Discovery of a functional, contracted heme-binding motif within a multiheme cytochrome

Anaerobic ammonium-oxidizing (anammox) bacteria convert nitrite and ammonium via nitric oxide (NO) and hydrazine into dinitrogen gas by using a diverse array of proteins, including numerous c-type cytochromes. Many new catalytic and spectroscopic properties of c-type cytochromes have been unraveled by studies on the biochemical pathways underlying the anammox process. The unique anammox intermediate hydrazine is produced by a multiheme cytochrome c protein, hydrazine synthase, through the comproportionation of ammonium and NO and the input of three electrons. It is unclear how these electrons are delivered to hydrazine synthase. Here, we report the discovery of a functional tetraheme c-type cytochrome from the anammox bacterium Kuenenia stuttgartiensis with a naturally-occurring contracted Cys–Lys–Cys–His (CKCH) heme-binding motif, which is encoded in the hydrazine synthase gene cluster. The purified tetraheme protein (named KsTH) exchanged electrons with hydrazine synthase. Complementary spectroscopic techniques revealed that this protein harbors four low-spin hexa-coordinated hemes with His/Lys (heme 1), His/Cys (heme 2), and two His/His ligations (hemes 3 and 4). A genomic database search revealed that c-type cytochromes with a contracted CXCH heme-binding motif are present throughout the bacterial and archaeal domains in the tree of life, suggesting that this heme recognition site may be employed by many different groups of microorganisms.

in their genomes, and their discovery enriched our knowledge on metabolic reactions that occur in nature (1,2,4,18). Investigation of the biochemical reactions that underlie the anammox metabolism through the characterization of key enzymes has led to the discovery of novel properties of c-type cytochromes and the reactions they can catalyze (1-4, 18, 19).
Cytochrome c proteins are characterized by one or more c-type heme cofactors that are covalently bound to the protein backbone. The two vinyl groups of a heme b moiety form thioether bonds with the sulfhydryls of two cysteine residues that are commonly arranged in the highly-conserved amino acid sequence CXXCH, the canonical heme-binding motif (20). Although the functional implications of covalent heme attachment are not fully understood, it has been proposed that it enhances protein stability and allows for more solvent-exposed heme centers (21). The histidine residue of the CXXCH motif typically serves as the proximal ligand to the heme iron, whereas the distal ligand is often a histidine or a methionine residue at a variable distance from the motif in the amino acid sequence (22). Even though the canonical CXXCH is by far the most common heme-binding motif, several cytochromes with more than two residues separating the cysteines have been described. The most extended heme-binding motif that has been identified so far is found in octaheme MccA from Shewanella and several ⑀-Proteobacteria and contains 15 or 17 residues between the two cysteines (CX 15/17 CH) (23). Several Desulfovibrio species express c 3 tetraheme cytochromes where one or two hemes are bound to a CX 4 CH motif, and a CX 3 CH motif has been observed in Desulfovibrio gigas (24). A CX 4 CH motif is also present in HDH of anammox bacteria, a hydroxylamine oxidoreductase-like octaheme cytochrome c protein that catalyzes hydrazine oxidation to N 2 (3,4). A functional hemecontaining holoprotein protein with a contracted heme-binding motif was produced synthetically only once, and in contrast to extensions, contractions of heme-binding motifs (i.e. CXCH or CCH) have not been observed in nature (25).
In addition to three canonical CXXCH heme-binding motifs, KsTH also contains a CKCH sequence that is fully conserved in all 14 anammox TH sequences that were investigated in this study and potentially serves as a fourth contracted heme-binding site. In this study, we addressed this hypothesis using an array of complementary methods and characterized the spectroscopic and redox properties of purified KsTH. We demonstrated that this contracted heme-binding motif bound a heme c cofactor; we assessed whether this contracted heme coordination resulted in any distinct properties for the heme, and we probed the interaction between purified KsTH and hydrazine synthase.

