Spin cascade and doming in ferric hemes: Femtosecond X-ray absorption and X-ray emission studies

Significance The structure–function relationship in central to biology, while the structural dynamics are driven by electronic changes. Doming of ferrous heme proteins, which is central to the respiratory function of hemoglobin, ensues from populating high-spin states. However, for ferric heme proteins, doming was excluded. Here, we show that high-spin states are populated in photoexcited ferric cytochrome c, and we present evidence for doming. We also conclude that photo- or thermally activated doming occurs in a wide variety of ferric heme proteins, calling for a deeper understanding of its role in their respective functions.


Sample preparation
The ferric redox state of the sample was prepared from horse heart Cytochrome c (Cyt c), measuring the steady-state UV-Visible (1) and X-ray (2) absorption spectra, which clearly differ from those of ferrous Cyt c. The latter was prepared by adding a molar excess of sodium dithionate (Na2S2O4) to reduce the sample.

Experimental Methods
The X-ray absorption (XAS) and X-ray emission spectroscopy (XES) experiments were carried out at the Alvra end-station of the swiss Free Electron Laser (SwissFEL) at the Paul-Scherrer-Institut (Villigen). (3) The XES experiments were then repeated at the FXE end-station of the European-XFEL (Eu-XFEL, Hamburg). (4) a) SwissFEL measurements: The sample was delivered to the interaction region by a cylindrical liquid jet of 100 µm diameter, running in a closed loop. The jet was pumped into the chamber by an HPLC pump and out by a peristaltic one. Right at the chamber exit, a mobile UV-visible spectrometer provides an on-line monitoring of the sample integrity. No sample damage was observed during the measurements.
The pump pulse was generated by an 800 nm Ti:Sapphire amplified laser system coupled to an optical parametric amplifier, running at 12.5 Hz (half the repetition rate of the X-ray pulses) and yielding ~60 fs/350 nm laser pulses to excite the sample. This pump wavelength was chosen in order to match the penetration depth of the laser to the jet thickness while providing high energy excitation of the haem group. The fluence dependence of the XANES transient signal at 7125 eV and 500 fs time delay after excitation at 350 nm is shown in figure S13 and it exhibits a linear response up to about 20 mJ/cm 2 . The measurements were carried out at a fluence range of 14 -22 mJ/cm 2 at the sample position. Such fluences are comparable to those used in the ps and fs Fe K-edge absorption studies of ferrous myoglobins and ferrous Cyt c.(5-7) The jet speed was such that the sample was completely replenished between consecutive pairs of laser pump/X-ray probe pulses.
The laser and X-ray beams hit the sample in a nearly collinear geometry. The X-ray beam was focused into a 30x40 µm 2 (FWHM) spot by a pair of Kirkpatrick-Baez mirrors. The laser focus was in the order of 100x130 µm 2 (beam waist). Tune-able monochromatic x-rays in the 7.1-7.2 keV range were used for the Fe K-edge absorption measurements, and were generated by scanning the FEL electron beam and the Si(111) monochromator central energies. For the XES experiment, the SASE beam was used and the energy set to 8 keV, away from any Fe shape resonances.
The relative time delay between the laser and x-ray pulses was tuned by a delay stage and scanned in the region of -5 to +20 ps. The instrument response function (IRF) of the experiment was obtained from the rise time of the Fe XANES signal and confirmed by a separate measurement of the rise time of a YAG crystal using an X-ray pump/optical probe combination, (8) in which the X-ray pulses are used to pump a YAG crystal, changing its reflectivity due to the creation of free charge carriers. The transient optical transmission of the 350 nm probe though the YAG is then detected by a diode and used to measure the X-ray/optical cross correlation. The IRF is ~140 ± 30 fs (FWHM), with the uncertainty being mainly due to the intrinsic jitter in arrival time between optical and X-ray pulses.
The XANES spectra were recorded by a Be shielded Si diode in total fluorescence yield (TFY) mode at approximately 90º from the incident X-ray beam and normalized by the incoming Xray pulse intensity (I0), measured by the scattering from a thin target placed about three meters upstream in the beam path. The FEL was running at a repetition rate of 25 Hz, twice that of the pump laser, such that the transient data were obtained by subtracting consecutive laser-on from laser-off shots. The XES data were measured by a spectrometer based on the von Hamos geometry.(9) Two cylindrical Ge(111) crystal analyzers were used to collect the Fe Kα1/Kα2 fluorescence and focus it onto a 4.5 Megapixel 2D Jungfrau detector. The experiments were carried on a 600 mbar He atmosphere to maximize X-ray transmission and minimize X-ray scattering noise.

