Stepwise Isotope Editing of [FeFe]-Hydrogenases Exposes Cofactor Dynam - PDF document

Stepwise Isotope Editing of [FeFe]-Hydro
Stepwise Isotope Editing of [FeFe]-Hydrogenases Exposes Cofactor Dynamics , Jifu Duan, Florian Wittkamp, Ulf-Peter Apfel, Joachim HeberleMichael Haumann, Sven T. StrippDepartment of Physics, Experimental Molecular Biophysics, Freie Universität Berlin, 14195 Berlin, Department of Physics, Biophysics of Metalloenzymes, Freie Universität Berlin, 14195 Berlin, Department of Biochemistry of Plants, Photobiotechnology, Ruhr-Universität Bochum, 44801 Bochum, Germany Department of Chemistry and Biochemistry, Inorganic Chemistry I, Ruhr-Universität Bochum, 44801 Bochum, GermanyATR-FTIR experiments. IR data evaluation and root-mean-square-deviation (rmsd) calculation. DFT calculations. otein film formation. Optical absorption spectra of HY ATR-FTIR real-time detection of spectral changes in HYDA1 films. Quantitative evaluation of experimental ATR-FTIR spectra. H-cluster model structures used in DFT calculations. ATR-FTIR spectra in the CN-CO. Figure S9: Vibrational couplings in Hox and Hox-CO. Figure S10: Possible H-bonding interactions of the CN- ligands of the H-cluster. Figure S11: Correlation of experimental and calculated CO/CN IR frequencies for H Correlation of experimental and calculated CO/CN IR frequencies for H Experimental CO band frequencies and amplitudes for Hnd intensities for H Correlation of experimental and calculated CO band frequencies for H Correlation of experimental and calculated CO band intensities for H Experimental CO band frequenc

ies and amplitudes for Hnd intensities f
ies and amplitudes for Hnd intensities for HCO) from DFT. nd intensities for HCO) from DFT. Correlation of experimental Correlation of experimental and calculated CO frequencies ( Correlation of exp. and cal. CO frequencies (TPSSh, larger models) for H Correlation of experimental and calculated CO band intensities for Honal modes to IR spectra of H-CO rotamer structures.ATR-FTIR experiments. ATR-FTIR measurements were performed with a Tensor27 spectrometer (Bruker) placed in an anaerobic glovebox and equipped with a mid–IR globar, a liquid-nitrogen cooled MCT detem with two active reflections (Smiths Detection), which was capped by a gas-tight and light-shielPCTFE head-space compartment. For irradiation of samples, the flow compartment was equipped with an acrylic glass window on top of the ATR prism and connected by a light guide fiber to the visible light source (Schott KL1500). The gas flow in the head-space compartment was adjusted and monitored using a multi-channel mass flow controller (Sierra Instruments) at room temperature. Aliquots (2 µL) of concentrated (0.5 mM) [FeFe]-hydrogenase HYDA1, CPI, or DDH protein samples in a buffer solution (e.g., 100 mM Tris/ HCl, 100 mM NaCl, pH 8.0) were pipetted onto the silicon prism of the ATR cell under an Natmosphere in the glovebox, the head-space compartment was closed, and samples were exposed to a constant stream (1L/min) of dry N gas for 5 min to remove most liquid water from the sample so that a thin protein

film was formed on the surface of the A
film was formed on the surface of the ATR prism. Film formation was monitored by the decrease of the relative amplitudes of liquid-water bands and increase of the amide I and amide II protein backbone bands and of the specific CO/CNbands of the H-cluster in successive IR scans during the sample dehydration procedure (Fig. S1). The resulting films contained only a small percentage of liquid water so that their IR spectra were dominated by the protein backbone bands and showed a high signal-to-noise �ratio (100:1) in the spectral region of the CO/CN bands of the H-cluster (Fig. S1). Such films were stable for days and facilitated successive rounds of CO editing experiments and reversible formation of the H-CO states of the H-cluster. For isotope editing experiments, the humidity of the gas stream was varied by passing either a smaller part (50 %, “semi-dry” sample condition) or a larger part (100 %, “humidified” sample condition) of the head-space N carrier gas through a bottle (500 mL) with deionized (Millipore) water. Depending on the ratio between “dry” and “wet” gas, the liquid water content of the protein films varied and for the chosen “wet” condition the relative IR band amplitudes of the CO/CN bands were decreased by about 25 %, which still provided excellent signal-to-noise ratio (Fig. S1). For IR absorption spectra of hydrogenase protein films meas of the empty ATR prism (I/I), the residual spectral background was approximated by smooth spline fu

nctions as defined using the OPUS softwa
nctions as defined using the OPUS software (Bruker) available with the spectrometer and subtracted (Fig. S1). The resulting IR spectra were set to zero and normalized to an area sum of unity in the brations for evaluation and comparison. Deposition of [FeFe]-hydrogenases under an N atmosphere initially yielded films containing the unlabeled H state. Unlabeled H-CO was quantitatively formed within tens of seconds by adding 5 % (50 mbar) CO gas to the N head-space gas stream. CO gas was provided to the sample films at partial pressures of 100-200 mbar. Using the mass flow controller, the composition of the headspace gas stream could be changed within a few seconds, i.e. considerably faster than the reaction of the gases with the hydrogenase films occurred. Irradiation of samples with red or blue light was achieved using band pass filters (center wavelengths 640 nm or 460 nm, for absorption spectra of the band pass filters and of HYDA1 protein see Fig. S2) placed in front of the fiber optics. Real-time formation of H-cluster isotopic species in response to controlled gas exposure, light irradiation, and humidification were monitored by series of rapid IR scans (typically 10–100 scans summed per spectrum, recorded at 80 kHz mirror velocity with a spectral resolution of 1 cmExamples for the formation of selected isotopic species within seconds to minutes in response to varying gas exposures and irradiation treatments as monitored in real-time ATR-FTIR experiments are