KsTH harbors four c-type heme cofactors
The gene product of kuste2854 was purified directly from K. stuttgartiensis biomass by a three-step purification protocol. The purified protein migrated as a monomeric species on both native and SDS-denaturing PAGE. Peptide mass fingerprinting using MALDI-TOF MS verified the identity of the protein as kuste2854. Tandem MS analysis of the intact, albeit denatured full-length protein, established a molecular mass of 26095.2744 Da for the monoisotopic species (Fig. 1A). The theoretical value of the apoprotein without the N-terminal targeting sequence (aa 1-29) was 23634.79 Da. The ϩ2460.4844-Da offset in mass could only be explained by the covalent binding of four heme groups to the protein backbone, resulting in a total theoretical mass of 26095.46 Da for the holoprotein, which was an exact match to the mass measured for the denatured full-length protein. Collision-induced dissociation tandem MS (CID MS/MS) detected only one dominant fragment ion of 617.18 [M ϩ H] ϩ Da, and its spectrum matched the simulated isotope envelope of Fe-protoporphyrin IX (heme b) (Fig. 1B). Consequently, the observed mass discrepancy was most likely due to the covalent attachment of four heme groups instead of the three that were predicted from the presence of three canonical CXXCH heme c-binding motifs in the KsTH sequence. Hence, this additional heme must be bound to a noncanonical heme-binding motif. Apart from the cysteine residues present in the three canonical heme-binding sites, there are four additional cysteines present in the KsTH sequence (Cys-12, Cys-14, Cys-20, and Cys-151). Only two of these cysteine residues could constitute a hemebinding site, the contracted CKCH motif, which has been designated as "heme 1-binding site" (Fig. 2).
To probe the covalent attachments formed between the heme moieties and the protein backbone, the reduced alkaline pyridine hemochromogen spectrum of KsTH was recorded. Bis-pyridine heme adducts that are linked to the protein via two thioether bonds result in an absorbance maximum of the ␣-band at 550 nm, whereas the presence of only a single thioether bond results in a red-shifted maximum at 553 nm (26). In perfect agreement, the reduced alkaline bis-pyridine adduct of KsTH displayed the typical absorbance maximum at 550 nm (Fig. 3), which was indicative of two thioether bonds linking the hemes to the protein backbone (26). Taken together,  Genome sequences corresponding to TH_ScaBr and TH_JetEc were used to manually curate the amino acid sequences to account for a frameshift and different assignments of the start codon, respectively.

Discovery of a functional, contracted heme-binding motif
these results further corroborated that the CKCH motif served as the additional heme-binding site in KsTH.

All four hemes are low-spin and hexa-coordinated
UV-visible and electron paramagnetic resonance (EPR) spectroscopy were used to investigate the spin state of the KsTH hemes as well as the nature of their ligation. In optical spectroscopy, oxidized high-spin hemes with a single axial ligand exhibit a Soret maximum around 390 nm and a charge transfer absorption band around 600 nm, whereas in the reduced state a broad signal around 430 nm is generally observed (13,27). Neither of these two features were observed in the KsTH spectra (Fig. 4). In EPR spectroscopy high-spin hemes give rise to a strong signal centered at g ϭ 6 (27, 28), which was absent from EPR spectra of KsTH recorded under different conditions (Fig. 5). Interactions between a high-spin heme and a neighboring co-factor can result in spectral features different from the g ϭ 6 signal. In the case of a multiheme protein such as KsTH, these interaction signals are situated between the signals of the individual hemes. For example, such a signal is observed for c 554 at g ϭ 3.9. Nothing similar could be observed for KsTH. Furthermore, the EPR spectrum of KsTH was very similar to the spectrum of c 554 at pH 12, where all hemes of c 554 have been shown to be low-spin (30).

