A distal ligand mutes the interaction of hydrogen sulfide with human neuroglobin

Hydrogen sulfide is a critical signaling molecule, but high concentrations cause cellular toxicity. A four-enzyme pathway in the mitochondrion detoxifies H2S by converting it to thiosulfate and sulfate. Recent studies have shown that globins like hemoglobin and myoglobin can also oxidize H2S to thiosulfate and hydropolysulfides. Neuroglobin, a globin enriched in the brain, was reported to bind H2S tightly and was postulated to play a role in modulating neuronal sensitivity to H2S in conditions such as stroke. However, the H2S reactivity of the coordinately saturated heme in neuroglobin is expected a priori to be substantially lower than that of the 5-coordinate hemes present in myoglobin and hemoglobin. To resolve this discrepancy, we explored the role of the distal histidine residue in muting the reactivity of human neuroglobin toward H2S. Ferric neuroglobin is slowly reduced by H2S and catalyzes its inefficient oxidative conversion to thiosulfate. Mutation of the distal His64 residue to alanine promotes rapid binding of H2S and its efficient conversion to oxidized products. X-ray absorption, EPR, and resonance Raman spectroscopy highlight the chemically different reaction options influenced by the distal histidine ligand. This study provides mechanistic insights into how the distal heme ligand in neuroglobin caps its reactivity toward H2S and identifies by cryo-mass spectrometry a range of sulfide oxidation products with 2–6 catenated sulfur atoms with or without oxygen insertion, which accumulate in the absence of the His64 ligand.

Hydrogen sulfide is a critical signaling molecule, but high concentrations cause cellular toxicity. A four-enzyme pathway in the mitochondrion detoxifies H 2 S by converting it to thiosulfate and sulfate. Recent studies have shown that globins like hemoglobin and myoglobin can also oxidize H 2 S to thiosulfate and hydropolysulfides. Neuroglobin, a globin enriched in the brain, was reported to bind H 2 S tightly and was postulated to play a role in modulating neuronal sensitivity to H 2 S in conditions such as stroke. However, the H 2 S reactivity of the coordinately saturated heme in neuroglobin is expected a priori to be substantially lower than that of the 5-coordinate hemes present in myoglobin and hemoglobin. To resolve this discrepancy, we explored the role of the distal histidine residue in muting the reactivity of human neuroglobin toward H 2 S. Ferric neuroglobin is slowly reduced by H 2 S and catalyzes its inefficient oxidative conversion to thiosulfate. Mutation of the distal His 64 residue to alanine promotes rapid binding of H 2 S and its efficient conversion to oxidized products. X-ray absorption, EPR, and resonance Raman spectroscopy highlight the chemically different reaction options influenced by the distal histidine ligand. This study provides mechanistic insights into how the distal heme ligand in neuroglobin caps its reactivity toward H 2 S and identifies by cryo-mass spectrometry a range of sulfide oxidation products with 2-6 catenated sulfur atoms with or without oxygen insertion, which accumulate in the absence of the His 64 ligand.
Neuroglobin is a member of the globin family (1) that is expressed primarily in neuronal tissues (2) but also in other metabolically highly active endocrine tissues, such as the adrenal gland and the pancreatic islets of Langerhans (3,4). Although the physiological role of neuroglobin is still elusive, early studies suggested that it might be involved in oxygen supply (1,5). However, neuroglobin is generally expressed at low levels and exhibits a high autoxidation rate, producing reactive oxygen species, which makes its role as an O 2 carrier unlikely (6). Neuroglobin can also scavenge reactive oxygen species and reduce nitrite to NO under hypoxic conditions (7)(8)(9).
In contrast to hemoglobin and myoglobin, two histidine residues coordinate the heme cofactor in neuroglobin in the ferrous and ferric states (Fig. 1A). The crystal structure of neuroglobin reveals a high structural similarity to myoglobin, despite Ͻ25% amino acid sequence similarity between the proteins. The structure reveals a large hydrophobic cavity, which connects the proximal and distal sides of the heme (7). His 64 and His 96 coordinate the heme on the distal and proximal sides, respectively (10). Exogenous ligands like O 2 , CO, and NO bind to ferrous neuroglobin (Fe II -Ngb) 2 on the distal side, and binding is limited by dissociation of His 64 , which is slow (6,(11)(12)(13)(14). Limited characterization of the binding of sulfide to ferric neuroglobin (Fe III -Ngb) has been reported; the bimolecular rate constant for H 2 S binding to neuroglobin was 35-and ϳ1000fold slower than for myoglobin and hemoglobin, respectively, at pH 5.5, where the comparisons were made (15). The ability of exogenous ligands to bind to the 6-coordinate heme in neuroglobin has been attributed to the electronic properties of the heme ligand environment and the flexibility of the helix bearing the distal heme ligand (16).
H 2 S is a signaling molecule (17,18), which is synthesized endogenously (19 -22) but is toxic at high concentrations by virtue of its inhibition of cytochrome c oxidase in the electron transfer chain (23). H 2 S is detoxified via a mitochondrial sulfide oxidation pathway comprising four enzymes that convert it to thiosulfate and sulfate. The electrons released by sulfide quinone oxidoreductase, which catalyzes the first step in the mitochondrial sulfide oxidation pathway (24,25), are transferred to complex III, thus coupling H 2 S oxidation to the mitochondrial electron transfer pathway (26). Recently, we have described a hemeprotein-dependent alternative to the mitochondrial pathway for catalytic H 2 S oxidation. Both ferric myoglobin and ferric hemoglobin oxidize sulfide to thiosulfate and to hydropolysulfides (27,28). H 2 S functions as a neuromodulator (29) and also protects neurons from oxidative stress (30). Several differences appear to exist in the enzymes involved in H 2 S homeostasis in brain versus liver, where H 2 S metabolism has been studied in greater detail. Thus, whereas ␥-cystathionase is the major contributor to H 2 S generation via the transsulfuration pathway in liver (31), cystathionase ␤-synthase appears to assume this role in brain (32). The rate of H 2 S consumption in brain is ϳ6-fold slower than in liver and is insensitive to stigmatellin, an inhibitor of complex III (33). Similarly, when [ 35 S]H 2 S was used to track sulfide oxidation in brain homogenate, an insensitivity of sulfate and thiosulfate formation to stigmatellin was observed (33). Very low levels of mRNA (and protein) encoding sulfide quinone oxidoreductase are found in brain (33). These observations raise questions about the mechanism by which H 2 S is oxidized in the brain and the possible involvement of hemeproteins such as neuroglobin in catalyzing H 2 S oxidation via a noncanonical pathway.
In this study, we have examined the interaction of human neuroglobin with H 2 S and find that it can oxidize H 2 S to thiosulfate, albeit inefficiently. The reaction of H 2 S with Fe III -Ngb results in slow reduction of the heme iron to the ferrous state. Replacement of the distal histidine ligand by alanine greatly accelerates sulfide oxidation by neuroglobin and supports its oxidation to thiosulfate and to a wide range of catenated sulfur products.