shown in Fig. S3. Evaluation of inf
shown in Fig. S3. Evaluation of infrared spectra and root-mean-square-deviation (rmsd) calculation. Experimental ATR-FTIR spectra in the carbonyl stretching vibration region were fitted using a least-squares algorithm and pseudo-Voigt functions (50 % Gaussian and 50 % Lorentzian line shape characters, individual full-width-at-half-maximum (FWHM) values for each band) to determine the CO band center frequencies and relative areas (Fig. S4). The rmsd was used as a quantitative criterion for judging the agreement between experimental and calculated frequencies and amplitudes of infrared bands due to CO/CN stretching vibrations. Experimental (exp) and DFT-calculated (cal) IR bands of CO for H and H-CO species were normalized and calculated IR bands for a given H-cluster model and all of its isotopic patterns were shifted by a common mean offset frequency (Tables S1 and S2), which depended on the structural model (rotamer species and model complexity) and on the DFT level (functional/basis-set combination), for alignment with the mean experimental frequencies determined for all labeled H or H-CO species. Eq. S1 was used for rmsd calculation (     Starting structures for geometry optimization were designed using the crystal structure of CO-inhibited [FeFe]-hydrogenase CPI (PDB entry 1C4C) as a template. Three different H-cluster models with increasing complexity were considered in the calculations (Fig. S5): small structures (~50 atoms) comprising solely the

H–cluster in which the cysteine residue
H–cluster in which the cysteine residues binding [4Fe4S] to the protein are represented by S–CH groups, medium structures (~140 atoms) including appropriate fragments of amino acids surrounding the diiron site, and large structures (~330 atoms) including further amino acids and part of the protein backbone surrounding the whole H-cluster. For the small-size models of the H and CO labeling patterns for all CO/CN rotational isomers were calculated; for the medium- and large-size models selected isotope and isomer species were calculated. Geometry optimizacalculations were performed using Gaussian09 on the Soroban computer cluster of the Freie Universität Berlin. We note that such calculations for the large-size models were a very time consuming task even when using our large-scale computer cluster facilities. Prior to geometry optimization, anti–ferromagnetic coupling was assigned to the iron atoms of the [4Fe4S] sub-complex by proper definition of molecular fragments (broken symmetry approach). During structural relaxation of the small models the positions of the proximal iron atom (Fe) and of the four C atoms of the cysteine CH–groups were kept constant. The positional constraint of the Featom was released for the medium models, whereas for the large models also the cysteine CH2-groups were optimized. The remaining positions of all non–H atoms of the surrounding amino acids were kept constant. For the small models, tight convergenmedium and large models, standard

convergence criteria were chosen in the
convergence criteria were chosen in the geometry optimizations. Vibrational frequency calculations were performed using the relaxed molecular geometries. Calculations were carried out using the BP86/TZVP functional/basis-set combination on all model structures or the TPSSh/TZVP functional/basis-set combination on selected model structures. [FeFe]-hydrogenase HYDA1 protein film formation. FTIR spectra in the liquid water (3500-3000 cm) and amide I and amide II protein backbone (1700-1200 cmregions for increasing spectrometer scan numbers (green to red spectra) after deposition of a drop of liquid protein solution on the silicon prism of the ATR cell with the sample exposed to a stream of dry N gas (“dehydration”). The relative amplitude decrease of the water band and concomitant increase of amide I/II bands indicates progressing water evaporation from the sample, leading to formation of a dehydrated protein film on the surface, selectively nds by the ATR effect. Same spectra as in (A), but in the carbonyl and cyanide stretching vibrations region of the H-cluster, emphasizing CO/CN band enhancement for the film. As-isolated enzyme shows several IR bands due to a mixture of oxidized, CO-inhibited, and reduced H-cluster states, which was converted by extended exposure to humidified N gas to pure H Controlled “rehydration” of a HYDA1 film under a humidified N stream from the “dry” (green) to the “wet” (red) condition. The protein contributions (amide-II bands) to t

he final spectrum are decreased by only
he final spectrum are decreased by only ~25 %. Same spectra as in (C), but in the carbonyl region of the H-cluster. ratio of CO bands is almost not affected and the strein constant during humidification. Note the N-induced conversion of more reduced H-cluster states (e.g. band at ~1891 cm (marker band at ~1940 cm) during humidification. (A)Optical absorption spectra of HYDA1 and band-pass filters. Filters were used for sample irradiation. Spectra were recorded with an USB2000 mini spectrometer equipped with a Mikropack DH2000 UV/vis/NIR light source (Ocean Optics). 2 µL of CO-inhibited HYDA1 (black trace) with a HYDA1 protein concentration of ~1 mM (H-CO state) were probed using a IMPLEN Sub-Microliter Cell. For irradiation a white light source was equipped with band-pass filters showing the indicated absorption characteristics (red light, center wavelength 640 nm, FWHM ~90 nm; blue light, center wavelength 470 nm, FWHM ~80 nm). Spectra were normalized to a maximum absorption of unity in the shown range for comparison. S8 (ii)(iii)(iv)(v)(vi)(viii)2(vii)(A) ATR-FTIR real-time detection of spectral changes in HYDA1 films. IR absorption spectra for increasing scan number(i) Transition from H-CO (12121212) to (121212) under 100 mbar CO gas, time resolution 10 s per spectrum. Transition from H) to (1212CO gas and irradiation at 640 nm (red light), 300 s per spectrum. Transition from H) to (12) under 100 mbar CO gas and irradiation at 470 nm (blue lig

ht), 60 s per spectrum. The IR bands at
ht), 60 s per spectrum. The IR bands at 1972/1960/1921 cm are attributed to transition (ii) that slightly blends into transition (iii), which is characterized by the difference band. (iv) Transition from H) to (12CO, 10 s per spectrum. Transition from H12) to 1212) under 100 mbar CO under irradiation at 640 nm (red), 300 s per spectrum. Transition from H1212) to (12121212) under 100 mbar CO and irradiation at 470 nm (blue), 60 s per spectrum. The IR bands at 2012/1968/1962 cm are attributed to transition (v) that slightly blends into transition (vi), which is ch difference band. Note that the bridging carbonyl (CO) has an extinction coefficient that is typically five times smaller than the one of the equatorial CO ligand at the distal iron ion. and (viii) Transition from H) or (12) to (CO gas and irradiation at 470 nm (blue light), 60 s per spectrum. Exchange of the proximal CO ligand was achieved only in more humidified protein films (“wet” sample Conversion of H-CO to H with 1 L/min N gas flow in the dark. Hformation from H-CO typically was slower than CO exchange and its completion took about 60 min. Reaction scheme for H-CO conversion transitions and isotope editing of CO ligands in response to the selected gas and light treatments (i) – (ix) as detailed Quantitative evaluation of experimental ATR-FTIR spectra. Example for background correction of the raw IR spectrum of unlabeled H-CO (black) using a smooth spline curve (red dashes). Due to the superio