Heme axial ligation is inferred based on spectroscopy and sequence homology
The UV-visible electronic absorption spectrum of the as-isolated (fully-oxidized) KsTH displayed a broad Soret band with a maximum at 409 nm, whereas, upon reduction, the Soret band maximum shifted to 418 nm with a pronounced increase in amplitude and a decrease in half-width. Absence of a shoulder at 437 and 564 nm (Fig. 4A) ruled out the possibility of a hydroxide ligand for any of the hemes of KsTH (31), and absence of a charge transfer band at 695 nm most likely excluded methionine as a heme axial ligand (27). Upon reduction, the Soret bands of His/Lys-, His/His-, and His/Cys-ligated hemes shift from 408 to 417 nm (32), 407 to 420 nm (30), and 418 to 416 nm (33), respectively. Although thiolate heme ligation in the ferric form usually exhibits red-shifted Soret bands around 418 nm (34), superposition of the optical contribution from a His/Cysligated heme to spectra of His/His-and/or His/Lys-ligated hemes could account for the broad Soret band of oxidized KsTH and the pronounced narrowing and rise in the Soret band upon reduction. EPR spectroscopic studies on purified KsTH agreed with the above conclusions. The EPR spectrum of oxidized KsTH at 15 K had broad spectral features in the g ϭ 2.67 to g ϭ 1.74 range that relaxed slowly. The shape of these signals slightly narrowed at higher temperature, above 70 K (Fig. 5A). The broad shape and temperature-dependent behavior were both characteristic of a spin-spin-interacting system of lowspin hemes. This was consistent with the EPR spectra of another tetraheme c-type cytochrome, cytochrome c 554 , where pairwise interaction between hemes 1 and 3 and hemes 2 and 4 induced interaction signals in EPR (29). Partial reduction of the sample by small amounts of dithionite induced disappearance of the interaction signal and the appearance of two heme signals (Fig. 5B). One of these had a g z value of 3.28, characteristic of highly-anisotropic low-spin hemes with His/Lys or His/His ligation (27). The other signal features g values at 2.54, 2.29, and 1.84, which were compatible with a His/Cys-ligated oxidized heme (33,35,36).
To retrieve more information about KsTH based on its primary structure, homology searches were performed against non-redundant protein sequences using BlastP. Although there was no significant hits to any characterized homolog, a conserved domain belonging to the c 554 protein family was identified (9). This domain was found close to the C terminus and accounted for approximately one-third of the total KsTH sequence length (i.e. amino acids 41-110). Multiple sequence alignments of KsTH orthologs with cytochrome c 554 (Fig. 2) combined with structural insights from the tertiary structure of c 554 allowed us to assign the above-mentioned identified heme ligations to specific hemes. For two of the hemes, the ligands that were detected in KsTH using EPR spectroscopy could be identified in the amino acid sequence of the protein by sequence alignment between anammox TH and c 554 . They were conserved between all TH sequences from all known anammox species and situated within the stretch that was homologous to c 554 . The distal histidine ligand to heme 4 was conserved among c 554 and TH homologs (His-30), and in TH there was a fully-conserved lysine residue (Lys-97) instead of the distal histidine ligand of heme 1 of c 554 . The His ligand to heme 3 of c 554 was outside the protein sequence that was homologous to TH. However, in addition to the above-discussed His ligands, there was only one further histidine fully conserved in TH sequences that was likely the sixth ligand to heme 3.
As discussed above, all hemes of KsTH had two axial ligands, whereas heme 2 of c 554 had only one. Based on the sequence alignment of KsTH with the structure of c 554 , we predicted a stretch of six amino acids (aa 147-152) at the distal side of heme 2 of KsTH. This loop was fully conserved in all sequenced anammox genomes and contained one cysteine and three lysine residues that could potentially serve as iron ligands. EPR and UV-visible spectroscopy point toward the presence of a Cys-

Discovery of a functional, contracted heme-binding motif
ligated heme, and we therefore propose Cys-150 as the sixth ligand for heme 2 in KsTH. Taken together, our results showed that KsTH harbored four hexa-coordinated hemes with His/ Lys (heme 1), His/Cys (heme 2), and two His/His ligations (hemes 3 and 4).