Purification and reaction of neuroglobin with sulfide
Wild-type and H64A human neuroglobin were expressed as fusion proteins with an N-terminal glutathione S-transferase (GST) tag that was cleaved by thrombin as described under "Experimental procedures." The proteins were judged to be Ͼ95% pure by SDS-PAGE analysis (Fig. 1B).
Human Fe III -Ngb exhibits a UV-visible spectrum typical of a bis-histidine coordinated hemeprotein with a Soret peak at 412 nm and a broad ␣/␤ band with a peak at 532 and a shoulder at 560 nm ( Fig. 2A). Upon the addition of Na 2 S, the Soret peak shifts to 415 nm with a concomitant decrease in intensity, whereas the ␣/␤ bands remain broad with a peak at 540 nm and a shoulder at 575 nm ( Fig. 2A). The spectrum of sulfide-treated neuroglobin under aerobic conditions appears to be a mixture of species and resembles that of O 2 -bound ferrous neuroglobin, which displays absorbance maxima at 413, 543, and 571 nm (34,35). However, accumulation of O 2 -Fe II -Ngb is excluded because similar spectral changes were induced by sulfide under anaerobic conditions with the exception that a slight increase in the intensity at 558 nm was observed, indicating the presence of a small amount of Fe II -Ngb (not shown).
The H64A mutation resulted in a shift in the Soret peak from 412 to 406 nm (at pH 7.4) compared with wild-type neuroglobin, and the ␣/␤ bands were at 565 and 531 nm (Fig. 2B). The addition of sulfide shifted the Soret band to 422 nm and the ␣/␤ bands to 570 and 543 nm (red spectrum). Similar absorbance changes were observed under anaerobic conditions (not shown). The maximum at 422 nm slowly decreased in intensity and slightly blue-shifted to 419 nm over the next 30 min (blue spectrum). A slight increase in intensity was observed at ϳ620 nm, suggesting that a small amount of sulfneuroglobin had formed from the addition of sulfur to the porphyrin ring ( Fig. 2B, inset).
A pH dependence for binding of exogenous ligands to neuroglobin has been reported previously and ascribed to protonation of the distal histidine, which would facilitate its dissociation (15,36). We therefore examined the pH dependence of sulfide binding to wild-type and H64A neuroglobin (Fig. 3). The interaction of Fe III -Ngb with sulfide exhibited biphasic kinetics at all pH values and was monitored at 412 or 406 nm for wild-type and H64A neuroglobin, respectively. From the pH dependence of the initial velocity for sulfide binding, pK a values of 7.19 Ϯ 0.04 and 6.85 Ϯ 0.07 for wild-type and H64A neuroglobin, respectively, were obtained (Fig.  3, A and B). Because these values are essentially the same within experimental error, it rules out the contribution of the distal His 64 ligand to the observed pK a .
Replots of the k obs (for the fast phase) on sulfide concentration yielded the following parameters for wild-type neuroglobin: k on ϭ 13.8 Ϯ 0.7 M Ϫ1 s Ϫ1 and k off ϭ 5.1 Ϯ 0.3 ϫ 10 Ϫ3 s Ϫ1 at pH 7.4 and 25°C, yielding a K D ϭ 370 M (Fig. 3C). The k obs for the slow phase (5.60 ϫ 10 Ϫ4 s Ϫ1 ) was independent of pH and sulfide concentration. The kinetics of sulfide binding to H64A Fe III -Ngb were considerably faster than for the wild-type protein and had to be monitored by stopped-flow spectroscopy. Analysis of the data for H64A neuroglobin yielded the following parameters: k on ϭ 58.8 Ϯ 3.8 ϫ 10 3 M Ϫ1 s Ϫ1 , k off ϭ 3.7 Ϯ 0.4 s Ϫ1 , and K D ϭ 63 M at pH 7.4 and 25°C (Fig. 3D).  the total sulfur (Fig. 4B). Under anaerobic conditions, the concentration of H 2 S lost was equal to the concentration of heme with both wild type and the H64A neuroglobin (Fig. 4, C and D).

Cryo-mass spectrometric characterization of sulfide oxidation intermediates
The products of the reaction between neuroglobin and H 2 S were detected by low-temperature ESI-TOF-MS, which preserves weak coordination species. Significant differences were observed between the mass spectra of wild-type and H64A neuroglobin following H 2 S treatment (Fig. 5, A and B). Analysis of the ⌬(m/z) between the control and H 2 S-treated samples revealed the actual speciation of the reaction products (Table  1). Whereas a limited number of sulfide oxidation products were observed with wild-type neuroglobin (SO . , SO 4 Ϫ , and S 2 O 3 2Ϫ ), a complex array of products was seen with the H64A mutant with 2-6 sulfur atoms (HS 2 Ϫ , HS 3 Ϫ , HS 4 Ϫ , HS 5 Ϫ , HS 6 Ϫ ) and variously oxygenated derivatives (Table 1).   Fig. 2 except that the buffer was varied over the pH range as described under "Experimental procedures." The data were fitted to Equation 2. Wild-type (C) or H64A (D) Fe III -Ngb in 100 mM HEPES buffer, pH 7.4, 25°C, was mixed with 0.05-1.0 mM Na 2 S. For H64A neuroglobin, the reaction was monitored in a stopped-flow spectrophotometer. The k obs for the fast phase was plotted as a function of sulfide concentration.
To investigate whether the observed species are covalently attached either to the heme or to the protein, samples were mixed with acetonitrile (with 0.5% formic acid) to denature the protein and sprayed into the mass spectrometer. No differences in the heme or protein mass spectra were observed (not shown), demonstrating that the sulfide oxidation products were not covalently attached to the heme or to the protein.