b) European XFEL measurements:
A very similar set-up to the above was utilized for recording the XES data at FXE. The sample delivery system is essentially identical and the jet thickness was also kept constant (100 µm) between measurements in the different facilities. While an on-line monitoring of the sample was not implemented at European XFEL, aliquots of the ferric Cyt c samples used were collected every couple of hours and their UV-Visible absorption spectra did not show signs of either laser of x-ray damage.
X-ray probe pulses with an energy of 9.3 keV were used. The X-ray delivery pattern consisted of pulse trains arriving at 10 Hz repetition rate, each train containing 128 pulses at a repetition rate of 564 kHz (i.e. 1280 pulses/s). The frequency-doubled output pulses (400 nm) of an amplified custom NOPA based laser system (10) were used to pump the sample. The pump laser was focused down to 90 x160 µm 2 (FWHM) while the x-rays were focused to 10x13 µm 2 (FWHM). The pump fluence was in the order of 9-10 mJ/cm 2 , chosen to maximize the S/N while still within the linear regime, as shown in Figure S14. The IRF was determined by the rise time of a [Fe(bpy)3] 2+ XES pump-probe signal to be ~150 fs (FWHM), mainly due to the intrinsic jitter between the arrival time of the laser and x-ray pulses and the group velocity mismatch between laser and x-ray pulses over the 100 µm sample thickness.
The data were acquired with laser-off and laser-on spectra recorded alternatively at 5 Hz. The XES signal was collected by a 16-crystal von Hamos spectrometer and detected by a Jungfrau 1M detector at 10 Hz, integrating all the signal within a pulse train by the detector. Both Kα and Kβ spectra were acquired simultaneously, such that eight 500 mm radius Si(531) cylindrically bent analyzers were used for the Kβ detection and another eight Ge(111) crystals were used for Kα1/Kα2 lines. While the sample/spectrometer setup were not enclosed in a vacuum chamber, a He bag was placed between the sample and spectrometer to minimize the XES attenuation by air. The following data analysis and treatment were identical to those performed on the SwissFEL results.

Data Collection and Analysis
XANES spectra were acquired at SwissFEL over 4000 FEL shots per energy point (2000 laseron, 2000 laser-off) and normalized by the incoming X-ray pulse intensities on a single shot basis. A threshold was set such that the low intensity X-ray shots, due to the intrinsic large flux fluctuations of monochromatized SASE beams, are discarded in order to improve the S/N of the measurements. Since the laser focus was substantially larger than the X-ray spot size, and presented a high shot-to-shot fluence stability, we assume uniform excitation across the entire probed volume. Three complete XANES spectra were averaged together, resulting in the data presented in this work. The energy axis was calibrated against a reference Cyt c spectrum from the literature ( Figure S2).(2)

Estimating the photolysis yield:
The excitation fraction was estimated based on the reference Cyt c Fe 2+ XANES spectra of ground state (GS) and excited state (ES) provided in ref. (7). By assuming the XANES spectrum at any given time to be a linear combination ES and GS contribution ( ( ) = ( ) * + �1 − ( )� * )), we can estimate the fraction of the excited state species in the pump-probe signal, by minimizing the following expression: At 500 fs, the photoexcitation fraction ≈ 50%. Since at this time delay 25% of the ES population has already decayed, we can extrapolate the excited state fraction immediately after photoexcitation to be in the order of ( = 0) ≈ 67%.
This number can be compared with the expected photoexcitation yield based on the excitation conditions. Considering the laser fluence (22 mJ/cm 2 ), the absorption cross section (1.9x10 4 M -1 cm -1 ) and the number of molecules in the irradiated volume (~10 13 molecules), approximately ~1.35 photons are absorbed per molecule, which would put the photoexcitation yield to 100%.
However, this estimate does not consider losses due to scattering and reflection from the liquid jet, which was round shaped, i.e. such losses are non-negligible but could not be quantified. We therefore believe that the two estimates are in good agreement, i.e. that the X-ray signal reflects the response of the entire population of excited species. The above numbers compare well with the ~3 photons/molecule used in the fs XAS and XAS study of ferrous cytochrome c(7) and the ~2.2 photons/molecule, used in the XAS study of MbCO.(6)

Time Scans
The temporal trace is an average of three individual scans collected over several thousands FEL shots at the maximum of the pump-probe signal at 7125.3 eV ( Figure 2). As mentioned above, the instrument response function (IRF~140 fs) was determined from pump-probe scans on a YAG substrate and it agrees with the rise time obtained by the fit of the XANES traces.
To fit the time traces, a bi-exponential decay function was used: where ( ) is the intensity of the pump-probe signal at a given time-delay, t; 0 is time-zero; is the FWHM of the IRF, 1 , 2 are the pre-exponential factors to the exponential decay time constant with 1 , 2 time constants, respectively and is a constant offset.
The above equation delivers the best fit with the parameters reported in Table S1. A threeexponential fit was also attempted but the first and third components were highly correlated and a very small amplitude was observed for the intermediate one, confirming that the biexponential model is the best suited to describe the results.

Fit of line shapes
The fits of the Kα1 XES data were performed using an asymmetric pseudo-Voigt function, as described in detail in ref. (11). Briefly, the XES line shape ( ( )) can be fit by equation (4) plus an additional linear offset (eq. 6) to correct for non-uniform background.
and = * + The physically relevant parameters are the FWHM ( 0 ) peak position in energy ( ). The asymmetry is introduced with the energy dependence of the FWHM, ( ) , and expressed by the asymmetry factor . It's worth noting that for large asymmetry parameters the fit parameter 0 can differ significantly from the actual FWHM of the line shape,(11) however for the fits presented in this work the numerically calculated FWHM agrees with the extracted fit parameter 0 within less than 2% deviation, due to the small asymmetric parameter of the XES line shape.
The fits were repeated for the Kα1 XES obtained at different time delays and the evolution of the FWHM ( 0 ) as a function of time is displayed on Fig. 3. The other fit parameters did not show any appreciable trend beyond the precision of the fit procedure. MbN3 (14) 2.4 18 Figure S1: Different haem conformations (reproduced from ref. (15)) and their associated orbital structure. ∆/λ is the axial ligand field parameter, with ∆ being the ligand-field splitting and λ the spin-orbit coupling parameter.