KsTH hemes have low redox midpoint potentials
To assess the redox properties of the heme cofactors of KsTH, electrochemical redox titration monitored by optical spectroscopy was performed (Fig. 6). KsTH titrated in two waves: one centered around Ϫ200 mV (versus SHE), accounting for 66 -85% of the total signal amplitude, and the remaining 15-34% around Ϫ400 mV (versus SHE).
EPR spectra recorded on samples prepared at a few selected potentials between the fully-oxidized KsTH and Ϫ400 mV (Fig.  5B) showed characteristic individual EPR signatures for two of the hemes and interaction signatures at higher potentials when all four hemes were oxidized.
The three hemes of the high-potential wave (Ϫ200 mV) exhibited slightly different redox midpoint potentials. Reduction of the highest potential heme induced loss of its interaction with the neighboring heme and therefore the appearance of a signal at g ϭ 3.28 (pair of hemes 1 and 3). Further decrease of the potential first brought up the signal of the Cys-ligated heme (heme 2), due to reduction of its interacting partner (heme 4), and then induced disappearance of the g ϭ 3.28. The low-potential wave (Ϫ400 mV) could be attributed to the Cys-ligated heme, which could only be partially reduced by dithionite. Such low redox midpoint potentials have been reported earlier for His/Cys-ligated hemes (33,35).
In optical spectroscopy, the difference spectra of the hemes contributing to the transition at Ϫ200 mV could not be resolved. The spectrum showed an ␣-band maximum at 552 nm and a Soret band at 418 nm. The spectrum of the lowpotential component exhibited a blue-shifted ␣-band at 549 nm accompanied by a split Soret band with a maximum at 418 nm and a shoulder at 425 nm (Fig. 6B). Splitting of variable intensity was observed, depending on the sample preparation and the redox direction. A possible explanation for the variability of these spectroscopic signals could involve either conformational changes around the heme pocket and/or ligand exchange of the respective heme iron that might have occurred upon electrochemical reduction. An interesting case of redox-dependent ligand change has been reported for cytochrome cЉ from

Discovery of a functional, contracted heme-binding motif
Methylophilus methylotrophus in which the detachment of the distal histidine ligand upon reduction leads to spin-state transition (37). Additionally, in the alkaline conformer of mitochondrial cytochrome c the native methionine distal ligand of the heme iron is replaced by a lysine (34).

Heme environments are flexible: external ligands can bind to the hemes
To further examine the heme ligand environment, potential external ligands were added to KsTH. Upon addition of CO to the reduced protein, 75% of the ␣-band changed its shape, indicating that CO became a ligand to three out of four hemes (Fig.  4B). The Soret band became narrower but, unexpectedly, did not increase in amplitude. Addition of an ϳ9-fold excess of NO/heme (600 M NO for 65 M heme) to the oxidized protein resulted in an optical spectrum of NO-bound heme (Fig. 4C) and in loss of all EPR features (Fig. 5C). Upon addition of an equimolar amount of NO, the EPR signal amplitude diminished by a factor of 2 (Fig. 5C). The reaction of NO with reduced TH was rapid as judged from the instant color change of the sample upon addition of NO-saturated buffer, even though extreme conditions such as low or high pH values or incubation with a high concentration of NO or CO for hours are usually necessary to achieve NO and CO binding to reduced and oxidized cytochrome c, respectively (38,39). Binding of NO to an oxidized heme can pull an electron from the heme iron resulting in loss of its paramagnetic properties. Our results suggested that NO became a ligand to all hemes of oxidized KsTH. However, exposure of pre-reduced KsTH to NO resulted in loss of the ␣-band, corresponding to approximately two-thirds of the hemes and

Discovery of a functional, contracted heme-binding motif
concomitant appearance of a typical signal for NO-bound heme at 540/560 nm (Fig. 4D). A heme-NO signal appeared in EPR accounting for less than one heme. The lines of the hyperfine split signal were separated by 17 G indicating that NO was bound as a fifth ligand to a reduced heme (Fig. 4C). Otherwise, the EPR spectra were nearly featureless, indicating that the NObound hemes were oxidized, except for the above-mentioned NO-heme signal around g ϭ 2, which could be a degradation product. Therefore, NO partially oxidizes KsTH with an unknown reaction product. Whether this reaction is of physiological importance remains to be determined. Additions of either imidazole or azide to the reduced protein did not result in observable spectroscopic changes.

