NirN Protein from Pseudomonas aeruginosa is a Novel Electron-bifurcating Dehydrogenase Catalyzing the Last Step of Heme d1 Biosynthesis*

Background: Cytochrome cd1 nitrite reductase contains heme d1 as an essential cofactor. Results: P. aeruginosa lacking the NirN protein produces the heme d1 precursor dihydro-heme d1, which is converted by NirN to heme d1. Conclusion: NirN catalyzes the last step of heme d1 biosynthesis. Significance: The so far unknown function of NirN is revealed. Heme d1 plays an important role in denitrification as the essential cofactor of the cytochrome cd1 nitrite reductase NirS. At present, the biosynthesis of heme d1 is only partially understood. The last step of heme d1 biosynthesis requires a so far unknown enzyme that catalyzes the introduction of a double bond into one of the propionate side chains of the tetrapyrrole yielding the corresponding acrylate side chain. In this study, we show that a Pseudomonas aeruginosa PAO1 strain lacking the NirN protein does not produce heme d1. Instead, the NirS purified from this strain contains the heme d1 precursor dihydro-heme d1 lacking the acrylic double bond, as indicated by UV-visible absorption spectroscopy and resonance Raman spectroscopy. Furthermore, the dihydro-heme d1 was extracted from purified NirS and characterized by UV-visible absorption spectroscopy and finally identified by high-resolution electrospray ionization mass spectrometry. Moreover, we show that purified NirN from P. aeruginosa binds the dihydro-heme d1 and catalyzes the introduction of the acrylic double bond in vitro. Strikingly, NirN uses an electron bifurcation mechanism for the two-electron oxidation reaction, during which one electron ends up on its heme c cofactor and the second electron reduces the substrate/product from the ferric to the ferrous state. On the basis of our results, we propose novel roles for the proteins NirN and NirF during the biosynthesis of heme d1.

denitrifying organisms such as the human pathogen Pseudomonas aeruginosa, denitrification serves as a very efficient strategy for energy generation under anaerobic growth conditions (1). In the second step of the denitrification process, nitrite is reduced to nitric oxide. In P. aeruginosa, this reaction is catalyzed by the periplasmic cytochrome cd 1 nitrite reductase NirS (2). This enzyme is a homodimer of two 60-kDa subunits, each carrying a heme c and a heme d 1 cofactor (3). The non-covalently bound heme d 1 represents the catalytic site of the enzyme where the reduction of nitrite to nitric oxide takes place (4).
Heme d 1 belongs to the tetrapyrrole subfamily of isobacteriochlorins. It carries two unusual carbonyl functions on pyrrole rings A and B and an acrylate side chain on ring D that make heme d 1 unique among the naturally occurring tetrapyrroles ( Fig. 1) (5,6). The biosynthesis of heme d 1 from the common tetrapyrrole precursor uroporphyrinogen III requires six enzymatic steps (7). Most of the involved enzymes are encoded by the nir genes, which are located together in the nir operon in P. aeruginosa (nirSMCFDLGHJEN) and also in other denitrifying bacteria (Fig. 1A) (8 -10). In the first step, the S-adenosylmethionine-dependent uroporphyrinogen III methyltransferase NirE catalyzes the methylation of uroporphyrinogen III to precorrin-2 (11,12). Next, precorrin 2 is converted to siroheme in two enzymatic steps catalyzed by a precorrin-2 dehydrogenase and a sirohydrochlorin ferrochelatase not encoded within the nir operon in P. aeruginosa. Siroheme is then decarboxylated to 12,18-didecarboxy-siroheme by the action of siroheme decarboxylase, which is composed of the four proteins NirD, NirL, NirG, and NirH (7). The penultimate step of heme d 1 biosynthesis consists of the removal of the two propionate side chains on pyrrole rings A and B, with the concomitant introduction of the carbonyl functions at these positions yielding the heme d 1 precursor dihydro-heme d 1 . This transformation is believed to be catalyzed by the radical S-adenosylmethionine enzyme NirJ (7). However, this hypothesis still requires experimental proof. Similarly, the enzyme responsible for the last step of heme d 1 biosynthesis has not yet been identified. In this step, the propionate side chain on pyrrole ring D is converted to the corresponding acrylate side chain (Fig. 1B). At some stage of its biosynthesis, heme d 1 , or a precursor thereof, has to be transported across the cytoplasmic membrane to be incorporated into its target enzyme NirS in the periplasm. Currently it is thought that all heme d 1 biosynthesis steps, except the last one, take place in the cytoplasm and that the final step is catalyzed in the periplasm. This hypothesis is on the basis of the finding that NirF, which has been proposed to catalyze the acrylate group formation, is a periplasmic protein (13), whereas NirE, NirDLGH, and NirJ are located in the cytoplasm.