r signal-to-noise ratio of spectra and r
r signal-to-noise ratio of spectra and relatively low background contributions, baseline correctiond did not affect the spectral shape of the IR bands. Simulation of spectra for unlabeled H and H-CO states of HYDA1. Fitting of experimental spectra in the shown carbonyl stretching vibration region using three or four pseudo-Voigt functions (see Methods, fit program facilitating simultaneous fitting of all CO bands in a spectrum with variable FWHM values; courtesy of Dr. P. Chernev, FU-Berlin) resulted in experimental CO band center frequencies with an , i.e. at the spectral resolution limit) and relative band areas (relative error ~10 % for individual CO bands), which were used for comparison with respective DFT-calculated values. Note low noise levels of the experimental spectra, negligible contributions from background and H-cluster states other than the ones of interest, H-cluster model structures used in the DFT calculations. Shown are examples of geometry-optimized structures for the BP86/TZVP functional/basis-set combination. Similar structures were used when applying the TPSSh/TZVP DFT approach. Color code: Fe, light blue, S, yellow; O, red; N, blue; C, grey; H, white. Small-size model structures comprising the H-cluster and its (truncated) cysteine ligands for three CO/CN configurations of the H-CO state (a, ”standard” geometry; b, CO/CN exchange at the distal iron atom; c, apical CN at the distal iron atom). Medium-size model for the H state including

(truncated) amino acid residues closest
(truncated) amino acid residues closest to the H-cluster. Large-size model for the Hincluding (truncated) amino acid residues at larger distances to the H-cluster. For conversion of H into H-CO structures, a CO ligand was added at the proximal iron ion. Coordinates of all calculated model structures in xyz-format are given in a separate text file in Supporting Information. ATR-FTIR spectra in the CN region for CO editing of H Spectra for spectra for CPI. Spectra in (A) show the lower-frequency CN band assigned to the cyanide at the distal iron ion in magnification, spectra in (B) show the bands of the proximal and distal CN ligands. Spectra were recorded at 1 cm resolution. Indicated carbonyl labeling patterns show the CO ligands in the order . Note that the frequency shifts of the CN bands are smaller than 1 cm even for CO editing of up to three of the four carbonyl ligands in H-CO, which indicates that the CN ligands are fully uncoupled from the vibrations of the CO ligands and therefore not decisive for the carbonyl isotopic labeling ) frequency differences between the CN bands of the two enzymes may reflect minor structural differences in the protein environment of the H-cluster. The small frequency shifts of the CN bands upon CO editing of H and HCO were reproduced by the DFT calculations, which showed shifts less than 2 cm of both CN bands for all isotopic species of a given model structure (A)(B)Uncoupled vibrational behavior of the bridging carbonyl

ligand (CO). Selected IR spectra of HYDA
ligand (CO). Selected IR spectra of HYDA1 were stacked to emphasize that CO editing of the bridging carbonyl in the 1808-1762 cm frequency range) does only marginally affect the higher band frequencies of the terminal carbonyls (bands ) at the proximal and distal iron ions and vice versa, which indicates that the bridging CO is fully uncoupled from the vibrations of the other carbonyl ligands. Shown are spectra for H-CO (left) with a CO (black) or (red) and with a CO ligand or with a CO ligand and spectra for H (right) with with a 2040198019201860180017401895192019051955176218022040198019201860180017401924193019211918191819421921196419681991201220061768180820401980192018601800174017681808197219621968196419641960199119421927192120062012(A)204019801920186018001740(B)17621802195519051964194017621802Isotope editing of bacterIsotope editing of bacter(A) CPI and DDH. Normalized spectra correspond to the following H-CO isotopic species of the H-cluster CO and CO (bold) ligands are given in the order ): black, (12 12 12 12); red, (12 ); blue, (12 ); magenta, (12 13 13 12). Note near-quantitative formation of respective isotopic species in the bacterial enzymes, similar to the HYDA1 enzyme from the green alga, comparable shifts of the carbonyl bands upon CO exchange in the bacterial and algal hydrogenases, and the small differences in CO band frequencies not exceeding 8 cmbetween CPI and DDH, which also were similar 20401980192018601800DDHfrequency / cm

197119471963A=102040198019201860180017
197119471963A=10204019801920186018001740CPIfrequency / cmA=10Vibrational couplings in and H-CO. Shown is the diiron site of the small-size model from DFT (BP86/TZVP) for H (A, B) and H-CO (C) “standard” structures and -CO with a distal CN in apical position (D). Approximate relative contributions and symmetry characters of vibrations of the individual CO ligands to the observed IR bands ( ) are denoted by lengths and colors of arrows. (A) HCO or CO in Fig. 2A or 2B). CO ligands show only minimal vibrational coupling. Band can be attributed to CO whereas band represents CO) is vibrationally isolated. CO, proximal CO (species and in Fig. 2B). Note the similar contributions of CO and CO to bands and due to extensive vibrational coupling, which CO substitution at the proximal carbonyl. (C) In the H“standard” geometry, bands and show equal contributions from symmetric and asymmetric vibrations of equatorial CO ligands at proximal and distal iron ions. (D) Only the -CO rotamer with a distal apical CNCN) reproduces the experimentally observed uncoupling of the proximal CO (band ) from the other stretching vibrations. (A)(B)(D) Possible hydrogen-bonding interactions of the CN ligands of the H-cluster. The “standard” configuration of the diatomic ligands is shown as modeled in crystal structures of CPI [FeFe]-hydrogenase attributed to oxidized enzyme. Potential H-bonds are marked by dashed lines. In this model (amino acid numbering for HYDA1),