KsTH and hydrazine synthase exhibit redox interaction
TH is conserved within the hydrazine synthase gene cluster across all sequenced anammox genomes and was hypothesized to provide three electrons for the first half-reaction of hydrazine synthesis, i.e. NO reduction to hydroxylamine (5). Therefore, electron transfer from KsTH to HZS was assessed. In a double-compartment cuvette, reduced KsTH and oxidized HZS gave rise to a Soret band at 415 nm with a shoulder at 425 nm and an ␣-band at 552 nm (Fig. 7A), in agreement with the simulated combined spectrum. Mixing of the two enzymes resulted in loss of the shoulder at 425 nm and in the appearance of Soret band maxima at 409 and 417 nm. The 417-nm absorbance decreased over a time period of 10 min accompanied by a shift of the 409-nm band to 407 nm.
EPR spectra on a 1:1 mixture of reduced KsTH and oxidized HZS (Fig. 7B) resulted in loss of lines in the g ϭ 3 region due to reduction of the corresponding heme(s) of HZS. The g ϭ 2.54 Figure 6. Optical redox titration. Using an optically transparent thin-layer electrochemical cell, redox potentials in the range from ϩ290 to Ϫ475 mV were applied in oxidizing and reducing directions. After equilibration, absorbance changes were monitored. A, sum of two n ϭ 1 Nernst equations was fitted to the difference in absorbance between 552 and 565 nm (solid black line), resulting in midpoint potentials of Ϫ190 Ϯ 20 and Ϫ400 Ϯ 20 mV, respectively. Open and closed squares represent data recorded in the oxidative or reducing direction, respectively. B, difference spectra corresponding to the redox transition at Ϫ400 mV (black line) and Ϫ190 mV (gray line).

Figure 7. Redox interaction between KsTH and hydrazine synthase.
A, combined spectrum of reduced KsTH and oxidized HZS (5:1 stoichiometry) before mixing (purple) and 0 (red), 3 (yellow), 6 (green), and 12 (blue) min after mixing. The inset shows a detailed view of the ␣-band region. B, EPR spectra of a mixture of reduced KsTH with oxidized HZS in a 1:1 stoichiometry (blue) and of as-isolated, fully-oxidized KsTH (orange) and HZS (purple). Spectra were recorded at 13 K, 2 milliwatts microwave power, and with 20 db modulation amplitude. The KsTH spectrum shows in this experiment a signal at g ϭ 2.95 that was not present on other KsTH preparations and may stem from a contaminant.

Discovery of a functional, contracted heme-binding motif
signal of the Cys-ligated heme 2 and the g ϭ 3.28 signal of either heme 1 or 3 of KsTH appeared. Disappearance of the shoulder at 425 nm together with the rise of the g ϭ 2.54 signal revealed oxidation of the low-potential heme of KsTH. Subsequent shift of the Soret band maximum from 409 to 407 nm indicated further oxidation of KsTH. Indeed, oxidized TH exhibited a blue-shifted Soret band maximum at 407 nm compared with HZS at 409 nm. This observation was corroborated by the g ϭ 3.28 signal of oxidized heme 1 or 3 that arose in EPR spectra. Therefore, about two electrons were transferred from KsTH to HZS, resulting in the rapid oxidation of heme 2 and the subsequent slower oxidation of either heme 1 or 3.

Contracted heme-binding motifs are present in a variety of genomes
The discovery of a naturally occurring contracted hemebinding motif in TH prompted us to investigate whether contracted binding sites also occurred in proteins other than KsTH and its orthologs. We found 107 multiheme proteins that also contained the CXCH motif, which were widely spread

Discovery of a functional, contracted heme-binding motif
across different phyla and present in both Archaea and Bacteria ( Fig. 8; Table S1), including environmentally-relevant microorganisms such as nitrate-dependent anaerobic methane-oxidizing archaea (40,41) and aerobic methanotrophic and methylotrophic bacteria (42). The residue separating the cysteines of the CXCH seemed to be conserved within phylogenetic groups; however, this was not the case for all groups. Notably, apart from the Lys residue being conserved within the anammox group, within the Desulfuromonadales group (containing the Geobacteraceae and Desulfuromonadaceae families), either a Gly or a Ser residue was conserved. Nonetheless, the prevalence of this contracted motif across more than 10 phyla from both archaeal and bacterial domains indicated that it was not limited to anammox bacteria. Whether the divergent Cys-separating residues confer different properties is still an open question.