Depending on the organism, NirF is either a soluble periplasmic protein or attached to the outer face of the cytoplasmic membrane, most likely by a lipid anchor (14,15). In addition to its periplasmic location, purified recombinant NirF from Paracoccus pantotrophus has been shown to bind heme d 1 in vitro (13). Moreover, early mutational studies with different denitrifying bacteria have shown that the deletion of the nirF gene results in bacterial strains unable to synthesize heme d 1 (8 -10, 13). On the basis of all of these findings, NirF has been proposed to catalyze the formation of the acrylic double bond of heme d 1 . However, so far there is no direct experimental evidence clearly attributing this enzymatic function to NirF.
Interestingly, another nir gene (nirN) is present in all denitrifying bacteria. The encoded NirN protein shares about 24% amino acid sequence identity with NirS. Moreover, NirN has been shown to be located in the periplasm and to contain a heme c cofactor (16). The monomeric NirN was able to bind heme d 1 in vitro and transferred the cofactor to NirS when the two proteins were mixed (12). The deletion of the nirN gene in different denitrifiers led to a mild growth phenotype, and cellfree extracts prepared from the corresponding ⌬nirN strains exhibited reduced NirS activity compared with the WT strains (9,12,16). From these observations it was deduced that NirN is not required for heme d 1 biosynthesis but, rather, might be involved in the insertion of the cofactor into NirS. In line with this proposal it has been shown that NirN from P. aeruginosa interacts with both NirS and NirF in vivo. However, in that study it was also observed by UV-visible absorption spectroscopy of periplasmic protein fractions that the cofactor content of NirS in the P. aeruginosa ⌬nirN strain was different from that in the P. aeruginosa WT strain (15). Later, the same observation was reported for a ⌬nirN strain of Magnetospirillum gryphiswaldense (17).
The apparently altered cofactor content of NirS in these ⌬nirN strains can be rationalized by two different hypothetical scenarios: (i) NirN is involved in heme d 1 insertion into NirS, and the absence of NirN leads to an improper incorporation of the cofactor causing the observed spectral differences, or (ii) NirN is involved in the last step of heme d 1 biosynthesis (together with NirF?), and the absence of NirN leads to the accumulation of dihydro-heme d 1 , which is nevertheless inserted into NirS, leading to a different form of holo NirS with different spectral properties and reduced enzyme activity. Therefore, the aim of this study was to investigate and clarify the still enigmatic function of NirN during either heme d 1 biosynthesis or insertion into NirS.

EXPERIMENTAL PROCEDURES
Chemicals-Chemicals and reagents were obtained from Sigma-Aldrich (Taufkirchen, Germany), Merck (Darmstadt, Germany), or Thermo Fisher Scientific Inc. (Schwerte, Germany). Restriction endonucleases were purchased from New England Biolabs (Frankfurt am Main, Germany). QIAquick PCR purification and gel extraction kits were purchased from Qiagen GmbH (Hilden, Germany). Q-Sepharose Fast Flow and SP-Sepharose Fast Flow were obtained from GE Healthcare. Strep-Tactin-HC resin, Desthiobiotin, and Avidin were purchased from IBA GmbH (Göttingen, Germany). All oligonucleotide primers were obtained from Metabion International AG (Martinsried, Germany).