EPR spectroscopy
The EPR spectrum of wild-type Fe III -Ngb at 10 K is complex (Fig. 6) and displays a low-spin ferric signal with g ϭ 3.09, 2.17, and 1.48, which is in good agreement with published data (37). The presence of a high-spin signal (g ϭ 6.0 and 1.99) could be due to the presence of a small population of Fe III -Ngb with a disulfide bridge between Cys 46 and Cys 55 , which promotes His 64 dissociation from the iron (38). In the presence of sulfide, a rhombic EPR signal was seen within 5 min with g ϭ 2.48, 2.21, and 1.85 (Fig. 6) representing ϳ28% of the initial spin concentration. The broad positive feature at g Ϸ 1.35 might represent an integer spin (open shell S ϭ 0) species or a ferric heme with very small dxy dyz splitting, as seen with the cyanide adduct of myoglobin (39). However, its origin is presently unknown. With longer incubation times, the EPR signal decayed further (supplemental Fig. S1).
The EPR spectrum of the H64A mutant showed a g ϭ 5.87 axial high-spin signal (Fig. 7). Treatment of H64A neuroglobin with sulfide for 5 min led to the quantitative conversion of the high-spin to a low-spin species, comprising at least two overlapping EPR signals. Simulations were used to deconvolute the spectra to a major signal (ϳ80%) with g ϭ 2.54, 2.24, and 1.87 and a minor signal (ϳ20%) with g ϭ 2.58, 2.21, and 1.83 ( Fig. 7 and supplemental Figs. S2 and S3). The EPR signal decreased in

Resonance Raman analysis of sulfide oxidation by Fe III -Ngb
The high-frequency region of the resonance Raman spectrum comprises in-plane porphyrin vibrational modes that hold information about the oxidation, coordination, and spin states of the heme iron. The resonance Raman spectrum of Fe III -Ngb (0 min) shows bands at 1375 cm Ϫ1 ( 4 ), characteristic of Fe III , and 1507 cm Ϫ1 ( 3 ), characteristic of low-spin Fe III (Fig.  8), which are similar to previously reported spectra of human (40) and murine (41) neuroglobin. Following the addition of sulfide, a gradual shift of 4 to 1362 cm Ϫ1 along with the appearance of 3 at 1482 cm Ϫ1 was observed, indicating the presence of ferrous heme, probably in the low-spin state (42). Increased fluorescence associated with the samples as the reaction progressed (especially between 30 -60 min) resulted in the loss of spectral quality. The spectra obtained under aerobic and anaerobic conditions were identical (not shown), ruling out the formation of O 2 -Fe II -Ngb, which is also not expected to accumulate due to its susceptibility to autoxidation (6).
The H64A mutant shows the oxidation state marker at 1375 cm Ϫ1 ( 4 ) and the spin state marker at 1475 cm Ϫ1 ( 3 ), indicating the presence of high-spin Fe III heme. This is similar to the resonance Raman spectrum of the H64V mutant reported previously at pH 6.4 with bands at 1374 cm Ϫ1 and 1484 cm Ϫ1 (40). In contrast to the wild-type protein, sulfide-treated H64A neuroglobin showed persistence of the oxidation state marker at 1375 cm Ϫ1 over 60 min (Fig. 9). However, the 3 marker at 1475 cm Ϫ1 decreased in intensity, whereas a band at 1498 cm Ϫ1 Figure 5. Mass spectrometric analysis of neuroglobin treated with sulfide. A, mass spectra of neuroglobin without (black) and with H 2 S (red) treatment. B, mass spectra of H64A neuroglobin without (black) and with H 2 S (red) treatment. Spectra were recorded using a low-temperature spraying mode as described under "Experimental procedures" and represent (8ϩ) species. The arrows indicate the appearance of major new species that are not seen in the spectra of wild-type or H64A neuroglobin alone. The full product list is given in Table 1. A.U. arbitrary units. a ⌬(m/z) was calculated by comparing the control spectrum of untreated wild-type or mutant neuroglobin with the sulfide-treated spectrum. The mass difference therefore reflects the necessary addition of protons to compensate for the charge difference caused by the coordination to the iron center. b ND, not determined. appeared to increase, suggesting the presence of a mixed population of high-and low-spin Fe III -Ngb. Although this assignment is supported by the observed increase in a new band at 1585 cm Ϫ1 , characteristic of low-spin heme (43), the assignment is tentative due to its low intensity.
The low-frequency region of the resonance Raman spectrum contains information about the heme iron-metal ligand vibrational modes. This region of the spectrum for wild-type neuroglobin shows minor differences between 0 and 60 min following sulfide addition (Fig. 10), indicating that no significant changes in the heme pocket occurred over this time course. In contrast, the low-energy region of the spectrum for the H64A mutant shows significant time-dependent changes following sulfide addition (Fig. 11). In particular, an intense new band appeared at 495 cm Ϫ1 , which was previously observed with ferric myoglobin treated with sulfide and could represent the S-S stretch in hydropolysulfide bound to the iron center. Further evidence for a change in the heme coordination environment is seen in the 351 cm Ϫ1 band, which corresponds to the totally symmetric Fe-N stretch of the heme (44,45). This mode loses intensity over time and shifts to 345 cm Ϫ1 , whereas a new feature appears at 360 cm Ϫ1 . A similar band at ϳ360 cm Ϫ1 was observed in sulfide-treated ferric myoglobin (28). The D 2 O insensitivity of this band, however, indicates that it does not represent an Fe-SH (or Fe-SH 2 ) vibration, although it could correspond to an Fe-S bond vibration in a hydropolysulfide derivative.

XAS characterization of neuroglobin treated with sulfide
Extended X-ray Absorption Fine Structure (EXAFS) analysis of wild-type Fe III -Ngb in the presence and absence of sulfide reveals the presence of 6-coordinate iron ( Table 2). Fe III -Ngb shows a first inflection point edge energy at 7125.5 (ϪH 2 S) and 7124.4 eV (ϩH 2 S). In comparison with the edge energies of the standards, ferrous sulfate (7122.9 eV) and ferric sulfate (7126.3 eV), the neuroglobin samples appear to consist of a mixture of ferrous and ferric iron (Fig. 12A). The pre-edge peak area of 9.16 10 Ϫ2 eV (neuroglobin) and 8. 16 10 Ϫ2 eV (neuroglobin ϩ sulfide) is consistent with a pseudosymmetric 6-coordinate iron center. The EXAFS data for both samples are best fit with six nitrogen ligands with Fe-N bond lengths of 1.99 Å (Fig.  13A). In addition, there are three Fe-C vectors that are due to single and multiple scattering signals of the porphyrin ring ( Table 2). Sulfur coordination to the heme iron was not detected with wild-type neuroglobin.
The XAS spectra of H64A Fe III -Ngb with and without sulfide show first inflection point energies at 7124.9 eV (ϪH 2 S) and 7124.0 eV (ϩH 2 S) (Fig. 12B). The EXAFS spectrum of H64A neuroglobin is best fit with five nitrogen ligands at a distance of 1.98 Å and three Fe-C vectors with values similar to those seen with wild-type neuroglobin (Fig. 13B). In the H 2 S-treated sample, in addition to Fe-N bonds with a 1.99-Å length, a sulfur ligand at an Fe-S distance of 2.23 Å was detected (Table 2), which is considerably longer than observed in H 2 S-treated myoglobin (2.16 Å) (28) but similar to the 2.25-Å bond length predicted for an FeN 5 S 1 species from the Cambridge structural database. Only two Fe-C vectors at bond distances of 3.02 and 4.08 Å were observed in H64A neuroglobin treated with sulfide.