Conclusions
Here. we characterized KsTH, a tetraheme cytochrome with one contracted CXCH heme-binding motif. Our results, combined with detailed analyses of the parts of KsTH protein sequence homologous to cyt c 554 , allowed us to attribute ligands, redox midpoint potentials, and spectral properties to the individual hemes (numbered by order of appearance in the sequence). Hemes 1 and 3 as well as hemes 2 and 4 were in magnetic interaction, indicating that the four hemes were arranged in plan-parallel pairs. Heme 1 was attached to the protein backbone via a CKCH motif and was the first observed naturally occurring representative of such a contracted hemebinding sequence. The CXCH heme-binding motif was conserved among anammox species, and our extensive database search revealed the presence of such a motif in cytochromes of a wide variety of phyla. In anammox bacteria, the heme attached to the contracted heme-binding site was His/Lys-coordinated, and in KsTH it exhibits a redox midpoint potential of around Ϫ200 mV (versus SHE). Its interaction partner, heme 3, had roughly the same redox midpoint potential and was His/ His-ligated. Both hemes had an ␣-band maximum at 418 nm and a Soret band at 552 nm. and one of them had an EPR signal at g ϭ 3.28. Heme 4, belonging to the second heme pair together with heme 2, had the same optical spectrum and midpoint potential as mentioned above, and it was His/His-ligated. Heme 2 had a redox midpoint potential of Ϫ400 mV, probably induced by its His/Cys ligation, and features EPR signals at g ϭ 2.54/2.29/1.84. It had a distinctive optical spectrum with a Soret band at 418 and 425 nm and an ␣-band at 549 nm. The protein environment of this heme appeared to be flexible because the split of the Soret band varied as a function of experimental conditions and the presence of hydrazine synthase. In line with the hypothesis that TH could be an electron donor to HZS, electron transfer from reduced TH to oxidized HZS was observed. The first intermolecular one-electron transfer was rapid, whereas electron transfer from a second heme (either 1 or 3) took several minutes. Interestingly, unlike other c-type cytochromes, addition of NO to KsTH resulted in NO binding to all hemes and in complete oxidation of KsTH when NO was added to the reduced protein. Spectroscopic analyses of KsTH together with the presence of contracted heme-binding motifs in a diverse group of microbial phyla suggest that such unusual cytochrome c proteins might have more functions in nature than just shuttling electrons.

Experimental procedures
All chemicals used were purchased from Sigma, unless stated otherwise. High-performance LC (HPLC)-grade chemicals were purchased from Mallinckrodt Baker. All purification steps took place in ambient air and at 4°C.

Cell-free extract preparation
Cells from a 10-liter laboratory scale-enriched (ϳ95% pure) K. stuttgartiensis continuous membrane bioreactor (43) were harvested and concentrated by centrifugation at 4,000 ϫ g for 15 min (Allegra X-15R, Swinging Bucket Rotor, Beckman Coulter). The pellet was resuspended with 1 volume of 20 mM Tris-HCl, pH 8.0. Cells were lysed by three subsequent passages through a French pressure cell operating at 120 MPa (American Instrument Company). Centrifugation at 4,000 ϫ g for 15 min (Allegra X-15R, Swinging Bucket Rotor, Beckman Coulter) removed cell debris, and the obtained supernatant was subjected to ultracentrifugation at 126,000 ϫ g for 1 h (Optima XE90, Fixed angle 90 Ti rotor, Beckman Coulter) to pellet cell membranes. The supernatant after ultracentrifugation constituted the cell-free extract.