Construction of Vectors-For the construction of a NirN expression vector, the nirN gene from P. aeruginosa PAO1 was PCR-amplified using the primers strepIInirN_fw (TAT ACC ATG GGC TGG AGC CAC CCG CAG TTC GAA AAA GCT AGC GGC GAA GCG CCG GGG GAA G) and nirN_rev (CG GGA TCC TCA GTG CGA GGT TC), which contained NcoI and BamHI restriction sites (underlined), respectively. The di- gested nirN insert was ligated into the correspondingly digested vector pET-22b(ϩ) to generate pET-22b(ϩ)StrepIInirN. For the construction of an expression vector for NirF from D. shibae, the plasmid pMK-RQrbspelBnirF (D.s.) PCShis 6 was purchased from Invitrogen. Through restriction of the plasmid with NdeI and Hin-dIII, the insert pelBnirF (D.s.) PCShis 6 was obtained and ligated into the correspondingly digested plasmid pET-22b(ϩ). A Strep tag coding insert was obtained by annealing the primers CTA GTA GCG CTT GGA GCC ACC CGC AGT TCG AAA AAG and GAT CCT TTT TCG AAC TGC GGG TGG CTC CAA GCG CTA. The His tag coding sequence was removed from pET-22b(ϩ)pelBnirF (D.s.) PCShis 6 using SpeI and BamHI restriction sites, and the Strep tag coding insert was ligated into pET-22b(ϩ)pel BnirF (D.s.) PCS to generate pET-22b(ϩ)nirF (D.s.) PCSStrepII.
Purification of Proteins-The purifications of recombinant P. aeruginosa NirS s.a. , native holo NirS WT , and NirS ⌬N were performed according to Parr et al. (22). Recombinant NirN and NirF were purified using Strep-Tactin-HC resin according to the instructions of the manufacturer (IBA GmbH) and with 50 mM Tris-HCl (pH 7.5), 150 mM NaCl as the buffer.
Determination of Protein Concentration-Bradford reagent (Sigma-Aldrich) was used to determine protein concentrations. The reagent was used according to the instructions of the manufacturer using BSA as a standard.
Extraction of Nitrite Reductase Cofactors-Heme d 1 and dihydro-heme d 1 were extracted from purified NirS WT and NirS ⌬N, respectively. Proteins at a concentration of 90 M in buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl) were precipitated with 6 mM HCl. Two volumes of ethyl acetate (HPLC grade) were added, and the samples were mixed for 1 min and then centrifuged for 2 min at 13000 ϫ g. The organic cofactorcontaining phase was removed and dried in vacuo. For HR-ESI-MS analysis, 3 proteins were transferred to double distilled water prior to the extraction of the cofactors, and the dried cofactors were dissolved in methanol.
UV-visible Absorption Spectroscopy-UV-visible absorption spectra of purified proteins and extracted cofactors were recorded using a V-650 spectrophotometer (Jasco, Gross-Um-stadt, Germany). The reduced forms of proteins and isolated cofactors were obtained by addition of a few grains of sodium dithionite.
Resonance Raman Spectroscopy-RR spectra with a spectral resolution of 2 cm Ϫ1 were recorded using a confocal Raman spectrometer (Lab Ram HR-800, Jobin Yvon) equipped with a liquid nitrogen-cooled charge-coupled device camera. The 442-nm excitation line of a He-Cd laser (Soliton) was focused onto the sample (2 m in diameter, 1 milliwatt at the sample). Using a Linkam Cryostage THMS600 cryostat, the temperature of the sample was set to 80 K throughout the measurement. Raman shifts were calibrated to an accuracy of 0.5 cm Ϫ1 using toluene as an external standard. All spectra were corrected by polynomial background subtraction.
Heme d 1 and Dihydro-Heme d 1 Binding to NirN and NirF-For binding of heme d 1 and dihydro-heme d 1 to NirN and NirF, the in vacuo dried cofactors were directly dissolved in 150 l of buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl) containing either NirN at a concentration of 20 M or NirF at a concentration of 38 M. Directly after mixing, UV-visible absorption spectra of the mixtures were recorded on a V-650 spectrophotometer (Jasco).