Reaction of Fe III -Ngb with cysteine persulfide
The transsulfuration pathway enzymes generate cysteine persulfide (Cys-SSH) from cystine (46), which is more reactive than H 2 S and might play a role in sulfide signaling. To test its reactivity toward neuroglobin, Cys-SSH was generated in situ by reaction of ␥-cystathionase with cystine (46). Under aerobic conditions, a small decrease in the 412 nm Soret absorption was seen, which was accompanied by small increases at 538 and 575 nm (Fig. 14A). In contrast, marked changes were observed in the spectrum of sulfide treated H64A neuroglobin (Fig. 14B). The Soret peak shifted from 406 to 422 nm with a simultaneous decrease in intensity, and the ␣/␤ bands increased in intensity at 543 and 570 nm. Under anaerobic conditions, the Soret peak of the wild-type protein shifted to 424 nm, and the ␤/␣ bands sharpened at 530 and 558 nm appear, indicating conversion to Fe II -Ngb (Fig. 14C). The spectrum of H64A neuroglobin was similar under aerobic and anaerobic conditions (Fig. 14, B and D).

Discussion
Neuroglobin is a divergent member of the globin superfamily whose cellular role continues to be a matter of debate (47). It is up-regulated in response to hypoxia and plays a role in protect- Figure 6. EPR spectroscopic characterization of Fe III -Ngb following the addition of Na 2 S. Shown is the spectrum of wild-type (400 M) Fe III -Ngb in aerobic 100 mM HEPES buffer, pH 7.4, containing 20% glycerol (v/v) after the addition of Na 2 S (10 mM) at 25°C at the indicated times. Fe III concentrations were quantified using 1 mM copper EDTA standard. The EPR settings are described under "Experimental procedures." The signal due to adventitious ferric iron is labeled with an asterisk. R.U., relative units.
ing neurons against hypoxic damage (48). Neuroglobin is unusual in several respects, including the presence of a commodious cavity around its 6-coordinate heme (10) that is not present in hemoglobin or myoglobin (Fig. 1A). Spectroscopic studies have revealed the presence of heme orientational isomers and complex kinetics for binding exogenous ligands, which is slowed by displacement of the endogenous distal heme ligand, His 64 (6,12,40,41,49). The oxidation state of a pair of cysteine residues (Cys 46 and Cys 55 in the human sequence), modulates the O 2 affinity of neuroglobin, decreasing it ϳ10-fold when the cysteines are reduced (38).
The interaction between neuroglobin and CO, reactive oxygen and reactive nitrogen species that accumulate during ischemia reperfusion has been studied (6,41,50). In contrast, the interaction between neuroglobin and H 2 S, which also accumulates during hypoxia (51) due to inhibition of the O 2 -dependent route for its disposal, has been minimally investigated (15). Treatment of Fe III -Ngb with NaHS under anaerobic conditions reportedly resulted in a shift in the Soret maximum from 412 to 426 nm (changes in the visible region of the spectrum were not reported) (15). This spectral change was unlike what we observed under similar conditions (Fig. 2) but like the changes that we saw when anaerobic Fe III -Ngb was mixed with Cys-SSH (Fig. 14C). The spectrum of Fe III -Ngb exposed to Cys-SSH under anaerobic conditions showed absorption maxima at 424, 530, and 558 nm, consistent with reduction to Fe II -Ngb. Hydrodisulfides are better reducing agents than the corresponding thiols and can reduce heme iron (52). Hence, it is likely that Cys-SSH reduced Fe III -Ngb via an outer-sphere process under these conditions. These results lead us to conclude that the interaction of Fe III -Ngb with spurious sulfide oxidation products was characterized previously (15). In contrast to Na 2 S used as a sulfide source in our study, NaHS, used in the previous study, is known to be highly contaminated with oxidized sulfide products (53).
When Fe III -Ngb was treated with H 2 S, we observed small changes in the absorption spectrum over a 30-min period, during which time the Soret peak shifted from 412 to 415 nm ( Fig.  2A). The resonance Raman spectra showed a slow conversion from the Fe III to Fe II species, and a mixture of oxidation states was present at 30 min (Fig. 8). In contrast, the EPR spectrum showed a rapid loss of paramagnetic signal at 5 min, and ϳ10% of the original spin concentration remained at 30 min. Collectively, these spectroscopic data suggest the presence of a mixture of (a) diamagnetic and paramagnetic states and (b) Fe III and Fe II species, albeit their relative proportions as reported by EPR and resonance Raman spectroscopy appear to be different. The X-ray absorption near edge structure (XANES) data are consistent with the presence of a mixture of Fe III and Fe II oxidation states both in the presence and absence of sulfide, which indicates the occurrence of photoreduction ( Table 2). The EXAFS data, however, do not support direct sulfur coordination to the iron in wild-type neuroglobin (Fig. 13A). It is important to note that the samples for XAS data collection were prepared under very different conditions in which the concentration of protein was high relative to O 2 , as described under "Experimental procedures." The low-energy Raman data show little change upon exposure of wild-type neuroglobin to sulfide, providing further evidence that sulfide is not directly coordinated to the heme iron.