Protein purification
KsTH (KSMBR1_3596) was brought to homogeneity in a three-step purification procedure. Cell-free extract was first fractionated with ammonium sulfate at 85% saturation. After stirring for about 1 h, the sample was allowed to settle overnight. The supernatant was collected by centrifugation (4,000 ϫ g for 20 min) and diluted to about 5% ammonium sulfate saturation with 20 mM KP i , pH 7.0. Liquid chromatography was performed on an Äkta Purifier (GE Healthcare). The sample of interest was loaded onto a 30-ml column packed with ceramic hydroxyapatite (Bio-Rad) and equilibrated with 20 mM KP i , pH 7.0. The column was packed at a flow rate of 10 ml⅐min Ϫ1 (XK 26/20 column, GE Healthcare) and eluted at 5 ml⅐min Ϫ1 ; the eluate was monitored at 280 nm. KsTH eluted as a near-symmetrical peak during a 30-min linear gradient (20 -500 mM KP i , pH 7.0) at a conductivity of about 40 mS⅐cm Ϫ1 . This peak was collected and concentrated with 10-kDa molecular mass cutoff polyether sulfone spin filters (Vivaspin 20; Sartorius Stedim Biotech). The concentrated sample was subsequently loaded on a 65-ml column packed with Superdex 200 (GE Healthcare) and equilibrated with 50 mM KP i , 150 mM NaCl, pH 7.0. The column was packed at a flow rate of 3 ml⅐min Ϫ1 (XK 26/20 column, GE Healthcare) and eluted at 1 ml⅐min Ϫ1 . Purity was checked throughout purification by SDSdenaturing PAGE (44). The identity of the protein was established by MALDI-TOF MS (Bruker Biflex III, Bruker Daltonik). Enzyme preparations were either used immediately or rapidly frozen in liquid nitrogen before storage.

Electronic absorbance spectra
All solutions were prepared freshly in serum bottles sealed with rubber stoppers and made anoxic by alternatively applying vacuum and argon seven times. UV-visible spectra were Discovery of a functional, contracted heme-binding motif recorded at room temperature in 1.4-or 0.2-ml quartz cuvettes (path length 1 cm; Hellma), sealed with rubber stoppers, using a Cary 60 (Agilent) that was placed inside an anaerobic glove box (N 2 /H 2 atmosphere; O 2 Ͻ2 ppm) or on a Safas spectrophotometer (Société Anonyme de Fabrication d'Appareillages Scientifiques) connected to the interior of the glove box by an optical fiber. Spectra taken outside the glove box were recorded on a Cary 60 or 4000 spectrophotometer (Agilent).

Addition of external ligands
NO-containing stock solution (0.9 mM) was prepared by sparging anoxic MOPS buffer (20 mM, pH 7.0) with an NO/He gas mixture (1:1, v/v) for 10 min. CO was added by directly flushing the assay cuvette with pure CO gas for 30 s. Imidazole and azide were added from 100-fold concentrated stock solutions made with the corresponding assay buffer. To prepare reduced KsTH in the presence of NO for EPR spectroscopy, an excess of dithionite was added to the EPR tube in the glove box followed by some grains for nitrite. The sample was frozen in the glove box. For the redox interaction experiment between KsTH and HZS, HZS was purified as described previously (1), and a double-chamber quartz cuvette was used (path length 0.875 cm; Hellma Analytics) under the same conditions described above.
Interaction with hydrazine synthase 5 eq of KsTH and 1 eq of HZS were added in two separated compartments of a double-compartment cell. Optical spectra were recorded on the oxidized proteins and after successive additions of dithionite, until KsTH was completely reduced without excess reductant being present in the cell. Their combined static spectra were recorded, and following mixing, three spectra of the merged samples were recorded with 3-min interval time. Upon mixing, the maximum intensity of the reduced Soret region decreased, which confirmed the absence of any excess reductant in the cell. Fully-reduced KsTH (without excess of reductant) was prepared in a glove box and mixed with oxidized HZS in a 1:1 stoichiometry for EPR spectroscopy.

EPR spectroscopy
Samples were prepared in an anaerobic glove box at 65 M heme, loaded into quartz EPR tubes sealed with butyl rubber stoppers, and then frozen in a cold finger inside the glove box. EPR spectra were recorded on ␣ Elexsys E500 X-band spectrometer (Bruker) fitted with a He-cryostat ESR900 (Oxford Instrument) and temperature control system. Spectra were recorded under conditions given in the figure legends. Several scans were accumulated to increase the signal/noise ratio.
Redox control samples were prepared in a glove box under optical control. Small amounts of 20 mM dithionite solution prepared in 1 M MOPS buffer, pH 7.0, were added to the sample until the optical spectra showed the desired change in the absorbance of the ␣-band. An aliquot of this sample was then transferred to an EPR tube and frozen in the glove box. In a second experiment, the samples were transferred back to the cuvette; 5 mM EDTA and redox mediators (same type and concentration as used for optical redox titrations; see above) were added, and the potential was monitored by a redox electrode calibrated against a saturated quinhydrone solution in 1 M MOPS buffer, pH 7.0. Potential was adjusted by addition of small amounts of 20 mM dithionite solution in MOPS buffer, pH 7.0, or 10 mM ferricyanide solution in water. At various potentials (Ϫ145, Ϫ200, Ϫ410, Ϫ286, and Ϫ225 mV), an aliquot of the sample was transferred to an EPR tube and frozen inside the glove box.