NirN Activity Assay, Transfer of Cofactors to NirS s.a. , and Reconstitution of NirS s.a. -The in vitro NirN activity assay was performed under aerobic or anaerobic conditions in an anaerobic chamber (Coy Laboratories, Grass Lake, MI). For the aerobic assay, dried dihydro-heme d 1 was dissolved in buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl) and transferred to a quartz cuvette (Hellma Analytics, Müllheim, Germany). NirN was subsequently added at a final concentration of 20 M to give a final volume of 150 l. For the anaerobic assay, the dried dihydro-heme d 1 was dissolved in degassed buffer in the anaerobic chamber, followed by the addition of 20 M NirN under anaerobic conditions. The NirN-catalyzed reaction was monitored using a V-650 spectrophotometer (Jasco). For the reconstitution of NirS s.a. with heme d 1 or dihydro-heme d 1 and the transfer of the NirN reaction product to NirS s.a. , the enzyme was added to the NirN assay or the cofactor solution at a final concentration of 20 M. For the oxidation of heme d 1 , sodium persulfate was added.

HR-ESI-MS (Accurate Mass Measurements) of Nitrite Reductase Cofactors-
The sample solutions obtained after cofactor extraction were used for accurate mass measurements on an LTQ-Orbitrap Velos (ThermoFisher Scientific, Bremen, Germany). Electrospray measurements in positive mode were performed in direct infusion mode using a custom-made microspray device mounted on a Proxeon nanospray ion source. The microspray device allows sample infusion without sheath gas through a stainless steel capillary (90-m inner diameter) at a flow rate of ϳ1 l/min. The spray voltage was set between 2.0 and 2.3 kV. The resolution of the Orbitrap mass analyzer was set to 100,000 full width at half-maximum (FWHM) at m/z 400. Accurate mass measurements were performed using the lock mass option of the instrument control software using the cation of tetradecyl-trimethylammonium bromide (256.29988 u) as an internal mass reference, which was added to the sample solution in appropriate amounts.

RESULTS AND DISCUSSION
Purification of Native NirS from the P. aeruginosa PAO1 ⌬nirN Strain RM361-To obtain more insights into the role of NirN, the first aim of this study was to purify NirS from the P. aeruginosa PAO1 ⌬nirN strain RM361 and to characterize the cofactor content of this apparently different form of holo NirS. Previous attempts to purify native NirS from the P. aeruginosa PAO1 strain RM361 only resulted in the isolation of semi-apo NirS (NirS s.a. ) completely lacking heme d 1 (15). Therefore, the purification protocol was changed slightly, including the reduction of the column volume of the anion exchange chromatography column. Using the new purification procedure, NirS was purified to apparent homogeneity ( Fig.  2A). The enzyme purified from strain RM361 (NirS ⌬N ) exhibited a less intense green color in contrast to the green color of NirS isolated from the WT strain (NirS WT ) and the red colored semi-apo form of NirS lacking heme d 1 . This observation clearly supports the idea that a different form of holo NirS was produced in the P. aeruginosa ⌬nirN strain.
UV-visible Absorption Spectroscopy of Purified NirS ⌬N and NirS WT -To characterize the cofactor content of NirS ⌬N , UVvisible absorption spectra of the purified enzyme were recorded of the oxidized (as isolated) and dithionite-reduced forms and compared with the corresponding spectra of purified NirS WT (Fig. 2, B and C). The absorption spectra of NirS ⌬N are clearly different from those of NirS WT in both the oxidized and reduced forms. In fact, the spectra of the two enzymes differ, particularly in the region of the characteristic heme d 1 absorption bands. Although the spectrum of oxidized NirS WT exhibits a typical heme d 1 peak at 642 nm, the spectrum of NirS ⌬N shows a peak at 626 nm. In the reduced form, the heme d 1 absorption band of NirS WT is shifted to 649 nm, whereas, for NirS ⌬N , a peak shift to 631 nm is observed. These spectral differences between the purified NirS ⌬N and NirS WT confirmed the presence of two different forms of holo NirS depending on whether the enzyme was isolated from P. aeruginosa PAO1 strain RM361 or from the WT strain. There were two possible explanations for the observed spectral differences. Either heme d 1 was not properly inserted into NirS ⌬N and the observed differences were due to a different ligation of the cofactor, or it was not heme d 1 but the precursor dihydro-heme d 1 that was present in NirS ⌬N . On the basis of the observed UV-visible absorption spectra of NirS ⌬N , we favored the second hypothesis. The conjugated system of dihydro-heme d 1 lacks one double bond compared with heme d 1 . For this reason, the absorption maxima of dihydro-heme d 1 can be expected to be located at shorter wavelengths than those of heme d 1 (hypsochromic shift). Indeed, such a shift was observed for the absorption maximum of NirS ⌬N (626 nm, oxidized) compared with the heme d 1 -containing NirS WT (642 nm, oxidized).