The chemistry of heme-dependent sulfide oxidation to a mixture of products is complex and poorly understood. To explain the spectroscopic data, we postulate that sulfide binding to the heme pocket in neuroglobin induces O 2 -dependent formation of reactive sulfur/oxygen radical species that is spincoupled to Fe III -Ngb, accounting for the initial loss of the EPR signal but retention of the Fe 3ϩ oxidation state marker as the  Shown is the high-energy region of the spectra monitoring the timedependent changes in H64A Fe III -Ngb (500 M) in 100 mM HEPES buffer, pH 7.4, containing 20% glycerol (v/v), incubated with Na 2 S (10 mM) at 25°C under aerobic conditions. The data were obtained using 413.1-nm excitation with 10 -25-milliwatt laser power. The measurement of each sample took ϳ10 min (various exposure times and accumulations). The spectra showed strong fluorescence, limiting their quality. The Fe III oxidation state marker band 4 at 1375 cm Ϫ1 is maintained throughout the reaction. The spin state marker band 3 is observed at 1475 cm Ϫ1 , indicating that the heme is high-spin. At longer reaction times, an additional signal at 1498 cm Ϫ1 appears, which could indicate the presence of a small fraction of low-spin ferric heme. major species at 5 and 15 min (Fig. 8). Detection of reactive sulfur/oxygen species by cryo-MS (Table 1) and limited O 2 consumption during the reaction (Fig. 4A) are consistent with this model. Spectroscopic studies provide evidence for the electrostatic stabilization of CO bound to neuroglobin as an FeϭCϭO-X ϩ species in which X ϩ could be His 64 and/or Lys 67 (41). These residues could potentially also stabilize sulfide and reactive sulfur species intermediates that are formed in the heme pocket. An alternative explanation to account for the apparent discrepancy at the early time points between the EPR (showing predominance of diamagnetic heme) and the resonance Raman (showing predominance of ferric heme) data is formation of Fe II -O 2 -Ngb. In analogy to Fe II -O 2 -myoglobin (54), Fe II -O 2 -Ngb should exhibit a ferric 4 signature while being EPR-silent. However, a significant accumulation of an Fe II -O 2 species seems unlikely because similar absorbance changes were seen under aerobic and anaerobic (not shown) conditions and because Fe II -O 2 -Ngb undergoes fairly rapid autoxidation (6).
Prolonged incubation (Ͼ60 min) of Fe III -Ngb with sulfide resulted in the full conversion of ferric to ferrous heme, as indi-cated by the 1362 cm Ϫ1 oxidation state marker in the resonance Raman spectrum (Fig. 8). However, the UV-visible spectra at 30 and 60 min (not shown) were the same and do not correspond to the spectrum described for deoxy-ferrous neuroglobin (i.e. Fe II -Ngb), which exhibits a Soret peak at 426 nm and sharp ␤/␣ bands at 528 and 558 nm (6). These results indicate that the 415 nm spectrum obtained in the presence of sulfide ( Fig. 2A) represents a mixture of species, which cannot be assigned at present. We can, however, rule out the presence of a dominant ironcoordinated sulfur species, which is associated with a significant red shift in the Soret peak. Such peak shifts were seen with myoglobin (409 3 428 nm) and hemoglobin (405 3 423 nm) together with sharpening of the ␣/␤ bands (27,28). As described below, spectral changes similar to those of myoglobin and hemoglobin are observed with the H64A mutant of neuroglobin.
The bimolecular rate constant for the interaction of sulfide with Fe III -Ngb (13.8 M Ϫ1 s Ϫ1 at pH 7.4 and 25°C) is considerably smaller than for sulfide binding to hemoglobin (3.   . Lower-energy region of the resonance Raman spectra of H64A Fe III -Ngb, incubated with Na 2 S. The spectra are of the same samples and were acquired under the same conditions as described in the legend to Fig. 9. The appearance of a new, intense band at 495 cm Ϫ1 is probably due to polysulfide formation. Other new signals are observed at 345 and 360 cm Ϫ1 after a 5-min reaction time (see "Results"). The spectrum at 60 min is hampered by strong fluorescence, which limits the data quality. due to the difference in their heme iron coordination states. We note that reduction of the Cys 46 -Cys 55 disulfide bond in neuroglobin stabilizes the 6-coordinate state of the heme iron and decreases its affinity for exogenous ligands (38). Under the conditions of our experiment, where the sulfide concentration was high, the Cys 46 -Cys 55 pair should have been in the reduced dithiol state.
Sulfide oxidation products (i.e. thiosulfate and polysulfides) were detected, albeit in lower amounts with neuroglobin ( Fig.  4A) than seen with myoglobin and hemoglobin (27,28). These results suggest that sulfide oxidation and oxygenation products can be formed even in the absence of direct coordination of the sulfur to the heme. However, we cannot rule out that a small proportion of Fe III -Ngb exists with direct sulfide coordination and is responsible for the observed oxidation products, which do not accumulate above background levels in the absence of O 2 (Fig. 4C).
Overall, the reaction and reactivity of sulfide with neuroglobin are substantially different from those of hemoglobin and myoglobin (27,28), which we attribute to the bis-histidine coordination of the heme iron in neuroglobin. To investigate the influence of the distal histidine on sulfide interaction with neuroglobin, His 64 was replaced by alanine. The latter substitution resulted in a Soret peak shift from 412 nm in wild-type Fe III -Ngb to 406 nm. In the presence of sulfide, the Soret peak of the mutant shifted from 406 to 422 nm, and sharpening of the ␤/␣ bands at 541 and 572 nm was seen (Fig. 2B). These spectral features are most similar to those of sulfide-treated hemoglobin (423, 541, and 577 nm) (27). Coordination of thiols to ferric hemes generally lowers the thiol pK a by ϳ4 units (55), predicting a pK a of ϳ3 for sulfide bound to heme. Hence, HS Ϫ -Fe III -H64A Ngb is expected to be the initial complex that is formed with a Soret peak at 422 nm.
The EXAFS data provide evidence for sulfur coordination to iron with an Fe-S bond distance of 2.23 Å ( Table 2). As with the wild-type protein, a mixture of iron oxidation states was observed by XANES spectroscopy for the mutant with and without H 2 S treatment (Fig. 12B), indicating photoreduction. The H64A mutant showed evidence for sulfur coordination to ferric iron in the presence of Cys-SSH (Fig. 14, B and D). The Soret peak shifted from 406 to 422 nm, and the ␣/␤ bands were observed at 570 and 543 nm. It is unclear whether Cys-SSH was directly coordinated to  B, normalized iron XANES of H64A neuroglobin with (blue) and without (red) sulfide treatment. In both panels, the XANES spectra of neuroglobin are compared with the Fe II SO 4 (green) and Fe III SO 4 (black) standards.