CN may interact with Pro231 (NH) and Ser
CN may interact with Pro231 (NH) and Ser232 (OH) side chains and the respective backbone nitrogen (NH) and CN may interact with backbone nitrogen (NH) or carbon (CH) groups of Ser323, Pro324, and Gln325 and with the Lys358 side chain (NH). All potential H-bonding partners are at distances ligands. Correlation of experimental and calculated CO and CN IR frequencies for H Data (points) stem from Tables S1A (experimental, F) and S1B (calculated, F); lines show linear regressions (F = y-intercept + slope x F) for the indicated DFT structures (std = “standard” model; d-inv, p-inv, p/d-inv = CO/CN inversion at Fe, or both; r-CN, r-CO = distal CO/CN rotated to an apical position with/without inversion); calculated data refer to the BP86 functional and small model if not indicated otherwise (med = medium model, lar = large model). Parameters for fit curves in (A). Similar fit qualities, i.e. a mean R-value of 0.987(9) were obtained for 6 out of 7 rotamer structures (e.g. for the std and r-CN models), for the three different model sizes (Fig. S5), and for pure or hybrid DFT functionals. Such structures are not clearly distinguished on basis of the IR frequencies. orientation of carbonyls) was slightly disfavored; r-CO showed larger deviations and was clearly disfavored. A mean fit slope of 1.12(3), i.e. slightly larger than 1 for ideal correlation (black line), was observed, which we attribute to systematic limitations of the DFT approach. The mean calculated CO/CN

frequency for the BP86 models (~1945 )
frequency for the BP86 models (~1945 ) deviated from the experimental value (~1961 cm) by ~16 cm (~54 cm for TPSSh), corresponding to a relative accuracy of ~1 % (~3 %), which we consider as good agreement. Correlation of experimental and calculated CO and CN IR frequencies for H Data (points) stem from Table S2A (experimental, F) and S2B and S2C (calculated, F); lines show linear regressions (F = y-intercept + slope x F) for the ” model; d-inv, p-inv, p/d-inv = CO/CN inversion at , or both; a-CN = distal CO/CN in apical position with/without inversion); calculated data refer to the BP86 functional and small model if not indicated otherwise (T = TPSSh, med = medium model, lar = large model). Parameters for fit curves in (A). A significantly increased fit quality was observed for the models with the distal CN ligand in apical position (mean R = 0.98(2)) compared to the “standard” model (mean R = 0.92(3)) for all three model sizes and for both DFT functionals. This supports the rmsd analysis. A mean fit slope of 1.13(5), i.e. slightly larger than 1 for ideal correlation (black line), was observed, which we attribute to systematic limitations of the DFT approach. The mean calculated frequency for the BP86 models (~1959 cm) deviated from the experimental value (~1973 cm) by ~14 cm (~51 cm for TPSSh), corresponding to (~3 %), which we consider as good agreement. Experimental CO band frequencies and amplitudes for HData correspond to experimental IR spectra in

Fig. 2 and were derived from least-squar
Fig. 2 and were derived from least-squares curve fitting of spectra. The frequency error is less than 1 cm, intensities (error ±5 %) correspond to integral areas of respective CO bands and were normalized to a sum of 100 % (spectrum) CO labeling p µ d d-1] intensity [%] 2A (i) 121212 1940 60 1964 21 2A (x) 121312 1940 66 1964 19 2A (iv) 121213 1906 62 1956 21 2A (vii) 121313 1905 72 1955 17 2B (vii) 131212 1905 35 1955 49 2B (iv) 131312 1905 33 1955 48 2B (x) 131213 1895 62 1920 22 2B (i) 131313 1895 63 1920 23 from DFT.structurepµd3333.01940.3734.52073.0352.93302.81939.6694.22073.0351.33167.91938.6833.02072.9336.83097.31938.0807.62072.9335.52106.51911.81949.72073.0353.82066.81910.51904.62073.0352.33212.11895.9777.92072.9337.63172.71894.6706.52072.9336.32178.21939.11167.52073.3356.62175.51938.51124.62073.2355.02271.31936.91010.02073.2341.52237.01936.4982.02073.1340.21161.61911.52159.92073.2359.51150.21910.32117.82073.2358.02077.41894.51177.62073.1344.52079.21893.31106.82073.1343.12638.21915.34202.22070.8659.92674.21914.64030.82070.8658.14076.91911.72684.32070.7634.53996.31911.22574.22070.7633.21192.81903.44033.02072.5415.91117.21898.55530.72070.8656.82456.11871.24256.22070.7633.72526.01869.83980.62070.7631.82645.71926.41424.42071.1496.62653.91925.21330.32071.1493.62598.61921.21422.62071.0475.62548.31920.41364.42071.0473.01831.91914.32238.82071.1494.31667.81910.42548.82070.9525.52498.01882.81521.12071.047

3.32515.01880.81367.22070.9470.82718.4
3.32515.01880.81367.22070.9470.82718.41943.1837.32074.1278.92686.01942.4798.02074.1277.92608.81941.8906.42074.0268.72538.91941.2878.42074.0267.91760.41913.71785.22074.1279.91730.11912.31729.12074.1278.92634.11898.5870.12074.0269.72591.51897.3800.92074.0268.83328.31943.8869.82079.9349.53433.11943.0842.92079.8347.82741.41943.0931.42079.7332.63116.21942.4905.72079.7331.13358.31902.2843.42079.9352.53497.11900.3756.62079.8350.82791.91899.3877.02079.7335.33175.51898.0825.22079.7333.81476.01923.32754.72073.0278.21504.11921.42682.12072.6276.42206.31917.71991.32072.5265.82421.11915.02074.72071.6288.0990.11907.03225.12072.6279.0989.01905.53144.42072.6277.81351.21878.82795.82072.5267.21413.81876.82637.52072.5266.23478.01955.2631.32077.1815.12632.51959.5634.32085.8503.32864.72015.1789.62161.1161.32860.32014.3755.32161.1159.12573.92013.2960.72161.0155.92569.72012.7934.52161.0153.9CO-lab = 12/13CO pattern; data for BP86 functional and small model if not indicated otherwise, for DFT model structure annotations see Fig. S11. Frequencies are given in cmintensities are given in arbitrary units. Correlation (rmsd) of experimental and (spectrum)invinvCN/pinvmediumstdstd(i)121212121212658121312121213121313131212131312131213131313(x)121312121212121312668121213121313131212131312131213131313(iv)1212131212121213121212137121313131212131312131213131313(vii)1213131212121213121212131213137131212131312131213131313(vii)131