Protein sequence analyses
Protein sequence homology searches were performed against selected entries using the BlastP program at the NCBI website. N-terminal signal cleavage sites were predicted with SignalP 4.1 (47). Multiple sequence alignments were performed using Muscle (48) as implemented in the EMBL webserver and then manually curated.

Mass spectrometry
Samples for MALDI-TOF MS were prepared as described previously (49). Each spectrum (900 -4,000 m/z) was analyzed using the Mascot Peptide Mass Fingerprint (Matrix Science) against the K. stuttgartiensis database, allowing methionine Discovery of a functional, contracted heme-binding motif oxidation as variable modification, 0.2-Da peptide tolerance, and at most one trypsin mis-cleavage.
Samples for ultrahigh-resolution quadrupole TOF MS were prepared as follows. Purified protein sample buffer was exchanged for 0.1% formic acid, 20% methanol using 3-kDa molecular mass cutoff filters (Amicon Ultracel) and subsequently analyzed by direct infusion electrospray ionization tandem MS (ESI-MS/MS) using an ultrahigh-resolution quadrupole TOF mass spectrometer (maXis 5G, Bruker Daltonics). The instrument was operated in positive ionization mode with the following parameters: capillary voltage, 4500 V; offset, 500 V; nebulizer gas pressure, 0.4 bar; N 2 drying gas, 4 liters⅐min Ϫ1 at 200°C, and in-source CID energy: 20 eV, mass range 300 -3,700 m/z, 1 Hz acquisition rate. The z ϭ 31 ϩ precursor ion (m/z 843.3) was isolated and subsequently fragmented using 25-eV collision-induced dissociation (CID) energy. Acquired data were processed in Data Analysis 4.2 software (Bruker Daltonics). Spectral peak detection and charge deconvolution for MS and MS/MS spectra were performed by the SNAP2 algorithm. In addition, a high-resolution charge-deconvoluted MS spectrum was generated using the maximum entropy charge deconvolution algorithm (MaxEnt), which was compared with a simulated isotope pattern of KsTH (aa 30 -236) plus four heme groups. The processed MS/MS spectrum was exported to BioTools 3.2 and compared against in silico b-and y-fragment ion masses of KsTH (aa 30 -236) plus four heme groups (0.05-Da mass tolerance).

Other analytical methods
Protein concentrations were measured with the Bio-Rad protein assay, based on the method of Bradford (50), using BSA as standard. The pyridine hemochrome assay was performed as described previously (51).

Bioinformatic analyses
The NCBI nonredundant (nr) database and all publicly available genomes from the NCBI were surveyed for the contracted heme-binding motif. Sequences that contained a CXCH motif, while also containing a CXXCH motif, were retrieved. Sequences were excluded if the CXCH motif was CCCH, CXCCH, or CCXCH. SignalP 4.1 (47) was used to predict whether the sequence contained a signal peptide. Positive hits that also contained a signal peptide were manually curated to exclude the following: (i) sequences of which either the annotation excluded that they were c-type cytochromes; (ii) sequences in which the contracted motif falls within the signal sequence; and (iii) sequences in which the contracted motif overlapped with the canonical motif (i.e. either CXCHXCH or CXXCHCH). The lineage of the organism in which the protein originated was obtained using the Entrez efetch protocol via the Entrez python package. To investigate the phylogenetic relationship of taxa containing the contracted heme-binding motif, their genomes were downloaded from the NCBI database for phylogenetic analysis. The UBCG pipeline was used to extract conserved phylogenetic marker genes and build multiple alignments (52). A final concatenated alignment of the marker genes was used to infer the phylogenetic relationship of the taxa and was visualized as an unrooted tree using the Interactive Tree of Life software (53).