Resonance Raman Spectroscopy of NirS ⌬N , NirS WT and NirS s.a. -To further characterize the unusual cofactor content of NirS ⌬N and to obtain more insights into the question of whether dihydro-heme d 1 is indeed bound in NirS ⌬N , resonance Raman (RR) spectroscopy was performed. RR spectra of purified NirS WT , NirS ⌬N , and semi-apo NirS were measured at low temperature to reduce the strong background scattering and to increase the signal-to-noise ratio. Although, upon 413-nm excitation, the heme c bands were selectively enhanced (data not shown), 442-nm excitation led to a substantial increase of the spectral mass of heme d 1 and the putative dihydro-heme d 1 in NirS WT and NirS ⌬N , respectively. Remaining contributions of heme c could readily be removed by subtracting the spectra of NirS s.a. from those of NirS WT and NirS ⌬N , using the characteristic bands of heme c as a reference (Fig. 3, C and D). Therefore, pure spectra of heme d 1 and the putative precursor were obtained in the oxidized (as-isolated) state (Fig. 3, A and B). The spectrum of ferric heme d 1 (Fig. 3A) differs substantially from that of heme c, which, because of its higher symmetry (D 4h ), is dominated by the totally symmetric A 1g modes (Fig. 3C) (23). In contrast, the low symmetry of heme d 1 leads to numerous RR-active modes that are not related to the heme c bands. As a consequence, well established empirical relationships for iron porphyrins between band frequencies above 1300 cm Ϫ1 and the oxidation, spin, and coordination state of iron (24) cannot be extended to heme d 1 and, therefore, most of its bands must remain unassigned. Remarkable excep-tions are the bands at 1687 and 1725 cm Ϫ1 , which originate from the CϭO stretching modes of the keto groups and the carboxyl function of the acrylic side chain, respectively. Note that differences between the spectrum in Fig. 3A and a spectrum of oxidized cd 1 published previously (25) are likely to be due to different excitation lines and temperatures.
As already expected from the weaker absorbance at 442 nm for NirS ⌬N compared with NirS WT (Fig. 2), the RR spectrum of the putative heme d 1 precursor suffers from a substantially weaker resonance enhancement compared with heme d 1 (Fig.  3B). Despite the relatively poor signal-to-noise ratio, the comparison of both spectra demonstrates that most of the bands of heme d 1 also have counterparts within Ϯ 3 cm Ϫ1 in the spectrum of the potential heme d 1 precursor, albeit with quite different relative intensities. Therefore, we conclude that the ground state structure of heme d 1 and its precursor are similar but not identical. This conclusion is consistent with the assignment of the heme d 1 precursor to a dihydro-heme d 1 . Moreover, among the few bands of NirS WT for which conjugate bands in NirS ⌬N are not immediately evident, the bands at 1611 and 1181 cm Ϫ1 may be attributed to modes involving the CϭC and C-(CϭC) stretching coordinates, taking into account previous assignments of modes involving the vinyl substituents in protoporphyin IX (26 -29). Furthermore, assuming a dihydroheme d 1 in NirS ⌬N , the different intensity pattern compared with the spectrum of heme d 1 can readily be understood because the extension of the delocalized -electron system affects both the absolute and relative RR intensities.