21212121212131212121312131313121298
21212121212131212121312131313121298131312131213131313(iv)1313121212121213121212131213131312121313129910131213131313(x)131213121212121312121213121313131212131312131213557131313(i)1313131212121213121212131213131312121313121312131313135683538303825rmsd (Eq. S1) values (in cm) were calculated with CO frequencies in Tables S1A (exp.) and S1B (calculated) and comparing all calculated with experimental isotope patterns; for DFT model annotations prior to rmsd calculation. Correlation (rmsd) of experimental and (spectrum) exp & cal CO pattern (p µ d) std d-invp-inv p/d-inv r-CO r-CN r-CN/p-inv 121212 7 6 24 6 3 3 26 121213 4 4 10 7 8 1 15 121313 4 12 15 13 9 8 20 121312 3 9 26 8 3 2 29 2B (vii) 131212 7 5 14 2 25 5 13 2B (iv) 131312 8 4 18 3 28 6 12 131213 5 8 26 8 10 1 29 131313 5 8 24 6 5 3 27 rmsd (Eq. S1) values (in %) were calculated with CO band intensities (= integrated band areas) in Tables S1A (experimental) and S1B (calculated, values for BP86 functional and small model structures) and comparing the indicated specific calculated and experimental isotope patterns; for DFT model annotations see Fig. S11. Calculated amplitudes were normalized to a sum of 100 % for the three CO bands of H prior to rmsd calculation. Larger rmsd values for the model structure with proximal CO/CN inversion (p-inv), leading to orientation of the proximal and distal carbonyls, and large rmsd values for structure r-CN/p-in

v disfavored equatorial CO/CN inversion
v disfavored equatorial CO/CN inversion at Fe. The increased rmsd values for the structure with rotated proximal CO (r-CO) in particular for a ligand arrangement. Similar rmsd values for the “standard” structure and the model with the distal CN rotated to a more apical position (r-CN) showed that these structures are similarly in agreement with the experiment and therefore equally likely according to the IR intensities. Experimental CO band frequencies and amplitudes for H(spectrum) CO labeling pattern -1] intensity[%] (spectrum) pattern -1] intensity [%] 12121212 1809 2913121212 1809 181963 171963 151969 221969 242013 312013 4212131212 1769 2713131212 1769 261962 221962 151968 191968 232012 322012 3612121312 1808 1913121312 1808 181928 341928 161965 211965 342006 262006 3212121213 1808 3013121213 1808 281942 151942 141964 191964 141992 361992 4512131312 1768 2413131312 1768 191928 341928 161965 171965 352006 242006 3012131213 1769 2313131213 1769 241943 141943 141964 201964 171991 431991 4512121313 1808 3113121313 1808 191921 241921 191960 321960 271973 131973 3512131313 1768 2813131313 1768 181921 261921 181960 341960 291972 131972 34Data correspond to experimental IR spectra in Fig. 2 and were derived from least-squares curve fitting of spectra. The frequency error is less than 1 cm, relative intensities (error ±5 %) correspond to integral areas of respective CO bands and were normalized to a sum of 100 CO) from DFT.st

ructurefreqfreqfreqfreqfreqfreq1212121
ructurefreqfreqfreqfreqfreqfreq121212121797.91326.21891.71204.31945.12440.31960.31618.42095.2199.32106.8322.7121312121757.21319.31891.71207.51945.02419.51959.51559.62095.1196.42106.7318.9121213121797.61231.21875.31265.71922.73361.71953.7626.52095.1196.62106.6300.2121212131797.81331.71852.81327.91942.62612.51957.31243.22095.1199.92106.6311.7121313121757.01268.91875.21230.31922.53323.81952.9578.62095.1193.82106.5296.8121312131757.11319.91852.71340.81942.62611.11956.51160.42095.1197.12106.6308.1121213131797.51243.21847.51120.41906.73337.11953.0708.92095.1197.22106.4290.7121313131757.01271.11847.51129.61906.43255.81952.3669.62095.1194.52106.4287.4121212121795.91249.11893.61186.01926.71267.61956.73470.32105.280.82109.1325.6121312121755.51259.11893.61187.21926.21257.71956.13382.92105.280.42109.1321.3121213121795.61174.91874.81145.11918.72342.31938.02404.42105.280.42108.9305.6121212131795.71252.11855.41333.81926.11023.51951.13489.22105.280.02109.0316.4121313121755.31217.31874.81127.01918.62337.21936.92298.82105.280.12108.9301.7121312131755.41259.11855.31340.81925.71027.31950.43385.32105.279.62109.0312.2121213131795.51183.41849.41096.01903.42493.01934.22218.72105.279.62108.8297.7121313131755.21218.91849.41102.61903.32484.01933.12099.02105.279.32108.8293.8121212121799.11416.71892.61253.51941.81350.61961.12088.12096.7151.12105.0442.0121312121758.51392.21892.61255.91941.61335.21960.42040.62096.7149.72105.0436.4121213121798.81348.91876.11080.31922.92710.51

