Nitrosylation of Nitric‐Oxide‐Sensing Regulatory Proteins Containing [4Fe‐4S] Clusters Gives Rise to Multiple Iron–Nitrosyl Complexes

Abstract The reaction of protein‐bound iron–sulfur (Fe‐S) clusters with nitric oxide (NO) plays key roles in NO‐mediated toxicity and signaling. Elucidation of the mechanism of the reaction of NO with DNA regulatory proteins that contain Fe‐S clusters has been hampered by a lack of information about the nature of the iron‐nitrosyl products formed. Herein, we report nuclear resonance vibrational spectroscopy (NRVS) and density functional theory (DFT) calculations that identify NO reaction products in WhiD and NsrR, regulatory proteins that use a [4Fe‐4S] cluster to sense NO. This work reveals that nitrosylation yields multiple products structurally related to Roussin's Red Ester (RRE, [Fe2(NO)4(Cys)2]) and Roussin's Black Salt (RBS, [Fe4(NO)7S3]. In the latter case, the absence of 32S/34S shifts in the Fe−S region of the NRVS spectra suggest that a new species, Roussin's Black Ester (RBE), may be formed, in which one or more of the sulfide ligands is replaced by Cys thiolates.


Table of Contents Page
Experimental section 3 Figure S1. Mössbauer spectra of Fe-S cluster regulatory protein before and after reaction with NO. 6 Table S1. Summary of refined Mössbauer parameters for WhiD and NsrR before and after nitrosylation 7 Figure S2. NRVS spectra of WhiD and NsrR 8 Figure S3. WhiD nitrosylation followed by absorbance spectroscopy 9 Figure S4. NRVS analysis of WhiD as a function of NO concentration 10 Figure S5. Overlay of WhiD and NsrR iron nitrosyl NRVS spectra with those of RRE, RBS and DNIC complexes 11 Figure S6. Overview of NRVS spectra for nitrosylated WhiD 12 Figure S7. Overview of NRVS spectra for nitrosylated NrsR 13 Figure S8. DFT calculated NRVS spectra of RRE 14 Table S2. DFT calculated vibrational modes and frequencies for RRE 15 Figure S9. DFT calculated NRVS spectra of RBS 16 Table S3. DFT calculated vibrational modes and frequencies for RBS 17 Figure S10. DFT calculated NRVS spectra of RBE with one thiolate bridge 18 Table S4. DFT calculated vibrational modes and frequencies for RBE with one thiolate bridge 19 Figure S11. DFT calculated NRVS spectra of RBE with two thiolate bridges 20 Table S5. DFT calculated vibrational modes and frequencies for RBE with two thiolate bridges 21 Figure S12. DFT calculated NRVS spectra of RBE with three thiolate bridges 22 Table S6. DFT calculated vibrational modes and frequencies for RBE with three thiolate bridges 23 Figure S13. DFT calculated NRVS spectra of RBE with one persulfide thiolate bridge 24 Table S7. DFT calculated vibrational modes and frequencies for RBE with one persulfide thiolate bridge 25 Figure S14. DFT calculated NRVS spectra of RRE with one persulfide bridge 26 Table S8. DFT calculated vibrational modes and frequencies for RRE with one persulfide bridge 27 Figure S15. DFT calculated NRVS spectra of RRE with two persulfide bridges 28 Table S9. DFT calculated vibrational modes and frequencies for RRE with two persulfide bridges 29 Supporting References 30 previously obtained by titration at lower concentration. [1c] Samples were loaded into NRVS cuvettes and frozen in liquid nitrogen. For samples containing 15 N-nitric oxide, 50 µl of holo-protein was placed in a 0.5 ml microtube and the headspace flushed with 1 ml of 15 N-nitric oxide gas (CIL, CKGas) and the sample gently agitated at ambient temperature for 5 min. The headspace of the sample was replaced a further two times before the sample was loaded into an NRVS cuvette and frozen.
The solubility properties of NO limited the total ratio of NO to original cluster in solution to ~10, with excess NO remaining in the headspace to ensure reaction went to completion. Samples prepared in the absence of nitric oxide donors or gas served as a control.  -8). Each scan took about 45 minutes and all scans were normalized to the intensity of the incident beam and then averaged according to their cts/s signal level. Partial vibrational densities of states (PVDOS) were calculated from the raw NRVS spectra using the PHOENIX software package. [7] During data collection the sample was maintained at low temperature using a liquid He cryostat (head temperature <10 K). Accurate sample temperatures were calculated from the ratio of the anti-Stokes to Stokes intensity by the expression − = and were 50-80 K. [8] Mössbauer Measurements. Measurements were performed using a MS4 spectrometer operating in the constant acceleration mode in transmission geometry, and at 10 K for all samples using a Janis SVT-400 cryostat. A 100 mCi 57 Co in Rh held at room temperature was used as source. All centroid shifts, δ, are given with respect to metallic-Fe at room temperature. The spectra were least square fitted using Recoil software. [9] The parameters from these fittings are centroid shifts (δ), quadrupole splitting (∆EQ), Lorenztian linewidth (Γ), and intensity (I). During the fitting procedure, all parameters were set to be free, for which the two signals were each locked to 50% intensity. This was done because each of these doublets is related to two identical Fe atoms in the cluster. However, even when the intensities were not locked, the hyperfine values were virtually unchanged. The discrepancy in the intensity ratios of the two doublets are likely due to recoil-free fractions of different Fe sites.