Reaction of H 2 S with neuroglobin
the iron in this sample or whether H 2 S produced by spontaneous decomposition of Cys-SSH (Reaction 1) as reported previously (46), was coordinated to the ferric iron.

REACTION 1
Interestingly, the spectrum of H64A Fe III -Ngb in the presence of cysteine was similar to that observed in the presence of Cys-SSH (not shown), indicating that the distal heme pocket of neuroglobin can accommodate cysteine and, presumably, Cys-SSH, supporting direct coordination of the persulfide to the ferric iron.
In contrast to myoglobin and hemoglobin, the 422 nm Soret band observed with H64A neuroglobin was not stable and, after ϳ5 min, started to shift to 419 nm with a slight increase in intensity (Fig. 2B). It is unclear what the 419 nm band represents, and it is likely to be a mixture of species. We speculate that the accumulation of catenated sulfur oxidation products coordinated to the heme iron might induce the slight blue shift. The H64A mutant exhibited a 4000-fold greater reactivity toward H 2 S (k on ϭ 58.8 Ϯ 3.8 ϫ 10 3 M Ϫ1 s Ϫ1 ) than wild-type neuroglobin. A similar enhancement of nitrite reduction was reported by mutation of the distal histidine residue to leucine or glutamine (7). The 700-fold higher k off for sulfide for H64A versus wild-type neuroglobin is consistent with the role of the distal histidine residue in stabilizing sulfide, as was previously noted with CO in the closed conformation of neuroglobin (41).
The EPR spectrum of the H64A mutant shows high-spin iron, consistent with the presence of 5-coordinate ferric heme (Fig. 7). The addition of sulfide resulted in the appearance of a complex mixture of low-spin ferric species within 5 min and a progressive loss of spin intensity, indicating conversion to an EPR-silent species. Multiple species are present in the EPR spectrum of sulfide-treated H64A neuroglobin (also seen in the 30-and 60-min spectra of the wild-type protein (Fig. 6)). This heterogeneity probably reflects distinct stable conformational substates of the distal pocket that have been spectroscopically characterized previously (12). Additionally, with the H64A mutant, the heterogeneity could reflect the presence of a mixture of oxidized sulfur species coordinated to the ferric iron. Similar EPR spectra have been reported for the multiheme protein SoxXA involved in bacterial sulfur oxidation (55). The complexity in SoxXA spectrum arises from the simultaneous presence of three hemes that are coordinated by histidine on one side and to a thiol, persulfide, or methionine on the other (56).
Interestingly, the resonance Raman spectra of the H64A mutant indicates that the iron remains in the ferric state upon sulfide addition for the duration of the experiment (Fig. 9), whereas the EPR spectrum shows a loss of ϳ50% of the spin concentration at 60 min (Fig. 7). However, the low quality of the resonance Raman spectra due to the high background fluorescence does not allow us to unequivocally exclude the presence of a small amount of ferrous heme between 15 and 60 min. In the low-energy region of the resonance Raman spectrum, a prominent 495 cm Ϫ1 band was observed upon the addition of sulfide (Fig. 11). We assign this band to the S-S vibration of an iron-bound hydropolysulfide product. This assignment is based on its similarity to the 498 cm Ϫ1 band in a Cys-SS-ligated [2Fe-2S] 2ϩ cluster in the FNR transcriptional factor, which was assigned using 34 S isotope labeling to the S-S stretching mode of the persulfide (57). In the resonance Raman spectrum of Na 2 S 4 , a 482 cm Ϫ1 band was assigned to the S-S vibration (58).
The H64A mutant catalyzed production of higher concentrations of sulfide oxidation products paralleled by higher O 2 consumption compared with wild-type neuroglobin (Fig. 4B). On the other hand, only stoichiometric loss of sulfide was observed under anaerobic conditions and corresponded to the heme iron concentration (Fig. 4D). MS analysis revealed a rich mixture of sulfide oxidation chemotypes with 1-4 oxygens and hydropolysulfides with 2-6 sulfur atoms (Table 1). Whereas the reaction of wild-type neuroglobin was associated with oxygenation at a single sulfur (HSO ⅐ , HSO 4 Ϫ , HS 2 O 3 Ϫ ), the H64A mutant was associated with hydropolysulfides (H 2 S 2 , H 2 S 3 , H 2 S 4 , H 2 S 5 , and H 2 S 6 ) with a subset containing oxygenated sulfur (HSO 3 , HS 2 O 3 Ϫ , H 2 S 3 O 3 , and H 2 S 4 O 3 ). In fact, the only product observed in common between wild-type and H64A neuroglobin was thiosulfate (HS 2 O 3 Ϫ ). The combination of spectroscopic approaches used in this study allows us to conclude that the reaction of Fe III -Ngb with sulfide is very different from that of the ferric forms of myoglobin, hemoglobin, and H64A neuroglobin and also different from that previously reported for neuroglobin (15). We speculate that with wild-type neuroglobin, outer-sphere electron transfer reactions involving sulfide, oxygen, and ferric iron lead to the limited formation of sulfide oxidation products. The redox potentials for the S ⅐ ,H ϩ /HS Ϫ couple and the 1 ⁄ 2O 2 , 2H ϩ / H 2 O couple are ϩ920 mV (59) and ϩ815 mV versus the standard hydrogen electrode at pH 7. The redox potential for the Fe 3ϩ /Fe 2ϩ couple in neuroglobin is Ϫ129 mV (6). The large excess of sulfide over Fe III -Ngb probably allows the mismatch in the redox potentials to be overcome in the in vitro experiments but suggests that this reaction is unlikely to occur in vivo. In contrast, persulfides are more reactive toward ferric neuroglobin, and under anaerobic conditions, accumulation of Fe II -Ngb is observed (Fig. 14).
In summary, our results indicate that Fe III -Ngb is relatively inert toward oxidizing sulfide, limiting the potential biological relevance of this interaction, especially when the cysteines are in the reduced dithiol state. The reactivity of wild-type neuroglobin toward H 2 S could, however, increase under cellular redox conditions that result in oxidation of both the cysteine pair and the heme iron. In the dithiol state, the O 2 affinity of human neuroglobin is reduced 10-fold primarily due to a decrease in the dissociation rate of the distal histidine ligand (38). By analogy, the presence of the disulfide bond should increase the affinity of Fe III -Ngb for H 2 S. Preservation of the disulfide bond in the presence of H 2 S is experimentally difficult to achieve in vitro, although it might be possible in vivo. Under these conditions, the interaction of H 2 S with neuroglobin is expected to be enhanced and could lead to sulfide oxidation. Unlike the globins with 5-coordinate hemes, including the H64A mutant of neuroglobin, accumulation of sulfide-coordinated heme is not observed with wild-type neuroglobin. The array of catenated sulfide oxidation products generated that were trapped by cryo-MS provides evidence for the rich ferric iron-dependent sulfide oxidation chemistry that can be catalyzed by iron-containing proteins. In contrast to H 2 S, Cys-SSH results in reduction Fe III -Ngb and accumulation of the bis-His coordinated ferrous species under anaerobic conditions.

Materials
Na 2 S nonahydrate, L-cystine, glutathione, H 2 S solution (0.8 M) in tetrahydrofuran, meta-phosphoric acid, and potassium ferricyanide were from Sigma. Monobromobimane was from Molecular Probes (Grand Island, NY). Dithiothreitol, isopropyl ␤-D-thiogalactopyranoside was from Gold Biotechnology (St. Louis, MO), and ␦-aminolevulinic acid was from Frontier Scientific (Logan, UT). Glutathione-agarose was from Thermo Scientific (Waltham, MA). All other chemicals were reagent grade and were purchased from Fisher. Recombinant human cystathionine ␥-lyase (CSE) was prepared as described previously (60). Protein molecular weight standards were from Bio-Rad.