NirS ⌬N Contains Dihydro-Heme d 1 -The results of the experiments described so far strongly suggest that the heme d 1 precursor dihydro-heme d 1 might indeed be bound in NirS ⌬N . To finally clarify this question, the non-covalently bound heme d 1 (or possibly dihydro-heme d 1 ) was extracted from the purified NirS WT and NirS ⌬N as described under "Experimental Procedures." Then, UV-visible absorption spectra of the oxidized and dithionite-reduced forms were recorded (Fig. 4, A and B). The absorption spectrum of the heme d 1 cofactor extracted from NirS WT exhibits characteristic peaks at 408 and 683 nm and a shoulder at about 476 nm in the oxidized state, as reported previously (30,31). In contrast, the spectrum of the oxidized form of the cofactor extracted from NirS ⌬N shows absorption maxima at 397 and 671 nm and also a shoulder at about 476 nm. After dithionite reduction, the absorption bands of the heme d 1 extracted from NirS WT are shifted to 457 and 627 nm, whereas those of the cofactor extracted from NirS ⌬N are shifted to 447 and 611 nm. Again, these observations strongly suggest that the cofactor isolated from NirS ⌬N was indeed different from the heme d 1 extracted from NirS WT . As described above, we suspected that the heme d 1 precursor dihydro-heme d 1 lacking the acrylic double bond might be present in NirS ⌬N . The UV-visible absorption spectra of the isolated cofactor are in line with this proposal because the characteristic heme d 1 absorption band at 683 nm in the oxidized state is shifted to a shorter wavelength by about 10 nm in the case of the cofactor from NirS ⌬N , and the corresponding band of heme d 1 in the reduced state at 627 nm is shifted by about 15 nm in this case. Such a hypsochromic shift would be expected for dihydro-heme d 1 because of the lacking double bond (6). The spectra in A and B were obtained by subtracting the heme c contributions (spectra C and D) from the spectra of the (as-isolated) enzymes NirS WT and NirS ⌬N , respectively. The spectra of the oxidized heme c (C) and reduced heme c (D) were generated by mutual spectra subtraction of the as-isolated (mainly oxidized) and dithionite-reduced semi-apo NirS because the spectrum of the pure oxidized semi-apo NirS cannot be measured directly because of photoreduction at low temperatures. Note that the ferrous heme c in semi-apo NirS undergoes an unprecedented reversible temperature-dependent transition from the low-spin to the high-spin configuration. The latter prevails at 80 K, as indicated by the characteristic marker bands at 1353, 1470, and 1607 cm Ϫ1 (24,32).

JOURNAL OF BIOLOGICAL CHEMISTRY 30757
Finally, we determined the exact mass of the cofactor extracted from NirS ⌬N by HR-ESI-MS. This method revealed a mass of 712.14675 for the [MϩH] ϩ ion, which is indeed in accordance with the calculated mass (712.14621), isotopic pattern and chemical formula of dihydro-heme d 1 (Fig. 4C). For comparison, the mass of the heme d 1 extracted from NirS WT was also determined by HR-ESI-MS. The measured value of 710.13075 for the [MϩH] ϩ ion is in excellent agreement with the theoretical mass (710.13065), isotopic pattern and chemical formula of heme d 1 as expected (Fig. 4D). Moreover, the mass difference between the two cofactors isolated from either NirS ⌬N or NirS WT corresponds to the two hydrogen atoms that are present in dihydro-heme d 1 (NirS ⌬N ) but absent in heme d 1 (NirS WT ) because of the dehydrogenation of the propionate side chain of pyrrole ring D.
On the basis of all of the results described so far, we conclude that, in the absence of NirN in the bacterial cell, the formation of the acrylic double bond of heme d 1 does not take place and that dihydro-heme d 1 is inserted into NirS ⌬N . This novel finding led to the question whether NirN itself catalyzes the dehydrogenation reaction, which would be completely unexpected, or whether NirF catalyzes the double bond formation, as proposed previously, but with the help of NirN. These two possibilities were tested in vitro, and the results are described in the following section.

Purified, Recombinant NirN from P. aeruginosa Catalyzes the Conversion of Dihydro-heme d 1 to Heme d 1 in Vitro-To test
whether NirN is able to catalyze the dehydrogenation of the propionate side chain of pyrrole ring D, yielding the acrylate side chain of heme d 1 , the extracted dihydro-heme d 1 was mixed with purified NirN, and UV-visible absorption spectra were recorded at different time points after mixing. As shown in Fig. 5A, the absorption band of the free dihydro-heme d 1 at 671 nm disappeared upon addition of NirN, and a new absorption band at 609 nm was visible immediately after mixing. This new absorption feature most likely represents the binding of dihydro-heme d 1 to NirN. However, after 15 s of incubation, the spectrum had changed, and the absorption band at 609 nm had developed into a broad band at about 635 nm that did not change further during prolonged incubation. On the basis of this observation, we speculated that NirN might have catalyzed the formation of the acrylic double bond after the initial binding of the dihydro-heme d 1 to the enzyme. Therefore, we compared the final absorption spectrum of the NirN/dihydroheme d 1 reaction mixture with the absorption spectra of a NirN/heme d 1 mixture in the oxidized and reduced states (Fig.  5B). Strikingly, the final spectrum of the NirN/dihydro-heme d 1 reaction mixture was identical to the spectrum of the NirN/ heme d 1 mixture in the reduced state, both showing an absorption band at 635 nm. Furthermore, we then oxidized the NirN reaction mixture initially containing dihydro-heme d 1 by the addition of sodium persulfate, which resulted in a shift of the 635 nm absorption band to 626 nm. The obtained absorption spectrum was now identical to the spectrum of the NirN/heme d 1 mixture in the oxidized state (Fig. 5B). These results suggest that NirN is indeed able to catalyze the conversion of dihydroheme d 1 to heme d 1 .