NRVS
DFT Calculations. All calculations were performed using the Gaussian 09 set of programs in the gas phase employing the PW91 functional. [10] Iron atoms were described using the LAN2LDZ basis set and effective core potential, with all other atoms being described with the all electron 6-311G** basis set as used by Mitra et al. [11] All structures were optimized and confirmed as minima during the frequency analysis. NRVS spectra were predicted from the DFT outputs using the approximation of Einstein-like modes, where there is no momentum dependence of vibrational frequency and polarization vector, and under the condition of random distribution of molecules. [5a, 12] Calculation of NRVS using harmonic DFT frequencies is a well established approach demonstrating a relatively accurate reproduction of experimental spectral features including isotopic shifts. [6e, 11, 13] As has been demonstrated previously [14] the effects of anharmonicity have much smaller impact on the difference between the experimental and calculated NRVS spectra compared to the error intrinsic to the chosen DFT/basis set combination.
The partial density of states, PVDOS, was calculated according to the following equation. [5a, 12a-h] (1) where composition factor, , for the resonant atom Fe in the normal mode α is given by: ,α Fe e (2) In the latter expression (∆rk,α) 2 is the mean square displacement of atom k in mode α and summation in the denominator is taken over all atoms; mk is the mass of the k th atom. N defines the total number of atoms and f(v-vα) is the line shape function which is a convolution of Gaussian and Lorentzian forms. The extracted normal modes were used to simulate the NRVS spectra according to equation (1) using an in-house built computer program. [12i] Calculated spectral lines were assigned a Lorentzian line shape with the line width of 6 cm -1 for PVDOS.  Figure S1. Mössbauer spectra of Fe-S cluster regulatory protein before and after reaction with NO. A) and B) Zero field Mössbauer spectra for WhiD and NsrR, respectively, as isolated and after addition of excess NO at 10 K. Due to a small asymmetry in the signals, the spectra of the two proteins were both fitted with two doublets, resulting in similar values in δ and ∆EQ (see Table S2), indicated that two pairs of high spin Fe 2+ /Fe 3+ ions exist in each [4Fe-4S] 2+ cluster. Here, a delocalized electron oscillates between the ions giving rise to an average oxidation state of Fe 2.5+ for each pair, where the δ value is the average of the two δ values for high-spin Fe 2+ and Fe 3+ . These results are also similar to those recently reported for [4Fe-4S] NsrR. [1c] Following addition of excess NO, the Mössbauer spectra of WhiD (A) and NsrR (B) were still fitted with two similar doublets with δ and ∆EQ parameters similar to each other but very different from those derived from the initial [4Fe-4S] 2+ spectra (see Table S1). Previous DFT calculations indicated that DNIC species are best described by two resonance structures consisting of high spin Fe 3+ bound to two NO¯ ligands and high spin Fe 2+ bound to an overall quartet 4 (NO)2̄ ligand, with antiferromagnetic coupling such that a total spin of S = ½ results. [15] Table S2. Summary of refined Mössbauer parameters for WhiD and NsrR before and after nitrosylation: Isomer shift, δ; quadrupole splitting, ∆EQ, Lorenztian linewidth (Γ), and intensity (I). NsrR; (top) oxidized ferrodoxin (D14C Fd). [16] WhiD NsrR D14C Fd Figure S3. WhiD nitrosylation followed by absorbance spectroscopy. UV-visible absorbance spectra of as isolated WhiD (grey) and following addition of ~11 NO per [4Fe-4S] cluster (black, previously reported [17] ), and the nitrosylated WhiD NRVS sample following ~40-fold dilution (red).    , [16] particularly in terms of the pattern of bands and their relative intensities right across the spectrum. The broad feature up to 100 cm -1 consists mainly of Fe-N=O bending modes with some S-Fe-S twisting. A sharp feature at ca. 110 cm -1 results from Fe-S-Fe scissoring.

Components
The slightly asymmetric band at 180 cm -1 is due to two vibrations, which are predominantly S-Fe-S scissoring/rocking motions. The band at ca. 280 cm -1 results from a combination of N-Fe-N twisting and S-Fe-S stretches, while those at ca. 350 cm -1 and 360 cm -1 are due predominantly to S-Fe-S stretching and rocking.
Agreement between the calculated and experimental bands in the higher energy part of the spectrum (above 500 cm -1 ) is less good, with the simulation approx. 50-60 cm -1 too high, but the overall pattern remains very well reproduced, with three clear bands. The first (ca. 595 cm -1 in the simulation) is due to N-Fe-N wagging with S-