: 7AYV), The structure was further refined using Phenix 7 and subsequently modelled with COOT 8 .
Difference maps (Fobsdark-Fobslight) were computed using PHENIX's Isomorphous difference map module 7 and loaded in Coot 8 as reference for residues position refinement for the light structure. The extrapolated structure factors were calculated as follows: (Fe = Fo(dark) + 1/r∆Fo 9,10 .
The extrapolated maps were used to estimate the percentage of activation (the percentage of the protein that get excited by the laser). The mean electron density of the residues that we see changes occurring at and the mean negative features at 3 rmsd were plotted against different level of activa-tion (Fig.S1) 11 . The point at which the two plots intersects gives us the percentage of protein that is activated by light. The activation factor was deemed to be 14 %. The extrapolated map was used to refine the light activated structure in COOT 8 . Further, the calculated difference maps were plotted using the reported methods 12 .
The dark structure has been deposited in PDB with the 7QUT ID, whereas the light structure can be found in the CXIDB (ID 202) .
UV/vis spectroscopy UV/VIS absorption measurements of Dm(6-4)photolyase in solution were performed to investigate which state is reached by illumination. The sample was excited for 100 ms with light from an LED at 453 nm and spectra were recorded 5 ms after the end of the illumination. The results are presented in Fig. 3, panel A shows the raw spectra for illuminated and dark measurement.
Panel B shows the background corrected spectra, calculated from the spectra in panel A. This corresponds to the absorption spectra of Dm(6-4)photolyase before and after illumination. For the dark measurement (blue) we observe a good agreement with the expected absorption profile for a photolyase in the oxidized FAD state and a reduced state in light (red). Panel C shows absorption spectra of the different oxidation states of the FAD taken from literature. 13 Panel D displays the difference absorption spectrum between light and dark (blue) which was computed by subtracting the two spectra in panel A. A matrix division was performed using the spectra in C and this yielded the contribution of the states to the difference scattering (see inset). The resulting difference spectrum is shown in red.
This verifies that after illumination for 100ms the photolyase is in the semi-reduced FAD .− state. The minor negative contribution of FADH is attributed to uncertainties in the measurement. where σt 2 is the variance of the difference between the intensities and their average and σ e 2 is the variance of random error of merged half datasets. 3 The R values are not meaningful because the structure was refined only in real space and only in selected areas. Residues conserved more than 25% are coloured according to clustal color scheme.The asterisks are placed on top of residues which show difference signal in our data.