To further substantiate this finding, we added NirS s.a. to the NirN/dihydro-heme d 1 reaction mixture after the potential conversion of the precursor to heme d 1 was completed. Because it has been reported previously that heme d 1 -loaded NirN is able to transfer the cofactor to NirS s.a. (12), we assumed that this transfer should also take place in our reaction mixture. After several minutes of incubation with NirS s.a. , UV-visible absorption spectra were recorded. As shown in Fig. 5C, the absorption band at 626 nm potentially representing heme d 1 -loaded NirN (oxidized) was shifted to 641 nm upon addition of NirS s.a. . Indeed, the newly obtained spectrum was identical to the spectrum of heme d 1 -containing NirS. Moreover, the same was true after the reduction of the cofactor with sodium dithionite (Fig. 5D). These observations clearly supported the idea that NirN catalyzed the conversion of dihydro-heme d 1 to heme d 1 . The subsequent transfer of the formed heme d 1 to NirS s.a. yielding holo-NirS took place as described previously (12).
Finally, to unambiguously demonstrate that heme d 1 was formed out of dihydro-heme d 1 by the action of NirN, we extracted the cofactor from the NirN/dihydro-heme d 1 reaction mixture after the reaction was completed and analyzed the formed reaction product by UV-visible absorption spectroscopy and HR-ESI-MS (Fig. 6). The extracted reaction product exhibited an absorption band at 683 nm identical to heme d 1 . Moreover, a mass of 710.13043 was determined for the reaction product ([MϩH] ϩ ), which is in agreement with the theoretical mass (710.13065) of heme d 1 . Therefore, we concluded that NirN indeed catalyzed the formation of the acrylic double bond, therefore converting dihydro-heme d 1 to heme d 1 .
Purified, Recombinant NirF from D. shibae Binds Dihydro-Heme d 1 without Conversion to Heme d 1 -The finding that NirN catalyzes the conversion of dihydro-heme d 1 to heme d 1 was completely unexpected because NirF was originally proposed to be responsible for this reaction. This proposal was made on the basis of the observation that the periplasmic NirF from P. pantotrophus was able to bind heme d 1 in vitro, and the obtained NirF-heme d 1 complex was interpreted to represent an enzyme-product com- plex (13). Therefore, we incubated purified, recombinant NirF from D. shibae with dihydro-heme d 1 and heme d 1 , and the binding of the cofactors to the protein was followed by UV-visible absorption spectroscopy (Fig. 7). The spectrum of the NirF/ heme d 1 mixture exhibits characteristic absorption bands at 429 and 629 nm in the oxidized state and at 448 and 627 nm in the reduced form, which is in agreement with the absorption features of heme d 1 -containing NirF from P. pantotrophus reported previously (13). The spectrum of the NirF/dihydroheme d 1 mixture displays absorption bands at 421 and 593 nm in the oxidized form and at 438 and 608 nm in the reduced form, showing that NirF is also able to bind the dihydro-heme d 1 precursor because the observed absorption features of the NirF/dihydro-heme d 1 mixture clearly differ from those of the free dihydro-heme d 1 . More importantly, the absorption spectrum of the NirF/dihydro-heme d 1 mixture recorded directly after mixing the protein with the cofactor did not change further upon prolonged incubation, in contrast to what was observed for the NirN/dihydro-heme d 1 mixture (see above). Therefore, we concluded that NirF was able to bind dihydroheme d 1 but did not convert it to heme d 1 .