Purification of human neuroglobin
A thrombin-cleavable GST tag was introduced at the N terminus of neuroglobin. For this, the neuroglobin cDNA was amplified by PCR using a pET28a plasmid containing the human neuroglobin cDNA as template (generously provided by Dr. Gladwin, University of Pittsburgh), using the following primers: 5Ј-TTATAGGATCCATGGAGCGCCCGGAG-3Ј containing the BamHI restriction site and 5Ј-TTATTCTCGA-GTTACTCGCCATCCCAGCC-3Ј with the XhoI restriction site. The PCR product was cut with BamHI and XhoI and ligated into the pEGX19 plasmid. The H64A mutation was introduced by site directed mutagenesis using the QuikChange kit with the following primers: 5Ј-TCGCCTGAGTTCCTGG-ACGCCATCAGGAAGGTGATGCT-3Ј and 5Ј-AGCATC-ACCTTCCTGATGGCGTCCAGGAACTCAGGCGA-3Ј. The presence of the mutation was confirmed by nucleotide sequencing (University of Michigan DNA Sequencing Core).
GST-tagged neuroglobin was expressed in Escherichia coli BL21(DE3) as described previously with some modifications (8). Briefly, starting cultures were grown overnight at 37°C in Luria-Bertani medium containing 100 g/ml ampicillin. The initial culture (1 ml) was transferred to 1 liter of Terrific Broth medium (with 100 g/ml ampicillin) and grown at 37°C until the A 600 reached 0.6 -0.8. Then 0.4 mM ␦-aminolevulinic acid was added, protein expression was induced by 0.1 mM isopropyl ␤-D-thiogalactopyranoside, and the culture was grown for 18 -20 h at 25°C. The cells were harvested by centrifugation, and the cell pellet was stored at Ϫ80°C until further use.
The cell pellet was suspended (5 ml/g wet weight) in lysis buffer (50 mM MOPS, pH 7.0, 0.5 mM DTT, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 2 mg/ml lysozyme) and stirred for 20 min at 4°C. The cells were disrupted by sonication and centrifuged at 38,500 ϫ g for 45 min. The supernatant was applied to a 10-ml glutathione-agarose column. The column was washed with 100 ml of phosphate-buffered saline (140 mM NaCl, 10 mM Na 2 HPO 4 , 2.8 mM KH 2 PO 4 , 1.8 mM KCl), pH 7.3, and GST-tagged neuroglobin was eluted with 10 mM glutathione in 50 mM Tris, pH 8.0, and concentrated to 10 ml with Amicon filters (MWCO ϭ 10,000). Thrombin (5 units/mg of protein) was added to the protein solution, which was dialyzed overnight at 4°C. Free GST was removed by applying the protein solution onto a GST trap column (GE Healthcare, Uppsala, Sweden). Cleaved recombinant neuroglobin was present in the column eluant and was concentrated using an Amicon filter (MWCO ϭ 10,000). The protein was further purified by sizeexclusion chromatography on a 1.6 ϫ 60-cm Hiload Superdex 200 column (GE Healthcare) pre-equilibrated with phosphatebuffered saline. Fractions containing neuroglobin were concentrated and flash-frozen in liquid nitrogen and stored at Ϫ80°C. The yield of wild-type and H64A neuroglobin was 25 mg and 9 mg/liter of culture, respectively.
Before experiments, the heme iron was converted to the ferric state by mixing with a 1.5-fold molar excess of potassium ferricyanide for 5 min at 0°C. Ferricyanide was removed by filtration using a 10-ml desalting column (MWCO ϭ 6000; Bio-Rad), pre-equilibrated with 100 mM HEPES buffer, pH 7.4.

Kinetics of sulfide binding to neuroglobin
The interaction between wild-type neuroglobin (5 M) in 100 mM HEPES, pH 7.4, and Na 2 S (0.05-1.5 mM) was monitored at 25°C under aerobic conditions. Binding of sulfide to H64A neuroglobin was carried out under anaerobic conditions using an Applied Photophysics stopped-flow spectrophotometer (SX.MV18) placed inside a glove box (Vacuum Atmospheres Co., Hawthorne, CA) with an O 2 concentration below 0.7 ppm. A solution of H64A neuroglobin (5 M) in 100 mM HEPES, pH 7.4, was mixed with Na 2 S solution (0.025-0.5 mM) to final concentrations ranging from 25 to 500 M at 25°C. Biphasic timedependent changes in absorbance at 412 nm (wild-type neuroglobin) and 406 nm (H64A neuroglobin) were observed and fitted using Equation 1 to obtain the rate constants associated with each reaction phase. In Equation 1, A 1 and A 2 represent the amplitude changes associated with the two phases, k 1 and k 2 represent the corresponding rate constants, and y 0 is the final absorbance.
The values for k off , k on , and K D , were obtained by replotting the dependence of k on on the concentration of sulfide.

Dependence of sulfide-binding kinetics on pH
Neuroglobin (5 M) in 100 mM aerobic buffer, pH ranging from 5.5 to 9.0 was mixed with 1 mM Na 2 S solution in the same buffer. Tris was used for pH 9.0, 8.5, 8.0, and 7.8; HEPES for pH 7.75, 7.5, and 7.0; and MES for 6.5, 6.0, and 5.5. The change in absorbance was monitored at 412 nm following the addition of sulfide, and the kinetic data were fit to Equation 1. Due to the faster kinetics of sulfide binding at lower pH, the experiments between pH 5.5 and 6.5 were performed on a stopped-flow spectrophotometer. The pH dependence of sulfide binding to H64A neuroglobin was monitored in a stopped-flow spectrophotometer at 406 nm. The data were fitted to Equation 1 to obtain an estimate of k 1 , the rate constant for the fast phase. The initial velocity (v 0 ) for the fast phase was then obtained using the relationship, v 0 ϭ k 1 ϫ A 1 . The dependence of v 0 on pH was fitted to Equation 2 to obtain the pK a value. y ϭ vH ϩ vL ϫ 10 n͑pKa Ϫ pH͒ 1 ϩ 10 n(pKa Ϫ pH) (Eq. 2)

Reaction of neuroglobin with cysteine persulfide
Neuroglobin (10 M wild type or H64A mutant) was mixed with CSE (2 M) in aerobic 100 mM HEPES buffer, pH 7.4, at 37°C. Persulfide formation was initiated by 1 mM cystine, and the absorbance was monitored from 300 to 750 nm. For monitoring the reaction under anaerobic conditions, it was set up similarly but in an anaerobic chamber, and anaerobic reagents were used.