951.0889.12096.7151.92104.8420.11212121
951.0889.12096.7151.92104.8420.1121212131799.01416.21853.81403.21940.01296.71957.41920.82096.7151.62104.8429.7121313121758.31355.41876.11060.11922.82697.01950.2830.92096.7150.42104.8415.0121312131758.41390.81853.71408.61939.91289.41956.61863.12096.7150.22104.8424.3121213131798.71354.31848.51170.01907.22497.71949.9933.42096.7152.42104.7409.2121313131758.21355.41848.41177.31906.92452.71949.1884.22096.7150.92104.6404.2121212121771.01366.91923.01071.41942.41036.11975.93124.12088.0289.22094.9208.9121312121730.81347.91923.01068.61942.11035.21975.53058.12088.0288.32094.8205.0121212131770.91318.41891.41629.41940.9910.81963.52642.02087.8266.32094.8201.6121213121770.91339.91891.51069.91941.61375.41962.72740.92087.8266.72094.8208.5121312131730.61318.51891.31606.01940.7911.21963.02575.72087.8265.62094.8198.1121313121730.71330.61891.51067.11941.41364.21962.22675.12087.8266.12094.8204.9121213131770.71295.11878.01051.11923.22785.01948.91293.02087.6247.12094.8202.1121313131730.61302.61878.01049.81923.12768.01948.21214.42087.6246.62094.8198.8121212121797.81249.71894.41292.11931.61809.31960.62321.32103.3270.22105.1240.0121312121757.41263.31894.41294.21931.11765.41960.02265.22103.3270.92105.1233.3121213121797.61183.21879.41106.31922.02844.41939.71460.72103.2280.22105.0206.1121212131797.71250.81855.01413.31931.41773.41955.82162.82103.2272.82105.0224.8121313121757.21226.61879.31086.81922.02845.91938.61350.42103.2280.12104.9200.7121312131757.31262.41854.91420.51931.0

1736.71955.22096.42103.2273.32105.0218.6
1736.71955.22096.42103.2273.32105.0218.6121213131797.41189.61850.21191.81909.62679.21936.81460.32103.1281.12104.9194.1121313131757.11227.11850.21200.11909.42648.51935.71360.12103.1281.12104.9188.8121212121770.41244.01921.9391.51926.52264.71976.93289.92088.2286.62101.498.0121312121730.41248.51921.7389.71926.22248.31976.53214.72088.1285.32101.496.1121212131770.31206.51888.71320.61925.41854.11965.12711.12087.9261.22101.497.9121213121770.31214.21895.81215.41923.71138.21960.23551.92087.9263.42101.496.1121312131730.21224.71888.61316.81924.91823.51965.32637.82087.9260.32101.496.1121313121730.21229.71895.71203.31923.21139.21959.73458.12087.9262.42101.494.3121213131770.11179.91878.61043.31917.81726.51937.63067.92087.7241.82101.496.2121313131730.21207.11878.61043.51917.71726.91936.82950.42087.6241.22101.494.5121212121809.11354.81891.11016.91951.12962.21972.4598.02075.5568.12099.6553.3121212121775.81398.11908.41007.61958.21941.11971.51953.32073.6621.32092.9227.4large121212121824.21364.31906.9973.91958.42872.91972.8633.32097.7417.82114.6534.1large121212121785.3572.31921.2518.11962.6942.31975.01027.32092.1251.92102.8173.3121212121845.01204.11968.81239.42018.23060.72037.8879.82167.5170.82178.0248.1121312121803.41186.41968.81239.52018.23060.22037.0826.32167.5169.22178.0245.5121213121844.71133.71949.91379.41996.83231.42031.7531.72167.5168.52177.9234.0121212131844.91213.91929.41330.72015.03009.62035.0752.12167.5172.02177.9241.0121313121803.21146.51949.71344.5199

6.73205.42031.0508.32167.4166.92177.9231
6.73205.42031.0508.32167.4166.92177.9231.7121312131803.31190.51929.31342.42015.03008.82034.2693.22167.4170.42177.9238.5121213131844.61145.21923.61157.61978.63195.52030.7699.62167.5169.72177.8227.9121313131803.11151.21923.61158.41978.33134.32030.0681.82167.4168.12177.8225.6121212121818.21166.31997.81115.12015.21576.42049.12481.62160.7218.32168.1154.3121312121777.11147.41997.71112.82014.91574.92048.52431.52160.7217.92168.0152.2121212131818.01130.21964.51673.02013.01337.42037.92097.32160.5204.22168.0150.2121213121818.11144.91964.81132.62013.82047.72036.81932.72160.5203.52168.0156.2121312131777.01124.71964.41653.72012.91339.32037.32047.12160.5203.92168.0148.3121313121777.01133.31964.71130.52013.72041.32036.11879.62160.5203.32168.0154.2121213131817.91110.91951.71104.31992.32781.72024.81157.02160.4191.42168.0152.5121313131776.91111.41951.71103.51992.12759.52024.11106.02160.4191.32168.0150.7CO-lab = 12/13CO pattern; data for BP86 functional and small model if not stated otherwise, for DFT , intensities in arbitrary units. structurebandsbandsintintintintintfint131212121797.01291.01891.51327.81905.7903.41954.93030.92095.0196.72106.7321.3131312121756.71301.41891.51329.61904.8866.71954.52972.22095.0194.12106.7317.6131212131796.71200.21875.21234.11903.81630.91927.72383.92095.0194.02106.5298.9131213121796.91294.21852.71375.31905.6932.01949.42878.02095.0197.32106.6310.2131312131756.51251.91875.11208.61903.21556.31926.92348.42095.0191.42106.5295.71313131217

56.61301.41852.61381.81904.8903.51949.02
56.61301.41852.61381.81904.8903.51949.02809.42095.0194.62106.6306.7131213131796.61209.61847.51141.01899.32370.91915.51652.52095.0194.62106.4289.3131313131756.51253.51847.51146.81899.22343.91914.11545.72094.9192.02106.4286.2inv131212121794.01217.61884.01486.61894.01004.21954.03308.12105.177.52109.0321.5131312121754.81217.61883.51486.61893.61004.21954.03308.12105.177.52109.0321.5131212131794.21107.21873.8543.01885.42292.51928.23053.72105.277.32108.9306.0131213121794.31170.01855.21464.71885.11105.91948.13288.12105.176.92109.0316.4131312131754.61178.91873.7531.01884.12225.41927.62974.52105.177.12108.9302.2131313121754.71217.01855.21458.71883.71071.51947.73195.02105.176.72108.9312.3131213131794.11111.91849.41127.41882.01222.81911.53458.82105.276.62108.8297.9131313131754.61179.71849.41128.71881.21207.41910.43318.12105.176.32108.8294.2inv131212121798.21383.01892.51444.31901.7643.01956.52600.52096.6149.62105.0441.7131312121757.91375.31892.51451.51900.8614.11956.12544.72096.5148.22104.9436.3131212131797.91321.11875.9946.01900.31417.31929.02306.12096.6150.32104.7419.8131213121798.01379.91853.61463.81901.7683.01951.02472.42096.6150.02104.8429.3131312131757.81340.11875.9941.81899.61361.51928.32261.92096.5149.02104.7415.0131313121757.91373.31853.61461.51900.8672.61950.62406.32096.5148.62104.8424.1131213131797.81323.61848.41192.11896.21326.31916.12075.42096.6150.82104.6408.9131313131757.71339.51848.41195.71896.01305.31914.81991.32096.5149.42104.6404.113121212

