Extended Data Fig. 3: Comparison of data truncation schemes and changes in retinal over time.
From: Femtosecond-to-millisecond structural changes in a light-driven sodium pump

a, Top, Fo − Fc simulated annealing omit maps of the resting state at 4σ. No truncation (left) shows the map using all data up to 1.6 Å resolution along a*, b* and c*. Spherical truncation (middle) shows the map using all data up to 2.2 Å resolution along a*, b* and c*. Anisotropic truncation (right) shows the map using data up to 2.3 Å, 2.2 Å and 1.6 Å resolution along a*, b* and c*, respectively, as truncated by the staraniso server47. Bottom, Fo(light) − Fo(dark) difference density maps of the 1-ms time delay at the region around retinal and V117 at 3σ. The structure is shown as sticks (salmon, resting state; cyan, 1 ms refined structure). No truncation (left) shows the map using all data up to 1.6 Å resolution. Spherical truncation (middle) shows the map using all data up to 2.2 Å resolution. Anisotropic truncation (right) shows the map using data up to 1.6 Å resolution in c* as truncated by the staraniso server. Overall, the truncated data result in better electron density maps (both for 2Fo − Fc maps and Fo(light) − Fo(dark) difference maps), with finer features being resolved. This effect is probably because noise along the missing directions is removed when compared to no truncation, while retaining the high-resolution data along c* when compared to spherical truncation. b, The evolution of electron density around the retinal chromophore over time. Retinal and K255 of the refined structures are shown as sticks and the electron density is displayed around them (blue mesh, 2σ). Top, original 2Fo − Fc electron density map; panels below show extrapolated 2Fextra − Fc maps. The extrapolated maps allow us to follow retinal isomerization in detail. In the dark state (top), the middle section or the retinal polyene chain is slightly bent downwards. In the picoseconds range, the isomerization is completed and the polyene appears to be straightened. In our ultrafast data, we did not observe retinal with a pronounced twist in the C13=C14 bond as in bR, with retinal in KR2 reaching a near planar 13-cis conformation much earlier along the activation pathway. In the time delays from nanoseconds to milliseconds, the electron density reveals a bend in the retinal molecule resulting from two planes that are twisted against each other. While the exact dihedral angles cannot be refined realistically on the basis of the extrapolated data, the bend seems to originate from the C9=C10–C11=C12 dihedral angle as suggested for the L, M and O intermediates based on time-resolved FTIR18 and resonance Raman spectroscopy32. After 20 ms, a definite conclusion concerning the retinal isomer is difficult. The extrapolated maps suggest that a fraction of the retinal molecules may have already re-isomerized to the all-trans conformation, while it is still bent sideways.