Analysis of neuroglobin-catalyzed sulfide oxidation
Oxygen consumption was monitored using a Clark O 2 electrode. For this, neuroglobin (25 M) in 100 mM HEPES buffer, pH 7.4, was placed in a 1.5-ml Gilson type chamber, and Na 2 S (1 mM in the same buffer) was injected into the chamber. O 2 consumption was recorded using a chart recorder.
The sulfide oxidation products, sulfite and thiosulfate, were detected using an HPLC method, and polysulfides were measured using cold cyanolysis as described previously (27). For this, neuroglobin (50 M, 1 ml) was incubated with or without Na 2 S (1 mM) in aerobic 100 mM HEPES, pH 7.4, at 25°C in a closed 1.5-ml sample tube. Aliquots were removed at the desired time points. For HPLC analysis, aliquots (45 l) were mixed with 2.5 l of 1 M Tris, pH 9.0, and 2.5 l of 60 mM monobromobimane in DMSO in black 0.5-ml sample tubes and incubated at room temperature for 10 min. Then 100 l of meta-phosphoric acid (16.8 mg/ml in H 2 O) was added, the mixture was centrifuged for 3 min at 15,900 ϫ g, and the supernatants were stored at Ϫ80°C until analysis.
For determining polysulfide concentrations, 100-l aliquots were incubated with 50 l of 62.5 mM potassium cyanide in 125 mM ammonium hydroxide for 45 min at 25°C. Then 150 l of Goldstein solution (1.25 g of Fe(NO 3 ) 3 ⅐9H 2 O in 50 ml of 17% HNO 3 ) was added, mixed, and centrifuged for 5 min at 15,900 ϫ g. The absorbance at 460 nm was recorded, and the concentration of polysulfides was determined using a calibration curve generated with potassium thiocyanate.

EPR spectroscopy
EPR spectra were recorded on a Bruker EMX 300 equipped with a Bruker 4102-ST cavity and an Oxford liquid helium cryostat. Spectra were measured at 10 K with the following parameters: 9.34-GHz microwave frequency; power, 10 milliwatts; modulation amplitude, 7.5 G; modulation frequency, 100 kHz; 5000-G sweep width centered at 3000 G; conversion time, 164 ms; time constant, 82 ms. Fe III -Ngb (400 M) and H64A Fe III -Ngb (500 M) samples were prepared aerobically in 100 mM HEPES, pH 7.4, containing 20% glycerol. Na 2 S prepared in the same buffer was added to a final concentration of 10 mM. At the desired time points, 300 l of the solution was transferred to an EPR tube, and the tube was sealed with a silicon septum and frozen in liquid nitrogen. Low spin signals were quantified using a 1 mM copper EDTA solution. Low spin (S ϭ 1 ⁄ 2) signals were fitted with the Easyspin toolbox (61) using the parameters described in supplemental Tables S1 and S2.

Resonance Raman spectroscopy
The resonance Raman measurements were performed using the 413.13 nm excitation line from a Kr ϩ gas ion laser (Spectra Physics Beam Lok 2060-RS). The frozen samples were prepared in EPR tubes, which were placed in a liquid N 2 -filled EPR cold finger Dewar for all measurements or stored in a Dewar filled with liquid N 2 in the dark until further use. Raman spectra were recorded at 77 K using an Acton two-stage TriVista 555 monochromator connected to a liquid nitrogen-cooled CCD camera (Princeton instruments Spec-10:400B/LN). The experiments were performed at different exposure times and accumulations due to the background fluorescence associated with neuroglobin, which increased with the reaction time. Typical laser powers were in the 10 -30-milliwatt range. Relative wavenumbers (Raman shift) were calibrated using sodium sulfate (Na 2 SO 4 ).

Mass spectrometry
MS measurements were performed on a UHR-TOF Bruker Daltonik maXis 4G (Bremen, Germany), coupled to a Bruker cryospray unit, an ESI-TOF-MS with a resolution of at least 40,000 full width at half-maximum. Detection was in the positive-ion mode. The flow rates were 300 l/h. The drying gas (N 2 ) was held at 10°C, and the spray gas was held at 10°C. The machine was calibrated before every experiment by direct infusion of the Agilent ESI-TOF low concentration tuning mixture, which provided an m/z range of singly charged peaks up to 2,700 Da in both ion modes. Wild-type or H64A neuroglobin (50 l of 400 M) in 200 mM ammonium carbonate buffer, pH 7.4, was mixed with 4 mM H 2 S (from 0.8 M H 2 S in tetrahydrofuran), diluted 10-fold with nano-pure water, and sprayed continuously into the MS. The spectra are recorded over 45 min. As a control, covalent modification of the protein or the heme cofactor was investigated by mixing 50 M neuroglobin in 20 mM ammonium carbonate buffer, pH 7.4, with 500 M H 2 S for 45 min. Samples were then mixed with acetonitrile (50/50, v/v) and formic acid (0.5%) and sprayed at 200°C with a flow rate of 180 l/h.

XAS analysis
Samples were prepared by mixing wild-type or H64A neuroglobin (1 mM in iron concentration) in 100 mM HEPES, pH 7.4, with H 2 S (20 mM final concentration) in tetrahydrofuran and incubated for 30 min at 25°C. An equivalent volume of tetrahydrofuran only was added to the control sample. Glycerol (30% (v/v) final concentration) was added to the samples, which were then loaded into Kapton-wrapped lucite XAS sample cells and frozen quickly in liquid N 2 until data collection at the beamline.
XAS data were collected at the Stanford Synchrotron Radiation Laboratory on beamlines 7-3 and 9-3. Beamline 7-3 is equipped with a single rhodium-coated silicon mirror and an Si[220] double crystal monochromator, and harmonic rejection was achieved by detuning the monochromator 50%. The 9-3 beamline is equipped with an Si[220] double crystal monochromator and a harmonic rejection mirror, so spectra were collected under fully tuned conditions. Samples were maintained at 10 K using Oxford Instrument continuous-flow liquid helium cryostats at both beamline locations. Protein fluorescence excitation spectra were collected using a 30-element germanium solid-state array detector at 7-3 and a 100-element germanium solid-state detector at 9-3. XAS spectra were measured using 5-eV steps in the pre-edge regions (6900 -7094 eV), 0.25 eV steps in the edge regions (7095-7135 eV), and 0.05 Å Ϫ1 increments in the EXAFS region (to k ϭ 13.5 Å Ϫ1 ), integrating from 1 to 20 s in a k 3 -weighted manner for a total scan length of ϳ40 min. X-ray energies were calibrated by collecting an iron foil absorption spectrum simultaneously with collection of protein data. The first inflection point for the iron foil edge was assigned at 7111.3 eV. Each fluorescence channel of each scan was examined for spectral anomalies before averaging, and spectra were closely monitored for photoreduction. Spectra collected at 7-3 represent the average of 7-8 scans, whereas spectra collected at 9-3 represent the average of 4 -5 scans.
XAS data processing and analysis were performed following protocols outlined previously (62). EXAFS spectra were simulated using both filtered and unfiltered data; however, simulation results are presented only for fits to raw (unfiltered) data. Simulation protocols and criteria for determining the best fit have been described previously (63).
Author contributions-M. R., J. K., B. E. L., and M. R. F. performed the experiments. N. L. helped analyze the resonance Raman data and co-wrote the manuscript. T. S. helped analyze the XAS data and cowrote the manuscript. All authors analyzed the experiments performed in their laboratories. M. R. and R. B were primarily responsible for writing the manuscript. All authors edited and approved the final version of the manuscript.