1770.41346.11898.9788.61923.11166.61974.
1770.41346.11898.9788.61923.11166.61974.43258.52088.0289.92094.7206.5131312121730.41336.11898.2775.11923.11166.51974.13193.22088.0289.02094.6202.9131212131770.21298.51890.71785.51899.5561.11960.62816.72087.8267.02094.6199.3131213121770.21320.51889.8750.91900.91263.91960.03152.02087.8267.52094.6206.2131312131730.31306.91890.61784.01898.8531.91960.32749.62087.8266.32094.6196.1131313121730.31319.31889.8745.71900.31251.31959.73081.72087.8266.82094.6202.8131213131770.11276.51878.01019.31897.1992.81930.03096.82087.6247.82094.6199.9131313131730.21291.51878.01014.91896.7990.91929.42998.22087.6247.32094.6196.8131212121796.51157.91888.11978.51895.1941.61959.12527.52103.2264.72105.0239.2131312121756.71216.21886.91838.41895.0992.51958.72473.82103.2264.92105.0233.3131212131796.31096.51879.11009.61888.11949.51929.62471.72103.2275.42104.9204.7131213121796.41157.41854.71491.41889.11528.61954.12356.02103.2267.62104.9223.6131312131756.51180.71879.11008.81886.91835.41928.92417.32103.2275.32104.8199.7131313121756.61214.81854.71491.51887.81446.41953.72295.92103.2267.52104.9218.3131213131796.31096.61879.11009.81888.11949.31929.52471.52103.2275.82104.9204.3131313131756.51180.71850.21233.71885.91716.71914.32237.42103.1276.42104.8187.6CN/pinv131212121769.51184.41880.81408.11923.91292.21976.03231.52088.1287.22101.394.5131312121729.81214.71879.91361.41923.91287.41975.73163.62088.1285.92101.392.7131212131769.31150.91878.4748.51892.12429.61964.72688.92087.9261.92101.394.513

1213121769.31155.51880.41565.61896.1943.
1213121769.31155.51880.41565.61896.1943.61958.53380.92087.9263.92101.392.6131312131729.71192.21877.8778.41891.72332.81964.42625.12087.9261.02101.392.9131313121729.71196.11879.51505.61896.0954.91958.23299.62087.9262.92101.391.0131213131769.21124.91877.0385.71881.62163.11930.93269.52087.6242.52101.392.9131313131729.61174.81876.7379.31881.12142.61930.33156.72087.6241.82101.391.4CO-lab = 12/13CO pattern; data for BP86 functional and small model, for DFT model structure Correlation (rmsd) of exp. and cal. CO band frequencies ((spectr.)invinvpIdinvCN/inv(spectr.)invinvinvCN/inv(xii)1212121212121212(viii)1213131212121212121312121213121212121312121213121212121312121213121313121213131212131213121312131212131312121313121313131213131313121212131212121313121213131212131213121312131213121213131212131313131213131312131312131313121313121313131213131313131313131313(ix)1213121212121212(xi)121312131212121212131212121312121212131212121312121212131212121312131312121313121213121312131213281212131312121313121313131213131313121212131212121313121213131212131213121312131213121213131212131313131213131312131312131313121313121313131213131313131313131313(v)1212131212121212(iii)1212131312121212121312121213121212121312121213121212121312121213121313121213131212131213121312131212131312121313121313131213131313121212131212121313121213131212131213121312131213121213131212131313131213131312131312131313121313121313131213131313131313131313(ii)12

12121312121212(vi)121313131212121212131
12121312121212(vi)1213131312121212121312121213121212121312121213121212121312121213121313121213131212131213121312131212131312121313121313131213131313121212131212121313121213131212131213121312131213121213131212131313131213131312131312131313121313121313131213131313131313131313rmsd (Eq. S1) values (in cm) were calculated with CO frequencies in Tables S2A (exp) and S2B (cal, (BP86, small model) and comparing all cal. with exp. isotope patterns; DFT model annotations in ) applied to calculated data prior to rmsd calculation. Correlation (rmsd) of exp. and cal. CO band frequencies ((spectr.)invinvinvCN/inv(spectr.)invinvinvCN/inv(vi)1312121212121212(ii)1313131212121212121312121213121212121312121213121212121312121213121313121213131212131213121312131212131312121313121313131213131313121212131212121313121213131212131213121312131213121213131212131313131213131312131312131313121313121313131213131313131313131313(iii)1313121212121212(v)1313121312121212121312121213121212121312121213121212121312121213121313121213131212131213121312131212131312121313121313131213131313121212131212121313121213131212131213121312131213121213131212131313131213131312131312131313121313121313131213131313131313131313(xi)1312131212121212(ix)1312131312121212121312121213121212121312121213121212121312121213121313121213131212131213121312131212131312121313121313131213131313121212131212121313121213131212131213121312131213121213131212131313131213131312