NirN Catalyzes the Double Bond Formation using Electron Bifurcation-As outlined above, our studies revealed a new and unexpected function of NirN during heme d 1 biosynthesis, namely that of the terminal dehydrogenase introducing the double bond into the propionate side chain of pyrrole ring D of dihydro-heme d 1 to yield the acrylate side chain of heme d 1 . However, NirN must be a quite unusual dehydrogenase because it did not require the addition of any cofactor such as NAD ϩ or FAD/FMN typically employed by common dehydrogenases catalyzing two-electron oxidation reactions. Interestingly, in our initial NirN/dihydro-heme d 1 reaction described above, we observed that the produced heme d 1 was in the reduced Fe(II) state, although the substrate dihydro-heme d 1 entered the reaction mixture in the oxidized Fe(III) form. Therefore, it seemed that one electron of the two-electron oxidation remained on the reaction product itself. Nevertheless, the fate of the second electron was still unclear. In this context, it is important to note that NirN contains a covalently attached heme c cofactor (16), and, therefore, we speculated that the second electron might be transferred to the heme c. In the initial reaction mixture, no reduction of the heme c in NirN was observed. However, this reaction was performed under aerobic conditions, which might have led to a rapid reoxidation of any reduced heme c. Therefore, we repeated the NirN activity assay under anaerobic conditions in an anaerobic chamber. NirN in its oxidized form (i.e. containing Fe(III)-heme c) was mixed with oxidized Fe(III)-dihydro-heme d 1 , and the reaction pro-  gress was followed by UV-visible absorption spectroscopy (Fig.  8). As before, a new absorption band at about 635 nm developed during the reaction, which represents the NirN-bound heme d 1 in its reduced Fe(II) form. Moreover, two new absorption peaks at 522 and 550 nm appeared simultaneously with the absorption band at 635 nm. These absorption features are indeed characteristic for the reduced Fe(II)-heme c in NirN (16). On the basis of this experiment, we conclude that NirN catalyzes the two-electron oxidation of dihydro-heme d 1 to heme d 1 using an electron bifurcation mechanism during which one electron ends up on the reaction product heme d 1 and the second electron is transferred to the heme c cofactor of the enzyme.
Novel Roles of NirN and NirF during Heme d 1 Biosynthesis and Insertion into NirS-On the basis of the experiments described in this study, we would like to assign a novel function to the NirN protein. So far, the role of NirN was unclear because nirN deletion strains did not show extreme growth phenotypes and still produced an active form of NirS, which led to the assumption that NirN was apparently not involved in heme d 1 biosynthesis. However, we observed that the deletion of the nirN gene in P. aeruginosa resulted in the production of a holo NirS with an obviously altered cofactor content (15). In this study, we could show that this form of holo NirS actually carried the heme d 1 precursor dihydro-heme d 1 . Further, we showed that NirN is able to catalyze the conversion of dihydroheme d 1 to heme d 1 in vitro. Therefore, we identified NirN as the terminal dehydrogenase catalyzing the last step of heme d 1 biosynthesis. Moreover, we propose that NirN represents a novel and unusual type of a periplasmic dehydrogenase that uses an electron bifurcation mechanism to catalyze a two-electron oxidation reaction carrying only a one-electron accepting redox cofactor (heme c). Such a mechanism is possible for NirN because its substrate/product itself is able to accept one electron.
Considering the finding that NirN catalyzes the last step of heme d 1 biosynthesis, the question arises what the function of NirF might be that was assumed previously to play the role now assigned to NirN. In this study, we observed that NirF is able to bind dihydro-heme d 1 . Together with the previous observation that NirF from P. aeruginosa is a periplasmic, membrane-anchored lipoprotein that interacts with both NirN and NirS (14,15), one might speculate that NirF could be involved in the uptake of dihydro-heme d 1 into the periplasm from a still unknown membrane transporter and then further transfer the precursor to NirN. NirN then catalyzes the formation of heme d 1 , which is finally transferred to NirS, yielding the active holo form of the enzyme. Such a role of NirF in the uptake of the heme d 1 precursor into the periplasm is, of course, speculative at this stage and will require further research.