1313121313131213131213131312131313131
131312131313121313121313131213131313131313131313(viii)1312121312121212(xii)1313131312121212121312121213121212121312121213121212121312121213121313121213131212131213121312131212131312121313121313131213131313121212131212121313121213131212131213121312131213121213131212131313131213131312131312131313121313121313131213131313131313131313rmsd (Eq. S1) values (in cm) were calculated with CO frequencies in Tables S2A (exp) and S2C (cal, (BP86, small model) and comparing all cal. with exp. isotope patterns; DFT model annotations in ) applied to calculated data prior to rmsd calculation. rmsd of exp. and cal. CO frequencies (TPSSh, larger models) for H-CO.1627141212121251212121212121212612121212offsetoffsetrmsd (Eq. S1) values (in cm) were calculated with CO frequencies in Tables S2A (exp) and S2B (cal) and comparing selected calculated with experimental isotope patterns; for DFT model annotations see Fig S12, cal-offset = frequency shift (in cm) applied to calculated data prior to rmsd calculation. Data for medium (med) and large (lar) model structures were Correlation (rmsd) of exp. and cal. CO band intensities for H(spectrum) exp & cal CO pattern (p µ d4376(xi)8626869758(xi)8895951rmsd (Eq. S1) values (in %) were calculated with CO band intensities (= integrated band areas) in Tables S2A (experimental) and S2B anand small model structures) and comparing the indicated specific calculated and experimental isotope patterns;

for DFT model annotations see Fig. S12.
for DFT model annotations see Fig. S12. Calculated amplitudes were normalized to a sum of 100 % for the four CO bands of H-CO prior to rmsd calculation. The smaller mean rmsd value (~10 % vs. ~15 %) for the structure with an apical CN at Fe (a-CN) compared to the “standard” structure favored the former configuration. We note that due to a considerably larger error in the CO band intensity (area) determination compared to the frequency determination, the intensities are less sensitive and hence less indicative of ligation onal modes to IR spectra of Hcarbonyl band apical CNcis 12121212 – – vs – – – vs – m m – – m w – – sym – – – s m w – m sym m m – – w m – m 121212 – – vs – – – vs – m w – – vs – – – sym – w – s – m – m sym w m – w – m – m – – vs – – – vs – w m – – m m – – sym w – – s m m – – sym m w – w – – – s 1313 – – vs – – – vs – m m – – m w – – sym w w – w w s – – sym – – – s – – – s structures with an apical and CO cis to CO and “standard” structures with an apical CO (Fig. 4). Data are for the four principal CO labeling patterns of H-CO (left) with a CO ligand at the proximal iron ion and isotopic exchange(s) at the indicated positions (CO ligands are given in the order ). Relative vibrational contributions to the IR are denoted: vs = very strong, s = strong, m = medium, w = weak. Symmetric (sym) and asymmetric (asym) contributions from the CO stretching vibrations of the individual carbonyls in apical or equatorial (equat) posit

ions at the proximal () iron ions are di
ions at the proximal () iron ions are distinguished from the vibrational mode of the bridging (uncoupled from the other ligands. -CO rotamer structures. parameter DFT model rmsd rmsd (int.) (int.) int. ratio int. invers.c 8 (+) 0.99 (++) 5 (++) 0.94 (++) yes (+) yes (++) 10+ 95 favored r-CN 9 (+) 0.99 (++) 4 (++) 0.95 (++) yes (+) yes (++) 10+ 95 favored 8 (+) 0.99 (++) 7 (+) 0.93 (+) yes (+) yes (++) 8+ 86 less likely 4 (++) 0.99 (++) 7 (+) 0.90 (-) yes (+) yes (++) 7+ 82 less likely 7 (+) 0.98 (+) 20 (--) 0.17 (--) no (-) yes (+) 2- 41 unlikely r-CN/p-inv 9 (+) 0.97 (-) 21 (--) 0.06 (--) no (-) yes (+) 4- 32 excluded 23 (--) 0.96 (-) 12 (-) 0.53 (-) yes (+) no (--) 6- 23 excluded -CO parameter sum DFT model rmsd rmsd (int.) (int.) int. ratio int. invers.c 7 (+) 0.99 (++) 8 (++) 0.29 (+) yes (+) yes (++) 9+ 91 favored 22 (--) 0.93 (--) 11 (+) 0.27 (+) yes (+) yes (++) 1+ 55 unlikely 22 (--) 0.93 (--) 9 (++) 0.40 (+) no (-) yes (+) 1- 45 unlikely 3 (++) 0.99 (++) 13 (-) 0.20 (-) yes (+) no (--) 3- 36 unlikely 23 (--) 0.94 (--) 11 (+) 0.21 (-) no (-) yes (+) 4- 32 excluded 23 (--) 0.94 (--) 14 (-) 0.06 (--) no (-) yes (+) 7- 18 excluded Shown data summarize results for BP86 calculations on small model structures in Figs. S11 and S12 and Tables S1 and S2. For DFT model annotations see Parameters: mean rmsd values (in cm or %) were calculated using Eq. S1 for IR band over rmsd values for all isotopic species), R-values correspond to fit qual

ities derived from linear fits to plots
ities derived from linear fits to plots of calculated vs. experimental frequencies (including CO and CN bands) or intensities (including only CO This parameter refers to the respective CO band intensity ratios of H band larger band, except for isotopic species (1212, in the experimental IR spectra) or Hp µ , 12131313, in the experimental IR spectra), which was reproduced (yes) or not (no) by the DFT approach. This parameter shows whether the inverted (invers.) intensity ratio of the and bands in H species and or of the and bands in H-CO species and was reproduced by the DFT approach. The sum adds the (+) or (–) rankings (in parenthesis) of all parameters. The probability (prob.) for each structure was calculated assuming that the maximal possible parameter sum range from 11- to 11+ corresponds to a probability span from 0-100 %. Ratings were given accordingly. For H, the two rotamers with “standard” CO/CN orientation or a distal more apical CN are equally likely and slightly or strongly favored over the other rotamer structures. For H-CO, the rotamer with an apical CO ranks at lowest probability. This holds also for data from TPSSh calculations and both larger model structures. Mutual comparison of parameters further seems to exclude a more apical CO in H and proximal CO/CN inversion in H and Hconfiguration of equatorial carbonyls). It also renders CO/CN inversion at both iron ions unlikely in H because CO/CN inversion at the proximal iron ion is unlikely